--- /dev/null
+Intel's Open Source Vulkan Driver
+Vulkan API Version: 0.210.1
+SPIR-V Version: 1.0
+
+Intro
+=====
+The Open Source Technology Center 3D graphics team at Intel has
+been working on a Vulkan implementation based on the Mesa open source
+OpenGL implementation. At this point we're ready to share what we have
+with our Khronos friends, in the hope that an early preview will be
+useful.
+
+When the Vulkan specification goes public, we will continue the work
+in the Mesa public git repository, but in the interim we will share
+our progress on the 'vulkan' branch in the 'mesa' repository in
+Khronos gitlab.
+
+The Mesa project source and our driver implementation is under the MIT
+license [1], but is also covered by the Khronos IP framework as it
+pertains to a specification under construction [2].
+
+We welcome all feedback and contibutions, as long as the contributions
+are MIT licensed and can be open sourced with the driver.
+
+[1] https://opensource.org/licenses/MIT
+[2] https://www.khronos.org/members/ip-framework
+
+
+Maintainers
+===========
+Kristian Høgsberg Kristensen <kristian.h.kristensen@intel.com>
+Jason Ekstrand <jason.ekstrand@intel.com>
+Chad Versace <chad.versace@intel.com>
+
+
+Supported Hardware
+==================
+- Broadwell, main development focus
+- Ivybridge
+- Sky Lake
+- Haswell
+- Bay Trail
+- Cherryview
+- Broxton
+
+
+Supported OS Platforms
+======================
+ - Linux, tested on Fedora 22 with kernel >= 4.1
+ - X11 with DRI3
+ - Wayland
+ - Android
+ - TODO
+
+
+Building and Installing
+=======================
+This driver is intended to be used directly from the build tree. Installing the
+driver into a system location is not yet fully supported. If you require support
+for system-wide installation, please contact a maintainer.
+
+Throughout the instructions, MESA_TOP refers to the top of the Mesa repository.
+
+First, install the usual dependencies needed to build Mesa.
+
+ Fedora:
+ $ sudo yum builddep mesa
+ Ubunutu:
+ $ FINISHME
+
+Next, configure and build. The below commands will build Mesa in release mode.
+If you wish to build Mesa in debug mode, add option '--enable-debug' to the
+configure command.
+
+ $ cd $MESA_TOP
+ $ autoreconf -vfi
+ $ ./configure --with-dri-drivers=i965 --with-gallium-drivers=
+ $ make
+
+To use the driver's libvulkan.so directly, without LunarG's loader, you must set
+an environment variable before running your Vulkan application:
+
+ $ export LD_LIBRARY_PATH="$MESA_TOP/lib"
+ $ your-vk-app
+
+Alternatively, to use the driver with LunarG's loader:
+
+ $ export VK_ICD_FILENAMES="$MESA_TOP/src/vulkan/anv_icd.json"
+ $ your-vk-app
+
+
+File Structure and Naming
+=========================
+The core code of Intel's Mesa Vulkan driver lives in src/vulkan. Files prefixed
+with "gen8" support Broadwell; files prefixed with "gen7" support Ivybridge;
+files prefixed with "anv" are common to all hardware generations.
+
+Mesa is an umbrella open source project containing many drivers for multiple
+APIs. The codename for Intel's Mesa Vulkan driver is "Anvil", hence the filename
+prefix "anv".
+
+
+Feature Status
+==============
+The driver is still a work-in-progress. We do our best to keep the below list of
+features up-to-date.
+
+Supported Features:
+ - Index buffers, instanced draw, indirect draw
+ - Nested command buffers
+ - Consumes SPIR-V (no GLSL "backdoor")
+ - Fragment, vertex, geometry, and compute shaders
+ - Uniform buffers, sampled images, dynamic uniform buffers
+ - Shader storage buffers
+ - Push constants (VS and FS only)
+ - Color, depth and stencil attachments
+ - 1D, 2D, 3D textures, texture arrays
+ - Memory barrier
+ - Optionally integrates with LunarGs loader
+ - WSI extension for X11
+ - Fences
+ - Most copy/blit commands for color and depth buffers,
+ vkCmdCopyImageToBuffer for stencil buffers
+ - Occlution query and timestamps
+ - VkkSemaphore and VkEvent
+
+Unsupported Features:
+ - Shader specialization
+ - Storage images
+ - Tesselation shaders
+ - Push constants in GS and CS (and VS on HSW and prior)
+ - Sparse resources
+ - MSAA
+ - vkCmdClear commands
+ - Input attachments
LIBDRM_RADEON_REQUIRED=2.4.56
LIBDRM_AMDGPU_REQUIRED=2.4.63
LIBDRM_INTEL_REQUIRED=2.4.61
-LIBDRM_NVVIEUX_REQUIRED=2.4.33
-LIBDRM_NOUVEAU_REQUIRED=2.4.62
+LIBDRM_NVVIEUX_REQUIRED=2.4.66
+LIBDRM_NOUVEAU_REQUIRED=2.4.66
LIBDRM_FREEDRENO_REQUIRED=2.4.65
DRI2PROTO_REQUIRED=2.6
DRI3PROTO_REQUIRED=1.0
AM_PROG_CC_C_O
AM_PROG_AS
AX_CHECK_GNU_MAKE
-AC_CHECK_PROGS([PYTHON2], [python2 python])
+AC_CHECK_PROGS([PYTHON2], [python2.7 python2 python])
AC_PROG_SED
AC_PROG_MKDIR_P
_SAVE_CPPFLAGS="$CPPFLAGS"
dnl Compiler macros
-DEFINES="-D__STDC_LIMIT_MACROS"
+DEFINES="-D__STDC_LIMIT_MACROS -D__STDC_CONSTANT_MACROS"
AC_SUBST([DEFINES])
case "$host_os" in
linux*|*-gnu*|gnu*)
CFLAGS="$SSE41_CFLAGS $CFLAGS"
AC_COMPILE_IFELSE([AC_LANG_SOURCE([[
#include <smmintrin.h>
+int param;
int main () {
- __m128i a = _mm_set1_epi32 (0), b = _mm_set1_epi32 (0), c;
+ __m128i a = _mm_set1_epi32 (param), b = _mm_set1_epi32 (param + 1), c;
c = _mm_max_epu32(a, b);
- return 0;
+ return _mm_cvtsi128_si32(c);
}]])], SSE41_SUPPORTED=1)
CFLAGS="$save_CFLAGS"
if test "x$SSE41_SUPPORTED" = x1; then
AM_CONDITIONAL([SSE41_SUPPORTED], [test x$SSE41_SUPPORTED = x1])
AC_SUBST([SSE41_CFLAGS], $SSE41_CFLAGS)
+dnl Check for Endianness
+AC_C_BIGENDIAN(
+ little_endian=no,
+ little_endian=yes,
+ little_endian=no,
+ little_endian=no
+)
+
+dnl Check for POWER8 Architecture
+PWR8_CFLAGS="-mpower8-vector"
+have_pwr8_intrinsics=no
+AC_MSG_CHECKING(whether gcc supports -mpower8-vector)
+save_CFLAGS=$CFLAGS
+CFLAGS="$PWR8_CFLAGS $CFLAGS"
+AC_COMPILE_IFELSE([AC_LANG_SOURCE([[
+#if defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 8))
+#error "Need GCC >= 4.8 for sane POWER8 support"
+#endif
+#include <altivec.h>
+int main () {
+ vector unsigned char r;
+ vector unsigned int v = vec_splat_u32 (1);
+ r = __builtin_vec_vgbbd ((vector unsigned char) v);
+ return 0;
+}]])], have_pwr8_intrinsics=yes)
+CFLAGS=$save_CFLAGS
+
+AC_ARG_ENABLE(pwr8,
+ [AC_HELP_STRING([--disable-pwr8-inst],
+ [disable POWER8-specific instructions])],
+ [enable_pwr8=$enableval], [enable_pwr8=auto])
+
+if test "x$enable_pwr8" = xno ; then
+ have_pwr8_intrinsics=disabled
+fi
+
+if test $have_pwr8_intrinsics = yes && test $little_endian = yes ; then
+ DEFINES="$DEFINES -D_ARCH_PWR8"
+ CXXFLAGS="$CXXFLAGS $PWR8_CFLAGS"
+ CFLAGS="$CFLAGS $PWR8_CFLAGS"
+else
+ PWR8_CFLAGS=
+fi
+
+AC_MSG_RESULT($have_pwr8_intrinsics)
+if test "x$enable_pwr8" = xyes && test $have_pwr8_intrinsics = no ; then
+ AC_MSG_ERROR([POWER8 compiler support not detected])
+fi
+
+if test $have_pwr8_intrinsics = yes && test $little_endian = no ; then
+ AC_MSG_WARN([POWER8 optimization is enabled only on POWER8 Little-Endian])
+fi
+
+AC_SUBST([PWR8_CFLAGS], $PWR8_CFLAGS)
+
dnl Can't have static and shared libraries, default to static if user
dnl explicitly requested. If both disabled, set to static since shared
dnl was explicitly requested.
[enable_debug="$enableval"],
[enable_debug=no]
)
+
+AC_ARG_ENABLE([profile],
+ [AS_HELP_STRING([--enable-profile],
+ [enable profiling of code @<:@default=disabled@:>@])],
+ [enable_profile="$enableval"],
+ [enable_profile=no]
+)
+
+if test "x$enable_profile" = xyes; then
+ DEFINES="$DEFINES -DPROFILE"
+ if test "x$GCC" = xyes; then
+ CFLAGS="$CFLAGS -fno-omit-frame-pointer"
+ fi
+ if test "x$GXX" = xyes; then
+ CXXFLAGS="$CXXFLAGS -fno-omit-frame-pointer"
+ fi
+fi
+
if test "x$enable_debug" = xyes; then
DEFINES="$DEFINES -DDEBUG"
+ if test "x$enable_profile" = xyes; then
+ AC_MSG_WARN([Debug and Profile are enabled at the same time])
+ fi
if test "x$GCC" = xyes; then
if ! echo "$CFLAGS" | grep -q -e '-g'; then
CFLAGS="$CFLAGS -g"
AC_SUBST([GBM_PC_REQ_PRIV])
AC_SUBST([GBM_PC_LIB_PRIV])
+AM_CONDITIONAL(HAVE_VULKAN, true)
+
dnl
dnl EGL configuration
dnl
[CLANG_LIBDIR=''])
PKG_CHECK_EXISTS([libclc], [have_libclc=yes], [have_libclc=no])
-AC_CHECK_LIB([elf], [elf_memory], [have_libelf=yes;ELF_LIB=-lelf])
+PKG_CHECK_MODULES([LIBELF], [libelf], [have_libelf=yes], [have_libelf=no])
+
+if test "x$have_libelf" = xno; then
+ LIBELF_LIBS=''
+ LIBELF_CFLAGS=''
+ AC_CHECK_LIB([elf], [elf_memory], [have_libelf=yes;LIBELF_LIBS=-lelf], [have_libelf=no])
+ AC_SUBST([LIBELF_LIBS])
+ AC_SUBST([LIBELF_CFLAGS])
+fi
if test "x$enable_opencl" = xyes; then
if test -z "$with_gallium_drivers"; then
AC_MSG_ERROR([VC4 simulator on x86 replaces i965 driver build, so ilo must be disabled.])
fi
-AC_SUBST([ELF_LIB])
-
AM_CONDITIONAL(HAVE_LIBDRM, test "x$have_libdrm" = xyes)
AM_CONDITIONAL(HAVE_X11_DRIVER, test "x$enable_xlib_glx" = xyes)
AM_CONDITIONAL(HAVE_OSMESA, test "x$enable_osmesa" = xyes)
AC_SUBST([XA_TINY], $XA_TINY)
AC_SUBST([XA_VERSION], "$XA_MAJOR.$XA_MINOR.$XA_TINY")
+PKG_CHECK_MODULES(VALGRIND, [valgrind],
+ [have_valgrind=yes], [have_valgrind=no])
+if test "x$have_valgrind" = "xyes"; then
+ AC_DEFINE([HAVE_VALGRIND], 1,
+ [Use valgrind intrinsics to suppress false warnings])
+fi
+
dnl Restore LDFLAGS and CPPFLAGS
LDFLAGS="$_SAVE_LDFLAGS"
CPPFLAGS="$_SAVE_CPPFLAGS"
src/glx/apple/Makefile
src/glx/tests/Makefile
src/gtest/Makefile
+ src/isl/Makefile
src/loader/Makefile
src/mapi/Makefile
src/mapi/es1api/glesv1_cm.pc
src/mesa/drivers/osmesa/osmesa.pc
src/mesa/drivers/x11/Makefile
src/mesa/main/tests/Makefile
+ src/vulkan/Makefile
+ src/vulkan/anv_icd.json
+ src/vulkan/tests/Makefile
src/util/Makefile
src/util/tests/hash_table/Makefile])
GL_ARB_gpu_shader_fp64 DONE (llvmpipe, softpipe)
GL_ARB_sample_shading DONE (i965, nv50)
GL_ARB_shader_subroutine DONE (i965, nv50, llvmpipe, softpipe)
- GL_ARB_tessellation_shader DONE ()
+ GL_ARB_tessellation_shader DONE (i965)
GL_ARB_texture_buffer_object_rgb32 DONE (i965, llvmpipe, softpipe)
GL_ARB_texture_cube_map_array DONE (i965, nv50, llvmpipe, softpipe)
GL_ARB_texture_gather DONE (i965, nv50, llvmpipe, softpipe)
- forced alignment within blocks in progress
- specified vec4-slot component numbers in progress
- specified transform/feedback layout in progress
- - input/output block locations in progress
+ - input/output block locations DONE
GL_ARB_multi_bind DONE (all drivers)
GL_ARB_query_buffer_object not started
GL_ARB_texture_mirror_clamp_to_edge DONE (i965, nv50, nvc0, r600, radeonsi, llvmpipe, softpipe)
glMemoryBarrierByRegion DONE
glGetTexLevelParameter[fi]v - needs updates DONE
glGetBooleani_v - restrict to GLES enums
- gl_HelperInvocation support
+ gl_HelperInvocation support DONE (i965, nvc0, r600)
GLES3.2, GLSL ES 3.2
GL_EXT_color_buffer_float DONE (all drivers)
<li><a href="http://www.opengl.org" target="_parent">OpenGL website</a>
<li><a href="http://dri.freedesktop.org" target="_parent">DRI website</a>
<li><a href="http://www.freedesktop.org" target="_parent">freedesktop.org</a>
+<li><a href="http://planet.freedesktop.org" target="_parent">Developer blogs</a>
</ul>
<b>Hosted by:</b>
<br>
<blockquote>
<a href="http://sourceforge.net"
-target="_parent"><img src="http://sourceforge.net/sflogo.php?group_id=3&type=1"
-width="88" height="31" align="bottom" alt="Sourceforge.net" border="0"></a>
+target="_parent">sourceforge.net</a>
</blockquote>
</body>
<h1>News</h1>
+<h2>January 13, 2015</h2>
+<p>
+<a href="relnotes/11.1.1.html">Mesa 11.1.1</a> is released.
+This is a bug-fix release.
+</p>
+
+<h2>December 21, 2015</h2>
+<p>
+<a href="relnotes/11.0.8.html">Mesa 11.0.8</a> is released.
+This is a bug-fix release.
+</p>
+
+<h2>December 15, 2015</h2>
+<p>
+<a href="relnotes/11.1.0.html">Mesa 11.1.0</a> is released. This is a new
+development release. See the release notes for more information about
+the release.
+</p>
+
<h2>December 9, 2015</h2>
<p>
<a href="relnotes/11.0.7.html">Mesa 11.0.7</a> is released.
</p>
<ul>
+<li><a href="relnotes/11.1.1.html">11.1.1 release notes</a>
+<li><a href="relnotes/11.0.8.html">11.0.8 release notes</a>
+<li><a href="relnotes/11.1.0.html">11.1.0 release notes</a>
<li><a href="relnotes/11.0.7.html">11.0.7 release notes</a>
<li><a href="relnotes/11.0.6.html">11.0.6 release notes</a>
<li><a href="relnotes/11.0.5.html">11.0.5 release notes</a>
--- /dev/null
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
+<html lang="en">
+<head>
+ <meta http-equiv="content-type" content="text/html; charset=utf-8">
+ <title>Mesa Release Notes</title>
+ <link rel="stylesheet" type="text/css" href="../mesa.css">
+</head>
+<body>
+
+<div class="header">
+ <h1>The Mesa 3D Graphics Library</h1>
+</div>
+
+<iframe src="../contents.html"></iframe>
+<div class="content">
+
+<h1>Mesa 11.0.8 Release Notes / December 9, 2015</h1>
+
+<p>
+Mesa 11.0.8 is a bug fix release which fixes bugs found since the 11.0.7 release.
+</p>
+<p>
+Mesa 11.0.8 implements the OpenGL 4.1 API, but the version reported by
+glGetString(GL_VERSION) or glGetIntegerv(GL_MAJOR_VERSION) /
+glGetIntegerv(GL_MINOR_VERSION) depends on the particular driver being used.
+Some drivers don't support all the features required in OpenGL 4.1. OpenGL
+4.1 is <strong>only</strong> available if requested at context creation
+because compatibility contexts are not supported.
+</p>
+
+
+<h2>SHA256 checksums</h2>
+<pre>
+ab9db87b54d7525e4b611b82577ea9a9eae55927558df57b190059d5ecd9406f mesa-11.0.8.tar.gz
+5696e4730518b6805d2ed5def393c4293f425a2c2c01bd5ed4bdd7ad62f7ad75 mesa-11.0.8.tar.xz
+</pre>
+
+
+<h2>New features</h2>
+<p>None</p>
+
+<h2>Bug fixes</h2>
+
+<p>This list is likely incomplete.</p>
+
+<ul>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=91806">Bug 91806</a> - configure does not test whether assembler supports sse4.1</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92849">Bug 92849</a> - [IVB HSW BDW] piglit image load/store load-from-cleared-image.shader_test fails</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92909">Bug 92909</a> - Offset/alignment issue with layout std140 and vec3</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=93004">Bug 93004</a> - Guild Wars 2 crash on nouveau DX11 cards</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=93215">Bug 93215</a> - [Regression bisected] Ogles1conform Automatic mipmap generation test is fail</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=93266">Bug 93266</a> - gl_arb_shading_language_420pack does not allow binding of image variables</li>
+
+</ul>
+
+
+<h2>Changes</h2>
+
+<p>Boyuan Zhang (1):</p>
+<ul>
+ <li>radeon/uvd: uv pitch separation for stoney</li>
+</ul>
+
+<p>Dave Airlie (9):</p>
+<ul>
+ <li>r600: do SQ flush ES ring rolling workaround</li>
+ <li>r600: SMX returns CONTEXT_DONE early workaround</li>
+ <li>r600/shader: split address get out to a function.</li>
+ <li>r600/shader: add utility functions to do single slot arithmatic</li>
+ <li>r600g: fix geom shader input indirect indexing.</li>
+ <li>r600: handle geometry dynamic input array index</li>
+ <li>radeonsi: handle doubles in lds load path.</li>
+ <li>mesa/varray: set double arrays to non-normalised.</li>
+ <li>mesa/shader: return correct attribute location for double matrix arrays</li>
+</ul>
+
+<p>Emil Velikov (8):</p>
+<ul>
+ <li>docs: add sha256 checksums for 11.0.7</li>
+ <li>cherry-ignore: don't pick a specific i965 formats patch</li>
+ <li>Revert "i965/nir: Remove unused indirect handling"</li>
+ <li>Revert "i965/state: Get rid of dword_pitch arguments to buffer functions"</li>
+ <li>Revert "i965/vec4: Use a stride of 1 and byte offsets for UBOs"</li>
+ <li>Revert "i965/fs: Use a stride of 1 and byte offsets for UBOs"</li>
+ <li>Revert "i965/vec4: Use byte offsets for UBO pulls on Sandy Bridge"</li>
+ <li>Update version to 11.0.8</li>
+</ul>
+
+<p>Francisco Jerez (1):</p>
+<ul>
+ <li>i965: Resolve color and flush for all active shader images in intel_update_state().</li>
+</ul>
+
+<p>Ian Romanick (1):</p>
+<ul>
+ <li>meta/generate_mipmap: Work-around GLES 1.x problem with GL_DRAW_FRAMEBUFFER</li>
+</ul>
+
+<p>Ilia Mirkin (17):</p>
+<ul>
+ <li>freedreno/a4xx: support lod_bias</li>
+ <li>freedreno/a4xx: fix 5_5_5_1 texture sampler format</li>
+ <li>freedreno/a4xx: point regid to "red" even for alpha-only rb formats</li>
+ <li>nvc0/ir: fold postfactor into immediate</li>
+ <li>nv50/ir: deal with loops with no breaks</li>
+ <li>nv50/ir: the mad source might not have a defining instruction</li>
+ <li>nv50/ir: fix instruction permutation logic</li>
+ <li>nv50/ir: don't forget to mark flagsDef on cvt in txb lowering</li>
+ <li>nv50/ir: fix DCE to not generate 96-bit loads</li>
+ <li>nv50/ir: avoid looking at uninitialized srcMods entries</li>
+ <li>gk110/ir: fix imul hi emission with limm arg</li>
+ <li>gk104/ir: sampler doesn't matter for txf</li>
+ <li>gk110/ir: fix imad sat/hi flag emission for immediate args</li>
+ <li>nv50/ir: fix cutoff for using r63 vs r127 when replacing zero</li>
+ <li>nv50/ir: can't have predication and immediates</li>
+ <li>glsl: assign varying locations to tess shaders when doing SSO</li>
+ <li>ttn: add TEX2 support</li>
+</ul>
+
+<p>Jason Ekstrand (5):</p>
+<ul>
+ <li>i965/vec4: Use byte offsets for UBO pulls on Sandy Bridge</li>
+ <li>i965/fs: Use a stride of 1 and byte offsets for UBOs</li>
+ <li>i965/vec4: Use a stride of 1 and byte offsets for UBOs</li>
+ <li>i965/state: Get rid of dword_pitch arguments to buffer functions</li>
+ <li>i965/nir: Remove unused indirect handling</li>
+</ul>
+
+<p>Jonathan Gray (2):</p>
+<ul>
+ <li>configure.ac: use pkg-config for libelf</li>
+ <li>configure: check for python2.7 for PYTHON2</li>
+</ul>
+
+<p>Kenneth Graunke (2):</p>
+<ul>
+ <li>i965: Fix fragment shader struct inputs.</li>
+ <li>i965: Fix scalar vertex shader struct outputs.</li>
+</ul>
+
+<p>Marek Olšák (8):</p>
+<ul>
+ <li>radeonsi: fix occlusion queries on Fiji</li>
+ <li>radeonsi: fix a hang due to uninitialized border color registers</li>
+ <li>radeonsi: fix Fiji for LLVM <= 3.7</li>
+ <li>radeonsi: don't call of u_prims_for_vertices for patches and rectangles</li>
+ <li>radeonsi: apply the streamout workaround to Fiji as well</li>
+ <li>gallium/radeon: fix Hyper-Z hangs by programming PA_SC_MODE_CNTL_1 correctly</li>
+ <li>tgsi/scan: add flag colors_written</li>
+ <li>r600g: write all MRTs only if there is exactly one output (fixes a hang)</li>
+</ul>
+
+<p>Matt Turner (1):</p>
+<ul>
+ <li>glsl: Allow binding of image variables with 420pack.</li>
+</ul>
+
+<p>Neil Roberts (2):</p>
+<ul>
+ <li>i965: Add MESA_FORMAT_B8G8R8X8_SRGB to brw_format_for_mesa_format</li>
+ <li>i965: Add B8G8R8X8_SRGB to the alpha format override</li>
+</ul>
+
+<p>Oded Gabbay (1):</p>
+<ul>
+ <li>configura.ac: fix test for SSE4.1 assembler support</li>
+</ul>
+
+<p>Patrick Rudolph (2):</p>
+<ul>
+ <li>nv50,nvc0: fix use-after-free when vertex buffers are unbound</li>
+ <li>gallium/util: return correct number of bound vertex buffers</li>
+</ul>
+
+<p>Samuel Pitoiset (1):</p>
+<ul>
+ <li>nvc0: free memory allocated by the prog which reads MP perf counters</li>
+</ul>
+
+<p>Tapani Pälli (1):</p>
+<ul>
+ <li>i965: use _Shader to get fragment program when updating surface state</li>
+</ul>
+
+<p>Tom Stellard (2):</p>
+<ul>
+ <li>radeonsi: Rename si_shader::ls_rsrc{1,2} to si_shader::rsrc{1,2}</li>
+ <li>radeonsi/compute: Use the compiler's COMPUTE_PGM_RSRC* register values</li>
+</ul>
+
+
+</div>
+</body>
+</html>
<iframe src="../contents.html"></iframe>
<div class="content">
-<h1>Mesa 11.1.0 Release Notes / TBD</h1>
+<h1>Mesa 11.1.0 Release Notes / 15 December 2015</h1>
<p>
Mesa 11.1.0 is a new development release.
<h2>SHA256 checksums</h2>
<pre>
-TBD.
+e3bc44be4df5e4dc728dfda7b55b1aaeadfce36eca6a367b76cc07598070cb2d mesa-11.1.0.tar.gz
+9befe03b04223eb1ede177fa8cac001e2850292c8c12a3ec9929106afad9cf1f mesa-11.1.0.tar.xz
</pre>
<h2>Bug fixes</h2>
-TBD.
+<p>This list is likely incomplete.</p>
+
+<ul>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=28130">Bug 28130</a> - vbo: premature flushing breaks GL_LINE_LOOP</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=38109">Bug 38109</a> - i915 driver crashes if too few vertices are submitted (Mesa 7.10.2)</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=49779">Bug 49779</a> - Extra line segments in GL_LINE_LOOP</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=55552">Bug 55552</a> - Compile errors with --enable-mangling</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=71789">Bug 71789</a> - [r300g] Visuals not found in (default) depth = 24</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=79783">Bug 79783</a> - Distorted output in obs-studio where other vendors "work"</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=80821">Bug 80821</a> - When LIBGL_ALWAYS_SOFTWARE is set, KHR_create_context is not supported</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=81174">Bug 81174</a> - Gallium: GL_LINE_LOOP broken with more than 512 points</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=83508">Bug 83508</a> - [UBO] Assertion for array of blocks</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=84677">Bug 84677</a> - Triangle disappears with glPolygonMode GL_LINE</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=86281">Bug 86281</a> - brw_meta_fast_clear (brw=brw@entry=0x7fffd4097a08, fb=fb@entry=0x7fffd40fa900, buffers=buffers@entry=2, partial_clear=partial_clear@entry=false)</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=86469">Bug 86469</a> - Unreal Engine demo doesn't run</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=86720">Bug 86720</a> - [radeon] Europa Universalis 4 freezing during game start (10.3.3+, still broken on 11.0.2)</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=89014">Bug 89014</a> - PIPE_QUERY_GPU_FINISHED is not acting as expected on SI</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=90175">Bug 90175</a> - [hsw bisected][PATCH] atomic counters doesn't work for a binding point different to zero</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=90348">Bug 90348</a> - Spilling failure of b96 merged value</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=90631">Bug 90631</a> - Compilation failure for fragment shader with many branches on Sandy Bridge</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=90734">Bug 90734</a> - glBufferSubData is corrupting data when buffer is > 32k</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=90887">Bug 90887</a> - PhiMovesPass in register allocator broken</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=91044">Bug 91044</a> - piglit spec/egl_khr_create_context/valid debug flag gles* fail</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=91114">Bug 91114</a> - ES3-CTS.gtf.GL3Tests.shadow.shadow_execution_vert fails</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=91254">Bug 91254</a> - (regresion) video using VA-API on Intel slow and freeze system with mesa 10.6 or 10.6.1</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=91292">Bug 91292</a> - [BDW+] glVertexAttribDivisor not working in combination with glPolygonMode</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=91342">Bug 91342</a> - Very dark textures on some objects in indoors environments in Postal 2</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=91526">Bug 91526</a> - World of Warcraft (on Wine) has UI corruption with nouveau</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=91551">Bug 91551</a> - DXTn compressed normal maps produce severe artifacts on all NV5x and NVDx chipsets</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=91596">Bug 91596</a> - EGL_KHR_gl_colorspace (v2) causes problem with Android-x86 GUI</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=91716">Bug 91716</a> - [bisected] piglit.shaders.glsl-vs-int-attrib regresses on 32 bit BYT, HSW, IVB, SNB</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=91718">Bug 91718</a> - piglit.spec.arb_shader_image_load_store.invalid causes intermittent GPU HANG</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=91719">Bug 91719</a> - [SNB,HSW,BYT] dEQP regressions associated with using NIR for vertex shaders</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=91726">Bug 91726</a> - R600 asserts in tgsi_cmp/make_src_for_op3</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=91780">Bug 91780</a> - Rendering issues with geometry shader</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=91785">Bug 91785</a> - make check DispatchSanity_test.GLES31 regression</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=91788">Bug 91788</a> - [HSW Regression] Synmark2_v6 Multithread performance case FPS reduced by 36%</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=91847">Bug 91847</a> - glGenerateTextureMipmap not working (no errors) unless glActiveTexture(GL_TEXTURE1) is called before</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=91857">Bug 91857</a> - Mesa 10.6.3 linker is slow</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=91881">Bug 91881</a> - regression: GPU lockups since mesa-11.0.0_rc1 on RV620 (r600) driver</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=91890">Bug 91890</a> - [nve7] witcher2: blurry image & DATA_ERRORs (class 0xa097 mthd 0x2380/0x238c)</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=91898">Bug 91898</a> - src/util/mesa-sha1.c:250:25: fatal error: openssl/sha.h: No such file or directory</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=91927">Bug 91927</a> - [SKL] [regression] piglit compressed textures tests fail with kernel upgrade</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=91930">Bug 91930</a> - Program with GtkGLArea widget does not redraw</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=91970">Bug 91970</a> - [BSW regression] dEQP-GLES3.functional.shaders.precision.int.highp_mul_vertex</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=91985">Bug 91985</a> - [regression, bisected] FTBFS with commit f9caabe8f1: R600_UCP_CONST_BUFFER is undefined</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=91993">Bug 91993</a> - Graphical glitch in Astromenace (open-source game).</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92009">Bug 92009</a> - ES3-CTS.gtf.GL3Tests.packed_pixels.packed_pixels fails</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92033">Bug 92033</a> - [SNB,regression,dEQP,bisected] functional.shaders.random tests regressed</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92052">Bug 92052</a> - nir/nir_builder.h:79: error: expected primary-expression before ‘.’ token</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92054">Bug 92054</a> - make check gbm-symbols-check regression</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92066">Bug 92066</a> - [ILK,G45,regression] New assertion on BRW_MAX_MRF breaks ilk and g45</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92072">Bug 92072</a> - Wine breakage since d082c5324 (st/mesa: don't call st_validate_state in BlitFramebuffer)</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92095">Bug 92095</a> - [Regression, bisected] arb_shader_atomic_counters.compiler.builtins.frag</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92122">Bug 92122</a> - [bisected, cts] Regression with Assault Android Cactus</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92124">Bug 92124</a> - shader_query.cpp:841:34: error: ‘strndup’ was not declared in this scope</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92183">Bug 92183</a> - linker.cpp:3187:46: error: ‘strtok_r’ was not declared in this scope</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92193">Bug 92193</a> - [SKL] ES2-CTS.gtf.GL2ExtensionTests.compressed_astc_texture.compressed_astc_texture fails</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92214">Bug 92214</a> - Flightgear crashes during splashboot with R600 driver, LLVM 3.7.0 and mesa 11.0.2</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92221">Bug 92221</a> - Unintended code changes in _mesa_base_tex_format commit</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92265">Bug 92265</a> - Black windows in weston after update mesa to 11.0.2-1</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92304">Bug 92304</a> - [cts] cts.shaders.negative conformance tests fail</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92363">Bug 92363</a> - [BSW/BDW] ogles1conform Gets test fails</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92437">Bug 92437</a> - osmesa: Expose GL entry points for Windows build, via .def file</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92438">Bug 92438</a> - Segfault in pushbuf_kref when running the android emulator (qemu) on nv50</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92476">Bug 92476</a> - [cts] ES2-CTS.gtf.GL2ExtensionTests.egl_image.egl_image fails</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92588">Bug 92588</a> - [HSW,BDW,BSW,SKL-Y][GLES 3.1 CTS] ES31-CTS.arrays_of_arrays.InteractionFunctionCalls2 - assert</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92621">Bug 92621</a> - [G965 ILK G45] Regression: 24 piglit regressions in glsl-1.10</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92623">Bug 92623</a> - Differences in prog_data ignored when caching fragment programs (causes hangs)</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92634">Bug 92634</a> - gallium's vl_mpeg12_decoder does not work with st/va</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92639">Bug 92639</a> - [Regression bisected] Ogles1conform mustpass.c fail</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92641">Bug 92641</a> - [SKL BSW] [Regression] Ogles1conform userclip.c fail</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92645">Bug 92645</a> - kodi vdpau interop fails since mesa,meta: move gl_texture_object::TargetIndex initializations</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92705">Bug 92705</a> - [clover] fail to build with llvm-svn/clang-svn 3.8</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92709">Bug 92709</a> - "LLVM triggered Diagnostic Handler: unsupported call to function ldexpf in main" when starting race in stuntrally</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92738">Bug 92738</a> - Randon R7 240 doesn't work on 16KiB page size platform</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92744">Bug 92744</a> - [g965 Regression bisected] Performance regression and piglit assertions due to liveness analysis</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92770">Bug 92770</a> - [SNB, regression, dEQP] deqp-gles3.functional.shaders.discard.dynamic_loop_texture</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92824">Bug 92824</a> - [regression, bisected] `make check` dispatch-sanity broken by GL_EXT_buffer_storage</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92849">Bug 92849</a> - [IVB HSW BDW] piglit image load/store load-from-cleared-image.shader_test fails</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92859">Bug 92859</a> - [regression, bisected] validate_intrinsic_instr: Assertion triggered</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92860">Bug 92860</a> - [radeonsi][bisected] st/mesa: implement ARB_copy_image - Corruption in ARK Survival Evolved</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92900">Bug 92900</a> - [regression bisected] About 700 piglit regressions is what could go wrong</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92909">Bug 92909</a> - Offset/alignment issue with layout std140 and vec3</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92985">Bug 92985</a> - Mac OS X build error "ar: no archive members specified"</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=93015">Bug 93015</a> - Tonga Elemental segfault + VM faults since radeon: implement r600_query_hw_get_result via function pointers</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=93048">Bug 93048</a> - [CTS regression] mesa af2723 breaks GL Conformance for debug extension</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=93063">Bug 93063</a> - drm_helper.h:227:1: error: static declaration of ‘pipe_virgl_create_screen’ follows non-static declaration</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=93091">Bug 93091</a> - [opencl] segfault when running any opencl programs (like clinfo)</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=93126">Bug 93126</a> - wrongly claim supporting GL_EXT_texture_rg</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=93180">Bug 93180</a> - [regression] arb_separate_shader_objects.active sampler conflict fails</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=93235">Bug 93235</a> - [regression] dispatch sanity broken by GetPointerv</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=93266">Bug 93266</a> - gl_arb_shading_language_420pack does not allow binding of image variables</li>
+
+</ul>
+
<h2>Changes</h2>
-TBD.
+<li>MPEG4 decoding has been disabled by default in the VAAPI driver</li>
</div>
</body>
--- /dev/null
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
+<html lang="en">
+<head>
+ <meta http-equiv="content-type" content="text/html; charset=utf-8">
+ <title>Mesa Release Notes</title>
+ <link rel="stylesheet" type="text/css" href="../mesa.css">
+</head>
+<body>
+
+<div class="header">
+ <h1>The Mesa 3D Graphics Library</h1>
+</div>
+
+<iframe src="../contents.html"></iframe>
+<div class="content">
+
+<h1>Mesa 11.1.1 Release Notes / January 13, 2016</h1>
+
+<p>
+Mesa 11.1.1 is a bug fix release which fixes bugs found since the 11.1.0 release.
+</p>
+<p>
+Mesa 11.1.1 implements the OpenGL 4.1 API, but the version reported by
+glGetString(GL_VERSION) or glGetIntegerv(GL_MAJOR_VERSION) /
+glGetIntegerv(GL_MINOR_VERSION) depends on the particular driver being used.
+Some drivers don't support all the features required in OpenGL 4.1. OpenGL
+4.1 is <strong>only</strong> available if requested at context creation
+because compatibility contexts are not supported.
+</p>
+
+
+<h2>SHA256 checksums</h2>
+<pre>
+b15089817540ba0bffd0aad323ecf3a8ff6779568451827c7274890b4a269d58 mesa-11.1.1.tar.gz
+64db074fc514136b5fb3890111f0d50604db52f0b1e94ba3fcb0fe8668a7fd20 mesa-11.1.1.tar.xz
+</pre>
+
+
+<h2>New features</h2>
+<p>None</p>
+
+<h2>Bug fixes</h2>
+
+<p>This list is likely incomplete.</p>
+
+<ul>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=91806">Bug 91806</a> - configure does not test whether assembler supports sse4.1</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92229">Bug 92229</a> - [APITRACE] SOMA have serious graphical errors</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=92233">Bug 92233</a> - Unigine Heaven 4.0 silhuette run</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=93004">Bug 93004</a> - Guild Wars 2 crash on nouveau DX11 cards</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=93215">Bug 93215</a> - [Regression bisected] Ogles1conform Automatic mipmap generation test is fail</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=93257">Bug 93257</a> - [SKL, bisected] ASTC dEQP tests segfault</li>
+
+</ul>
+
+
+<h2>Changes</h2>
+
+<p>Brian Paul (1):</p>
+<ul>
+ <li>st/mesa: check state->mesa in early return check in st_validate_state()</li>
+</ul>
+
+<p>Dave Airlie (6):</p>
+<ul>
+ <li>mesa/varray: set double arrays to non-normalised.</li>
+ <li>mesa/shader: return correct attribute location for double matrix arrays</li>
+ <li>glsl: pass stage into mark function</li>
+ <li>glsl/fp64: add helper for dual slot double detection.</li>
+ <li>glsl: fix count_attribute_slots to allow for different 64-bit handling</li>
+ <li>glsl: only update doubles inputs for vertex inputs.</li>
+</ul>
+
+<p>Emil Velikov (4):</p>
+<ul>
+ <li>docs: add sha256 checksums for 11.0.1</li>
+ <li>cherry-ignore: drop the "re-enable" DCC on Stoney</li>
+ <li>cherry-ignore: don't pick a specific i965 formats patch</li>
+ <li>Update version to 11.1.1</li>
+</ul>
+
+<p>Eric Anholt (2):</p>
+<ul>
+ <li>vc4: Warn instead of abort()ing on exec ioctl failures.</li>
+ <li>vc4: Keep sample mask writes from being reordered after TLB writes</li>
+</ul>
+
+<p>Grazvydas Ignotas (1):</p>
+<ul>
+ <li>r600: fix constant buffer size programming</li>
+</ul>
+
+<p>Ian Romanick (1):</p>
+<ul>
+ <li>meta/generate_mipmap: Work-around GLES 1.x problem with GL_DRAW_FRAMEBUFFER</li>
+</ul>
+
+<p>Ilia Mirkin (9):</p>
+<ul>
+ <li>nv50/ir: can't have predication and immediates</li>
+ <li>gk104/ir: simplify and fool-proof texbar algorithm</li>
+ <li>glsl: assign varying locations to tess shaders when doing SSO</li>
+ <li>glx/dri3: a drawable might not be bound at wait time</li>
+ <li>nvc0: don't forget to reset VTX_TMP bufctx slot after blit completion</li>
+ <li>nv50/ir: float(s32 & 0xff) = float(u8), not s8</li>
+ <li>nv50,nvc0: make sure there's pushbuf space and that we ref the bo early</li>
+ <li>nv50,nvc0: fix crash when increasing bsp bo size for h264</li>
+ <li>nvc0: scale up inter_bo size so that it's 16M for a 4K video</li>
+</ul>
+
+<p>Jonathan Gray (2):</p>
+<ul>
+ <li>configure.ac: use pkg-config for libelf</li>
+ <li>configure: check for python2.7 for PYTHON2</li>
+</ul>
+
+<p>Kenneth Graunke (5):</p>
+<ul>
+ <li>ralloc: Fix ralloc_adopt() to the old context's last child's parent.</li>
+ <li>drirc: Disable ARB_blend_func_extended for Heaven 4.0/Valley 1.0.</li>
+ <li>glsl: Fix varying struct locations when varying packing is disabled.</li>
+ <li>nvc0: Set winding order regardless of domain.</li>
+ <li>nir: Add a lower_fdiv option, turn fdiv into fmul/frcp.</li>
+</ul>
+
+<p>Marek Olšák (7):</p>
+<ul>
+ <li>tgsi/scan: add flag colors_written</li>
+ <li>r600g: write all MRTs only if there is exactly one output (fixes a hang)</li>
+ <li>radeonsi: don't call of u_prims_for_vertices for patches and rectangles</li>
+ <li>radeonsi: apply the streamout workaround to Fiji as well</li>
+ <li>gallium/radeon: fix Hyper-Z hangs by programming PA_SC_MODE_CNTL_1 correctly</li>
+ <li>program: add _mesa_reserve_parameter_storage</li>
+ <li>st/mesa: fix GLSL uniform updates for glBitmap & glDrawPixels (v2)</li>
+</ul>
+
+<p>Mark Janes (1):</p>
+<ul>
+ <li>Add missing platform information for KBL</li>
+</ul>
+
+<p>Miklós Máté (1):</p>
+<ul>
+ <li>mesa: Don't leak ATIfs instructions in DeleteFragmentShader</li>
+</ul>
+
+<p>Neil Roberts (3):</p>
+<ul>
+ <li>i965: Add MESA_FORMAT_B8G8R8X8_SRGB to brw_format_for_mesa_format</li>
+ <li>i965: Add B8G8R8X8_SRGB to the alpha format override</li>
+ <li>i965: Fix crash when calling glViewport with no surface bound</li>
+</ul>
+
+<p>Nicolai Hähnle (2):</p>
+<ul>
+ <li>gallium/radeon: only dispose locally created target machine in radeon_llvm_compile</li>
+ <li>gallium/radeon: fix regression in a number of driver queries</li>
+</ul>
+
+<p>Oded Gabbay (1):</p>
+<ul>
+ <li>configura.ac: fix test for SSE4.1 assembler support</li>
+</ul>
+
+<p>Patrick Rudolph (2):</p>
+<ul>
+ <li>nv50,nvc0: fix use-after-free when vertex buffers are unbound</li>
+ <li>gallium/util: return correct number of bound vertex buffers</li>
+</ul>
+
+<p>Rob Herring (1):</p>
+<ul>
+ <li>freedreno/ir3: fix 32-bit builds with pointer-to-int-cast error enabled</li>
+</ul>
+
+<p>Samuel Pitoiset (3):</p>
+<ul>
+ <li>nvc0: free memory allocated by the prog which reads MP perf counters</li>
+ <li>nv50,nvc0: free memory allocated by performance metrics</li>
+ <li>nv50: free memory allocated by the prog which reads MP perf counters</li>
+</ul>
+
+<p>Sarah Sharp (1):</p>
+<ul>
+ <li>mesa: Add KBL PCI IDs and platform information.</li>
+</ul>
+
+
+</div>
+</body>
+</html>
<li>GL_ARB_base_instance on freedreno/a4xx</li>
<li>GL_ARB_compute_shader on i965</li>
<li>GL_ARB_copy_image on r600</li>
-<li>GL_ARB_tessellation_shader on r600 (evergreen/cayman only)</li>
+<li>GL_ARB_indirect_parameters on nvc0</li>
+<li>GL_ARB_shader_draw_parameters on i965, nvc0</li>
+<li>GL_ARB_tessellation_shader on i965 and r600 (evergreen/cayman only)</li>
<li>GL_ARB_texture_buffer_object_rgb32 on freedreno/a4xx</li>
<li>GL_ARB_texture_buffer_range on freedreno/a4xx</li>
<li>GL_ARB_texture_query_lod on freedreno/a4xx</li>
<li>GL_ARB_vertex_type_10f_11f_11f_rev on freedreno/a4xx</li>
<li>GL_KHR_texture_compression_astc_ldr on freedreno/a4xx</li>
<li>GL_AMD_performance_monitor on radeonsi (CIK+ only)</li>
+<li>New OSMesaCreateContextAttribs() function (for creating core profile
+ contexts)</li>
</ul>
<h2>Bug fixes</h2>
<li>The
<a href="http://www.mesa3d.org">Mesa</a>
website is hosted by
-<a href="http://sourceforge.net">
-<img src="http://sourceforge.net/sflogo.php?group_id=3&type=1"
-width="88" height="31" align="bottom" alt="Sourceforge.net" border="0"></a>
+<a href="http://sourceforge.net">sourceforge.net</a>.
<br>
<br>
#include <GL/gl.h>
-#define OSMESA_MAJOR_VERSION 10
-#define OSMESA_MINOR_VERSION 0
+#define OSMESA_MAJOR_VERSION 11
+#define OSMESA_MINOR_VERSION 2
#define OSMESA_PATCH_VERSION 0
#define OSMESA_MAX_WIDTH 0x24 /* new in 4.0 */
#define OSMESA_MAX_HEIGHT 0x25 /* new in 4.0 */
+/*
+ * Accepted in OSMesaCreateContextAttrib's attribute list.
+ */
+#define OSMESA_DEPTH_BITS 0x30
+#define OSMESA_STENCIL_BITS 0x31
+#define OSMESA_ACCUM_BITS 0x32
+#define OSMESA_PROFILE 0x33
+#define OSMESA_CORE_PROFILE 0x34
+#define OSMESA_COMPAT_PROFILE 0x35
+#define OSMESA_CONTEXT_MAJOR_VERSION 0x36
+#define OSMESA_CONTEXT_MINOR_VERSION 0x37
+
typedef struct osmesa_context *OSMesaContext;
/*
+ * Create an Off-Screen Mesa rendering context with attribute list.
+ * The list is composed of (attribute, value) pairs and terminated with
+ * attribute==0. Supported Attributes:
+ *
+ * Attributes Values
+ * --------------------------------------------------------------------------
+ * OSMESA_FORMAT OSMESA_RGBA*, OSMESA_BGRA, OSMESA_ARGB, etc.
+ * OSMESA_DEPTH_BITS 0*, 16, 24, 32
+ * OSMESA_STENCIL_BITS 0*, 8
+ * OSMESA_ACCUM_BITS 0*, 16
+ * OSMESA_PROFILE OSMESA_COMPAT_PROFILE*, OSMESA_CORE_PROFILE
+ * OSMESA_CONTEXT_MAJOR_VERSION 1*, 2, 3
+ * OSMESA_CONTEXT_MINOR_VERSION 0+
+ *
+ * Note: * = default value
+ *
+ * We return a context version >= what's specified by OSMESA_CONTEXT_MAJOR/
+ * MINOR_VERSION for the given profile. For example, if you request a GL 1.4
+ * compat profile, you might get a GL 3.0 compat profile.
+ * Otherwise, null is returned if the version/profile is not supported.
+ *
+ * New in Mesa 11.2
+ */
+GLAPI OSMesaContext GLAPIENTRY
+OSMesaCreateContextAttribs( const int *attribList, OSMesaContext sharelist );
+
+
+
+/*
* Destroy an Off-Screen Mesa rendering context.
*
* Input: ctx - the context to destroy
CHIPSET(0x193A, skl_gt4, "Intel(R) Skylake GT4")
CHIPSET(0x193B, skl_gt4, "Intel(R) Skylake GT4")
CHIPSET(0x193D, skl_gt4, "Intel(R) Skylake GT4")
+CHIPSET(0x5902, kbl_gt1, "Intel(R) Kabylake GT1")
+CHIPSET(0x5906, kbl_gt1, "Intel(R) Kabylake GT1")
+CHIPSET(0x590A, kbl_gt1, "Intel(R) Kabylake GT1")
+CHIPSET(0x590B, kbl_gt1, "Intel(R) Kabylake GT1")
+CHIPSET(0x590E, kbl_gt1, "Intel(R) Kabylake GT1")
+CHIPSET(0x5913, kbl_gt1_5, "Intel(R) Kabylake GT1.5")
+CHIPSET(0x5915, kbl_gt1_5, "Intel(R) Kabylake GT1.5")
+CHIPSET(0x5917, kbl_gt1_5, "Intel(R) Kabylake GT1.5")
+CHIPSET(0x5912, kbl_gt2, "Intel(R) Kabylake GT2")
+CHIPSET(0x5916, kbl_gt2, "Intel(R) Kabylake GT2")
+CHIPSET(0x591A, kbl_gt2, "Intel(R) Kabylake GT2")
+CHIPSET(0x591B, kbl_gt2, "Intel(R) Kabylake GT2")
+CHIPSET(0x591D, kbl_gt2, "Intel(R) Kabylake GT2")
+CHIPSET(0x591E, kbl_gt2, "Intel(R) Kabylake GT2")
+CHIPSET(0x5921, kbl_gt2, "Intel(R) Kabylake GT2F")
+CHIPSET(0x5926, kbl_gt3, "Intel(R) Kabylake GT3")
+CHIPSET(0x592A, kbl_gt3, "Intel(R) Kabylake GT3")
+CHIPSET(0x592B, kbl_gt3, "Intel(R) Kabylake GT3")
+CHIPSET(0x5932, kbl_gt4, "Intel(R) Kabylake GT4")
+CHIPSET(0x593A, kbl_gt4, "Intel(R) Kabylake GT4")
+CHIPSET(0x593B, kbl_gt4, "Intel(R) Kabylake GT4")
+CHIPSET(0x593D, kbl_gt4, "Intel(R) Kabylake GT4")
CHIPSET(0x22B0, chv, "Intel(R) HD Graphics (Cherryview)")
CHIPSET(0x22B1, chv, "Intel(R) HD Graphics (Cherryview)")
CHIPSET(0x22B2, chv, "Intel(R) HD Graphics (Cherryview)")
--- /dev/null
+//
+// File: vk_platform.h
+//
+/*
+** Copyright (c) 2014-2015 The Khronos Group Inc.
+**
+** Permission is hereby granted, free of charge, to any person obtaining a
+** copy of this software and/or associated documentation files (the
+** "Materials"), to deal in the Materials without restriction, including
+** without limitation the rights to use, copy, modify, merge, publish,
+** distribute, sublicense, and/or sell copies of the Materials, and to
+** permit persons to whom the Materials are furnished to do so, subject to
+** the following conditions:
+**
+** The above copyright notice and this permission notice shall be included
+** in all copies or substantial portions of the Materials.
+**
+** THE MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+** EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+** MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+** IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
+** CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+** TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+** MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS.
+*/
+
+
+#ifndef __VK_PLATFORM_H__
+#define __VK_PLATFORM_H__
+
+#ifdef __cplusplus
+extern "C"
+{
+#endif // __cplusplus
+
+/*
+***************************************************************************************************
+* Platform-specific directives and type declarations
+***************************************************************************************************
+*/
+
+/* Platform-specific calling convention macros.
+ *
+ * Platforms should define these so that Vulkan clients call Vulkan commands
+ * with the same calling conventions that the Vulkan implementation expects.
+ *
+ * VKAPI_ATTR - Placed before the return type in function declarations.
+ * Useful for C++11 and GCC/Clang-style function attribute syntax.
+ * VKAPI_CALL - Placed after the return type in function declarations.
+ * Useful for MSVC-style calling convention syntax.
+ * VKAPI_PTR - Placed between the '(' and '*' in function pointer types.
+ *
+ * Function declaration: VKAPI_ATTR void VKAPI_CALL vkCommand(void);
+ * Function pointer type: typedef void (VKAPI_PTR *PFN_vkCommand)(void);
+ */
+#if defined(_WIN32)
+ // On Windows, Vulkan commands use the stdcall convention
+ #define VKAPI_ATTR
+ #define VKAPI_CALL __stdcall
+ #define VKAPI_PTR VKAPI_CALL
+#elif defined(__ANDROID__) && defined(__ARM_EABI__) && !defined(__ARM_ARCH_7A__)
+ // Android does not support Vulkan in native code using the "armeabi" ABI.
+ #error "Vulkan requires the 'armeabi-v7a' or 'armeabi-v7a-hard' ABI on 32-bit ARM CPUs"
+#elif defined(__ANDROID__) && defined(__ARM_ARCH_7A__)
+ // On Android/ARMv7a, Vulkan functions use the armeabi-v7a-hard calling
+ // convention, even if the application's native code is compiled with the
+ // armeabi-v7a calling convention.
+ #define VKAPI_ATTR __attribute__((pcs("aapcs-vfp")))
+ #define VKAPI_CALL
+ #define VKAPI_PTR VKAPI_ATTR
+#else
+ // On other platforms, use the default calling convention
+ #define VKAPI_ATTR
+ #define VKAPI_CALL
+ #define VKAPI_PTR
+#endif
+
+#include <stddef.h>
+
+#if !defined(VK_NO_STDINT_H)
+ #if defined(_MSC_VER) && (_MSC_VER < 1600)
+ typedef signed __int8 int8_t;
+ typedef unsigned __int8 uint8_t;
+ typedef signed __int16 int16_t;
+ typedef unsigned __int16 uint16_t;
+ typedef signed __int32 int32_t;
+ typedef unsigned __int32 uint32_t;
+ typedef signed __int64 int64_t;
+ typedef unsigned __int64 uint64_t;
+ #else
+ #include <stdint.h>
+ #endif
+#endif // !defined(VK_NO_STDINT_H)
+
+#ifdef __cplusplus
+} // extern "C"
+#endif // __cplusplus
+
+// Platform-specific headers required by platform window system extensions.
+// These are enabled prior to #including "vulkan.h". The same enable then
+// controls inclusion of the extension interfaces in vulkan.h.
+
+#ifdef VK_USE_PLATFORM_ANDROID_KHR
+#include <android/native_window.h>
+#endif
+
+#ifdef VK_USE_PLATFORM_MIR_KHR
+#include <mir_toolkit/client_types.h>
+#endif
+
+#ifdef VK_USE_PLATFORM_WAYLAND_KHR
+#include <wayland-client.h>
+#endif
+
+#ifdef VK_USE_PLATFORM_WIN32_KHR
+#include <windows.h>
+#endif
+
+#ifdef VK_USE_PLATFORM_XLIB_KHR
+#include <X11/Xlib.h>
+#endif
+
+#ifdef VK_USE_PLATFORM_XCB_KHR
+#include <xcb/xcb.h>
+#endif
+
+#endif // __VK_PLATFORM_H__
--- /dev/null
+#ifndef __vulkan_h_
+#define __vulkan_h_ 1
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+** Copyright (c) 2015 The Khronos Group Inc.
+**
+** Permission is hereby granted, free of charge, to any person obtaining a
+** copy of this software and/or associated documentation files (the
+** "Materials"), to deal in the Materials without restriction, including
+** without limitation the rights to use, copy, modify, merge, publish,
+** distribute, sublicense, and/or sell copies of the Materials, and to
+** permit persons to whom the Materials are furnished to do so, subject to
+** the following conditions:
+**
+** The above copyright notice and this permission notice shall be included
+** in all copies or substantial portions of the Materials.
+**
+** THE MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+** EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+** MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+** IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
+** CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+** TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+** MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS.
+*/
+
+/*
+** This header is generated from the Khronos Vulkan XML API Registry.
+**
+*/
+
+
+#define VK_VERSION_1_0 1
+#include "vk_platform.h"
+
+#define VK_MAKE_VERSION(major, minor, patch) \
+ ((major << 22) | (minor << 12) | patch)
+
+// Vulkan API version supported by this file
+#define VK_API_VERSION VK_MAKE_VERSION(1, 0, 1)
+
+
+#define VK_NULL_HANDLE 0
+
+
+
+#define VK_DEFINE_HANDLE(object) typedef struct object##_T* object;
+
+
+#if defined(__LP64__) || defined(_WIN64) || defined(__x86_64__) || defined(_M_X64) || defined(__ia64) || defined (_M_IA64) || defined(__aarch64__) || defined(__powerpc64__)
+ #define VK_DEFINE_NON_DISPATCHABLE_HANDLE(object) typedef struct object##_T *object;
+#else
+ #define VK_DEFINE_NON_DISPATCHABLE_HANDLE(object) typedef uint64_t object;
+#endif
+
+
+
+typedef uint32_t VkFlags;
+typedef uint32_t VkBool32;
+typedef uint64_t VkDeviceSize;
+typedef uint32_t VkSampleMask;
+
+VK_DEFINE_HANDLE(VkInstance)
+VK_DEFINE_HANDLE(VkPhysicalDevice)
+VK_DEFINE_HANDLE(VkDevice)
+VK_DEFINE_HANDLE(VkQueue)
+VK_DEFINE_NON_DISPATCHABLE_HANDLE(VkSemaphore)
+VK_DEFINE_HANDLE(VkCommandBuffer)
+VK_DEFINE_NON_DISPATCHABLE_HANDLE(VkFence)
+VK_DEFINE_NON_DISPATCHABLE_HANDLE(VkDeviceMemory)
+VK_DEFINE_NON_DISPATCHABLE_HANDLE(VkBuffer)
+VK_DEFINE_NON_DISPATCHABLE_HANDLE(VkImage)
+VK_DEFINE_NON_DISPATCHABLE_HANDLE(VkEvent)
+VK_DEFINE_NON_DISPATCHABLE_HANDLE(VkQueryPool)
+VK_DEFINE_NON_DISPATCHABLE_HANDLE(VkBufferView)
+VK_DEFINE_NON_DISPATCHABLE_HANDLE(VkImageView)
+VK_DEFINE_NON_DISPATCHABLE_HANDLE(VkShaderModule)
+VK_DEFINE_NON_DISPATCHABLE_HANDLE(VkPipelineCache)
+VK_DEFINE_NON_DISPATCHABLE_HANDLE(VkPipelineLayout)
+VK_DEFINE_NON_DISPATCHABLE_HANDLE(VkRenderPass)
+VK_DEFINE_NON_DISPATCHABLE_HANDLE(VkPipeline)
+VK_DEFINE_NON_DISPATCHABLE_HANDLE(VkDescriptorSetLayout)
+VK_DEFINE_NON_DISPATCHABLE_HANDLE(VkSampler)
+VK_DEFINE_NON_DISPATCHABLE_HANDLE(VkDescriptorPool)
+VK_DEFINE_NON_DISPATCHABLE_HANDLE(VkDescriptorSet)
+VK_DEFINE_NON_DISPATCHABLE_HANDLE(VkFramebuffer)
+VK_DEFINE_NON_DISPATCHABLE_HANDLE(VkCommandPool)
+
+#define VK_LOD_CLAMP_NONE 1000.0f
+#define VK_REMAINING_MIP_LEVELS (~0U)
+#define VK_REMAINING_ARRAY_LAYERS (~0U)
+#define VK_WHOLE_SIZE (~0ULL)
+#define VK_ATTACHMENT_UNUSED (~0U)
+#define VK_TRUE 1
+#define VK_FALSE 0
+#define VK_QUEUE_FAMILY_IGNORED (~0U)
+#define VK_SUBPASS_EXTERNAL (~0U)
+#define VK_MAX_PHYSICAL_DEVICE_NAME_SIZE 256
+#define VK_UUID_SIZE 16
+#define VK_MAX_MEMORY_TYPES 32
+#define VK_MAX_MEMORY_HEAPS 16
+#define VK_MAX_EXTENSION_NAME_SIZE 256
+#define VK_MAX_DESCRIPTION_SIZE 256
+
+
+typedef enum VkPipelineCacheHeaderVersion {
+ VK_PIPELINE_CACHE_HEADER_VERSION_ONE = 1,
+ VK_PIPELINE_CACHE_HEADER_VERSION_BEGIN_RANGE = VK_PIPELINE_CACHE_HEADER_VERSION_ONE,
+ VK_PIPELINE_CACHE_HEADER_VERSION_END_RANGE = VK_PIPELINE_CACHE_HEADER_VERSION_ONE,
+ VK_PIPELINE_CACHE_HEADER_VERSION_RANGE_SIZE = (VK_PIPELINE_CACHE_HEADER_VERSION_ONE - VK_PIPELINE_CACHE_HEADER_VERSION_ONE + 1),
+ VK_PIPELINE_CACHE_HEADER_VERSION_MAX_ENUM = 0x7FFFFFFF
+} VkPipelineCacheHeaderVersion;
+
+typedef enum VkResult {
+ VK_SUCCESS = 0,
+ VK_NOT_READY = 1,
+ VK_TIMEOUT = 2,
+ VK_EVENT_SET = 3,
+ VK_EVENT_RESET = 4,
+ VK_INCOMPLETE = 5,
+ VK_ERROR_OUT_OF_HOST_MEMORY = -1,
+ VK_ERROR_OUT_OF_DEVICE_MEMORY = -2,
+ VK_ERROR_INITIALIZATION_FAILED = -3,
+ VK_ERROR_DEVICE_LOST = -4,
+ VK_ERROR_MEMORY_MAP_FAILED = -5,
+ VK_ERROR_LAYER_NOT_PRESENT = -6,
+ VK_ERROR_EXTENSION_NOT_PRESENT = -7,
+ VK_ERROR_FEATURE_NOT_PRESENT = -8,
+ VK_ERROR_INCOMPATIBLE_DRIVER = -9,
+ VK_ERROR_TOO_MANY_OBJECTS = -10,
+ VK_ERROR_FORMAT_NOT_SUPPORTED = -11,
+ VK_ERROR_SURFACE_LOST_KHR = -1000000000,
+ VK_SUBOPTIMAL_KHR = 1000001003,
+ VK_ERROR_OUT_OF_DATE_KHR = -1000001004,
+ VK_ERROR_INCOMPATIBLE_DISPLAY_KHR = -1000003001,
+ VK_ERROR_NATIVE_WINDOW_IN_USE_KHR = -1000008000,
+ VK_ERROR_VALIDATION_FAILED_EXT = -1000011001,
+ VK_RESULT_BEGIN_RANGE = VK_ERROR_FORMAT_NOT_SUPPORTED,
+ VK_RESULT_END_RANGE = VK_INCOMPLETE,
+ VK_RESULT_RANGE_SIZE = (VK_INCOMPLETE - VK_ERROR_FORMAT_NOT_SUPPORTED + 1),
+ VK_RESULT_MAX_ENUM = 0x7FFFFFFF
+} VkResult;
+
+typedef enum VkStructureType {
+ VK_STRUCTURE_TYPE_APPLICATION_INFO = 0,
+ VK_STRUCTURE_TYPE_INSTANCE_CREATE_INFO = 1,
+ VK_STRUCTURE_TYPE_DEVICE_QUEUE_CREATE_INFO = 2,
+ VK_STRUCTURE_TYPE_DEVICE_CREATE_INFO = 3,
+ VK_STRUCTURE_TYPE_SUBMIT_INFO = 4,
+ VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO = 5,
+ VK_STRUCTURE_TYPE_MAPPED_MEMORY_RANGE = 6,
+ VK_STRUCTURE_TYPE_BIND_SPARSE_INFO = 7,
+ VK_STRUCTURE_TYPE_FENCE_CREATE_INFO = 8,
+ VK_STRUCTURE_TYPE_SEMAPHORE_CREATE_INFO = 9,
+ VK_STRUCTURE_TYPE_EVENT_CREATE_INFO = 10,
+ VK_STRUCTURE_TYPE_QUERY_POOL_CREATE_INFO = 11,
+ VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO = 12,
+ VK_STRUCTURE_TYPE_BUFFER_VIEW_CREATE_INFO = 13,
+ VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO = 14,
+ VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO = 15,
+ VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO = 16,
+ VK_STRUCTURE_TYPE_PIPELINE_CACHE_CREATE_INFO = 17,
+ VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO = 18,
+ VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO = 19,
+ VK_STRUCTURE_TYPE_PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO = 20,
+ VK_STRUCTURE_TYPE_PIPELINE_TESSELLATION_STATE_CREATE_INFO = 21,
+ VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO = 22,
+ VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_STATE_CREATE_INFO = 23,
+ VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO = 24,
+ VK_STRUCTURE_TYPE_PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO = 25,
+ VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO = 26,
+ VK_STRUCTURE_TYPE_PIPELINE_DYNAMIC_STATE_CREATE_INFO = 27,
+ VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO = 28,
+ VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO = 29,
+ VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO = 30,
+ VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO = 31,
+ VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO = 32,
+ VK_STRUCTURE_TYPE_DESCRIPTOR_POOL_CREATE_INFO = 33,
+ VK_STRUCTURE_TYPE_DESCRIPTOR_SET_ALLOCATE_INFO = 34,
+ VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET = 35,
+ VK_STRUCTURE_TYPE_COPY_DESCRIPTOR_SET = 36,
+ VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO = 37,
+ VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO = 38,
+ VK_STRUCTURE_TYPE_COMMAND_POOL_CREATE_INFO = 39,
+ VK_STRUCTURE_TYPE_COMMAND_BUFFER_ALLOCATE_INFO = 40,
+ VK_STRUCTURE_TYPE_COMMAND_BUFFER_INHERITANCE_INFO = 41,
+ VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO = 42,
+ VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO = 43,
+ VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER = 44,
+ VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER = 45,
+ VK_STRUCTURE_TYPE_MEMORY_BARRIER = 46,
+ VK_STRUCTURE_TYPE_LOADER_INSTANCE_CREATE_INFO = 47,
+ VK_STRUCTURE_TYPE_LOADER_DEVICE_CREATE_INFO = 48,
+ VK_STRUCTURE_TYPE_SWAPCHAIN_CREATE_INFO_KHR = 1000001000,
+ VK_STRUCTURE_TYPE_PRESENT_INFO_KHR = 1000001001,
+ VK_STRUCTURE_TYPE_DISPLAY_MODE_CREATE_INFO_KHR = 1000002000,
+ VK_STRUCTURE_TYPE_DISPLAY_SURFACE_CREATE_INFO_KHR = 1000002001,
+ VK_STRUCTURE_TYPE_DISPLAY_PRESENT_INFO_KHR = 1000003000,
+ VK_STRUCTURE_TYPE_XLIB_SURFACE_CREATE_INFO_KHR = 1000004000,
+ VK_STRUCTURE_TYPE_XCB_SURFACE_CREATE_INFO_KHR = 1000005000,
+ VK_STRUCTURE_TYPE_WAYLAND_SURFACE_CREATE_INFO_KHR = 1000006000,
+ VK_STRUCTURE_TYPE_MIR_SURFACE_CREATE_INFO_KHR = 1000007000,
+ VK_STRUCTURE_TYPE_ANDROID_SURFACE_CREATE_INFO_KHR = 1000008000,
+ VK_STRUCTURE_TYPE_WIN32_SURFACE_CREATE_INFO_KHR = 1000009000,
+ VK_STRUCTURE_TYPE_DEBUG_REPORT_CREATE_INFO_EXT = 1000011000,
+ VK_STRUCTURE_TYPE_BEGIN_RANGE = VK_STRUCTURE_TYPE_APPLICATION_INFO,
+ VK_STRUCTURE_TYPE_END_RANGE = VK_STRUCTURE_TYPE_LOADER_DEVICE_CREATE_INFO,
+ VK_STRUCTURE_TYPE_RANGE_SIZE = (VK_STRUCTURE_TYPE_LOADER_DEVICE_CREATE_INFO - VK_STRUCTURE_TYPE_APPLICATION_INFO + 1),
+ VK_STRUCTURE_TYPE_MAX_ENUM = 0x7FFFFFFF
+} VkStructureType;
+
+typedef enum VkSystemAllocationScope {
+ VK_SYSTEM_ALLOCATION_SCOPE_COMMAND = 0,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT = 1,
+ VK_SYSTEM_ALLOCATION_SCOPE_CACHE = 2,
+ VK_SYSTEM_ALLOCATION_SCOPE_DEVICE = 3,
+ VK_SYSTEM_ALLOCATION_SCOPE_INSTANCE = 4,
+ VK_SYSTEM_ALLOCATION_SCOPE_BEGIN_RANGE = VK_SYSTEM_ALLOCATION_SCOPE_COMMAND,
+ VK_SYSTEM_ALLOCATION_SCOPE_END_RANGE = VK_SYSTEM_ALLOCATION_SCOPE_INSTANCE,
+ VK_SYSTEM_ALLOCATION_SCOPE_RANGE_SIZE = (VK_SYSTEM_ALLOCATION_SCOPE_INSTANCE - VK_SYSTEM_ALLOCATION_SCOPE_COMMAND + 1),
+ VK_SYSTEM_ALLOCATION_SCOPE_MAX_ENUM = 0x7FFFFFFF
+} VkSystemAllocationScope;
+
+typedef enum VkInternalAllocationType {
+ VK_INTERNAL_ALLOCATION_TYPE_EXECUTABLE = 0,
+ VK_INTERNAL_ALLOCATION_TYPE_BEGIN_RANGE = VK_INTERNAL_ALLOCATION_TYPE_EXECUTABLE,
+ VK_INTERNAL_ALLOCATION_TYPE_END_RANGE = VK_INTERNAL_ALLOCATION_TYPE_EXECUTABLE,
+ VK_INTERNAL_ALLOCATION_TYPE_RANGE_SIZE = (VK_INTERNAL_ALLOCATION_TYPE_EXECUTABLE - VK_INTERNAL_ALLOCATION_TYPE_EXECUTABLE + 1),
+ VK_INTERNAL_ALLOCATION_TYPE_MAX_ENUM = 0x7FFFFFFF
+} VkInternalAllocationType;
+
+typedef enum VkFormat {
+ VK_FORMAT_UNDEFINED = 0,
+ VK_FORMAT_R4G4_UNORM_PACK8 = 1,
+ VK_FORMAT_R4G4B4A4_UNORM_PACK16 = 2,
+ VK_FORMAT_B4G4R4A4_UNORM_PACK16 = 3,
+ VK_FORMAT_R5G6B5_UNORM_PACK16 = 4,
+ VK_FORMAT_B5G6R5_UNORM_PACK16 = 5,
+ VK_FORMAT_R5G5B5A1_UNORM_PACK16 = 6,
+ VK_FORMAT_B5G5R5A1_UNORM_PACK16 = 7,
+ VK_FORMAT_A1R5G5B5_UNORM_PACK16 = 8,
+ VK_FORMAT_R8_UNORM = 9,
+ VK_FORMAT_R8_SNORM = 10,
+ VK_FORMAT_R8_USCALED = 11,
+ VK_FORMAT_R8_SSCALED = 12,
+ VK_FORMAT_R8_UINT = 13,
+ VK_FORMAT_R8_SINT = 14,
+ VK_FORMAT_R8_SRGB = 15,
+ VK_FORMAT_R8G8_UNORM = 16,
+ VK_FORMAT_R8G8_SNORM = 17,
+ VK_FORMAT_R8G8_USCALED = 18,
+ VK_FORMAT_R8G8_SSCALED = 19,
+ VK_FORMAT_R8G8_UINT = 20,
+ VK_FORMAT_R8G8_SINT = 21,
+ VK_FORMAT_R8G8_SRGB = 22,
+ VK_FORMAT_R8G8B8_UNORM = 23,
+ VK_FORMAT_R8G8B8_SNORM = 24,
+ VK_FORMAT_R8G8B8_USCALED = 25,
+ VK_FORMAT_R8G8B8_SSCALED = 26,
+ VK_FORMAT_R8G8B8_UINT = 27,
+ VK_FORMAT_R8G8B8_SINT = 28,
+ VK_FORMAT_R8G8B8_SRGB = 29,
+ VK_FORMAT_B8G8R8_UNORM = 30,
+ VK_FORMAT_B8G8R8_SNORM = 31,
+ VK_FORMAT_B8G8R8_USCALED = 32,
+ VK_FORMAT_B8G8R8_SSCALED = 33,
+ VK_FORMAT_B8G8R8_UINT = 34,
+ VK_FORMAT_B8G8R8_SINT = 35,
+ VK_FORMAT_B8G8R8_SRGB = 36,
+ VK_FORMAT_R8G8B8A8_UNORM = 37,
+ VK_FORMAT_R8G8B8A8_SNORM = 38,
+ VK_FORMAT_R8G8B8A8_USCALED = 39,
+ VK_FORMAT_R8G8B8A8_SSCALED = 40,
+ VK_FORMAT_R8G8B8A8_UINT = 41,
+ VK_FORMAT_R8G8B8A8_SINT = 42,
+ VK_FORMAT_R8G8B8A8_SRGB = 43,
+ VK_FORMAT_B8G8R8A8_UNORM = 44,
+ VK_FORMAT_B8G8R8A8_SNORM = 45,
+ VK_FORMAT_B8G8R8A8_USCALED = 46,
+ VK_FORMAT_B8G8R8A8_SSCALED = 47,
+ VK_FORMAT_B8G8R8A8_UINT = 48,
+ VK_FORMAT_B8G8R8A8_SINT = 49,
+ VK_FORMAT_B8G8R8A8_SRGB = 50,
+ VK_FORMAT_A8B8G8R8_UNORM_PACK32 = 51,
+ VK_FORMAT_A8B8G8R8_SNORM_PACK32 = 52,
+ VK_FORMAT_A8B8G8R8_USCALED_PACK32 = 53,
+ VK_FORMAT_A8B8G8R8_SSCALED_PACK32 = 54,
+ VK_FORMAT_A8B8G8R8_UINT_PACK32 = 55,
+ VK_FORMAT_A8B8G8R8_SINT_PACK32 = 56,
+ VK_FORMAT_A8B8G8R8_SRGB_PACK32 = 57,
+ VK_FORMAT_A2R10G10B10_UNORM_PACK32 = 58,
+ VK_FORMAT_A2R10G10B10_SNORM_PACK32 = 59,
+ VK_FORMAT_A2R10G10B10_USCALED_PACK32 = 60,
+ VK_FORMAT_A2R10G10B10_SSCALED_PACK32 = 61,
+ VK_FORMAT_A2R10G10B10_UINT_PACK32 = 62,
+ VK_FORMAT_A2R10G10B10_SINT_PACK32 = 63,
+ VK_FORMAT_A2B10G10R10_UNORM_PACK32 = 64,
+ VK_FORMAT_A2B10G10R10_SNORM_PACK32 = 65,
+ VK_FORMAT_A2B10G10R10_USCALED_PACK32 = 66,
+ VK_FORMAT_A2B10G10R10_SSCALED_PACK32 = 67,
+ VK_FORMAT_A2B10G10R10_UINT_PACK32 = 68,
+ VK_FORMAT_A2B10G10R10_SINT_PACK32 = 69,
+ VK_FORMAT_R16_UNORM = 70,
+ VK_FORMAT_R16_SNORM = 71,
+ VK_FORMAT_R16_USCALED = 72,
+ VK_FORMAT_R16_SSCALED = 73,
+ VK_FORMAT_R16_UINT = 74,
+ VK_FORMAT_R16_SINT = 75,
+ VK_FORMAT_R16_SFLOAT = 76,
+ VK_FORMAT_R16G16_UNORM = 77,
+ VK_FORMAT_R16G16_SNORM = 78,
+ VK_FORMAT_R16G16_USCALED = 79,
+ VK_FORMAT_R16G16_SSCALED = 80,
+ VK_FORMAT_R16G16_UINT = 81,
+ VK_FORMAT_R16G16_SINT = 82,
+ VK_FORMAT_R16G16_SFLOAT = 83,
+ VK_FORMAT_R16G16B16_UNORM = 84,
+ VK_FORMAT_R16G16B16_SNORM = 85,
+ VK_FORMAT_R16G16B16_USCALED = 86,
+ VK_FORMAT_R16G16B16_SSCALED = 87,
+ VK_FORMAT_R16G16B16_UINT = 88,
+ VK_FORMAT_R16G16B16_SINT = 89,
+ VK_FORMAT_R16G16B16_SFLOAT = 90,
+ VK_FORMAT_R16G16B16A16_UNORM = 91,
+ VK_FORMAT_R16G16B16A16_SNORM = 92,
+ VK_FORMAT_R16G16B16A16_USCALED = 93,
+ VK_FORMAT_R16G16B16A16_SSCALED = 94,
+ VK_FORMAT_R16G16B16A16_UINT = 95,
+ VK_FORMAT_R16G16B16A16_SINT = 96,
+ VK_FORMAT_R16G16B16A16_SFLOAT = 97,
+ VK_FORMAT_R32_UINT = 98,
+ VK_FORMAT_R32_SINT = 99,
+ VK_FORMAT_R32_SFLOAT = 100,
+ VK_FORMAT_R32G32_UINT = 101,
+ VK_FORMAT_R32G32_SINT = 102,
+ VK_FORMAT_R32G32_SFLOAT = 103,
+ VK_FORMAT_R32G32B32_UINT = 104,
+ VK_FORMAT_R32G32B32_SINT = 105,
+ VK_FORMAT_R32G32B32_SFLOAT = 106,
+ VK_FORMAT_R32G32B32A32_UINT = 107,
+ VK_FORMAT_R32G32B32A32_SINT = 108,
+ VK_FORMAT_R32G32B32A32_SFLOAT = 109,
+ VK_FORMAT_R64_UINT = 110,
+ VK_FORMAT_R64_SINT = 111,
+ VK_FORMAT_R64_SFLOAT = 112,
+ VK_FORMAT_R64G64_UINT = 113,
+ VK_FORMAT_R64G64_SINT = 114,
+ VK_FORMAT_R64G64_SFLOAT = 115,
+ VK_FORMAT_R64G64B64_UINT = 116,
+ VK_FORMAT_R64G64B64_SINT = 117,
+ VK_FORMAT_R64G64B64_SFLOAT = 118,
+ VK_FORMAT_R64G64B64A64_UINT = 119,
+ VK_FORMAT_R64G64B64A64_SINT = 120,
+ VK_FORMAT_R64G64B64A64_SFLOAT = 121,
+ VK_FORMAT_B10G11R11_UFLOAT_PACK32 = 122,
+ VK_FORMAT_E5B9G9R9_UFLOAT_PACK32 = 123,
+ VK_FORMAT_D16_UNORM = 124,
+ VK_FORMAT_X8_D24_UNORM_PACK32 = 125,
+ VK_FORMAT_D32_SFLOAT = 126,
+ VK_FORMAT_S8_UINT = 127,
+ VK_FORMAT_D16_UNORM_S8_UINT = 128,
+ VK_FORMAT_D24_UNORM_S8_UINT = 129,
+ VK_FORMAT_D32_SFLOAT_S8_UINT = 130,
+ VK_FORMAT_BC1_RGB_UNORM_BLOCK = 131,
+ VK_FORMAT_BC1_RGB_SRGB_BLOCK = 132,
+ VK_FORMAT_BC1_RGBA_UNORM_BLOCK = 133,
+ VK_FORMAT_BC1_RGBA_SRGB_BLOCK = 134,
+ VK_FORMAT_BC2_UNORM_BLOCK = 135,
+ VK_FORMAT_BC2_SRGB_BLOCK = 136,
+ VK_FORMAT_BC3_UNORM_BLOCK = 137,
+ VK_FORMAT_BC3_SRGB_BLOCK = 138,
+ VK_FORMAT_BC4_UNORM_BLOCK = 139,
+ VK_FORMAT_BC4_SNORM_BLOCK = 140,
+ VK_FORMAT_BC5_UNORM_BLOCK = 141,
+ VK_FORMAT_BC5_SNORM_BLOCK = 142,
+ VK_FORMAT_BC6H_UFLOAT_BLOCK = 143,
+ VK_FORMAT_BC6H_SFLOAT_BLOCK = 144,
+ VK_FORMAT_BC7_UNORM_BLOCK = 145,
+ VK_FORMAT_BC7_SRGB_BLOCK = 146,
+ VK_FORMAT_ETC2_R8G8B8_UNORM_BLOCK = 147,
+ VK_FORMAT_ETC2_R8G8B8_SRGB_BLOCK = 148,
+ VK_FORMAT_ETC2_R8G8B8A1_UNORM_BLOCK = 149,
+ VK_FORMAT_ETC2_R8G8B8A1_SRGB_BLOCK = 150,
+ VK_FORMAT_ETC2_R8G8B8A8_UNORM_BLOCK = 151,
+ VK_FORMAT_ETC2_R8G8B8A8_SRGB_BLOCK = 152,
+ VK_FORMAT_EAC_R11_UNORM_BLOCK = 153,
+ VK_FORMAT_EAC_R11_SNORM_BLOCK = 154,
+ VK_FORMAT_EAC_R11G11_UNORM_BLOCK = 155,
+ VK_FORMAT_EAC_R11G11_SNORM_BLOCK = 156,
+ VK_FORMAT_ASTC_4x4_UNORM_BLOCK = 157,
+ VK_FORMAT_ASTC_4x4_SRGB_BLOCK = 158,
+ VK_FORMAT_ASTC_5x4_UNORM_BLOCK = 159,
+ VK_FORMAT_ASTC_5x4_SRGB_BLOCK = 160,
+ VK_FORMAT_ASTC_5x5_UNORM_BLOCK = 161,
+ VK_FORMAT_ASTC_5x5_SRGB_BLOCK = 162,
+ VK_FORMAT_ASTC_6x5_UNORM_BLOCK = 163,
+ VK_FORMAT_ASTC_6x5_SRGB_BLOCK = 164,
+ VK_FORMAT_ASTC_6x6_UNORM_BLOCK = 165,
+ VK_FORMAT_ASTC_6x6_SRGB_BLOCK = 166,
+ VK_FORMAT_ASTC_8x5_UNORM_BLOCK = 167,
+ VK_FORMAT_ASTC_8x5_SRGB_BLOCK = 168,
+ VK_FORMAT_ASTC_8x6_UNORM_BLOCK = 169,
+ VK_FORMAT_ASTC_8x6_SRGB_BLOCK = 170,
+ VK_FORMAT_ASTC_8x8_UNORM_BLOCK = 171,
+ VK_FORMAT_ASTC_8x8_SRGB_BLOCK = 172,
+ VK_FORMAT_ASTC_10x5_UNORM_BLOCK = 173,
+ VK_FORMAT_ASTC_10x5_SRGB_BLOCK = 174,
+ VK_FORMAT_ASTC_10x6_UNORM_BLOCK = 175,
+ VK_FORMAT_ASTC_10x6_SRGB_BLOCK = 176,
+ VK_FORMAT_ASTC_10x8_UNORM_BLOCK = 177,
+ VK_FORMAT_ASTC_10x8_SRGB_BLOCK = 178,
+ VK_FORMAT_ASTC_10x10_UNORM_BLOCK = 179,
+ VK_FORMAT_ASTC_10x10_SRGB_BLOCK = 180,
+ VK_FORMAT_ASTC_12x10_UNORM_BLOCK = 181,
+ VK_FORMAT_ASTC_12x10_SRGB_BLOCK = 182,
+ VK_FORMAT_ASTC_12x12_UNORM_BLOCK = 183,
+ VK_FORMAT_ASTC_12x12_SRGB_BLOCK = 184,
+ VK_FORMAT_BEGIN_RANGE = VK_FORMAT_UNDEFINED,
+ VK_FORMAT_END_RANGE = VK_FORMAT_ASTC_12x12_SRGB_BLOCK,
+ VK_FORMAT_RANGE_SIZE = (VK_FORMAT_ASTC_12x12_SRGB_BLOCK - VK_FORMAT_UNDEFINED + 1),
+ VK_FORMAT_MAX_ENUM = 0x7FFFFFFF
+} VkFormat;
+
+typedef enum VkImageType {
+ VK_IMAGE_TYPE_1D = 0,
+ VK_IMAGE_TYPE_2D = 1,
+ VK_IMAGE_TYPE_3D = 2,
+ VK_IMAGE_TYPE_BEGIN_RANGE = VK_IMAGE_TYPE_1D,
+ VK_IMAGE_TYPE_END_RANGE = VK_IMAGE_TYPE_3D,
+ VK_IMAGE_TYPE_RANGE_SIZE = (VK_IMAGE_TYPE_3D - VK_IMAGE_TYPE_1D + 1),
+ VK_IMAGE_TYPE_MAX_ENUM = 0x7FFFFFFF
+} VkImageType;
+
+typedef enum VkImageTiling {
+ VK_IMAGE_TILING_OPTIMAL = 0,
+ VK_IMAGE_TILING_LINEAR = 1,
+ VK_IMAGE_TILING_BEGIN_RANGE = VK_IMAGE_TILING_OPTIMAL,
+ VK_IMAGE_TILING_END_RANGE = VK_IMAGE_TILING_LINEAR,
+ VK_IMAGE_TILING_RANGE_SIZE = (VK_IMAGE_TILING_LINEAR - VK_IMAGE_TILING_OPTIMAL + 1),
+ VK_IMAGE_TILING_MAX_ENUM = 0x7FFFFFFF
+} VkImageTiling;
+
+typedef enum VkPhysicalDeviceType {
+ VK_PHYSICAL_DEVICE_TYPE_OTHER = 0,
+ VK_PHYSICAL_DEVICE_TYPE_INTEGRATED_GPU = 1,
+ VK_PHYSICAL_DEVICE_TYPE_DISCRETE_GPU = 2,
+ VK_PHYSICAL_DEVICE_TYPE_VIRTUAL_GPU = 3,
+ VK_PHYSICAL_DEVICE_TYPE_CPU = 4,
+ VK_PHYSICAL_DEVICE_TYPE_BEGIN_RANGE = VK_PHYSICAL_DEVICE_TYPE_OTHER,
+ VK_PHYSICAL_DEVICE_TYPE_END_RANGE = VK_PHYSICAL_DEVICE_TYPE_CPU,
+ VK_PHYSICAL_DEVICE_TYPE_RANGE_SIZE = (VK_PHYSICAL_DEVICE_TYPE_CPU - VK_PHYSICAL_DEVICE_TYPE_OTHER + 1),
+ VK_PHYSICAL_DEVICE_TYPE_MAX_ENUM = 0x7FFFFFFF
+} VkPhysicalDeviceType;
+
+typedef enum VkQueryType {
+ VK_QUERY_TYPE_OCCLUSION = 0,
+ VK_QUERY_TYPE_PIPELINE_STATISTICS = 1,
+ VK_QUERY_TYPE_TIMESTAMP = 2,
+ VK_QUERY_TYPE_BEGIN_RANGE = VK_QUERY_TYPE_OCCLUSION,
+ VK_QUERY_TYPE_END_RANGE = VK_QUERY_TYPE_TIMESTAMP,
+ VK_QUERY_TYPE_RANGE_SIZE = (VK_QUERY_TYPE_TIMESTAMP - VK_QUERY_TYPE_OCCLUSION + 1),
+ VK_QUERY_TYPE_MAX_ENUM = 0x7FFFFFFF
+} VkQueryType;
+
+typedef enum VkSharingMode {
+ VK_SHARING_MODE_EXCLUSIVE = 0,
+ VK_SHARING_MODE_CONCURRENT = 1,
+ VK_SHARING_MODE_BEGIN_RANGE = VK_SHARING_MODE_EXCLUSIVE,
+ VK_SHARING_MODE_END_RANGE = VK_SHARING_MODE_CONCURRENT,
+ VK_SHARING_MODE_RANGE_SIZE = (VK_SHARING_MODE_CONCURRENT - VK_SHARING_MODE_EXCLUSIVE + 1),
+ VK_SHARING_MODE_MAX_ENUM = 0x7FFFFFFF
+} VkSharingMode;
+
+typedef enum VkImageLayout {
+ VK_IMAGE_LAYOUT_UNDEFINED = 0,
+ VK_IMAGE_LAYOUT_GENERAL = 1,
+ VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL = 2,
+ VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL = 3,
+ VK_IMAGE_LAYOUT_DEPTH_STENCIL_READ_ONLY_OPTIMAL = 4,
+ VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL = 5,
+ VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL = 6,
+ VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL = 7,
+ VK_IMAGE_LAYOUT_PREINITIALIZED = 8,
+ VK_IMAGE_LAYOUT_PRESENT_SRC_KHR = 1000001002,
+ VK_IMAGE_LAYOUT_BEGIN_RANGE = VK_IMAGE_LAYOUT_UNDEFINED,
+ VK_IMAGE_LAYOUT_END_RANGE = VK_IMAGE_LAYOUT_PREINITIALIZED,
+ VK_IMAGE_LAYOUT_RANGE_SIZE = (VK_IMAGE_LAYOUT_PREINITIALIZED - VK_IMAGE_LAYOUT_UNDEFINED + 1),
+ VK_IMAGE_LAYOUT_MAX_ENUM = 0x7FFFFFFF
+} VkImageLayout;
+
+typedef enum VkImageViewType {
+ VK_IMAGE_VIEW_TYPE_1D = 0,
+ VK_IMAGE_VIEW_TYPE_2D = 1,
+ VK_IMAGE_VIEW_TYPE_3D = 2,
+ VK_IMAGE_VIEW_TYPE_CUBE = 3,
+ VK_IMAGE_VIEW_TYPE_1D_ARRAY = 4,
+ VK_IMAGE_VIEW_TYPE_2D_ARRAY = 5,
+ VK_IMAGE_VIEW_TYPE_CUBE_ARRAY = 6,
+ VK_IMAGE_VIEW_TYPE_BEGIN_RANGE = VK_IMAGE_VIEW_TYPE_1D,
+ VK_IMAGE_VIEW_TYPE_END_RANGE = VK_IMAGE_VIEW_TYPE_CUBE_ARRAY,
+ VK_IMAGE_VIEW_TYPE_RANGE_SIZE = (VK_IMAGE_VIEW_TYPE_CUBE_ARRAY - VK_IMAGE_VIEW_TYPE_1D + 1),
+ VK_IMAGE_VIEW_TYPE_MAX_ENUM = 0x7FFFFFFF
+} VkImageViewType;
+
+typedef enum VkComponentSwizzle {
+ VK_COMPONENT_SWIZZLE_IDENTITY = 0,
+ VK_COMPONENT_SWIZZLE_ZERO = 1,
+ VK_COMPONENT_SWIZZLE_ONE = 2,
+ VK_COMPONENT_SWIZZLE_R = 3,
+ VK_COMPONENT_SWIZZLE_G = 4,
+ VK_COMPONENT_SWIZZLE_B = 5,
+ VK_COMPONENT_SWIZZLE_A = 6,
+ VK_COMPONENT_SWIZZLE_BEGIN_RANGE = VK_COMPONENT_SWIZZLE_IDENTITY,
+ VK_COMPONENT_SWIZZLE_END_RANGE = VK_COMPONENT_SWIZZLE_A,
+ VK_COMPONENT_SWIZZLE_RANGE_SIZE = (VK_COMPONENT_SWIZZLE_A - VK_COMPONENT_SWIZZLE_IDENTITY + 1),
+ VK_COMPONENT_SWIZZLE_MAX_ENUM = 0x7FFFFFFF
+} VkComponentSwizzle;
+
+typedef enum VkVertexInputRate {
+ VK_VERTEX_INPUT_RATE_VERTEX = 0,
+ VK_VERTEX_INPUT_RATE_INSTANCE = 1,
+ VK_VERTEX_INPUT_RATE_BEGIN_RANGE = VK_VERTEX_INPUT_RATE_VERTEX,
+ VK_VERTEX_INPUT_RATE_END_RANGE = VK_VERTEX_INPUT_RATE_INSTANCE,
+ VK_VERTEX_INPUT_RATE_RANGE_SIZE = (VK_VERTEX_INPUT_RATE_INSTANCE - VK_VERTEX_INPUT_RATE_VERTEX + 1),
+ VK_VERTEX_INPUT_RATE_MAX_ENUM = 0x7FFFFFFF
+} VkVertexInputRate;
+
+typedef enum VkPrimitiveTopology {
+ VK_PRIMITIVE_TOPOLOGY_POINT_LIST = 0,
+ VK_PRIMITIVE_TOPOLOGY_LINE_LIST = 1,
+ VK_PRIMITIVE_TOPOLOGY_LINE_STRIP = 2,
+ VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST = 3,
+ VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP = 4,
+ VK_PRIMITIVE_TOPOLOGY_TRIANGLE_FAN = 5,
+ VK_PRIMITIVE_TOPOLOGY_LINE_LIST_WITH_ADJACENCY = 6,
+ VK_PRIMITIVE_TOPOLOGY_LINE_STRIP_WITH_ADJACENCY = 7,
+ VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST_WITH_ADJACENCY = 8,
+ VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP_WITH_ADJACENCY = 9,
+ VK_PRIMITIVE_TOPOLOGY_PATCH_LIST = 10,
+ VK_PRIMITIVE_TOPOLOGY_BEGIN_RANGE = VK_PRIMITIVE_TOPOLOGY_POINT_LIST,
+ VK_PRIMITIVE_TOPOLOGY_END_RANGE = VK_PRIMITIVE_TOPOLOGY_PATCH_LIST,
+ VK_PRIMITIVE_TOPOLOGY_RANGE_SIZE = (VK_PRIMITIVE_TOPOLOGY_PATCH_LIST - VK_PRIMITIVE_TOPOLOGY_POINT_LIST + 1),
+ VK_PRIMITIVE_TOPOLOGY_MAX_ENUM = 0x7FFFFFFF
+} VkPrimitiveTopology;
+
+typedef enum VkPolygonMode {
+ VK_POLYGON_MODE_FILL = 0,
+ VK_POLYGON_MODE_LINE = 1,
+ VK_POLYGON_MODE_POINT = 2,
+ VK_POLYGON_MODE_BEGIN_RANGE = VK_POLYGON_MODE_FILL,
+ VK_POLYGON_MODE_END_RANGE = VK_POLYGON_MODE_POINT,
+ VK_POLYGON_MODE_RANGE_SIZE = (VK_POLYGON_MODE_POINT - VK_POLYGON_MODE_FILL + 1),
+ VK_POLYGON_MODE_MAX_ENUM = 0x7FFFFFFF
+} VkPolygonMode;
+
+typedef enum VkFrontFace {
+ VK_FRONT_FACE_COUNTER_CLOCKWISE = 0,
+ VK_FRONT_FACE_CLOCKWISE = 1,
+ VK_FRONT_FACE_BEGIN_RANGE = VK_FRONT_FACE_COUNTER_CLOCKWISE,
+ VK_FRONT_FACE_END_RANGE = VK_FRONT_FACE_CLOCKWISE,
+ VK_FRONT_FACE_RANGE_SIZE = (VK_FRONT_FACE_CLOCKWISE - VK_FRONT_FACE_COUNTER_CLOCKWISE + 1),
+ VK_FRONT_FACE_MAX_ENUM = 0x7FFFFFFF
+} VkFrontFace;
+
+typedef enum VkCompareOp {
+ VK_COMPARE_OP_NEVER = 0,
+ VK_COMPARE_OP_LESS = 1,
+ VK_COMPARE_OP_EQUAL = 2,
+ VK_COMPARE_OP_LESS_OR_EQUAL = 3,
+ VK_COMPARE_OP_GREATER = 4,
+ VK_COMPARE_OP_NOT_EQUAL = 5,
+ VK_COMPARE_OP_GREATER_OR_EQUAL = 6,
+ VK_COMPARE_OP_ALWAYS = 7,
+ VK_COMPARE_OP_BEGIN_RANGE = VK_COMPARE_OP_NEVER,
+ VK_COMPARE_OP_END_RANGE = VK_COMPARE_OP_ALWAYS,
+ VK_COMPARE_OP_RANGE_SIZE = (VK_COMPARE_OP_ALWAYS - VK_COMPARE_OP_NEVER + 1),
+ VK_COMPARE_OP_MAX_ENUM = 0x7FFFFFFF
+} VkCompareOp;
+
+typedef enum VkStencilOp {
+ VK_STENCIL_OP_KEEP = 0,
+ VK_STENCIL_OP_ZERO = 1,
+ VK_STENCIL_OP_REPLACE = 2,
+ VK_STENCIL_OP_INCREMENT_AND_CLAMP = 3,
+ VK_STENCIL_OP_DECREMENT_AND_CLAMP = 4,
+ VK_STENCIL_OP_INVERT = 5,
+ VK_STENCIL_OP_INCREMENT_AND_WRAP = 6,
+ VK_STENCIL_OP_DECREMENT_AND_WRAP = 7,
+ VK_STENCIL_OP_BEGIN_RANGE = VK_STENCIL_OP_KEEP,
+ VK_STENCIL_OP_END_RANGE = VK_STENCIL_OP_DECREMENT_AND_WRAP,
+ VK_STENCIL_OP_RANGE_SIZE = (VK_STENCIL_OP_DECREMENT_AND_WRAP - VK_STENCIL_OP_KEEP + 1),
+ VK_STENCIL_OP_MAX_ENUM = 0x7FFFFFFF
+} VkStencilOp;
+
+typedef enum VkLogicOp {
+ VK_LOGIC_OP_CLEAR = 0,
+ VK_LOGIC_OP_AND = 1,
+ VK_LOGIC_OP_AND_REVERSE = 2,
+ VK_LOGIC_OP_COPY = 3,
+ VK_LOGIC_OP_AND_INVERTED = 4,
+ VK_LOGIC_OP_NO_OP = 5,
+ VK_LOGIC_OP_XOR = 6,
+ VK_LOGIC_OP_OR = 7,
+ VK_LOGIC_OP_NOR = 8,
+ VK_LOGIC_OP_EQUIVALENT = 9,
+ VK_LOGIC_OP_INVERT = 10,
+ VK_LOGIC_OP_OR_REVERSE = 11,
+ VK_LOGIC_OP_COPY_INVERTED = 12,
+ VK_LOGIC_OP_OR_INVERTED = 13,
+ VK_LOGIC_OP_NAND = 14,
+ VK_LOGIC_OP_SET = 15,
+ VK_LOGIC_OP_BEGIN_RANGE = VK_LOGIC_OP_CLEAR,
+ VK_LOGIC_OP_END_RANGE = VK_LOGIC_OP_SET,
+ VK_LOGIC_OP_RANGE_SIZE = (VK_LOGIC_OP_SET - VK_LOGIC_OP_CLEAR + 1),
+ VK_LOGIC_OP_MAX_ENUM = 0x7FFFFFFF
+} VkLogicOp;
+
+typedef enum VkBlendFactor {
+ VK_BLEND_FACTOR_ZERO = 0,
+ VK_BLEND_FACTOR_ONE = 1,
+ VK_BLEND_FACTOR_SRC_COLOR = 2,
+ VK_BLEND_FACTOR_ONE_MINUS_SRC_COLOR = 3,
+ VK_BLEND_FACTOR_DST_COLOR = 4,
+ VK_BLEND_FACTOR_ONE_MINUS_DST_COLOR = 5,
+ VK_BLEND_FACTOR_SRC_ALPHA = 6,
+ VK_BLEND_FACTOR_ONE_MINUS_SRC_ALPHA = 7,
+ VK_BLEND_FACTOR_DST_ALPHA = 8,
+ VK_BLEND_FACTOR_ONE_MINUS_DST_ALPHA = 9,
+ VK_BLEND_FACTOR_CONSTANT_COLOR = 10,
+ VK_BLEND_FACTOR_ONE_MINUS_CONSTANT_COLOR = 11,
+ VK_BLEND_FACTOR_CONSTANT_ALPHA = 12,
+ VK_BLEND_FACTOR_ONE_MINUS_CONSTANT_ALPHA = 13,
+ VK_BLEND_FACTOR_SRC_ALPHA_SATURATE = 14,
+ VK_BLEND_FACTOR_SRC1_COLOR = 15,
+ VK_BLEND_FACTOR_ONE_MINUS_SRC1_COLOR = 16,
+ VK_BLEND_FACTOR_SRC1_ALPHA = 17,
+ VK_BLEND_FACTOR_ONE_MINUS_SRC1_ALPHA = 18,
+ VK_BLEND_FACTOR_BEGIN_RANGE = VK_BLEND_FACTOR_ZERO,
+ VK_BLEND_FACTOR_END_RANGE = VK_BLEND_FACTOR_ONE_MINUS_SRC1_ALPHA,
+ VK_BLEND_FACTOR_RANGE_SIZE = (VK_BLEND_FACTOR_ONE_MINUS_SRC1_ALPHA - VK_BLEND_FACTOR_ZERO + 1),
+ VK_BLEND_FACTOR_MAX_ENUM = 0x7FFFFFFF
+} VkBlendFactor;
+
+typedef enum VkBlendOp {
+ VK_BLEND_OP_ADD = 0,
+ VK_BLEND_OP_SUBTRACT = 1,
+ VK_BLEND_OP_REVERSE_SUBTRACT = 2,
+ VK_BLEND_OP_MIN = 3,
+ VK_BLEND_OP_MAX = 4,
+ VK_BLEND_OP_BEGIN_RANGE = VK_BLEND_OP_ADD,
+ VK_BLEND_OP_END_RANGE = VK_BLEND_OP_MAX,
+ VK_BLEND_OP_RANGE_SIZE = (VK_BLEND_OP_MAX - VK_BLEND_OP_ADD + 1),
+ VK_BLEND_OP_MAX_ENUM = 0x7FFFFFFF
+} VkBlendOp;
+
+typedef enum VkDynamicState {
+ VK_DYNAMIC_STATE_VIEWPORT = 0,
+ VK_DYNAMIC_STATE_SCISSOR = 1,
+ VK_DYNAMIC_STATE_LINE_WIDTH = 2,
+ VK_DYNAMIC_STATE_DEPTH_BIAS = 3,
+ VK_DYNAMIC_STATE_BLEND_CONSTANTS = 4,
+ VK_DYNAMIC_STATE_DEPTH_BOUNDS = 5,
+ VK_DYNAMIC_STATE_STENCIL_COMPARE_MASK = 6,
+ VK_DYNAMIC_STATE_STENCIL_WRITE_MASK = 7,
+ VK_DYNAMIC_STATE_STENCIL_REFERENCE = 8,
+ VK_DYNAMIC_STATE_BEGIN_RANGE = VK_DYNAMIC_STATE_VIEWPORT,
+ VK_DYNAMIC_STATE_END_RANGE = VK_DYNAMIC_STATE_STENCIL_REFERENCE,
+ VK_DYNAMIC_STATE_RANGE_SIZE = (VK_DYNAMIC_STATE_STENCIL_REFERENCE - VK_DYNAMIC_STATE_VIEWPORT + 1),
+ VK_DYNAMIC_STATE_MAX_ENUM = 0x7FFFFFFF
+} VkDynamicState;
+
+typedef enum VkFilter {
+ VK_FILTER_NEAREST = 0,
+ VK_FILTER_LINEAR = 1,
+ VK_FILTER_BEGIN_RANGE = VK_FILTER_NEAREST,
+ VK_FILTER_END_RANGE = VK_FILTER_LINEAR,
+ VK_FILTER_RANGE_SIZE = (VK_FILTER_LINEAR - VK_FILTER_NEAREST + 1),
+ VK_FILTER_MAX_ENUM = 0x7FFFFFFF
+} VkFilter;
+
+typedef enum VkSamplerMipmapMode {
+ VK_SAMPLER_MIPMAP_MODE_NEAREST = 0,
+ VK_SAMPLER_MIPMAP_MODE_LINEAR = 1,
+ VK_SAMPLER_MIPMAP_MODE_BEGIN_RANGE = VK_SAMPLER_MIPMAP_MODE_NEAREST,
+ VK_SAMPLER_MIPMAP_MODE_END_RANGE = VK_SAMPLER_MIPMAP_MODE_LINEAR,
+ VK_SAMPLER_MIPMAP_MODE_RANGE_SIZE = (VK_SAMPLER_MIPMAP_MODE_LINEAR - VK_SAMPLER_MIPMAP_MODE_NEAREST + 1),
+ VK_SAMPLER_MIPMAP_MODE_MAX_ENUM = 0x7FFFFFFF
+} VkSamplerMipmapMode;
+
+typedef enum VkSamplerAddressMode {
+ VK_SAMPLER_ADDRESS_MODE_REPEAT = 0,
+ VK_SAMPLER_ADDRESS_MODE_MIRRORED_REPEAT = 1,
+ VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE = 2,
+ VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER = 3,
+ VK_SAMPLER_ADDRESS_MODE_MIRROR_CLAMP_TO_EDGE = 4,
+ VK_SAMPLER_ADDRESS_MODE_BEGIN_RANGE = VK_SAMPLER_ADDRESS_MODE_REPEAT,
+ VK_SAMPLER_ADDRESS_MODE_END_RANGE = VK_SAMPLER_ADDRESS_MODE_MIRROR_CLAMP_TO_EDGE,
+ VK_SAMPLER_ADDRESS_MODE_RANGE_SIZE = (VK_SAMPLER_ADDRESS_MODE_MIRROR_CLAMP_TO_EDGE - VK_SAMPLER_ADDRESS_MODE_REPEAT + 1),
+ VK_SAMPLER_ADDRESS_MODE_MAX_ENUM = 0x7FFFFFFF
+} VkSamplerAddressMode;
+
+typedef enum VkBorderColor {
+ VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK = 0,
+ VK_BORDER_COLOR_INT_TRANSPARENT_BLACK = 1,
+ VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK = 2,
+ VK_BORDER_COLOR_INT_OPAQUE_BLACK = 3,
+ VK_BORDER_COLOR_FLOAT_OPAQUE_WHITE = 4,
+ VK_BORDER_COLOR_INT_OPAQUE_WHITE = 5,
+ VK_BORDER_COLOR_BEGIN_RANGE = VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK,
+ VK_BORDER_COLOR_END_RANGE = VK_BORDER_COLOR_INT_OPAQUE_WHITE,
+ VK_BORDER_COLOR_RANGE_SIZE = (VK_BORDER_COLOR_INT_OPAQUE_WHITE - VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK + 1),
+ VK_BORDER_COLOR_MAX_ENUM = 0x7FFFFFFF
+} VkBorderColor;
+
+typedef enum VkDescriptorType {
+ VK_DESCRIPTOR_TYPE_SAMPLER = 0,
+ VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER = 1,
+ VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE = 2,
+ VK_DESCRIPTOR_TYPE_STORAGE_IMAGE = 3,
+ VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER = 4,
+ VK_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER = 5,
+ VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER = 6,
+ VK_DESCRIPTOR_TYPE_STORAGE_BUFFER = 7,
+ VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC = 8,
+ VK_DESCRIPTOR_TYPE_STORAGE_BUFFER_DYNAMIC = 9,
+ VK_DESCRIPTOR_TYPE_INPUT_ATTACHMENT = 10,
+ VK_DESCRIPTOR_TYPE_BEGIN_RANGE = VK_DESCRIPTOR_TYPE_SAMPLER,
+ VK_DESCRIPTOR_TYPE_END_RANGE = VK_DESCRIPTOR_TYPE_INPUT_ATTACHMENT,
+ VK_DESCRIPTOR_TYPE_RANGE_SIZE = (VK_DESCRIPTOR_TYPE_INPUT_ATTACHMENT - VK_DESCRIPTOR_TYPE_SAMPLER + 1),
+ VK_DESCRIPTOR_TYPE_MAX_ENUM = 0x7FFFFFFF
+} VkDescriptorType;
+
+typedef enum VkAttachmentLoadOp {
+ VK_ATTACHMENT_LOAD_OP_LOAD = 0,
+ VK_ATTACHMENT_LOAD_OP_CLEAR = 1,
+ VK_ATTACHMENT_LOAD_OP_DONT_CARE = 2,
+ VK_ATTACHMENT_LOAD_OP_BEGIN_RANGE = VK_ATTACHMENT_LOAD_OP_LOAD,
+ VK_ATTACHMENT_LOAD_OP_END_RANGE = VK_ATTACHMENT_LOAD_OP_DONT_CARE,
+ VK_ATTACHMENT_LOAD_OP_RANGE_SIZE = (VK_ATTACHMENT_LOAD_OP_DONT_CARE - VK_ATTACHMENT_LOAD_OP_LOAD + 1),
+ VK_ATTACHMENT_LOAD_OP_MAX_ENUM = 0x7FFFFFFF
+} VkAttachmentLoadOp;
+
+typedef enum VkAttachmentStoreOp {
+ VK_ATTACHMENT_STORE_OP_STORE = 0,
+ VK_ATTACHMENT_STORE_OP_DONT_CARE = 1,
+ VK_ATTACHMENT_STORE_OP_BEGIN_RANGE = VK_ATTACHMENT_STORE_OP_STORE,
+ VK_ATTACHMENT_STORE_OP_END_RANGE = VK_ATTACHMENT_STORE_OP_DONT_CARE,
+ VK_ATTACHMENT_STORE_OP_RANGE_SIZE = (VK_ATTACHMENT_STORE_OP_DONT_CARE - VK_ATTACHMENT_STORE_OP_STORE + 1),
+ VK_ATTACHMENT_STORE_OP_MAX_ENUM = 0x7FFFFFFF
+} VkAttachmentStoreOp;
+
+typedef enum VkPipelineBindPoint {
+ VK_PIPELINE_BIND_POINT_GRAPHICS = 0,
+ VK_PIPELINE_BIND_POINT_COMPUTE = 1,
+ VK_PIPELINE_BIND_POINT_BEGIN_RANGE = VK_PIPELINE_BIND_POINT_GRAPHICS,
+ VK_PIPELINE_BIND_POINT_END_RANGE = VK_PIPELINE_BIND_POINT_COMPUTE,
+ VK_PIPELINE_BIND_POINT_RANGE_SIZE = (VK_PIPELINE_BIND_POINT_COMPUTE - VK_PIPELINE_BIND_POINT_GRAPHICS + 1),
+ VK_PIPELINE_BIND_POINT_MAX_ENUM = 0x7FFFFFFF
+} VkPipelineBindPoint;
+
+typedef enum VkCommandBufferLevel {
+ VK_COMMAND_BUFFER_LEVEL_PRIMARY = 0,
+ VK_COMMAND_BUFFER_LEVEL_SECONDARY = 1,
+ VK_COMMAND_BUFFER_LEVEL_BEGIN_RANGE = VK_COMMAND_BUFFER_LEVEL_PRIMARY,
+ VK_COMMAND_BUFFER_LEVEL_END_RANGE = VK_COMMAND_BUFFER_LEVEL_SECONDARY,
+ VK_COMMAND_BUFFER_LEVEL_RANGE_SIZE = (VK_COMMAND_BUFFER_LEVEL_SECONDARY - VK_COMMAND_BUFFER_LEVEL_PRIMARY + 1),
+ VK_COMMAND_BUFFER_LEVEL_MAX_ENUM = 0x7FFFFFFF
+} VkCommandBufferLevel;
+
+typedef enum VkIndexType {
+ VK_INDEX_TYPE_UINT16 = 0,
+ VK_INDEX_TYPE_UINT32 = 1,
+ VK_INDEX_TYPE_BEGIN_RANGE = VK_INDEX_TYPE_UINT16,
+ VK_INDEX_TYPE_END_RANGE = VK_INDEX_TYPE_UINT32,
+ VK_INDEX_TYPE_RANGE_SIZE = (VK_INDEX_TYPE_UINT32 - VK_INDEX_TYPE_UINT16 + 1),
+ VK_INDEX_TYPE_MAX_ENUM = 0x7FFFFFFF
+} VkIndexType;
+
+typedef enum VkSubpassContents {
+ VK_SUBPASS_CONTENTS_INLINE = 0,
+ VK_SUBPASS_CONTENTS_SECONDARY_COMMAND_BUFFERS = 1,
+ VK_SUBPASS_CONTENTS_BEGIN_RANGE = VK_SUBPASS_CONTENTS_INLINE,
+ VK_SUBPASS_CONTENTS_END_RANGE = VK_SUBPASS_CONTENTS_SECONDARY_COMMAND_BUFFERS,
+ VK_SUBPASS_CONTENTS_RANGE_SIZE = (VK_SUBPASS_CONTENTS_SECONDARY_COMMAND_BUFFERS - VK_SUBPASS_CONTENTS_INLINE + 1),
+ VK_SUBPASS_CONTENTS_MAX_ENUM = 0x7FFFFFFF
+} VkSubpassContents;
+
+typedef VkFlags VkInstanceCreateFlags;
+
+typedef enum VkFormatFeatureFlagBits {
+ VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT = 0x00000001,
+ VK_FORMAT_FEATURE_STORAGE_IMAGE_BIT = 0x00000002,
+ VK_FORMAT_FEATURE_STORAGE_IMAGE_ATOMIC_BIT = 0x00000004,
+ VK_FORMAT_FEATURE_UNIFORM_TEXEL_BUFFER_BIT = 0x00000008,
+ VK_FORMAT_FEATURE_STORAGE_TEXEL_BUFFER_BIT = 0x00000010,
+ VK_FORMAT_FEATURE_STORAGE_TEXEL_BUFFER_ATOMIC_BIT = 0x00000020,
+ VK_FORMAT_FEATURE_VERTEX_BUFFER_BIT = 0x00000040,
+ VK_FORMAT_FEATURE_COLOR_ATTACHMENT_BIT = 0x00000080,
+ VK_FORMAT_FEATURE_COLOR_ATTACHMENT_BLEND_BIT = 0x00000100,
+ VK_FORMAT_FEATURE_DEPTH_STENCIL_ATTACHMENT_BIT = 0x00000200,
+ VK_FORMAT_FEATURE_BLIT_SRC_BIT = 0x00000400,
+ VK_FORMAT_FEATURE_BLIT_DST_BIT = 0x00000800,
+} VkFormatFeatureFlagBits;
+typedef VkFlags VkFormatFeatureFlags;
+
+typedef enum VkImageUsageFlagBits {
+ VK_IMAGE_USAGE_TRANSFER_SRC_BIT = 0x00000001,
+ VK_IMAGE_USAGE_TRANSFER_DST_BIT = 0x00000002,
+ VK_IMAGE_USAGE_SAMPLED_BIT = 0x00000004,
+ VK_IMAGE_USAGE_STORAGE_BIT = 0x00000008,
+ VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT = 0x00000010,
+ VK_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT = 0x00000020,
+ VK_IMAGE_USAGE_TRANSIENT_ATTACHMENT_BIT = 0x00000040,
+ VK_IMAGE_USAGE_INPUT_ATTACHMENT_BIT = 0x00000080,
+} VkImageUsageFlagBits;
+typedef VkFlags VkImageUsageFlags;
+
+typedef enum VkImageCreateFlagBits {
+ VK_IMAGE_CREATE_SPARSE_BINDING_BIT = 0x00000001,
+ VK_IMAGE_CREATE_SPARSE_RESIDENCY_BIT = 0x00000002,
+ VK_IMAGE_CREATE_SPARSE_ALIASED_BIT = 0x00000004,
+ VK_IMAGE_CREATE_MUTABLE_FORMAT_BIT = 0x00000008,
+ VK_IMAGE_CREATE_CUBE_COMPATIBLE_BIT = 0x00000010,
+} VkImageCreateFlagBits;
+typedef VkFlags VkImageCreateFlags;
+
+typedef enum VkSampleCountFlagBits {
+ VK_SAMPLE_COUNT_1_BIT = 0x00000001,
+ VK_SAMPLE_COUNT_2_BIT = 0x00000002,
+ VK_SAMPLE_COUNT_4_BIT = 0x00000004,
+ VK_SAMPLE_COUNT_8_BIT = 0x00000008,
+ VK_SAMPLE_COUNT_16_BIT = 0x00000010,
+ VK_SAMPLE_COUNT_32_BIT = 0x00000020,
+ VK_SAMPLE_COUNT_64_BIT = 0x00000040,
+} VkSampleCountFlagBits;
+typedef VkFlags VkSampleCountFlags;
+
+typedef enum VkQueueFlagBits {
+ VK_QUEUE_GRAPHICS_BIT = 0x00000001,
+ VK_QUEUE_COMPUTE_BIT = 0x00000002,
+ VK_QUEUE_TRANSFER_BIT = 0x00000004,
+ VK_QUEUE_SPARSE_BINDING_BIT = 0x00000008,
+} VkQueueFlagBits;
+typedef VkFlags VkQueueFlags;
+
+typedef enum VkMemoryPropertyFlagBits {
+ VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT = 0x00000001,
+ VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT = 0x00000002,
+ VK_MEMORY_PROPERTY_HOST_COHERENT_BIT = 0x00000004,
+ VK_MEMORY_PROPERTY_HOST_CACHED_BIT = 0x00000008,
+ VK_MEMORY_PROPERTY_LAZILY_ALLOCATED_BIT = 0x00000010,
+} VkMemoryPropertyFlagBits;
+typedef VkFlags VkMemoryPropertyFlags;
+
+typedef enum VkMemoryHeapFlagBits {
+ VK_MEMORY_HEAP_DEVICE_LOCAL_BIT = 0x00000001,
+} VkMemoryHeapFlagBits;
+typedef VkFlags VkMemoryHeapFlags;
+typedef VkFlags VkDeviceCreateFlags;
+typedef VkFlags VkDeviceQueueCreateFlags;
+
+typedef enum VkPipelineStageFlagBits {
+ VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT = 0x00000001,
+ VK_PIPELINE_STAGE_DRAW_INDIRECT_BIT = 0x00000002,
+ VK_PIPELINE_STAGE_VERTEX_INPUT_BIT = 0x00000004,
+ VK_PIPELINE_STAGE_VERTEX_SHADER_BIT = 0x00000008,
+ VK_PIPELINE_STAGE_TESSELLATION_CONTROL_SHADER_BIT = 0x00000010,
+ VK_PIPELINE_STAGE_TESSELLATION_EVALUATION_SHADER_BIT = 0x00000020,
+ VK_PIPELINE_STAGE_GEOMETRY_SHADER_BIT = 0x00000040,
+ VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT = 0x00000080,
+ VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT = 0x00000100,
+ VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT = 0x00000200,
+ VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT = 0x00000400,
+ VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT = 0x00000800,
+ VK_PIPELINE_STAGE_TRANSFER_BIT = 0x00001000,
+ VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT = 0x00002000,
+ VK_PIPELINE_STAGE_HOST_BIT = 0x00004000,
+ VK_PIPELINE_STAGE_ALL_GRAPHICS_BIT = 0x00008000,
+ VK_PIPELINE_STAGE_ALL_COMMANDS_BIT = 0x00010000,
+} VkPipelineStageFlagBits;
+typedef VkFlags VkPipelineStageFlags;
+typedef VkFlags VkMemoryMapFlags;
+
+typedef enum VkImageAspectFlagBits {
+ VK_IMAGE_ASPECT_COLOR_BIT = 0x00000001,
+ VK_IMAGE_ASPECT_DEPTH_BIT = 0x00000002,
+ VK_IMAGE_ASPECT_STENCIL_BIT = 0x00000004,
+ VK_IMAGE_ASPECT_METADATA_BIT = 0x00000008,
+} VkImageAspectFlagBits;
+typedef VkFlags VkImageAspectFlags;
+
+typedef enum VkSparseImageFormatFlagBits {
+ VK_SPARSE_IMAGE_FORMAT_SINGLE_MIPTAIL_BIT = 0x00000001,
+ VK_SPARSE_IMAGE_FORMAT_ALIGNED_MIP_SIZE_BIT = 0x00000002,
+ VK_SPARSE_IMAGE_FORMAT_NONSTANDARD_BLOCK_SIZE_BIT = 0x00000004,
+} VkSparseImageFormatFlagBits;
+typedef VkFlags VkSparseImageFormatFlags;
+
+typedef enum VkSparseMemoryBindFlagBits {
+ VK_SPARSE_MEMORY_BIND_METADATA_BIT = 0x00000001,
+} VkSparseMemoryBindFlagBits;
+typedef VkFlags VkSparseMemoryBindFlags;
+
+typedef enum VkFenceCreateFlagBits {
+ VK_FENCE_CREATE_SIGNALED_BIT = 0x00000001,
+} VkFenceCreateFlagBits;
+typedef VkFlags VkFenceCreateFlags;
+typedef VkFlags VkSemaphoreCreateFlags;
+typedef VkFlags VkEventCreateFlags;
+typedef VkFlags VkQueryPoolCreateFlags;
+
+typedef enum VkQueryPipelineStatisticFlagBits {
+ VK_QUERY_PIPELINE_STATISTIC_INPUT_ASSEMBLY_VERTICES_BIT = 0x00000001,
+ VK_QUERY_PIPELINE_STATISTIC_INPUT_ASSEMBLY_PRIMITIVES_BIT = 0x00000002,
+ VK_QUERY_PIPELINE_STATISTIC_VERTEX_SHADER_INVOCATIONS_BIT = 0x00000004,
+ VK_QUERY_PIPELINE_STATISTIC_GEOMETRY_SHADER_INVOCATIONS_BIT = 0x00000008,
+ VK_QUERY_PIPELINE_STATISTIC_GEOMETRY_SHADER_PRIMITIVES_BIT = 0x00000010,
+ VK_QUERY_PIPELINE_STATISTIC_CLIPPING_INVOCATIONS_BIT = 0x00000020,
+ VK_QUERY_PIPELINE_STATISTIC_CLIPPING_PRIMITIVES_BIT = 0x00000040,
+ VK_QUERY_PIPELINE_STATISTIC_FRAGMENT_SHADER_INVOCATIONS_BIT = 0x00000080,
+ VK_QUERY_PIPELINE_STATISTIC_TESSELLATION_CONTROL_SHADER_PATCHES_BIT = 0x00000100,
+ VK_QUERY_PIPELINE_STATISTIC_TESSELLATION_EVALUATION_SHADER_INVOCATIONS_BIT = 0x00000200,
+ VK_QUERY_PIPELINE_STATISTIC_COMPUTE_SHADER_INVOCATIONS_BIT = 0x00000400,
+} VkQueryPipelineStatisticFlagBits;
+typedef VkFlags VkQueryPipelineStatisticFlags;
+
+typedef enum VkQueryResultFlagBits {
+ VK_QUERY_RESULT_64_BIT = 0x00000001,
+ VK_QUERY_RESULT_WAIT_BIT = 0x00000002,
+ VK_QUERY_RESULT_WITH_AVAILABILITY_BIT = 0x00000004,
+ VK_QUERY_RESULT_PARTIAL_BIT = 0x00000008,
+} VkQueryResultFlagBits;
+typedef VkFlags VkQueryResultFlags;
+
+typedef enum VkBufferCreateFlagBits {
+ VK_BUFFER_CREATE_SPARSE_BINDING_BIT = 0x00000001,
+ VK_BUFFER_CREATE_SPARSE_RESIDENCY_BIT = 0x00000002,
+ VK_BUFFER_CREATE_SPARSE_ALIASED_BIT = 0x00000004,
+} VkBufferCreateFlagBits;
+typedef VkFlags VkBufferCreateFlags;
+
+typedef enum VkBufferUsageFlagBits {
+ VK_BUFFER_USAGE_TRANSFER_SRC_BIT = 0x00000001,
+ VK_BUFFER_USAGE_TRANSFER_DST_BIT = 0x00000002,
+ VK_BUFFER_USAGE_UNIFORM_TEXEL_BUFFER_BIT = 0x00000004,
+ VK_BUFFER_USAGE_STORAGE_TEXEL_BUFFER_BIT = 0x00000008,
+ VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT = 0x00000010,
+ VK_BUFFER_USAGE_STORAGE_BUFFER_BIT = 0x00000020,
+ VK_BUFFER_USAGE_INDEX_BUFFER_BIT = 0x00000040,
+ VK_BUFFER_USAGE_VERTEX_BUFFER_BIT = 0x00000080,
+ VK_BUFFER_USAGE_INDIRECT_BUFFER_BIT = 0x00000100,
+} VkBufferUsageFlagBits;
+typedef VkFlags VkBufferUsageFlags;
+typedef VkFlags VkBufferViewCreateFlags;
+typedef VkFlags VkImageViewCreateFlags;
+typedef VkFlags VkShaderModuleCreateFlags;
+typedef VkFlags VkPipelineCacheCreateFlags;
+
+typedef enum VkPipelineCreateFlagBits {
+ VK_PIPELINE_CREATE_DISABLE_OPTIMIZATION_BIT = 0x00000001,
+ VK_PIPELINE_CREATE_ALLOW_DERIVATIVES_BIT = 0x00000002,
+ VK_PIPELINE_CREATE_DERIVATIVE_BIT = 0x00000004,
+} VkPipelineCreateFlagBits;
+typedef VkFlags VkPipelineCreateFlags;
+typedef VkFlags VkPipelineShaderStageCreateFlags;
+
+typedef enum VkShaderStageFlagBits {
+ VK_SHADER_STAGE_VERTEX_BIT = 0x00000001,
+ VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT = 0x00000002,
+ VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT = 0x00000004,
+ VK_SHADER_STAGE_GEOMETRY_BIT = 0x00000008,
+ VK_SHADER_STAGE_FRAGMENT_BIT = 0x00000010,
+ VK_SHADER_STAGE_COMPUTE_BIT = 0x00000020,
+ VK_SHADER_STAGE_ALL_GRAPHICS = 0x1F,
+ VK_SHADER_STAGE_ALL = 0x7FFFFFFF,
+} VkShaderStageFlagBits;
+typedef VkFlags VkPipelineVertexInputStateCreateFlags;
+typedef VkFlags VkPipelineInputAssemblyStateCreateFlags;
+typedef VkFlags VkPipelineTessellationStateCreateFlags;
+typedef VkFlags VkPipelineViewportStateCreateFlags;
+typedef VkFlags VkPipelineRasterizationStateCreateFlags;
+
+typedef enum VkCullModeFlagBits {
+ VK_CULL_MODE_NONE = 0,
+ VK_CULL_MODE_FRONT_BIT = 0x00000001,
+ VK_CULL_MODE_BACK_BIT = 0x00000002,
+ VK_CULL_MODE_FRONT_AND_BACK = 0x3,
+} VkCullModeFlagBits;
+typedef VkFlags VkCullModeFlags;
+typedef VkFlags VkPipelineMultisampleStateCreateFlags;
+typedef VkFlags VkPipelineDepthStencilStateCreateFlags;
+typedef VkFlags VkPipelineColorBlendStateCreateFlags;
+
+typedef enum VkColorComponentFlagBits {
+ VK_COLOR_COMPONENT_R_BIT = 0x00000001,
+ VK_COLOR_COMPONENT_G_BIT = 0x00000002,
+ VK_COLOR_COMPONENT_B_BIT = 0x00000004,
+ VK_COLOR_COMPONENT_A_BIT = 0x00000008,
+} VkColorComponentFlagBits;
+typedef VkFlags VkColorComponentFlags;
+typedef VkFlags VkPipelineDynamicStateCreateFlags;
+typedef VkFlags VkPipelineLayoutCreateFlags;
+typedef VkFlags VkShaderStageFlags;
+typedef VkFlags VkSamplerCreateFlags;
+typedef VkFlags VkDescriptorSetLayoutCreateFlags;
+
+typedef enum VkDescriptorPoolCreateFlagBits {
+ VK_DESCRIPTOR_POOL_CREATE_FREE_DESCRIPTOR_SET_BIT = 0x00000001,
+} VkDescriptorPoolCreateFlagBits;
+typedef VkFlags VkDescriptorPoolCreateFlags;
+typedef VkFlags VkDescriptorPoolResetFlags;
+typedef VkFlags VkFramebufferCreateFlags;
+typedef VkFlags VkRenderPassCreateFlags;
+
+typedef enum VkAttachmentDescriptionFlagBits {
+ VK_ATTACHMENT_DESCRIPTION_MAY_ALIAS_BIT = 0x00000001,
+} VkAttachmentDescriptionFlagBits;
+typedef VkFlags VkAttachmentDescriptionFlags;
+typedef VkFlags VkSubpassDescriptionFlags;
+
+typedef enum VkAccessFlagBits {
+ VK_ACCESS_INDIRECT_COMMAND_READ_BIT = 0x00000001,
+ VK_ACCESS_INDEX_READ_BIT = 0x00000002,
+ VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT = 0x00000004,
+ VK_ACCESS_UNIFORM_READ_BIT = 0x00000008,
+ VK_ACCESS_INPUT_ATTACHMENT_READ_BIT = 0x00000010,
+ VK_ACCESS_SHADER_READ_BIT = 0x00000020,
+ VK_ACCESS_SHADER_WRITE_BIT = 0x00000040,
+ VK_ACCESS_COLOR_ATTACHMENT_READ_BIT = 0x00000080,
+ VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT = 0x00000100,
+ VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT = 0x00000200,
+ VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT = 0x00000400,
+ VK_ACCESS_TRANSFER_READ_BIT = 0x00000800,
+ VK_ACCESS_TRANSFER_WRITE_BIT = 0x00001000,
+ VK_ACCESS_HOST_READ_BIT = 0x00002000,
+ VK_ACCESS_HOST_WRITE_BIT = 0x00004000,
+ VK_ACCESS_MEMORY_READ_BIT = 0x00008000,
+ VK_ACCESS_MEMORY_WRITE_BIT = 0x00010000,
+} VkAccessFlagBits;
+typedef VkFlags VkAccessFlags;
+
+typedef enum VkDependencyFlagBits {
+ VK_DEPENDENCY_BY_REGION_BIT = 0x00000001,
+} VkDependencyFlagBits;
+typedef VkFlags VkDependencyFlags;
+
+typedef enum VkCommandPoolCreateFlagBits {
+ VK_COMMAND_POOL_CREATE_TRANSIENT_BIT = 0x00000001,
+ VK_COMMAND_POOL_CREATE_RESET_COMMAND_BUFFER_BIT = 0x00000002,
+} VkCommandPoolCreateFlagBits;
+typedef VkFlags VkCommandPoolCreateFlags;
+
+typedef enum VkCommandPoolResetFlagBits {
+ VK_COMMAND_POOL_RESET_RELEASE_RESOURCES_BIT = 0x00000001,
+} VkCommandPoolResetFlagBits;
+typedef VkFlags VkCommandPoolResetFlags;
+
+typedef enum VkCommandBufferUsageFlagBits {
+ VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT = 0x00000001,
+ VK_COMMAND_BUFFER_USAGE_RENDER_PASS_CONTINUE_BIT = 0x00000002,
+ VK_COMMAND_BUFFER_USAGE_SIMULTANEOUS_USE_BIT = 0x00000004,
+} VkCommandBufferUsageFlagBits;
+typedef VkFlags VkCommandBufferUsageFlags;
+
+typedef enum VkQueryControlFlagBits {
+ VK_QUERY_CONTROL_PRECISE_BIT = 0x00000001,
+} VkQueryControlFlagBits;
+typedef VkFlags VkQueryControlFlags;
+
+typedef enum VkCommandBufferResetFlagBits {
+ VK_COMMAND_BUFFER_RESET_RELEASE_RESOURCES_BIT = 0x00000001,
+} VkCommandBufferResetFlagBits;
+typedef VkFlags VkCommandBufferResetFlags;
+
+typedef enum VkStencilFaceFlagBits {
+ VK_STENCIL_FACE_FRONT_BIT = 0x00000001,
+ VK_STENCIL_FACE_BACK_BIT = 0x00000002,
+ VK_STENCIL_FRONT_AND_BACK = 0x3,
+} VkStencilFaceFlagBits;
+typedef VkFlags VkStencilFaceFlags;
+
+typedef void* (VKAPI_PTR *PFN_vkAllocationFunction)(
+ void* pUserData,
+ size_t size,
+ size_t alignment,
+ VkSystemAllocationScope allocationScope);
+
+typedef void* (VKAPI_PTR *PFN_vkReallocationFunction)(
+ void* pUserData,
+ void* pOriginal,
+ size_t size,
+ size_t alignment,
+ VkSystemAllocationScope allocationScope);
+
+typedef void (VKAPI_PTR *PFN_vkFreeFunction)(
+ void* pUserData,
+ void* pMemory);
+
+typedef void (VKAPI_PTR *PFN_vkInternalAllocationNotification)(
+ void* pUserData,
+ size_t size,
+ VkInternalAllocationType allocationType,
+ VkSystemAllocationScope allocationScope);
+
+typedef void (VKAPI_PTR *PFN_vkInternalFreeNotification)(
+ void* pUserData,
+ size_t size,
+ VkInternalAllocationType allocationType,
+ VkSystemAllocationScope allocationScope);
+
+typedef void (VKAPI_PTR *PFN_vkVoidFunction)(void);
+
+typedef struct VkApplicationInfo {
+ VkStructureType sType;
+ const void* pNext;
+ const char* pApplicationName;
+ uint32_t applicationVersion;
+ const char* pEngineName;
+ uint32_t engineVersion;
+ uint32_t apiVersion;
+} VkApplicationInfo;
+
+typedef struct VkInstanceCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkInstanceCreateFlags flags;
+ const VkApplicationInfo* pApplicationInfo;
+ uint32_t enabledLayerCount;
+ const char* const* ppEnabledLayerNames;
+ uint32_t enabledExtensionCount;
+ const char* const* ppEnabledExtensionNames;
+} VkInstanceCreateInfo;
+
+typedef struct VkAllocationCallbacks {
+ void* pUserData;
+ PFN_vkAllocationFunction pfnAllocation;
+ PFN_vkReallocationFunction pfnReallocation;
+ PFN_vkFreeFunction pfnFree;
+ PFN_vkInternalAllocationNotification pfnInternalAllocation;
+ PFN_vkInternalFreeNotification pfnInternalFree;
+} VkAllocationCallbacks;
+
+typedef struct VkPhysicalDeviceFeatures {
+ VkBool32 robustBufferAccess;
+ VkBool32 fullDrawIndexUint32;
+ VkBool32 imageCubeArray;
+ VkBool32 independentBlend;
+ VkBool32 geometryShader;
+ VkBool32 tessellationShader;
+ VkBool32 sampleRateShading;
+ VkBool32 dualSrcBlend;
+ VkBool32 logicOp;
+ VkBool32 multiDrawIndirect;
+ VkBool32 drawIndirectFirstInstance;
+ VkBool32 depthClamp;
+ VkBool32 depthBiasClamp;
+ VkBool32 fillModeNonSolid;
+ VkBool32 depthBounds;
+ VkBool32 wideLines;
+ VkBool32 largePoints;
+ VkBool32 alphaToOne;
+ VkBool32 multiViewport;
+ VkBool32 samplerAnisotropy;
+ VkBool32 textureCompressionETC2;
+ VkBool32 textureCompressionASTC_LDR;
+ VkBool32 textureCompressionBC;
+ VkBool32 occlusionQueryPrecise;
+ VkBool32 pipelineStatisticsQuery;
+ VkBool32 vertexPipelineStoresAndAtomics;
+ VkBool32 fragmentStoresAndAtomics;
+ VkBool32 shaderTessellationAndGeometryPointSize;
+ VkBool32 shaderImageGatherExtended;
+ VkBool32 shaderStorageImageExtendedFormats;
+ VkBool32 shaderStorageImageMultisample;
+ VkBool32 shaderStorageImageReadWithoutFormat;
+ VkBool32 shaderStorageImageWriteWithoutFormat;
+ VkBool32 shaderUniformBufferArrayDynamicIndexing;
+ VkBool32 shaderSampledImageArrayDynamicIndexing;
+ VkBool32 shaderStorageBufferArrayDynamicIndexing;
+ VkBool32 shaderStorageImageArrayDynamicIndexing;
+ VkBool32 shaderClipDistance;
+ VkBool32 shaderCullDistance;
+ VkBool32 shaderFloat64;
+ VkBool32 shaderInt64;
+ VkBool32 shaderInt16;
+ VkBool32 shaderResourceResidency;
+ VkBool32 shaderResourceMinLod;
+ VkBool32 sparseBinding;
+ VkBool32 sparseResidencyBuffer;
+ VkBool32 sparseResidencyImage2D;
+ VkBool32 sparseResidencyImage3D;
+ VkBool32 sparseResidency2Samples;
+ VkBool32 sparseResidency4Samples;
+ VkBool32 sparseResidency8Samples;
+ VkBool32 sparseResidency16Samples;
+ VkBool32 sparseResidencyAliased;
+ VkBool32 variableMultisampleRate;
+ VkBool32 inheritedQueries;
+} VkPhysicalDeviceFeatures;
+
+typedef struct VkFormatProperties {
+ VkFormatFeatureFlags linearTilingFeatures;
+ VkFormatFeatureFlags optimalTilingFeatures;
+ VkFormatFeatureFlags bufferFeatures;
+} VkFormatProperties;
+
+typedef struct VkExtent3D {
+ uint32_t width;
+ uint32_t height;
+ uint32_t depth;
+} VkExtent3D;
+
+typedef struct VkImageFormatProperties {
+ VkExtent3D maxExtent;
+ uint32_t maxMipLevels;
+ uint32_t maxArrayLayers;
+ VkSampleCountFlags sampleCounts;
+ VkDeviceSize maxResourceSize;
+} VkImageFormatProperties;
+
+typedef struct VkPhysicalDeviceLimits {
+ uint32_t maxImageDimension1D;
+ uint32_t maxImageDimension2D;
+ uint32_t maxImageDimension3D;
+ uint32_t maxImageDimensionCube;
+ uint32_t maxImageArrayLayers;
+ uint32_t maxTexelBufferElements;
+ uint32_t maxUniformBufferRange;
+ uint32_t maxStorageBufferRange;
+ uint32_t maxPushConstantsSize;
+ uint32_t maxMemoryAllocationCount;
+ uint32_t maxSamplerAllocationCount;
+ VkDeviceSize bufferImageGranularity;
+ VkDeviceSize sparseAddressSpaceSize;
+ uint32_t maxBoundDescriptorSets;
+ uint32_t maxPerStageDescriptorSamplers;
+ uint32_t maxPerStageDescriptorUniformBuffers;
+ uint32_t maxPerStageDescriptorStorageBuffers;
+ uint32_t maxPerStageDescriptorSampledImages;
+ uint32_t maxPerStageDescriptorStorageImages;
+ uint32_t maxPerStageDescriptorInputAttachments;
+ uint32_t maxPerStageResources;
+ uint32_t maxDescriptorSetSamplers;
+ uint32_t maxDescriptorSetUniformBuffers;
+ uint32_t maxDescriptorSetUniformBuffersDynamic;
+ uint32_t maxDescriptorSetStorageBuffers;
+ uint32_t maxDescriptorSetStorageBuffersDynamic;
+ uint32_t maxDescriptorSetSampledImages;
+ uint32_t maxDescriptorSetStorageImages;
+ uint32_t maxDescriptorSetInputAttachments;
+ uint32_t maxVertexInputAttributes;
+ uint32_t maxVertexInputBindings;
+ uint32_t maxVertexInputAttributeOffset;
+ uint32_t maxVertexInputBindingStride;
+ uint32_t maxVertexOutputComponents;
+ uint32_t maxTessellationGenerationLevel;
+ uint32_t maxTessellationPatchSize;
+ uint32_t maxTessellationControlPerVertexInputComponents;
+ uint32_t maxTessellationControlPerVertexOutputComponents;
+ uint32_t maxTessellationControlPerPatchOutputComponents;
+ uint32_t maxTessellationControlTotalOutputComponents;
+ uint32_t maxTessellationEvaluationInputComponents;
+ uint32_t maxTessellationEvaluationOutputComponents;
+ uint32_t maxGeometryShaderInvocations;
+ uint32_t maxGeometryInputComponents;
+ uint32_t maxGeometryOutputComponents;
+ uint32_t maxGeometryOutputVertices;
+ uint32_t maxGeometryTotalOutputComponents;
+ uint32_t maxFragmentInputComponents;
+ uint32_t maxFragmentOutputAttachments;
+ uint32_t maxFragmentDualSrcAttachments;
+ uint32_t maxFragmentCombinedOutputResources;
+ uint32_t maxComputeSharedMemorySize;
+ uint32_t maxComputeWorkGroupCount[3];
+ uint32_t maxComputeWorkGroupInvocations;
+ uint32_t maxComputeWorkGroupSize[3];
+ uint32_t subPixelPrecisionBits;
+ uint32_t subTexelPrecisionBits;
+ uint32_t mipmapPrecisionBits;
+ uint32_t maxDrawIndexedIndexValue;
+ uint32_t maxDrawIndirectCount;
+ float maxSamplerLodBias;
+ float maxSamplerAnisotropy;
+ uint32_t maxViewports;
+ uint32_t maxViewportDimensions[2];
+ float viewportBoundsRange[2];
+ uint32_t viewportSubPixelBits;
+ size_t minMemoryMapAlignment;
+ VkDeviceSize minTexelBufferOffsetAlignment;
+ VkDeviceSize minUniformBufferOffsetAlignment;
+ VkDeviceSize minStorageBufferOffsetAlignment;
+ int32_t minTexelOffset;
+ uint32_t maxTexelOffset;
+ int32_t minTexelGatherOffset;
+ uint32_t maxTexelGatherOffset;
+ float minInterpolationOffset;
+ float maxInterpolationOffset;
+ uint32_t subPixelInterpolationOffsetBits;
+ uint32_t maxFramebufferWidth;
+ uint32_t maxFramebufferHeight;
+ uint32_t maxFramebufferLayers;
+ VkSampleCountFlags framebufferColorSampleCounts;
+ VkSampleCountFlags framebufferDepthSampleCounts;
+ VkSampleCountFlags framebufferStencilSampleCounts;
+ VkSampleCountFlags framebufferNoAttachmentsSampleCounts;
+ uint32_t maxColorAttachments;
+ VkSampleCountFlags sampledImageColorSampleCounts;
+ VkSampleCountFlags sampledImageIntegerSampleCounts;
+ VkSampleCountFlags sampledImageDepthSampleCounts;
+ VkSampleCountFlags sampledImageStencilSampleCounts;
+ VkSampleCountFlags storageImageSampleCounts;
+ uint32_t maxSampleMaskWords;
+ VkBool32 timestampComputeAndGraphics;
+ float timestampPeriod;
+ uint32_t maxClipDistances;
+ uint32_t maxCullDistances;
+ uint32_t maxCombinedClipAndCullDistances;
+ uint32_t discreteQueuePriorities;
+ float pointSizeRange[2];
+ float lineWidthRange[2];
+ float pointSizeGranularity;
+ float lineWidthGranularity;
+ VkBool32 strictLines;
+ VkBool32 standardSampleLocations;
+ VkDeviceSize optimalBufferCopyOffsetAlignment;
+ VkDeviceSize optimalBufferCopyRowPitchAlignment;
+ VkDeviceSize nonCoherentAtomSize;
+} VkPhysicalDeviceLimits;
+
+typedef struct VkPhysicalDeviceSparseProperties {
+ VkBool32 residencyStandard2DBlockShape;
+ VkBool32 residencyStandard2DMultisampleBlockShape;
+ VkBool32 residencyStandard3DBlockShape;
+ VkBool32 residencyAlignedMipSize;
+ VkBool32 residencyNonResidentStrict;
+} VkPhysicalDeviceSparseProperties;
+
+typedef struct VkPhysicalDeviceProperties {
+ uint32_t apiVersion;
+ uint32_t driverVersion;
+ uint32_t vendorID;
+ uint32_t deviceID;
+ VkPhysicalDeviceType deviceType;
+ char deviceName[VK_MAX_PHYSICAL_DEVICE_NAME_SIZE];
+ uint8_t pipelineCacheUUID[VK_UUID_SIZE];
+ VkPhysicalDeviceLimits limits;
+ VkPhysicalDeviceSparseProperties sparseProperties;
+} VkPhysicalDeviceProperties;
+
+typedef struct VkQueueFamilyProperties {
+ VkQueueFlags queueFlags;
+ uint32_t queueCount;
+ uint32_t timestampValidBits;
+ VkExtent3D minImageTransferGranularity;
+} VkQueueFamilyProperties;
+
+typedef struct VkMemoryType {
+ VkMemoryPropertyFlags propertyFlags;
+ uint32_t heapIndex;
+} VkMemoryType;
+
+typedef struct VkMemoryHeap {
+ VkDeviceSize size;
+ VkMemoryHeapFlags flags;
+} VkMemoryHeap;
+
+typedef struct VkPhysicalDeviceMemoryProperties {
+ uint32_t memoryTypeCount;
+ VkMemoryType memoryTypes[VK_MAX_MEMORY_TYPES];
+ uint32_t memoryHeapCount;
+ VkMemoryHeap memoryHeaps[VK_MAX_MEMORY_HEAPS];
+} VkPhysicalDeviceMemoryProperties;
+
+typedef struct VkDeviceQueueCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkDeviceQueueCreateFlags flags;
+ uint32_t queueFamilyIndex;
+ uint32_t queueCount;
+ const float* pQueuePriorities;
+} VkDeviceQueueCreateInfo;
+
+typedef struct VkDeviceCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkDeviceCreateFlags flags;
+ uint32_t queueCreateInfoCount;
+ const VkDeviceQueueCreateInfo* pQueueCreateInfos;
+ uint32_t enabledLayerCount;
+ const char* const* ppEnabledLayerNames;
+ uint32_t enabledExtensionCount;
+ const char* const* ppEnabledExtensionNames;
+ const VkPhysicalDeviceFeatures* pEnabledFeatures;
+} VkDeviceCreateInfo;
+
+typedef struct VkExtensionProperties {
+ char extensionName[VK_MAX_EXTENSION_NAME_SIZE];
+ uint32_t specVersion;
+} VkExtensionProperties;
+
+typedef struct VkLayerProperties {
+ char layerName[VK_MAX_EXTENSION_NAME_SIZE];
+ uint32_t specVersion;
+ uint32_t implementationVersion;
+ char description[VK_MAX_DESCRIPTION_SIZE];
+} VkLayerProperties;
+
+typedef struct VkSubmitInfo {
+ VkStructureType sType;
+ const void* pNext;
+ uint32_t waitSemaphoreCount;
+ const VkSemaphore* pWaitSemaphores;
+ const VkPipelineStageFlags* pWaitDstStageMask;
+ uint32_t commandBufferCount;
+ const VkCommandBuffer* pCommandBuffers;
+ uint32_t signalSemaphoreCount;
+ const VkSemaphore* pSignalSemaphores;
+} VkSubmitInfo;
+
+typedef struct VkMemoryAllocateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkDeviceSize allocationSize;
+ uint32_t memoryTypeIndex;
+} VkMemoryAllocateInfo;
+
+typedef struct VkMappedMemoryRange {
+ VkStructureType sType;
+ const void* pNext;
+ VkDeviceMemory memory;
+ VkDeviceSize offset;
+ VkDeviceSize size;
+} VkMappedMemoryRange;
+
+typedef struct VkMemoryRequirements {
+ VkDeviceSize size;
+ VkDeviceSize alignment;
+ uint32_t memoryTypeBits;
+} VkMemoryRequirements;
+
+typedef struct VkSparseImageFormatProperties {
+ VkImageAspectFlags aspectMask;
+ VkExtent3D imageGranularity;
+ VkSparseImageFormatFlags flags;
+} VkSparseImageFormatProperties;
+
+typedef struct VkSparseImageMemoryRequirements {
+ VkSparseImageFormatProperties formatProperties;
+ uint32_t imageMipTailFirstLod;
+ VkDeviceSize imageMipTailSize;
+ VkDeviceSize imageMipTailOffset;
+ VkDeviceSize imageMipTailStride;
+} VkSparseImageMemoryRequirements;
+
+typedef struct VkSparseMemoryBind {
+ VkDeviceSize resourceOffset;
+ VkDeviceSize size;
+ VkDeviceMemory memory;
+ VkDeviceSize memoryOffset;
+ VkSparseMemoryBindFlags flags;
+} VkSparseMemoryBind;
+
+typedef struct VkSparseBufferMemoryBindInfo {
+ VkBuffer buffer;
+ uint32_t bindCount;
+ const VkSparseMemoryBind* pBinds;
+} VkSparseBufferMemoryBindInfo;
+
+typedef struct VkSparseImageOpaqueMemoryBindInfo {
+ VkImage image;
+ uint32_t bindCount;
+ const VkSparseMemoryBind* pBinds;
+} VkSparseImageOpaqueMemoryBindInfo;
+
+typedef struct VkImageSubresource {
+ VkImageAspectFlags aspectMask;
+ uint32_t mipLevel;
+ uint32_t arrayLayer;
+} VkImageSubresource;
+
+typedef struct VkOffset3D {
+ int32_t x;
+ int32_t y;
+ int32_t z;
+} VkOffset3D;
+
+typedef struct VkSparseImageMemoryBind {
+ VkImageSubresource subresource;
+ VkOffset3D offset;
+ VkExtent3D extent;
+ VkDeviceMemory memory;
+ VkDeviceSize memoryOffset;
+ VkSparseMemoryBindFlags flags;
+} VkSparseImageMemoryBind;
+
+typedef struct VkSparseImageMemoryBindInfo {
+ VkImage image;
+ uint32_t bindCount;
+ const VkSparseImageMemoryBind* pBinds;
+} VkSparseImageMemoryBindInfo;
+
+typedef struct VkBindSparseInfo {
+ VkStructureType sType;
+ const void* pNext;
+ uint32_t waitSemaphoreCount;
+ const VkSemaphore* pWaitSemaphores;
+ uint32_t bufferBindCount;
+ const VkSparseBufferMemoryBindInfo* pBufferBinds;
+ uint32_t imageOpaqueBindCount;
+ const VkSparseImageOpaqueMemoryBindInfo* pImageOpaqueBinds;
+ uint32_t imageBindCount;
+ const VkSparseImageMemoryBindInfo* pImageBinds;
+ uint32_t signalSemaphoreCount;
+ const VkSemaphore* pSignalSemaphores;
+} VkBindSparseInfo;
+
+typedef struct VkFenceCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkFenceCreateFlags flags;
+} VkFenceCreateInfo;
+
+typedef struct VkSemaphoreCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkSemaphoreCreateFlags flags;
+} VkSemaphoreCreateInfo;
+
+typedef struct VkEventCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkEventCreateFlags flags;
+} VkEventCreateInfo;
+
+typedef struct VkQueryPoolCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkQueryPoolCreateFlags flags;
+ VkQueryType queryType;
+ uint32_t queryCount;
+ VkQueryPipelineStatisticFlags pipelineStatistics;
+} VkQueryPoolCreateInfo;
+
+typedef struct VkBufferCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkBufferCreateFlags flags;
+ VkDeviceSize size;
+ VkBufferUsageFlags usage;
+ VkSharingMode sharingMode;
+ uint32_t queueFamilyIndexCount;
+ const uint32_t* pQueueFamilyIndices;
+} VkBufferCreateInfo;
+
+typedef struct VkBufferViewCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkBufferViewCreateFlags flags;
+ VkBuffer buffer;
+ VkFormat format;
+ VkDeviceSize offset;
+ VkDeviceSize range;
+} VkBufferViewCreateInfo;
+
+typedef struct VkImageCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkImageCreateFlags flags;
+ VkImageType imageType;
+ VkFormat format;
+ VkExtent3D extent;
+ uint32_t mipLevels;
+ uint32_t arrayLayers;
+ VkSampleCountFlagBits samples;
+ VkImageTiling tiling;
+ VkImageUsageFlags usage;
+ VkSharingMode sharingMode;
+ uint32_t queueFamilyIndexCount;
+ const uint32_t* pQueueFamilyIndices;
+ VkImageLayout initialLayout;
+} VkImageCreateInfo;
+
+typedef struct VkSubresourceLayout {
+ VkDeviceSize offset;
+ VkDeviceSize size;
+ VkDeviceSize rowPitch;
+ VkDeviceSize arrayPitch;
+ VkDeviceSize depthPitch;
+} VkSubresourceLayout;
+
+typedef struct VkComponentMapping {
+ VkComponentSwizzle r;
+ VkComponentSwizzle g;
+ VkComponentSwizzle b;
+ VkComponentSwizzle a;
+} VkComponentMapping;
+
+typedef struct VkImageSubresourceRange {
+ VkImageAspectFlags aspectMask;
+ uint32_t baseMipLevel;
+ uint32_t levelCount;
+ uint32_t baseArrayLayer;
+ uint32_t layerCount;
+} VkImageSubresourceRange;
+
+typedef struct VkImageViewCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkImageViewCreateFlags flags;
+ VkImage image;
+ VkImageViewType viewType;
+ VkFormat format;
+ VkComponentMapping components;
+ VkImageSubresourceRange subresourceRange;
+} VkImageViewCreateInfo;
+
+typedef struct VkShaderModuleCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkShaderModuleCreateFlags flags;
+ size_t codeSize;
+ const uint32_t* pCode;
+} VkShaderModuleCreateInfo;
+
+typedef struct VkPipelineCacheCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkPipelineCacheCreateFlags flags;
+ size_t initialDataSize;
+ const void* pInitialData;
+} VkPipelineCacheCreateInfo;
+
+typedef struct VkSpecializationMapEntry {
+ uint32_t constantID;
+ uint32_t offset;
+ size_t size;
+} VkSpecializationMapEntry;
+
+typedef struct VkSpecializationInfo {
+ uint32_t mapEntryCount;
+ const VkSpecializationMapEntry* pMapEntries;
+ size_t dataSize;
+ const void* pData;
+} VkSpecializationInfo;
+
+typedef struct VkPipelineShaderStageCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkPipelineShaderStageCreateFlags flags;
+ VkShaderStageFlagBits stage;
+ VkShaderModule module;
+ const char* pName;
+ const VkSpecializationInfo* pSpecializationInfo;
+} VkPipelineShaderStageCreateInfo;
+
+typedef struct VkVertexInputBindingDescription {
+ uint32_t binding;
+ uint32_t stride;
+ VkVertexInputRate inputRate;
+} VkVertexInputBindingDescription;
+
+typedef struct VkVertexInputAttributeDescription {
+ uint32_t location;
+ uint32_t binding;
+ VkFormat format;
+ uint32_t offset;
+} VkVertexInputAttributeDescription;
+
+typedef struct VkPipelineVertexInputStateCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkPipelineVertexInputStateCreateFlags flags;
+ uint32_t vertexBindingDescriptionCount;
+ const VkVertexInputBindingDescription* pVertexBindingDescriptions;
+ uint32_t vertexAttributeDescriptionCount;
+ const VkVertexInputAttributeDescription* pVertexAttributeDescriptions;
+} VkPipelineVertexInputStateCreateInfo;
+
+typedef struct VkPipelineInputAssemblyStateCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkPipelineInputAssemblyStateCreateFlags flags;
+ VkPrimitiveTopology topology;
+ VkBool32 primitiveRestartEnable;
+} VkPipelineInputAssemblyStateCreateInfo;
+
+typedef struct VkPipelineTessellationStateCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkPipelineTessellationStateCreateFlags flags;
+ uint32_t patchControlPoints;
+} VkPipelineTessellationStateCreateInfo;
+
+typedef struct VkViewport {
+ float x;
+ float y;
+ float width;
+ float height;
+ float minDepth;
+ float maxDepth;
+} VkViewport;
+
+typedef struct VkOffset2D {
+ int32_t x;
+ int32_t y;
+} VkOffset2D;
+
+typedef struct VkExtent2D {
+ uint32_t width;
+ uint32_t height;
+} VkExtent2D;
+
+typedef struct VkRect2D {
+ VkOffset2D offset;
+ VkExtent2D extent;
+} VkRect2D;
+
+typedef struct VkPipelineViewportStateCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkPipelineViewportStateCreateFlags flags;
+ uint32_t viewportCount;
+ const VkViewport* pViewports;
+ uint32_t scissorCount;
+ const VkRect2D* pScissors;
+} VkPipelineViewportStateCreateInfo;
+
+typedef struct VkPipelineRasterizationStateCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkPipelineRasterizationStateCreateFlags flags;
+ VkBool32 depthClampEnable;
+ VkBool32 rasterizerDiscardEnable;
+ VkPolygonMode polygonMode;
+ VkCullModeFlags cullMode;
+ VkFrontFace frontFace;
+ VkBool32 depthBiasEnable;
+ float depthBiasConstantFactor;
+ float depthBiasClamp;
+ float depthBiasSlopeFactor;
+ float lineWidth;
+} VkPipelineRasterizationStateCreateInfo;
+
+typedef struct VkPipelineMultisampleStateCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkPipelineMultisampleStateCreateFlags flags;
+ VkSampleCountFlagBits rasterizationSamples;
+ VkBool32 sampleShadingEnable;
+ float minSampleShading;
+ const VkSampleMask* pSampleMask;
+ VkBool32 alphaToCoverageEnable;
+ VkBool32 alphaToOneEnable;
+} VkPipelineMultisampleStateCreateInfo;
+
+typedef struct VkStencilOpState {
+ VkStencilOp failOp;
+ VkStencilOp passOp;
+ VkStencilOp depthFailOp;
+ VkCompareOp compareOp;
+ uint32_t compareMask;
+ uint32_t writeMask;
+ uint32_t reference;
+} VkStencilOpState;
+
+typedef struct VkPipelineDepthStencilStateCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkPipelineDepthStencilStateCreateFlags flags;
+ VkBool32 depthTestEnable;
+ VkBool32 depthWriteEnable;
+ VkCompareOp depthCompareOp;
+ VkBool32 depthBoundsTestEnable;
+ VkBool32 stencilTestEnable;
+ VkStencilOpState front;
+ VkStencilOpState back;
+ float minDepthBounds;
+ float maxDepthBounds;
+} VkPipelineDepthStencilStateCreateInfo;
+
+typedef struct VkPipelineColorBlendAttachmentState {
+ VkBool32 blendEnable;
+ VkBlendFactor srcColorBlendFactor;
+ VkBlendFactor dstColorBlendFactor;
+ VkBlendOp colorBlendOp;
+ VkBlendFactor srcAlphaBlendFactor;
+ VkBlendFactor dstAlphaBlendFactor;
+ VkBlendOp alphaBlendOp;
+ VkColorComponentFlags colorWriteMask;
+} VkPipelineColorBlendAttachmentState;
+
+typedef struct VkPipelineColorBlendStateCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkPipelineColorBlendStateCreateFlags flags;
+ VkBool32 logicOpEnable;
+ VkLogicOp logicOp;
+ uint32_t attachmentCount;
+ const VkPipelineColorBlendAttachmentState* pAttachments;
+ float blendConstants[4];
+} VkPipelineColorBlendStateCreateInfo;
+
+typedef struct VkPipelineDynamicStateCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkPipelineDynamicStateCreateFlags flags;
+ uint32_t dynamicStateCount;
+ const VkDynamicState* pDynamicStates;
+} VkPipelineDynamicStateCreateInfo;
+
+typedef struct VkGraphicsPipelineCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkPipelineCreateFlags flags;
+ uint32_t stageCount;
+ const VkPipelineShaderStageCreateInfo* pStages;
+ const VkPipelineVertexInputStateCreateInfo* pVertexInputState;
+ const VkPipelineInputAssemblyStateCreateInfo* pInputAssemblyState;
+ const VkPipelineTessellationStateCreateInfo* pTessellationState;
+ const VkPipelineViewportStateCreateInfo* pViewportState;
+ const VkPipelineRasterizationStateCreateInfo* pRasterizationState;
+ const VkPipelineMultisampleStateCreateInfo* pMultisampleState;
+ const VkPipelineDepthStencilStateCreateInfo* pDepthStencilState;
+ const VkPipelineColorBlendStateCreateInfo* pColorBlendState;
+ const VkPipelineDynamicStateCreateInfo* pDynamicState;
+ VkPipelineLayout layout;
+ VkRenderPass renderPass;
+ uint32_t subpass;
+ VkPipeline basePipelineHandle;
+ int32_t basePipelineIndex;
+} VkGraphicsPipelineCreateInfo;
+
+typedef struct VkComputePipelineCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkPipelineCreateFlags flags;
+ VkPipelineShaderStageCreateInfo stage;
+ VkPipelineLayout layout;
+ VkPipeline basePipelineHandle;
+ int32_t basePipelineIndex;
+} VkComputePipelineCreateInfo;
+
+typedef struct VkPushConstantRange {
+ VkShaderStageFlags stageFlags;
+ uint32_t offset;
+ uint32_t size;
+} VkPushConstantRange;
+
+typedef struct VkPipelineLayoutCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkPipelineLayoutCreateFlags flags;
+ uint32_t setLayoutCount;
+ const VkDescriptorSetLayout* pSetLayouts;
+ uint32_t pushConstantRangeCount;
+ const VkPushConstantRange* pPushConstantRanges;
+} VkPipelineLayoutCreateInfo;
+
+typedef struct VkSamplerCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkSamplerCreateFlags flags;
+ VkFilter magFilter;
+ VkFilter minFilter;
+ VkSamplerMipmapMode mipmapMode;
+ VkSamplerAddressMode addressModeU;
+ VkSamplerAddressMode addressModeV;
+ VkSamplerAddressMode addressModeW;
+ float mipLodBias;
+ VkBool32 anisotropyEnable;
+ float maxAnisotropy;
+ VkBool32 compareEnable;
+ VkCompareOp compareOp;
+ float minLod;
+ float maxLod;
+ VkBorderColor borderColor;
+ VkBool32 unnormalizedCoordinates;
+} VkSamplerCreateInfo;
+
+typedef struct VkDescriptorSetLayoutBinding {
+ uint32_t binding;
+ VkDescriptorType descriptorType;
+ uint32_t descriptorCount;
+ VkShaderStageFlags stageFlags;
+ const VkSampler* pImmutableSamplers;
+} VkDescriptorSetLayoutBinding;
+
+typedef struct VkDescriptorSetLayoutCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkDescriptorSetLayoutCreateFlags flags;
+ uint32_t bindingCount;
+ const VkDescriptorSetLayoutBinding* pBindings;
+} VkDescriptorSetLayoutCreateInfo;
+
+typedef struct VkDescriptorPoolSize {
+ VkDescriptorType type;
+ uint32_t descriptorCount;
+} VkDescriptorPoolSize;
+
+typedef struct VkDescriptorPoolCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkDescriptorPoolCreateFlags flags;
+ uint32_t maxSets;
+ uint32_t poolSizeCount;
+ const VkDescriptorPoolSize* pPoolSizes;
+} VkDescriptorPoolCreateInfo;
+
+typedef struct VkDescriptorSetAllocateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkDescriptorPool descriptorPool;
+ uint32_t descriptorSetCount;
+ const VkDescriptorSetLayout* pSetLayouts;
+} VkDescriptorSetAllocateInfo;
+
+typedef struct VkDescriptorImageInfo {
+ VkSampler sampler;
+ VkImageView imageView;
+ VkImageLayout imageLayout;
+} VkDescriptorImageInfo;
+
+typedef struct VkDescriptorBufferInfo {
+ VkBuffer buffer;
+ VkDeviceSize offset;
+ VkDeviceSize range;
+} VkDescriptorBufferInfo;
+
+typedef struct VkWriteDescriptorSet {
+ VkStructureType sType;
+ const void* pNext;
+ VkDescriptorSet dstSet;
+ uint32_t dstBinding;
+ uint32_t dstArrayElement;
+ uint32_t descriptorCount;
+ VkDescriptorType descriptorType;
+ const VkDescriptorImageInfo* pImageInfo;
+ const VkDescriptorBufferInfo* pBufferInfo;
+ const VkBufferView* pTexelBufferView;
+} VkWriteDescriptorSet;
+
+typedef struct VkCopyDescriptorSet {
+ VkStructureType sType;
+ const void* pNext;
+ VkDescriptorSet srcSet;
+ uint32_t srcBinding;
+ uint32_t srcArrayElement;
+ VkDescriptorSet dstSet;
+ uint32_t dstBinding;
+ uint32_t dstArrayElement;
+ uint32_t descriptorCount;
+} VkCopyDescriptorSet;
+
+typedef struct VkFramebufferCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkFramebufferCreateFlags flags;
+ VkRenderPass renderPass;
+ uint32_t attachmentCount;
+ const VkImageView* pAttachments;
+ uint32_t width;
+ uint32_t height;
+ uint32_t layers;
+} VkFramebufferCreateInfo;
+
+typedef struct VkAttachmentDescription {
+ VkAttachmentDescriptionFlags flags;
+ VkFormat format;
+ VkSampleCountFlagBits samples;
+ VkAttachmentLoadOp loadOp;
+ VkAttachmentStoreOp storeOp;
+ VkAttachmentLoadOp stencilLoadOp;
+ VkAttachmentStoreOp stencilStoreOp;
+ VkImageLayout initialLayout;
+ VkImageLayout finalLayout;
+} VkAttachmentDescription;
+
+typedef struct VkAttachmentReference {
+ uint32_t attachment;
+ VkImageLayout layout;
+} VkAttachmentReference;
+
+typedef struct VkSubpassDescription {
+ VkSubpassDescriptionFlags flags;
+ VkPipelineBindPoint pipelineBindPoint;
+ uint32_t inputAttachmentCount;
+ const VkAttachmentReference* pInputAttachments;
+ uint32_t colorAttachmentCount;
+ const VkAttachmentReference* pColorAttachments;
+ const VkAttachmentReference* pResolveAttachments;
+ const VkAttachmentReference* pDepthStencilAttachment;
+ uint32_t preserveAttachmentCount;
+ const uint32_t* pPreserveAttachments;
+} VkSubpassDescription;
+
+typedef struct VkSubpassDependency {
+ uint32_t srcSubpass;
+ uint32_t dstSubpass;
+ VkPipelineStageFlags srcStageMask;
+ VkPipelineStageFlags dstStageMask;
+ VkAccessFlags srcAccessMask;
+ VkAccessFlags dstAccessMask;
+ VkDependencyFlags dependencyFlags;
+} VkSubpassDependency;
+
+typedef struct VkRenderPassCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkRenderPassCreateFlags flags;
+ uint32_t attachmentCount;
+ const VkAttachmentDescription* pAttachments;
+ uint32_t subpassCount;
+ const VkSubpassDescription* pSubpasses;
+ uint32_t dependencyCount;
+ const VkSubpassDependency* pDependencies;
+} VkRenderPassCreateInfo;
+
+typedef struct VkCommandPoolCreateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkCommandPoolCreateFlags flags;
+ uint32_t queueFamilyIndex;
+} VkCommandPoolCreateInfo;
+
+typedef struct VkCommandBufferAllocateInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkCommandPool commandPool;
+ VkCommandBufferLevel level;
+ uint32_t commandBufferCount;
+} VkCommandBufferAllocateInfo;
+
+typedef struct VkCommandBufferInheritanceInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkRenderPass renderPass;
+ uint32_t subpass;
+ VkFramebuffer framebuffer;
+ VkBool32 occlusionQueryEnable;
+ VkQueryControlFlags queryFlags;
+ VkQueryPipelineStatisticFlags pipelineStatistics;
+} VkCommandBufferInheritanceInfo;
+
+typedef struct VkCommandBufferBeginInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkCommandBufferUsageFlags flags;
+ const VkCommandBufferInheritanceInfo* pInheritanceInfo;
+} VkCommandBufferBeginInfo;
+
+typedef struct VkBufferCopy {
+ VkDeviceSize srcOffset;
+ VkDeviceSize dstOffset;
+ VkDeviceSize size;
+} VkBufferCopy;
+
+typedef struct VkImageSubresourceLayers {
+ VkImageAspectFlags aspectMask;
+ uint32_t mipLevel;
+ uint32_t baseArrayLayer;
+ uint32_t layerCount;
+} VkImageSubresourceLayers;
+
+typedef struct VkImageCopy {
+ VkImageSubresourceLayers srcSubresource;
+ VkOffset3D srcOffset;
+ VkImageSubresourceLayers dstSubresource;
+ VkOffset3D dstOffset;
+ VkExtent3D extent;
+} VkImageCopy;
+
+typedef struct VkImageBlit {
+ VkImageSubresourceLayers srcSubresource;
+ VkOffset3D srcOffsets[2];
+ VkImageSubresourceLayers dstSubresource;
+ VkOffset3D dstOffsets[2];
+} VkImageBlit;
+
+typedef struct VkBufferImageCopy {
+ VkDeviceSize bufferOffset;
+ uint32_t bufferRowLength;
+ uint32_t bufferImageHeight;
+ VkImageSubresourceLayers imageSubresource;
+ VkOffset3D imageOffset;
+ VkExtent3D imageExtent;
+} VkBufferImageCopy;
+
+typedef union VkClearColorValue {
+ float float32[4];
+ int32_t int32[4];
+ uint32_t uint32[4];
+} VkClearColorValue;
+
+typedef struct VkClearDepthStencilValue {
+ float depth;
+ uint32_t stencil;
+} VkClearDepthStencilValue;
+
+typedef union VkClearValue {
+ VkClearColorValue color;
+ VkClearDepthStencilValue depthStencil;
+} VkClearValue;
+
+typedef struct VkClearAttachment {
+ VkImageAspectFlags aspectMask;
+ uint32_t colorAttachment;
+ VkClearValue clearValue;
+} VkClearAttachment;
+
+typedef struct VkClearRect {
+ VkRect2D rect;
+ uint32_t baseArrayLayer;
+ uint32_t layerCount;
+} VkClearRect;
+
+typedef struct VkImageResolve {
+ VkImageSubresourceLayers srcSubresource;
+ VkOffset3D srcOffset;
+ VkImageSubresourceLayers dstSubresource;
+ VkOffset3D dstOffset;
+ VkExtent3D extent;
+} VkImageResolve;
+
+typedef struct VkMemoryBarrier {
+ VkStructureType sType;
+ const void* pNext;
+ VkAccessFlags srcAccessMask;
+ VkAccessFlags dstAccessMask;
+} VkMemoryBarrier;
+
+typedef struct VkBufferMemoryBarrier {
+ VkStructureType sType;
+ const void* pNext;
+ VkAccessFlags srcAccessMask;
+ VkAccessFlags dstAccessMask;
+ uint32_t srcQueueFamilyIndex;
+ uint32_t dstQueueFamilyIndex;
+ VkBuffer buffer;
+ VkDeviceSize offset;
+ VkDeviceSize size;
+} VkBufferMemoryBarrier;
+
+typedef struct VkImageMemoryBarrier {
+ VkStructureType sType;
+ const void* pNext;
+ VkAccessFlags srcAccessMask;
+ VkAccessFlags dstAccessMask;
+ VkImageLayout oldLayout;
+ VkImageLayout newLayout;
+ uint32_t srcQueueFamilyIndex;
+ uint32_t dstQueueFamilyIndex;
+ VkImage image;
+ VkImageSubresourceRange subresourceRange;
+} VkImageMemoryBarrier;
+
+typedef struct VkRenderPassBeginInfo {
+ VkStructureType sType;
+ const void* pNext;
+ VkRenderPass renderPass;
+ VkFramebuffer framebuffer;
+ VkRect2D renderArea;
+ uint32_t clearValueCount;
+ const VkClearValue* pClearValues;
+} VkRenderPassBeginInfo;
+
+typedef struct VkDispatchIndirectCommand {
+ uint32_t x;
+ uint32_t y;
+ uint32_t z;
+} VkDispatchIndirectCommand;
+
+typedef struct VkDrawIndexedIndirectCommand {
+ uint32_t indexCount;
+ uint32_t instanceCount;
+ uint32_t firstIndex;
+ int32_t vertexOffset;
+ uint32_t firstInstance;
+} VkDrawIndexedIndirectCommand;
+
+typedef struct VkDrawIndirectCommand {
+ uint32_t vertexCount;
+ uint32_t instanceCount;
+ uint32_t firstVertex;
+ uint32_t firstInstance;
+} VkDrawIndirectCommand;
+
+
+typedef VkResult (VKAPI_PTR *PFN_vkCreateInstance)(const VkInstanceCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkInstance* pInstance);
+typedef void (VKAPI_PTR *PFN_vkDestroyInstance)(VkInstance instance, const VkAllocationCallbacks* pAllocator);
+typedef VkResult (VKAPI_PTR *PFN_vkEnumeratePhysicalDevices)(VkInstance instance, uint32_t* pPhysicalDeviceCount, VkPhysicalDevice* pPhysicalDevices);
+typedef void (VKAPI_PTR *PFN_vkGetPhysicalDeviceFeatures)(VkPhysicalDevice physicalDevice, VkPhysicalDeviceFeatures* pFeatures);
+typedef void (VKAPI_PTR *PFN_vkGetPhysicalDeviceFormatProperties)(VkPhysicalDevice physicalDevice, VkFormat format, VkFormatProperties* pFormatProperties);
+typedef VkResult (VKAPI_PTR *PFN_vkGetPhysicalDeviceImageFormatProperties)(VkPhysicalDevice physicalDevice, VkFormat format, VkImageType type, VkImageTiling tiling, VkImageUsageFlags usage, VkImageCreateFlags flags, VkImageFormatProperties* pImageFormatProperties);
+typedef void (VKAPI_PTR *PFN_vkGetPhysicalDeviceProperties)(VkPhysicalDevice physicalDevice, VkPhysicalDeviceProperties* pProperties);
+typedef void (VKAPI_PTR *PFN_vkGetPhysicalDeviceQueueFamilyProperties)(VkPhysicalDevice physicalDevice, uint32_t* pQueueFamilyPropertyCount, VkQueueFamilyProperties* pQueueFamilyProperties);
+typedef void (VKAPI_PTR *PFN_vkGetPhysicalDeviceMemoryProperties)(VkPhysicalDevice physicalDevice, VkPhysicalDeviceMemoryProperties* pMemoryProperties);
+typedef PFN_vkVoidFunction (VKAPI_PTR *PFN_vkGetInstanceProcAddr)(VkInstance instance, const char* pName);
+typedef PFN_vkVoidFunction (VKAPI_PTR *PFN_vkGetDeviceProcAddr)(VkDevice device, const char* pName);
+typedef VkResult (VKAPI_PTR *PFN_vkCreateDevice)(VkPhysicalDevice physicalDevice, const VkDeviceCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkDevice* pDevice);
+typedef void (VKAPI_PTR *PFN_vkDestroyDevice)(VkDevice device, const VkAllocationCallbacks* pAllocator);
+typedef VkResult (VKAPI_PTR *PFN_vkEnumerateInstanceExtensionProperties)(const char* pLayerName, uint32_t* pPropertyCount, VkExtensionProperties* pProperties);
+typedef VkResult (VKAPI_PTR *PFN_vkEnumerateDeviceExtensionProperties)(VkPhysicalDevice physicalDevice, const char* pLayerName, uint32_t* pPropertyCount, VkExtensionProperties* pProperties);
+typedef VkResult (VKAPI_PTR *PFN_vkEnumerateInstanceLayerProperties)(uint32_t* pPropertyCount, VkLayerProperties* pProperties);
+typedef VkResult (VKAPI_PTR *PFN_vkEnumerateDeviceLayerProperties)(VkPhysicalDevice physicalDevice, uint32_t* pPropertyCount, VkLayerProperties* pProperties);
+typedef void (VKAPI_PTR *PFN_vkGetDeviceQueue)(VkDevice device, uint32_t queueFamilyIndex, uint32_t queueIndex, VkQueue* pQueue);
+typedef VkResult (VKAPI_PTR *PFN_vkQueueSubmit)(VkQueue queue, uint32_t submitCount, const VkSubmitInfo* pSubmits, VkFence fence);
+typedef VkResult (VKAPI_PTR *PFN_vkQueueWaitIdle)(VkQueue queue);
+typedef VkResult (VKAPI_PTR *PFN_vkDeviceWaitIdle)(VkDevice device);
+typedef VkResult (VKAPI_PTR *PFN_vkAllocateMemory)(VkDevice device, const VkMemoryAllocateInfo* pAllocateInfo, const VkAllocationCallbacks* pAllocator, VkDeviceMemory* pMemory);
+typedef void (VKAPI_PTR *PFN_vkFreeMemory)(VkDevice device, VkDeviceMemory memory, const VkAllocationCallbacks* pAllocator);
+typedef VkResult (VKAPI_PTR *PFN_vkMapMemory)(VkDevice device, VkDeviceMemory memory, VkDeviceSize offset, VkDeviceSize size, VkMemoryMapFlags flags, void** ppData);
+typedef void (VKAPI_PTR *PFN_vkUnmapMemory)(VkDevice device, VkDeviceMemory memory);
+typedef VkResult (VKAPI_PTR *PFN_vkFlushMappedMemoryRanges)(VkDevice device, uint32_t memoryRangeCount, const VkMappedMemoryRange* pMemoryRanges);
+typedef VkResult (VKAPI_PTR *PFN_vkInvalidateMappedMemoryRanges)(VkDevice device, uint32_t memoryRangeCount, const VkMappedMemoryRange* pMemoryRanges);
+typedef void (VKAPI_PTR *PFN_vkGetDeviceMemoryCommitment)(VkDevice device, VkDeviceMemory memory, VkDeviceSize* pCommittedMemoryInBytes);
+typedef VkResult (VKAPI_PTR *PFN_vkBindBufferMemory)(VkDevice device, VkBuffer buffer, VkDeviceMemory memory, VkDeviceSize memoryOffset);
+typedef VkResult (VKAPI_PTR *PFN_vkBindImageMemory)(VkDevice device, VkImage image, VkDeviceMemory memory, VkDeviceSize memoryOffset);
+typedef void (VKAPI_PTR *PFN_vkGetBufferMemoryRequirements)(VkDevice device, VkBuffer buffer, VkMemoryRequirements* pMemoryRequirements);
+typedef void (VKAPI_PTR *PFN_vkGetImageMemoryRequirements)(VkDevice device, VkImage image, VkMemoryRequirements* pMemoryRequirements);
+typedef void (VKAPI_PTR *PFN_vkGetImageSparseMemoryRequirements)(VkDevice device, VkImage image, uint32_t* pSparseMemoryRequirementCount, VkSparseImageMemoryRequirements* pSparseMemoryRequirements);
+typedef void (VKAPI_PTR *PFN_vkGetPhysicalDeviceSparseImageFormatProperties)(VkPhysicalDevice physicalDevice, VkFormat format, VkImageType type, VkSampleCountFlagBits samples, VkImageUsageFlags usage, VkImageTiling tiling, uint32_t* pPropertyCount, VkSparseImageFormatProperties* pProperties);
+typedef VkResult (VKAPI_PTR *PFN_vkQueueBindSparse)(VkQueue queue, uint32_t bindInfoCount, const VkBindSparseInfo* pBindInfo, VkFence fence);
+typedef VkResult (VKAPI_PTR *PFN_vkCreateFence)(VkDevice device, const VkFenceCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkFence* pFence);
+typedef void (VKAPI_PTR *PFN_vkDestroyFence)(VkDevice device, VkFence fence, const VkAllocationCallbacks* pAllocator);
+typedef VkResult (VKAPI_PTR *PFN_vkResetFences)(VkDevice device, uint32_t fenceCount, const VkFence* pFences);
+typedef VkResult (VKAPI_PTR *PFN_vkGetFenceStatus)(VkDevice device, VkFence fence);
+typedef VkResult (VKAPI_PTR *PFN_vkWaitForFences)(VkDevice device, uint32_t fenceCount, const VkFence* pFences, VkBool32 waitAll, uint64_t timeout);
+typedef VkResult (VKAPI_PTR *PFN_vkCreateSemaphore)(VkDevice device, const VkSemaphoreCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkSemaphore* pSemaphore);
+typedef void (VKAPI_PTR *PFN_vkDestroySemaphore)(VkDevice device, VkSemaphore semaphore, const VkAllocationCallbacks* pAllocator);
+typedef VkResult (VKAPI_PTR *PFN_vkCreateEvent)(VkDevice device, const VkEventCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkEvent* pEvent);
+typedef void (VKAPI_PTR *PFN_vkDestroyEvent)(VkDevice device, VkEvent event, const VkAllocationCallbacks* pAllocator);
+typedef VkResult (VKAPI_PTR *PFN_vkGetEventStatus)(VkDevice device, VkEvent event);
+typedef VkResult (VKAPI_PTR *PFN_vkSetEvent)(VkDevice device, VkEvent event);
+typedef VkResult (VKAPI_PTR *PFN_vkResetEvent)(VkDevice device, VkEvent event);
+typedef VkResult (VKAPI_PTR *PFN_vkCreateQueryPool)(VkDevice device, const VkQueryPoolCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkQueryPool* pQueryPool);
+typedef void (VKAPI_PTR *PFN_vkDestroyQueryPool)(VkDevice device, VkQueryPool queryPool, const VkAllocationCallbacks* pAllocator);
+typedef VkResult (VKAPI_PTR *PFN_vkGetQueryPoolResults)(VkDevice device, VkQueryPool queryPool, uint32_t firstQuery, uint32_t queryCount, size_t dataSize, void* pData, VkDeviceSize stride, VkQueryResultFlags flags);
+typedef VkResult (VKAPI_PTR *PFN_vkCreateBuffer)(VkDevice device, const VkBufferCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkBuffer* pBuffer);
+typedef void (VKAPI_PTR *PFN_vkDestroyBuffer)(VkDevice device, VkBuffer buffer, const VkAllocationCallbacks* pAllocator);
+typedef VkResult (VKAPI_PTR *PFN_vkCreateBufferView)(VkDevice device, const VkBufferViewCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkBufferView* pView);
+typedef void (VKAPI_PTR *PFN_vkDestroyBufferView)(VkDevice device, VkBufferView bufferView, const VkAllocationCallbacks* pAllocator);
+typedef VkResult (VKAPI_PTR *PFN_vkCreateImage)(VkDevice device, const VkImageCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkImage* pImage);
+typedef void (VKAPI_PTR *PFN_vkDestroyImage)(VkDevice device, VkImage image, const VkAllocationCallbacks* pAllocator);
+typedef void (VKAPI_PTR *PFN_vkGetImageSubresourceLayout)(VkDevice device, VkImage image, const VkImageSubresource* pSubresource, VkSubresourceLayout* pLayout);
+typedef VkResult (VKAPI_PTR *PFN_vkCreateImageView)(VkDevice device, const VkImageViewCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkImageView* pView);
+typedef void (VKAPI_PTR *PFN_vkDestroyImageView)(VkDevice device, VkImageView imageView, const VkAllocationCallbacks* pAllocator);
+typedef VkResult (VKAPI_PTR *PFN_vkCreateShaderModule)(VkDevice device, const VkShaderModuleCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkShaderModule* pShaderModule);
+typedef void (VKAPI_PTR *PFN_vkDestroyShaderModule)(VkDevice device, VkShaderModule shaderModule, const VkAllocationCallbacks* pAllocator);
+typedef VkResult (VKAPI_PTR *PFN_vkCreatePipelineCache)(VkDevice device, const VkPipelineCacheCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkPipelineCache* pPipelineCache);
+typedef void (VKAPI_PTR *PFN_vkDestroyPipelineCache)(VkDevice device, VkPipelineCache pipelineCache, const VkAllocationCallbacks* pAllocator);
+typedef VkResult (VKAPI_PTR *PFN_vkGetPipelineCacheData)(VkDevice device, VkPipelineCache pipelineCache, size_t* pDataSize, void* pData);
+typedef VkResult (VKAPI_PTR *PFN_vkMergePipelineCaches)(VkDevice device, VkPipelineCache dstCache, uint32_t srcCacheCount, const VkPipelineCache* pSrcCaches);
+typedef VkResult (VKAPI_PTR *PFN_vkCreateGraphicsPipelines)(VkDevice device, VkPipelineCache pipelineCache, uint32_t createInfoCount, const VkGraphicsPipelineCreateInfo* pCreateInfos, const VkAllocationCallbacks* pAllocator, VkPipeline* pPipelines);
+typedef VkResult (VKAPI_PTR *PFN_vkCreateComputePipelines)(VkDevice device, VkPipelineCache pipelineCache, uint32_t createInfoCount, const VkComputePipelineCreateInfo* pCreateInfos, const VkAllocationCallbacks* pAllocator, VkPipeline* pPipelines);
+typedef void (VKAPI_PTR *PFN_vkDestroyPipeline)(VkDevice device, VkPipeline pipeline, const VkAllocationCallbacks* pAllocator);
+typedef VkResult (VKAPI_PTR *PFN_vkCreatePipelineLayout)(VkDevice device, const VkPipelineLayoutCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkPipelineLayout* pPipelineLayout);
+typedef void (VKAPI_PTR *PFN_vkDestroyPipelineLayout)(VkDevice device, VkPipelineLayout pipelineLayout, const VkAllocationCallbacks* pAllocator);
+typedef VkResult (VKAPI_PTR *PFN_vkCreateSampler)(VkDevice device, const VkSamplerCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkSampler* pSampler);
+typedef void (VKAPI_PTR *PFN_vkDestroySampler)(VkDevice device, VkSampler sampler, const VkAllocationCallbacks* pAllocator);
+typedef VkResult (VKAPI_PTR *PFN_vkCreateDescriptorSetLayout)(VkDevice device, const VkDescriptorSetLayoutCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkDescriptorSetLayout* pSetLayout);
+typedef void (VKAPI_PTR *PFN_vkDestroyDescriptorSetLayout)(VkDevice device, VkDescriptorSetLayout descriptorSetLayout, const VkAllocationCallbacks* pAllocator);
+typedef VkResult (VKAPI_PTR *PFN_vkCreateDescriptorPool)(VkDevice device, const VkDescriptorPoolCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkDescriptorPool* pDescriptorPool);
+typedef void (VKAPI_PTR *PFN_vkDestroyDescriptorPool)(VkDevice device, VkDescriptorPool descriptorPool, const VkAllocationCallbacks* pAllocator);
+typedef VkResult (VKAPI_PTR *PFN_vkResetDescriptorPool)(VkDevice device, VkDescriptorPool descriptorPool, VkDescriptorPoolResetFlags flags);
+typedef VkResult (VKAPI_PTR *PFN_vkAllocateDescriptorSets)(VkDevice device, const VkDescriptorSetAllocateInfo* pAllocateInfo, VkDescriptorSet* pDescriptorSets);
+typedef VkResult (VKAPI_PTR *PFN_vkFreeDescriptorSets)(VkDevice device, VkDescriptorPool descriptorPool, uint32_t descriptorSetCount, const VkDescriptorSet* pDescriptorSets);
+typedef void (VKAPI_PTR *PFN_vkUpdateDescriptorSets)(VkDevice device, uint32_t descriptorWriteCount, const VkWriteDescriptorSet* pDescriptorWrites, uint32_t descriptorCopyCount, const VkCopyDescriptorSet* pDescriptorCopies);
+typedef VkResult (VKAPI_PTR *PFN_vkCreateFramebuffer)(VkDevice device, const VkFramebufferCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkFramebuffer* pFramebuffer);
+typedef void (VKAPI_PTR *PFN_vkDestroyFramebuffer)(VkDevice device, VkFramebuffer framebuffer, const VkAllocationCallbacks* pAllocator);
+typedef VkResult (VKAPI_PTR *PFN_vkCreateRenderPass)(VkDevice device, const VkRenderPassCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkRenderPass* pRenderPass);
+typedef void (VKAPI_PTR *PFN_vkDestroyRenderPass)(VkDevice device, VkRenderPass renderPass, const VkAllocationCallbacks* pAllocator);
+typedef void (VKAPI_PTR *PFN_vkGetRenderAreaGranularity)(VkDevice device, VkRenderPass renderPass, VkExtent2D* pGranularity);
+typedef VkResult (VKAPI_PTR *PFN_vkCreateCommandPool)(VkDevice device, const VkCommandPoolCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkCommandPool* pCommandPool);
+typedef void (VKAPI_PTR *PFN_vkDestroyCommandPool)(VkDevice device, VkCommandPool commandPool, const VkAllocationCallbacks* pAllocator);
+typedef VkResult (VKAPI_PTR *PFN_vkResetCommandPool)(VkDevice device, VkCommandPool commandPool, VkCommandPoolResetFlags flags);
+typedef VkResult (VKAPI_PTR *PFN_vkAllocateCommandBuffers)(VkDevice device, const VkCommandBufferAllocateInfo* pAllocateInfo, VkCommandBuffer* pCommandBuffers);
+typedef void (VKAPI_PTR *PFN_vkFreeCommandBuffers)(VkDevice device, VkCommandPool commandPool, uint32_t commandBufferCount, const VkCommandBuffer* pCommandBuffers);
+typedef VkResult (VKAPI_PTR *PFN_vkBeginCommandBuffer)(VkCommandBuffer commandBuffer, const VkCommandBufferBeginInfo* pBeginInfo);
+typedef VkResult (VKAPI_PTR *PFN_vkEndCommandBuffer)(VkCommandBuffer commandBuffer);
+typedef VkResult (VKAPI_PTR *PFN_vkResetCommandBuffer)(VkCommandBuffer commandBuffer, VkCommandBufferResetFlags flags);
+typedef void (VKAPI_PTR *PFN_vkCmdBindPipeline)(VkCommandBuffer commandBuffer, VkPipelineBindPoint pipelineBindPoint, VkPipeline pipeline);
+typedef void (VKAPI_PTR *PFN_vkCmdSetViewport)(VkCommandBuffer commandBuffer, uint32_t firstViewport, uint32_t viewportCount, const VkViewport* pViewports);
+typedef void (VKAPI_PTR *PFN_vkCmdSetScissor)(VkCommandBuffer commandBuffer, uint32_t firstScissor, uint32_t scissorCount, const VkRect2D* pScissors);
+typedef void (VKAPI_PTR *PFN_vkCmdSetLineWidth)(VkCommandBuffer commandBuffer, float lineWidth);
+typedef void (VKAPI_PTR *PFN_vkCmdSetDepthBias)(VkCommandBuffer commandBuffer, float depthBiasConstantFactor, float depthBiasClamp, float depthBiasSlopeFactor);
+typedef void (VKAPI_PTR *PFN_vkCmdSetBlendConstants)(VkCommandBuffer commandBuffer, const float blendConstants[4]);
+typedef void (VKAPI_PTR *PFN_vkCmdSetDepthBounds)(VkCommandBuffer commandBuffer, float minDepthBounds, float maxDepthBounds);
+typedef void (VKAPI_PTR *PFN_vkCmdSetStencilCompareMask)(VkCommandBuffer commandBuffer, VkStencilFaceFlags faceMask, uint32_t compareMask);
+typedef void (VKAPI_PTR *PFN_vkCmdSetStencilWriteMask)(VkCommandBuffer commandBuffer, VkStencilFaceFlags faceMask, uint32_t writeMask);
+typedef void (VKAPI_PTR *PFN_vkCmdSetStencilReference)(VkCommandBuffer commandBuffer, VkStencilFaceFlags faceMask, uint32_t reference);
+typedef void (VKAPI_PTR *PFN_vkCmdBindDescriptorSets)(VkCommandBuffer commandBuffer, VkPipelineBindPoint pipelineBindPoint, VkPipelineLayout layout, uint32_t firstSet, uint32_t descriptorSetCount, const VkDescriptorSet* pDescriptorSets, uint32_t dynamicOffsetCount, const uint32_t* pDynamicOffsets);
+typedef void (VKAPI_PTR *PFN_vkCmdBindIndexBuffer)(VkCommandBuffer commandBuffer, VkBuffer buffer, VkDeviceSize offset, VkIndexType indexType);
+typedef void (VKAPI_PTR *PFN_vkCmdBindVertexBuffers)(VkCommandBuffer commandBuffer, uint32_t firstBinding, uint32_t bindingCount, const VkBuffer* pBuffers, const VkDeviceSize* pOffsets);
+typedef void (VKAPI_PTR *PFN_vkCmdDraw)(VkCommandBuffer commandBuffer, uint32_t vertexCount, uint32_t instanceCount, uint32_t firstVertex, uint32_t firstInstance);
+typedef void (VKAPI_PTR *PFN_vkCmdDrawIndexed)(VkCommandBuffer commandBuffer, uint32_t indexCount, uint32_t instanceCount, uint32_t firstIndex, int32_t vertexOffset, uint32_t firstInstance);
+typedef void (VKAPI_PTR *PFN_vkCmdDrawIndirect)(VkCommandBuffer commandBuffer, VkBuffer buffer, VkDeviceSize offset, uint32_t drawCount, uint32_t stride);
+typedef void (VKAPI_PTR *PFN_vkCmdDrawIndexedIndirect)(VkCommandBuffer commandBuffer, VkBuffer buffer, VkDeviceSize offset, uint32_t drawCount, uint32_t stride);
+typedef void (VKAPI_PTR *PFN_vkCmdDispatch)(VkCommandBuffer commandBuffer, uint32_t x, uint32_t y, uint32_t z);
+typedef void (VKAPI_PTR *PFN_vkCmdDispatchIndirect)(VkCommandBuffer commandBuffer, VkBuffer buffer, VkDeviceSize offset);
+typedef void (VKAPI_PTR *PFN_vkCmdCopyBuffer)(VkCommandBuffer commandBuffer, VkBuffer srcBuffer, VkBuffer dstBuffer, uint32_t regionCount, const VkBufferCopy* pRegions);
+typedef void (VKAPI_PTR *PFN_vkCmdCopyImage)(VkCommandBuffer commandBuffer, VkImage srcImage, VkImageLayout srcImageLayout, VkImage dstImage, VkImageLayout dstImageLayout, uint32_t regionCount, const VkImageCopy* pRegions);
+typedef void (VKAPI_PTR *PFN_vkCmdBlitImage)(VkCommandBuffer commandBuffer, VkImage srcImage, VkImageLayout srcImageLayout, VkImage dstImage, VkImageLayout dstImageLayout, uint32_t regionCount, const VkImageBlit* pRegions, VkFilter filter);
+typedef void (VKAPI_PTR *PFN_vkCmdCopyBufferToImage)(VkCommandBuffer commandBuffer, VkBuffer srcBuffer, VkImage dstImage, VkImageLayout dstImageLayout, uint32_t regionCount, const VkBufferImageCopy* pRegions);
+typedef void (VKAPI_PTR *PFN_vkCmdCopyImageToBuffer)(VkCommandBuffer commandBuffer, VkImage srcImage, VkImageLayout srcImageLayout, VkBuffer dstBuffer, uint32_t regionCount, const VkBufferImageCopy* pRegions);
+typedef void (VKAPI_PTR *PFN_vkCmdUpdateBuffer)(VkCommandBuffer commandBuffer, VkBuffer dstBuffer, VkDeviceSize dstOffset, VkDeviceSize dataSize, const uint32_t* pData);
+typedef void (VKAPI_PTR *PFN_vkCmdFillBuffer)(VkCommandBuffer commandBuffer, VkBuffer dstBuffer, VkDeviceSize dstOffset, VkDeviceSize size, uint32_t data);
+typedef void (VKAPI_PTR *PFN_vkCmdClearColorImage)(VkCommandBuffer commandBuffer, VkImage image, VkImageLayout imageLayout, const VkClearColorValue* pColor, uint32_t rangeCount, const VkImageSubresourceRange* pRanges);
+typedef void (VKAPI_PTR *PFN_vkCmdClearDepthStencilImage)(VkCommandBuffer commandBuffer, VkImage image, VkImageLayout imageLayout, const VkClearDepthStencilValue* pDepthStencil, uint32_t rangeCount, const VkImageSubresourceRange* pRanges);
+typedef void (VKAPI_PTR *PFN_vkCmdClearAttachments)(VkCommandBuffer commandBuffer, uint32_t attachmentCount, const VkClearAttachment* pAttachments, uint32_t rectCount, const VkClearRect* pRects);
+typedef void (VKAPI_PTR *PFN_vkCmdResolveImage)(VkCommandBuffer commandBuffer, VkImage srcImage, VkImageLayout srcImageLayout, VkImage dstImage, VkImageLayout dstImageLayout, uint32_t regionCount, const VkImageResolve* pRegions);
+typedef void (VKAPI_PTR *PFN_vkCmdSetEvent)(VkCommandBuffer commandBuffer, VkEvent event, VkPipelineStageFlags stageMask);
+typedef void (VKAPI_PTR *PFN_vkCmdResetEvent)(VkCommandBuffer commandBuffer, VkEvent event, VkPipelineStageFlags stageMask);
+typedef void (VKAPI_PTR *PFN_vkCmdWaitEvents)(VkCommandBuffer commandBuffer, uint32_t eventCount, const VkEvent* pEvents, VkPipelineStageFlags srcStageMask, VkPipelineStageFlags dstStageMask, uint32_t memoryBarrierCount, const VkMemoryBarrier* pMemoryBarriers, uint32_t bufferMemoryBarrierCount, const VkBufferMemoryBarrier* pBufferMemoryBarriers, uint32_t imageMemoryBarrierCount, const VkImageMemoryBarrier* pImageMemoryBarriers);
+typedef void (VKAPI_PTR *PFN_vkCmdPipelineBarrier)(VkCommandBuffer commandBuffer, VkPipelineStageFlags srcStageMask, VkPipelineStageFlags dstStageMask, VkDependencyFlags dependencyFlags, uint32_t memoryBarrierCount, const VkMemoryBarrier* pMemoryBarriers, uint32_t bufferMemoryBarrierCount, const VkBufferMemoryBarrier* pBufferMemoryBarriers, uint32_t imageMemoryBarrierCount, const VkImageMemoryBarrier* pImageMemoryBarriers);
+typedef void (VKAPI_PTR *PFN_vkCmdBeginQuery)(VkCommandBuffer commandBuffer, VkQueryPool queryPool, uint32_t query, VkQueryControlFlags flags);
+typedef void (VKAPI_PTR *PFN_vkCmdEndQuery)(VkCommandBuffer commandBuffer, VkQueryPool queryPool, uint32_t query);
+typedef void (VKAPI_PTR *PFN_vkCmdResetQueryPool)(VkCommandBuffer commandBuffer, VkQueryPool queryPool, uint32_t firstQuery, uint32_t queryCount);
+typedef void (VKAPI_PTR *PFN_vkCmdWriteTimestamp)(VkCommandBuffer commandBuffer, VkPipelineStageFlagBits pipelineStage, VkQueryPool queryPool, uint32_t query);
+typedef void (VKAPI_PTR *PFN_vkCmdCopyQueryPoolResults)(VkCommandBuffer commandBuffer, VkQueryPool queryPool, uint32_t firstQuery, uint32_t queryCount, VkBuffer dstBuffer, VkDeviceSize dstOffset, VkDeviceSize stride, VkQueryResultFlags flags);
+typedef void (VKAPI_PTR *PFN_vkCmdPushConstants)(VkCommandBuffer commandBuffer, VkPipelineLayout layout, VkShaderStageFlags stageFlags, uint32_t offset, uint32_t size, const void* pValues);
+typedef void (VKAPI_PTR *PFN_vkCmdBeginRenderPass)(VkCommandBuffer commandBuffer, const VkRenderPassBeginInfo* pRenderPassBegin, VkSubpassContents contents);
+typedef void (VKAPI_PTR *PFN_vkCmdNextSubpass)(VkCommandBuffer commandBuffer, VkSubpassContents contents);
+typedef void (VKAPI_PTR *PFN_vkCmdEndRenderPass)(VkCommandBuffer commandBuffer);
+typedef void (VKAPI_PTR *PFN_vkCmdExecuteCommands)(VkCommandBuffer commandBuffer, uint32_t commandBufferCount, const VkCommandBuffer* pCommandBuffers);
+
+#ifndef VK_NO_PROTOTYPES
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateInstance(
+ const VkInstanceCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkInstance* pInstance);
+
+VKAPI_ATTR void VKAPI_CALL vkDestroyInstance(
+ VkInstance instance,
+ const VkAllocationCallbacks* pAllocator);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkEnumeratePhysicalDevices(
+ VkInstance instance,
+ uint32_t* pPhysicalDeviceCount,
+ VkPhysicalDevice* pPhysicalDevices);
+
+VKAPI_ATTR void VKAPI_CALL vkGetPhysicalDeviceFeatures(
+ VkPhysicalDevice physicalDevice,
+ VkPhysicalDeviceFeatures* pFeatures);
+
+VKAPI_ATTR void VKAPI_CALL vkGetPhysicalDeviceFormatProperties(
+ VkPhysicalDevice physicalDevice,
+ VkFormat format,
+ VkFormatProperties* pFormatProperties);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkGetPhysicalDeviceImageFormatProperties(
+ VkPhysicalDevice physicalDevice,
+ VkFormat format,
+ VkImageType type,
+ VkImageTiling tiling,
+ VkImageUsageFlags usage,
+ VkImageCreateFlags flags,
+ VkImageFormatProperties* pImageFormatProperties);
+
+VKAPI_ATTR void VKAPI_CALL vkGetPhysicalDeviceProperties(
+ VkPhysicalDevice physicalDevice,
+ VkPhysicalDeviceProperties* pProperties);
+
+VKAPI_ATTR void VKAPI_CALL vkGetPhysicalDeviceQueueFamilyProperties(
+ VkPhysicalDevice physicalDevice,
+ uint32_t* pQueueFamilyPropertyCount,
+ VkQueueFamilyProperties* pQueueFamilyProperties);
+
+VKAPI_ATTR void VKAPI_CALL vkGetPhysicalDeviceMemoryProperties(
+ VkPhysicalDevice physicalDevice,
+ VkPhysicalDeviceMemoryProperties* pMemoryProperties);
+
+VKAPI_ATTR PFN_vkVoidFunction VKAPI_CALL vkGetInstanceProcAddr(
+ VkInstance instance,
+ const char* pName);
+
+VKAPI_ATTR PFN_vkVoidFunction VKAPI_CALL vkGetDeviceProcAddr(
+ VkDevice device,
+ const char* pName);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateDevice(
+ VkPhysicalDevice physicalDevice,
+ const VkDeviceCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkDevice* pDevice);
+
+VKAPI_ATTR void VKAPI_CALL vkDestroyDevice(
+ VkDevice device,
+ const VkAllocationCallbacks* pAllocator);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkEnumerateInstanceExtensionProperties(
+ const char* pLayerName,
+ uint32_t* pPropertyCount,
+ VkExtensionProperties* pProperties);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkEnumerateDeviceExtensionProperties(
+ VkPhysicalDevice physicalDevice,
+ const char* pLayerName,
+ uint32_t* pPropertyCount,
+ VkExtensionProperties* pProperties);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkEnumerateInstanceLayerProperties(
+ uint32_t* pPropertyCount,
+ VkLayerProperties* pProperties);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkEnumerateDeviceLayerProperties(
+ VkPhysicalDevice physicalDevice,
+ uint32_t* pPropertyCount,
+ VkLayerProperties* pProperties);
+
+VKAPI_ATTR void VKAPI_CALL vkGetDeviceQueue(
+ VkDevice device,
+ uint32_t queueFamilyIndex,
+ uint32_t queueIndex,
+ VkQueue* pQueue);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkQueueSubmit(
+ VkQueue queue,
+ uint32_t submitCount,
+ const VkSubmitInfo* pSubmits,
+ VkFence fence);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkQueueWaitIdle(
+ VkQueue queue);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkDeviceWaitIdle(
+ VkDevice device);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkAllocateMemory(
+ VkDevice device,
+ const VkMemoryAllocateInfo* pAllocateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkDeviceMemory* pMemory);
+
+VKAPI_ATTR void VKAPI_CALL vkFreeMemory(
+ VkDevice device,
+ VkDeviceMemory memory,
+ const VkAllocationCallbacks* pAllocator);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkMapMemory(
+ VkDevice device,
+ VkDeviceMemory memory,
+ VkDeviceSize offset,
+ VkDeviceSize size,
+ VkMemoryMapFlags flags,
+ void** ppData);
+
+VKAPI_ATTR void VKAPI_CALL vkUnmapMemory(
+ VkDevice device,
+ VkDeviceMemory memory);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkFlushMappedMemoryRanges(
+ VkDevice device,
+ uint32_t memoryRangeCount,
+ const VkMappedMemoryRange* pMemoryRanges);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkInvalidateMappedMemoryRanges(
+ VkDevice device,
+ uint32_t memoryRangeCount,
+ const VkMappedMemoryRange* pMemoryRanges);
+
+VKAPI_ATTR void VKAPI_CALL vkGetDeviceMemoryCommitment(
+ VkDevice device,
+ VkDeviceMemory memory,
+ VkDeviceSize* pCommittedMemoryInBytes);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkBindBufferMemory(
+ VkDevice device,
+ VkBuffer buffer,
+ VkDeviceMemory memory,
+ VkDeviceSize memoryOffset);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkBindImageMemory(
+ VkDevice device,
+ VkImage image,
+ VkDeviceMemory memory,
+ VkDeviceSize memoryOffset);
+
+VKAPI_ATTR void VKAPI_CALL vkGetBufferMemoryRequirements(
+ VkDevice device,
+ VkBuffer buffer,
+ VkMemoryRequirements* pMemoryRequirements);
+
+VKAPI_ATTR void VKAPI_CALL vkGetImageMemoryRequirements(
+ VkDevice device,
+ VkImage image,
+ VkMemoryRequirements* pMemoryRequirements);
+
+VKAPI_ATTR void VKAPI_CALL vkGetImageSparseMemoryRequirements(
+ VkDevice device,
+ VkImage image,
+ uint32_t* pSparseMemoryRequirementCount,
+ VkSparseImageMemoryRequirements* pSparseMemoryRequirements);
+
+VKAPI_ATTR void VKAPI_CALL vkGetPhysicalDeviceSparseImageFormatProperties(
+ VkPhysicalDevice physicalDevice,
+ VkFormat format,
+ VkImageType type,
+ VkSampleCountFlagBits samples,
+ VkImageUsageFlags usage,
+ VkImageTiling tiling,
+ uint32_t* pPropertyCount,
+ VkSparseImageFormatProperties* pProperties);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkQueueBindSparse(
+ VkQueue queue,
+ uint32_t bindInfoCount,
+ const VkBindSparseInfo* pBindInfo,
+ VkFence fence);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateFence(
+ VkDevice device,
+ const VkFenceCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkFence* pFence);
+
+VKAPI_ATTR void VKAPI_CALL vkDestroyFence(
+ VkDevice device,
+ VkFence fence,
+ const VkAllocationCallbacks* pAllocator);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkResetFences(
+ VkDevice device,
+ uint32_t fenceCount,
+ const VkFence* pFences);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkGetFenceStatus(
+ VkDevice device,
+ VkFence fence);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkWaitForFences(
+ VkDevice device,
+ uint32_t fenceCount,
+ const VkFence* pFences,
+ VkBool32 waitAll,
+ uint64_t timeout);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateSemaphore(
+ VkDevice device,
+ const VkSemaphoreCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkSemaphore* pSemaphore);
+
+VKAPI_ATTR void VKAPI_CALL vkDestroySemaphore(
+ VkDevice device,
+ VkSemaphore semaphore,
+ const VkAllocationCallbacks* pAllocator);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateEvent(
+ VkDevice device,
+ const VkEventCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkEvent* pEvent);
+
+VKAPI_ATTR void VKAPI_CALL vkDestroyEvent(
+ VkDevice device,
+ VkEvent event,
+ const VkAllocationCallbacks* pAllocator);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkGetEventStatus(
+ VkDevice device,
+ VkEvent event);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkSetEvent(
+ VkDevice device,
+ VkEvent event);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkResetEvent(
+ VkDevice device,
+ VkEvent event);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateQueryPool(
+ VkDevice device,
+ const VkQueryPoolCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkQueryPool* pQueryPool);
+
+VKAPI_ATTR void VKAPI_CALL vkDestroyQueryPool(
+ VkDevice device,
+ VkQueryPool queryPool,
+ const VkAllocationCallbacks* pAllocator);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkGetQueryPoolResults(
+ VkDevice device,
+ VkQueryPool queryPool,
+ uint32_t firstQuery,
+ uint32_t queryCount,
+ size_t dataSize,
+ void* pData,
+ VkDeviceSize stride,
+ VkQueryResultFlags flags);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateBuffer(
+ VkDevice device,
+ const VkBufferCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkBuffer* pBuffer);
+
+VKAPI_ATTR void VKAPI_CALL vkDestroyBuffer(
+ VkDevice device,
+ VkBuffer buffer,
+ const VkAllocationCallbacks* pAllocator);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateBufferView(
+ VkDevice device,
+ const VkBufferViewCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkBufferView* pView);
+
+VKAPI_ATTR void VKAPI_CALL vkDestroyBufferView(
+ VkDevice device,
+ VkBufferView bufferView,
+ const VkAllocationCallbacks* pAllocator);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateImage(
+ VkDevice device,
+ const VkImageCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkImage* pImage);
+
+VKAPI_ATTR void VKAPI_CALL vkDestroyImage(
+ VkDevice device,
+ VkImage image,
+ const VkAllocationCallbacks* pAllocator);
+
+VKAPI_ATTR void VKAPI_CALL vkGetImageSubresourceLayout(
+ VkDevice device,
+ VkImage image,
+ const VkImageSubresource* pSubresource,
+ VkSubresourceLayout* pLayout);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateImageView(
+ VkDevice device,
+ const VkImageViewCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkImageView* pView);
+
+VKAPI_ATTR void VKAPI_CALL vkDestroyImageView(
+ VkDevice device,
+ VkImageView imageView,
+ const VkAllocationCallbacks* pAllocator);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateShaderModule(
+ VkDevice device,
+ const VkShaderModuleCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkShaderModule* pShaderModule);
+
+VKAPI_ATTR void VKAPI_CALL vkDestroyShaderModule(
+ VkDevice device,
+ VkShaderModule shaderModule,
+ const VkAllocationCallbacks* pAllocator);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkCreatePipelineCache(
+ VkDevice device,
+ const VkPipelineCacheCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkPipelineCache* pPipelineCache);
+
+VKAPI_ATTR void VKAPI_CALL vkDestroyPipelineCache(
+ VkDevice device,
+ VkPipelineCache pipelineCache,
+ const VkAllocationCallbacks* pAllocator);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkGetPipelineCacheData(
+ VkDevice device,
+ VkPipelineCache pipelineCache,
+ size_t* pDataSize,
+ void* pData);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkMergePipelineCaches(
+ VkDevice device,
+ VkPipelineCache dstCache,
+ uint32_t srcCacheCount,
+ const VkPipelineCache* pSrcCaches);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateGraphicsPipelines(
+ VkDevice device,
+ VkPipelineCache pipelineCache,
+ uint32_t createInfoCount,
+ const VkGraphicsPipelineCreateInfo* pCreateInfos,
+ const VkAllocationCallbacks* pAllocator,
+ VkPipeline* pPipelines);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateComputePipelines(
+ VkDevice device,
+ VkPipelineCache pipelineCache,
+ uint32_t createInfoCount,
+ const VkComputePipelineCreateInfo* pCreateInfos,
+ const VkAllocationCallbacks* pAllocator,
+ VkPipeline* pPipelines);
+
+VKAPI_ATTR void VKAPI_CALL vkDestroyPipeline(
+ VkDevice device,
+ VkPipeline pipeline,
+ const VkAllocationCallbacks* pAllocator);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkCreatePipelineLayout(
+ VkDevice device,
+ const VkPipelineLayoutCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkPipelineLayout* pPipelineLayout);
+
+VKAPI_ATTR void VKAPI_CALL vkDestroyPipelineLayout(
+ VkDevice device,
+ VkPipelineLayout pipelineLayout,
+ const VkAllocationCallbacks* pAllocator);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateSampler(
+ VkDevice device,
+ const VkSamplerCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkSampler* pSampler);
+
+VKAPI_ATTR void VKAPI_CALL vkDestroySampler(
+ VkDevice device,
+ VkSampler sampler,
+ const VkAllocationCallbacks* pAllocator);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateDescriptorSetLayout(
+ VkDevice device,
+ const VkDescriptorSetLayoutCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkDescriptorSetLayout* pSetLayout);
+
+VKAPI_ATTR void VKAPI_CALL vkDestroyDescriptorSetLayout(
+ VkDevice device,
+ VkDescriptorSetLayout descriptorSetLayout,
+ const VkAllocationCallbacks* pAllocator);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateDescriptorPool(
+ VkDevice device,
+ const VkDescriptorPoolCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkDescriptorPool* pDescriptorPool);
+
+VKAPI_ATTR void VKAPI_CALL vkDestroyDescriptorPool(
+ VkDevice device,
+ VkDescriptorPool descriptorPool,
+ const VkAllocationCallbacks* pAllocator);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkResetDescriptorPool(
+ VkDevice device,
+ VkDescriptorPool descriptorPool,
+ VkDescriptorPoolResetFlags flags);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkAllocateDescriptorSets(
+ VkDevice device,
+ const VkDescriptorSetAllocateInfo* pAllocateInfo,
+ VkDescriptorSet* pDescriptorSets);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkFreeDescriptorSets(
+ VkDevice device,
+ VkDescriptorPool descriptorPool,
+ uint32_t descriptorSetCount,
+ const VkDescriptorSet* pDescriptorSets);
+
+VKAPI_ATTR void VKAPI_CALL vkUpdateDescriptorSets(
+ VkDevice device,
+ uint32_t descriptorWriteCount,
+ const VkWriteDescriptorSet* pDescriptorWrites,
+ uint32_t descriptorCopyCount,
+ const VkCopyDescriptorSet* pDescriptorCopies);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateFramebuffer(
+ VkDevice device,
+ const VkFramebufferCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkFramebuffer* pFramebuffer);
+
+VKAPI_ATTR void VKAPI_CALL vkDestroyFramebuffer(
+ VkDevice device,
+ VkFramebuffer framebuffer,
+ const VkAllocationCallbacks* pAllocator);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateRenderPass(
+ VkDevice device,
+ const VkRenderPassCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkRenderPass* pRenderPass);
+
+VKAPI_ATTR void VKAPI_CALL vkDestroyRenderPass(
+ VkDevice device,
+ VkRenderPass renderPass,
+ const VkAllocationCallbacks* pAllocator);
+
+VKAPI_ATTR void VKAPI_CALL vkGetRenderAreaGranularity(
+ VkDevice device,
+ VkRenderPass renderPass,
+ VkExtent2D* pGranularity);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateCommandPool(
+ VkDevice device,
+ const VkCommandPoolCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkCommandPool* pCommandPool);
+
+VKAPI_ATTR void VKAPI_CALL vkDestroyCommandPool(
+ VkDevice device,
+ VkCommandPool commandPool,
+ const VkAllocationCallbacks* pAllocator);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkResetCommandPool(
+ VkDevice device,
+ VkCommandPool commandPool,
+ VkCommandPoolResetFlags flags);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkAllocateCommandBuffers(
+ VkDevice device,
+ const VkCommandBufferAllocateInfo* pAllocateInfo,
+ VkCommandBuffer* pCommandBuffers);
+
+VKAPI_ATTR void VKAPI_CALL vkFreeCommandBuffers(
+ VkDevice device,
+ VkCommandPool commandPool,
+ uint32_t commandBufferCount,
+ const VkCommandBuffer* pCommandBuffers);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkBeginCommandBuffer(
+ VkCommandBuffer commandBuffer,
+ const VkCommandBufferBeginInfo* pBeginInfo);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkEndCommandBuffer(
+ VkCommandBuffer commandBuffer);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkResetCommandBuffer(
+ VkCommandBuffer commandBuffer,
+ VkCommandBufferResetFlags flags);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdBindPipeline(
+ VkCommandBuffer commandBuffer,
+ VkPipelineBindPoint pipelineBindPoint,
+ VkPipeline pipeline);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdSetViewport(
+ VkCommandBuffer commandBuffer,
+ uint32_t firstViewport,
+ uint32_t viewportCount,
+ const VkViewport* pViewports);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdSetScissor(
+ VkCommandBuffer commandBuffer,
+ uint32_t firstScissor,
+ uint32_t scissorCount,
+ const VkRect2D* pScissors);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdSetLineWidth(
+ VkCommandBuffer commandBuffer,
+ float lineWidth);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdSetDepthBias(
+ VkCommandBuffer commandBuffer,
+ float depthBiasConstantFactor,
+ float depthBiasClamp,
+ float depthBiasSlopeFactor);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdSetBlendConstants(
+ VkCommandBuffer commandBuffer,
+ const float blendConstants[4]);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdSetDepthBounds(
+ VkCommandBuffer commandBuffer,
+ float minDepthBounds,
+ float maxDepthBounds);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdSetStencilCompareMask(
+ VkCommandBuffer commandBuffer,
+ VkStencilFaceFlags faceMask,
+ uint32_t compareMask);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdSetStencilWriteMask(
+ VkCommandBuffer commandBuffer,
+ VkStencilFaceFlags faceMask,
+ uint32_t writeMask);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdSetStencilReference(
+ VkCommandBuffer commandBuffer,
+ VkStencilFaceFlags faceMask,
+ uint32_t reference);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdBindDescriptorSets(
+ VkCommandBuffer commandBuffer,
+ VkPipelineBindPoint pipelineBindPoint,
+ VkPipelineLayout layout,
+ uint32_t firstSet,
+ uint32_t descriptorSetCount,
+ const VkDescriptorSet* pDescriptorSets,
+ uint32_t dynamicOffsetCount,
+ const uint32_t* pDynamicOffsets);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdBindIndexBuffer(
+ VkCommandBuffer commandBuffer,
+ VkBuffer buffer,
+ VkDeviceSize offset,
+ VkIndexType indexType);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdBindVertexBuffers(
+ VkCommandBuffer commandBuffer,
+ uint32_t firstBinding,
+ uint32_t bindingCount,
+ const VkBuffer* pBuffers,
+ const VkDeviceSize* pOffsets);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdDraw(
+ VkCommandBuffer commandBuffer,
+ uint32_t vertexCount,
+ uint32_t instanceCount,
+ uint32_t firstVertex,
+ uint32_t firstInstance);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdDrawIndexed(
+ VkCommandBuffer commandBuffer,
+ uint32_t indexCount,
+ uint32_t instanceCount,
+ uint32_t firstIndex,
+ int32_t vertexOffset,
+ uint32_t firstInstance);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdDrawIndirect(
+ VkCommandBuffer commandBuffer,
+ VkBuffer buffer,
+ VkDeviceSize offset,
+ uint32_t drawCount,
+ uint32_t stride);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdDrawIndexedIndirect(
+ VkCommandBuffer commandBuffer,
+ VkBuffer buffer,
+ VkDeviceSize offset,
+ uint32_t drawCount,
+ uint32_t stride);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdDispatch(
+ VkCommandBuffer commandBuffer,
+ uint32_t x,
+ uint32_t y,
+ uint32_t z);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdDispatchIndirect(
+ VkCommandBuffer commandBuffer,
+ VkBuffer buffer,
+ VkDeviceSize offset);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdCopyBuffer(
+ VkCommandBuffer commandBuffer,
+ VkBuffer srcBuffer,
+ VkBuffer dstBuffer,
+ uint32_t regionCount,
+ const VkBufferCopy* pRegions);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdCopyImage(
+ VkCommandBuffer commandBuffer,
+ VkImage srcImage,
+ VkImageLayout srcImageLayout,
+ VkImage dstImage,
+ VkImageLayout dstImageLayout,
+ uint32_t regionCount,
+ const VkImageCopy* pRegions);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdBlitImage(
+ VkCommandBuffer commandBuffer,
+ VkImage srcImage,
+ VkImageLayout srcImageLayout,
+ VkImage dstImage,
+ VkImageLayout dstImageLayout,
+ uint32_t regionCount,
+ const VkImageBlit* pRegions,
+ VkFilter filter);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdCopyBufferToImage(
+ VkCommandBuffer commandBuffer,
+ VkBuffer srcBuffer,
+ VkImage dstImage,
+ VkImageLayout dstImageLayout,
+ uint32_t regionCount,
+ const VkBufferImageCopy* pRegions);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdCopyImageToBuffer(
+ VkCommandBuffer commandBuffer,
+ VkImage srcImage,
+ VkImageLayout srcImageLayout,
+ VkBuffer dstBuffer,
+ uint32_t regionCount,
+ const VkBufferImageCopy* pRegions);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdUpdateBuffer(
+ VkCommandBuffer commandBuffer,
+ VkBuffer dstBuffer,
+ VkDeviceSize dstOffset,
+ VkDeviceSize dataSize,
+ const uint32_t* pData);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdFillBuffer(
+ VkCommandBuffer commandBuffer,
+ VkBuffer dstBuffer,
+ VkDeviceSize dstOffset,
+ VkDeviceSize size,
+ uint32_t data);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdClearColorImage(
+ VkCommandBuffer commandBuffer,
+ VkImage image,
+ VkImageLayout imageLayout,
+ const VkClearColorValue* pColor,
+ uint32_t rangeCount,
+ const VkImageSubresourceRange* pRanges);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdClearDepthStencilImage(
+ VkCommandBuffer commandBuffer,
+ VkImage image,
+ VkImageLayout imageLayout,
+ const VkClearDepthStencilValue* pDepthStencil,
+ uint32_t rangeCount,
+ const VkImageSubresourceRange* pRanges);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdClearAttachments(
+ VkCommandBuffer commandBuffer,
+ uint32_t attachmentCount,
+ const VkClearAttachment* pAttachments,
+ uint32_t rectCount,
+ const VkClearRect* pRects);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdResolveImage(
+ VkCommandBuffer commandBuffer,
+ VkImage srcImage,
+ VkImageLayout srcImageLayout,
+ VkImage dstImage,
+ VkImageLayout dstImageLayout,
+ uint32_t regionCount,
+ const VkImageResolve* pRegions);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdSetEvent(
+ VkCommandBuffer commandBuffer,
+ VkEvent event,
+ VkPipelineStageFlags stageMask);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdResetEvent(
+ VkCommandBuffer commandBuffer,
+ VkEvent event,
+ VkPipelineStageFlags stageMask);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdWaitEvents(
+ VkCommandBuffer commandBuffer,
+ uint32_t eventCount,
+ const VkEvent* pEvents,
+ VkPipelineStageFlags srcStageMask,
+ VkPipelineStageFlags dstStageMask,
+ uint32_t memoryBarrierCount,
+ const VkMemoryBarrier* pMemoryBarriers,
+ uint32_t bufferMemoryBarrierCount,
+ const VkBufferMemoryBarrier* pBufferMemoryBarriers,
+ uint32_t imageMemoryBarrierCount,
+ const VkImageMemoryBarrier* pImageMemoryBarriers);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdPipelineBarrier(
+ VkCommandBuffer commandBuffer,
+ VkPipelineStageFlags srcStageMask,
+ VkPipelineStageFlags dstStageMask,
+ VkDependencyFlags dependencyFlags,
+ uint32_t memoryBarrierCount,
+ const VkMemoryBarrier* pMemoryBarriers,
+ uint32_t bufferMemoryBarrierCount,
+ const VkBufferMemoryBarrier* pBufferMemoryBarriers,
+ uint32_t imageMemoryBarrierCount,
+ const VkImageMemoryBarrier* pImageMemoryBarriers);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdBeginQuery(
+ VkCommandBuffer commandBuffer,
+ VkQueryPool queryPool,
+ uint32_t query,
+ VkQueryControlFlags flags);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdEndQuery(
+ VkCommandBuffer commandBuffer,
+ VkQueryPool queryPool,
+ uint32_t query);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdResetQueryPool(
+ VkCommandBuffer commandBuffer,
+ VkQueryPool queryPool,
+ uint32_t firstQuery,
+ uint32_t queryCount);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdWriteTimestamp(
+ VkCommandBuffer commandBuffer,
+ VkPipelineStageFlagBits pipelineStage,
+ VkQueryPool queryPool,
+ uint32_t query);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdCopyQueryPoolResults(
+ VkCommandBuffer commandBuffer,
+ VkQueryPool queryPool,
+ uint32_t firstQuery,
+ uint32_t queryCount,
+ VkBuffer dstBuffer,
+ VkDeviceSize dstOffset,
+ VkDeviceSize stride,
+ VkQueryResultFlags flags);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdPushConstants(
+ VkCommandBuffer commandBuffer,
+ VkPipelineLayout layout,
+ VkShaderStageFlags stageFlags,
+ uint32_t offset,
+ uint32_t size,
+ const void* pValues);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdBeginRenderPass(
+ VkCommandBuffer commandBuffer,
+ const VkRenderPassBeginInfo* pRenderPassBegin,
+ VkSubpassContents contents);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdNextSubpass(
+ VkCommandBuffer commandBuffer,
+ VkSubpassContents contents);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdEndRenderPass(
+ VkCommandBuffer commandBuffer);
+
+VKAPI_ATTR void VKAPI_CALL vkCmdExecuteCommands(
+ VkCommandBuffer commandBuffer,
+ uint32_t commandBufferCount,
+ const VkCommandBuffer* pCommandBuffers);
+#endif
+
+#define VK_KHR_surface 1
+VK_DEFINE_NON_DISPATCHABLE_HANDLE(VkSurfaceKHR)
+
+#define VK_KHR_SURFACE_SPEC_VERSION 24
+#define VK_KHR_SURFACE_EXTENSION_NAME "VK_KHR_surface"
+
+
+typedef enum VkColorSpaceKHR {
+ VK_COLORSPACE_SRGB_NONLINEAR_KHR = 0,
+ VK_COLORSPACE_BEGIN_RANGE = VK_COLORSPACE_SRGB_NONLINEAR_KHR,
+ VK_COLORSPACE_END_RANGE = VK_COLORSPACE_SRGB_NONLINEAR_KHR,
+ VK_COLORSPACE_RANGE_SIZE = (VK_COLORSPACE_SRGB_NONLINEAR_KHR - VK_COLORSPACE_SRGB_NONLINEAR_KHR + 1),
+ VK_COLORSPACE_MAX_ENUM = 0x7FFFFFFF
+} VkColorSpaceKHR;
+
+typedef enum VkPresentModeKHR {
+ VK_PRESENT_MODE_IMMEDIATE_KHR = 0,
+ VK_PRESENT_MODE_MAILBOX_KHR = 1,
+ VK_PRESENT_MODE_FIFO_KHR = 2,
+ VK_PRESENT_MODE_FIFO_RELAXED_KHR = 3,
+ VK_PRESENT_MODE_BEGIN_RANGE = VK_PRESENT_MODE_IMMEDIATE_KHR,
+ VK_PRESENT_MODE_END_RANGE = VK_PRESENT_MODE_FIFO_RELAXED_KHR,
+ VK_PRESENT_MODE_RANGE_SIZE = (VK_PRESENT_MODE_FIFO_RELAXED_KHR - VK_PRESENT_MODE_IMMEDIATE_KHR + 1),
+ VK_PRESENT_MODE_MAX_ENUM = 0x7FFFFFFF
+} VkPresentModeKHR;
+
+
+typedef enum VkSurfaceTransformFlagBitsKHR {
+ VK_SURFACE_TRANSFORM_IDENTITY_BIT_KHR = 0x00000001,
+ VK_SURFACE_TRANSFORM_ROTATE_90_BIT_KHR = 0x00000002,
+ VK_SURFACE_TRANSFORM_ROTATE_180_BIT_KHR = 0x00000004,
+ VK_SURFACE_TRANSFORM_ROTATE_270_BIT_KHR = 0x00000008,
+ VK_SURFACE_TRANSFORM_HORIZONTAL_MIRROR_BIT_KHR = 0x00000010,
+ VK_SURFACE_TRANSFORM_HORIZONTAL_MIRROR_ROTATE_90_BIT_KHR = 0x00000020,
+ VK_SURFACE_TRANSFORM_HORIZONTAL_MIRROR_ROTATE_180_BIT_KHR = 0x00000040,
+ VK_SURFACE_TRANSFORM_HORIZONTAL_MIRROR_ROTATE_270_BIT_KHR = 0x00000080,
+ VK_SURFACE_TRANSFORM_INHERIT_BIT_KHR = 0x00000100,
+} VkSurfaceTransformFlagBitsKHR;
+typedef VkFlags VkSurfaceTransformFlagsKHR;
+
+typedef enum VkCompositeAlphaFlagBitsKHR {
+ VK_COMPOSITE_ALPHA_OPAQUE_BIT_KHR = 0x00000001,
+ VK_COMPOSITE_ALPHA_PRE_MULTIPLIED_BIT_KHR = 0x00000002,
+ VK_COMPOSITE_ALPHA_POST_MULTIPLIED_BIT_KHR = 0x00000004,
+ VK_COMPOSITE_ALPHA_INHERIT_BIT_KHR = 0x00000008,
+} VkCompositeAlphaFlagBitsKHR;
+typedef VkFlags VkCompositeAlphaFlagsKHR;
+
+typedef struct VkSurfaceCapabilitiesKHR {
+ uint32_t minImageCount;
+ uint32_t maxImageCount;
+ VkExtent2D currentExtent;
+ VkExtent2D minImageExtent;
+ VkExtent2D maxImageExtent;
+ uint32_t maxImageArrayLayers;
+ VkSurfaceTransformFlagsKHR supportedTransforms;
+ VkSurfaceTransformFlagBitsKHR currentTransform;
+ VkCompositeAlphaFlagsKHR supportedCompositeAlpha;
+ VkImageUsageFlags supportedUsageFlags;
+} VkSurfaceCapabilitiesKHR;
+
+typedef struct VkSurfaceFormatKHR {
+ VkFormat format;
+ VkColorSpaceKHR colorSpace;
+} VkSurfaceFormatKHR;
+
+
+typedef void (VKAPI_PTR *PFN_vkDestroySurfaceKHR)(VkInstance instance, VkSurfaceKHR surface, const VkAllocationCallbacks* pAllocator);
+typedef VkResult (VKAPI_PTR *PFN_vkGetPhysicalDeviceSurfaceSupportKHR)(VkPhysicalDevice physicalDevice, uint32_t queueFamilyIndex, VkSurfaceKHR surface, VkBool32* pSupported);
+typedef VkResult (VKAPI_PTR *PFN_vkGetPhysicalDeviceSurfaceCapabilitiesKHR)(VkPhysicalDevice physicalDevice, VkSurfaceKHR surface, VkSurfaceCapabilitiesKHR* pSurfaceCapabilities);
+typedef VkResult (VKAPI_PTR *PFN_vkGetPhysicalDeviceSurfaceFormatsKHR)(VkPhysicalDevice physicalDevice, VkSurfaceKHR surface, uint32_t* pSurfaceFormatCount, VkSurfaceFormatKHR* pSurfaceFormats);
+typedef VkResult (VKAPI_PTR *PFN_vkGetPhysicalDeviceSurfacePresentModesKHR)(VkPhysicalDevice physicalDevice, VkSurfaceKHR surface, uint32_t* pPresentModeCount, VkPresentModeKHR* pPresentModes);
+
+#ifndef VK_NO_PROTOTYPES
+VKAPI_ATTR void VKAPI_CALL vkDestroySurfaceKHR(
+ VkInstance instance,
+ VkSurfaceKHR surface,
+ const VkAllocationCallbacks* pAllocator);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkGetPhysicalDeviceSurfaceSupportKHR(
+ VkPhysicalDevice physicalDevice,
+ uint32_t queueFamilyIndex,
+ VkSurfaceKHR surface,
+ VkBool32* pSupported);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkGetPhysicalDeviceSurfaceCapabilitiesKHR(
+ VkPhysicalDevice physicalDevice,
+ VkSurfaceKHR surface,
+ VkSurfaceCapabilitiesKHR* pSurfaceCapabilities);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkGetPhysicalDeviceSurfaceFormatsKHR(
+ VkPhysicalDevice physicalDevice,
+ VkSurfaceKHR surface,
+ uint32_t* pSurfaceFormatCount,
+ VkSurfaceFormatKHR* pSurfaceFormats);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkGetPhysicalDeviceSurfacePresentModesKHR(
+ VkPhysicalDevice physicalDevice,
+ VkSurfaceKHR surface,
+ uint32_t* pPresentModeCount,
+ VkPresentModeKHR* pPresentModes);
+#endif
+
+#define VK_KHR_swapchain 1
+VK_DEFINE_NON_DISPATCHABLE_HANDLE(VkSwapchainKHR)
+
+#define VK_KHR_SWAPCHAIN_SPEC_VERSION 67
+#define VK_KHR_SWAPCHAIN_EXTENSION_NAME "VK_KHR_swapchain"
+
+typedef VkFlags VkSwapchainCreateFlagsKHR;
+
+typedef struct VkSwapchainCreateInfoKHR {
+ VkStructureType sType;
+ const void* pNext;
+ VkSwapchainCreateFlagsKHR flags;
+ VkSurfaceKHR surface;
+ uint32_t minImageCount;
+ VkFormat imageFormat;
+ VkColorSpaceKHR imageColorSpace;
+ VkExtent2D imageExtent;
+ uint32_t imageArrayLayers;
+ VkImageUsageFlags imageUsage;
+ VkSharingMode imageSharingMode;
+ uint32_t queueFamilyIndexCount;
+ const uint32_t* pQueueFamilyIndices;
+ VkSurfaceTransformFlagBitsKHR preTransform;
+ VkCompositeAlphaFlagBitsKHR compositeAlpha;
+ VkPresentModeKHR presentMode;
+ VkBool32 clipped;
+ VkSwapchainKHR oldSwapchain;
+} VkSwapchainCreateInfoKHR;
+
+typedef struct VkPresentInfoKHR {
+ VkStructureType sType;
+ const void* pNext;
+ uint32_t waitSemaphoreCount;
+ const VkSemaphore* pWaitSemaphores;
+ uint32_t swapchainCount;
+ const VkSwapchainKHR* pSwapchains;
+ const uint32_t* pImageIndices;
+ VkResult* pResults;
+} VkPresentInfoKHR;
+
+
+typedef VkResult (VKAPI_PTR *PFN_vkCreateSwapchainKHR)(VkDevice device, const VkSwapchainCreateInfoKHR* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkSwapchainKHR* pSwapchain);
+typedef void (VKAPI_PTR *PFN_vkDestroySwapchainKHR)(VkDevice device, VkSwapchainKHR swapchain, const VkAllocationCallbacks* pAllocator);
+typedef VkResult (VKAPI_PTR *PFN_vkGetSwapchainImagesKHR)(VkDevice device, VkSwapchainKHR swapchain, uint32_t* pSwapchainImageCount, VkImage* pSwapchainImages);
+typedef VkResult (VKAPI_PTR *PFN_vkAcquireNextImageKHR)(VkDevice device, VkSwapchainKHR swapchain, uint64_t timeout, VkSemaphore semaphore, VkFence fence, uint32_t* pImageIndex);
+typedef VkResult (VKAPI_PTR *PFN_vkQueuePresentKHR)(VkQueue queue, const VkPresentInfoKHR* pPresentInfo);
+
+#ifndef VK_NO_PROTOTYPES
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateSwapchainKHR(
+ VkDevice device,
+ const VkSwapchainCreateInfoKHR* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkSwapchainKHR* pSwapchain);
+
+VKAPI_ATTR void VKAPI_CALL vkDestroySwapchainKHR(
+ VkDevice device,
+ VkSwapchainKHR swapchain,
+ const VkAllocationCallbacks* pAllocator);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkGetSwapchainImagesKHR(
+ VkDevice device,
+ VkSwapchainKHR swapchain,
+ uint32_t* pSwapchainImageCount,
+ VkImage* pSwapchainImages);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkAcquireNextImageKHR(
+ VkDevice device,
+ VkSwapchainKHR swapchain,
+ uint64_t timeout,
+ VkSemaphore semaphore,
+ VkFence fence,
+ uint32_t* pImageIndex);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkQueuePresentKHR(
+ VkQueue queue,
+ const VkPresentInfoKHR* pPresentInfo);
+#endif
+
+#define VK_KHR_display 1
+VK_DEFINE_NON_DISPATCHABLE_HANDLE(VkDisplayKHR)
+VK_DEFINE_NON_DISPATCHABLE_HANDLE(VkDisplayModeKHR)
+
+#define VK_KHR_DISPLAY_SPEC_VERSION 21
+#define VK_KHR_DISPLAY_EXTENSION_NAME "VK_KHR_display"
+
+
+typedef enum VkDisplayPlaneAlphaFlagBitsKHR {
+ VK_DISPLAY_PLANE_ALPHA_OPAQUE_BIT_KHR = 0x00000001,
+ VK_DISPLAY_PLANE_ALPHA_GLOBAL_BIT_KHR = 0x00000002,
+ VK_DISPLAY_PLANE_ALPHA_PER_PIXEL_BIT_KHR = 0x00000004,
+ VK_DISPLAY_PLANE_ALPHA_PER_PIXEL_PREMULTIPLIED_BIT_KHR = 0x00000008,
+} VkDisplayPlaneAlphaFlagBitsKHR;
+typedef VkFlags VkDisplayModeCreateFlagsKHR;
+typedef VkFlags VkDisplayPlaneAlphaFlagsKHR;
+typedef VkFlags VkDisplaySurfaceCreateFlagsKHR;
+
+typedef struct VkDisplayPropertiesKHR {
+ VkDisplayKHR display;
+ const char* displayName;
+ VkExtent2D physicalDimensions;
+ VkExtent2D physicalResolution;
+ VkSurfaceTransformFlagsKHR supportedTransforms;
+ VkBool32 planeReorderPossible;
+ VkBool32 persistentContent;
+} VkDisplayPropertiesKHR;
+
+typedef struct VkDisplayModeParametersKHR {
+ VkExtent2D visibleRegion;
+ uint32_t refreshRate;
+} VkDisplayModeParametersKHR;
+
+typedef struct VkDisplayModePropertiesKHR {
+ VkDisplayModeKHR displayMode;
+ VkDisplayModeParametersKHR parameters;
+} VkDisplayModePropertiesKHR;
+
+typedef struct VkDisplayModeCreateInfoKHR {
+ VkStructureType sType;
+ const void* pNext;
+ VkDisplayModeCreateFlagsKHR flags;
+ VkDisplayModeParametersKHR parameters;
+} VkDisplayModeCreateInfoKHR;
+
+typedef struct VkDisplayPlaneCapabilitiesKHR {
+ VkDisplayPlaneAlphaFlagsKHR supportedAlpha;
+ VkOffset2D minSrcPosition;
+ VkOffset2D maxSrcPosition;
+ VkExtent2D minSrcExtent;
+ VkExtent2D maxSrcExtent;
+ VkOffset2D minDstPosition;
+ VkOffset2D maxDstPosition;
+ VkExtent2D minDstExtent;
+ VkExtent2D maxDstExtent;
+} VkDisplayPlaneCapabilitiesKHR;
+
+typedef struct VkDisplayPlanePropertiesKHR {
+ VkDisplayKHR currentDisplay;
+ uint32_t currentStackIndex;
+} VkDisplayPlanePropertiesKHR;
+
+typedef struct VkDisplaySurfaceCreateInfoKHR {
+ VkStructureType sType;
+ const void* pNext;
+ VkDisplaySurfaceCreateFlagsKHR flags;
+ VkDisplayModeKHR displayMode;
+ uint32_t planeIndex;
+ uint32_t planeStackIndex;
+ VkSurfaceTransformFlagBitsKHR transform;
+ float globalAlpha;
+ VkDisplayPlaneAlphaFlagBitsKHR alphaMode;
+ VkExtent2D imageExtent;
+} VkDisplaySurfaceCreateInfoKHR;
+
+
+typedef VkResult (VKAPI_PTR *PFN_vkGetPhysicalDeviceDisplayPropertiesKHR)(VkPhysicalDevice physicalDevice, uint32_t* pPropertyCount, VkDisplayPropertiesKHR* pProperties);
+typedef VkResult (VKAPI_PTR *PFN_vkGetPhysicalDeviceDisplayPlanePropertiesKHR)(VkPhysicalDevice physicalDevice, uint32_t* pPropertyCount, VkDisplayPlanePropertiesKHR* pProperties);
+typedef VkResult (VKAPI_PTR *PFN_vkGetDisplayPlaneSupportedDisplaysKHR)(VkPhysicalDevice physicalDevice, uint32_t planeIndex, uint32_t* pDisplayCount, VkDisplayKHR* pDisplays);
+typedef VkResult (VKAPI_PTR *PFN_vkGetDisplayModePropertiesKHR)(VkPhysicalDevice physicalDevice, VkDisplayKHR display, uint32_t* pPropertyCount, VkDisplayModePropertiesKHR* pProperties);
+typedef VkResult (VKAPI_PTR *PFN_vkCreateDisplayModeKHR)(VkPhysicalDevice physicalDevice, VkDisplayKHR display, const VkDisplayModeCreateInfoKHR*pCreateInfo, const VkAllocationCallbacks* pAllocator, VkDisplayModeKHR* pMode);
+typedef VkResult (VKAPI_PTR *PFN_vkGetDisplayPlaneCapabilitiesKHR)(VkPhysicalDevice physicalDevice, VkDisplayModeKHR mode, uint32_t planeIndex, VkDisplayPlaneCapabilitiesKHR* pCapabilities);
+typedef VkResult (VKAPI_PTR *PFN_vkCreateDisplayPlaneSurfaceKHR)(VkInstance instance, const VkDisplaySurfaceCreateInfoKHR* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkSurfaceKHR* pSurface);
+
+#ifndef VK_NO_PROTOTYPES
+VKAPI_ATTR VkResult VKAPI_CALL vkGetPhysicalDeviceDisplayPropertiesKHR(
+ VkPhysicalDevice physicalDevice,
+ uint32_t* pPropertyCount,
+ VkDisplayPropertiesKHR* pProperties);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkGetPhysicalDeviceDisplayPlanePropertiesKHR(
+ VkPhysicalDevice physicalDevice,
+ uint32_t* pPropertyCount,
+ VkDisplayPlanePropertiesKHR* pProperties);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkGetDisplayPlaneSupportedDisplaysKHR(
+ VkPhysicalDevice physicalDevice,
+ uint32_t planeIndex,
+ uint32_t* pDisplayCount,
+ VkDisplayKHR* pDisplays);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkGetDisplayModePropertiesKHR(
+ VkPhysicalDevice physicalDevice,
+ VkDisplayKHR display,
+ uint32_t* pPropertyCount,
+ VkDisplayModePropertiesKHR* pProperties);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateDisplayModeKHR(
+ VkPhysicalDevice physicalDevice,
+ VkDisplayKHR display,
+ const VkDisplayModeCreateInfoKHR* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkDisplayModeKHR* pMode);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkGetDisplayPlaneCapabilitiesKHR(
+ VkPhysicalDevice physicalDevice,
+ VkDisplayModeKHR mode,
+ uint32_t planeIndex,
+ VkDisplayPlaneCapabilitiesKHR* pCapabilities);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateDisplayPlaneSurfaceKHR(
+ VkInstance instance,
+ const VkDisplaySurfaceCreateInfoKHR* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkSurfaceKHR* pSurface);
+#endif
+
+#define VK_KHR_display_swapchain 1
+#define VK_KHR_DISPLAY_SWAPCHAIN_SPEC_VERSION 9
+#define VK_KHR_DISPLAY_SWAPCHAIN_EXTENSION_NAME "VK_KHR_display_swapchain"
+
+typedef struct VkDisplayPresentInfoKHR {
+ VkStructureType sType;
+ const void* pNext;
+ VkRect2D srcRect;
+ VkRect2D dstRect;
+ VkBool32 persistent;
+} VkDisplayPresentInfoKHR;
+
+
+typedef VkResult (VKAPI_PTR *PFN_vkCreateSharedSwapchainsKHR)(VkDevice device, uint32_t swapchainCount, const VkSwapchainCreateInfoKHR* pCreateInfos, const VkAllocationCallbacks* pAllocator, VkSwapchainKHR* pSwapchains);
+
+#ifndef VK_NO_PROTOTYPES
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateSharedSwapchainsKHR(
+ VkDevice device,
+ uint32_t swapchainCount,
+ const VkSwapchainCreateInfoKHR* pCreateInfos,
+ const VkAllocationCallbacks* pAllocator,
+ VkSwapchainKHR* pSwapchains);
+#endif
+
+#ifdef VK_USE_PLATFORM_XLIB_KHR
+#define VK_KHR_xlib_surface 1
+#include <X11/Xlib.h>
+
+#define VK_KHR_XLIB_SURFACE_SPEC_VERSION 6
+#define VK_KHR_XLIB_SURFACE_EXTENSION_NAME "VK_KHR_xlib_surface"
+
+typedef VkFlags VkXlibSurfaceCreateFlagsKHR;
+
+typedef struct VkXlibSurfaceCreateInfoKHR {
+ VkStructureType sType;
+ const void* pNext;
+ VkXlibSurfaceCreateFlagsKHR flags;
+ Display* dpy;
+ Window window;
+} VkXlibSurfaceCreateInfoKHR;
+
+
+typedef VkResult (VKAPI_PTR *PFN_vkCreateXlibSurfaceKHR)(VkInstance instance, const VkXlibSurfaceCreateInfoKHR* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkSurfaceKHR* pSurface);
+typedef VkBool32 (VKAPI_PTR *PFN_vkGetPhysicalDeviceXlibPresentationSupportKHR)(VkPhysicalDevice physicalDevice, uint32_t queueFamilyIndex, Display* dpy, VisualID visualID);
+
+#ifndef VK_NO_PROTOTYPES
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateXlibSurfaceKHR(
+ VkInstance instance,
+ const VkXlibSurfaceCreateInfoKHR* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkSurfaceKHR* pSurface);
+
+VKAPI_ATTR VkBool32 VKAPI_CALL vkGetPhysicalDeviceXlibPresentationSupportKHR(
+ VkPhysicalDevice physicalDevice,
+ uint32_t queueFamilyIndex,
+ Display* dpy,
+ VisualID visualID);
+#endif
+#endif /* VK_USE_PLATFORM_XLIB_KHR */
+
+#ifdef VK_USE_PLATFORM_XCB_KHR
+#define VK_KHR_xcb_surface 1
+#include <xcb/xcb.h>
+
+#define VK_KHR_XCB_SURFACE_SPEC_VERSION 6
+#define VK_KHR_XCB_SURFACE_EXTENSION_NAME "VK_KHR_xcb_surface"
+
+typedef VkFlags VkXcbSurfaceCreateFlagsKHR;
+
+typedef struct VkXcbSurfaceCreateInfoKHR {
+ VkStructureType sType;
+ const void* pNext;
+ VkXcbSurfaceCreateFlagsKHR flags;
+ xcb_connection_t* connection;
+ xcb_window_t window;
+} VkXcbSurfaceCreateInfoKHR;
+
+
+typedef VkResult (VKAPI_PTR *PFN_vkCreateXcbSurfaceKHR)(VkInstance instance, const VkXcbSurfaceCreateInfoKHR* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkSurfaceKHR* pSurface);
+typedef VkBool32 (VKAPI_PTR *PFN_vkGetPhysicalDeviceXcbPresentationSupportKHR)(VkPhysicalDevice physicalDevice, uint32_t queueFamilyIndex, xcb_connection_t* connection, xcb_visualid_t visual_id);
+
+#ifndef VK_NO_PROTOTYPES
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateXcbSurfaceKHR(
+ VkInstance instance,
+ const VkXcbSurfaceCreateInfoKHR* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkSurfaceKHR* pSurface);
+
+VKAPI_ATTR VkBool32 VKAPI_CALL vkGetPhysicalDeviceXcbPresentationSupportKHR(
+ VkPhysicalDevice physicalDevice,
+ uint32_t queueFamilyIndex,
+ xcb_connection_t* connection,
+ xcb_visualid_t visual_id);
+#endif
+#endif /* VK_USE_PLATFORM_XCB_KHR */
+
+#ifdef VK_USE_PLATFORM_WAYLAND_KHR
+#define VK_KHR_wayland_surface 1
+#include <wayland-client.h>
+
+#define VK_KHR_WAYLAND_SURFACE_SPEC_VERSION 5
+#define VK_KHR_WAYLAND_SURFACE_EXTENSION_NAME "VK_KHR_wayland_surface"
+
+typedef VkFlags VkWaylandSurfaceCreateFlagsKHR;
+
+typedef struct VkWaylandSurfaceCreateInfoKHR {
+ VkStructureType sType;
+ const void* pNext;
+ VkWaylandSurfaceCreateFlagsKHR flags;
+ struct wl_display* display;
+ struct wl_surface* surface;
+} VkWaylandSurfaceCreateInfoKHR;
+
+
+typedef VkResult (VKAPI_PTR *PFN_vkCreateWaylandSurfaceKHR)(VkInstance instance, const VkWaylandSurfaceCreateInfoKHR* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkSurfaceKHR* pSurface);
+typedef VkBool32 (VKAPI_PTR *PFN_vkGetPhysicalDeviceWaylandPresentationSupportKHR)(VkPhysicalDevice physicalDevice, uint32_t queueFamilyIndex, struct wl_display* display);
+
+#ifndef VK_NO_PROTOTYPES
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateWaylandSurfaceKHR(
+ VkInstance instance,
+ const VkWaylandSurfaceCreateInfoKHR* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkSurfaceKHR* pSurface);
+
+VKAPI_ATTR VkBool32 VKAPI_CALL vkGetPhysicalDeviceWaylandPresentationSupportKHR(
+ VkPhysicalDevice physicalDevice,
+ uint32_t queueFamilyIndex,
+ struct wl_display* display);
+#endif
+#endif /* VK_USE_PLATFORM_WAYLAND_KHR */
+
+#ifdef VK_USE_PLATFORM_MIR_KHR
+#define VK_KHR_mir_surface 1
+#include <mir_toolkit/client_types.h>
+
+#define VK_KHR_MIR_SURFACE_SPEC_VERSION 4
+#define VK_KHR_MIR_SURFACE_EXTENSION_NAME "VK_KHR_mir_surface"
+
+typedef VkFlags VkMirSurfaceCreateFlagsKHR;
+
+typedef struct VkMirSurfaceCreateInfoKHR {
+ VkStructureType sType;
+ const void* pNext;
+ VkMirSurfaceCreateFlagsKHR flags;
+ MirConnection* connection;
+ MirSurface* mirSurface;
+} VkMirSurfaceCreateInfoKHR;
+
+
+typedef VkResult (VKAPI_PTR *PFN_vkCreateMirSurfaceKHR)(VkInstance instance, const VkMirSurfaceCreateInfoKHR* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkSurfaceKHR* pSurface);
+typedef VkBool32 (VKAPI_PTR *PFN_vkGetPhysicalDeviceMirPresentationSupportKHR)(VkPhysicalDevice physicalDevice, uint32_t queueFamilyIndex, MirConnection* connection);
+
+#ifndef VK_NO_PROTOTYPES
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateMirSurfaceKHR(
+ VkInstance instance,
+ const VkMirSurfaceCreateInfoKHR* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkSurfaceKHR* pSurface);
+
+VKAPI_ATTR VkBool32 VKAPI_CALL vkGetPhysicalDeviceMirPresentationSupportKHR(
+ VkPhysicalDevice physicalDevice,
+ uint32_t queueFamilyIndex,
+ MirConnection* connection);
+#endif
+#endif /* VK_USE_PLATFORM_MIR_KHR */
+
+#ifdef VK_USE_PLATFORM_ANDROID_KHR
+#define VK_KHR_android_surface 1
+#include <android/native_window.h>
+
+#define VK_KHR_ANDROID_SURFACE_SPEC_VERSION 5
+#define VK_KHR_ANDROID_SURFACE_EXTENSION_NAME "VK_KHR_android_surface"
+
+typedef VkFlags VkAndroidSurfaceCreateFlagsKHR;
+
+typedef struct VkAndroidSurfaceCreateInfoKHR {
+ VkStructureType sType;
+ const void* pNext;
+ VkAndroidSurfaceCreateFlagsKHR flags;
+ ANativeWindow* window;
+} VkAndroidSurfaceCreateInfoKHR;
+
+
+typedef VkResult (VKAPI_PTR *PFN_vkCreateAndroidSurfaceKHR)(VkInstance instance, const VkAndroidSurfaceCreateInfoKHR* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkSurfaceKHR* pSurface);
+
+#ifndef VK_NO_PROTOTYPES
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateAndroidSurfaceKHR(
+ VkInstance instance,
+ const VkAndroidSurfaceCreateInfoKHR* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkSurfaceKHR* pSurface);
+#endif
+#endif /* VK_USE_PLATFORM_ANDROID_KHR */
+
+#ifdef VK_USE_PLATFORM_WIN32_KHR
+#define VK_KHR_win32_surface 1
+#include <windows.h>
+
+#define VK_KHR_WIN32_SURFACE_SPEC_VERSION 5
+#define VK_KHR_WIN32_SURFACE_EXTENSION_NAME "VK_KHR_win32_surface"
+
+typedef VkFlags VkWin32SurfaceCreateFlagsKHR;
+
+typedef struct VkWin32SurfaceCreateInfoKHR {
+ VkStructureType sType;
+ const void* pNext;
+ VkWin32SurfaceCreateFlagsKHR flags;
+ HINSTANCE hinstance;
+ HWND hwnd;
+} VkWin32SurfaceCreateInfoKHR;
+
+
+typedef VkResult (VKAPI_PTR *PFN_vkCreateWin32SurfaceKHR)(VkInstance instance, const VkWin32SurfaceCreateInfoKHR* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkSurfaceKHR* pSurface);
+typedef VkBool32 (VKAPI_PTR *PFN_vkGetPhysicalDeviceWin32PresentationSupportKHR)(VkPhysicalDevice physicalDevice, uint32_t queueFamilyIndex);
+
+#ifndef VK_NO_PROTOTYPES
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateWin32SurfaceKHR(
+ VkInstance instance,
+ const VkWin32SurfaceCreateInfoKHR* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkSurfaceKHR* pSurface);
+
+VKAPI_ATTR VkBool32 VKAPI_CALL vkGetPhysicalDeviceWin32PresentationSupportKHR(
+ VkPhysicalDevice physicalDevice,
+ uint32_t queueFamilyIndex);
+#endif
+#endif /* VK_USE_PLATFORM_WIN32_KHR */
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#ifndef __VULKAN_INTEL_H__
+#define __VULKAN_INTEL_H__
+
+#include "vulkan.h"
+
+#ifdef __cplusplus
+extern "C"
+{
+#endif // __cplusplus
+
+#define VK_STRUCTURE_TYPE_DMA_BUF_IMAGE_CREATE_INFO_INTEL 1024
+typedef struct VkDmaBufImageCreateInfo_
+{
+ VkStructureType sType; // Must be VK_STRUCTURE_TYPE_DMA_BUF_IMAGE_CREATE_INFO_INTEL
+ const void* pNext; // Pointer to next structure.
+ int fd;
+ VkFormat format;
+ VkExtent3D extent; // Depth must be 1
+ uint32_t strideInBytes;
+} VkDmaBufImageCreateInfo;
+
+typedef VkResult (VKAPI_PTR *PFN_vkCreateDmaBufImageINTEL)(VkDevice device, const VkDmaBufImageCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkDeviceMemory* pMem, VkImage* pImage);
+
+#ifdef VK_PROTOTYPES
+
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateDmaBufImageINTEL(
+ VkDevice _device,
+ const VkDmaBufImageCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkDeviceMemory* pMem,
+ VkImage* pImage);
+
+#endif
+
+#ifdef __cplusplus
+} // extern "C"
+#endif // __cplusplus
+
+#endif // __VULKAN_INTEL_H__
# C preprocessor options
cppdefines = []
- cppdefines += ['__STDC_LIMIT_MACROS']
+ cppdefines += ['__STDC_LIMIT_MACROS', '__STDC_CONSTANT_MACROS']
if env['build'] in ('debug', 'checked'):
cppdefines += ['DEBUG']
else:
AM_CFLAGS = $(VISIBILITY_CFLAGS)
AM_CXXFLAGS = $(VISIBILITY_CXXFLAGS)
+if HAVE_VULKAN
+SUBDIRS += isl
+SUBDIRS += vulkan
+endif
+
AM_CPPFLAGS = \
-I$(top_srcdir)/include/ \
-I$(top_srcdir)/src/mapi/ \
include Makefile.sources
include $(top_srcdir)/src/gallium/Automake.inc
-noinst_LTLIBRARIES = libgallium.la
+noinst_LTLIBRARIES = libgallium_nir.la
AM_CFLAGS = \
-I$(top_srcdir)/src/loader \
- -I$(top_builddir)/src/glsl/nir \
-I$(top_srcdir)/src/gallium/auxiliary/util \
$(GALLIUM_CFLAGS) \
$(VISIBILITY_CFLAGS) \
$(VISIBILITY_CXXFLAGS) \
$(MSVC2008_COMPAT_CXXFLAGS)
+libgallium_nir_la_SOURCES = \
+ $(NIR_SOURCES)
+
+libgallium_nir_la_CFLAGS = \
+ -I$(top_builddir)/src/glsl/nir \
+ $(GALLIUM_CFLAGS) \
+ $(VISIBILITY_CFLAGS) \
+ $(MSVC2013_COMPAT_CFLAGS)
+
+noinst_LTLIBRARIES += libgallium.la
+
libgallium_la_SOURCES = \
$(C_SOURCES) \
- $(NIR_SOURCES) \
$(GENERATED_SOURCES)
+libgallium_la_LIBADD = \
+ libgallium_nir.la
+
if HAVE_MESA_LLVM
AM_CFLAGS += \
util/u_helpers.h \
util/u_index_modify.c \
util/u_index_modify.h \
- util/u_init.h \
util/u_inlines.h \
util/u_keymap.c \
util/u_keymap.h \
}
for (i = 0; i < 4; i++) {
- out->clip[i] = clipvertex[i];
- out->pre_clip_pos[i] = position[i];
+ out->clip_pos[i] = position[i];
}
+ /* Be careful with NaNs. Comparisons must be true for them. */
/* Do the hardwired planes first:
*/
if (flags & DO_CLIP_XY_GUARD_BAND) {
- if (-0.50 * position[0] + position[3] < 0) mask |= (1<<0);
- if ( 0.50 * position[0] + position[3] < 0) mask |= (1<<1);
- if (-0.50 * position[1] + position[3] < 0) mask |= (1<<2);
- if ( 0.50 * position[1] + position[3] < 0) mask |= (1<<3);
+ if (!(-0.50 * position[0] + position[3] >= 0)) mask |= (1<<0);
+ if (!( 0.50 * position[0] + position[3] >= 0)) mask |= (1<<1);
+ if (!(-0.50 * position[1] + position[3] >= 0)) mask |= (1<<2);
+ if (!( 0.50 * position[1] + position[3] >= 0)) mask |= (1<<3);
}
else if (flags & DO_CLIP_XY) {
- if (-position[0] + position[3] < 0) mask |= (1<<0);
- if ( position[0] + position[3] < 0) mask |= (1<<1);
- if (-position[1] + position[3] < 0) mask |= (1<<2);
- if ( position[1] + position[3] < 0) mask |= (1<<3);
+ if (!(-position[0] + position[3] >= 0)) mask |= (1<<0);
+ if (!( position[0] + position[3] >= 0)) mask |= (1<<1);
+ if (!(-position[1] + position[3] >= 0)) mask |= (1<<2);
+ if (!( position[1] + position[3] >= 0)) mask |= (1<<3);
}
/* Clip Z planes according to full cube, half cube or none.
*/
if (flags & DO_CLIP_FULL_Z) {
- if ( position[2] + position[3] < 0) mask |= (1<<4);
- if (-position[2] + position[3] < 0) mask |= (1<<5);
+ if (!( position[2] + position[3] >= 0)) mask |= (1<<4);
+ if (!(-position[2] + position[3] >= 0)) mask |= (1<<5);
}
else if (flags & DO_CLIP_HALF_Z) {
- if ( position[2] < 0) mask |= (1<<4);
- if (-position[2] + position[3] < 0) mask |= (1<<5);
+ if (!( position[2] >= 0)) mask |= (1<<4);
+ if (!(-position[2] + position[3] >= 0)) mask |= (1<<5);
}
if (flags & DO_CLIP_USER) {
if (have_cd && num_written_clipdistance) {
float clipdist;
i = plane_idx - 6;
- out->have_clipdist = 1;
/* first four clip distance in first vector etc. */
if (i < 4)
clipdist = out->data[cd[0]][i];
if (clipdist < 0 || util_is_inf_or_nan(clipdist))
mask |= 1 << plane_idx;
} else {
- if (dot4(clipvertex, plane[plane_idx]) < 0)
+ if (!(dot4(clipvertex, plane[plane_idx]) >= 0))
mask |= 1 << plane_idx;
}
}
out = (struct vertex_header *)( (char *)out + info->stride );
}
-
return need_pipeline != 0;
}
sampler_type = LLVMStructTypeInContext(gallivm->context, elem_types,
Elements(elem_types), 0);
+ (void) target; /* silence unused var warning for non-debug build */
LP_CHECK_MEMBER_OFFSET(struct draw_jit_sampler, min_lod,
target, sampler_type,
DRAW_JIT_SAMPLER_MIN_LOD);
PIPE_MAX_SAMPLERS); /* samplers */
context_type = LLVMStructTypeInContext(gallivm->context, elem_types,
Elements(elem_types), 0);
+
+ (void) target; /* silence unused var warning for non-debug build */
LP_CHECK_MEMBER_OFFSET(struct draw_jit_context, vs_constants,
target, context_type, DRAW_JIT_CTX_CONSTANTS);
LP_CHECK_MEMBER_OFFSET(struct draw_jit_context, num_vs_constants,
create_jit_vertex_header(struct gallivm_state *gallivm, int data_elems)
{
LLVMTargetDataRef target = gallivm->target;
- LLVMTypeRef elem_types[4];
+ LLVMTypeRef elem_types[3];
LLVMTypeRef vertex_header;
char struct_name[24];
util_snprintf(struct_name, 23, "vertex_header%d", data_elems);
elem_types[DRAW_JIT_VERTEX_VERTEX_ID] = LLVMIntTypeInContext(gallivm->context, 32);
- elem_types[DRAW_JIT_VERTEX_CLIP] = LLVMArrayType(LLVMFloatTypeInContext(gallivm->context), 4);
- elem_types[DRAW_JIT_VERTEX_PRE_CLIP_POS] = LLVMArrayType(LLVMFloatTypeInContext(gallivm->context), 4);
+ elem_types[DRAW_JIT_VERTEX_CLIP_POS] = LLVMArrayType(LLVMFloatTypeInContext(gallivm->context), 4);
elem_types[DRAW_JIT_VERTEX_DATA] = LLVMArrayType(elem_types[1], data_elems);
vertex_header = LLVMStructTypeInContext(gallivm->context, elem_types,
target, vertex_header,
DRAW_JIT_VERTEX_VERTEX_ID);
*/
- LP_CHECK_MEMBER_OFFSET(struct vertex_header, clip,
- target, vertex_header,
- DRAW_JIT_VERTEX_CLIP);
- LP_CHECK_MEMBER_OFFSET(struct vertex_header, pre_clip_pos,
+ (void) target; /* silence unused var warning for non-debug build */
+ LP_CHECK_MEMBER_OFFSET(struct vertex_header, clip_pos,
target, vertex_header,
- DRAW_JIT_VERTEX_PRE_CLIP_POS);
+ DRAW_JIT_VERTEX_CLIP_POS);
LP_CHECK_MEMBER_OFFSET(struct vertex_header, data,
target, vertex_header,
DRAW_JIT_VERTEX_DATA);
* struct vertex_header {
* unsigned clipmask:DRAW_TOTAL_CLIP_PLANES;
* unsigned edgeflag:1;
- * unsigned have_clipdist:1;
+ * unsigned pad:1;
* unsigned vertex_id:16;
* [...]
* }
* {
* return (x >> 16) | // vertex_id
* ((x & 0x3fff) << 18) | // clipmask
- * ((x & 0x4000) << 3) | // have_clipdist
+ * ((x & 0x4000) << 3) | // pad
* ((x & 0x8000) << 1); // edgeflag
* }
*/
LLVMBuilderRef builder = gallivm->builder;
LLVMValueRef vertex_id;
LLVMValueRef clipmask;
- LLVMValueRef have_clipdist;
+ LLVMValueRef pad;
LLVMValueRef edgeflag;
vertex_id = LLVMBuildLShr(builder, mask, lp_build_const_int32(gallivm, 16), "");
clipmask = LLVMBuildAnd(builder, mask, lp_build_const_int32(gallivm, 0x3fff), "");
clipmask = LLVMBuildShl(builder, clipmask, lp_build_const_int32(gallivm, 18), "");
- have_clipdist = LLVMBuildAnd(builder, mask, lp_build_const_int32(gallivm, 0x4000), "");
- have_clipdist = LLVMBuildShl(builder, have_clipdist, lp_build_const_int32(gallivm, 3), "");
+ if (0) {
+ pad = LLVMBuildAnd(builder, mask, lp_build_const_int32(gallivm, 0x4000), "");
+ pad = LLVMBuildShl(builder, pad, lp_build_const_int32(gallivm, 3), "");
+ }
edgeflag = LLVMBuildAnd(builder, mask, lp_build_const_int32(gallivm, 0x8000), "");
edgeflag = LLVMBuildShl(builder, edgeflag, lp_build_const_int32(gallivm, 1), "");
mask = LLVMBuildOr(builder, vertex_id, clipmask, "");
- mask = LLVMBuildOr(builder, mask, have_clipdist, "");
+ if (0) {
+ mask = LLVMBuildOr(builder, mask, pad, "");
+ }
mask = LLVMBuildOr(builder, mask, edgeflag, "");
#endif
return mask;
int attrib,
int num_outputs,
LLVMValueRef clipmask,
- boolean have_clipdist)
+ boolean need_edgeflag)
{
LLVMBuilderRef builder = gallivm->builder;
LLVMValueRef attr_index = lp_build_const_int32(gallivm, attrib);
* code here. See struct vertex_header in draw_private.h.
*/
assert(DRAW_TOTAL_CLIP_PLANES==14);
- /* initialize vertex id:16 = 0xffff, have_clipdist:1 = 0, edgeflag:1 = 1 */
- vertex_id_pad_edgeflag = (0xffff << 16) | (1 << DRAW_TOTAL_CLIP_PLANES);
- if (have_clipdist)
- vertex_id_pad_edgeflag |= 1 << (DRAW_TOTAL_CLIP_PLANES+1);
- val = lp_build_const_int_vec(gallivm, lp_int_type(soa_type), vertex_id_pad_edgeflag);
+ /* initialize vertex id:16 = 0xffff, pad:1 = 0, edgeflag:1 = 1 */
+ if (!need_edgeflag) {
+ vertex_id_pad_edgeflag = (0xffff << 16) | (1 << DRAW_TOTAL_CLIP_PLANES);
+ }
+ else {
+ vertex_id_pad_edgeflag = (0xffff << 16);
+ }
+ val = lp_build_const_int_vec(gallivm, lp_int_type(soa_type),
+ vertex_id_pad_edgeflag);
/* OR with the clipmask */
cliptmp = LLVMBuildOr(builder, val, clipmask, "");
for (i = 0; i < vector_length; i++) {
LLVMValueRef clipmask,
int num_outputs,
struct lp_type soa_type,
- boolean have_clipdist)
+ boolean need_edgeflag)
{
LLVMBuilderRef builder = gallivm->builder;
unsigned chan, attrib, i;
aos,
attrib,
num_outputs,
- clipmask, have_clipdist);
+ clipmask,
+ need_edgeflag);
}
#if DEBUG_STORE
lp_build_printf(gallivm, " # storing end\n");
const struct lp_type vs_type,
LLVMValueRef io_ptr,
LLVMValueRef (*outputs)[TGSI_NUM_CHANNELS],
- boolean pre_clip_pos, int idx)
+ int idx)
{
LLVMBuilderRef builder = gallivm->builder;
LLVMValueRef soa[4];
soa[2] = LLVMBuildLoad(builder, outputs[idx][2], ""); /*z0 z1 .. zn*/
soa[3] = LLVMBuildLoad(builder, outputs[idx][3], ""); /*w0 w1 .. wn*/
- if (!pre_clip_pos) {
- for (i = 0; i < vs_type.length; i++) {
- clip_ptrs[i] = draw_jit_header_clip(gallivm, io_ptrs[i]);
- }
- } else {
- for (i = 0; i < vs_type.length; i++) {
- clip_ptrs[i] = draw_jit_header_pre_clip_pos(gallivm, io_ptrs[i]);
- }
+ for (i = 0; i < vs_type.length; i++) {
+ clip_ptrs[i] = draw_jit_header_clip_pos(gallivm, io_ptrs[i]);
}
lp_build_transpose_aos(gallivm, vs_type, soa, soa);
struct gallivm_state *gallivm,
struct lp_type vs_type,
LLVMValueRef (*outputs)[TGSI_NUM_CHANNELS],
- boolean clip_xy,
- boolean clip_z,
- boolean clip_user,
- boolean clip_halfz,
- unsigned ucp_enable,
+ struct draw_llvm_variant_key *key,
LLVMValueRef context_ptr,
boolean *have_clipdist)
{
const unsigned pos = llvm->draw->vs.position_output;
const unsigned cv = llvm->draw->vs.clipvertex_output;
int num_written_clipdistance = llvm->draw->vs.vertex_shader->info.num_written_clipdistance;
- bool have_cd = false;
+ boolean have_cd = false;
+ boolean clip_user = key->clip_user;
+ unsigned ucp_enable = key->ucp_enable;
unsigned cd[2];
cd[0] = llvm->draw->vs.clipdistance_output[0];
cv_w = pos_w;
}
+ /*
+ * Be careful with the comparisons and NaNs (using llvm's unordered
+ * comparisons here).
+ */
/* Cliptest, for hardwired planes */
- if (clip_xy) {
+ /*
+ * XXX should take guardband into account (currently not in key).
+ * Otherwise might run the draw pipeline stages for nothing.
+ */
+ if (key->clip_xy) {
/* plane 1 */
test = lp_build_compare(gallivm, f32_type, PIPE_FUNC_GREATER, pos_x , pos_w);
temp = shift;
mask = LLVMBuildOr(builder, mask, test, "");
}
- if (clip_z) {
+ if (key->clip_z) {
temp = lp_build_const_int_vec(gallivm, i32_type, 16);
- if (clip_halfz) {
+ if (key->clip_halfz) {
/* plane 5 */
test = lp_build_compare(gallivm, f32_type, PIPE_FUNC_GREATER, zero, pos_z);
test = LLVMBuildAnd(builder, test, temp, "");
}
}
}
+ if (key->need_edgeflags) {
+ /*
+ * This isn't really part of clipmask but stored the same in vertex
+ * header later, so do it here.
+ */
+ unsigned edge_attr = llvm->draw->vs.edgeflag_output;
+ LLVMValueRef one = lp_build_const_vec(gallivm, f32_type, 1.0);
+ LLVMValueRef edgeflag = LLVMBuildLoad(builder, outputs[edge_attr][0], "");
+ test = lp_build_compare(gallivm, f32_type, PIPE_FUNC_EQUAL, one, edgeflag);
+ temp = lp_build_const_int_vec(gallivm, i32_type,
+ 1LL << DRAW_TOTAL_CLIP_PLANES);
+ test = LLVMBuildAnd(builder, test, temp, "");
+ mask = LLVMBuildOr(builder, mask, test, "");
+ }
return mask;
}
static LLVMValueRef
clipmask_booli32(struct gallivm_state *gallivm,
const struct lp_type vs_type,
- LLVMValueRef clipmask_bool_ptr)
+ LLVMValueRef clipmask_bool_ptr,
+ boolean edgeflag_in_clipmask)
{
LLVMBuilderRef builder = gallivm->builder;
LLVMTypeRef int32_type = LLVMInt32TypeInContext(gallivm->context);
int i;
/*
- * Can do this with log2(vector length) pack instructions and one extract
- * (as we don't actually need a or) with sse2 which would be way better.
+ * We need to invert the edgeflag bit from the clipmask here
+ * (because the result is really if we want to run the pipeline or not
+ * and we (may) need it if edgeflag was 0).
+ */
+ if (edgeflag_in_clipmask) {
+ struct lp_type i32_type = lp_int_type(vs_type);
+ LLVMValueRef edge = lp_build_const_int_vec(gallivm, i32_type,
+ 1LL << DRAW_TOTAL_CLIP_PLANES);
+ clipmask_bool = LLVMBuildXor(builder, clipmask_bool, edge, "");
+ }
+ /*
+ * Could do much better with just cmp/movmskps.
*/
for (i=0; i < vs_type.length; i++) {
temp = LLVMBuildExtractElement(builder, clipmask_bool,
const boolean bypass_viewport = key->has_gs || key->bypass_viewport ||
llvm->draw->vs.vertex_shader->info.writes_viewport_index;
const boolean enable_cliptest = !key->has_gs && (key->clip_xy ||
- key->clip_z ||
- key->clip_user);
+ key->clip_z ||
+ key->clip_user ||
+ key->need_edgeflags);
LLVMValueRef variant_func;
const unsigned pos = llvm->draw->vs.position_output;
const unsigned cv = llvm->draw->vs.clipvertex_output;
if (pos != -1 && cv != -1) {
/* store original positions in clip before further manipulation */
- store_clip(gallivm, vs_type, io, outputs, FALSE, key->clip_user ? cv : pos);
- store_clip(gallivm, vs_type, io, outputs, TRUE, pos);
+ store_clip(gallivm, vs_type, io, outputs, pos);
/* do cliptest */
if (enable_cliptest) {
gallivm,
vs_type,
outputs,
- key->clip_xy,
- key->clip_z,
- key->clip_user,
- key->clip_halfz,
- key->ucp_enable,
+ key,
context_ptr, &have_clipdist);
temp = LLVMBuildOr(builder, clipmask, temp, "");
/* store temporary clipping boolean value */
*/
convert_to_aos(gallivm, io, NULL, outputs, clipmask,
vs_info->num_outputs, vs_type,
- have_clipdist);
+ enable_cliptest && key->need_edgeflags);
}
lp_build_loop_end_cond(&lp_loop, count, step, LLVMIntUGE);
sampler->destroy(sampler);
/* return clipping boolean value for function */
- ret = clipmask_booli32(gallivm, vs_type, clipmask_bool_ptr);
+ ret = clipmask_booli32(gallivm, vs_type, clipmask_bool_ptr,
+ enable_cliptest && key->need_edgeflags);
LLVMBuildRet(builder, ret);
key = (struct draw_llvm_variant_key *)store;
+ memset(key, 0, offsetof(struct draw_llvm_variant_key, vertex_element[0]));
+
key->clamp_vertex_color = llvm->draw->rasterizer->clamp_vertex_color; /**/
/* Presumably all variants of the shader should have the same
key->clip_user = llvm->draw->clip_user;
key->bypass_viewport = llvm->draw->bypass_viewport;
key->clip_halfz = llvm->draw->rasterizer->clip_halfz;
+ /* XXX assumes edgeflag output not at 0 */
key->need_edgeflags = (llvm->draw->vs.edgeflag_output ? TRUE : FALSE);
key->ucp_enable = llvm->draw->rasterizer->clip_plane_enable;
key->has_gs = llvm->draw->gs.geometry_shader != NULL;
key->num_outputs = draw_total_vs_outputs(llvm->draw);
- key->pad1 = 0;
/* All variants of this shader will have the same value for
* nr_samplers. Not yet trying to compact away holes in the
key = (struct draw_gs_llvm_variant_key *)store;
+ memset(key, 0, offsetof(struct draw_gs_llvm_variant_key, samplers[0]));
+
key->num_outputs = draw_total_gs_outputs(llvm->draw);
/* All variants of this shader will have the same value for
enum {
DRAW_JIT_VERTEX_VERTEX_ID = 0,
- DRAW_JIT_VERTEX_CLIP,
- DRAW_JIT_VERTEX_PRE_CLIP_POS,
+ DRAW_JIT_VERTEX_CLIP_POS,
DRAW_JIT_VERTEX_DATA
};
#define draw_jit_header_id(_gallivm, _ptr) \
lp_build_struct_get_ptr(_gallivm, _ptr, DRAW_JIT_VERTEX_VERTEX_ID, "id")
-#define draw_jit_header_clip(_gallivm, _ptr) \
- lp_build_struct_get_ptr(_gallivm, _ptr, DRAW_JIT_VERTEX_CLIP, "clip")
-
-#define draw_jit_header_pre_clip_pos(_gallivm, _ptr) \
- lp_build_struct_get_ptr(_gallivm, _ptr, DRAW_JIT_VERTEX_PRE_CLIP_POS, "pre_clip_pos")
+#define draw_jit_header_clip_pos(_gallivm, _ptr) \
+ lp_build_struct_get_ptr(_gallivm, _ptr, DRAW_JIT_VERTEX_CLIP_POS, "clip_pos")
#define draw_jit_header_data(_gallivm, _ptr) \
lp_build_struct_get_ptr(_gallivm, _ptr, DRAW_JIT_VERTEX_DATA, "data")
unsigned need_edgeflags:1;
unsigned has_gs:1;
unsigned num_outputs:8;
- /*
- * it is important there are no holes in this struct
- * (and all padding gets zeroed).
- */
unsigned ucp_enable:PIPE_MAX_CLIP_PLANES;
- unsigned pad1:24-PIPE_MAX_CLIP_PLANES;
+ /* note padding here - must use memset */
/* Variable number of vertex elements:
*/
unsigned nr_samplers:8;
unsigned nr_sampler_views:8;
unsigned num_outputs:8;
+ /* note padding here - must use memset */
struct draw_sampler_static_state samplers[1];
};
struct pipe_context *pipe = draw->pipe;
const struct pipe_rasterizer_state *rast = draw->rasterizer;
uint num_samplers;
+ uint num_sampler_views;
void *r;
assert(draw->rasterizer->line_smooth);
draw_aaline_prepare_outputs(draw, draw->pipeline.aaline);
/* how many samplers? */
- /* we'll use sampler/texture[pstip->sampler_unit] for the stipple */
- num_samplers = MAX2(aaline->num_sampler_views, aaline->num_samplers);
- num_samplers = MAX2(num_samplers, aaline->fs->sampler_unit + 1);
+ /* we'll use sampler/texture[aaline->sampler_unit] for the alpha texture */
+ num_samplers = MAX2(aaline->num_samplers, aaline->fs->sampler_unit + 1);
+ num_sampler_views = MAX2(num_samplers, aaline->num_sampler_views);
aaline->state.sampler[aaline->fs->sampler_unit] = aaline->sampler_cso;
pipe_sampler_view_reference(&aaline->state.sampler_views[aaline->fs->sampler_unit],
num_samplers, aaline->state.sampler);
aaline->driver_set_sampler_views(pipe, PIPE_SHADER_FRAGMENT, 0,
- num_samplers, aaline->state.sampler_views);
+ num_sampler_views, aaline->state.sampler_views);
/* Disable triangle culling, stippling, unfilled mode etc. */
r = draw_get_rasterizer_no_cull(draw, rast->scissor, rast->flatshade);
const struct pipe_rasterizer_state *rast = draw->rasterizer;
/* update vertex attrib info */
- aaline->pos_slot = draw_current_shader_position_output(draw);;
+ aaline->pos_slot = draw_current_shader_position_output(draw);
if (!rast->line_smooth)
return;
struct draw_stage stage; /**< base class */
unsigned pos_attr;
+ boolean have_clipdist;
+ int cv_attr;
/* List of the attributes to be constant interpolated. */
uint num_const_attribs;
*/
dst->clipmask = 0;
dst->edgeflag = 0; /* will get overwritten later */
- dst->have_clipdist = in->have_clipdist;
+ dst->pad = 0;
dst->vertex_id = UNDEFINED_VERTEX_ID;
/* Interpolate the clip-space coords.
*/
- interp_attr(dst->clip, t, in->clip, out->clip);
+ if (clip->cv_attr >= 0) {
+ interp_attr(dst->data[clip->cv_attr], t,
+ in->data[clip->cv_attr], out->data[clip->cv_attr]);
+ }
/* interpolate the clip-space position */
- interp_attr(dst->pre_clip_pos, t, in->pre_clip_pos, out->pre_clip_pos);
+ interp_attr(dst->clip_pos, t, in->clip_pos, out->clip_pos);
/* Do the projective divide and viewport transformation to get
* new window coordinates:
*/
{
- const float *pos = dst->pre_clip_pos;
+ const float *pos = dst->clip_pos;
const float *scale =
clip->stage.draw->viewports[viewport_index].scale;
const float *trans =
t_nopersp = t;
/* find either in.x != out.x or in.y != out.y */
for (k = 0; k < 2; k++) {
- if (in->pre_clip_pos[k] != out->pre_clip_pos[k]) {
+ if (in->clip_pos[k] != out->clip_pos[k]) {
/* do divide by W, then compute linear interpolation factor */
- float in_coord = in->pre_clip_pos[k] / in->pre_clip_pos[3];
- float out_coord = out->pre_clip_pos[k] / out->pre_clip_pos[3];
- float dst_coord = dst->pre_clip_pos[k] / dst->pre_clip_pos[3];
+ float in_coord = in->clip_pos[k] / in->clip_pos[3];
+ float out_coord = out->clip_pos[k] / out->clip_pos[3];
+ float dst_coord = dst->clip_pos[k] / dst->clip_pos[3];
t_nopersp = (dst_coord - out_coord) / (in_coord - out_coord);
break;
}
* Triangle is considered null/empty if its area is equal to zero.
*/
static inline boolean
-is_tri_null(struct draw_context *draw, const struct prim_header *header)
+is_tri_null(const struct clip_stage *clip, const struct prim_header *header)
{
- const unsigned pos_attr = draw_current_shader_position_output(draw);
+ const unsigned pos_attr = clip->pos_attr;
float x1 = header->v[1]->data[pos_attr][0] - header->v[0]->data[pos_attr][0];
float y1 = header->v[1]->data[pos_attr][1] - header->v[0]->data[pos_attr][1];
float z1 = header->v[1]->data[pos_attr][2] - header->v[0]->data[pos_attr][2];
unsigned n,
const struct prim_header *origPrim)
{
+ const struct clip_stage *clipper = clip_stage(stage);
struct prim_header header;
unsigned i;
ushort edge_first, edge_middle, edge_last;
header.v[2] = inlist[0]; /* the provoking vertex */
}
- tri_null = is_tri_null(stage->draw, &header);
+ tri_null = is_tri_null(clipper, &header);
/* If we generated a triangle with an area, aka. non-null triangle,
* or if the previous triangle was also null then skip all subsequent
* null triangles */
debug_printf("Clipped tri: (flat-shade-first = %d)\n",
stage->draw->rasterizer->flatshade_first);
for (j = 0; j < 3; j++) {
- debug_printf(" Vert %d: clip: %f %f %f %f\n", j,
- header.v[j]->clip[0],
- header.v[j]->clip[1],
- header.v[j]->clip[2],
- header.v[j]->clip[3]);
+ debug_printf(" Vert %d: clip pos: %f %f %f %f\n", j,
+ header.v[j]->clip_pos[0],
+ header.v[j]->clip_pos[1],
+ header.v[j]->clip_pos[2],
+ header.v[j]->clip_pos[3]);
+ if (clipper->cv_attr >= 0) {
+ debug_printf(" Vert %d: cv: %f %f %f %f\n", j,
+ header.v[j]->data[clipper->cv_attr][0],
+ header.v[j]->data[clipper->cv_attr][1],
+ header.v[j]->data[clipper->cv_attr][2],
+ header.v[j]->data[clipper->cv_attr][3]);
+ }
for (k = 0; k < draw_num_shader_outputs(stage->draw); k++) {
debug_printf(" Vert %d: Attr %d: %f %f %f %f\n", j, k,
header.v[j]->data[k][0],
}
}
}
- stage->next->tri( stage->next, &header );
+ stage->next->tri(stage->next, &header);
}
}
{
const float *plane;
float dp;
- if (vert->have_clipdist && plane_idx >= 6) {
+ if (plane_idx < 6) {
+ /* ordinary xyz view volume clipping uses pos output */
+ plane = clipper->plane[plane_idx];
+ dp = dot4(vert->clip_pos, plane);
+ }
+ else if (clipper->have_clipdist) {
/* pick the correct clipdistance element from the output vectors */
int _idx = plane_idx - 6;
int cdi = _idx >= 4;
int vidx = cdi ? _idx - 4 : _idx;
dp = vert->data[draw_current_shader_clipdistance_output(clipper->stage.draw, cdi)][vidx];
} else {
+ /*
+ * legacy user clip planes or gl_ClipVertex
+ */
plane = clipper->plane[plane_idx];
- dp = dot4(vert->clip, plane);
+ if (clipper->cv_attr >= 0) {
+ dp = dot4(vert->data[clipper->cv_attr], plane);
+ }
+ else {
+ dp = dot4(vert->clip_pos, plane);
+ }
}
return dp;
}
viewport_index = draw_viewport_index(clipper->stage.draw, prov_vertex);
if (DEBUG_CLIP) {
- const float *v0 = header->v[0]->clip;
- const float *v1 = header->v[1]->clip;
- const float *v2 = header->v[2]->clip;
- debug_printf("Clip triangle:\n");
+ const float *v0 = header->v[0]->clip_pos;
+ const float *v1 = header->v[1]->clip_pos;
+ const float *v2 = header->v[2]->clip_pos;
+ debug_printf("Clip triangle pos:\n");
debug_printf(" %f, %f, %f, %f\n", v0[0], v0[1], v0[2], v0[3]);
debug_printf(" %f, %f, %f, %f\n", v1[0], v1[1], v1[2], v1[3]);
debug_printf(" %f, %f, %f, %f\n", v2[0], v2[1], v2[2], v2[3]);
+ if (clipper->cv_attr >= 0) {
+ const float *v0 = header->v[0]->data[clipper->cv_attr];
+ const float *v1 = header->v[1]->data[clipper->cv_attr];
+ const float *v2 = header->v[2]->data[clipper->cv_attr];
+ debug_printf("Clip triangle cv:\n");
+ debug_printf(" %f, %f, %f, %f\n", v0[0], v0[1], v0[2], v0[3]);
+ debug_printf(" %f, %f, %f, %f\n", v1[0], v1[1], v1[2], v1[3]);
+ debug_printf(" %f, %f, %f, %f\n", v2[0], v2[1], v2[2], v2[3]);
+ }
}
/*
/* Emit the polygon as triangles to the setup stage:
*/
- emit_poly( stage, inlist, inEdges, n, header );
+ emit_poly(stage, inlist, inEdges, n, header);
}
}
struct prim_header *header,
unsigned clipmask)
{
- const struct clip_stage *clipper = clip_stage( stage );
+ const struct clip_stage *clipper = clip_stage(stage);
struct vertex_header *v0 = header->v[0];
struct vertex_header *v1 = header->v[1];
struct vertex_header *prov_vertex;
struct prim_header newprim;
int viewport_index;
+ newprim.flags = header->flags;
+
if (stage->draw->rasterizer->flatshade_first) {
prov_vertex = v0;
}
* automatically). These would usually be captured by depth clip
* too but this can be disabled.
*/
- if (header->v[0]->clip[3] <= 0.0f ||
- util_is_inf_or_nan(header->v[0]->clip[0]) ||
- util_is_inf_or_nan(header->v[0]->clip[1]))
+ if (header->v[0]->clip_pos[3] <= 0.0f ||
+ util_is_inf_or_nan(header->v[0]->clip_pos[0]) ||
+ util_is_inf_or_nan(header->v[0]->clip_pos[1]))
return;
}
stage->next->point(stage->next, header);
static void
clip_init_state(struct draw_stage *stage)
{
- struct clip_stage *clipper = clip_stage( stage );
+ struct clip_stage *clipper = clip_stage(stage);
const struct draw_context *draw = stage->draw;
const struct draw_fragment_shader *fs = draw->fs.fragment_shader;
const struct tgsi_shader_info *info = draw_get_shader_info(draw);
int indexed_interp[2];
clipper->pos_attr = draw_current_shader_position_output(draw);
+ clipper->have_clipdist = draw_current_shader_num_written_clipdistances(draw) > 0;
+ if (draw_current_shader_clipvertex_output(draw) != clipper->pos_attr) {
+ clipper->cv_attr = (int)draw_current_shader_clipvertex_output(draw);
+ }
+ else {
+ clipper->cv_attr = -1;
+ }
/* We need to know for each attribute what kind of interpolation is
* done on it (flat, smooth or noperspective). But the information
struct pipe_context *pipe = pstip->pipe;
struct draw_context *draw = stage->draw;
uint num_samplers;
+ uint num_sampler_views;
assert(stage->draw->rasterizer->poly_stipple_enable);
/* how many samplers? */
/* we'll use sampler/texture[pstip->sampler_unit] for the stipple */
- num_samplers = MAX2(pstip->num_sampler_views, pstip->num_samplers);
- num_samplers = MAX2(num_samplers, pstip->fs->sampler_unit + 1);
+ num_samplers = MAX2(pstip->num_samplers, pstip->fs->sampler_unit + 1);
+ num_sampler_views = MAX2(pstip->num_sampler_views, num_samplers);
/* plug in our sampler, texture */
pstip->state.samplers[pstip->fs->sampler_unit] = pstip->sampler_cso;
num_samplers, pstip->state.samplers);
pstip->driver_set_sampler_views(pipe, PIPE_SHADER_FRAGMENT, 0,
- num_samplers, pstip->state.sampler_views);
+ num_sampler_views, pstip->state.sampler_views);
draw->suspend_flushing = FALSE;
}
-static void point( struct draw_stage *stage,
- struct vertex_header *v0 )
+static void point(struct draw_stage *stage,
+ struct prim_header *header,
+ struct vertex_header *v0)
{
struct prim_header tmp;
+ tmp.det = header->det;
+ tmp.flags = 0;
tmp.v[0] = v0;
- stage->next->point( stage->next, &tmp );
+ stage->next->point(stage->next, &tmp);
}
-static void line( struct draw_stage *stage,
- struct vertex_header *v0,
- struct vertex_header *v1 )
+static void line(struct draw_stage *stage,
+ struct prim_header *header,
+ struct vertex_header *v0,
+ struct vertex_header *v1)
{
struct prim_header tmp;
+ tmp.det = header->det;
+ tmp.flags = 0;
tmp.v[0] = v0;
tmp.v[1] = v1;
- stage->next->line( stage->next, &tmp );
+ stage->next->line(stage->next, &tmp);
}
-static void points( struct draw_stage *stage,
- struct prim_header *header )
+static void points(struct draw_stage *stage,
+ struct prim_header *header)
{
struct vertex_header *v0 = header->v[0];
struct vertex_header *v1 = header->v[1];
inject_front_face_info(stage, header);
- if ((header->flags & DRAW_PIPE_EDGE_FLAG_0) && v0->edgeflag) point( stage, v0 );
- if ((header->flags & DRAW_PIPE_EDGE_FLAG_1) && v1->edgeflag) point( stage, v1 );
- if ((header->flags & DRAW_PIPE_EDGE_FLAG_2) && v2->edgeflag) point( stage, v2 );
+ if ((header->flags & DRAW_PIPE_EDGE_FLAG_0) && v0->edgeflag)
+ point(stage, header, v0);
+ if ((header->flags & DRAW_PIPE_EDGE_FLAG_1) && v1->edgeflag)
+ point(stage, header, v1);
+ if ((header->flags & DRAW_PIPE_EDGE_FLAG_2) && v2->edgeflag)
+ point(stage, header, v2);
}
-static void lines( struct draw_stage *stage,
- struct prim_header *header )
+static void lines(struct draw_stage *stage,
+ struct prim_header *header)
{
struct vertex_header *v0 = header->v[0];
struct vertex_header *v1 = header->v[1];
struct vertex_header *v2 = header->v[2];
if (header->flags & DRAW_PIPE_RESET_STIPPLE)
- stage->next->reset_stipple_counter( stage->next );
+ /*
+ * XXX could revisit this. The only stage which cares is the line
+ * stipple stage. Could just emit correct reset flags here and not
+ * bother about all the calling through reset_stipple_counter
+ * stages. Though technically it is necessary if line stipple is
+ * handled by the driver, but this is not actually hooked up when
+ * using vbuf (vbuf stage reset_stipple_counter does nothing).
+ */
+ stage->next->reset_stipple_counter(stage->next);
inject_front_face_info(stage, header);
- if ((header->flags & DRAW_PIPE_EDGE_FLAG_2) && v2->edgeflag) line( stage, v2, v0 );
- if ((header->flags & DRAW_PIPE_EDGE_FLAG_0) && v0->edgeflag) line( stage, v0, v1 );
- if ((header->flags & DRAW_PIPE_EDGE_FLAG_1) && v1->edgeflag) line( stage, v1, v2 );
+ if ((header->flags & DRAW_PIPE_EDGE_FLAG_2) && v2->edgeflag)
+ line(stage, header, v2, v0);
+ if ((header->flags & DRAW_PIPE_EDGE_FLAG_0) && v0->edgeflag)
+ line(stage, header, v0, v1);
+ if ((header->flags & DRAW_PIPE_EDGE_FLAG_1) && v1->edgeflag)
+ line(stage, header, v1, v2);
}
unsigned max_indices;
unsigned nr_indices;
- /* Cache point size somewhere it's address won't change:
+ /* Cache point size somewhere its address won't change:
*/
float point_size;
+ float zero4[4];
struct translate_cache *cache;
};
struct translate_key hw_key;
unsigned dst_offset;
unsigned i;
+ const struct vertex_info *vinfo;
vbuf->render->set_primitive(vbuf->render, prim);
* state change.
*/
vbuf->vinfo = vbuf->render->get_vertex_info(vbuf->render);
- vbuf->vertex_size = vbuf->vinfo->size * sizeof(float);
+ vinfo = vbuf->vinfo;
+ vbuf->vertex_size = vinfo->size * sizeof(float);
/* Translate from pipeline vertices to hw vertices.
*/
dst_offset = 0;
- for (i = 0; i < vbuf->vinfo->num_attribs; i++) {
+ for (i = 0; i < vinfo->num_attribs; i++) {
unsigned emit_sz = 0;
unsigned src_buffer = 0;
enum pipe_format output_format;
- unsigned src_offset = (vbuf->vinfo->attrib[i].src_index * 4 * sizeof(float) );
+ unsigned src_offset = (vinfo->attrib[i].src_index * 4 * sizeof(float) );
- output_format = draw_translate_vinfo_format(vbuf->vinfo->attrib[i].emit);
- emit_sz = draw_translate_vinfo_size(vbuf->vinfo->attrib[i].emit);
+ output_format = draw_translate_vinfo_format(vinfo->attrib[i].emit);
+ emit_sz = draw_translate_vinfo_size(vinfo->attrib[i].emit);
/* doesn't handle EMIT_OMIT */
assert(emit_sz != 0);
- if (vbuf->vinfo->attrib[i].emit == EMIT_1F_PSIZE) {
- src_buffer = 1;
- src_offset = 0;
+ if (vinfo->attrib[i].emit == EMIT_1F_PSIZE) {
+ src_buffer = 1;
+ src_offset = 0;
+ }
+ else if (vinfo->attrib[i].src_index == DRAW_ATTR_NONEXIST) {
+ /* elements which don't exist will get assigned zeros */
+ src_buffer = 2;
+ src_offset = 0;
}
hw_key.element[i].type = TRANSLATE_ELEMENT_NORMAL;
dst_offset += emit_sz;
}
- hw_key.nr_elements = vbuf->vinfo->num_attribs;
+ hw_key.nr_elements = vinfo->num_attribs;
hw_key.output_stride = vbuf->vertex_size;
/* Don't bother with caching at this stage:
vbuf->translate = translate_cache_find(vbuf->cache, &hw_key);
vbuf->translate->set_buffer(vbuf->translate, 1, &vbuf->point_size, 0, ~0);
+ vbuf->translate->set_buffer(vbuf->translate, 2, &vbuf->zero4[0], 0, ~0);
}
vbuf->point_size = vbuf->stage.draw->rasterizer->point_size;
struct vbuf_stage *vbuf = CALLOC_STRUCT(vbuf_stage);
if (!vbuf)
goto fail;
-
+
vbuf->stage.draw = draw;
vbuf->stage.name = "vbuf";
vbuf->stage.point = vbuf_first_point;
vbuf->stage.flush = vbuf_flush;
vbuf->stage.reset_stipple_counter = vbuf_reset_stipple_counter;
vbuf->stage.destroy = vbuf_destroy;
-
+
vbuf->render = render;
vbuf->max_indices = MIN2(render->max_indices, UNDEFINED_VERTEX_ID-1);
- vbuf->indices = (ushort *) align_malloc( vbuf->max_indices *
- sizeof(vbuf->indices[0]),
- 16 );
+ vbuf->indices = (ushort *) align_malloc(vbuf->max_indices *
+ sizeof(vbuf->indices[0]),
+ 16);
if (!vbuf->indices)
goto fail;
vbuf->cache = translate_cache_create();
- if (!vbuf->cache)
+ if (!vbuf->cache)
goto fail;
-
-
+
vbuf->vertices = NULL;
vbuf->vertex_ptr = vbuf->vertices;
-
+
+ vbuf->zero4[0] = vbuf->zero4[1] = vbuf->zero4[2] = vbuf->zero4[3] = 0.0f;
+
return &vbuf->stage;
- fail:
+fail:
if (vbuf)
vbuf_destroy(&vbuf->stage);
-
+
return NULL;
}
struct vertex_header {
unsigned clipmask:DRAW_TOTAL_CLIP_PLANES;
unsigned edgeflag:1;
- unsigned have_clipdist:1;
+ unsigned pad:1;
unsigned vertex_id:16;
- float clip[4];
- float pre_clip_pos[4];
+ float clip_pos[4];
/* This will probably become float (*data)[4] soon:
*/
#endif
{
if (index_limit == 0) {
- /* one of the buffers is too small to do any valid drawing */
+ /* one of the buffers is too small to do any valid drawing */
debug_warning("draw: VBO too small to draw anything\n");
util_fpstate_set(fpstate);
return;
unsigned prim;
const struct vertex_info *vinfo;
+
+ float zero4[4];
+
};
/* XXX: need to flush to get prim_vbuf.c to release its allocation??
*/
- draw_do_flush( draw, DRAW_FLUSH_BACKEND );
+ draw_do_flush(draw, DRAW_FLUSH_BACKEND);
/* XXX: may need to defensively reset this later on as clipping can
* clobber this state in the render backend.
unsigned emit_sz = 0;
unsigned src_buffer = 0;
unsigned output_format;
- unsigned src_offset = (vinfo->attrib[i].src_index * 4 * sizeof(float) );
+ unsigned src_offset = vinfo->attrib[i].src_index * 4 * sizeof(float);
output_format = draw_translate_vinfo_format(vinfo->attrib[i].emit);
emit_sz = draw_translate_vinfo_size(vinfo->attrib[i].emit);
assert(emit_sz != 0);
if (vinfo->attrib[i].emit == EMIT_1F_PSIZE) {
- src_buffer = 1;
- src_offset = 0;
+ src_buffer = 1;
+ src_offset = 0;
+ }
+ else if (vinfo->attrib[i].src_index == DRAW_ATTR_NONEXIST) {
+ /* elements which don't exist will get assigned zeros */
+ src_buffer = 2;
+ src_offset = 0;
}
hw_key.element[i].type = TRANSLATE_ELEMENT_NORMAL;
translate_key_compare(&emit->translate->key, &hw_key) != 0) {
translate_key_sanitize(&hw_key);
emit->translate = translate_cache_find(emit->cache, &hw_key);
+
+ emit->translate->set_buffer(emit->translate, 2, &emit->zero4[0], 0, ~0);
}
if (!vinfo->size)
/* XXX: need to flush to get prim_vbuf.c to release its allocation??
*/
- draw_do_flush( draw, DRAW_FLUSH_BACKEND );
+ draw_do_flush(draw, DRAW_FLUSH_BACKEND);
if (vertex_count == 0)
return;
(ushort)translate->key.output_stride,
(ushort)vertex_count);
- hw_verts = render->map_vertices( render );
+ hw_verts = render->map_vertices(render);
if (!hw_verts) {
debug_warn_once("map of vertex buffer failed (out of memory?)");
return;
}
translate->set_buffer(translate,
- 0,
- vertex_data,
- stride,
- ~0);
+ 0,
+ vertex_data,
+ stride,
+ ~0);
translate->set_buffer(translate,
- 1,
- &draw->rasterizer->point_size,
- 0,
- ~0);
+ 1,
+ &draw->rasterizer->point_size,
+ 0,
+ ~0);
/* fetch/translate vertex attribs to fill hw_verts[] */
translate->run(translate,
- 0,
- vertex_count,
- draw->start_instance,
- draw->instance_id,
- hw_verts );
+ 0,
+ vertex_count,
+ 0,
+ 0,
+ hw_verts);
render->unmap_vertices(render, 0, vertex_count - 1);
#endif
/* XXX: need to flush to get prim_vbuf.c to release its allocation??
*/
- draw_do_flush( draw, DRAW_FLUSH_BACKEND );
+ draw_do_flush(draw, DRAW_FLUSH_BACKEND);
/* XXX: and work out some way to coordinate the render primitive
* between vbuf.c and here...
(ushort)count))
goto fail;
- hw_verts = render->map_vertices( render );
+ hw_verts = render->map_vertices(render);
if (!hw_verts)
goto fail;
translate->set_buffer(translate, 0,
- vertex_data, stride, count - 1);
+ vertex_data, stride, count - 1);
translate->set_buffer(translate, 1,
- &draw->rasterizer->point_size,
- 0, ~0);
+ &draw->rasterizer->point_size,
+ 0, ~0);
translate->run(translate,
0,
count,
- draw->start_instance,
- draw->instance_id,
+ 0,
+ 0,
hw_verts);
if (0) {
unsigned i;
for (i = 0; i < count; i++) {
debug_printf("\n\n%s vertex %d:\n", __FUNCTION__, i);
- draw_dump_emitted_vertex( emit->vinfo,
- (const uint8_t *)hw_verts +
- translate->key.output_stride * i );
+ draw_dump_emitted_vertex(emit->vinfo,
+ (const uint8_t *)hw_verts +
+ translate->key.output_stride * i);
}
}
- render->unmap_vertices( render, 0, count - 1 );
+ render->unmap_vertices(render, 0, count - 1);
for (start = i = 0;
i < prim_info->primitive_count;
start,
prim_info->primitive_lengths[i]);
}
-
+
render->release_vertices(render);
return;
return NULL;
}
+ emit->zero4[0] = emit->zero4[1] = emit->zero4[2] = emit->zero4[3] = 0.0f;
+
return emit;
}
draw->vs.vertex_shader->info.writes_viewport_index)) {
clipped = draw_pt_post_vs_run( fpme->post_vs, vert_info, prim_info );
}
+ /* "clipped" also includes non-one edgeflag */
if (clipped) {
opt |= PT_PIPELINE;
}
{
header->clipmask = 0;
header->edgeflag = 1;
- header->have_clipdist = 0;
+ header->pad = 0;
header->vertex_id = UNDEFINED_VERTEX_ID;
}
emit->generated_primitives = 0;
emit->input_vertex_stride = input_verts->stride;
if (emit->use_pre_clip_pos)
- emit->pre_clip_pos = input_verts->verts->pre_clip_pos;
+ emit->pre_clip_pos = input_verts->verts->clip_pos;
emit->inputs = (const float (*)[4])input_verts->verts->data;
#include "util/u_debug.h"
#include "util/u_memory.h"
+#define DRAW_ATTR_NONEXIST 255
/**
* Vertex attribute emit modes
/**
- * Attribute interpolation mode
- */
-enum interp_mode {
- INTERP_NONE, /**< never interpolate vertex header info */
- INTERP_POS, /**< special case for frag position */
- INTERP_CONSTANT,
- INTERP_LINEAR,
- INTERP_PERSPECTIVE
-};
-
-
-/**
* Information about hardware/rasterization vertex layout.
*/
struct vertex_info
* memcmp() comparisons.
*/
struct {
- unsigned interp_mode:4; /**< INTERP_x */
- unsigned emit:4; /**< EMIT_x */
+ unsigned emit:8; /**< EMIT_x */
unsigned src_index:8; /**< map to post-xform attribs */
} attrib[PIPE_MAX_SHADER_OUTPUTS];
};
static inline uint
draw_emit_vertex_attr(struct vertex_info *vinfo,
enum attrib_emit emit,
- enum interp_mode interp, /* only used by softpipe??? */
int src_index)
{
const uint n = vinfo->num_attribs;
/* If the src_index is negative, meaning it hasn't been found
- * lets just redirect it to the first output slot */
+ * we'll assign it all zeros later - set to DRAW_ATTR_NONEXIST */
if (src_index < 0) {
- src_index = 0;
+ src_index = DRAW_ATTR_NONEXIST;
}
assert(n < Elements(vinfo->attrib));
vinfo->attrib[n].emit = emit;
- vinfo->attrib[n].interp_mode = interp;
vinfo->attrib[n].src_index = src_index;
vinfo->num_attribs++;
return n;
{
/* Special case 4x4f --> 1x16ub */
if (src_type.length == 4 &&
- util_cpu_caps.has_sse2)
+ (util_cpu_caps.has_sse2 || util_cpu_caps.has_altivec))
{
num_dsts = (num_srcs + 3) / 4;
dst_type->length = num_srcs * 4 >= 16 ? 16 : num_srcs * 4;
((dst_type.length == 16 && 4 * num_dsts == num_srcs) ||
(num_dsts == 1 && dst_type.length * num_srcs == 16 && num_srcs != 3)) &&
- util_cpu_caps.has_sse2)
+ (util_cpu_caps.has_sse2 || util_cpu_caps.has_altivec))
{
struct lp_build_context bld;
struct lp_type int16_type, int32_type;
case PIPE_SHADER_CAP_TGSI_DROUND_SUPPORTED:
case PIPE_SHADER_CAP_TGSI_DFRACEXP_DLDEXP_SUPPORTED:
case PIPE_SHADER_CAP_TGSI_FMA_SUPPORTED:
+ case PIPE_SHADER_CAP_MAX_SHADER_BUFFERS:
return 0;
case PIPE_SHADER_CAP_MAX_UNROLL_ITERATIONS_HINT:
return 32;
default:
assert(0);
- };
+ }
}
if (need_perelement_offset) {
LLVMValueRef pixel_offsets;
- int i;
+ unsigned i;
/* build pixel offset vector: {0, 1, 2, 3, ...} */
pixel_offsets = uint_bld->undef;
for (i = 0; i < uint_bld->type.length; i++) {
struct gallivm_state *gallivm = bld_base->base.gallivm;
LLVMBuilderRef builder = gallivm->builder;
struct lp_build_context *float_bld = &bld_base->base;
- int i;
+ unsigned i;
LLVMValueRef temp, temp2;
LLVMValueRef shuffles[8];
LLVMValueRef shuffles2[8];
near_end_of_shader(struct lp_build_tgsi_soa_context *bld,
int pc)
{
- int i;
+ unsigned i;
for (i = 0; i < 5; i++) {
unsigned opcode;
v->max_num_vertices = num_vertices;
v->vbuf.stride = stride;
u_upload_alloc(hud->uploader, 0, v->vbuf.stride * v->max_num_vertices,
- &v->vbuf.buffer_offset, &v->vbuf.buffer,
+ 16, &v->vbuf.buffer_offset, &v->vbuf.buffer,
(void**)&v->vertices);
}
hud->pipe = pipe;
hud->cso = cso;
- hud->uploader = u_upload_create(pipe, 256 * 1024, 16,
- PIPE_BIND_VERTEX_BUFFER);
+ hud->uploader = u_upload_create(pipe, 256 * 1024,
+ PIPE_BIND_VERTEX_BUFFER, PIPE_USAGE_STREAM);
/* font */
if (!util_font_create(pipe, UTIL_FONT_FIXED_8X13, &hud->font)) {
}
if (!pc->upload) {
- pc->upload = u_upload_create(pc->pipe, 4096, 4, PIPE_BIND_INDEX_BUFFER);
+ pc->upload = u_upload_create(pc->pipe, 4096, PIPE_BIND_INDEX_BUFFER,
+ PIPE_USAGE_STREAM);
}
- u_upload_alloc(pc->upload, 0, new_ib.index_size * new_info.count,
+ u_upload_alloc(pc->upload, 0, new_ib.index_size * new_info.count, 4,
&new_ib.offset, &new_ib.buffer, &dst);
if (info->indexed) {
* IN THE SOFTWARE.
*/
-#ifdef __GNUC__
-#pragma GCC diagnostic ignored "-Wdeclaration-after-statement"
-#endif
-
#include "util/ralloc.h"
#include "glsl/nir/nir.h"
#include "glsl/nir/nir_control_flow.h"
static void
ttn_kill_if(nir_builder *b, nir_op op, nir_alu_dest dest, nir_ssa_def **src)
{
- nir_ssa_def *cmp = nir_bany4(b, nir_flt(b, src[0], nir_imm_float(b, 0.0)));
+ nir_ssa_def *cmp = nir_bany_inequal4(b, nir_flt(b, src[0],
+ nir_imm_float(b, 0.0)),
+ nir_imm_int(b, 0));
nir_intrinsic_instr *discard =
nir_intrinsic_instr_create(b->shader, nir_intrinsic_discard_if);
discard->src[0] = nir_src_for_ssa(cmp);
&tgsi_dst->Indirect : NULL;
store->num_components = 4;
+ store->const_index[0] = 0xf;
store->variables[0] = ttn_array_deref(c, store, var, offset, indirect);
store->src[0] = nir_src_for_reg(dest.dest.reg.reg);
case TGSI_PROCESSOR_COMPUTE: return MESA_SHADER_COMPUTE;
default:
unreachable("invalid TGSI processor");
- };
+ }
}
struct nir_shader *
tgsi_scan_shader(tgsi_tokens, &scan);
c->scan = &scan;
- s = nir_shader_create(NULL, tgsi_processor_to_shader_stage(scan.processor),
- options);
-
- nir_function *func = nir_function_create(s, "main");
- nir_function_overload *overload = nir_function_overload_create(func);
- nir_function_impl *impl = nir_function_impl_create(overload);
-
- nir_builder_init(&c->build, impl);
- c->build.cursor = nir_after_cf_list(&impl->body);
+ nir_builder_init_simple_shader(&c->build, NULL,
+ tgsi_processor_to_shader_stage(scan.processor),
+ options);
+ s = c->build.shader;
s->num_inputs = scan.file_max[TGSI_FILE_INPUT] + 1;
s->num_uniforms = scan.const_file_max[0] + 1;
return ds;
}
-static struct tgsi_declaration_resource
-tgsi_default_declaration_resource(void)
+static struct tgsi_declaration_image
+tgsi_default_declaration_image(void)
{
- struct tgsi_declaration_resource dr;
+ struct tgsi_declaration_image di;
- dr.Resource = TGSI_TEXTURE_BUFFER;
- dr.Raw = 0;
- dr.Writable = 0;
- dr.Padding = 0;
+ di.Resource = TGSI_TEXTURE_BUFFER;
+ di.Raw = 0;
+ di.Writable = 0;
+ di.Format = 0;
+ di.Padding = 0;
- return dr;
+ return di;
}
-static struct tgsi_declaration_resource
-tgsi_build_declaration_resource(unsigned texture,
- unsigned raw,
- unsigned writable,
- struct tgsi_declaration *declaration,
- struct tgsi_header *header)
+static struct tgsi_declaration_image
+tgsi_build_declaration_image(unsigned texture,
+ unsigned format,
+ unsigned raw,
+ unsigned writable,
+ struct tgsi_declaration *declaration,
+ struct tgsi_header *header)
{
- struct tgsi_declaration_resource dr;
+ struct tgsi_declaration_image di;
- dr = tgsi_default_declaration_resource();
- dr.Resource = texture;
- dr.Raw = raw;
- dr.Writable = writable;
+ di = tgsi_default_declaration_image();
+ di.Resource = texture;
+ di.Format = format;
+ di.Raw = raw;
+ di.Writable = writable;
declaration_grow(declaration, header);
- return dr;
+ return di;
}
static struct tgsi_declaration_sampler_view
full_declaration.Range = tgsi_default_declaration_range();
full_declaration.Semantic = tgsi_default_declaration_semantic();
full_declaration.Interp = tgsi_default_declaration_interp();
- full_declaration.Resource = tgsi_default_declaration_resource();
+ full_declaration.Image = tgsi_default_declaration_image();
full_declaration.SamplerView = tgsi_default_declaration_sampler_view();
full_declaration.Array = tgsi_default_declaration_array();
header );
}
- if (full_decl->Declaration.File == TGSI_FILE_RESOURCE) {
- struct tgsi_declaration_resource *dr;
+ if (full_decl->Declaration.File == TGSI_FILE_IMAGE) {
+ struct tgsi_declaration_image *di;
if (maxsize <= size) {
return 0;
}
- dr = (struct tgsi_declaration_resource *)&tokens[size];
+ di = (struct tgsi_declaration_image *)&tokens[size];
size++;
- *dr = tgsi_build_declaration_resource(full_decl->Resource.Resource,
- full_decl->Resource.Raw,
- full_decl->Resource.Writable,
- declaration,
- header);
+ *di = tgsi_build_declaration_image(full_decl->Image.Resource,
+ full_decl->Image.Format,
+ full_decl->Image.Raw,
+ full_decl->Image.Writable,
+ declaration,
+ header);
}
if (full_decl->Declaration.File == TGSI_FILE_SAMPLER_VIEW) {
instruction.NumSrcRegs = 1;
instruction.Label = 0;
instruction.Texture = 0;
- instruction.Padding = 0;
+ instruction.Memory = 0;
+ instruction.Padding = 0;
return instruction;
}
return instruction_texture;
}
+static struct tgsi_instruction_memory
+tgsi_default_instruction_memory( void )
+{
+ struct tgsi_instruction_memory instruction_memory;
+
+ instruction_memory.Qualifier = 0;
+ instruction_memory.Padding = 0;
+
+ return instruction_memory;
+}
+
+static struct tgsi_instruction_memory
+tgsi_build_instruction_memory(
+ unsigned qualifier,
+ struct tgsi_token *prev_token,
+ struct tgsi_instruction *instruction,
+ struct tgsi_header *header )
+{
+ struct tgsi_instruction_memory instruction_memory;
+
+ instruction_memory.Qualifier = qualifier;
+ instruction_memory.Padding = 0;
+ instruction->Memory = 1;
+
+ instruction_grow( instruction, header );
+
+ return instruction_memory;
+}
static struct tgsi_texture_offset
tgsi_default_texture_offset( void )
full_instruction.Predicate = tgsi_default_instruction_predicate();
full_instruction.Label = tgsi_default_instruction_label();
full_instruction.Texture = tgsi_default_instruction_texture();
+ full_instruction.Memory = tgsi_default_instruction_memory();
for( i = 0; i < TGSI_FULL_MAX_TEX_OFFSETS; i++ ) {
full_instruction.TexOffsets[i] = tgsi_default_texture_offset();
}
prev_token = (struct tgsi_token *) texture_offset;
}
}
+
+ if (full_inst->Instruction.Memory) {
+ struct tgsi_instruction_memory *instruction_memory;
+
+ if( maxsize <= size )
+ return 0;
+ instruction_memory =
+ (struct tgsi_instruction_memory *) &tokens[size];
+ size++;
+
+ *instruction_memory = tgsi_build_instruction_memory(
+ full_inst->Memory.Qualifier,
+ prev_token,
+ instruction,
+ header );
+ prev_token = (struct tgsi_token *) instruction_memory;
+ }
+
for( i = 0; i < full_inst->Instruction.NumDstRegs; i++ ) {
const struct tgsi_full_dst_register *reg = &full_inst->Dst[i];
struct tgsi_dst_register *dst_register;
}
}
- if (decl->Declaration.File == TGSI_FILE_RESOURCE) {
+ if (decl->Declaration.File == TGSI_FILE_IMAGE) {
TXT(", ");
- ENM(decl->Resource.Resource, tgsi_texture_names);
- if (decl->Resource.Writable)
+ ENM(decl->Image.Resource, tgsi_texture_names);
+ TXT(", ");
+ UID(decl->Image.Format);
+ if (decl->Image.Writable)
TXT(", WR");
- if (decl->Resource.Raw)
+ if (decl->Image.Raw)
TXT(", RAW");
}
+ if (decl->Declaration.File == TGSI_FILE_BUFFER) {
+ if (decl->Declaration.Atomic)
+ TXT(", ATOMIC");
+ }
+
if (decl->Declaration.File == TGSI_FILE_SAMPLER_VIEW) {
TXT(", ");
ENM(decl->SamplerView.Resource, tgsi_texture_names);
}
}
+ if (inst->Instruction.Memory) {
+ uint32_t qualifier = inst->Memory.Qualifier;
+ while (qualifier) {
+ int bit = ffs(qualifier) - 1;
+ qualifier &= ~(1U << bit);
+ TXT(", ");
+ ENM(bit, tgsi_memory_names);
+ }
+ }
+
switch (inst->Instruction.Opcode) {
case TGSI_OPCODE_IF:
case TGSI_OPCODE_UIF:
return 1;
case PIPE_SHADER_CAP_TGSI_DROUND_SUPPORTED:
case PIPE_SHADER_CAP_TGSI_FMA_SUPPORTED:
+ case PIPE_SHADER_CAP_MAX_SHADER_BUFFERS:
return 0;
case PIPE_SHADER_CAP_MAX_UNROLL_ITERATIONS_HINT:
return 32;
static const struct tgsi_opcode_info opcode_info[TGSI_OPCODE_LAST] =
{
- { 1, 1, 0, 0, 0, 0, COMP, "ARL", TGSI_OPCODE_ARL },
- { 1, 1, 0, 0, 0, 0, COMP, "MOV", TGSI_OPCODE_MOV },
- { 1, 1, 0, 0, 0, 0, CHAN, "LIT", TGSI_OPCODE_LIT },
- { 1, 1, 0, 0, 0, 0, REPL, "RCP", TGSI_OPCODE_RCP },
- { 1, 1, 0, 0, 0, 0, REPL, "RSQ", TGSI_OPCODE_RSQ },
- { 1, 1, 0, 0, 0, 0, CHAN, "EXP", TGSI_OPCODE_EXP },
- { 1, 1, 0, 0, 0, 0, CHAN, "LOG", TGSI_OPCODE_LOG },
- { 1, 2, 0, 0, 0, 0, COMP, "MUL", TGSI_OPCODE_MUL },
- { 1, 2, 0, 0, 0, 0, COMP, "ADD", TGSI_OPCODE_ADD },
- { 1, 2, 0, 0, 0, 0, REPL, "DP3", TGSI_OPCODE_DP3 },
- { 1, 2, 0, 0, 0, 0, REPL, "DP4", TGSI_OPCODE_DP4 },
- { 1, 2, 0, 0, 0, 0, CHAN, "DST", TGSI_OPCODE_DST },
- { 1, 2, 0, 0, 0, 0, COMP, "MIN", TGSI_OPCODE_MIN },
- { 1, 2, 0, 0, 0, 0, COMP, "MAX", TGSI_OPCODE_MAX },
- { 1, 2, 0, 0, 0, 0, COMP, "SLT", TGSI_OPCODE_SLT },
- { 1, 2, 0, 0, 0, 0, COMP, "SGE", TGSI_OPCODE_SGE },
- { 1, 3, 0, 0, 0, 0, COMP, "MAD", TGSI_OPCODE_MAD },
- { 1, 2, 0, 0, 0, 0, COMP, "SUB", TGSI_OPCODE_SUB },
- { 1, 3, 0, 0, 0, 0, COMP, "LRP", TGSI_OPCODE_LRP },
- { 1, 3, 0, 0, 0, 0, COMP, "FMA", TGSI_OPCODE_FMA },
- { 1, 1, 0, 0, 0, 0, REPL, "SQRT", TGSI_OPCODE_SQRT },
- { 1, 3, 0, 0, 0, 0, REPL, "DP2A", TGSI_OPCODE_DP2A },
- { 0, 0, 0, 0, 0, 0, NONE, "", 22 }, /* removed */
- { 0, 0, 0, 0, 0, 0, NONE, "", 23 }, /* removed */
- { 1, 1, 0, 0, 0, 0, COMP, "FRC", TGSI_OPCODE_FRC },
- { 1, 3, 0, 0, 0, 0, COMP, "CLAMP", TGSI_OPCODE_CLAMP },
- { 1, 1, 0, 0, 0, 0, COMP, "FLR", TGSI_OPCODE_FLR },
- { 1, 1, 0, 0, 0, 0, COMP, "ROUND", TGSI_OPCODE_ROUND },
- { 1, 1, 0, 0, 0, 0, REPL, "EX2", TGSI_OPCODE_EX2 },
- { 1, 1, 0, 0, 0, 0, REPL, "LG2", TGSI_OPCODE_LG2 },
- { 1, 2, 0, 0, 0, 0, REPL, "POW", TGSI_OPCODE_POW },
- { 1, 2, 0, 0, 0, 0, COMP, "XPD", TGSI_OPCODE_XPD },
- { 0, 0, 0, 0, 0, 0, NONE, "", 32 }, /* removed */
- { 1, 1, 0, 0, 0, 0, COMP, "ABS", TGSI_OPCODE_ABS },
- { 0, 0, 0, 0, 0, 0, NONE, "", 34 }, /* removed */
- { 1, 2, 0, 0, 0, 0, REPL, "DPH", TGSI_OPCODE_DPH },
- { 1, 1, 0, 0, 0, 0, REPL, "COS", TGSI_OPCODE_COS },
- { 1, 1, 0, 0, 0, 0, COMP, "DDX", TGSI_OPCODE_DDX },
- { 1, 1, 0, 0, 0, 0, COMP, "DDY", TGSI_OPCODE_DDY },
- { 0, 0, 0, 0, 0, 0, NONE, "KILL", TGSI_OPCODE_KILL },
- { 1, 1, 0, 0, 0, 0, COMP, "PK2H", TGSI_OPCODE_PK2H },
- { 1, 1, 0, 0, 0, 0, COMP, "PK2US", TGSI_OPCODE_PK2US },
- { 1, 1, 0, 0, 0, 0, COMP, "PK4B", TGSI_OPCODE_PK4B },
- { 1, 1, 0, 0, 0, 0, COMP, "PK4UB", TGSI_OPCODE_PK4UB },
- { 0, 1, 0, 0, 0, 1, NONE, "", 44 }, /* removed */
- { 1, 2, 0, 0, 0, 0, COMP, "SEQ", TGSI_OPCODE_SEQ },
- { 0, 1, 0, 0, 0, 1, NONE, "", 46 }, /* removed */
- { 1, 2, 0, 0, 0, 0, COMP, "SGT", TGSI_OPCODE_SGT },
- { 1, 1, 0, 0, 0, 0, REPL, "SIN", TGSI_OPCODE_SIN },
- { 1, 2, 0, 0, 0, 0, COMP, "SLE", TGSI_OPCODE_SLE },
- { 1, 2, 0, 0, 0, 0, COMP, "SNE", TGSI_OPCODE_SNE },
- { 0, 1, 0, 0, 0, 1, NONE, "", 51 }, /* removed */
- { 1, 2, 1, 0, 0, 0, OTHR, "TEX", TGSI_OPCODE_TEX },
- { 1, 4, 1, 0, 0, 0, OTHR, "TXD", TGSI_OPCODE_TXD },
- { 1, 2, 1, 0, 0, 0, OTHR, "TXP", TGSI_OPCODE_TXP },
- { 1, 1, 0, 0, 0, 0, COMP, "UP2H", TGSI_OPCODE_UP2H },
- { 1, 1, 0, 0, 0, 0, COMP, "UP2US", TGSI_OPCODE_UP2US },
- { 1, 1, 0, 0, 0, 0, COMP, "UP4B", TGSI_OPCODE_UP4B },
- { 1, 1, 0, 0, 0, 0, COMP, "UP4UB", TGSI_OPCODE_UP4UB },
- { 0, 1, 0, 0, 0, 1, NONE, "", 59 }, /* removed */
- { 0, 1, 0, 0, 0, 1, NONE, "", 60 }, /* removed */
- { 1, 1, 0, 0, 0, 0, COMP, "ARR", TGSI_OPCODE_ARR },
- { 0, 1, 0, 0, 0, 1, NONE, "", 62 }, /* removed */
- { 0, 0, 0, 1, 0, 0, NONE, "CAL", TGSI_OPCODE_CAL },
- { 0, 0, 0, 0, 0, 0, NONE, "RET", TGSI_OPCODE_RET },
- { 1, 1, 0, 0, 0, 0, COMP, "SSG", TGSI_OPCODE_SSG },
- { 1, 3, 0, 0, 0, 0, COMP, "CMP", TGSI_OPCODE_CMP },
- { 1, 1, 0, 0, 0, 0, CHAN, "SCS", TGSI_OPCODE_SCS },
- { 1, 2, 1, 0, 0, 0, OTHR, "TXB", TGSI_OPCODE_TXB },
- { 0, 1, 0, 0, 0, 1, NONE, "", 69 }, /* removed */
- { 1, 2, 0, 0, 0, 0, COMP, "DIV", TGSI_OPCODE_DIV },
- { 1, 2, 0, 0, 0, 0, REPL, "DP2", TGSI_OPCODE_DP2 },
- { 1, 2, 1, 0, 0, 0, OTHR, "TXL", TGSI_OPCODE_TXL },
- { 0, 0, 0, 0, 0, 0, NONE, "BRK", TGSI_OPCODE_BRK },
- { 0, 1, 0, 1, 0, 1, NONE, "IF", TGSI_OPCODE_IF },
- { 0, 1, 0, 1, 0, 1, NONE, "UIF", TGSI_OPCODE_UIF },
- { 0, 1, 0, 0, 0, 1, NONE, "", 76 }, /* removed */
- { 0, 0, 0, 1, 1, 1, NONE, "ELSE", TGSI_OPCODE_ELSE },
- { 0, 0, 0, 0, 1, 0, NONE, "ENDIF", TGSI_OPCODE_ENDIF },
- { 1, 1, 0, 0, 0, 0, COMP, "DDX_FINE", TGSI_OPCODE_DDX_FINE },
- { 1, 1, 0, 0, 0, 0, COMP, "DDY_FINE", TGSI_OPCODE_DDY_FINE },
- { 0, 1, 0, 0, 0, 0, NONE, "PUSHA", TGSI_OPCODE_PUSHA },
- { 1, 0, 0, 0, 0, 0, NONE, "POPA", TGSI_OPCODE_POPA },
- { 1, 1, 0, 0, 0, 0, COMP, "CEIL", TGSI_OPCODE_CEIL },
- { 1, 1, 0, 0, 0, 0, COMP, "I2F", TGSI_OPCODE_I2F },
- { 1, 1, 0, 0, 0, 0, COMP, "NOT", TGSI_OPCODE_NOT },
- { 1, 1, 0, 0, 0, 0, COMP, "TRUNC", TGSI_OPCODE_TRUNC },
- { 1, 2, 0, 0, 0, 0, COMP, "SHL", TGSI_OPCODE_SHL },
- { 0, 0, 0, 0, 0, 0, NONE, "", 88 }, /* removed */
- { 1, 2, 0, 0, 0, 0, COMP, "AND", TGSI_OPCODE_AND },
- { 1, 2, 0, 0, 0, 0, COMP, "OR", TGSI_OPCODE_OR },
- { 1, 2, 0, 0, 0, 0, COMP, "MOD", TGSI_OPCODE_MOD },
- { 1, 2, 0, 0, 0, 0, COMP, "XOR", TGSI_OPCODE_XOR },
- { 1, 3, 0, 0, 0, 0, COMP, "SAD", TGSI_OPCODE_SAD },
- { 1, 2, 1, 0, 0, 0, OTHR, "TXF", TGSI_OPCODE_TXF },
- { 1, 2, 1, 0, 0, 0, OTHR, "TXQ", TGSI_OPCODE_TXQ },
- { 0, 0, 0, 0, 0, 0, NONE, "CONT", TGSI_OPCODE_CONT },
- { 0, 1, 0, 0, 0, 0, NONE, "EMIT", TGSI_OPCODE_EMIT },
- { 0, 1, 0, 0, 0, 0, NONE, "ENDPRIM", TGSI_OPCODE_ENDPRIM },
- { 0, 0, 0, 1, 0, 1, NONE, "BGNLOOP", TGSI_OPCODE_BGNLOOP },
- { 0, 0, 0, 0, 0, 1, NONE, "BGNSUB", TGSI_OPCODE_BGNSUB },
- { 0, 0, 0, 1, 1, 0, NONE, "ENDLOOP", TGSI_OPCODE_ENDLOOP },
- { 0, 0, 0, 0, 1, 0, NONE, "ENDSUB", TGSI_OPCODE_ENDSUB },
- { 1, 1, 1, 0, 0, 0, OTHR, "TXQ_LZ", TGSI_OPCODE_TXQ_LZ },
- { 1, 1, 1, 0, 0, 0, OTHR, "TXQS", TGSI_OPCODE_TXQS },
- { 0, 0, 0, 0, 0, 0, NONE, "", 105 }, /* removed */
- { 0, 0, 0, 0, 0, 0, NONE, "", 106 }, /* removed */
- { 0, 0, 0, 0, 0, 0, NONE, "NOP", TGSI_OPCODE_NOP },
- { 1, 2, 0, 0, 0, 0, COMP, "FSEQ", TGSI_OPCODE_FSEQ },
- { 1, 2, 0, 0, 0, 0, COMP, "FSGE", TGSI_OPCODE_FSGE },
- { 1, 2, 0, 0, 0, 0, COMP, "FSLT", TGSI_OPCODE_FSLT },
- { 1, 2, 0, 0, 0, 0, COMP, "FSNE", TGSI_OPCODE_FSNE },
- { 0, 1, 0, 0, 0, 1, NONE, "", 112 }, /* removed */
- { 0, 1, 0, 0, 0, 0, NONE, "CALLNZ", TGSI_OPCODE_CALLNZ },
- { 0, 1, 0, 0, 0, 0, NONE, "", 114 }, /* removed */
- { 0, 1, 0, 0, 0, 0, NONE, "BREAKC", TGSI_OPCODE_BREAKC },
- { 0, 1, 0, 0, 0, 0, NONE, "KILL_IF", TGSI_OPCODE_KILL_IF },
- { 0, 0, 0, 0, 0, 0, NONE, "END", TGSI_OPCODE_END },
- { 1, 3, 0, 0, 0, 0, COMP, "DFMA", TGSI_OPCODE_DFMA },
- { 1, 1, 0, 0, 0, 0, COMP, "F2I", TGSI_OPCODE_F2I },
- { 1, 2, 0, 0, 0, 0, COMP, "IDIV", TGSI_OPCODE_IDIV },
- { 1, 2, 0, 0, 0, 0, COMP, "IMAX", TGSI_OPCODE_IMAX },
- { 1, 2, 0, 0, 0, 0, COMP, "IMIN", TGSI_OPCODE_IMIN },
- { 1, 1, 0, 0, 0, 0, COMP, "INEG", TGSI_OPCODE_INEG },
- { 1, 2, 0, 0, 0, 0, COMP, "ISGE", TGSI_OPCODE_ISGE },
- { 1, 2, 0, 0, 0, 0, COMP, "ISHR", TGSI_OPCODE_ISHR },
- { 1, 2, 0, 0, 0, 0, COMP, "ISLT", TGSI_OPCODE_ISLT },
- { 1, 1, 0, 0, 0, 0, COMP, "F2U", TGSI_OPCODE_F2U },
- { 1, 1, 0, 0, 0, 0, COMP, "U2F", TGSI_OPCODE_U2F },
- { 1, 2, 0, 0, 0, 0, COMP, "UADD", TGSI_OPCODE_UADD },
- { 1, 2, 0, 0, 0, 0, COMP, "UDIV", TGSI_OPCODE_UDIV },
- { 1, 3, 0, 0, 0, 0, COMP, "UMAD", TGSI_OPCODE_UMAD },
- { 1, 2, 0, 0, 0, 0, COMP, "UMAX", TGSI_OPCODE_UMAX },
- { 1, 2, 0, 0, 0, 0, COMP, "UMIN", TGSI_OPCODE_UMIN },
- { 1, 2, 0, 0, 0, 0, COMP, "UMOD", TGSI_OPCODE_UMOD },
- { 1, 2, 0, 0, 0, 0, COMP, "UMUL", TGSI_OPCODE_UMUL },
- { 1, 2, 0, 0, 0, 0, COMP, "USEQ", TGSI_OPCODE_USEQ },
- { 1, 2, 0, 0, 0, 0, COMP, "USGE", TGSI_OPCODE_USGE },
- { 1, 2, 0, 0, 0, 0, COMP, "USHR", TGSI_OPCODE_USHR },
- { 1, 2, 0, 0, 0, 0, COMP, "USLT", TGSI_OPCODE_USLT },
- { 1, 2, 0, 0, 0, 0, COMP, "USNE", TGSI_OPCODE_USNE },
- { 0, 1, 0, 0, 0, 0, NONE, "SWITCH", TGSI_OPCODE_SWITCH },
- { 0, 1, 0, 0, 0, 0, NONE, "CASE", TGSI_OPCODE_CASE },
- { 0, 0, 0, 0, 0, 0, NONE, "DEFAULT", TGSI_OPCODE_DEFAULT },
- { 0, 0, 0, 0, 0, 0, NONE, "ENDSWITCH", TGSI_OPCODE_ENDSWITCH },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "ARL", TGSI_OPCODE_ARL },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "MOV", TGSI_OPCODE_MOV },
+ { 1, 1, 0, 0, 0, 0, 0, CHAN, "LIT", TGSI_OPCODE_LIT },
+ { 1, 1, 0, 0, 0, 0, 0, REPL, "RCP", TGSI_OPCODE_RCP },
+ { 1, 1, 0, 0, 0, 0, 0, REPL, "RSQ", TGSI_OPCODE_RSQ },
+ { 1, 1, 0, 0, 0, 0, 0, CHAN, "EXP", TGSI_OPCODE_EXP },
+ { 1, 1, 0, 0, 0, 0, 0, CHAN, "LOG", TGSI_OPCODE_LOG },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "MUL", TGSI_OPCODE_MUL },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "ADD", TGSI_OPCODE_ADD },
+ { 1, 2, 0, 0, 0, 0, 0, REPL, "DP3", TGSI_OPCODE_DP3 },
+ { 1, 2, 0, 0, 0, 0, 0, REPL, "DP4", TGSI_OPCODE_DP4 },
+ { 1, 2, 0, 0, 0, 0, 0, CHAN, "DST", TGSI_OPCODE_DST },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "MIN", TGSI_OPCODE_MIN },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "MAX", TGSI_OPCODE_MAX },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "SLT", TGSI_OPCODE_SLT },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "SGE", TGSI_OPCODE_SGE },
+ { 1, 3, 0, 0, 0, 0, 0, COMP, "MAD", TGSI_OPCODE_MAD },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "SUB", TGSI_OPCODE_SUB },
+ { 1, 3, 0, 0, 0, 0, 0, COMP, "LRP", TGSI_OPCODE_LRP },
+ { 1, 3, 0, 0, 0, 0, 0, COMP, "FMA", TGSI_OPCODE_FMA },
+ { 1, 1, 0, 0, 0, 0, 0, REPL, "SQRT", TGSI_OPCODE_SQRT },
+ { 1, 3, 0, 0, 0, 0, 0, REPL, "DP2A", TGSI_OPCODE_DP2A },
+ { 0, 0, 0, 0, 0, 0, 0, NONE, "", 22 }, /* removed */
+ { 0, 0, 0, 0, 0, 0, 0, NONE, "", 23 }, /* removed */
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "FRC", TGSI_OPCODE_FRC },
+ { 1, 3, 0, 0, 0, 0, 0, COMP, "CLAMP", TGSI_OPCODE_CLAMP },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "FLR", TGSI_OPCODE_FLR },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "ROUND", TGSI_OPCODE_ROUND },
+ { 1, 1, 0, 0, 0, 0, 0, REPL, "EX2", TGSI_OPCODE_EX2 },
+ { 1, 1, 0, 0, 0, 0, 0, REPL, "LG2", TGSI_OPCODE_LG2 },
+ { 1, 2, 0, 0, 0, 0, 0, REPL, "POW", TGSI_OPCODE_POW },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "XPD", TGSI_OPCODE_XPD },
+ { 0, 0, 0, 0, 0, 0, 0, NONE, "", 32 }, /* removed */
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "ABS", TGSI_OPCODE_ABS },
+ { 0, 0, 0, 0, 0, 0, 0, NONE, "", 34 }, /* removed */
+ { 1, 2, 0, 0, 0, 0, 0, REPL, "DPH", TGSI_OPCODE_DPH },
+ { 1, 1, 0, 0, 0, 0, 0, REPL, "COS", TGSI_OPCODE_COS },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "DDX", TGSI_OPCODE_DDX },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "DDY", TGSI_OPCODE_DDY },
+ { 0, 0, 0, 0, 0, 0, 0, NONE, "KILL", TGSI_OPCODE_KILL },
+ { 1, 1, 0, 0, 0, 0, 0, REPL, "PK2H", TGSI_OPCODE_PK2H },
+ { 1, 1, 0, 0, 0, 0, 0, REPL, "PK2US", TGSI_OPCODE_PK2US },
+ { 1, 1, 0, 0, 0, 0, 0, REPL, "PK4B", TGSI_OPCODE_PK4B },
+ { 1, 1, 0, 0, 0, 0, 0, REPL, "PK4UB", TGSI_OPCODE_PK4UB },
+ { 0, 1, 0, 0, 0, 0, 1, NONE, "", 44 }, /* removed */
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "SEQ", TGSI_OPCODE_SEQ },
+ { 0, 1, 0, 0, 0, 0, 1, NONE, "", 46 }, /* removed */
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "SGT", TGSI_OPCODE_SGT },
+ { 1, 1, 0, 0, 0, 0, 0, REPL, "SIN", TGSI_OPCODE_SIN },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "SLE", TGSI_OPCODE_SLE },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "SNE", TGSI_OPCODE_SNE },
+ { 0, 1, 0, 0, 0, 0, 1, NONE, "", 51 }, /* removed */
+ { 1, 2, 1, 0, 0, 0, 0, OTHR, "TEX", TGSI_OPCODE_TEX },
+ { 1, 4, 1, 0, 0, 0, 0, OTHR, "TXD", TGSI_OPCODE_TXD },
+ { 1, 2, 1, 0, 0, 0, 0, OTHR, "TXP", TGSI_OPCODE_TXP },
+ { 1, 1, 0, 0, 0, 0, 0, CHAN, "UP2H", TGSI_OPCODE_UP2H },
+ { 1, 1, 0, 0, 0, 0, 0, CHAN, "UP2US", TGSI_OPCODE_UP2US },
+ { 1, 1, 0, 0, 0, 0, 0, CHAN, "UP4B", TGSI_OPCODE_UP4B },
+ { 1, 1, 0, 0, 0, 0, 0, CHAN, "UP4UB", TGSI_OPCODE_UP4UB },
+ { 0, 1, 0, 0, 0, 0, 1, NONE, "", 59 }, /* removed */
+ { 0, 1, 0, 0, 0, 0, 1, NONE, "", 60 }, /* removed */
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "ARR", TGSI_OPCODE_ARR },
+ { 0, 1, 0, 0, 0, 0, 1, NONE, "", 62 }, /* removed */
+ { 0, 0, 0, 0, 1, 0, 0, NONE, "CAL", TGSI_OPCODE_CAL },
+ { 0, 0, 0, 0, 0, 0, 0, NONE, "RET", TGSI_OPCODE_RET },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "SSG", TGSI_OPCODE_SSG },
+ { 1, 3, 0, 0, 0, 0, 0, COMP, "CMP", TGSI_OPCODE_CMP },
+ { 1, 1, 0, 0, 0, 0, 0, CHAN, "SCS", TGSI_OPCODE_SCS },
+ { 1, 2, 1, 0, 0, 0, 0, OTHR, "TXB", TGSI_OPCODE_TXB },
+ { 0, 1, 0, 0, 0, 0, 1, NONE, "", 69 }, /* removed */
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "DIV", TGSI_OPCODE_DIV },
+ { 1, 2, 0, 0, 0, 0, 0, REPL, "DP2", TGSI_OPCODE_DP2 },
+ { 1, 2, 1, 0, 0, 0, 0, OTHR, "TXL", TGSI_OPCODE_TXL },
+ { 0, 0, 0, 0, 0, 0, 0, NONE, "BRK", TGSI_OPCODE_BRK },
+ { 0, 1, 0, 0, 1, 0, 1, NONE, "IF", TGSI_OPCODE_IF },
+ { 0, 1, 0, 0, 1, 0, 1, NONE, "UIF", TGSI_OPCODE_UIF },
+ { 0, 1, 0, 0, 0, 0, 1, NONE, "", 76 }, /* removed */
+ { 0, 0, 0, 0, 1, 1, 1, NONE, "ELSE", TGSI_OPCODE_ELSE },
+ { 0, 0, 0, 0, 0, 1, 0, NONE, "ENDIF", TGSI_OPCODE_ENDIF },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "DDX_FINE", TGSI_OPCODE_DDX_FINE },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "DDY_FINE", TGSI_OPCODE_DDY_FINE },
+ { 0, 1, 0, 0, 0, 0, 0, NONE, "PUSHA", TGSI_OPCODE_PUSHA },
+ { 1, 0, 0, 0, 0, 0, 0, NONE, "POPA", TGSI_OPCODE_POPA },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "CEIL", TGSI_OPCODE_CEIL },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "I2F", TGSI_OPCODE_I2F },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "NOT", TGSI_OPCODE_NOT },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "TRUNC", TGSI_OPCODE_TRUNC },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "SHL", TGSI_OPCODE_SHL },
+ { 0, 0, 0, 0, 0, 0, 0, NONE, "", 88 }, /* removed */
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "AND", TGSI_OPCODE_AND },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "OR", TGSI_OPCODE_OR },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "MOD", TGSI_OPCODE_MOD },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "XOR", TGSI_OPCODE_XOR },
+ { 1, 3, 0, 0, 0, 0, 0, COMP, "SAD", TGSI_OPCODE_SAD },
+ { 1, 2, 1, 0, 0, 0, 0, OTHR, "TXF", TGSI_OPCODE_TXF },
+ { 1, 2, 1, 0, 0, 0, 0, OTHR, "TXQ", TGSI_OPCODE_TXQ },
+ { 0, 0, 0, 0, 0, 0, 0, NONE, "CONT", TGSI_OPCODE_CONT },
+ { 0, 1, 0, 0, 0, 0, 0, NONE, "EMIT", TGSI_OPCODE_EMIT },
+ { 0, 1, 0, 0, 0, 0, 0, NONE, "ENDPRIM", TGSI_OPCODE_ENDPRIM },
+ { 0, 0, 0, 0, 1, 0, 1, NONE, "BGNLOOP", TGSI_OPCODE_BGNLOOP },
+ { 0, 0, 0, 0, 0, 0, 1, NONE, "BGNSUB", TGSI_OPCODE_BGNSUB },
+ { 0, 0, 0, 0, 1, 1, 0, NONE, "ENDLOOP", TGSI_OPCODE_ENDLOOP },
+ { 0, 0, 0, 0, 0, 1, 0, NONE, "ENDSUB", TGSI_OPCODE_ENDSUB },
+ { 1, 1, 1, 0, 0, 0, 0, OTHR, "TXQ_LZ", TGSI_OPCODE_TXQ_LZ },
+ { 1, 1, 1, 0, 0, 0, 0, OTHR, "TXQS", TGSI_OPCODE_TXQS },
+ { 1, 1, 0, 0, 0, 0, 0, NONE, "RESQ", TGSI_OPCODE_RESQ },
+ { 0, 0, 0, 0, 0, 0, 0, NONE, "", 106 }, /* removed */
+ { 0, 0, 0, 0, 0, 0, 0, NONE, "NOP", TGSI_OPCODE_NOP },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "FSEQ", TGSI_OPCODE_FSEQ },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "FSGE", TGSI_OPCODE_FSGE },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "FSLT", TGSI_OPCODE_FSLT },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "FSNE", TGSI_OPCODE_FSNE },
+ { 0, 1, 0, 0, 0, 0, 1, NONE, "", 112 }, /* removed */
+ { 0, 1, 0, 0, 0, 0, 0, NONE, "CALLNZ", TGSI_OPCODE_CALLNZ },
+ { 0, 1, 0, 0, 0, 0, 0, NONE, "", 114 }, /* removed */
+ { 0, 1, 0, 0, 0, 0, 0, NONE, "BREAKC", TGSI_OPCODE_BREAKC },
+ { 0, 1, 0, 0, 0, 0, 0, NONE, "KILL_IF", TGSI_OPCODE_KILL_IF },
+ { 0, 0, 0, 0, 0, 0, 0, NONE, "END", TGSI_OPCODE_END },
+ { 1, 3, 0, 0, 0, 0, 0, COMP, "DFMA", TGSI_OPCODE_DFMA },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "F2I", TGSI_OPCODE_F2I },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "IDIV", TGSI_OPCODE_IDIV },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "IMAX", TGSI_OPCODE_IMAX },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "IMIN", TGSI_OPCODE_IMIN },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "INEG", TGSI_OPCODE_INEG },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "ISGE", TGSI_OPCODE_ISGE },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "ISHR", TGSI_OPCODE_ISHR },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "ISLT", TGSI_OPCODE_ISLT },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "F2U", TGSI_OPCODE_F2U },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "U2F", TGSI_OPCODE_U2F },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "UADD", TGSI_OPCODE_UADD },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "UDIV", TGSI_OPCODE_UDIV },
+ { 1, 3, 0, 0, 0, 0, 0, COMP, "UMAD", TGSI_OPCODE_UMAD },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "UMAX", TGSI_OPCODE_UMAX },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "UMIN", TGSI_OPCODE_UMIN },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "UMOD", TGSI_OPCODE_UMOD },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "UMUL", TGSI_OPCODE_UMUL },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "USEQ", TGSI_OPCODE_USEQ },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "USGE", TGSI_OPCODE_USGE },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "USHR", TGSI_OPCODE_USHR },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "USLT", TGSI_OPCODE_USLT },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "USNE", TGSI_OPCODE_USNE },
+ { 0, 1, 0, 0, 0, 0, 0, NONE, "SWITCH", TGSI_OPCODE_SWITCH },
+ { 0, 1, 0, 0, 0, 0, 0, NONE, "CASE", TGSI_OPCODE_CASE },
+ { 0, 0, 0, 0, 0, 0, 0, NONE, "DEFAULT", TGSI_OPCODE_DEFAULT },
+ { 0, 0, 0, 0, 0, 0, 0, NONE, "ENDSWITCH", TGSI_OPCODE_ENDSWITCH },
- { 1, 3, 0, 0, 0, 0, OTHR, "SAMPLE", TGSI_OPCODE_SAMPLE },
- { 1, 2, 0, 0, 0, 0, OTHR, "SAMPLE_I", TGSI_OPCODE_SAMPLE_I },
- { 1, 3, 0, 0, 0, 0, OTHR, "SAMPLE_I_MS", TGSI_OPCODE_SAMPLE_I_MS },
- { 1, 4, 0, 0, 0, 0, OTHR, "SAMPLE_B", TGSI_OPCODE_SAMPLE_B },
- { 1, 4, 0, 0, 0, 0, OTHR, "SAMPLE_C", TGSI_OPCODE_SAMPLE_C },
- { 1, 4, 0, 0, 0, 0, OTHR, "SAMPLE_C_LZ", TGSI_OPCODE_SAMPLE_C_LZ },
- { 1, 5, 0, 0, 0, 0, OTHR, "SAMPLE_D", TGSI_OPCODE_SAMPLE_D },
- { 1, 4, 0, 0, 0, 0, OTHR, "SAMPLE_L", TGSI_OPCODE_SAMPLE_L },
- { 1, 3, 0, 0, 0, 0, OTHR, "GATHER4", TGSI_OPCODE_GATHER4 },
- { 1, 2, 0, 0, 0, 0, OTHR, "SVIEWINFO", TGSI_OPCODE_SVIEWINFO },
- { 1, 2, 0, 0, 0, 0, OTHR, "SAMPLE_POS", TGSI_OPCODE_SAMPLE_POS },
- { 1, 2, 0, 0, 0, 0, OTHR, "SAMPLE_INFO", TGSI_OPCODE_SAMPLE_INFO },
- { 1, 1, 0, 0, 0, 0, COMP, "UARL", TGSI_OPCODE_UARL },
- { 1, 3, 0, 0, 0, 0, COMP, "UCMP", TGSI_OPCODE_UCMP },
- { 1, 1, 0, 0, 0, 0, COMP, "IABS", TGSI_OPCODE_IABS },
- { 1, 1, 0, 0, 0, 0, COMP, "ISSG", TGSI_OPCODE_ISSG },
- { 1, 2, 0, 0, 0, 0, OTHR, "LOAD", TGSI_OPCODE_LOAD },
- { 1, 2, 0, 0, 0, 0, OTHR, "STORE", TGSI_OPCODE_STORE },
- { 1, 0, 0, 0, 0, 0, OTHR, "MFENCE", TGSI_OPCODE_MFENCE },
- { 1, 0, 0, 0, 0, 0, OTHR, "LFENCE", TGSI_OPCODE_LFENCE },
- { 1, 0, 0, 0, 0, 0, OTHR, "SFENCE", TGSI_OPCODE_SFENCE },
- { 0, 0, 0, 0, 0, 0, OTHR, "BARRIER", TGSI_OPCODE_BARRIER },
+ { 1, 3, 0, 0, 0, 0, 0, OTHR, "SAMPLE", TGSI_OPCODE_SAMPLE },
+ { 1, 2, 0, 0, 0, 0, 0, OTHR, "SAMPLE_I", TGSI_OPCODE_SAMPLE_I },
+ { 1, 3, 0, 0, 0, 0, 0, OTHR, "SAMPLE_I_MS", TGSI_OPCODE_SAMPLE_I_MS },
+ { 1, 4, 0, 0, 0, 0, 0, OTHR, "SAMPLE_B", TGSI_OPCODE_SAMPLE_B },
+ { 1, 4, 0, 0, 0, 0, 0, OTHR, "SAMPLE_C", TGSI_OPCODE_SAMPLE_C },
+ { 1, 4, 0, 0, 0, 0, 0, OTHR, "SAMPLE_C_LZ", TGSI_OPCODE_SAMPLE_C_LZ },
+ { 1, 5, 0, 0, 0, 0, 0, OTHR, "SAMPLE_D", TGSI_OPCODE_SAMPLE_D },
+ { 1, 4, 0, 0, 0, 0, 0, OTHR, "SAMPLE_L", TGSI_OPCODE_SAMPLE_L },
+ { 1, 3, 0, 0, 0, 0, 0, OTHR, "GATHER4", TGSI_OPCODE_GATHER4 },
+ { 1, 2, 0, 0, 0, 0, 0, OTHR, "SVIEWINFO", TGSI_OPCODE_SVIEWINFO },
+ { 1, 2, 0, 0, 0, 0, 0, OTHR, "SAMPLE_POS", TGSI_OPCODE_SAMPLE_POS },
+ { 1, 2, 0, 0, 0, 0, 0, OTHR, "SAMPLE_INFO", TGSI_OPCODE_SAMPLE_INFO },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "UARL", TGSI_OPCODE_UARL },
+ { 1, 3, 0, 0, 0, 0, 0, COMP, "UCMP", TGSI_OPCODE_UCMP },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "IABS", TGSI_OPCODE_IABS },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "ISSG", TGSI_OPCODE_ISSG },
+ { 1, 2, 0, 0, 0, 0, 0, OTHR, "LOAD", TGSI_OPCODE_LOAD },
+ { 1, 2, 0, 1, 0, 0, 0, OTHR, "STORE", TGSI_OPCODE_STORE },
+ { 1, 0, 0, 0, 0, 0, 0, OTHR, "MFENCE", TGSI_OPCODE_MFENCE },
+ { 1, 0, 0, 0, 0, 0, 0, OTHR, "LFENCE", TGSI_OPCODE_LFENCE },
+ { 1, 0, 0, 0, 0, 0, 0, OTHR, "SFENCE", TGSI_OPCODE_SFENCE },
+ { 0, 0, 0, 0, 0, 0, 0, OTHR, "BARRIER", TGSI_OPCODE_BARRIER },
- { 1, 3, 0, 0, 0, 0, OTHR, "ATOMUADD", TGSI_OPCODE_ATOMUADD },
- { 1, 3, 0, 0, 0, 0, OTHR, "ATOMXCHG", TGSI_OPCODE_ATOMXCHG },
- { 1, 4, 0, 0, 0, 0, OTHR, "ATOMCAS", TGSI_OPCODE_ATOMCAS },
- { 1, 3, 0, 0, 0, 0, OTHR, "ATOMAND", TGSI_OPCODE_ATOMAND },
- { 1, 3, 0, 0, 0, 0, OTHR, "ATOMOR", TGSI_OPCODE_ATOMOR },
- { 1, 3, 0, 0, 0, 0, OTHR, "ATOMXOR", TGSI_OPCODE_ATOMXOR },
- { 1, 3, 0, 0, 0, 0, OTHR, "ATOMUMIN", TGSI_OPCODE_ATOMUMIN },
- { 1, 3, 0, 0, 0, 0, OTHR, "ATOMUMAX", TGSI_OPCODE_ATOMUMAX },
- { 1, 3, 0, 0, 0, 0, OTHR, "ATOMIMIN", TGSI_OPCODE_ATOMIMIN },
- { 1, 3, 0, 0, 0, 0, OTHR, "ATOMIMAX", TGSI_OPCODE_ATOMIMAX },
- { 1, 3, 1, 0, 0, 0, OTHR, "TEX2", TGSI_OPCODE_TEX2 },
- { 1, 3, 1, 0, 0, 0, OTHR, "TXB2", TGSI_OPCODE_TXB2 },
- { 1, 3, 1, 0, 0, 0, OTHR, "TXL2", TGSI_OPCODE_TXL2 },
- { 1, 2, 0, 0, 0, 0, COMP, "IMUL_HI", TGSI_OPCODE_IMUL_HI },
- { 1, 2, 0, 0, 0, 0, COMP, "UMUL_HI", TGSI_OPCODE_UMUL_HI },
- { 1, 3, 1, 0, 0, 0, OTHR, "TG4", TGSI_OPCODE_TG4 },
- { 1, 2, 1, 0, 0, 0, OTHR, "LODQ", TGSI_OPCODE_LODQ },
- { 1, 3, 0, 0, 0, 0, COMP, "IBFE", TGSI_OPCODE_IBFE },
- { 1, 3, 0, 0, 0, 0, COMP, "UBFE", TGSI_OPCODE_UBFE },
- { 1, 4, 0, 0, 0, 0, COMP, "BFI", TGSI_OPCODE_BFI },
- { 1, 1, 0, 0, 0, 0, COMP, "BREV", TGSI_OPCODE_BREV },
- { 1, 1, 0, 0, 0, 0, COMP, "POPC", TGSI_OPCODE_POPC },
- { 1, 1, 0, 0, 0, 0, COMP, "LSB", TGSI_OPCODE_LSB },
- { 1, 1, 0, 0, 0, 0, COMP, "IMSB", TGSI_OPCODE_IMSB },
- { 1, 1, 0, 0, 0, 0, COMP, "UMSB", TGSI_OPCODE_UMSB },
- { 1, 1, 0, 0, 0, 0, OTHR, "INTERP_CENTROID", TGSI_OPCODE_INTERP_CENTROID },
- { 1, 2, 0, 0, 0, 0, OTHR, "INTERP_SAMPLE", TGSI_OPCODE_INTERP_SAMPLE },
- { 1, 2, 0, 0, 0, 0, OTHR, "INTERP_OFFSET", TGSI_OPCODE_INTERP_OFFSET },
- { 1, 1, 0, 0, 0, 0, COMP, "F2D", TGSI_OPCODE_F2D },
- { 1, 1, 0, 0, 0, 0, COMP, "D2F", TGSI_OPCODE_D2F },
- { 1, 1, 0, 0, 0, 0, COMP, "DABS", TGSI_OPCODE_DABS },
- { 1, 1, 0, 0, 0, 0, COMP, "DNEG", TGSI_OPCODE_DNEG },
- { 1, 2, 0, 0, 0, 0, COMP, "DADD", TGSI_OPCODE_DADD },
- { 1, 2, 0, 0, 0, 0, COMP, "DMUL", TGSI_OPCODE_DMUL },
- { 1, 2, 0, 0, 0, 0, COMP, "DMAX", TGSI_OPCODE_DMAX },
- { 1, 2, 0, 0, 0, 0, COMP, "DMIN", TGSI_OPCODE_DMIN },
- { 1, 2, 0, 0, 0, 0, COMP, "DSLT", TGSI_OPCODE_DSLT },
- { 1, 2, 0, 0, 0, 0, COMP, "DSGE", TGSI_OPCODE_DSGE },
- { 1, 2, 0, 0, 0, 0, COMP, "DSEQ", TGSI_OPCODE_DSEQ },
- { 1, 2, 0, 0, 0, 0, COMP, "DSNE", TGSI_OPCODE_DSNE },
- { 1, 1, 0, 0, 0, 0, COMP, "DRCP", TGSI_OPCODE_DRCP },
- { 1, 1, 0, 0 ,0, 0, COMP, "DSQRT", TGSI_OPCODE_DSQRT },
- { 1, 3, 0, 0 ,0, 0, COMP, "DMAD", TGSI_OPCODE_DMAD },
- { 1, 1, 0, 0, 0, 0, COMP, "DFRAC", TGSI_OPCODE_DFRAC},
- { 1, 2, 0, 0, 0, 0, COMP, "DLDEXP", TGSI_OPCODE_DLDEXP},
- { 2, 1, 0, 0, 0, 0, COMP, "DFRACEXP", TGSI_OPCODE_DFRACEXP},
- { 1, 1, 0, 0, 0, 0, COMP, "D2I", TGSI_OPCODE_D2I },
- { 1, 1, 0, 0, 0, 0, COMP, "I2D", TGSI_OPCODE_I2D },
- { 1, 1, 0, 0, 0, 0, COMP, "D2U", TGSI_OPCODE_D2U },
- { 1, 1, 0, 0, 0, 0, COMP, "U2D", TGSI_OPCODE_U2D },
- { 1, 1, 0, 0 ,0, 0, COMP, "DRSQ", TGSI_OPCODE_DRSQ },
- { 1, 1, 0, 0, 0, 0, COMP, "DTRUNC", TGSI_OPCODE_DTRUNC },
- { 1, 1, 0, 0, 0, 0, COMP, "DCEIL", TGSI_OPCODE_DCEIL },
- { 1, 1, 0, 0, 0, 0, COMP, "DFLR", TGSI_OPCODE_DFLR },
- { 1, 1, 0, 0, 0, 0, COMP, "DROUND", TGSI_OPCODE_DROUND },
- { 1, 1, 0, 0, 0, 0, COMP, "DSSG", TGSI_OPCODE_DSSG },
+ { 1, 3, 0, 1, 0, 0, 0, OTHR, "ATOMUADD", TGSI_OPCODE_ATOMUADD },
+ { 1, 3, 0, 1, 0, 0, 0, OTHR, "ATOMXCHG", TGSI_OPCODE_ATOMXCHG },
+ { 1, 4, 0, 1, 0, 0, 0, OTHR, "ATOMCAS", TGSI_OPCODE_ATOMCAS },
+ { 1, 3, 0, 1, 0, 0, 0, OTHR, "ATOMAND", TGSI_OPCODE_ATOMAND },
+ { 1, 3, 0, 1, 0, 0, 0, OTHR, "ATOMOR", TGSI_OPCODE_ATOMOR },
+ { 1, 3, 0, 1, 0, 0, 0, OTHR, "ATOMXOR", TGSI_OPCODE_ATOMXOR },
+ { 1, 3, 0, 1, 0, 0, 0, OTHR, "ATOMUMIN", TGSI_OPCODE_ATOMUMIN },
+ { 1, 3, 0, 1, 0, 0, 0, OTHR, "ATOMUMAX", TGSI_OPCODE_ATOMUMAX },
+ { 1, 3, 0, 1, 0, 0, 0, OTHR, "ATOMIMIN", TGSI_OPCODE_ATOMIMIN },
+ { 1, 3, 0, 1, 0, 0, 0, OTHR, "ATOMIMAX", TGSI_OPCODE_ATOMIMAX },
+ { 1, 3, 1, 0, 0, 0, 0, OTHR, "TEX2", TGSI_OPCODE_TEX2 },
+ { 1, 3, 1, 0, 0, 0, 0, OTHR, "TXB2", TGSI_OPCODE_TXB2 },
+ { 1, 3, 1, 0, 0, 0, 0, OTHR, "TXL2", TGSI_OPCODE_TXL2 },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "IMUL_HI", TGSI_OPCODE_IMUL_HI },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "UMUL_HI", TGSI_OPCODE_UMUL_HI },
+ { 1, 3, 1, 0, 0, 0, 0, OTHR, "TG4", TGSI_OPCODE_TG4 },
+ { 1, 2, 1, 0, 0, 0, 0, OTHR, "LODQ", TGSI_OPCODE_LODQ },
+ { 1, 3, 0, 0, 0, 0, 0, COMP, "IBFE", TGSI_OPCODE_IBFE },
+ { 1, 3, 0, 0, 0, 0, 0, COMP, "UBFE", TGSI_OPCODE_UBFE },
+ { 1, 4, 0, 0, 0, 0, 0, COMP, "BFI", TGSI_OPCODE_BFI },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "BREV", TGSI_OPCODE_BREV },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "POPC", TGSI_OPCODE_POPC },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "LSB", TGSI_OPCODE_LSB },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "IMSB", TGSI_OPCODE_IMSB },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "UMSB", TGSI_OPCODE_UMSB },
+ { 1, 1, 0, 0, 0, 0, 0, OTHR, "INTERP_CENTROID", TGSI_OPCODE_INTERP_CENTROID },
+ { 1, 2, 0, 0, 0, 0, 0, OTHR, "INTERP_SAMPLE", TGSI_OPCODE_INTERP_SAMPLE },
+ { 1, 2, 0, 0, 0, 0, 0, OTHR, "INTERP_OFFSET", TGSI_OPCODE_INTERP_OFFSET },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "F2D", TGSI_OPCODE_F2D },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "D2F", TGSI_OPCODE_D2F },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "DABS", TGSI_OPCODE_DABS },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "DNEG", TGSI_OPCODE_DNEG },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "DADD", TGSI_OPCODE_DADD },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "DMUL", TGSI_OPCODE_DMUL },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "DMAX", TGSI_OPCODE_DMAX },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "DMIN", TGSI_OPCODE_DMIN },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "DSLT", TGSI_OPCODE_DSLT },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "DSGE", TGSI_OPCODE_DSGE },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "DSEQ", TGSI_OPCODE_DSEQ },
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "DSNE", TGSI_OPCODE_DSNE },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "DRCP", TGSI_OPCODE_DRCP },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "DSQRT", TGSI_OPCODE_DSQRT },
+ { 1, 3, 0, 0, 0, 0, 0, COMP, "DMAD", TGSI_OPCODE_DMAD },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "DFRAC", TGSI_OPCODE_DFRAC},
+ { 1, 2, 0, 0, 0, 0, 0, COMP, "DLDEXP", TGSI_OPCODE_DLDEXP},
+ { 2, 1, 0, 0, 0, 0, 0, COMP, "DFRACEXP", TGSI_OPCODE_DFRACEXP},
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "D2I", TGSI_OPCODE_D2I },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "I2D", TGSI_OPCODE_I2D },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "D2U", TGSI_OPCODE_D2U },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "U2D", TGSI_OPCODE_U2D },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "DRSQ", TGSI_OPCODE_DRSQ },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "DTRUNC", TGSI_OPCODE_DTRUNC },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "DCEIL", TGSI_OPCODE_DCEIL },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "DFLR", TGSI_OPCODE_DFLR },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "DROUND", TGSI_OPCODE_DROUND },
+ { 1, 1, 0, 0, 0, 0, 0, COMP, "DSSG", TGSI_OPCODE_DSSG },
};
const struct tgsi_opcode_info *
unsigned num_dst:3;
unsigned num_src:3;
unsigned is_tex:1;
+ unsigned is_store:1;
unsigned is_branch:1;
int pre_dedent:2;
int post_indent:2;
next_token( ctx, &decl->Semantic );
}
- if (decl->Declaration.File == TGSI_FILE_RESOURCE) {
- next_token(ctx, &decl->Resource);
+ if (decl->Declaration.File == TGSI_FILE_IMAGE) {
+ next_token(ctx, &decl->Image);
}
if (decl->Declaration.File == TGSI_FILE_SAMPLER_VIEW) {
}
}
+ if (inst->Instruction.Memory) {
+ next_token(ctx, &inst->Memory);
+ }
+
assert( inst->Instruction.NumDstRegs <= TGSI_FULL_MAX_DST_REGISTERS );
for (i = 0; i < inst->Instruction.NumDstRegs; i++) {
struct tgsi_declaration_dimension Dim;
struct tgsi_declaration_interp Interp;
struct tgsi_declaration_semantic Semantic;
- struct tgsi_declaration_resource Resource;
+ struct tgsi_declaration_image Image;
struct tgsi_declaration_sampler_view SamplerView;
struct tgsi_declaration_array Array;
};
struct tgsi_instruction_predicate Predicate;
struct tgsi_instruction_label Label;
struct tgsi_instruction_texture Texture;
+ struct tgsi_instruction_memory Memory;
struct tgsi_full_dst_register Dst[TGSI_FULL_MAX_DST_REGISTERS];
struct tgsi_full_src_register Src[TGSI_FULL_MAX_SRC_REGISTERS];
struct tgsi_texture_offset TexOffsets[TGSI_FULL_MAX_TEX_OFFSETS];
}
if (procType == TGSI_PROCESSOR_FRAGMENT &&
- info->reads_position &&
- src->Register.Index == 0 &&
- (src->Register.SwizzleX == TGSI_SWIZZLE_Z ||
- src->Register.SwizzleY == TGSI_SWIZZLE_Z ||
- src->Register.SwizzleZ == TGSI_SWIZZLE_Z ||
- src->Register.SwizzleW == TGSI_SWIZZLE_Z)) {
- info->reads_z = TRUE;
+ !src->Register.Indirect) {
+ unsigned name =
+ info->input_semantic_name[src->Register.Index];
+ unsigned index =
+ info->input_semantic_index[src->Register.Index];
+
+ if (name == TGSI_SEMANTIC_POSITION &&
+ (src->Register.SwizzleX == TGSI_SWIZZLE_Z ||
+ src->Register.SwizzleY == TGSI_SWIZZLE_Z ||
+ src->Register.SwizzleZ == TGSI_SWIZZLE_Z ||
+ src->Register.SwizzleW == TGSI_SWIZZLE_Z))
+ info->reads_z = TRUE;
+
+ if (name == TGSI_SEMANTIC_COLOR) {
+ unsigned mask =
+ (1 << src->Register.SwizzleX) |
+ (1 << src->Register.SwizzleY) |
+ (1 << src->Register.SwizzleZ) |
+ (1 << src->Register.SwizzleW);
+
+ info->colors_read |= mask << (index * 4);
+ }
}
}
info->uses_primid = TRUE;
} else if (semName == TGSI_SEMANTIC_INVOCATIONID) {
info->uses_invocationid = TRUE;
- }
+ } else if (semName == TGSI_SEMANTIC_POSITION)
+ info->reads_position = TRUE;
+ else if (semName == TGSI_SEMANTIC_FACE)
+ info->uses_frontface = TRUE;
}
else if (file == TGSI_FILE_OUTPUT) {
info->output_semantic_name[reg] = (ubyte) semName;
}
else if (semName == TGSI_SEMANTIC_STENCIL) {
info->writes_stencil = TRUE;
+ } else if (semName == TGSI_SEMANTIC_SAMPLEMASK) {
+ info->writes_samplemask = TRUE;
}
}
uint opcode_count[TGSI_OPCODE_LAST]; /**< opcode histogram */
+ ubyte colors_read; /**< which color components are read by the FS */
ubyte colors_written;
boolean reads_position; /**< does fragment shader read position? */
boolean reads_z; /**< does fragment shader read depth? */
boolean writes_z; /**< does fragment shader write Z value? */
boolean writes_stencil; /**< does fragment shader write stencil value? */
+ boolean writes_samplemask; /**< does fragment shader write sample mask? */
boolean writes_edgeflag; /**< vertex shader outputs edgeflag */
boolean uses_kill; /**< KILL or KILL_IF instruction used? */
boolean uses_persp_center;
"IMM",
"PRED",
"SV",
- "RES",
- "SVIEW"
+ "IMAGE",
+ "SVIEW",
+ "BUFFER",
};
const char *tgsi_semantic_names[TGSI_SEMANTIC_COUNT] =
"TESSINNER",
"VERTICESIN",
"HELPER_INVOCATION",
+ "BASEINSTANCE",
+ "DRAWID",
};
const char *tgsi_texture_names[TGSI_TEXTURE_COUNT] =
"FLT64"
};
+const char *tgsi_memory_names[3] =
+{
+ "COHERENT",
+ "RESTRICT",
+ "VOLATILE",
+};
+
static inline void
tgsi_strings_check(void)
extern const char *tgsi_immediate_type_names[4];
+extern const char *tgsi_memory_names[3];
+
const char *
tgsi_file_name(unsigned file);
inst.Texture.Texture = TGSI_TEXTURE_UNKNOWN;
}
+ if ((i >= TGSI_OPCODE_LOAD && i <= TGSI_OPCODE_ATOMIMAX) ||
+ i == TGSI_OPCODE_RESQ) {
+ inst.Instruction.Memory = 1;
+ inst.Memory.Qualifier = 0;
+ }
+
/* Parse instruction operands.
*/
for (i = 0; i < info->num_dst + info->num_src + info->is_tex; i++) {
inst.Texture.NumOffsets = i;
cur = ctx->cur;
+ eat_opt_white(&cur);
+ for (i = 0; inst.Instruction.Memory && *cur == ','; i++) {
+ uint j;
+ cur++;
+ eat_opt_white(&cur);
+ ctx->cur = cur;
+ for (j = 0; j < 3; j++) {
+ if (str_match_nocase_whole(&ctx->cur, tgsi_memory_names[j])) {
+ inst.Memory.Qualifier |= 1U << j;
+ break;
+ }
+ }
+ if (j == 3) {
+ report_error(ctx, "Expected memory qualifier");
+ return FALSE;
+ }
+ cur = ctx->cur;
+ eat_opt_white(&cur);
+ }
+
+ cur = ctx->cur;
eat_opt_white( &cur );
if (info->is_branch && *cur == ':') {
uint target;
cur++;
eat_opt_white( &cur );
- if (file == TGSI_FILE_RESOURCE) {
+ if (file == TGSI_FILE_IMAGE) {
for (i = 0; i < TGSI_TEXTURE_COUNT; i++) {
if (str_match_nocase_whole(&cur, tgsi_texture_names[i])) {
- decl.Resource.Resource = i;
+ decl.Image.Resource = i;
break;
}
}
return FALSE;
}
+ /* XXX format */
+
cur2 = cur;
eat_opt_white(&cur2);
while (*cur2 == ',') {
cur2++;
eat_opt_white(&cur2);
if (str_match_nocase_whole(&cur2, "RAW")) {
- decl.Resource.Raw = 1;
+ decl.Image.Raw = 1;
} else if (str_match_nocase_whole(&cur2, "WR")) {
- decl.Resource.Writable = 1;
+ decl.Image.Writable = 1;
} else {
break;
decl.SamplerView.ReturnTypeX;
}
ctx->cur = cur;
+ } else if (file == TGSI_FILE_BUFFER) {
+ if (str_match_nocase_whole(&cur, "ATOMIC")) {
+ decl.Declaration.Atomic = 1;
+ ctx->cur = cur;
+ }
} else {
if (str_match_nocase_whole(&cur, "LOCAL")) {
decl.Declaration.Local = 1;
struct tgsi_declaration_range decl_range;
struct tgsi_declaration_dimension decl_dim;
struct tgsi_declaration_interp decl_interp;
+ struct tgsi_declaration_image decl_image;
struct tgsi_declaration_semantic decl_semantic;
struct tgsi_declaration_sampler_view decl_sampler_view;
struct tgsi_declaration_array array;
struct tgsi_instruction_predicate insn_predicate;
struct tgsi_instruction_label insn_label;
struct tgsi_instruction_texture insn_texture;
+ struct tgsi_instruction_memory insn_memory;
struct tgsi_texture_offset insn_texture_offset;
struct tgsi_src_register src;
struct tgsi_ind_register ind;
unsigned vs_inputs[PIPE_MAX_ATTRIBS/32];
struct {
- unsigned index;
unsigned semantic_name;
unsigned semantic_index;
} system_value[UREG_MAX_SYSTEM_VALUE];
} sampler_view[PIPE_MAX_SHADER_SAMPLER_VIEWS];
unsigned nr_sampler_views;
+ struct {
+ unsigned index;
+ unsigned target;
+ unsigned format;
+ boolean wr;
+ boolean raw;
+ } image[PIPE_MAX_SHADER_IMAGES];
+ unsigned nr_images;
+
+ struct {
+ unsigned index;
+ bool atomic;
+ } buffer[PIPE_MAX_SHADER_BUFFERS];
+ unsigned nr_buffers;
+
struct util_bitmask *free_temps;
struct util_bitmask *local_temps;
struct util_bitmask *decl_temps;
struct ureg_src
ureg_DECL_system_value(struct ureg_program *ureg,
- unsigned index,
unsigned semantic_name,
unsigned semantic_index)
{
+ unsigned i;
+
+ for (i = 0; i < ureg->nr_system_values; i++) {
+ if (ureg->system_value[i].semantic_name == semantic_name &&
+ ureg->system_value[i].semantic_index == semantic_index) {
+ goto out;
+ }
+ }
+
if (ureg->nr_system_values < UREG_MAX_SYSTEM_VALUE) {
- ureg->system_value[ureg->nr_system_values].index = index;
ureg->system_value[ureg->nr_system_values].semantic_name = semantic_name;
ureg->system_value[ureg->nr_system_values].semantic_index = semantic_index;
+ i = ureg->nr_system_values;
ureg->nr_system_values++;
} else {
set_bad(ureg);
}
- return ureg_src_register(TGSI_FILE_SYSTEM_VALUE, index);
+out:
+ return ureg_src_register(TGSI_FILE_SYSTEM_VALUE, i);
}
return reg;
}
+/* Allocate a new image.
+ */
+struct ureg_src
+ureg_DECL_image(struct ureg_program *ureg,
+ unsigned index,
+ unsigned target,
+ unsigned format,
+ boolean wr,
+ boolean raw)
+{
+ struct ureg_src reg = ureg_src_register(TGSI_FILE_IMAGE, index);
+ unsigned i;
+
+ for (i = 0; i < ureg->nr_images; i++)
+ if (ureg->image[i].index == index)
+ return reg;
+
+ if (i < PIPE_MAX_SHADER_IMAGES) {
+ ureg->image[i].index = index;
+ ureg->image[i].target = target;
+ ureg->image[i].wr = wr;
+ ureg->image[i].raw = raw;
+ ureg->image[i].format = format;
+ ureg->nr_images++;
+ return reg;
+ }
+
+ assert(0);
+ return reg;
+}
+
+/* Allocate a new buffer.
+ */
+struct ureg_src ureg_DECL_buffer(struct ureg_program *ureg, unsigned nr,
+ bool atomic)
+{
+ struct ureg_src reg = ureg_src_register(TGSI_FILE_BUFFER, nr);
+ unsigned i;
+
+ for (i = 0; i < ureg->nr_buffers; i++)
+ if (ureg->buffer[i].index == nr)
+ return reg;
+
+ if (i < PIPE_MAX_SHADER_BUFFERS) {
+ ureg->buffer[i].index = nr;
+ ureg->buffer[i].atomic = atomic;
+ ureg->nr_buffers++;
+ return reg;
+ }
+
+ assert(0);
+ return reg;
+}
+
static int
match_or_expand_immediate64( const unsigned *v,
int type,
}
+void
+ureg_emit_memory(struct ureg_program *ureg,
+ unsigned extended_token,
+ unsigned qualifier)
+{
+ union tgsi_any_token *out, *insn;
+
+ out = get_tokens( ureg, DOMAIN_INSN, 1 );
+ insn = retrieve_token( ureg, DOMAIN_INSN, extended_token );
+
+ insn->insn.Memory = 1;
+
+ out[0].value = 0;
+ out[0].insn_memory.Qualifier = qualifier;
+}
void
ureg_fixup_insn_size(struct ureg_program *ureg,
}
+void
+ureg_memory_insn(struct ureg_program *ureg,
+ unsigned opcode,
+ const struct ureg_dst *dst,
+ unsigned nr_dst,
+ const struct ureg_src *src,
+ unsigned nr_src,
+ unsigned qualifier)
+{
+ struct ureg_emit_insn_result insn;
+ unsigned i;
+
+ insn = ureg_emit_insn(ureg,
+ opcode,
+ FALSE,
+ FALSE,
+ FALSE,
+ TGSI_SWIZZLE_X,
+ TGSI_SWIZZLE_Y,
+ TGSI_SWIZZLE_Z,
+ TGSI_SWIZZLE_W,
+ nr_dst,
+ nr_src);
+
+ ureg_emit_memory(ureg, insn.extended_token, qualifier);
+
+ for (i = 0; i < nr_dst; i++)
+ ureg_emit_dst(ureg, dst[i]);
+
+ for (i = 0; i < nr_src; i++)
+ ureg_emit_src(ureg, src[i]);
+
+ ureg_fixup_insn_size(ureg, insn.insn_token);
+}
+
+
static void
emit_decl_semantic(struct ureg_program *ureg,
unsigned file,
}
static void
+emit_decl_image(struct ureg_program *ureg,
+ unsigned index,
+ unsigned target,
+ unsigned format,
+ boolean wr,
+ boolean raw)
+{
+ union tgsi_any_token *out = get_tokens(ureg, DOMAIN_DECL, 3);
+
+ out[0].value = 0;
+ out[0].decl.Type = TGSI_TOKEN_TYPE_DECLARATION;
+ out[0].decl.NrTokens = 3;
+ out[0].decl.File = TGSI_FILE_IMAGE;
+ out[0].decl.UsageMask = 0xf;
+
+ out[1].value = 0;
+ out[1].decl_range.First = index;
+ out[1].decl_range.Last = index;
+
+ out[2].value = 0;
+ out[2].decl_image.Resource = target;
+ out[2].decl_image.Writable = wr;
+ out[2].decl_image.Raw = raw;
+ out[2].decl_image.Format = format;
+}
+
+static void
+emit_decl_buffer(struct ureg_program *ureg,
+ unsigned index,
+ bool atomic)
+{
+ union tgsi_any_token *out = get_tokens(ureg, DOMAIN_DECL, 2);
+
+ out[0].value = 0;
+ out[0].decl.Type = TGSI_TOKEN_TYPE_DECLARATION;
+ out[0].decl.NrTokens = 2;
+ out[0].decl.File = TGSI_FILE_BUFFER;
+ out[0].decl.UsageMask = 0xf;
+ out[0].decl.Atomic = atomic;
+
+ out[1].value = 0;
+ out[1].decl_range.First = index;
+ out[1].decl_range.Last = index;
+}
+
+static void
emit_immediate( struct ureg_program *ureg,
const unsigned *v,
unsigned type )
for (i = 0; i < ureg->nr_system_values; i++) {
emit_decl_semantic(ureg,
TGSI_FILE_SYSTEM_VALUE,
- ureg->system_value[i].index,
- ureg->system_value[i].index,
+ i,
+ i,
ureg->system_value[i].semantic_name,
ureg->system_value[i].semantic_index,
TGSI_WRITEMASK_XYZW, 0);
ureg->sampler_view[i].return_type_w);
}
+ for (i = 0; i < ureg->nr_images; i++) {
+ emit_decl_image(ureg,
+ ureg->image[i].index,
+ ureg->image[i].target,
+ ureg->image[i].format,
+ ureg->image[i].wr,
+ ureg->image[i].raw);
+ }
+
+ for (i = 0; i < ureg->nr_buffers; i++) {
+ emit_decl_buffer(ureg, ureg->buffer[i].index, ureg->buffer[i].atomic);
+ }
+
if (ureg->const_decls.nr_constant_ranges) {
for (i = 0; i < ureg->const_decls.nr_constant_ranges; i++) {
emit_decl_range(ureg,
struct ureg_src
ureg_DECL_system_value(struct ureg_program *,
- unsigned index,
unsigned semantic_name,
unsigned semantic_index);
unsigned return_type_z,
unsigned return_type_w );
+struct ureg_src
+ureg_DECL_image(struct ureg_program *ureg,
+ unsigned index,
+ unsigned target,
+ unsigned format,
+ boolean wr,
+ boolean raw);
+
+struct ureg_src
+ureg_DECL_buffer(struct ureg_program *ureg, unsigned nr, bool atomic);
static inline struct ureg_src
ureg_imm4f( struct ureg_program *ureg,
unsigned nr_src,
unsigned *label);
+void
+ureg_memory_insn(struct ureg_program *ureg,
+ unsigned opcode,
+ const struct ureg_dst *dst,
+ unsigned nr_dst,
+ const struct ureg_src *src,
+ unsigned nr_src,
+ unsigned qualifier);
/***********************************************************************
* Internal instruction helpers, don't call these directly:
ureg_emit_texture_offset(struct ureg_program *ureg,
const struct tgsi_texture_offset *offset);
+void
+ureg_emit_memory(struct ureg_program *ureg,
+ unsigned insn_token,
+ unsigned qualifier);
+
void
ureg_emit_dst( struct ureg_program *ureg,
struct ureg_dst dst );
#include "pipe/p_shader_tokens.h"
#include "tgsi_parse.h"
#include "tgsi_util.h"
+#include "tgsi_exec.h"
union pointer_hack
{
const struct tgsi_src_register *reg,
unsigned component )
{
- switch( component ) {
- case 0:
+ switch (component) {
+ case TGSI_CHAN_X:
return reg->SwizzleX;
- case 1:
+ case TGSI_CHAN_Y:
return reg->SwizzleY;
- case 2:
+ case TGSI_CHAN_Z:
return reg->SwizzleZ;
- case 3:
+ case TGSI_CHAN_W:
return reg->SwizzleW;
default:
- assert( 0 );
+ assert(0);
}
return 0;
}
for (i = 0; i < 4; i++)
ctx->vertices[i][0][3] = 1; /*v.w*/
- ctx->upload = u_upload_create(pipe, 65536, 4, PIPE_BIND_VERTEX_BUFFER);
+ ctx->upload = u_upload_create(pipe, 65536, PIPE_BIND_VERTEX_BUFFER,
+ PIPE_USAGE_STREAM);
return &ctx->base;
}
vb.stride = 8 * sizeof(float);
- u_upload_data(ctx->upload, 0, sizeof(ctx->vertices), ctx->vertices,
+ u_upload_data(ctx->upload, 0, sizeof(ctx->vertices), 4, ctx->vertices,
&vb.buffer_offset, &vb.buffer);
if (!vb.buffer)
return;
return;
}
- u_upload_data(ctx->upload, 0, num_channels*4, clear_value,
+ u_upload_data(ctx->upload, 0, num_channels*4, 4, clear_value,
&vb.buffer_offset, &vb.buffer);
if (!vb.buffer)
goto out;
;
}
+void
+debug_dump_ubyte_rgba_bmp(const char *filename,
+ unsigned width, unsigned height,
+ const ubyte *rgba, unsigned stride)
+{
+ FILE *stream;
+ struct bmp_file_header bmfh;
+ struct bmp_info_header bmih;
+ unsigned x, y;
+
+ assert(rgba);
+ if(!rgba)
+ goto error1;
+
+ bmfh.bfType = 0x4d42;
+ bmfh.bfSize = 14 + 40 + height*width*4;
+ bmfh.bfReserved1 = 0;
+ bmfh.bfReserved2 = 0;
+ bmfh.bfOffBits = 14 + 40;
+
+ bmih.biSize = 40;
+ bmih.biWidth = width;
+ bmih.biHeight = height;
+ bmih.biPlanes = 1;
+ bmih.biBitCount = 32;
+ bmih.biCompression = 0;
+ bmih.biSizeImage = height*width*4;
+ bmih.biXPelsPerMeter = 0;
+ bmih.biYPelsPerMeter = 0;
+ bmih.biClrUsed = 0;
+ bmih.biClrImportant = 0;
+
+ stream = fopen(filename, "wb");
+ assert(stream);
+ if(!stream)
+ goto error1;
+
+ fwrite(&bmfh, 14, 1, stream);
+ fwrite(&bmih, 40, 1, stream);
+
+ y = height;
+ while(y--) {
+ const ubyte *ptr = rgba + (stride * y * 4);
+ for(x = 0; x < width; ++x)
+ {
+ struct bmp_rgb_quad pixel;
+ pixel.rgbRed = ptr[x*4 + 0];
+ pixel.rgbGreen = ptr[x*4 + 1];
+ pixel.rgbBlue = ptr[x*4 + 2];
+ pixel.rgbAlpha = ptr[x*4 + 3];
+ fwrite(&pixel, 1, 4, stream);
+ }
+ }
+
+ fclose(stream);
+error1:
+ ;
+}
+
/**
* Print PIPE_TRANSFER_x flags with a message.
const struct debug_named_value *flags,
uint64_t dfault);
+#define DEBUG_GET_ONCE_OPTION(suffix, name, dfault) \
+static const char * \
+debug_get_option_ ## suffix (void) \
+{ \
+ static boolean first = TRUE; \
+ static const char * value; \
+ if (first) { \
+ first = FALSE; \
+ value = debug_get_option(name, dfault); \
+ } \
+ return value; \
+}
+
#define DEBUG_GET_ONCE_BOOL_OPTION(sufix, name, dfault) \
static boolean \
debug_get_option_ ## sufix (void) \
void debug_dump_float_rgba_bmp(const char *filename,
unsigned width, unsigned height,
float *rgba, unsigned stride);
+void debug_dump_ubyte_rgba_bmp(const char *filename,
+ unsigned width, unsigned height,
+ const ubyte *rgba, unsigned stride);
#else
#define debug_dump_image(prefix, format, cpp, width, height, stride, data) ((void)0)
#define debug_dump_surface(pipe, prefix, surface) ((void)0)
#define debug_dump_surface_bmp(pipe, filename, surface) ((void)0)
#define debug_dump_transfer_bmp(filename, transfer, ptr) ((void)0)
#define debug_dump_float_rgba_bmp(filename, width, height, rgba, stride) ((void)0)
+#define debug_dump_ubyte_rgba_bmp(filename, width, height, rgba, stride) ((void)0)
#endif
+++ /dev/null
-/*
- * Copyright 2010 Luca Barbieri
- *
- * Permission is hereby granted, free of charge, to any person obtaining
- * a copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sublicense, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- *
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial
- * portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
- * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
- * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
- * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
- * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- *
- **************************************************************************/
-
-#ifndef U_INIT_H
-#define U_INIT_H
-
-/* Use UTIL_INIT(f) to have f called at program initialization.
- Note that it is only guaranteed to be called if any symbol in the
- .c file it is in sis referenced by the program.
-
- UTIL_INIT functions are called in arbitrary order.
-*/
-
-#ifdef __cplusplus
-/* use a C++ global constructor */
-#define UTIL_INIT(f) struct f##__gctor_t {f##__gctor_t() {x();}} f##__gctor;
-#elif defined(_MSC_VER)
-/* add a pointer to the section where MSVC stores global constructor pointers */
-/* see http://blogs.msdn.com/vcblog/archive/2006/10/20/crt-initialization.aspx and
- http://stackoverflow.com/questions/1113409/attribute-constructor-equivalent-in-vc */
-#pragma section(".CRT$XCU",read)
-#define UTIL_INIT(f) static void __cdecl f##__init(void) {f();}; __declspec(allocate(".CRT$XCU")) void (__cdecl* f##__xcu)(void) = f##__init;
-#elif defined(__GNUC__)
-#define UTIL_INIT(f) static void f##__init(void) __attribute__((constructor)); static void f##__init(void) {f();}
-#else
-#error Unsupported compiler: please find out how to implement global initializers in C on it
-#endif
-
-#endif
-
struct tgsi_shader_info info;
uint tempsUsed; /**< bitmask */
int wincoordInput;
+ unsigned wincoordFile;
int maxInput;
uint samplersUsed; /**< bitfield of samplers used */
int freeSampler; /** an available sampler for the pstipple */
pctx->samplersUsed |= 1 << i;
}
}
- else if (decl->Declaration.File == TGSI_FILE_INPUT) {
+ else if (decl->Declaration.File == pctx->wincoordFile) {
pctx->maxInput = MAX2(pctx->maxInput, (int) decl->Range.Last);
if (decl->Semantic.Name == TGSI_SEMANTIC_POSITION)
pctx->wincoordInput = (int) decl->Range.First;
wincoordInput = pctx->wincoordInput;
if (pctx->wincoordInput < 0) {
+ struct tgsi_full_declaration decl;
+
+ decl = tgsi_default_full_declaration();
/* declare new position input reg */
- tgsi_transform_input_decl(ctx, wincoordInput,
- TGSI_SEMANTIC_POSITION, 1,
- TGSI_INTERPOLATE_LINEAR);
+ decl.Declaration.File = pctx->wincoordFile;
+ decl.Declaration.Semantic = 1;
+ decl.Semantic.Name = TGSI_SEMANTIC_POSITION;
+ decl.Range.First =
+ decl.Range.Last = wincoordInput;
+
+ if (pctx->wincoordFile == TGSI_FILE_INPUT) {
+ decl.Declaration.Interpolate = 1;
+ decl.Interp.Interpolate = TGSI_INTERPOLATE_LINEAR;
+ }
+
+ ctx->emit_declaration(ctx, &decl);
}
sampIdx = pctx->hasFixedUnit ? pctx->fixedUnit : pctx->freeSampler;
tgsi_transform_op2_inst(ctx, TGSI_OPCODE_MUL,
TGSI_FILE_TEMPORARY, texTemp,
TGSI_WRITEMASK_XYZW,
- TGSI_FILE_INPUT, wincoordInput,
+ pctx->wincoordFile, wincoordInput,
TGSI_FILE_IMMEDIATE, pctx->numImmed);
/* TEX texTemp, texTemp, sampler; */
* will be used to sample the stipple texture;
* if NULL, the fixed unit is used
* \param fixedUnit fixed texture unit used for the stipple texture
+ * \param wincoordFile TGSI_FILE_INPUT or TGSI_FILE_SYSTEM_VALUE,
+ * depending on which one is supported by the driver
+ * for TGSI_SEMANTIC_POSITION in the fragment shader
*/
struct tgsi_token *
util_pstipple_create_fragment_shader(const struct tgsi_token *tokens,
unsigned *samplerUnitOut,
- unsigned fixedUnit)
+ unsigned fixedUnit,
+ unsigned wincoordFile)
{
struct pstip_transform_context transform;
const uint newLen = tgsi_num_tokens(tokens) + NUM_NEW_TOKENS;
*/
memset(&transform, 0, sizeof(transform));
transform.wincoordInput = -1;
+ transform.wincoordFile = wincoordFile;
transform.maxInput = -1;
transform.coordOrigin = TGSI_FS_COORD_ORIGIN_UPPER_LEFT;
transform.hasFixedUnit = !samplerUnitOut;
struct tgsi_token *
util_pstipple_create_fragment_shader(const struct tgsi_token *tokens,
unsigned *samplerUnitOut,
- unsigned fixed_unit);
+ unsigned fixed_unit,
+ unsigned wincoordFile);
#endif
--- /dev/null
+/*
+ * Copyright 2015 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Author: Oded Gabbay <oded.gabbay@redhat.com>
+ */
+
+/**
+ * @file
+ * POWER8 intrinsics portability header.
+ *
+ */
+
+#ifndef U_PWR8_H_
+#define U_PWR8_H_
+
+#if defined(_ARCH_PWR8) && defined(PIPE_ARCH_LITTLE_ENDIAN)
+
+#define VECTOR_ALIGN_16 __attribute__ ((__aligned__ (16)))
+
+typedef VECTOR_ALIGN_16 vector unsigned char __m128i;
+
+typedef VECTOR_ALIGN_16 union m128i {
+ __m128i m128i;
+ vector signed int m128si;
+ vector unsigned int m128ui;
+ ubyte ub[16];
+ ushort us[8];
+ int i[4];
+ uint ui[4];
+} __m128i_union;
+
+static inline __m128i
+vec_set_epi32 (int i3, int i2, int i1, int i0)
+{
+ __m128i_union vdst;
+
+#ifdef PIPE_ARCH_LITTLE_ENDIAN
+ vdst.i[0] = i0;
+ vdst.i[1] = i1;
+ vdst.i[2] = i2;
+ vdst.i[3] = i3;
+#else
+ vdst.i[3] = i0;
+ vdst.i[2] = i1;
+ vdst.i[1] = i2;
+ vdst.i[0] = i3;
+#endif
+
+ return (__m128i) vdst.m128si;
+}
+
+static inline __m128i
+vec_setr_epi32 (int i0, int i1, int i2, int i3)
+{
+ return vec_set_epi32 (i3, i2, i1, i0);
+}
+
+static inline __m128i
+vec_unpacklo_epi32 (__m128i even, __m128i odd)
+{
+ static const __m128i perm_mask =
+#ifdef PIPE_ARCH_LITTLE_ENDIAN
+ { 0, 1, 2, 3, 16, 17, 18, 19, 4, 5, 6, 7, 20, 21, 22, 23};
+#else
+ {24, 25, 26, 27, 8, 9, 10, 11, 28, 29, 30, 31, 12, 13, 14, 15};
+#endif
+
+ return vec_perm (even, odd, perm_mask);
+}
+
+static inline __m128i
+vec_unpackhi_epi32 (__m128i even, __m128i odd)
+{
+ static const __m128i perm_mask =
+#ifdef PIPE_ARCH_LITTLE_ENDIAN
+ { 8, 9, 10, 11, 24, 25, 26, 27, 12, 13, 14, 15, 28, 29, 30, 31};
+#else
+ {16, 17, 18, 19, 0, 1, 2, 3, 20, 21, 22, 23, 4, 5, 6, 7};
+#endif
+
+ return vec_perm (even, odd, perm_mask);
+}
+
+static inline __m128i
+vec_unpacklo_epi64 (__m128i even, __m128i odd)
+{
+ static const __m128i perm_mask =
+#ifdef PIPE_ARCH_LITTLE_ENDIAN
+ { 0, 1, 2, 3, 4, 5, 6, 7, 16, 17, 18, 19, 20, 21, 22, 23};
+#else
+ {24, 25, 26, 27, 28, 29, 30, 31, 8, 9, 10, 11, 12, 13, 14, 15};
+#endif
+
+ return vec_perm (even, odd, perm_mask);
+}
+
+static inline __m128i
+vec_unpackhi_epi64 (__m128i even, __m128i odd)
+{
+ static const __m128i perm_mask =
+#ifdef PIPE_ARCH_LITTLE_ENDIAN
+ { 8, 9, 10, 11, 12, 13, 14, 15, 24, 25, 26, 27, 28, 29, 30, 31};
+#else
+ {16, 17, 18, 19, 20, 21, 22, 23, 0, 1, 2, 3, 4, 5, 6, 7};
+#endif
+
+ return vec_perm (even, odd, perm_mask);
+}
+
+static inline __m128i
+vec_add_epi32 (__m128i a, __m128i b)
+{
+ return (__m128i) vec_add ((vector signed int) a, (vector signed int) b);
+}
+
+static inline __m128i
+vec_sub_epi32 (__m128i a, __m128i b)
+{
+ return (__m128i) vec_sub ((vector signed int) a, (vector signed int) b);
+}
+
+/* Call this function ONLY on POWER8 and newer platforms */
+static inline __m128i
+vec_mullo_epi32 (__m128i a, __m128i b)
+{
+ __m128i v;
+
+ __asm__(
+ "vmuluwm %0, %1, %2 \n"
+ : "=v" (v)
+ : "v" (a), "v" (b)
+ );
+
+ return v;
+}
+
+static inline void
+transpose4_epi32(const __m128i * restrict a,
+ const __m128i * restrict b,
+ const __m128i * restrict c,
+ const __m128i * restrict d,
+ __m128i * restrict o,
+ __m128i * restrict p,
+ __m128i * restrict q,
+ __m128i * restrict r)
+{
+ __m128i t0 = vec_unpacklo_epi32(*a, *b);
+ __m128i t1 = vec_unpacklo_epi32(*c, *d);
+ __m128i t2 = vec_unpackhi_epi32(*a, *b);
+ __m128i t3 = vec_unpackhi_epi32(*c, *d);
+
+ *o = vec_unpacklo_epi64(t0, t1);
+ *p = vec_unpackhi_epi64(t0, t1);
+ *q = vec_unpacklo_epi64(t2, t3);
+ *r = vec_unpackhi_epi64(t2, t3);
+}
+
+static inline __m128i
+vec_slli_epi32 (__m128i vsrc, unsigned int count)
+{
+ __m128i_union vec_count;
+
+ if (count >= 32)
+ return (__m128i) vec_splats (0);
+ else if (count == 0)
+ return vsrc;
+
+ /* In VMX, all shift count fields must contain the same value */
+ vec_count.m128si = (vector signed int) vec_splats (count);
+ return (__m128i) vec_sl ((vector signed int) vsrc, vec_count.m128ui);
+}
+
+static inline __m128i
+vec_srli_epi32 (__m128i vsrc, unsigned int count)
+{
+ __m128i_union vec_count;
+
+ if (count >= 32)
+ return (__m128i) vec_splats (0);
+ else if (count == 0)
+ return vsrc;
+
+ /* In VMX, all shift count fields must contain the same value */
+ vec_count.m128si = (vector signed int) vec_splats (count);
+ return (__m128i) vec_sr ((vector signed int) vsrc, vec_count.m128ui);
+}
+
+static inline __m128i
+vec_srai_epi32 (__m128i vsrc, unsigned int count)
+{
+ __m128i_union vec_count;
+
+ if (count >= 32)
+ return (__m128i) vec_splats (0);
+ else if (count == 0)
+ return vsrc;
+
+ /* In VMX, all shift count fields must contain the same value */
+ vec_count.m128si = (vector signed int) vec_splats (count);
+ return (__m128i) vec_sra ((vector signed int) vsrc, vec_count.m128ui);
+}
+
+static inline __m128i
+vec_cmpeq_epi32 (__m128i a, __m128i b)
+{
+ return (__m128i) vec_cmpeq ((vector signed int) a, (vector signed int) b);
+}
+
+static inline __m128i
+vec_loadu_si128 (const uint32_t* src)
+{
+ __m128i_union vsrc;
+
+#ifdef PIPE_ARCH_LITTLE_ENDIAN
+
+ vsrc.m128ui = *((vector unsigned int *) src);
+
+#else
+
+ __m128i vmask, tmp1, tmp2;
+
+ vmask = vec_lvsl(0, src);
+
+ tmp1 = (__m128i) vec_ld (0, src);
+ tmp2 = (__m128i) vec_ld (15, src);
+ vsrc.m128ui = (vector unsigned int) vec_perm (tmp1, tmp2, vmask);
+
+#endif
+
+ return vsrc.m128i;
+}
+
+static inline __m128i
+vec_load_si128 (const uint32_t* src)
+{
+ __m128i_union vsrc;
+
+ vsrc.m128ui = *((vector unsigned int *) src);
+
+ return vsrc.m128i;
+}
+
+static inline void
+vec_store_si128 (uint32_t* dest, __m128i vdata)
+{
+ vec_st ((vector unsigned int) vdata, 0, dest);
+}
+
+/* Call this function ONLY on POWER8 and newer platforms */
+static inline int
+vec_movemask_epi8 (__m128i vsrc)
+{
+ __m128i_union vtemp;
+ int result;
+
+ vtemp.m128i = vec_vgbbd(vsrc);
+
+#ifdef PIPE_ARCH_LITTLE_ENDIAN
+ result = vtemp.ub[15] << 8 | vtemp.ub[7];
+#else
+ result = vtemp.ub[0] << 8 | vtemp.ub[8];
+#endif
+
+ return result;
+}
+
+static inline __m128i
+vec_packs_epi16 (__m128i a, __m128i b)
+{
+#ifdef PIPE_ARCH_LITTLE_ENDIAN
+ return (__m128i) vec_packs ((vector signed short) a,
+ (vector signed short) b);
+#else
+ return (__m128i) vec_packs ((vector signed short) b,
+ (vector signed short) a);
+#endif
+}
+
+static inline __m128i
+vec_packs_epi32 (__m128i a, __m128i b)
+{
+#ifdef PIPE_ARCH_LITTLE_ENDIAN
+ return (__m128i) vec_packs ((vector signed int) a, (vector signed int) b);
+#else
+ return (__m128i) vec_packs ((vector signed int) b, (vector signed int) a);
+#endif
+}
+
+#endif /* _ARCH_PWR8 && PIPE_ARCH_LITTLE_ENDIAN */
+
+#endif /* U_PWR8_H_ */
#include "pipe/p_compiler.h"
+#include "pipe/p_shader_tokens.h"
struct pipe_context;
#endif /* !PIPE_ARCH_SSSE3 */
+/*
+ * Provide an SSE implementation of _mm_mul_epi32() in terms of
+ * _mm_mul_epu32().
+ *
+ * Basically, albeit surprising at first (and second, and third...) look
+ * if a * b is done signed instead of unsigned, can just
+ * subtract b from the high bits of the result if a is negative
+ * (and the same for a if b is negative). Modular arithmetic at its best!
+ *
+ * So for int32 a,b in crude pseudo-code ("*" here denoting a widening mul)
+ * fixupb = (signmask(b) & a) << 32ULL
+ * fixupa = (signmask(a) & b) << 32ULL
+ * a * b = (unsigned)a * (unsigned)b - fixupb - fixupa
+ * = (unsigned)a * (unsigned)b -(fixupb + fixupa)
+ *
+ * This does both lo (dwords 0/2) and hi parts (1/3) at the same time due
+ * to some optimization potential.
+ */
+static inline __m128i
+mm_mullohi_epi32(const __m128i a, const __m128i b, __m128i *res13)
+{
+ __m128i a13, b13, mul02, mul13;
+ __m128i anegmask, bnegmask, fixup, fixup02, fixup13;
+ a13 = _mm_shuffle_epi32(a, _MM_SHUFFLE(2,3,0,1));
+ b13 = _mm_shuffle_epi32(b, _MM_SHUFFLE(2,3,0,1));
+ anegmask = _mm_srai_epi32(a, 31);
+ bnegmask = _mm_srai_epi32(b, 31);
+ fixup = _mm_add_epi32(_mm_and_si128(anegmask, b),
+ _mm_and_si128(bnegmask, a));
+ mul02 = _mm_mul_epu32(a, b);
+ mul13 = _mm_mul_epu32(a13, b13);
+ fixup02 = _mm_slli_epi64(fixup, 32);
+ fixup13 = _mm_and_si128(fixup, _mm_set_epi32(-1,0,-1,0));
+ *res13 = _mm_sub_epi64(mul13, fixup13);
+ return _mm_sub_epi64(mul02, fixup02);
+}
/* Provide an SSE2 implementation of _mm_mullo_epi32() in terms of
* _mm_mul_epu32().
*
- * I suspect this works fine for us because one of our operands is
- * always positive, but not sure that this can be used for general
- * signed integer multiplication.
+ * This always works regardless the signs of the operands, since
+ * the high bits (which would be different) aren't used.
*
* This seems close enough to the speed of SSE4 and the real
* _mm_mullo_epi32() intrinsic as to not justify adding an sse4
/* Interleave the results, either with shuffles or (slightly
* faster) direct bit operations:
+ * XXX: might be only true for some cpus (in particular 65nm
+ * Core 2). On most cpus (including that Core 2, but not Nehalem...)
+ * using _mm_shuffle_ps/_mm_shuffle_epi32 might also be faster
+ * than using the 3 instructions below. But logic should be fine
+ * as well, we can't have optimal solution for all cpus (if anything,
+ * should just use _mm_mullo_epi32() if sse41 is available...).
*/
#if 0
__m128i ba8 = _mm_shuffle_epi32(ba, 8);
__m128i * restrict q,
__m128i * restrict r)
{
- __m128i t0 = _mm_unpacklo_epi32(*a, *b);
- __m128i t1 = _mm_unpacklo_epi32(*c, *d);
- __m128i t2 = _mm_unpackhi_epi32(*a, *b);
- __m128i t3 = _mm_unpackhi_epi32(*c, *d);
-
- *o = _mm_unpacklo_epi64(t0, t1);
- *p = _mm_unpackhi_epi64(t0, t1);
- *q = _mm_unpacklo_epi64(t2, t3);
- *r = _mm_unpackhi_epi64(t2, t3);
+ __m128i t0 = _mm_unpacklo_epi32(*a, *b);
+ __m128i t1 = _mm_unpacklo_epi32(*c, *d);
+ __m128i t2 = _mm_unpackhi_epi32(*a, *b);
+ __m128i t3 = _mm_unpackhi_epi32(*c, *d);
+
+ *o = _mm_unpacklo_epi64(t0, t1);
+ *p = _mm_unpackhi_epi64(t0, t1);
+ *q = _mm_unpacklo_epi64(t2, t3);
+ *r = _mm_unpackhi_epi64(t2, t3);
}
+
+/*
+ * Same as above, except the first two values are already interleaved
+ * (i.e. contain 64bit values).
+ */
+static inline void
+transpose2_64_2_32(const __m128i * restrict a01,
+ const __m128i * restrict a23,
+ const __m128i * restrict c,
+ const __m128i * restrict d,
+ __m128i * restrict o,
+ __m128i * restrict p,
+ __m128i * restrict q,
+ __m128i * restrict r)
+{
+ __m128i t0 = *a01;
+ __m128i t1 = _mm_unpacklo_epi32(*c, *d);
+ __m128i t2 = *a23;
+ __m128i t3 = _mm_unpackhi_epi32(*c, *d);
+
+ *o = _mm_unpacklo_epi64(t0, t1);
+ *p = _mm_unpackhi_epi64(t0, t1);
+ *q = _mm_unpacklo_epi64(t2, t3);
+ *r = _mm_unpackhi_epi64(t2, t3);
+}
+
+
#define SCALAR_EPI32(m, i) _mm_shuffle_epi32((m), _MM_SHUFFLE(i,i,i,i))
depth = res->array_size;
assert(res->array_size % 6 == 0);
break;
- case PIPE_MAX_TEXTURE_TYPES:;
+ case PIPE_MAX_TEXTURE_TYPES:
+ break;
}
return box->x >= 0 &&
struct pipe_context *pipe;
unsigned default_size; /* Minimum size of the upload buffer, in bytes. */
- unsigned alignment; /* Alignment of each sub-allocation. */
unsigned bind; /* Bitmask of PIPE_BIND_* flags. */
+ unsigned usage; /* PIPE_USAGE_* */
unsigned map_flags; /* Bitmask of PIPE_TRANSFER_* flags. */
boolean map_persistent; /* If persistent mappings are supported. */
};
-struct u_upload_mgr *u_upload_create( struct pipe_context *pipe,
- unsigned default_size,
- unsigned alignment,
- unsigned bind )
+struct u_upload_mgr *
+u_upload_create(struct pipe_context *pipe, unsigned default_size,
+ unsigned bind, unsigned usage)
{
struct u_upload_mgr *upload = CALLOC_STRUCT( u_upload_mgr );
if (!upload)
upload->pipe = pipe;
upload->default_size = default_size;
- upload->alignment = alignment;
upload->bind = bind;
+ upload->usage = usage;
upload->map_persistent =
pipe->screen->get_param(pipe->screen,
buffer.target = PIPE_BUFFER;
buffer.format = PIPE_FORMAT_R8_UNORM; /* want TYPELESS or similar */
buffer.bind = upload->bind;
- buffer.usage = PIPE_USAGE_STREAM;
+ buffer.usage = upload->usage;
buffer.width0 = size;
buffer.height0 = 1;
buffer.depth0 = 1;
u_upload_alloc(struct u_upload_mgr *upload,
unsigned min_out_offset,
unsigned size,
+ unsigned alignment,
unsigned *out_offset,
struct pipe_resource **outbuf,
void **ptr)
{
- unsigned alloc_size = align(size, upload->alignment);
- unsigned alloc_offset = align(min_out_offset, upload->alignment);
unsigned buffer_size = upload->buffer ? upload->buffer->width0 : 0;
unsigned offset;
+ min_out_offset = align(min_out_offset, alignment);
+
+ offset = align(upload->offset, alignment);
+ offset = MAX2(offset, min_out_offset);
+
/* Make sure we have enough space in the upload buffer
- * for the sub-allocation. */
- if (unlikely(MAX2(upload->offset, alloc_offset) + alloc_size > buffer_size)) {
- u_upload_alloc_buffer(upload, alloc_offset + alloc_size);
+ * for the sub-allocation.
+ */
+ if (unlikely(!upload->buffer || offset + size > buffer_size)) {
+ u_upload_alloc_buffer(upload, min_out_offset + size);
if (unlikely(!upload->buffer)) {
*out_offset = ~0;
return;
}
+ offset = min_out_offset;
buffer_size = upload->buffer->width0;
}
- offset = MAX2(upload->offset, alloc_offset);
-
if (unlikely(!upload->map)) {
upload->map = pipe_buffer_map_range(upload->pipe, upload->buffer,
offset,
upload->map -= offset;
}
- assert(offset < upload->buffer->width0);
- assert(offset + size <= upload->buffer->width0);
+ assert(offset < buffer_size);
+ assert(offset + size <= buffer_size);
assert(size);
/* Emit the return values: */
pipe_resource_reference(outbuf, upload->buffer);
*out_offset = offset;
- upload->offset = offset + alloc_size;
+ upload->offset = offset + size;
}
void u_upload_data(struct u_upload_mgr *upload,
unsigned min_out_offset,
unsigned size,
+ unsigned alignment,
const void *data,
unsigned *out_offset,
struct pipe_resource **outbuf)
{
uint8_t *ptr;
- u_upload_alloc(upload, min_out_offset, size,
+ u_upload_alloc(upload, min_out_offset, size, alignment,
out_offset, outbuf,
(void**)&ptr);
if (ptr)
unsigned min_out_offset,
unsigned offset,
unsigned size,
+ unsigned alignment,
struct pipe_resource *inbuf,
unsigned *out_offset,
struct pipe_resource **outbuf)
if (0)
debug_printf("upload ptr %p ofs %d sz %d\n", map, offset, size);
- u_upload_data(upload, min_out_offset, size, map, out_offset, outbuf);
+ u_upload_data(upload, min_out_offset, size, alignment,
+ map, out_offset, outbuf);
pipe_buffer_unmap( upload->pipe, transfer );
}
*
* \param pipe Pipe driver.
* \param default_size Minimum size of the upload buffer, in bytes.
- * \param alignment Alignment of each suballocation in the upload buffer.
* \param bind Bitmask of PIPE_BIND_* flags.
+ * \param usage PIPE_USAGE_*
*/
-struct u_upload_mgr *u_upload_create( struct pipe_context *pipe,
- unsigned default_size,
- unsigned alignment,
- unsigned bind );
+struct u_upload_mgr *
+u_upload_create(struct pipe_context *pipe, unsigned default_size,
+ unsigned bind, unsigned usage);
/**
* Destroy the upload manager.
* \param upload Upload manager
* \param min_out_offset Minimum offset that should be returned in out_offset.
* \param size Size of the allocation.
+ * \param alignment Alignment of the suballocation within the buffer
* \param out_offset Pointer to where the new buffer offset will be returned.
* \param outbuf Pointer to where the upload buffer will be returned.
* \param ptr Pointer to the allocated memory that is returned.
void u_upload_alloc(struct u_upload_mgr *upload,
unsigned min_out_offset,
unsigned size,
+ unsigned alignment,
unsigned *out_offset,
struct pipe_resource **outbuf,
void **ptr);
void u_upload_data(struct u_upload_mgr *upload,
unsigned min_out_offset,
unsigned size,
+ unsigned alignment,
const void *data,
unsigned *out_offset,
struct pipe_resource **outbuf);
unsigned min_out_offset,
unsigned offset,
unsigned size,
+ unsigned alignment,
struct pipe_resource *inbuf,
unsigned *out_offset,
struct pipe_resource **outbuf);
mgr->translate_cache = translate_cache_create();
memset(mgr->fallback_vbs, ~0, sizeof(mgr->fallback_vbs));
- mgr->uploader = u_upload_create(pipe, 1024 * 1024, 4,
- PIPE_BIND_VERTEX_BUFFER);
+ mgr->uploader = u_upload_create(pipe, 1024 * 1024,
+ PIPE_BIND_VERTEX_BUFFER,
+ PIPE_USAGE_STREAM);
return mgr;
}
/* Create and map the output buffer. */
u_upload_alloc(mgr->uploader, 0,
- key->output_stride * num_indices,
+ key->output_stride * num_indices, 4,
&out_offset, &out_buffer,
(void**)&out_map);
if (!out_buffer)
/* Create and map the output buffer. */
u_upload_alloc(mgr->uploader,
key->output_stride * start_vertex,
- key->output_stride * num_vertices,
+ key->output_stride * num_vertices, 4,
&out_offset, &out_buffer,
(void**)&out_map);
if (!out_buffer)
real_vb = &mgr->real_vertex_buffer[i];
ptr = mgr->vertex_buffer[i].user_buffer;
- u_upload_data(mgr->uploader, start, end - start, ptr + start,
+ u_upload_data(mgr->uploader, start, end - start, 4, ptr + start,
&real_vb->buffer_offset, &real_vb->buffer);
if (!real_vb->buffer)
return PIPE_ERROR_OUT_OF_MEMORY;
/* Allocate new memory for vertices. */
u_upload_alloc(c->upload, 0,
c->vertex_buf.stride * VL_COMPOSITOR_MAX_LAYERS * 4, /* size */
+ 4, /* alignment */
&c->vertex_buf.buffer_offset, &c->vertex_buf.buffer,
(void**)&vb);
c->pipe = pipe;
- c->upload = u_upload_create(pipe, 128 * 1024, 4, PIPE_BIND_VERTEX_BUFFER);
+ c->upload = u_upload_create(pipe, 128 * 1024, PIPE_BIND_VERTEX_BUFFER,
+ PIPE_USAGE_STREAM);
if (!c->upload)
return false;
#include "util/u_draw.h"
#include "util/u_memory.h"
#include "util/u_math.h"
+#include "util/u_format.h"
#include "vl_types.h"
#include "vl_video_buffer.h"
ureg_imm4f(shader, -0.02353f, 0, 0, 0));
ureg_MUL(shader, ureg_saturate(ureg_writemask(t_diff, TGSI_WRITEMASK_X)),
ureg_src(t_diff), ureg_imm4f(shader, 31.8750f, 0, 0, 0));
- ureg_LRP(shader, ureg_writemask(o_fragment, TGSI_WRITEMASK_X), ureg_src(t_diff),
+ ureg_LRP(shader, ureg_writemask(t_tex, TGSI_WRITEMASK_X), ureg_src(t_diff),
ureg_src(t_linear), ureg_src(t_weave));
+ ureg_MOV(shader, o_fragment, ureg_scalar(ureg_src(t_tex), TGSI_SWIZZLE_X));
ureg_release_temporary(shader, t_tex);
ureg_release_temporary(shader, t_comp_top);
/* TODO: handle other than 4:2:0 subsampling */
memset(&templ, 0, sizeof(templ));
- templ.buffer_format = PIPE_FORMAT_YV12;
+ templ.buffer_format = pipe->screen->get_video_param
+ (
+ pipe->screen,
+ PIPE_VIDEO_PROFILE_UNKNOWN,
+ PIPE_VIDEO_ENTRYPOINT_UNKNOWN,
+ PIPE_VIDEO_CAP_PREFERED_FORMAT
+ );
templ.chroma_format = PIPE_VIDEO_CHROMA_FORMAT_420;
templ.width = video_width;
templ.height = video_height;
goto error_rs_state;
memset(&blend, 0, sizeof blend);
- blend.rt[0].colormask = PIPE_MASK_RGBA;
- filter->blend = pipe->create_blend_state(pipe, &blend);
- if (!filter->blend)
- goto error_blend;
+ blend.rt[0].colormask = PIPE_MASK_R;
+ filter->blend[0] = pipe->create_blend_state(pipe, &blend);
+ if (!filter->blend[0])
+ goto error_blendR;
+
+ blend.rt[0].colormask = PIPE_MASK_G;
+ filter->blend[1] = pipe->create_blend_state(pipe, &blend);
+ if (!filter->blend[1])
+ goto error_blendG;
+
+ blend.rt[0].colormask = PIPE_MASK_B;
+ filter->blend[2] = pipe->create_blend_state(pipe, &blend);
+ if (!filter->blend[2])
+ goto error_blendB;
memset(&sampler, 0, sizeof(sampler));
sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
pipe->delete_sampler_state(pipe, filter->sampler);
error_sampler:
- pipe->delete_blend_state(pipe, filter->blend);
+ pipe->delete_blend_state(pipe, filter->blend[2]);
+
+error_blendB:
+ pipe->delete_blend_state(pipe, filter->blend[1]);
+
+error_blendG:
+ pipe->delete_blend_state(pipe, filter->blend[0]);
-error_blend:
+error_blendR:
pipe->delete_rasterizer_state(pipe, filter->rs_state);
error_rs_state:
assert(filter);
filter->pipe->delete_sampler_state(filter->pipe, filter->sampler[0]);
- filter->pipe->delete_blend_state(filter->pipe, filter->blend);
+ filter->pipe->delete_blend_state(filter->pipe, filter->blend[0]);
+ filter->pipe->delete_blend_state(filter->pipe, filter->blend[1]);
+ filter->pipe->delete_blend_state(filter->pipe, filter->blend[2]);
filter->pipe->delete_rasterizer_state(filter->pipe, filter->rs_state);
filter->pipe->delete_vertex_elements_state(filter->pipe, filter->ves);
pipe_resource_reference(&filter->quad.buffer, NULL);
struct pipe_sampler_view **next_sv;
struct pipe_sampler_view *sampler_views[4];
struct pipe_surface **dst_surfaces;
- int j;
+ const unsigned *plane_order;
+ int i, j;
assert(filter && prevprev && prev && cur && next && field <= 1);
/* set up destination and source */
dst_surfaces = filter->video_buffer->get_surfaces(filter->video_buffer);
+ plane_order = vl_video_buffer_plane_order(filter->video_buffer->buffer_format);
cur_sv = cur->get_sampler_view_components(cur);
prevprev_sv = prevprev->get_sampler_view_components(prevprev);
prev_sv = prev->get_sampler_view_components(prev);
/* set up pipe state */
filter->pipe->bind_rasterizer_state(filter->pipe, filter->rs_state);
- filter->pipe->bind_blend_state(filter->pipe, filter->blend);
filter->pipe->set_vertex_buffers(filter->pipe, 0, 1, &filter->quad);
filter->pipe->bind_vertex_elements_state(filter->pipe, filter->ves);
filter->pipe->bind_vs_state(filter->pipe, filter->vs);
fb_state.nr_cbufs = 1;
/* process each plane separately */
- for (j = 0; j < 3; j++) {
- /* select correct YV12 surfaces */
- int k = j == 1 ? 2 :
- j == 2 ? 1 : 0;
- struct pipe_surface *blit_surf = dst_surfaces[2 * k + field];
- struct pipe_surface *dst_surf = dst_surfaces[2 * k + 1 - field];
+ for (i = 0, j = 0; i < VL_NUM_COMPONENTS; ++i) {
+ struct pipe_surface *blit_surf = dst_surfaces[field];
+ struct pipe_surface *dst_surf = dst_surfaces[1 - field];
+ int k = plane_order[i];
+
+ /* bind blend state for this component in the plane */
+ filter->pipe->bind_blend_state(filter->pipe, filter->blend[j]);
/* update render target state */
viewport.scale[0] = blit_surf->texture->width0;
fb_state.height = blit_surf->texture->height0;
/* update sampler view sources */
- sampler_views[0] = prevprev_sv[j];
- sampler_views[1] = prev_sv[j];
- sampler_views[2] = cur_sv[j];
- sampler_views[3] = next_sv[j];
+ sampler_views[0] = prevprev_sv[k];
+ sampler_views[1] = prev_sv[k];
+ sampler_views[2] = cur_sv[k];
+ sampler_views[3] = next_sv[k];
filter->pipe->set_sampler_views(filter->pipe, PIPE_SHADER_FRAGMENT, 0, 4, sampler_views);
/* blit current field */
/* blit or interpolate other field */
fb_state.cbufs[0] = dst_surf;
filter->pipe->set_framebuffer_state(filter->pipe, &fb_state);
- if (j > 0 && filter->skip_chroma) {
+ if (i > 0 && filter->skip_chroma) {
util_draw_arrays(filter->pipe, PIPE_PRIM_QUADS, 0, 4);
} else {
filter->pipe->bind_fs_state(filter->pipe, field ? filter->fs_deint_top : filter->fs_deint_bottom);
util_draw_arrays(filter->pipe, PIPE_PRIM_QUADS, 0, 4);
}
+
+ if (++j >= util_format_get_nr_components(dst_surf->format)) {
+ dst_surfaces += 2;
+ j = 0;
+ }
}
}
struct pipe_vertex_buffer quad;
void *rs_state;
- void *blend;
+ void *blend[3];
void *sampler[4];
void *ves;
void *vs;
}
static struct ureg_dst
-calc_line(struct ureg_program *shader)
+calc_line(struct pipe_screen *screen, struct ureg_program *shader)
{
struct ureg_dst tmp;
struct ureg_src pos;
tmp = ureg_DECL_temporary(shader);
- pos = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_POSITION, VS_O_VPOS, TGSI_INTERPOLATE_LINEAR);
+ if (screen->get_param(screen, PIPE_CAP_TGSI_FS_POSITION_IS_SYSVAL))
+ pos = ureg_DECL_system_value(shader, TGSI_SEMANTIC_POSITION, 0);
+ else
+ pos = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_POSITION, VS_O_VPOS,
+ TGSI_INTERPOLATE_LINEAR);
/*
* tmp.y = fraction(pos.y / 2) >= 0.5 ? 1 : 0
fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
- field = calc_line(shader);
+ field = calc_line(r->pipe->screen, shader);
/*
* ref = field.z ? tc[1] : tc[0]
fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
- tmp = calc_line(shader);
+ tmp = calc_line(r->pipe->screen, shader);
/*
* if (field == tc.w)
for (j = 0; j < VL_MAX_REF_FRAMES; ++j) {
if (!ref_frames[j] || !ref_frames[j][i]) continue;
- vb[2] = vl_vb_get_mv(&buf->vertex_stream, j);;
+ vb[2] = vl_vb_get_mv(&buf->vertex_stream, j);
dec->context->set_vertex_buffers(dec->context, 0, 3, vb);
vl_mc_render_ref(i ? &dec->mc_c : &dec->mc_y, &buf->mc[i], ref_frames[j][i]);
templ->bind = PIPE_BIND_SAMPLER_VIEW | PIPE_BIND_RENDER_TARGET;
templ->usage = usage;
- if (plane > 0) {
- if (tmpl->chroma_format == PIPE_VIDEO_CHROMA_FORMAT_420) {
- templ->width0 /= 2;
- templ->height0 /= 2;
- } else if (tmpl->chroma_format == PIPE_VIDEO_CHROMA_FORMAT_422) {
- templ->width0 /= 2;
- }
- }
+ vl_video_buffer_adjust_size(&templ->width0, &templ->height0, plane,
+ tmpl->chroma_format, false);
}
static void
struct pipe_surface *surfaces[VL_MAX_SURFACES];
};
+static inline void
+vl_video_buffer_adjust_size(unsigned *width, unsigned *height, unsigned plane,
+ enum pipe_video_chroma_format chroma_format,
+ bool interlaced)
+{
+ if (interlaced) {
+ *height /= 2;
+ }
+ if (plane > 0) {
+ if (chroma_format == PIPE_VIDEO_CHROMA_FORMAT_420) {
+ *width /= 2;
+ *height /= 2;
+ } else if (chroma_format == PIPE_VIDEO_CHROMA_FORMAT_422) {
+ *width /= 2;
+ }
+ }
+}
+
/**
* get subformats for each plane
*/
program: ``bind_sampler_states`` may be used to set up texture
samplers for the compute stage and ``set_sampler_views`` may
be used to bind a number of sampler views to it.
+
+Mipmap generation
+^^^^^^^^^^^^^^^^^
+
+If PIPE_CAP_GENERATE_MIPMAP is true, ``generate_mipmap`` can be used
+to generate mipmaps for the specified texture resource.
+It replaces texel image levels base_level+1 through
+last_level for layers range from first_layer through last_layer.
+It returns TRUE if mipmap generation succeeds, otherwise it
+returns FALSE. Mipmap generation may fail when it is not supported
+for particular texture types or formats.
* ``PIPE_CAP_DRAW_INDIRECT``: Whether the driver supports taking draw arguments
{ count, instance_count, start, index_bias } from a PIPE_BUFFER resource.
See pipe_draw_info.
+* ``PIPE_CAP_MULTI_DRAW_INDIRECT``: Whether the driver supports
+ pipe_draw_info::indirect_stride and ::indirect_count
+* ``PIPE_CAP_MULTI_DRAW_INDIRECT_PARAMS``: Whether the driver supports
+ taking the number of indirect draws from a separate parameter
+ buffer, see pipe_draw_info::indirect_params.
* ``PIPE_CAP_TGSI_FS_FINE_DERIVATIVE``: Whether the fragment shader supports
the FINE versions of DDX/DDY.
* ``PIPE_CAP_VENDOR_ID``: The vendor ID of the underlying hardware. If it's
will need to lower TGSI_SEMANTIC_VERTEXID to TGSI_SEMANTIC_VERTEXID_NOBASE
and TGSI_SEMANTIC_BASEVERTEX, so drivers setting this must handle both these
semantics. Only relevant if geometry shaders are supported.
- (Currently not possible to query availability of these two semantics outside
- this, at least BASEVERTEX should be exposed separately too).
+ (BASEVERTEX could be exposed separately too via ``PIPE_CAP_DRAW_PARAMETERS``).
* ``PIPE_CAP_POLYGON_OFFSET_CLAMP``: If true, the driver implements support
for ``pipe_rasterizer_state::offset_clamp``.
* ``PIPE_CAP_MULTISAMPLE_Z_RESOLVE``: Whether the driver supports blitting
a compressed block is copied to/from a plain pixel of the same size.
* ``PIPE_CAP_CLEAR_TEXTURE``: Whether `clear_texture` will be
available in contexts.
+* ``PIPE_CAP_DRAW_PARAMETERS``: Whether ``TGSI_SEMANTIC_BASEVERTEX``,
+ ``TGSI_SEMANTIC_BASEINSTANCE``, and ``TGSI_SEMANTIC_DRAWID`` are
+ supported in vertex shaders.
+* ``PIPE_CAP_TGSI_PACK_HALF_FLOAT``: Whether the ``UP2H`` and ``PK2H``
+ TGSI opcodes are supported.
+* ``PIPE_CAP_TGSI_FS_POSITION_IS_SYSVAL``: If state trackers should use
+ a system value for the POSITION fragment shader input.
+* ``PIPE_CAP_TGSI_FS_FACE_IS_INTEGER_SYSVAL``: If state trackers should use
+ a system value for the FACE fragment shader input.
+ Also, the FACE system value is integer, not float.
+* ``PIPE_CAP_SHADER_BUFFER_OFFSET_ALIGNMENT``: Describes the required
+ alignment for pipe_shader_buffer::buffer_offset, in bytes. Maximum
+ value allowed is 256 (for GL conformance). 0 is only allowed if
+ shader buffers are not supported.
+* ``PIPE_CAP_INVALIDATE_BUFFER``: Whether the use of ``invalidate_resource``
+ for buffers is supported.
+* ``PIPE_CAP_GENERATE_MIPMAP``: Indicates whether pipe_context::generate_mipmap
+ is supported.
.. _pipe_capf:
of iterations that loops are allowed to have to be unrolled. It is only
a hint to state trackers. Whether any loops will be unrolled is not
guaranteed.
+* ``PIPE_SHADER_CAP_MAX_SHADER_BUFFERS``: Maximum number of memory buffers
+ (also used to implement atomic counters). Having this be non-0 also
+ implies support for the ``LOAD``, ``STORE``, and ``ATOM*`` TGSI
+ opcodes.
.. _pipe_compute_cap:
.. opcode:: PK2H - Pack Two 16-bit Floats
- TBD
+This instruction replicates its result.
+
+.. math::
+
+ dst = f32\_to\_f16(src.x) | f32\_to\_f16(src.y) << 16
.. opcode:: PK2US - Pack Two Unsigned 16-bit Scalars
.. opcode:: UP2H - Unpack Two 16-Bit Floats
- TBD
+.. math::
+
+ dst.x = f16\_to\_f32(src0.x \& 0xffff)
+
+ dst.y = f16\_to\_f32(src0.x >> 16)
+
+ dst.z = f16\_to\_f32(src0.x \& 0xffff)
+
+ dst.w = f16\_to\_f32(src0.x >> 16)
.. note::
Resource Access Opcodes
^^^^^^^^^^^^^^^^^^^^^^^
-.. opcode:: LOAD - Fetch data from a shader resource
+.. opcode:: LOAD - Fetch data from a shader buffer or image
Syntax: ``LOAD dst, resource, address``
- Example: ``LOAD TEMP[0], RES[0], TEMP[1]``
+ Example: ``LOAD TEMP[0], BUFFER[0], TEMP[1]``
Using the provided integer address, LOAD fetches data
from the specified buffer or texture without any
Syntax: ``STORE resource, address, src``
- Example: ``STORE RES[0], TEMP[0], TEMP[1]``
+ Example: ``STORE BUFFER[0], TEMP[0], TEMP[1]``
Using the provided integer address, STORE writes data
to the specified buffer or texture.
texture arrays and 2D textures. address.w is always
ignored.
+.. opcode:: RESQ - Query information about a resource
+
+ Syntax: ``RESQ dst, resource``
+
+ Example: ``RESQ TEMP[0], BUFFER[0]``
+
+ Returns information about the buffer or image resource. For buffer
+ resources, the size (in bytes) is returned in the x component. For
+ image resources, .xyz will contain the width/height/layers of the
+ image, while .w will contain the number of samples for multi-sampled
+ images.
+
.. _threadsyncopcodes:
logical operations. In this context atomicity means that another
concurrent memory access operation that affects the same memory
location is guaranteed to be performed strictly before or after the
-entire execution of the atomic operation.
-
-For the moment they're only valid in compute programs.
+entire execution of the atomic operation. The resource may be a buffer
+or an image. In the case of an image, the offset works the same as for
+``LOAD`` and ``STORE``, specified above. These atomic operations may
+only be used with 32-bit integer image formats.
.. opcode:: ATOMUADD - Atomic integer addition
Syntax: ``ATOMUADD dst, resource, offset, src``
- Example: ``ATOMUADD TEMP[0], RES[0], TEMP[1], TEMP[2]``
+ Example: ``ATOMUADD TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
- The following operation is performed atomically on each component:
+ The following operation is performed atomically:
.. math::
- dst_i = resource[offset]_i
+ dst_x = resource[offset]
- resource[offset]_i = dst_i + src_i
+ resource[offset] = dst_x + src_x
.. opcode:: ATOMXCHG - Atomic exchange
Syntax: ``ATOMXCHG dst, resource, offset, src``
- Example: ``ATOMXCHG TEMP[0], RES[0], TEMP[1], TEMP[2]``
+ Example: ``ATOMXCHG TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
- The following operation is performed atomically on each component:
+ The following operation is performed atomically:
.. math::
- dst_i = resource[offset]_i
+ dst_x = resource[offset]
- resource[offset]_i = src_i
+ resource[offset] = src_x
.. opcode:: ATOMCAS - Atomic compare-and-exchange
Syntax: ``ATOMCAS dst, resource, offset, cmp, src``
- Example: ``ATOMCAS TEMP[0], RES[0], TEMP[1], TEMP[2], TEMP[3]``
+ Example: ``ATOMCAS TEMP[0], BUFFER[0], TEMP[1], TEMP[2], TEMP[3]``
- The following operation is performed atomically on each component:
+ The following operation is performed atomically:
.. math::
- dst_i = resource[offset]_i
+ dst_x = resource[offset]
- resource[offset]_i = (dst_i == cmp_i ? src_i : dst_i)
+ resource[offset] = (dst_x == cmp_x ? src_x : dst_x)
.. opcode:: ATOMAND - Atomic bitwise And
Syntax: ``ATOMAND dst, resource, offset, src``
- Example: ``ATOMAND TEMP[0], RES[0], TEMP[1], TEMP[2]``
+ Example: ``ATOMAND TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
- The following operation is performed atomically on each component:
+ The following operation is performed atomically:
.. math::
- dst_i = resource[offset]_i
+ dst_x = resource[offset]
- resource[offset]_i = dst_i \& src_i
+ resource[offset] = dst_x \& src_x
.. opcode:: ATOMOR - Atomic bitwise Or
Syntax: ``ATOMOR dst, resource, offset, src``
- Example: ``ATOMOR TEMP[0], RES[0], TEMP[1], TEMP[2]``
+ Example: ``ATOMOR TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
- The following operation is performed atomically on each component:
+ The following operation is performed atomically:
.. math::
- dst_i = resource[offset]_i
+ dst_x = resource[offset]
- resource[offset]_i = dst_i | src_i
+ resource[offset] = dst_x | src_x
.. opcode:: ATOMXOR - Atomic bitwise Xor
Syntax: ``ATOMXOR dst, resource, offset, src``
- Example: ``ATOMXOR TEMP[0], RES[0], TEMP[1], TEMP[2]``
+ Example: ``ATOMXOR TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
- The following operation is performed atomically on each component:
+ The following operation is performed atomically:
.. math::
- dst_i = resource[offset]_i
+ dst_x = resource[offset]
- resource[offset]_i = dst_i \oplus src_i
+ resource[offset] = dst_x \oplus src_x
.. opcode:: ATOMUMIN - Atomic unsigned minimum
Syntax: ``ATOMUMIN dst, resource, offset, src``
- Example: ``ATOMUMIN TEMP[0], RES[0], TEMP[1], TEMP[2]``
+ Example: ``ATOMUMIN TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
- The following operation is performed atomically on each component:
+ The following operation is performed atomically:
.. math::
- dst_i = resource[offset]_i
+ dst_x = resource[offset]
- resource[offset]_i = (dst_i < src_i ? dst_i : src_i)
+ resource[offset] = (dst_x < src_x ? dst_x : src_x)
.. opcode:: ATOMUMAX - Atomic unsigned maximum
Syntax: ``ATOMUMAX dst, resource, offset, src``
- Example: ``ATOMUMAX TEMP[0], RES[0], TEMP[1], TEMP[2]``
+ Example: ``ATOMUMAX TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
- The following operation is performed atomically on each component:
+ The following operation is performed atomically:
.. math::
- dst_i = resource[offset]_i
+ dst_x = resource[offset]
- resource[offset]_i = (dst_i > src_i ? dst_i : src_i)
+ resource[offset] = (dst_x > src_x ? dst_x : src_x)
.. opcode:: ATOMIMIN - Atomic signed minimum
Syntax: ``ATOMIMIN dst, resource, offset, src``
- Example: ``ATOMIMIN TEMP[0], RES[0], TEMP[1], TEMP[2]``
+ Example: ``ATOMIMIN TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
- The following operation is performed atomically on each component:
+ The following operation is performed atomically:
.. math::
- dst_i = resource[offset]_i
+ dst_x = resource[offset]
- resource[offset]_i = (dst_i < src_i ? dst_i : src_i)
+ resource[offset] = (dst_x < src_x ? dst_x : src_x)
.. opcode:: ATOMIMAX - Atomic signed maximum
Syntax: ``ATOMIMAX dst, resource, offset, src``
- Example: ``ATOMIMAX TEMP[0], RES[0], TEMP[1], TEMP[2]``
+ Example: ``ATOMIMAX TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
- The following operation is performed atomically on each component:
+ The following operation is performed atomically:
.. math::
- dst_i = resource[offset]_i
+ dst_x = resource[offset]
- resource[offset]_i = (dst_i > src_i ? dst_i : src_i)
+ resource[offset] = (dst_x > src_x ? dst_x : src_x)
vertex will be divided by the W value to get normalized device coordinates.
For fragment shaders, TGSI_SEMANTIC_POSITION is used to indicate that
-fragment shader input contains the fragment's window position. The X
+fragment shader input (or system value, depending on which one is
+supported by the driver) contains the fragment's window position. The X
component starts at zero and always increases from left to right.
The Y component starts at zero and always increases but Y=0 may either
indicate the top of the window or the bottom depending on the fragment
TGSI_SEMANTIC_FACE
""""""""""""""""""
-This label applies to fragment shader inputs only and indicates that
-the register contains front/back-face information of the form (F, 0,
-0, 1). The first component will be positive when the fragment belongs
-to a front-facing polygon, and negative when the fragment belongs to a
-back-facing polygon.
+This label applies to fragment shader inputs (or system values,
+depending on which one is supported by the driver) and indicates that
+the register contains front/back-face information.
+
+If it is an input, it will be a floating-point vector in the form (F, 0, 0, 1),
+where F will be positive when the fragment belongs to a front-facing polygon,
+and negative when the fragment belongs to a back-facing polygon.
+
+If it is a system value, it will be an integer vector in the form (F, 0, 0, 1),
+where F is 0xffffffff when the fragment belongs to a front-facing polygon and
+0 when the fragment belongs to a back-facing polygon.
TGSI_SEMANTIC_EDGEFLAG
to properly compute derivatives, however it may be desirable to skip
some of the logic in those cases. See ``gl_HelperInvocation`` documentation.
+TGSI_SEMANTIC_BASEINSTANCE
+""""""""""""""""""""""""""
+
+For vertex shaders, the base instance argument supplied for this
+draw. This is an integer value, and only the X component is used.
+
+TGSI_SEMANTIC_DRAWID
+""""""""""""""""""""
+
+For vertex shaders, the zero-based index of the current draw in a
+``glMultiDraw*`` invocation. This is an integer value, and only the X
+component is used.
+
Declaration Interpolate
^^^^^^^^^^^^^^^^^^^^^^^
--- /dev/null
+ir3_compiler
ir3/ir3_group.c \
ir3/ir3.h \
ir3/ir3_legalize.c \
+ ir3/ir3_nir.c \
ir3/ir3_nir.h \
ir3/ir3_nir_lower_if_else.c \
ir3/ir3_print.c \
- /home/robclark/src/freedreno/envytools/rnndb/adreno.xml ( 398 bytes, from 2015-09-24 17:25:31)
- /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml ( 1453 bytes, from 2015-05-20 20:03:07)
- /home/robclark/src/freedreno/envytools/rnndb/adreno/a2xx.xml ( 32901 bytes, from 2015-05-20 20:03:14)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml ( 10755 bytes, from 2015-09-14 20:46:55)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml ( 14968 bytes, from 2015-05-20 20:12:27)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml ( 68291 bytes, from 2015-11-17 16:39:59)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml ( 64038 bytes, from 2015-11-17 16:37:36)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml ( 11518 bytes, from 2015-11-24 14:39:00)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml ( 15149 bytes, from 2015-11-20 16:22:25)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml ( 69600 bytes, from 2015-11-24 14:39:00)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml ( 67220 bytes, from 2015-12-13 17:58:09)
- /home/robclark/src/freedreno/envytools/rnndb/adreno/ocmem.xml ( 1773 bytes, from 2015-09-24 17:30:00)
Copyright (C) 2013-2015 by the following authors:
- /home/robclark/src/freedreno/envytools/rnndb/adreno.xml ( 398 bytes, from 2015-09-24 17:25:31)
- /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml ( 1453 bytes, from 2015-05-20 20:03:07)
- /home/robclark/src/freedreno/envytools/rnndb/adreno/a2xx.xml ( 32901 bytes, from 2015-05-20 20:03:14)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml ( 10755 bytes, from 2015-09-14 20:46:55)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml ( 14968 bytes, from 2015-05-20 20:12:27)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml ( 68291 bytes, from 2015-11-17 16:39:59)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml ( 64038 bytes, from 2015-11-17 16:37:36)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml ( 11518 bytes, from 2015-11-24 14:39:00)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml ( 15149 bytes, from 2015-11-20 16:22:25)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml ( 69600 bytes, from 2015-11-24 14:39:00)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml ( 67220 bytes, from 2015-12-13 17:58:09)
- /home/robclark/src/freedreno/envytools/rnndb/adreno/ocmem.xml ( 1773 bytes, from 2015-09-24 17:30:00)
Copyright (C) 2013-2015 by the following authors:
#define REG_A3XX_PC_RESTART_INDEX 0x000021ed
#define REG_A3XX_HLSQ_CONTROL_0_REG 0x00002200
-#define A3XX_HLSQ_CONTROL_0_REG_FSTHREADSIZE__MASK 0x00000010
+#define A3XX_HLSQ_CONTROL_0_REG_FSTHREADSIZE__MASK 0x00000030
#define A3XX_HLSQ_CONTROL_0_REG_FSTHREADSIZE__SHIFT 4
static inline uint32_t A3XX_HLSQ_CONTROL_0_REG_FSTHREADSIZE(enum a3xx_threadsize val)
{
return ((val) << A3XX_HLSQ_CONTROL_0_REG_FSTHREADSIZE__SHIFT) & A3XX_HLSQ_CONTROL_0_REG_FSTHREADSIZE__MASK;
}
#define A3XX_HLSQ_CONTROL_0_REG_FSSUPERTHREADENABLE 0x00000040
+#define A3XX_HLSQ_CONTROL_0_REG_COMPUTEMODE 0x00000100
#define A3XX_HLSQ_CONTROL_0_REG_SPSHADERRESTART 0x00000200
#define A3XX_HLSQ_CONTROL_0_REG_RESERVED2 0x00000400
+#define A3XX_HLSQ_CONTROL_0_REG_CYCLETIMEOUTLIMITVPC__MASK 0x00fff000
+#define A3XX_HLSQ_CONTROL_0_REG_CYCLETIMEOUTLIMITVPC__SHIFT 12
+static inline uint32_t A3XX_HLSQ_CONTROL_0_REG_CYCLETIMEOUTLIMITVPC(uint32_t val)
+{
+ return ((val) << A3XX_HLSQ_CONTROL_0_REG_CYCLETIMEOUTLIMITVPC__SHIFT) & A3XX_HLSQ_CONTROL_0_REG_CYCLETIMEOUTLIMITVPC__MASK;
+}
+#define A3XX_HLSQ_CONTROL_0_REG_FSONLYTEX 0x02000000
#define A3XX_HLSQ_CONTROL_0_REG_CHUNKDISABLE 0x04000000
#define A3XX_HLSQ_CONTROL_0_REG_CONSTMODE__MASK 0x08000000
#define A3XX_HLSQ_CONTROL_0_REG_CONSTMODE__SHIFT 27
#define A3XX_HLSQ_CONTROL_0_REG_SINGLECONTEXT 0x80000000
#define REG_A3XX_HLSQ_CONTROL_1_REG 0x00002201
-#define A3XX_HLSQ_CONTROL_1_REG_VSTHREADSIZE__MASK 0x00000040
+#define A3XX_HLSQ_CONTROL_1_REG_VSTHREADSIZE__MASK 0x000000c0
#define A3XX_HLSQ_CONTROL_1_REG_VSTHREADSIZE__SHIFT 6
static inline uint32_t A3XX_HLSQ_CONTROL_1_REG_VSTHREADSIZE(enum a3xx_threadsize val)
{
return ((val) << A3XX_HLSQ_CONTROL_1_REG_VSTHREADSIZE__SHIFT) & A3XX_HLSQ_CONTROL_1_REG_VSTHREADSIZE__MASK;
}
#define A3XX_HLSQ_CONTROL_1_REG_VSSUPERTHREADENABLE 0x00000100
-#define A3XX_HLSQ_CONTROL_1_REG_RESERVED1 0x00000200
-#define A3XX_HLSQ_CONTROL_1_REG_ZWCOORD 0x02000000
+#define A3XX_HLSQ_CONTROL_1_REG_FRAGCOORDXYREGID__MASK 0x00ff0000
+#define A3XX_HLSQ_CONTROL_1_REG_FRAGCOORDXYREGID__SHIFT 16
+static inline uint32_t A3XX_HLSQ_CONTROL_1_REG_FRAGCOORDXYREGID(uint32_t val)
+{
+ return ((val) << A3XX_HLSQ_CONTROL_1_REG_FRAGCOORDXYREGID__SHIFT) & A3XX_HLSQ_CONTROL_1_REG_FRAGCOORDXYREGID__MASK;
+}
+#define A3XX_HLSQ_CONTROL_1_REG_FRAGCOORDZWREGID__MASK 0xff000000
+#define A3XX_HLSQ_CONTROL_1_REG_FRAGCOORDZWREGID__SHIFT 24
+static inline uint32_t A3XX_HLSQ_CONTROL_1_REG_FRAGCOORDZWREGID(uint32_t val)
+{
+ return ((val) << A3XX_HLSQ_CONTROL_1_REG_FRAGCOORDZWREGID__SHIFT) & A3XX_HLSQ_CONTROL_1_REG_FRAGCOORDZWREGID__MASK;
+}
#define REG_A3XX_HLSQ_CONTROL_2_REG 0x00002202
+#define A3XX_HLSQ_CONTROL_2_REG_FACENESSREGID__MASK 0x000003fc
+#define A3XX_HLSQ_CONTROL_2_REG_FACENESSREGID__SHIFT 2
+static inline uint32_t A3XX_HLSQ_CONTROL_2_REG_FACENESSREGID(uint32_t val)
+{
+ return ((val) << A3XX_HLSQ_CONTROL_2_REG_FACENESSREGID__SHIFT) & A3XX_HLSQ_CONTROL_2_REG_FACENESSREGID__MASK;
+}
+#define A3XX_HLSQ_CONTROL_2_REG_COVVALUEREGID__MASK 0x03fc0000
+#define A3XX_HLSQ_CONTROL_2_REG_COVVALUEREGID__SHIFT 18
+static inline uint32_t A3XX_HLSQ_CONTROL_2_REG_COVVALUEREGID(uint32_t val)
+{
+ return ((val) << A3XX_HLSQ_CONTROL_2_REG_COVVALUEREGID__SHIFT) & A3XX_HLSQ_CONTROL_2_REG_COVVALUEREGID__MASK;
+}
#define A3XX_HLSQ_CONTROL_2_REG_PRIMALLOCTHRESHOLD__MASK 0xfc000000
#define A3XX_HLSQ_CONTROL_2_REG_PRIMALLOCTHRESHOLD__SHIFT 26
static inline uint32_t A3XX_HLSQ_CONTROL_2_REG_PRIMALLOCTHRESHOLD(uint32_t val)
}
#define REG_A3XX_HLSQ_VS_CONTROL_REG 0x00002204
-#define A3XX_HLSQ_VS_CONTROL_REG_CONSTLENGTH__MASK 0x00000fff
+#define A3XX_HLSQ_VS_CONTROL_REG_CONSTLENGTH__MASK 0x000003ff
#define A3XX_HLSQ_VS_CONTROL_REG_CONSTLENGTH__SHIFT 0
static inline uint32_t A3XX_HLSQ_VS_CONTROL_REG_CONSTLENGTH(uint32_t val)
{
return ((val) << A3XX_HLSQ_VS_CONTROL_REG_CONSTLENGTH__SHIFT) & A3XX_HLSQ_VS_CONTROL_REG_CONSTLENGTH__MASK;
}
-#define A3XX_HLSQ_VS_CONTROL_REG_CONSTSTARTOFFSET__MASK 0x00fff000
+#define A3XX_HLSQ_VS_CONTROL_REG_CONSTSTARTOFFSET__MASK 0x001ff000
#define A3XX_HLSQ_VS_CONTROL_REG_CONSTSTARTOFFSET__SHIFT 12
static inline uint32_t A3XX_HLSQ_VS_CONTROL_REG_CONSTSTARTOFFSET(uint32_t val)
{
}
#define REG_A3XX_HLSQ_FS_CONTROL_REG 0x00002205
-#define A3XX_HLSQ_FS_CONTROL_REG_CONSTLENGTH__MASK 0x00000fff
+#define A3XX_HLSQ_FS_CONTROL_REG_CONSTLENGTH__MASK 0x000003ff
#define A3XX_HLSQ_FS_CONTROL_REG_CONSTLENGTH__SHIFT 0
static inline uint32_t A3XX_HLSQ_FS_CONTROL_REG_CONSTLENGTH(uint32_t val)
{
return ((val) << A3XX_HLSQ_FS_CONTROL_REG_CONSTLENGTH__SHIFT) & A3XX_HLSQ_FS_CONTROL_REG_CONSTLENGTH__MASK;
}
-#define A3XX_HLSQ_FS_CONTROL_REG_CONSTSTARTOFFSET__MASK 0x00fff000
+#define A3XX_HLSQ_FS_CONTROL_REG_CONSTSTARTOFFSET__MASK 0x001ff000
#define A3XX_HLSQ_FS_CONTROL_REG_CONSTSTARTOFFSET__SHIFT 12
static inline uint32_t A3XX_HLSQ_FS_CONTROL_REG_CONSTSTARTOFFSET(uint32_t val)
{
}
#define REG_A3XX_HLSQ_CONST_VSPRESV_RANGE_REG 0x00002206
-#define A3XX_HLSQ_CONST_VSPRESV_RANGE_REG_STARTENTRY__MASK 0x0000ffff
+#define A3XX_HLSQ_CONST_VSPRESV_RANGE_REG_STARTENTRY__MASK 0x000001ff
#define A3XX_HLSQ_CONST_VSPRESV_RANGE_REG_STARTENTRY__SHIFT 0
static inline uint32_t A3XX_HLSQ_CONST_VSPRESV_RANGE_REG_STARTENTRY(uint32_t val)
{
return ((val) << A3XX_HLSQ_CONST_VSPRESV_RANGE_REG_STARTENTRY__SHIFT) & A3XX_HLSQ_CONST_VSPRESV_RANGE_REG_STARTENTRY__MASK;
}
-#define A3XX_HLSQ_CONST_VSPRESV_RANGE_REG_ENDENTRY__MASK 0xffff0000
+#define A3XX_HLSQ_CONST_VSPRESV_RANGE_REG_ENDENTRY__MASK 0x01ff0000
#define A3XX_HLSQ_CONST_VSPRESV_RANGE_REG_ENDENTRY__SHIFT 16
static inline uint32_t A3XX_HLSQ_CONST_VSPRESV_RANGE_REG_ENDENTRY(uint32_t val)
{
}
#define REG_A3XX_HLSQ_CONST_FSPRESV_RANGE_REG 0x00002207
-#define A3XX_HLSQ_CONST_FSPRESV_RANGE_REG_STARTENTRY__MASK 0x0000ffff
+#define A3XX_HLSQ_CONST_FSPRESV_RANGE_REG_STARTENTRY__MASK 0x000001ff
#define A3XX_HLSQ_CONST_FSPRESV_RANGE_REG_STARTENTRY__SHIFT 0
static inline uint32_t A3XX_HLSQ_CONST_FSPRESV_RANGE_REG_STARTENTRY(uint32_t val)
{
return ((val) << A3XX_HLSQ_CONST_FSPRESV_RANGE_REG_STARTENTRY__SHIFT) & A3XX_HLSQ_CONST_FSPRESV_RANGE_REG_STARTENTRY__MASK;
}
-#define A3XX_HLSQ_CONST_FSPRESV_RANGE_REG_ENDENTRY__MASK 0xffff0000
+#define A3XX_HLSQ_CONST_FSPRESV_RANGE_REG_ENDENTRY__MASK 0x01ff0000
#define A3XX_HLSQ_CONST_FSPRESV_RANGE_REG_ENDENTRY__SHIFT 16
static inline uint32_t A3XX_HLSQ_CONST_FSPRESV_RANGE_REG_ENDENTRY(uint32_t val)
{
return ((val) << A3XX_SP_VS_CTRL_REG0_INSTRBUFFERMODE__SHIFT) & A3XX_SP_VS_CTRL_REG0_INSTRBUFFERMODE__MASK;
}
#define A3XX_SP_VS_CTRL_REG0_CACHEINVALID 0x00000004
+#define A3XX_SP_VS_CTRL_REG0_ALUSCHMODE 0x00000008
#define A3XX_SP_VS_CTRL_REG0_HALFREGFOOTPRINT__MASK 0x000003f0
#define A3XX_SP_VS_CTRL_REG0_HALFREGFOOTPRINT__SHIFT 4
static inline uint32_t A3XX_SP_VS_CTRL_REG0_HALFREGFOOTPRINT(uint32_t val)
{
return ((val) << A3XX_SP_VS_CTRL_REG0_HALFREGFOOTPRINT__SHIFT) & A3XX_SP_VS_CTRL_REG0_HALFREGFOOTPRINT__MASK;
}
-#define A3XX_SP_VS_CTRL_REG0_FULLREGFOOTPRINT__MASK 0x0003fc00
+#define A3XX_SP_VS_CTRL_REG0_FULLREGFOOTPRINT__MASK 0x0000fc00
#define A3XX_SP_VS_CTRL_REG0_FULLREGFOOTPRINT__SHIFT 10
static inline uint32_t A3XX_SP_VS_CTRL_REG0_FULLREGFOOTPRINT(uint32_t val)
{
return ((val) << A3XX_SP_VS_CTRL_REG0_FULLREGFOOTPRINT__SHIFT) & A3XX_SP_VS_CTRL_REG0_FULLREGFOOTPRINT__MASK;
}
-#define A3XX_SP_VS_CTRL_REG0_INOUTREGOVERLAP__MASK 0x000c0000
-#define A3XX_SP_VS_CTRL_REG0_INOUTREGOVERLAP__SHIFT 18
-static inline uint32_t A3XX_SP_VS_CTRL_REG0_INOUTREGOVERLAP(uint32_t val)
-{
- return ((val) << A3XX_SP_VS_CTRL_REG0_INOUTREGOVERLAP__SHIFT) & A3XX_SP_VS_CTRL_REG0_INOUTREGOVERLAP__MASK;
-}
#define A3XX_SP_VS_CTRL_REG0_THREADSIZE__MASK 0x00100000
#define A3XX_SP_VS_CTRL_REG0_THREADSIZE__SHIFT 20
static inline uint32_t A3XX_SP_VS_CTRL_REG0_THREADSIZE(enum a3xx_threadsize val)
return ((val) << A3XX_SP_VS_CTRL_REG0_THREADSIZE__SHIFT) & A3XX_SP_VS_CTRL_REG0_THREADSIZE__MASK;
}
#define A3XX_SP_VS_CTRL_REG0_SUPERTHREADMODE 0x00200000
-#define A3XX_SP_VS_CTRL_REG0_PIXLODENABLE 0x00400000
-#define A3XX_SP_VS_CTRL_REG0_COMPUTEMODE 0x00800000
#define A3XX_SP_VS_CTRL_REG0_LENGTH__MASK 0xff000000
#define A3XX_SP_VS_CTRL_REG0_LENGTH__SHIFT 24
static inline uint32_t A3XX_SP_VS_CTRL_REG0_LENGTH(uint32_t val)
{
return ((val) << A3XX_SP_VS_PARAM_REG_PSIZEREGID__SHIFT) & A3XX_SP_VS_PARAM_REG_PSIZEREGID__MASK;
}
-#define A3XX_SP_VS_PARAM_REG_TOTALVSOUTVAR__MASK 0xfff00000
+#define A3XX_SP_VS_PARAM_REG_POS2DMODE 0x00010000
+#define A3XX_SP_VS_PARAM_REG_TOTALVSOUTVAR__MASK 0x01f00000
#define A3XX_SP_VS_PARAM_REG_TOTALVSOUTVAR__SHIFT 20
static inline uint32_t A3XX_SP_VS_PARAM_REG_TOTALVSOUTVAR(uint32_t val)
{
static inline uint32_t REG_A3XX_SP_VS_OUT(uint32_t i0) { return 0x000022c7 + 0x1*i0; }
static inline uint32_t REG_A3XX_SP_VS_OUT_REG(uint32_t i0) { return 0x000022c7 + 0x1*i0; }
-#define A3XX_SP_VS_OUT_REG_A_REGID__MASK 0x000001ff
+#define A3XX_SP_VS_OUT_REG_A_REGID__MASK 0x000000ff
#define A3XX_SP_VS_OUT_REG_A_REGID__SHIFT 0
static inline uint32_t A3XX_SP_VS_OUT_REG_A_REGID(uint32_t val)
{
return ((val) << A3XX_SP_VS_OUT_REG_A_REGID__SHIFT) & A3XX_SP_VS_OUT_REG_A_REGID__MASK;
}
+#define A3XX_SP_VS_OUT_REG_A_HALF 0x00000100
#define A3XX_SP_VS_OUT_REG_A_COMPMASK__MASK 0x00001e00
#define A3XX_SP_VS_OUT_REG_A_COMPMASK__SHIFT 9
static inline uint32_t A3XX_SP_VS_OUT_REG_A_COMPMASK(uint32_t val)
{
return ((val) << A3XX_SP_VS_OUT_REG_A_COMPMASK__SHIFT) & A3XX_SP_VS_OUT_REG_A_COMPMASK__MASK;
}
-#define A3XX_SP_VS_OUT_REG_B_REGID__MASK 0x01ff0000
+#define A3XX_SP_VS_OUT_REG_B_REGID__MASK 0x00ff0000
#define A3XX_SP_VS_OUT_REG_B_REGID__SHIFT 16
static inline uint32_t A3XX_SP_VS_OUT_REG_B_REGID(uint32_t val)
{
return ((val) << A3XX_SP_VS_OUT_REG_B_REGID__SHIFT) & A3XX_SP_VS_OUT_REG_B_REGID__MASK;
}
+#define A3XX_SP_VS_OUT_REG_B_HALF 0x01000000
#define A3XX_SP_VS_OUT_REG_B_COMPMASK__MASK 0x1e000000
#define A3XX_SP_VS_OUT_REG_B_COMPMASK__SHIFT 25
static inline uint32_t A3XX_SP_VS_OUT_REG_B_COMPMASK(uint32_t val)
static inline uint32_t REG_A3XX_SP_VS_VPC_DST(uint32_t i0) { return 0x000022d0 + 0x1*i0; }
static inline uint32_t REG_A3XX_SP_VS_VPC_DST_REG(uint32_t i0) { return 0x000022d0 + 0x1*i0; }
-#define A3XX_SP_VS_VPC_DST_REG_OUTLOC0__MASK 0x000000ff
+#define A3XX_SP_VS_VPC_DST_REG_OUTLOC0__MASK 0x0000007f
#define A3XX_SP_VS_VPC_DST_REG_OUTLOC0__SHIFT 0
static inline uint32_t A3XX_SP_VS_VPC_DST_REG_OUTLOC0(uint32_t val)
{
return ((val) << A3XX_SP_VS_VPC_DST_REG_OUTLOC0__SHIFT) & A3XX_SP_VS_VPC_DST_REG_OUTLOC0__MASK;
}
-#define A3XX_SP_VS_VPC_DST_REG_OUTLOC1__MASK 0x0000ff00
+#define A3XX_SP_VS_VPC_DST_REG_OUTLOC1__MASK 0x00007f00
#define A3XX_SP_VS_VPC_DST_REG_OUTLOC1__SHIFT 8
static inline uint32_t A3XX_SP_VS_VPC_DST_REG_OUTLOC1(uint32_t val)
{
return ((val) << A3XX_SP_VS_VPC_DST_REG_OUTLOC1__SHIFT) & A3XX_SP_VS_VPC_DST_REG_OUTLOC1__MASK;
}
-#define A3XX_SP_VS_VPC_DST_REG_OUTLOC2__MASK 0x00ff0000
+#define A3XX_SP_VS_VPC_DST_REG_OUTLOC2__MASK 0x007f0000
#define A3XX_SP_VS_VPC_DST_REG_OUTLOC2__SHIFT 16
static inline uint32_t A3XX_SP_VS_VPC_DST_REG_OUTLOC2(uint32_t val)
{
return ((val) << A3XX_SP_VS_VPC_DST_REG_OUTLOC2__SHIFT) & A3XX_SP_VS_VPC_DST_REG_OUTLOC2__MASK;
}
-#define A3XX_SP_VS_VPC_DST_REG_OUTLOC3__MASK 0xff000000
+#define A3XX_SP_VS_VPC_DST_REG_OUTLOC3__MASK 0x7f000000
#define A3XX_SP_VS_VPC_DST_REG_OUTLOC3__SHIFT 24
static inline uint32_t A3XX_SP_VS_VPC_DST_REG_OUTLOC3(uint32_t val)
{
}
#define REG_A3XX_SP_VS_OBJ_OFFSET_REG 0x000022d4
+#define A3XX_SP_VS_OBJ_OFFSET_REG_FIRSTEXECINSTROFFSET__MASK 0x0000ffff
+#define A3XX_SP_VS_OBJ_OFFSET_REG_FIRSTEXECINSTROFFSET__SHIFT 0
+static inline uint32_t A3XX_SP_VS_OBJ_OFFSET_REG_FIRSTEXECINSTROFFSET(uint32_t val)
+{
+ return ((val) << A3XX_SP_VS_OBJ_OFFSET_REG_FIRSTEXECINSTROFFSET__SHIFT) & A3XX_SP_VS_OBJ_OFFSET_REG_FIRSTEXECINSTROFFSET__MASK;
+}
#define A3XX_SP_VS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__MASK 0x01ff0000
#define A3XX_SP_VS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__SHIFT 16
static inline uint32_t A3XX_SP_VS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET(uint32_t val)
#define REG_A3XX_SP_VS_OBJ_START_REG 0x000022d5
#define REG_A3XX_SP_VS_PVT_MEM_PARAM_REG 0x000022d6
+#define A3XX_SP_VS_PVT_MEM_PARAM_REG_MEMSIZEPERITEM__MASK 0x000000ff
+#define A3XX_SP_VS_PVT_MEM_PARAM_REG_MEMSIZEPERITEM__SHIFT 0
+static inline uint32_t A3XX_SP_VS_PVT_MEM_PARAM_REG_MEMSIZEPERITEM(uint32_t val)
+{
+ return ((val) << A3XX_SP_VS_PVT_MEM_PARAM_REG_MEMSIZEPERITEM__SHIFT) & A3XX_SP_VS_PVT_MEM_PARAM_REG_MEMSIZEPERITEM__MASK;
+}
+#define A3XX_SP_VS_PVT_MEM_PARAM_REG_HWSTACKOFFSET__MASK 0x00ffff00
+#define A3XX_SP_VS_PVT_MEM_PARAM_REG_HWSTACKOFFSET__SHIFT 8
+static inline uint32_t A3XX_SP_VS_PVT_MEM_PARAM_REG_HWSTACKOFFSET(uint32_t val)
+{
+ return ((val) << A3XX_SP_VS_PVT_MEM_PARAM_REG_HWSTACKOFFSET__SHIFT) & A3XX_SP_VS_PVT_MEM_PARAM_REG_HWSTACKOFFSET__MASK;
+}
+#define A3XX_SP_VS_PVT_MEM_PARAM_REG_HWSTACKSIZEPERTHREAD__MASK 0xff000000
+#define A3XX_SP_VS_PVT_MEM_PARAM_REG_HWSTACKSIZEPERTHREAD__SHIFT 24
+static inline uint32_t A3XX_SP_VS_PVT_MEM_PARAM_REG_HWSTACKSIZEPERTHREAD(uint32_t val)
+{
+ return ((val) << A3XX_SP_VS_PVT_MEM_PARAM_REG_HWSTACKSIZEPERTHREAD__SHIFT) & A3XX_SP_VS_PVT_MEM_PARAM_REG_HWSTACKSIZEPERTHREAD__MASK;
+}
#define REG_A3XX_SP_VS_PVT_MEM_ADDR_REG 0x000022d7
+#define A3XX_SP_VS_PVT_MEM_ADDR_REG_BURSTLEN__MASK 0x0000001f
+#define A3XX_SP_VS_PVT_MEM_ADDR_REG_BURSTLEN__SHIFT 0
+static inline uint32_t A3XX_SP_VS_PVT_MEM_ADDR_REG_BURSTLEN(uint32_t val)
+{
+ return ((val) << A3XX_SP_VS_PVT_MEM_ADDR_REG_BURSTLEN__SHIFT) & A3XX_SP_VS_PVT_MEM_ADDR_REG_BURSTLEN__MASK;
+}
+#define A3XX_SP_VS_PVT_MEM_ADDR_REG_SHADERSTARTADDRESS__MASK 0xffffffe0
+#define A3XX_SP_VS_PVT_MEM_ADDR_REG_SHADERSTARTADDRESS__SHIFT 5
+static inline uint32_t A3XX_SP_VS_PVT_MEM_ADDR_REG_SHADERSTARTADDRESS(uint32_t val)
+{
+ return ((val >> 5) << A3XX_SP_VS_PVT_MEM_ADDR_REG_SHADERSTARTADDRESS__SHIFT) & A3XX_SP_VS_PVT_MEM_ADDR_REG_SHADERSTARTADDRESS__MASK;
+}
#define REG_A3XX_SP_VS_PVT_MEM_SIZE_REG 0x000022d8
return ((val) << A3XX_SP_FS_CTRL_REG0_INSTRBUFFERMODE__SHIFT) & A3XX_SP_FS_CTRL_REG0_INSTRBUFFERMODE__MASK;
}
#define A3XX_SP_FS_CTRL_REG0_CACHEINVALID 0x00000004
+#define A3XX_SP_FS_CTRL_REG0_ALUSCHMODE 0x00000008
#define A3XX_SP_FS_CTRL_REG0_HALFREGFOOTPRINT__MASK 0x000003f0
#define A3XX_SP_FS_CTRL_REG0_HALFREGFOOTPRINT__SHIFT 4
static inline uint32_t A3XX_SP_FS_CTRL_REG0_HALFREGFOOTPRINT(uint32_t val)
{
return ((val) << A3XX_SP_FS_CTRL_REG0_HALFREGFOOTPRINT__SHIFT) & A3XX_SP_FS_CTRL_REG0_HALFREGFOOTPRINT__MASK;
}
-#define A3XX_SP_FS_CTRL_REG0_FULLREGFOOTPRINT__MASK 0x0003fc00
+#define A3XX_SP_FS_CTRL_REG0_FULLREGFOOTPRINT__MASK 0x0000fc00
#define A3XX_SP_FS_CTRL_REG0_FULLREGFOOTPRINT__SHIFT 10
static inline uint32_t A3XX_SP_FS_CTRL_REG0_FULLREGFOOTPRINT(uint32_t val)
{
return ((val) << A3XX_SP_FS_CTRL_REG0_FULLREGFOOTPRINT__SHIFT) & A3XX_SP_FS_CTRL_REG0_FULLREGFOOTPRINT__MASK;
}
-#define A3XX_SP_FS_CTRL_REG0_INOUTREGOVERLAP__MASK 0x000c0000
-#define A3XX_SP_FS_CTRL_REG0_INOUTREGOVERLAP__SHIFT 18
-static inline uint32_t A3XX_SP_FS_CTRL_REG0_INOUTREGOVERLAP(uint32_t val)
-{
- return ((val) << A3XX_SP_FS_CTRL_REG0_INOUTREGOVERLAP__SHIFT) & A3XX_SP_FS_CTRL_REG0_INOUTREGOVERLAP__MASK;
-}
+#define A3XX_SP_FS_CTRL_REG0_FSBYPASSENABLE 0x00020000
+#define A3XX_SP_FS_CTRL_REG0_INOUTREGOVERLAP 0x00040000
+#define A3XX_SP_FS_CTRL_REG0_OUTORDERED 0x00080000
#define A3XX_SP_FS_CTRL_REG0_THREADSIZE__MASK 0x00100000
#define A3XX_SP_FS_CTRL_REG0_THREADSIZE__SHIFT 20
static inline uint32_t A3XX_SP_FS_CTRL_REG0_THREADSIZE(enum a3xx_threadsize val)
{
return ((val) << A3XX_SP_FS_CTRL_REG1_INITIALOUTSTANDING__SHIFT) & A3XX_SP_FS_CTRL_REG1_INITIALOUTSTANDING__MASK;
}
-#define A3XX_SP_FS_CTRL_REG1_HALFPRECVAROFFSET__MASK 0x3f000000
+#define A3XX_SP_FS_CTRL_REG1_HALFPRECVAROFFSET__MASK 0x7f000000
#define A3XX_SP_FS_CTRL_REG1_HALFPRECVAROFFSET__SHIFT 24
static inline uint32_t A3XX_SP_FS_CTRL_REG1_HALFPRECVAROFFSET(uint32_t val)
{
}
#define REG_A3XX_SP_FS_OBJ_OFFSET_REG 0x000022e2
+#define A3XX_SP_FS_OBJ_OFFSET_REG_FIRSTEXECINSTROFFSET__MASK 0x0000ffff
+#define A3XX_SP_FS_OBJ_OFFSET_REG_FIRSTEXECINSTROFFSET__SHIFT 0
+static inline uint32_t A3XX_SP_FS_OBJ_OFFSET_REG_FIRSTEXECINSTROFFSET(uint32_t val)
+{
+ return ((val) << A3XX_SP_FS_OBJ_OFFSET_REG_FIRSTEXECINSTROFFSET__SHIFT) & A3XX_SP_FS_OBJ_OFFSET_REG_FIRSTEXECINSTROFFSET__MASK;
+}
#define A3XX_SP_FS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__MASK 0x01ff0000
#define A3XX_SP_FS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__SHIFT 16
static inline uint32_t A3XX_SP_FS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET(uint32_t val)
#define REG_A3XX_SP_FS_OBJ_START_REG 0x000022e3
#define REG_A3XX_SP_FS_PVT_MEM_PARAM_REG 0x000022e4
+#define A3XX_SP_FS_PVT_MEM_PARAM_REG_MEMSIZEPERITEM__MASK 0x000000ff
+#define A3XX_SP_FS_PVT_MEM_PARAM_REG_MEMSIZEPERITEM__SHIFT 0
+static inline uint32_t A3XX_SP_FS_PVT_MEM_PARAM_REG_MEMSIZEPERITEM(uint32_t val)
+{
+ return ((val) << A3XX_SP_FS_PVT_MEM_PARAM_REG_MEMSIZEPERITEM__SHIFT) & A3XX_SP_FS_PVT_MEM_PARAM_REG_MEMSIZEPERITEM__MASK;
+}
+#define A3XX_SP_FS_PVT_MEM_PARAM_REG_HWSTACKOFFSET__MASK 0x00ffff00
+#define A3XX_SP_FS_PVT_MEM_PARAM_REG_HWSTACKOFFSET__SHIFT 8
+static inline uint32_t A3XX_SP_FS_PVT_MEM_PARAM_REG_HWSTACKOFFSET(uint32_t val)
+{
+ return ((val) << A3XX_SP_FS_PVT_MEM_PARAM_REG_HWSTACKOFFSET__SHIFT) & A3XX_SP_FS_PVT_MEM_PARAM_REG_HWSTACKOFFSET__MASK;
+}
+#define A3XX_SP_FS_PVT_MEM_PARAM_REG_HWSTACKSIZEPERTHREAD__MASK 0xff000000
+#define A3XX_SP_FS_PVT_MEM_PARAM_REG_HWSTACKSIZEPERTHREAD__SHIFT 24
+static inline uint32_t A3XX_SP_FS_PVT_MEM_PARAM_REG_HWSTACKSIZEPERTHREAD(uint32_t val)
+{
+ return ((val) << A3XX_SP_FS_PVT_MEM_PARAM_REG_HWSTACKSIZEPERTHREAD__SHIFT) & A3XX_SP_FS_PVT_MEM_PARAM_REG_HWSTACKSIZEPERTHREAD__MASK;
+}
#define REG_A3XX_SP_FS_PVT_MEM_ADDR_REG 0x000022e5
+#define A3XX_SP_FS_PVT_MEM_ADDR_REG_BURSTLEN__MASK 0x0000001f
+#define A3XX_SP_FS_PVT_MEM_ADDR_REG_BURSTLEN__SHIFT 0
+static inline uint32_t A3XX_SP_FS_PVT_MEM_ADDR_REG_BURSTLEN(uint32_t val)
+{
+ return ((val) << A3XX_SP_FS_PVT_MEM_ADDR_REG_BURSTLEN__SHIFT) & A3XX_SP_FS_PVT_MEM_ADDR_REG_BURSTLEN__MASK;
+}
+#define A3XX_SP_FS_PVT_MEM_ADDR_REG_SHADERSTARTADDRESS__MASK 0xffffffe0
+#define A3XX_SP_FS_PVT_MEM_ADDR_REG_SHADERSTARTADDRESS__SHIFT 5
+static inline uint32_t A3XX_SP_FS_PVT_MEM_ADDR_REG_SHADERSTARTADDRESS(uint32_t val)
+{
+ return ((val >> 5) << A3XX_SP_FS_PVT_MEM_ADDR_REG_SHADERSTARTADDRESS__SHIFT) & A3XX_SP_FS_PVT_MEM_ADDR_REG_SHADERSTARTADDRESS__MASK;
+}
#define REG_A3XX_SP_FS_PVT_MEM_SIZE_REG 0x000022e6
fd3_query_context_init(pctx);
- fd3_ctx->border_color_uploader = u_upload_create(pctx, 4096,
- 2 * PIPE_MAX_SAMPLERS * BORDERCOLOR_SIZE, 0);
+ fd3_ctx->border_color_uploader = u_upload_create(pctx, 4096, 0,
+ PIPE_USAGE_STREAM);
return pctx;
}
void *ptr;
u_upload_alloc(fd3_ctx->border_color_uploader,
- 0, 2 * PIPE_MAX_SAMPLERS * BORDERCOLOR_SIZE, &off,
+ 0, BORDER_COLOR_UPLOAD_SIZE,
+ BORDER_COLOR_UPLOAD_SIZE, &off,
&fd3_ctx->border_color_buf,
&ptr);
A3XX_HLSQ_CONTROL_0_REG_SPCONSTFULLUPDATE);
OUT_RING(ring, A3XX_HLSQ_CONTROL_1_REG_VSTHREADSIZE(TWO_QUADS) |
A3XX_HLSQ_CONTROL_1_REG_VSSUPERTHREADENABLE |
- COND(fp->frag_coord, A3XX_HLSQ_CONTROL_1_REG_ZWCOORD));
+ COND(fp->frag_coord, A3XX_HLSQ_CONTROL_1_REG_FRAGCOORDXYREGID(regid(0,0)) |
+ A3XX_HLSQ_CONTROL_1_REG_FRAGCOORDZWREGID(regid(0,2))));
OUT_RING(ring, A3XX_HLSQ_CONTROL_2_REG_PRIMALLOCTHRESHOLD(31));
OUT_RING(ring, A3XX_HLSQ_CONTROL_3_REG_REGID(fp->pos_regid));
OUT_RING(ring, A3XX_HLSQ_VS_CONTROL_REG_CONSTLENGTH(vp->constlen) |
COND(vpbuffer == CACHE, A3XX_SP_VS_CTRL_REG0_CACHEINVALID) |
A3XX_SP_VS_CTRL_REG0_HALFREGFOOTPRINT(vsi->max_half_reg + 1) |
A3XX_SP_VS_CTRL_REG0_FULLREGFOOTPRINT(vsi->max_reg + 1) |
- A3XX_SP_VS_CTRL_REG0_INOUTREGOVERLAP(0) |
A3XX_SP_VS_CTRL_REG0_THREADSIZE(TWO_QUADS) |
A3XX_SP_VS_CTRL_REG0_SUPERTHREADMODE |
- COND(vp->has_samp, A3XX_SP_VS_CTRL_REG0_PIXLODENABLE) |
A3XX_SP_VS_CTRL_REG0_LENGTH(vpbuffersz));
OUT_RING(ring, A3XX_SP_VS_CTRL_REG1_CONSTLENGTH(vp->constlen) |
A3XX_SP_VS_CTRL_REG1_INITIALOUTSTANDING(vp->total_in) |
COND(fpbuffer == CACHE, A3XX_SP_FS_CTRL_REG0_CACHEINVALID) |
A3XX_SP_FS_CTRL_REG0_HALFREGFOOTPRINT(fsi->max_half_reg + 1) |
A3XX_SP_FS_CTRL_REG0_FULLREGFOOTPRINT(fsi->max_reg + 1) |
- A3XX_SP_FS_CTRL_REG0_INOUTREGOVERLAP(1) |
+ A3XX_SP_FS_CTRL_REG0_INOUTREGOVERLAP |
A3XX_SP_FS_CTRL_REG0_THREADSIZE(FOUR_QUADS) |
A3XX_SP_FS_CTRL_REG0_SUPERTHREADMODE |
COND(fp->has_samp > 0, A3XX_SP_FS_CTRL_REG0_PIXLODENABLE) |
- /home/robclark/src/freedreno/envytools/rnndb/adreno.xml ( 398 bytes, from 2015-09-24 17:25:31)
- /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml ( 1453 bytes, from 2015-05-20 20:03:07)
- /home/robclark/src/freedreno/envytools/rnndb/adreno/a2xx.xml ( 32901 bytes, from 2015-05-20 20:03:14)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml ( 10755 bytes, from 2015-09-14 20:46:55)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml ( 14968 bytes, from 2015-05-20 20:12:27)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml ( 68291 bytes, from 2015-11-17 16:39:59)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml ( 64038 bytes, from 2015-11-17 16:37:36)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml ( 11518 bytes, from 2015-11-24 14:39:00)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml ( 15149 bytes, from 2015-11-20 16:22:25)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml ( 69600 bytes, from 2015-11-24 14:39:00)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml ( 67220 bytes, from 2015-12-13 17:58:09)
- /home/robclark/src/freedreno/envytools/rnndb/adreno/ocmem.xml ( 1773 bytes, from 2015-09-24 17:30:00)
Copyright (C) 2013-2015 by the following authors:
VFMT4_10_10_10_2_UNORM = 57,
VFMT4_10_10_10_2_SINT = 58,
VFMT4_10_10_10_2_SNORM = 59,
+ VFMT4_2_10_10_10_UINT = 60,
+ VFMT4_2_10_10_10_UNORM = 61,
+ VFMT4_2_10_10_10_SINT = 62,
+ VFMT4_2_10_10_10_SNORM = 63,
};
enum a4xx_tex_fmt {
- TFMT4_5_6_5_UNORM = 11,
- TFMT4_5_5_5_1_UNORM = 9,
- TFMT4_4_4_4_4_UNORM = 8,
- TFMT4_X8Z24_UNORM = 71,
- TFMT4_10_10_10_2_UNORM = 33,
- TFMT4_10_10_10_2_UINT = 34,
TFMT4_A8_UNORM = 3,
- TFMT4_L8_A8_UNORM = 13,
TFMT4_8_UNORM = 4,
- TFMT4_8_8_UNORM = 14,
- TFMT4_8_8_8_8_UNORM = 28,
TFMT4_8_SNORM = 5,
- TFMT4_8_8_SNORM = 15,
- TFMT4_8_8_8_8_SNORM = 29,
TFMT4_8_UINT = 6,
- TFMT4_8_8_UINT = 16,
- TFMT4_8_8_8_8_UINT = 30,
TFMT4_8_SINT = 7,
+ TFMT4_4_4_4_4_UNORM = 8,
+ TFMT4_5_5_5_1_UNORM = 9,
+ TFMT4_5_6_5_UNORM = 11,
+ TFMT4_L8_A8_UNORM = 13,
+ TFMT4_8_8_UNORM = 14,
+ TFMT4_8_8_SNORM = 15,
+ TFMT4_8_8_UINT = 16,
TFMT4_8_8_SINT = 17,
- TFMT4_8_8_8_8_SINT = 31,
TFMT4_16_UNORM = 18,
- TFMT4_16_16_UNORM = 38,
- TFMT4_16_16_16_16_UNORM = 51,
TFMT4_16_SNORM = 19,
- TFMT4_16_16_SNORM = 39,
- TFMT4_16_16_16_16_SNORM = 52,
+ TFMT4_16_FLOAT = 20,
TFMT4_16_UINT = 21,
- TFMT4_16_16_UINT = 41,
- TFMT4_16_16_16_16_UINT = 54,
TFMT4_16_SINT = 22,
+ TFMT4_8_8_8_8_UNORM = 28,
+ TFMT4_8_8_8_8_SNORM = 29,
+ TFMT4_8_8_8_8_UINT = 30,
+ TFMT4_8_8_8_8_SINT = 31,
+ TFMT4_9_9_9_E5_FLOAT = 32,
+ TFMT4_10_10_10_2_UNORM = 33,
+ TFMT4_10_10_10_2_UINT = 34,
+ TFMT4_11_11_10_FLOAT = 37,
+ TFMT4_16_16_UNORM = 38,
+ TFMT4_16_16_SNORM = 39,
+ TFMT4_16_16_FLOAT = 40,
+ TFMT4_16_16_UINT = 41,
TFMT4_16_16_SINT = 42,
- TFMT4_16_16_16_16_SINT = 55,
+ TFMT4_32_FLOAT = 43,
TFMT4_32_UINT = 44,
- TFMT4_32_32_UINT = 57,
- TFMT4_32_32_32_32_UINT = 64,
TFMT4_32_SINT = 45,
- TFMT4_32_32_SINT = 58,
- TFMT4_32_32_32_32_SINT = 65,
- TFMT4_16_FLOAT = 20,
- TFMT4_16_16_FLOAT = 40,
+ TFMT4_16_16_16_16_UNORM = 51,
+ TFMT4_16_16_16_16_SNORM = 52,
TFMT4_16_16_16_16_FLOAT = 53,
- TFMT4_32_FLOAT = 43,
+ TFMT4_16_16_16_16_UINT = 54,
+ TFMT4_16_16_16_16_SINT = 55,
TFMT4_32_32_FLOAT = 56,
- TFMT4_32_32_32_32_FLOAT = 63,
+ TFMT4_32_32_UINT = 57,
+ TFMT4_32_32_SINT = 58,
TFMT4_32_32_32_FLOAT = 59,
TFMT4_32_32_32_UINT = 60,
TFMT4_32_32_32_SINT = 61,
- TFMT4_9_9_9_E5_FLOAT = 32,
- TFMT4_11_11_10_FLOAT = 37,
+ TFMT4_32_32_32_32_FLOAT = 63,
+ TFMT4_32_32_32_32_UINT = 64,
+ TFMT4_32_32_32_32_SINT = 65,
+ TFMT4_X8Z24_UNORM = 71,
TFMT4_DXT1 = 86,
TFMT4_DXT3 = 87,
TFMT4_DXT5 = 88,
}
#define A4XX_RB_DEPTH_CONTROL_BF_ENABLE 0x00000080
#define A4XX_RB_DEPTH_CONTROL_EARLY_Z_DISABLE 0x00010000
+#define A4XX_RB_DEPTH_CONTROL_FORCE_FRAGZ_TO_FS 0x00020000
#define A4XX_RB_DEPTH_CONTROL_Z_TEST_ENABLE 0x80000000
#define REG_A4XX_RB_DEPTH_CLEAR 0x00002102
#define REG_A4XX_RBBM_CFG_DEBBUS_SEL_D 0x0000004d
+#define REG_A4XX_RBBM_POWER_CNTL_IP 0x00000098
+#define A4XX_RBBM_POWER_CNTL_IP_SW_COLLAPSE 0x00000001
+
#define REG_A4XX_RBBM_PERFCTR_CP_0_LO 0x0000009c
static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL_SP(uint32_t i0) { return 0x00000068 + 0x1*i0; }
static inline uint32_t REG_A4XX_RBBM_CLOCK_DELAY_RB_MARB_CCU_L1_REG(uint32_t i0) { return 0x0000008e + 0x1*i0; }
+#define REG_A4XX_RBBM_SP_REGFILE_SLEEP_CNTL_0 0x00000099
+
+#define REG_A4XX_RBBM_SP_REGFILE_SLEEP_CNTL_1 0x0000009a
+
#define REG_A4XX_RBBM_PERFCTR_PWR_1_LO 0x00000168
#define REG_A4XX_RBBM_PERFCTR_CTL 0x00000170
#define REG_A4XX_RBBM_INTERFACE_RRDY_STATUS5 0x0000019f
+#define REG_A4XX_RBBM_POWER_STATUS 0x000001b0
+#define A4XX_RBBM_POWER_STATUS_SP_TP_PWR_ON 0x00100000
+
+#define REG_A4XX_RBBM_WAIT_IDLE_CLOCKS_CTL2 0x000001b8
+
#define REG_A4XX_CP_SCRATCH_UMASK 0x00000228
#define REG_A4XX_CP_SCRATCH_ADDR 0x00000229
#define REG_A4XX_SP_MODE_CONTROL 0x00000ec3
+#define REG_A4XX_SP_PERFCTR_SP_SEL_0 0x00000ec4
+
+#define REG_A4XX_SP_PERFCTR_SP_SEL_1 0x00000ec5
+
+#define REG_A4XX_SP_PERFCTR_SP_SEL_2 0x00000ec6
+
+#define REG_A4XX_SP_PERFCTR_SP_SEL_3 0x00000ec7
+
+#define REG_A4XX_SP_PERFCTR_SP_SEL_4 0x00000ec8
+
+#define REG_A4XX_SP_PERFCTR_SP_SEL_5 0x00000ec9
+
+#define REG_A4XX_SP_PERFCTR_SP_SEL_6 0x00000eca
+
+#define REG_A4XX_SP_PERFCTR_SP_SEL_7 0x00000ecb
+
+#define REG_A4XX_SP_PERFCTR_SP_SEL_8 0x00000ecc
+
+#define REG_A4XX_SP_PERFCTR_SP_SEL_9 0x00000ecd
+
+#define REG_A4XX_SP_PERFCTR_SP_SEL_10 0x00000ece
+
#define REG_A4XX_SP_PERFCTR_SP_SEL_11 0x00000ecf
#define REG_A4XX_SP_SP_CTRL_REG 0x000022c0
#define REG_A4XX_GRAS_ALPHA_CONTROL 0x00002073
#define A4XX_GRAS_ALPHA_CONTROL_ALPHA_TEST_ENABLE 0x00000004
+#define A4XX_GRAS_ALPHA_CONTROL_FORCE_FRAGZ_TO_FS 0x00000008
#define REG_A4XX_GRAS_SU_POLY_OFFSET_SCALE 0x00002074
#define A4XX_GRAS_SU_POLY_OFFSET_SCALE__MASK 0xffffffff
fd4_query_context_init(pctx);
- fd4_ctx->border_color_uploader = u_upload_create(pctx, 4096,
- 2 * PIPE_MAX_SAMPLERS * BORDERCOLOR_SIZE, 0);
+ fd4_ctx->border_color_uploader = u_upload_create(pctx, 4096, 0,
+ PIPE_USAGE_STREAM);
return pctx;
}
void *ptr;
u_upload_alloc(fd4_ctx->border_color_uploader,
- 0, 2 * PIPE_MAX_SAMPLERS * BORDERCOLOR_SIZE, &off,
+ 0, BORDER_COLOR_UPLOAD_SIZE,
+ BORDER_COLOR_UPLOAD_SIZE, &off,
&fd4_ctx->border_color_buf,
&ptr);
OUT_PKT0(ring, REG_A4XX_RB_DEPTH_CONTROL, 1);
OUT_RING(ring, zsa->rb_depth_control |
- COND(fragz, A4XX_RB_DEPTH_CONTROL_EARLY_Z_DISABLE));
+ COND(fragz, A4XX_RB_DEPTH_CONTROL_EARLY_Z_DISABLE) |
+ COND(fragz && fp->frag_coord, A4XX_RB_DEPTH_CONTROL_FORCE_FRAGZ_TO_FS));
/* maybe this register/bitfield needs a better name.. this
* appears to be just disabling early-z
*/
OUT_PKT0(ring, REG_A4XX_GRAS_ALPHA_CONTROL, 1);
OUT_RING(ring, zsa->gras_alpha_control |
- COND(fragz, A4XX_GRAS_ALPHA_CONTROL_ALPHA_TEST_ENABLE));
+ COND(fragz, A4XX_GRAS_ALPHA_CONTROL_ALPHA_TEST_ENABLE) |
+ COND(fragz && fp->frag_coord, A4XX_GRAS_ALPHA_CONTROL_FORCE_FRAGZ_TO_FS));
}
if (dirty & FD_DIRTY_RASTERIZER) {
COND(s[FS].v->frag_face, A4XX_RB_RENDER_CONTROL2_FACENESS) |
COND(s[FS].v->frag_coord, A4XX_RB_RENDER_CONTROL2_XCOORD |
A4XX_RB_RENDER_CONTROL2_YCOORD |
-// TODO enabling gl_FragCoord.z is causing lockups on 0ad (but seems
-// to work everywhere else).
-// A4XX_RB_RENDER_CONTROL2_ZCOORD |
+ A4XX_RB_RENDER_CONTROL2_ZCOORD |
A4XX_RB_RENDER_CONTROL2_WCOORD));
OUT_PKT0(ring, REG_A4XX_RB_FS_OUTPUT_REG, 1);
- /home/robclark/src/freedreno/envytools/rnndb/adreno.xml ( 398 bytes, from 2015-09-24 17:25:31)
- /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml ( 1453 bytes, from 2015-05-20 20:03:07)
- /home/robclark/src/freedreno/envytools/rnndb/adreno/a2xx.xml ( 32901 bytes, from 2015-05-20 20:03:14)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml ( 10755 bytes, from 2015-09-14 20:46:55)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml ( 14968 bytes, from 2015-05-20 20:12:27)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml ( 68291 bytes, from 2015-11-17 16:39:59)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml ( 64038 bytes, from 2015-11-17 16:37:36)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml ( 11518 bytes, from 2015-11-24 14:39:00)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml ( 15149 bytes, from 2015-11-20 16:22:25)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml ( 69600 bytes, from 2015-11-24 14:39:00)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml ( 67220 bytes, from 2015-12-13 17:58:09)
- /home/robclark/src/freedreno/envytools/rnndb/adreno/ocmem.xml ( 1773 bytes, from 2015-09-24 17:30:00)
Copyright (C) 2013-2015 by the following authors:
- /home/robclark/src/freedreno/envytools/rnndb/adreno.xml ( 398 bytes, from 2015-09-24 17:25:31)
- /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml ( 1453 bytes, from 2015-05-20 20:03:07)
- /home/robclark/src/freedreno/envytools/rnndb/adreno/a2xx.xml ( 32901 bytes, from 2015-05-20 20:03:14)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml ( 10755 bytes, from 2015-09-14 20:46:55)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml ( 14968 bytes, from 2015-05-20 20:12:27)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml ( 68291 bytes, from 2015-11-17 16:39:59)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml ( 64038 bytes, from 2015-11-17 16:37:36)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml ( 11518 bytes, from 2015-11-24 14:39:00)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml ( 15149 bytes, from 2015-11-20 16:22:25)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml ( 69600 bytes, from 2015-11-24 14:39:00)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml ( 67220 bytes, from 2015-12-13 17:58:09)
- /home/robclark/src/freedreno/envytools/rnndb/adreno/ocmem.xml ( 1773 bytes, from 2015-09-24 17:30:00)
Copyright (C) 2013-2015 by the following authors:
enum adreno_state_src {
SS_DIRECT = 0,
+ SS_INVALID_ALL_IC = 2,
+ SS_INVALID_PART_IC = 3,
SS_INDIRECT = 4,
+ SS_INDIRECT_TCM = 5,
+ SS_INDIRECT_STM = 6,
};
enum a4xx_index_size {
{
return ((val) << CP_LOAD_STATE_0_STATE_BLOCK__SHIFT) & CP_LOAD_STATE_0_STATE_BLOCK__MASK;
}
-#define CP_LOAD_STATE_0_NUM_UNIT__MASK 0x7fc00000
+#define CP_LOAD_STATE_0_NUM_UNIT__MASK 0xffc00000
#define CP_LOAD_STATE_0_NUM_UNIT__SHIFT 22
static inline uint32_t CP_LOAD_STATE_0_NUM_UNIT(uint32_t val)
{
#include "freedreno_gmem.h"
#include "freedreno_util.h"
+#define BORDER_COLOR_UPLOAD_SIZE (2 * PIPE_MAX_SAMPLERS * BORDERCOLOR_SIZE)
+
struct fd_vertex_stateobj;
struct fd_texture_stateobj {
case PIPE_CAP_TEXTURE_GATHER_OFFSETS:
case PIPE_CAP_TGSI_VS_WINDOW_SPACE_POSITION:
case PIPE_CAP_DRAW_INDIRECT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT_PARAMS:
case PIPE_CAP_TGSI_FS_FINE_DERIVATIVE:
case PIPE_CAP_POLYGON_OFFSET_CLAMP:
case PIPE_CAP_MULTISAMPLE_Z_RESOLVE:
case PIPE_CAP_SHAREABLE_SHADERS:
case PIPE_CAP_COPY_BETWEEN_COMPRESSED_AND_PLAIN_FORMATS:
case PIPE_CAP_CLEAR_TEXTURE:
+ case PIPE_CAP_DRAW_PARAMETERS:
+ case PIPE_CAP_TGSI_PACK_HALF_FLOAT:
+ case PIPE_CAP_TGSI_FS_POSITION_IS_SYSVAL:
+ case PIPE_CAP_TGSI_FS_FACE_IS_INTEGER_SYSVAL:
+ case PIPE_CAP_SHADER_BUFFER_OFFSET_ALIGNMENT:
+ case PIPE_CAP_INVALIDATE_BUFFER:
+ case PIPE_CAP_GENERATE_MIPMAP:
return 0;
case PIPE_CAP_MAX_VIEWPORTS:
return PIPE_SHADER_IR_TGSI;
case PIPE_SHADER_CAP_MAX_UNROLL_ITERATIONS_HINT:
return 32;
+ case PIPE_SHADER_CAP_MAX_SHADER_BUFFERS:
+ return 0;
}
debug_printf("unknown shader param %d\n", param);
return 0;
#include "freedreno_util.h"
#include "ir3_compiler.h"
+#include "ir3_nir.h"
#include "instr-a3xx.h"
#include "ir3.h"
const char *filename;
struct tgsi_token toks[65536];
struct tgsi_parse_context parse;
- struct ir3_compiler *compiler;
struct ir3_shader_variant v;
struct ir3_shader s;
struct ir3_shader_key key = {};
+ /* TODO cmdline option to target different gpus: */
unsigned gpu_id = 320;
const char *info;
void *ptr;
size_t size;
- fd_mesa_debug |= FD_DBG_DISASM;
-
memset(&s, 0, sizeof(s));
memset(&v, 0, sizeof(v));
while (n < argc) {
if (!strcmp(argv[n], "--verbose")) {
- fd_mesa_debug |= FD_DBG_MSGS | FD_DBG_OPTMSGS;
+ fd_mesa_debug |= FD_DBG_MSGS | FD_DBG_OPTMSGS | FD_DBG_DISASM;
n++;
continue;
}
if (!tgsi_text_translate(ptr, toks, Elements(toks)))
errx(1, "could not parse `%s'", filename);
- s.tokens = toks;
+ if (fd_mesa_debug & FD_DBG_OPTMSGS)
+ tgsi_dump(toks, 0);
+
+ nir_shader *nir = ir3_tgsi_to_nir(toks);
+ s.compiler = ir3_compiler_create(gpu_id);
+ s.nir = ir3_optimize_nir(&s, nir, NULL);
v.key = key;
v.shader = &s;
break;
}
- /* TODO cmdline option to target different gpus: */
- compiler = ir3_compiler_create(gpu_id);
-
info = "NIR compiler";
- ret = ir3_compile_shader_nir(compiler, &v);
+ ret = ir3_compile_shader_nir(s.compiler, &v);
if (ret) {
fprintf(stderr, "compiler failed!\n");
return ret;
#include "util/u_string.h"
#include "util/u_memory.h"
#include "util/u_inlines.h"
-#include "tgsi/tgsi_lowering.h"
-#include "tgsi/tgsi_strings.h"
-
-#include "nir/tgsi_to_nir.h"
#include "freedreno_util.h"
static struct ir3_instruction * create_immed(struct ir3_block *block, uint32_t val);
static struct ir3_block * get_block(struct ir3_compile *ctx, nir_block *nblock);
-static struct nir_shader *to_nir(struct ir3_compile *ctx,
- const struct tgsi_token *tokens, struct ir3_shader_variant *so)
-{
- static const nir_shader_compiler_options options = {
- .lower_fpow = true,
- .lower_fsat = true,
- .lower_scmp = true,
- .lower_flrp = true,
- .lower_ffract = true,
- .native_integers = true,
- };
- struct nir_lower_tex_options tex_options = {
- .lower_rect = 0,
- };
- bool progress;
-
- switch (so->type) {
- case SHADER_FRAGMENT:
- case SHADER_COMPUTE:
- tex_options.saturate_s = so->key.fsaturate_s;
- tex_options.saturate_t = so->key.fsaturate_t;
- tex_options.saturate_r = so->key.fsaturate_r;
- break;
- case SHADER_VERTEX:
- tex_options.saturate_s = so->key.vsaturate_s;
- tex_options.saturate_t = so->key.vsaturate_t;
- tex_options.saturate_r = so->key.vsaturate_r;
- break;
- }
-
- if (ctx->compiler->gpu_id >= 400) {
- /* a4xx seems to have *no* sam.p */
- tex_options.lower_txp = ~0; /* lower all txp */
- } else {
- /* a3xx just needs to avoid sam.p for 3d tex */
- tex_options.lower_txp = (1 << GLSL_SAMPLER_DIM_3D);
- }
-
- struct nir_shader *s = tgsi_to_nir(tokens, &options);
-
- if (fd_mesa_debug & FD_DBG_DISASM) {
- debug_printf("----------------------\n");
- nir_print_shader(s, stdout);
- debug_printf("----------------------\n");
- }
-
- nir_opt_global_to_local(s);
- nir_convert_to_ssa(s);
- if (s->stage == MESA_SHADER_VERTEX) {
- nir_lower_clip_vs(s, so->key.ucp_enables);
- } else if (s->stage == MESA_SHADER_FRAGMENT) {
- nir_lower_clip_fs(s, so->key.ucp_enables);
- }
- nir_lower_tex(s, &tex_options);
- if (so->key.color_two_side)
- nir_lower_two_sided_color(s);
- nir_lower_idiv(s);
- nir_lower_load_const_to_scalar(s);
-
- do {
- progress = false;
-
- nir_lower_vars_to_ssa(s);
- nir_lower_alu_to_scalar(s);
- nir_lower_phis_to_scalar(s);
-
- progress |= nir_copy_prop(s);
- progress |= nir_opt_dce(s);
- progress |= nir_opt_cse(s);
- progress |= ir3_nir_lower_if_else(s);
- progress |= nir_opt_algebraic(s);
- progress |= nir_opt_constant_folding(s);
-
- } while (progress);
-
- nir_remove_dead_variables(s);
- nir_validate_shader(s);
-
- if (fd_mesa_debug & FD_DBG_DISASM) {
- debug_printf("----------------------\n");
- nir_print_shader(s, stdout);
- debug_printf("----------------------\n");
- }
-
- return s;
-}
static struct ir3_compile *
compile_init(struct ir3_compiler *compiler,
- struct ir3_shader_variant *so,
- const struct tgsi_token *tokens)
+ struct ir3_shader_variant *so)
{
struct ir3_compile *ctx = rzalloc(NULL, struct ir3_compile);
ctx->block_ht = _mesa_hash_table_create(ctx,
_mesa_hash_pointer, _mesa_key_pointer_equal);
- ctx->s = to_nir(ctx, tokens, so);
+ /* TODO: maybe generate some sort of bitmask of what key
+ * lowers vs what shader has (ie. no need to lower
+ * texture clamp lowering if no texture sample instrs)..
+ * although should be done further up the stack to avoid
+ * creating duplicate variants..
+ */
+
+ if (ir3_key_lowers_nir(&so->key)) {
+ nir_shader *s = nir_shader_clone(ctx, so->shader->nir);
+ ctx->s = ir3_optimize_nir(so->shader, s, &so->key);
+ } else {
+ /* fast-path for shader key that lowers nothing in NIR: */
+ ctx->s = so->shader->nir;
+ }
+
+ if (fd_mesa_debug & FD_DBG_DISASM) {
+ DBG("dump nir%dv%d: type=%d, k={bp=%u,cts=%u,hp=%u}",
+ so->shader->id, so->id, so->type,
+ so->key.binning_pass, so->key.color_two_side,
+ so->key.half_precision);
+ nir_print_shader(ctx->s, stdout);
+ }
so->first_driver_param = so->first_immediate = ctx->s->num_uniforms;
/* handles array reads: */
static void
-emit_intrinisic_load_var(struct ir3_compile *ctx, nir_intrinsic_instr *intr,
+emit_intrinsic_load_var(struct ir3_compile *ctx, nir_intrinsic_instr *intr,
struct ir3_instruction **dst)
{
nir_deref_var *dvar = intr->variables[0];
/* handles array writes: */
static void
-emit_intrinisic_store_var(struct ir3_compile *ctx, nir_intrinsic_instr *intr)
+emit_intrinsic_store_var(struct ir3_compile *ctx, nir_intrinsic_instr *intr)
{
nir_deref_var *dvar = intr->variables[0];
nir_deref_array *darr = nir_deref_as_array(dvar->deref.child);
case nir_deref_array_type_direct:
/* direct access does not require anything special: */
for (int i = 0; i < intr->num_components; i++) {
+ /* ttn doesn't generate partial writemasks */
+ assert(intr->const_index[0] ==
+ (1 << intr->num_components) - 1);
+
unsigned n = darr->base_offset * 4 + i;
compile_assert(ctx, n < arr->length);
arr->arr[n] = src[i];
struct ir3_instruction *addr =
get_addr(ctx, get_src(ctx, &darr->indirect)[0]);
for (int i = 0; i < intr->num_components; i++) {
+ /* ttn doesn't generate partial writemasks */
+ assert(intr->const_index[0] ==
+ (1 << intr->num_components) - 1);
+
struct ir3_instruction *store;
unsigned n = darr->base_offset * 4 + i;
compile_assert(ctx, n < arr->length);
}
static void
-emit_intrinisic(struct ir3_compile *ctx, nir_intrinsic_instr *intr)
+emit_intrinsic(struct ir3_compile *ctx, nir_intrinsic_instr *intr)
{
const nir_intrinsic_info *info = &nir_intrinsic_infos[intr->intrinsic];
struct ir3_instruction **dst, **src;
}
break;
case nir_intrinsic_load_var:
- emit_intrinisic_load_var(ctx, intr, dst);
+ emit_intrinsic_load_var(ctx, intr, dst);
break;
case nir_intrinsic_store_var:
- emit_intrinisic_store_var(ctx, intr);
+ emit_intrinsic_store_var(ctx, intr);
break;
case nir_intrinsic_store_output:
const_offset = nir_src_as_const_value(intr->src[1]);
emit_alu(ctx, nir_instr_as_alu(instr));
break;
case nir_instr_type_intrinsic:
- emit_intrinisic(ctx, nir_instr_as_intrinsic(instr));
+ emit_intrinsic(ctx, nir_instr_as_intrinsic(instr));
break;
case nir_instr_type_load_const:
emit_load_const(ctx, nir_instr_as_load_const(instr));
case nir_texop_query_levels:
emit_tex_query_levels(ctx, tex);
break;
- case nir_texop_samples_identical:
- unreachable("nir_texop_samples_identical");
default:
emit_tex(ctx, tex);
break;
static void
emit_function(struct ir3_compile *ctx, nir_function_impl *impl)
{
+ nir_metadata_require(impl, nir_metadata_block_index);
+
emit_cf_list(ctx, &impl->body);
emit_block(ctx, impl->end_block);
nir_function_impl *fxn = NULL;
/* Find the main function: */
- nir_foreach_overload(ctx->s, overload) {
- compile_assert(ctx, strcmp(overload->function->name, "main") == 0);
- compile_assert(ctx, overload->impl);
- fxn = overload->impl;
+ nir_foreach_function(ctx->s, function) {
+ compile_assert(ctx, strcmp(function->name, "main") == 0);
+ compile_assert(ctx, function->impl);
+ fxn = function->impl;
break;
}
assert(!so->ir);
- ctx = compile_init(compiler, so, so->shader->tokens);
+ ctx = compile_init(compiler, so);
if (!ctx) {
DBG("INIT failed!");
ret = -1;
--- /dev/null
+/* -*- mode: C; c-file-style: "k&r"; tab-width 4; indent-tabs-mode: t; -*- */
+
+/*
+ * Copyright (C) 2015 Rob Clark <robclark@freedesktop.org>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ * Authors:
+ * Rob Clark <robclark@freedesktop.org>
+ */
+
+
+#include "freedreno_util.h"
+
+#include "ir3_nir.h"
+#include "ir3_compiler.h"
+#include "ir3_shader.h"
+
+#include "nir/tgsi_to_nir.h"
+
+struct nir_shader *
+ir3_tgsi_to_nir(const struct tgsi_token *tokens)
+{
+ static const nir_shader_compiler_options options = {
+ .lower_fpow = true,
+ .lower_fsat = true,
+ .lower_scmp = true,
+ .lower_flrp = true,
+ .lower_ffract = true,
+ .native_integers = true,
+ };
+ return tgsi_to_nir(tokens, &options);
+}
+
+/* for given shader key, are any steps handled in nir? */
+bool
+ir3_key_lowers_nir(const struct ir3_shader_key *key)
+{
+ return key->fsaturate_s | key->fsaturate_t | key->fsaturate_r |
+ key->vsaturate_s | key->vsaturate_t | key->vsaturate_r |
+ key->ucp_enables | key->color_two_side;
+}
+
+#define OPT(nir, pass, ...) ({ \
+ bool this_progress = false; \
+ NIR_PASS(this_progress, nir, pass, ##__VA_ARGS__); \
+ this_progress; \
+})
+
+#define OPT_V(nir, pass, ...) NIR_PASS_V(nir, pass, ##__VA_ARGS__)
+
+struct nir_shader *
+ir3_optimize_nir(struct ir3_shader *shader, nir_shader *s,
+ const struct ir3_shader_key *key)
+{
+ struct nir_lower_tex_options tex_options = {
+ .lower_rect = 0,
+ };
+ bool progress;
+
+ if (key) {
+ switch (shader->type) {
+ case SHADER_FRAGMENT:
+ case SHADER_COMPUTE:
+ tex_options.saturate_s = key->fsaturate_s;
+ tex_options.saturate_t = key->fsaturate_t;
+ tex_options.saturate_r = key->fsaturate_r;
+ break;
+ case SHADER_VERTEX:
+ tex_options.saturate_s = key->vsaturate_s;
+ tex_options.saturate_t = key->vsaturate_t;
+ tex_options.saturate_r = key->vsaturate_r;
+ break;
+ }
+ }
+
+ if (shader->compiler->gpu_id >= 400) {
+ /* a4xx seems to have *no* sam.p */
+ tex_options.lower_txp = ~0; /* lower all txp */
+ } else {
+ /* a3xx just needs to avoid sam.p for 3d tex */
+ tex_options.lower_txp = (1 << GLSL_SAMPLER_DIM_3D);
+ }
+
+ if (fd_mesa_debug & FD_DBG_DISASM) {
+ debug_printf("----------------------\n");
+ nir_print_shader(s, stdout);
+ debug_printf("----------------------\n");
+ }
+
+ OPT_V(s, nir_opt_global_to_local);
+ OPT_V(s, nir_convert_to_ssa);
+
+ if (key) {
+ if (s->stage == MESA_SHADER_VERTEX) {
+ OPT_V(s, nir_lower_clip_vs, key->ucp_enables);
+ } else if (s->stage == MESA_SHADER_FRAGMENT) {
+ OPT_V(s, nir_lower_clip_fs, key->ucp_enables);
+ }
+ if (key->color_two_side) {
+ OPT_V(s, nir_lower_two_sided_color);
+ }
+ }
+
+ OPT_V(s, nir_lower_tex, &tex_options);
+ OPT_V(s, nir_lower_idiv);
+ OPT_V(s, nir_lower_load_const_to_scalar);
+
+ do {
+ progress = false;
+
+ OPT_V(s, nir_lower_vars_to_ssa);
+ OPT_V(s, nir_lower_alu_to_scalar);
+ OPT_V(s, nir_lower_phis_to_scalar);
+
+ progress |= OPT(s, nir_copy_prop);
+ progress |= OPT(s, nir_opt_dce);
+ progress |= OPT(s, nir_opt_cse);
+ progress |= OPT(s, ir3_nir_lower_if_else);
+ progress |= OPT(s, nir_opt_algebraic);
+ progress |= OPT(s, nir_opt_constant_folding);
+
+ } while (progress);
+
+ OPT_V(s, nir_remove_dead_variables);
+
+ if (fd_mesa_debug & FD_DBG_DISASM) {
+ debug_printf("----------------------\n");
+ nir_print_shader(s, stdout);
+ debug_printf("----------------------\n");
+ }
+
+ nir_sweep(s);
+
+ return s;
+}
#include "glsl/nir/nir.h"
#include "glsl/nir/shader_enums.h"
+#include "ir3_shader.h"
+
bool ir3_nir_lower_if_else(nir_shader *shader);
+struct nir_shader * ir3_tgsi_to_nir(const struct tgsi_token *tokens);
+bool ir3_key_lowers_nir(const struct ir3_shader_key *key);
+struct nir_shader * ir3_optimize_nir(struct ir3_shader *shader, nir_shader *s,
+ const struct ir3_shader_key *key);
+
#endif /* IR3_NIR_H_ */
{
bool progress = false;
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- progress |= lower_if_else_impl(overload->impl);
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ progress |= lower_if_else_impl(function->impl);
}
return progress;
#ifdef DEBUG
return block->serialno;
#else
- return (uint32_t)(uint64_t)block;
+ return (uint32_t)(unsigned long)block;
#endif
}
#include "ir3_shader.h"
#include "ir3_compiler.h"
-
+#include "ir3_nir.h"
static void
delete_variant(struct ir3_shader_variant *v)
v->key = key;
v->type = shader->type;
- if (fd_mesa_debug & FD_DBG_DISASM) {
- DBG("dump tgsi: type=%d, k={bp=%u,cts=%u,hp=%u}", shader->type,
- key.binning_pass, key.color_two_side, key.half_precision);
- tgsi_dump(shader->tokens, 0);
- }
-
ret = ir3_compile_shader_nir(shader->compiler, v);
if (ret) {
debug_error("compile failed!");
v = v->next;
delete_variant(t);
}
- free((void *)shader->tokens);
+ ralloc_free(shader->nir);
free(shader);
}
shader->id = ++shader->compiler->shader_count;
shader->pctx = pctx;
shader->type = type;
- shader->tokens = tgsi_dup_tokens(cso->tokens);
+ if (fd_mesa_debug & FD_DBG_DISASM) {
+ DBG("dump tgsi: type=%d", shader->type);
+ tgsi_dump(cso->tokens, 0);
+ }
+ nir_shader *nir = ir3_tgsi_to_nir(cso->tokens);
+ /* do first pass optimization, ignoring the key: */
+ shader->nir = ir3_optimize_nir(shader, nir, NULL);
+ if (fd_mesa_debug & FD_DBG_DISASM) {
+ DBG("dump nir%d: type=%d", shader->id, shader->type);
+ nir_print_shader(shader->nir, stdout);
+ }
shader->stream_output = cso->stream_output;
if (fd_mesa_debug & FD_DBG_SHADERDB) {
/* if shader-db run, create a standard variant immediately
* (as otherwise nothing will trigger the shader to be
* actually compiled)
*/
- static struct ir3_shader_key key = {};
+ static struct ir3_shader_key key = {0};
ir3_shader_variant(shader, key);
}
return shader;
struct ir3_shader *shader;
};
+typedef struct nir_shader nir_shader;
+
struct ir3_shader {
enum shader_t type;
struct ir3_compiler *compiler;
struct pipe_context *pctx; /* TODO replace w/ pipe_screen */
- const struct tgsi_token *tokens;
+ nir_shader *nir;
struct pipe_stream_output_info stream_output;
struct ir3_shader_variant *variants;
unsigned light_twoside : 1;
unsigned st;
- enum interp_mode color_interp;
unsigned LIS4;
unsigned LIS7;
case PIPE_CAP_SHAREABLE_SHADERS:
case PIPE_CAP_COPY_BETWEEN_COMPRESSED_AND_PLAIN_FORMATS:
case PIPE_CAP_CLEAR_TEXTURE:
+ case PIPE_CAP_DRAW_PARAMETERS:
+ case PIPE_CAP_TGSI_PACK_HALF_FLOAT:
+ case PIPE_CAP_TGSI_FS_POSITION_IS_SYSVAL:
+ case PIPE_CAP_TGSI_FS_FACE_IS_INTEGER_SYSVAL:
+ case PIPE_CAP_SHADER_BUFFER_OFFSET_ALIGNMENT:
+ case PIPE_CAP_INVALIDATE_BUFFER:
+ case PIPE_CAP_GENERATE_MIPMAP:
return 0;
case PIPE_CAP_MAX_DUAL_SOURCE_RENDER_TARGETS:
case PIPE_CAP_TGSI_VS_LAYER_VIEWPORT:
case PIPE_CAP_BUFFER_MAP_PERSISTENT_COHERENT:
case PIPE_CAP_DRAW_INDIRECT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT_PARAMS:
case PIPE_CAP_TGSI_FS_FINE_DERIVATIVE:
case PIPE_CAP_SAMPLER_VIEW_TARGET:
return 0;
for (j = view->u.tex.first_level; j <= tex->last_level; j++) {
mip_offsets[j] = i915_texture_offset(i915_tex, j , 0 /* FIXME depth */);
row_stride[j] = i915_tex->stride;
- img_stride[j] = 0; /* FIXME */;
+ img_stride[j] = 0; /* FIXME */
}
draw_set_mapped_texture(i915->draw,
struct i915_rasterizer_state *cso = CALLOC_STRUCT( i915_rasterizer_state );
cso->templ = *rasterizer;
- cso->color_interp = rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR;
cso->light_twoside = rasterizer->light_twoside;
cso->ds[0].u = _3DSTATE_DEPTH_OFFSET_SCALE;
cso->ds[1].f = rasterizer->offset_scale;
static void calculate_vertex_layout(struct i915_context *i915)
{
const struct i915_fragment_shader *fs = i915->fs;
- const enum interp_mode colorInterp = i915->rasterizer->color_interp;
struct vertex_info vinfo;
boolean texCoords[I915_TEX_UNITS], colors[2], fog, needW, face;
uint i;
/* pos */
src = draw_find_shader_output(i915->draw, TGSI_SEMANTIC_POSITION, 0);
if (needW) {
- draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_LINEAR, src);
+ draw_emit_vertex_attr(&vinfo, EMIT_4F, src);
vinfo.hwfmt[0] |= S4_VFMT_XYZW;
vinfo.attrib[0].emit = EMIT_4F;
}
else {
- draw_emit_vertex_attr(&vinfo, EMIT_3F, INTERP_LINEAR, src);
+ draw_emit_vertex_attr(&vinfo, EMIT_3F, src);
vinfo.hwfmt[0] |= S4_VFMT_XYZ;
vinfo.attrib[0].emit = EMIT_3F;
}
/* primary color */
if (colors[0]) {
src = draw_find_shader_output(i915->draw, TGSI_SEMANTIC_COLOR, 0);
- draw_emit_vertex_attr(&vinfo, EMIT_4UB_BGRA, colorInterp, src);
+ draw_emit_vertex_attr(&vinfo, EMIT_4UB_BGRA, src);
vinfo.hwfmt[0] |= S4_VFMT_COLOR;
}
/* secondary color */
if (colors[1]) {
src = draw_find_shader_output(i915->draw, TGSI_SEMANTIC_COLOR, 1);
- draw_emit_vertex_attr(&vinfo, EMIT_4UB_BGRA, colorInterp, src);
+ draw_emit_vertex_attr(&vinfo, EMIT_4UB_BGRA, src);
vinfo.hwfmt[0] |= S4_VFMT_SPEC_FOG;
}
/* fog coord, not fog blend factor */
if (fog) {
src = draw_find_shader_output(i915->draw, TGSI_SEMANTIC_FOG, 0);
- draw_emit_vertex_attr(&vinfo, EMIT_1F, INTERP_PERSPECTIVE, src);
+ draw_emit_vertex_attr(&vinfo, EMIT_1F, src);
vinfo.hwfmt[0] |= S4_VFMT_FOG_PARAM;
}
if (texCoords[i]) {
hwtc = TEXCOORDFMT_4D;
src = draw_find_shader_output(i915->draw, TGSI_SEMANTIC_GENERIC, fs->generic_mapping[i]);
- draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_PERSPECTIVE, src);
+ draw_emit_vertex_attr(&vinfo, EMIT_4F, src);
}
else {
hwtc = TEXCOORDFMT_NOT_PRESENT;
* module by adding an extra shader output.
*/
src = draw_find_shader_output(i915->draw, TGSI_SEMANTIC_FACE, 0);
- draw_emit_vertex_attr(&vinfo, EMIT_1F, INTERP_CONSTANT, src);
+ draw_emit_vertex_attr(&vinfo, EMIT_1F, src);
vinfo.hwfmt[1] &= ~(TEXCOORDFMT_NOT_PRESENT << (slot * 4));
vinfo.hwfmt[1] |= TEXCOORDFMT_1D << (slot * 4);
}
struct i915_tracked_state i915_update_vertex_layout = {
"vertex_layout",
calculate_vertex_layout,
- I915_NEW_RASTERIZER | I915_NEW_FS | I915_NEW_VS
+ I915_NEW_FS | I915_NEW_VS
};
const struct ilo_dev *dev,
struct intel_winsys *winsys)
{
- int i;
+ unsigned i;
assert(ilo_is_zeroed(builder, sizeof(*builder)));
void
ilo_builder_reset(struct ilo_builder *builder)
{
- int i;
+ unsigned i;
for (i = 0; i < ILO_BUILDER_WRITER_COUNT; i++)
ilo_builder_writer_reset(builder, i);
bool
ilo_builder_begin(struct ilo_builder *builder)
{
- int i;
+ unsigned i;
for (i = 0; i < ILO_BUILDER_WRITER_COUNT; i++) {
if (!ilo_builder_writer_alloc_and_map(builder, i)) {
ilo_builder_end(struct ilo_builder *builder, unsigned *used)
{
struct ilo_builder_writer *bat;
- int i;
+ unsigned i;
ilo_builder_batch_patch_sba(builder);
* These must be called last as u_upload/u_blitter are clients of the pipe
* context.
*/
- ilo->uploader = u_upload_create(&ilo->base, 1024 * 1024, 16,
- PIPE_BIND_CONSTANT_BUFFER | PIPE_BIND_INDEX_BUFFER);
+ ilo->uploader = u_upload_create(&ilo->base, 1024 * 1024,
+ PIPE_BIND_CONSTANT_BUFFER | PIPE_BIND_INDEX_BUFFER,
+ PIPE_USAGE_STREAM);
if (!ilo->uploader) {
ilo_context_destroy(&ilo->base);
return NULL;
input_buf.buffer_size =
ilo_shader_get_kernel_param(cs, ILO_KERNEL_CS_INPUT_SIZE);
if (input_buf.buffer_size) {
- u_upload_data(ilo->uploader, 0, input_buf.buffer_size, input,
+ u_upload_data(ilo->uploader, 0, input_buf.buffer_size, 16, input,
&input_buf.buffer_offset, &input_buf.buffer);
}
case PIPE_CAP_TGSI_VS_WINDOW_SPACE_POSITION:
case PIPE_CAP_MAX_VERTEX_STREAMS:
case PIPE_CAP_DRAW_INDIRECT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT_PARAMS:
case PIPE_CAP_TGSI_FS_FINE_DERIVATIVE:
case PIPE_CAP_CONDITIONAL_RENDER_INVERTED:
case PIPE_CAP_SAMPLER_VIEW_TARGET:
case PIPE_CAP_SHAREABLE_SHADERS:
case PIPE_CAP_COPY_BETWEEN_COMPRESSED_AND_PLAIN_FORMATS:
case PIPE_CAP_CLEAR_TEXTURE:
+ case PIPE_CAP_DRAW_PARAMETERS:
+ case PIPE_CAP_TGSI_PACK_HALF_FLOAT:
+ case PIPE_CAP_TGSI_FS_POSITION_IS_SYSVAL:
+ case PIPE_CAP_TGSI_FS_FACE_IS_INTEGER_SYSVAL:
+ case PIPE_CAP_SHADER_BUFFER_OFFSET_ALIGNMENT:
+ case PIPE_CAP_INVALIDATE_BUFFER:
+ case PIPE_CAP_GENERATE_MIPMAP:
return 0;
case PIPE_CAP_VENDOR_ID:
if (cbuf->cso[i].resource)
continue;
- u_upload_data(ilo->uploader, 0, cbuf->cso[i].info.size,
+ u_upload_data(ilo->uploader, 0, cbuf->cso[i].info.size, 16,
cbuf->cso[i].user_buffer, &offset, &cbuf->cso[i].resource);
cbuf->cso[i].info.vma = ilo_resource_get_vma(cbuf->cso[i].resource);
unsigned hw_offset;
if (vec->ib.state.user_buffer) {
- u_upload_data(ilo->uploader, 0, size,
+ u_upload_data(ilo->uploader, 0, size, 16,
vec->ib.state.user_buffer + offset,
&hw_offset, &vec->ib.hw_resource);
} else {
u_upload_buffer(ilo->uploader, 0,
- vec->ib.state.offset + offset, size, vec->ib.state.buffer,
+ vec->ib.state.offset + offset, size, 16, vec->ib.state.buffer,
&hw_offset, &vec->ib.hw_resource);
}
struct toy_inst *inst;
struct toy_src desc, real_src[4];
struct toy_dst tmp, real_dst[4];
- int i;
+ unsigned i;
tsrc_transpose(idx, real_src);
const int grf_subreg = (idx.val32 & 1) * 16;
struct toy_src src;
struct toy_dst real_dst[4];
- int i;
+ unsigned i;
if (!fcc->variant->use_pcb || dim != 0 || idx.file != TOY_FILE_IMM ||
grf >= fcc->first_attr_grf)
struct toy_inst *inst;
struct toy_src desc;
struct toy_dst tmp, real_dst[4];
- int i;
+ unsigned i;
if (fs_lower_opcode_tgsi_const_pcb(fcc, dst, dim, idx))
return;
struct toy_src desc;
struct toy_inst *inst;
struct toy_dst tmp, real_dst[4];
- int i;
+ unsigned i;
if (fs_lower_opcode_tgsi_const_pcb(fcc, dst, dim, idx))
return;
{
struct toy_dst dst[4];
struct toy_src src[4];
- int i;
+ unsigned i;
tdst_transpose(inst->dst, dst);
tsrc_transpose(inst->src[0], src);
}
else {
struct toy_src src[4];
- int i;
+ unsigned i;
tsrc_transpose(inst->src[0], src);
/* mask out killed pixels */
static void
fs_setup_shader_out(struct ilo_shader *sh, const struct toy_tgsi *tgsi)
{
- int i;
+ unsigned i;
sh->out.count = tgsi->num_outputs;
for (i = 0; i < tgsi->num_outputs; i++) {
fs_setup_shader_in(struct ilo_shader *sh, const struct toy_tgsi *tgsi,
bool flatshade)
{
- int i;
+ unsigned i;
sh->in.count = tgsi->num_inputs;
for (i = 0; i < tgsi->num_inputs; i++) {
tc_MOV(tc, block_offsets, idx);
msg_type = GEN6_MSG_DP_OWORD_DUAL_BLOCK_READ;
- msg_ctrl = GEN6_MSG_DP_OWORD_DUAL_BLOCK_SIZE_1;;
+ msg_ctrl = GEN6_MSG_DP_OWORD_DUAL_BLOCK_SIZE_1;
msg_len = 2;
desc = tsrc_imm_mdesc_data_port(tc, false, msg_len, 1, true, false,
if (num_coords >= 3) {
struct toy_dst tmp, max;
struct toy_src abs_coords[3];
- int i;
+ unsigned i;
tmp = tc_alloc_tmp(tc);
max = tdst_writemask(tmp, TOY_WRITEMASK_W);
vs_collect_outputs(struct vs_compile_context *vcc, struct toy_src *outs)
{
const struct toy_tgsi *tgsi = &vcc->tgsi;
- int i;
+ unsigned i;
for (i = 0; i < vcc->shader->out.count; i++) {
const int slot = vcc->output_map[i];
static void
linear_scan_free_regs(struct linear_scan *ls, int reg, int count)
{
- int i;
+ unsigned i;
for (i = 0; i < count; i++)
ls->free_regs[ls->num_free_regs++] = reg + count - 1 - i;
static void
linear_scan_spill_range(struct linear_scan *ls, int first, int count)
{
- int i;
+ unsigned i;
for (i = 0; i < count; i++) {
struct linear_scan_live_interval *interval = &ls->intervals[first + i];
tgsi_inst->Src[operand].Register.File;
switch (file) {
case TGSI_FILE_SAMPLER:
- case TGSI_FILE_RESOURCE:
+ case TGSI_FILE_IMAGE:
case TGSI_FILE_SAMPLER_VIEW:
type = TOY_TYPE_D;
break;
src = tsrc_null();
break;
case TGSI_FILE_SAMPLER:
- case TGSI_FILE_RESOURCE:
+ case TGSI_FILE_IMAGE:
case TGSI_FILE_SAMPLER_VIEW:
is_resource = true;
/* fall through */
need_vrf = true;
break;
case TGSI_FILE_SAMPLER:
- case TGSI_FILE_RESOURCE:
+ case TGSI_FILE_IMAGE:
case TGSI_FILE_SAMPLER_VIEW:
assert(!s->Register.Dimension);
src = tsrc_imm_d(s->Register.Index);
case TGSI_FILE_SAMPLER:
case TGSI_FILE_PREDICATE:
case TGSI_FILE_ADDRESS:
- case TGSI_FILE_RESOURCE:
+ case TGSI_FILE_IMAGE:
case TGSI_FILE_SAMPLER_VIEW:
/* nothing to do */
break;
LP_INTERP_LINEAR,
LP_INTERP_PERSPECTIVE,
LP_INTERP_POSITION,
- LP_INTERP_FACING,
- LP_INTERP_ZERO
+ LP_INTERP_FACING
};
struct lp_shader_input {
struct vertex_info vertex_info;
/** Which vertex shader output slot contains color */
- uint8_t color_slot[2];
+ int8_t color_slot[2];
/** Which vertex shader output slot contains bcolor */
- uint8_t bcolor_slot[2];
+ int8_t bcolor_slot[2];
/** Which vertex shader output slot contains point size */
- uint8_t psize_slot;
+ int8_t psize_slot;
/** Which vertex shader output slot contains viewport index */
- uint8_t viewport_index_slot;
+ int8_t viewport_index_slot;
/** Which geometry shader output slot contains layer */
- uint8_t layer_slot;
+ int8_t layer_slot;
/** A fake frontface output for unfilled primitives */
- uint8_t face_slot;
-
- /** Which output slot is used for the fake vp index info */
- uint8_t fake_vpindex_slot;
-
- /** Which output slot is used for the fake layer info */
- uint8_t fake_layer_slot;
+ int8_t face_slot;
/** Depth format and bias settings. */
boolean floating_point_depth;
int32_t dcdy;
/* one-pixel sized trivial reject offsets for each plane */
- int64_t eo;
+ uint32_t eo;
};
/**
}
static inline unsigned
-build_mask_linear(int64_t c, int64_t dcdx, int64_t dcdy)
+build_mask_linear(int32_t c, int32_t dcdx, int32_t dcdy)
{
unsigned mask = 0;
- int64_t c0 = c;
- int64_t c1 = c0 + dcdy;
- int64_t c2 = c1 + dcdy;
- int64_t c3 = c2 + dcdy;
-
- mask |= ((c0 + 0 * dcdx) >> FIXED_SHIFT) & (1 << 0);
- mask |= ((c0 + 1 * dcdx) >> FIXED_SHIFT) & (1 << 1);
- mask |= ((c0 + 2 * dcdx) >> FIXED_SHIFT) & (1 << 2);
- mask |= ((c0 + 3 * dcdx) >> FIXED_SHIFT) & (1 << 3);
- mask |= ((c1 + 0 * dcdx) >> FIXED_SHIFT) & (1 << 4);
- mask |= ((c1 + 1 * dcdx) >> FIXED_SHIFT) & (1 << 5);
- mask |= ((c1 + 2 * dcdx) >> FIXED_SHIFT) & (1 << 6);
- mask |= ((c1 + 3 * dcdx) >> FIXED_SHIFT) & (1 << 7);
- mask |= ((c2 + 0 * dcdx) >> FIXED_SHIFT) & (1 << 8);
- mask |= ((c2 + 1 * dcdx) >> FIXED_SHIFT) & (1 << 9);
- mask |= ((c2 + 2 * dcdx) >> FIXED_SHIFT) & (1 << 10);
- mask |= ((c2 + 3 * dcdx) >> FIXED_SHIFT) & (1 << 11);
- mask |= ((c3 + 0 * dcdx) >> FIXED_SHIFT) & (1 << 12);
- mask |= ((c3 + 1 * dcdx) >> FIXED_SHIFT) & (1 << 13);
- mask |= ((c3 + 2 * dcdx) >> FIXED_SHIFT) & (1 << 14);
- mask |= ((c3 + 3 * dcdx) >> FIXED_SHIFT) & (1 << 15);
+ int32_t c0 = c;
+ int32_t c1 = c0 + dcdy;
+ int32_t c2 = c1 + dcdy;
+ int32_t c3 = c2 + dcdy;
+
+ mask |= ((c0 + 0 * dcdx) >> 31) & (1 << 0);
+ mask |= ((c0 + 1 * dcdx) >> 31) & (1 << 1);
+ mask |= ((c0 + 2 * dcdx) >> 31) & (1 << 2);
+ mask |= ((c0 + 3 * dcdx) >> 31) & (1 << 3);
+ mask |= ((c1 + 0 * dcdx) >> 31) & (1 << 4);
+ mask |= ((c1 + 1 * dcdx) >> 31) & (1 << 5);
+ mask |= ((c1 + 2 * dcdx) >> 31) & (1 << 6);
+ mask |= ((c1 + 3 * dcdx) >> 31) & (1 << 7);
+ mask |= ((c2 + 0 * dcdx) >> 31) & (1 << 8);
+ mask |= ((c2 + 1 * dcdx) >> 31) & (1 << 9);
+ mask |= ((c2 + 2 * dcdx) >> 31) & (1 << 10);
+ mask |= ((c2 + 3 * dcdx) >> 31) & (1 << 11);
+ mask |= ((c3 + 0 * dcdx) >> 31) & (1 << 12);
+ mask |= ((c3 + 1 * dcdx) >> 31) & (1 << 13);
+ mask |= ((c3 + 2 * dcdx) >> 31) & (1 << 14);
+ mask |= ((c3 + 3 * dcdx) >> 31) & (1 << 15);
return mask;
}
static inline void
-build_masks(int64_t c,
- int64_t cdiff,
- int64_t dcdx,
- int64_t dcdy,
- unsigned *outmask,
- unsigned *partmask)
+build_masks(int32_t c,
+ int32_t cdiff,
+ int32_t dcdx,
+ int32_t dcdy,
+ unsigned *outmask,
+ unsigned *partmask)
{
*outmask |= build_mask_linear(c, dcdx, dcdy);
*partmask |= build_mask_linear(c + cdiff, dcdx, dcdy);
lp_rast_triangle_4(task, arg2);
}
-#if !defined(PIPE_ARCH_SSE)
+#if defined(PIPE_ARCH_SSE)
-void
-lp_rast_triangle_32_3_16(struct lp_rasterizer_task *task,
- const union lp_rast_cmd_arg arg)
-{
- union lp_rast_cmd_arg arg2;
- arg2.triangle.tri = arg.triangle.tri;
- arg2.triangle.plane_mask = (1<<3)-1;
- lp_rast_triangle_32_3(task, arg2);
-}
-
-void
-lp_rast_triangle_32_4_16(struct lp_rasterizer_task *task,
- const union lp_rast_cmd_arg arg)
-{
- union lp_rast_cmd_arg arg2;
- arg2.triangle.tri = arg.triangle.tri;
- arg2.triangle.plane_mask = (1<<4)-1;
- lp_rast_triangle_32_4(task, arg2);
-}
-
-void
-lp_rast_triangle_32_3_4(struct lp_rasterizer_task *task,
- const union lp_rast_cmd_arg arg)
-{
- lp_rast_triangle_32_3_16(task, arg);
-}
-
-#else
#include <emmintrin.h>
#include "util/u_sse.h"
static inline void
-build_masks_32(int c,
- int cdiff,
- int dcdx,
- int dcdy,
- unsigned *outmask,
- unsigned *partmask)
+build_masks_sse(int c,
+ int cdiff,
+ int dcdx,
+ int dcdy,
+ unsigned *outmask,
+ unsigned *partmask)
{
__m128i cstep0 = _mm_setr_epi32(c, c+dcdx, c+dcdx*2, c+dcdx*3);
__m128i xdcdy = _mm_set1_epi32(dcdy);
static inline unsigned
-build_mask_linear_32(int c, int dcdx, int dcdy)
+build_mask_linear_sse(int c, int dcdx, int dcdy)
{
__m128i cstep0 = _mm_setr_epi32(c, c+dcdx, c+dcdx*2, c+dcdx*3);
__m128i xdcdy = _mm_set1_epi32(dcdy);
#define NR_PLANES 3
-
-
-
-
-
-
void
lp_rast_triangle_32_3_16(struct lp_rasterizer_task *task,
const union lp_rast_cmd_arg arg)
0xffff & ~out[i].mask);
}
-
-
-
-
void
lp_rast_triangle_32_3_4(struct lp_rasterizer_task *task,
const union lp_rast_cmd_arg arg)
}
#undef NR_PLANES
+
+#else
+
+#if defined(_ARCH_PWR8) && defined(PIPE_ARCH_LITTLE_ENDIAN)
+
+#include <altivec.h>
+#include "util/u_pwr8.h"
+
+static inline void
+build_masks_ppc(int c,
+ int cdiff,
+ int dcdx,
+ int dcdy,
+ unsigned *outmask,
+ unsigned *partmask)
+{
+ __m128i cstep0 = vec_setr_epi32(c, c+dcdx, c+dcdx*2, c+dcdx*3);
+ __m128i xdcdy = (__m128i) vec_splats(dcdy);
+
+ /* Get values across the quad
+ */
+ __m128i cstep1 = vec_add_epi32(cstep0, xdcdy);
+ __m128i cstep2 = vec_add_epi32(cstep1, xdcdy);
+ __m128i cstep3 = vec_add_epi32(cstep2, xdcdy);
+
+ {
+ __m128i cstep01, cstep23, result;
+
+ cstep01 = vec_packs_epi32(cstep0, cstep1);
+ cstep23 = vec_packs_epi32(cstep2, cstep3);
+ result = vec_packs_epi16(cstep01, cstep23);
+
+ *outmask |= vec_movemask_epi8(result);
+ }
+
+
+ {
+ __m128i cio4 = (__m128i) vec_splats(cdiff);
+ __m128i cstep01, cstep23, result;
+
+ cstep0 = vec_add_epi32(cstep0, cio4);
+ cstep1 = vec_add_epi32(cstep1, cio4);
+ cstep2 = vec_add_epi32(cstep2, cio4);
+ cstep3 = vec_add_epi32(cstep3, cio4);
+
+ cstep01 = vec_packs_epi32(cstep0, cstep1);
+ cstep23 = vec_packs_epi32(cstep2, cstep3);
+ result = vec_packs_epi16(cstep01, cstep23);
+
+ *partmask |= vec_movemask_epi8(result);
+ }
+}
+
+static inline unsigned
+build_mask_linear_ppc(int c, int dcdx, int dcdy)
+{
+ __m128i cstep0 = vec_setr_epi32(c, c+dcdx, c+dcdx*2, c+dcdx*3);
+ __m128i xdcdy = (__m128i) vec_splats(dcdy);
+
+ /* Get values across the quad
+ */
+ __m128i cstep1 = vec_add_epi32(cstep0, xdcdy);
+ __m128i cstep2 = vec_add_epi32(cstep1, xdcdy);
+ __m128i cstep3 = vec_add_epi32(cstep2, xdcdy);
+
+ /* pack pairs of results into epi16
+ */
+ __m128i cstep01 = vec_packs_epi32(cstep0, cstep1);
+ __m128i cstep23 = vec_packs_epi32(cstep2, cstep3);
+
+ /* pack into epi8, preserving sign bits
+ */
+ __m128i result = vec_packs_epi16(cstep01, cstep23);
+
+ /* extract sign bits to create mask
+ */
+ return vec_movemask_epi8(result);
+}
+
+static inline __m128i
+lp_plane_to_m128i(const struct lp_rast_plane *plane)
+{
+ return vec_setr_epi32((int32_t)plane->c, (int32_t)plane->dcdx,
+ (int32_t)plane->dcdy, (int32_t)plane->eo);
+}
+
+#define NR_PLANES 3
+
+void
+lp_rast_triangle_32_3_16(struct lp_rasterizer_task *task,
+ const union lp_rast_cmd_arg arg)
+{
+ const struct lp_rast_triangle *tri = arg.triangle.tri;
+ const struct lp_rast_plane *plane = GET_PLANES(tri);
+ int x = (arg.triangle.plane_mask & 0xff) + task->x;
+ int y = (arg.triangle.plane_mask >> 8) + task->y;
+ unsigned i, j;
+
+ struct { unsigned mask:16; unsigned i:8; unsigned j:8; } out[16];
+ unsigned nr = 0;
+
+ __m128i p0 = lp_plane_to_m128i(&plane[0]); /* c, dcdx, dcdy, eo */
+ __m128i p1 = lp_plane_to_m128i(&plane[1]); /* c, dcdx, dcdy, eo */
+ __m128i p2 = lp_plane_to_m128i(&plane[2]); /* c, dcdx, dcdy, eo */
+ __m128i zero = vec_splats((unsigned char) 0);
+
+ __m128i c;
+ __m128i dcdx;
+ __m128i dcdy;
+ __m128i rej4;
+
+ __m128i dcdx2;
+ __m128i dcdx3;
+
+ __m128i span_0; /* 0,dcdx,2dcdx,3dcdx for plane 0 */
+ __m128i span_1; /* 0,dcdx,2dcdx,3dcdx for plane 1 */
+ __m128i span_2; /* 0,dcdx,2dcdx,3dcdx for plane 2 */
+ __m128i unused;
+
+ __m128i vshuf_mask0;
+ __m128i vshuf_mask1;
+ __m128i vshuf_mask2;
+
+#ifdef PIPE_ARCH_LITTLE_ENDIAN
+ vshuf_mask0 = (__m128i) vec_splats((unsigned int) 0x03020100);
+ vshuf_mask1 = (__m128i) vec_splats((unsigned int) 0x07060504);
+ vshuf_mask2 = (__m128i) vec_splats((unsigned int) 0x0B0A0908);
+#else
+ vshuf_mask0 = (__m128i) vec_splats((unsigned int) 0x0C0D0E0F);
+ vshuf_mask1 = (__m128i) vec_splats((unsigned int) 0x08090A0B);
+ vshuf_mask2 = (__m128i) vec_splats((unsigned int) 0x04050607);
+#endif
+
+ transpose4_epi32(&p0, &p1, &p2, &zero,
+ &c, &dcdx, &dcdy, &rej4);
+
+ /* Adjust dcdx;
+ */
+ dcdx = vec_sub_epi32(zero, dcdx);
+
+ c = vec_add_epi32(c, vec_mullo_epi32(dcdx, (__m128i) vec_splats(x)));
+ c = vec_add_epi32(c, vec_mullo_epi32(dcdy, (__m128i) vec_splats(y)));
+ rej4 = vec_slli_epi32(rej4, 2);
+
+ /*
+ * Adjust so we can just check the sign bit (< 0 comparison),
+ * instead of having to do a less efficient <= 0 comparison
+ */
+ c = vec_sub_epi32(c, (__m128i) vec_splats((unsigned int) 1));
+ rej4 = vec_add_epi32(rej4, (__m128i) vec_splats((unsigned int) 1));
+
+ dcdx2 = vec_add_epi32(dcdx, dcdx);
+ dcdx3 = vec_add_epi32(dcdx2, dcdx);
+
+ transpose4_epi32(&zero, &dcdx, &dcdx2, &dcdx3,
+ &span_0, &span_1, &span_2, &unused);
+
+ for (i = 0; i < 4; i++) {
+ __m128i cx = c;
+
+ for (j = 0; j < 4; j++) {
+ __m128i c4rej = vec_add_epi32(cx, rej4);
+ __m128i rej_masks = vec_srai_epi32(c4rej, 31);
+
+ /* if (is_zero(rej_masks)) */
+ if (vec_movemask_epi8(rej_masks) == 0) {
+ __m128i c0_0 = vec_add_epi32(vec_perm(cx, cx, vshuf_mask0), span_0);
+ __m128i c1_0 = vec_add_epi32(vec_perm(cx, cx, vshuf_mask1), span_1);
+ __m128i c2_0 = vec_add_epi32(vec_perm(cx, cx, vshuf_mask2), span_2);
+
+ __m128i c_0 = vec_or(vec_or(c0_0, c1_0), c2_0);
+
+ __m128i c0_1 = vec_add_epi32(c0_0, vec_perm(dcdy, dcdy, vshuf_mask0));
+ __m128i c1_1 = vec_add_epi32(c1_0, vec_perm(dcdy, dcdy, vshuf_mask1));
+ __m128i c2_1 = vec_add_epi32(c2_0, vec_perm(dcdy, dcdy, vshuf_mask2));
+
+ __m128i c_1 = vec_or(vec_or(c0_1, c1_1), c2_1);
+ __m128i c_01 = vec_packs_epi32(c_0, c_1);
+
+ __m128i c0_2 = vec_add_epi32(c0_1, vec_perm(dcdy, dcdy, vshuf_mask0));
+ __m128i c1_2 = vec_add_epi32(c1_1, vec_perm(dcdy, dcdy, vshuf_mask1));
+ __m128i c2_2 = vec_add_epi32(c2_1, vec_perm(dcdy, dcdy, vshuf_mask2));
+
+ __m128i c_2 = vec_or(vec_or(c0_2, c1_2), c2_2);
+
+ __m128i c0_3 = vec_add_epi32(c0_2, vec_perm(dcdy, dcdy, vshuf_mask0));
+ __m128i c1_3 = vec_add_epi32(c1_2, vec_perm(dcdy, dcdy, vshuf_mask1));
+ __m128i c2_3 = vec_add_epi32(c2_2, vec_perm(dcdy, dcdy, vshuf_mask2));
+
+ __m128i c_3 = vec_or(vec_or(c0_3, c1_3), c2_3);
+ __m128i c_23 = vec_packs_epi32(c_2, c_3);
+ __m128i c_0123 = vec_packs_epi16(c_01, c_23);
+
+ unsigned mask = vec_movemask_epi8(c_0123);
+
+ out[nr].i = i;
+ out[nr].j = j;
+ out[nr].mask = mask;
+ if (mask != 0xffff)
+ nr++;
+ }
+ cx = vec_add_epi32(cx, vec_slli_epi32(dcdx, 2));
+ }
+
+ c = vec_add_epi32(c, vec_slli_epi32(dcdy, 2));
+ }
+
+ for (i = 0; i < nr; i++)
+ lp_rast_shade_quads_mask(task,
+ &tri->inputs,
+ x + 4 * out[i].j,
+ y + 4 * out[i].i,
+ 0xffff & ~out[i].mask);
+}
+
+#undef NR_PLANES
+
+#else
+
+void
+lp_rast_triangle_32_3_16(struct lp_rasterizer_task *task,
+ const union lp_rast_cmd_arg arg)
+{
+ union lp_rast_cmd_arg arg2;
+ arg2.triangle.tri = arg.triangle.tri;
+ arg2.triangle.plane_mask = (1<<3)-1;
+ lp_rast_triangle_32_3(task, arg2);
+}
+
+#endif /* _ARCH_PWR8 && PIPE_ARCH_LITTLE_ENDIAN */
+
+void
+lp_rast_triangle_32_4_16(struct lp_rasterizer_task *task,
+ const union lp_rast_cmd_arg arg)
+{
+ union lp_rast_cmd_arg arg2;
+ arg2.triangle.tri = arg.triangle.tri;
+ arg2.triangle.plane_mask = (1<<4)-1;
+ lp_rast_triangle_32_4(task, arg2);
+}
+
+void
+lp_rast_triangle_32_3_4(struct lp_rasterizer_task *task,
+ const union lp_rast_cmd_arg arg)
+{
+ lp_rast_triangle_32_3_16(task, arg);
+}
+
#endif
+#if defined PIPE_ARCH_SSE
+#define BUILD_MASKS(c, cdiff, dcdx, dcdy, omask, pmask) build_masks_sse((int)c, (int)cdiff, dcdx, dcdy, omask, pmask)
+#define BUILD_MASK_LINEAR(c, dcdx, dcdy) build_mask_linear_sse((int)c, dcdx, dcdy)
+#elif (defined(_ARCH_PWR8) && defined(PIPE_ARCH_LITTLE_ENDIAN))
+#define BUILD_MASKS(c, cdiff, dcdx, dcdy, omask, pmask) build_masks_ppc((int)c, (int)cdiff, dcdx, dcdy, omask, pmask)
+#define BUILD_MASK_LINEAR(c, dcdx, dcdy) build_mask_linear_ppc((int)c, dcdx, dcdy)
+#else
#define BUILD_MASKS(c, cdiff, dcdx, dcdy, omask, pmask) build_masks(c, cdiff, dcdx, dcdy, omask, pmask)
#define BUILD_MASK_LINEAR(c, dcdx, dcdy) build_mask_linear(c, dcdx, dcdy)
+#endif
+
+#define RASTER_64 1
#define TAG(x) x##_1
#define NR_PLANES 1
#define NR_PLANES 8
#include "lp_rast_tri_tmp.h"
-#ifdef PIPE_ARCH_SSE
-#undef BUILD_MASKS
-#undef BUILD_MASK_LINEAR
-#define BUILD_MASKS(c, cdiff, dcdx, dcdy, omask, pmask) build_masks_32((int)c, (int)cdiff, dcdx, dcdy, omask, pmask)
-#define BUILD_MASK_LINEAR(c, dcdx, dcdy) build_mask_linear_32((int)c, dcdx, dcdy)
-#endif
+#undef RASTER_64
#define TAG(x) x##_32_1
#define NR_PLANES 1
int j;
for (j = 0; j < NR_PLANES; j++) {
- mask &= ~BUILD_MASK_LINEAR(c[j] - 1,
- -plane[j].dcdx,
- plane[j].dcdy);
+#ifdef RASTER_64
+ mask &= ~BUILD_MASK_LINEAR(((c[j] - 1) >> (int64_t)FIXED_ORDER),
+ -plane[j].dcdx >> FIXED_ORDER,
+ plane[j].dcdy >> FIXED_ORDER);
+#else
+ mask &= ~BUILD_MASK_LINEAR((c[j] - 1),
+ -plane[j].dcdx,
+ plane[j].dcdy);
+#endif
}
/* Now pass to the shader:
partmask = 0; /* outside one or more trivial accept planes */
for (j = 0; j < NR_PLANES; j++) {
+#ifdef RASTER_64
+ int32_t dcdx = -plane[j].dcdx >> FIXED_ORDER;
+ int32_t dcdy = plane[j].dcdy >> FIXED_ORDER;
+ const int32_t cox = plane[j].eo >> FIXED_ORDER;
+ const int32_t ei = (dcdy + dcdx - cox) << 2;
+ const int32_t cox_s = cox << 2;
+ const int32_t co = (int32_t)(c[j] >> (int64_t)FIXED_ORDER) + cox_s;
+ int32_t cdiff;
+ cdiff = ei - cox_s + ((int32_t)((c[j] - 1) >> (int64_t)FIXED_ORDER) -
+ (int32_t)(c[j] >> (int64_t)FIXED_ORDER));
+ dcdx <<= 2;
+ dcdy <<= 2;
+#else
const int64_t dcdx = -IMUL64(plane[j].dcdx, 4);
const int64_t dcdy = IMUL64(plane[j].dcdy, 4);
const int64_t cox = IMUL64(plane[j].eo, 4);
- const int64_t ei = plane[j].dcdy - plane[j].dcdx - plane[j].eo;
+ const int32_t ei = plane[j].dcdy - plane[j].dcdx - (int64_t)plane[j].eo;
const int64_t cio = IMUL64(ei, 4) - 1;
+ int32_t co, cdiff;
+ co = c[j] + cox;
+ cdiff = cio - cox;
+#endif
- BUILD_MASKS(c[j] + cox,
- cio - cox,
- dcdx, dcdy,
- &outmask, /* sign bits from c[i][0..15] + cox */
- &partmask); /* sign bits from c[i][0..15] + cio */
+ BUILD_MASKS(co, cdiff,
+ dcdx, dcdy,
+ &outmask, /* sign bits from c[i][0..15] + cox */
+ &partmask); /* sign bits from c[i][0..15] + cio */
}
if (outmask == 0xffff)
c[j] = plane[j].c + IMUL64(plane[j].dcdy, y) - IMUL64(plane[j].dcdx, x);
{
- const int64_t dcdx = -IMUL64(plane[j].dcdx, 16);
- const int64_t dcdy = IMUL64(plane[j].dcdy, 16);
- const int64_t cox = IMUL64(plane[j].eo, 16);
- const int64_t ei = plane[j].dcdy - plane[j].dcdx - plane[j].eo;
- const int64_t cio = IMUL64(ei, 16) - 1;
-
- BUILD_MASKS(c[j] + cox,
- cio - cox,
+#ifdef RASTER_64
+ /*
+ * Strip off lower FIXED_ORDER bits. Note that those bits from
+ * dcdx, dcdy, eo are always 0 (by definition).
+ * c values, however, are not. This means that for every
+ * addition of the form c + n*dcdx the lower FIXED_ORDER bits will
+ * NOT change. And those bits are not relevant to the sign bit (which
+ * is only what we need!) that is,
+ * sign(c + n*dcdx) == sign((c >> FIXED_ORDER) + n*(dcdx >> FIXED_ORDER))
+ * This means we can get away with using 32bit math for the most part.
+ * Only tricky part is the -1 adjustment for cdiff.
+ */
+ int32_t dcdx = -plane[j].dcdx >> FIXED_ORDER;
+ int32_t dcdy = plane[j].dcdy >> FIXED_ORDER;
+ const int32_t cox = plane[j].eo >> FIXED_ORDER;
+ const int32_t ei = (dcdy + dcdx - cox) << 4;
+ const int32_t cox_s = cox << 4;
+ const int32_t co = (int32_t)(c[j] >> (int64_t)FIXED_ORDER) + cox_s;
+ int32_t cdiff;
+ /*
+ * Plausibility check to ensure the 32bit math works.
+ * Note that within a tile, the max we can move the edge function
+ * is essentially dcdx * TILE_SIZE + dcdy * TILE_SIZE.
+ * TILE_SIZE is 64, dcdx/dcdy are nominally 21 bit (for 8192 max size
+ * and 8 subpixel bits), I'd be happy with 2 bits more too (1 for
+ * increasing fb size to 16384, the required d3d11 value, another one
+ * because I'm not quite sure we can't be _just_ above the max value
+ * here). This gives us 30 bits max - hence if c would exceed that here
+ * that means the plane is either trivial reject for the whole tile
+ * (in which case the tri will not get binned), or trivial accept for
+ * the whole tile (in which case plane_mask will not include it).
+ */
+ assert((c[j] >> (int64_t)FIXED_ORDER) > (int32_t)0xb0000000 &&
+ (c[j] >> (int64_t)FIXED_ORDER) < (int32_t)0x3fffffff);
+ /*
+ * Note the fixup part is constant throughout the tile - thus could
+ * just calculate this and avoid _all_ 64bit math in rasterization
+ * (except exactly this fixup calc).
+ * In fact theoretically could move that even to setup, albeit that
+ * seems tricky (pre-bin certainly can have values larger than 32bit,
+ * and would need to communicate that fixup value through).
+ * And if we want to support msaa, we'd probably don't want to do the
+ * downscaling in setup in any case...
+ */
+ cdiff = ei - cox_s + ((int32_t)((c[j] - 1) >> (int64_t)FIXED_ORDER) -
+ (int32_t)(c[j] >> (int64_t)FIXED_ORDER));
+ dcdx <<= 4;
+ dcdy <<= 4;
+#else
+ const int32_t dcdx = -plane[j].dcdx << 4;
+ const int32_t dcdy = plane[j].dcdy << 4;
+ const int32_t cox = plane[j].eo << 4;
+ const int32_t ei = plane[j].dcdy - plane[j].dcdx - (int32_t)plane[j].eo;
+ const int32_t cio = (ei << 4) - 1;
+ int32_t co, cdiff;
+ co = c[j] + cox;
+ cdiff = cio - cox;
+#endif
+ BUILD_MASKS(co, cdiff,
dcdx, dcdy,
&outmask, /* sign bits from c[i][0..15] + cox */
&partmask); /* sign bits from c[i][0..15] + cio */
case PIPE_CAP_SHAREABLE_SHADERS:
case PIPE_CAP_COPY_BETWEEN_COMPRESSED_AND_PLAIN_FORMATS:
case PIPE_CAP_CLEAR_TEXTURE:
+ case PIPE_CAP_DRAW_PARAMETERS:
+ case PIPE_CAP_TGSI_PACK_HALF_FLOAT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT_PARAMS:
+ case PIPE_CAP_TGSI_FS_POSITION_IS_SYSVAL:
+ case PIPE_CAP_TGSI_FS_FACE_IS_INTEGER_SYSVAL:
+ case PIPE_CAP_SHADER_BUFFER_OFFSET_ALIGNMENT:
+ case PIPE_CAP_INVALIDATE_BUFFER:
+ case PIPE_CAP_GENERATE_MIPMAP:
return 0;
}
/* should only get here on unhandled cases */
depth,
stencil);
+ /*
+ * XXX: should make a full mask here for things like D24X8,
+ * otherwise we'll do a read-modify-write clear later which
+ * should be unnecessary.
+ */
zsmask = util_pack64_mask_z_stencil(setup->fb.zsbuf->format,
zmask32,
smask8);
float pixel_offset;
float line_width;
float point_size;
- uint8_t psize_slot;
- uint8_t viewport_index_slot;
- uint8_t layer_slot;
- uint8_t face_slot;
+ int8_t psize_slot;
+ int8_t viewport_index_slot;
+ int8_t layer_slot;
+ int8_t face_slot;
struct pipe_framebuffer_state fb;
struct u_rect framebuffer;
line->inputs.layer = layer;
line->inputs.viewport_index = viewport_index;
+ /*
+ * XXX: this code is mostly identical to the one in lp_setup_tri, except it
+ * uses 4 planes instead of 3. Could share the code (including the sse
+ * assembly, in fact we'd get the 4th plane for free).
+ * The only difference apart from storing the 4th plane would be some
+ * different shuffle for calculating dcdx/dcdy.
+ */
for (i = 0; i < 4; i++) {
- /* half-edge constants, will be interated over the whole render
+ /* half-edge constants, will be iterated over the whole render
* target.
*/
plane[i].c = IMUL64(plane[i].dcdx, x[i]) - IMUL64(plane[i].dcdy, y[i]);
-
- /* correct for top-left vs. bottom-left fill convention.
- */
+ /* correct for top-left vs. bottom-left fill convention.
+ */
if (plane[i].dcdx < 0) {
/* both fill conventions want this - adjust for left edges */
- plane[i].c++;
+ plane[i].c++;
}
else if (plane[i].dcdx == 0) {
if (setup->pixel_offset == 0) {
const struct u_rect *scissor =
&setup->scissors[viewport_index];
- plane[4].dcdx = -1;
+ plane[4].dcdx = -1 << 8;
plane[4].dcdy = 0;
- plane[4].c = 1-scissor->x0;
- plane[4].eo = 1;
+ plane[4].c = (1-scissor->x0) << 8;
+ plane[4].eo = 1 << 8;
- plane[5].dcdx = 1;
+ plane[5].dcdx = 1 << 8;
plane[5].dcdy = 0;
- plane[5].c = scissor->x1+1;
+ plane[5].c = (scissor->x1+1) << 8;
plane[5].eo = 0;
plane[6].dcdx = 0;
- plane[6].dcdy = 1;
- plane[6].c = 1-scissor->y0;
- plane[6].eo = 1;
+ plane[6].dcdy = 1 << 8;
+ plane[6].c = (1-scissor->y0) << 8;
+ plane[6].eo = 1 << 8;
plane[7].dcdx = 0;
- plane[7].dcdy = -1;
- plane[7].c = scissor->y1+1;
+ plane[7].dcdy = -1 << 8;
+ plane[7].c = (scissor->y1+1) << 8;
plane[7].eo = 0;
}
{
struct lp_rast_plane *plane = GET_PLANES(point);
- plane[0].dcdx = -1;
+ plane[0].dcdx = -1 << 8;
plane[0].dcdy = 0;
- plane[0].c = 1-bbox.x0;
- plane[0].eo = 1;
+ plane[0].c = (1-bbox.x0) << 8;
+ plane[0].eo = 1 << 8;
- plane[1].dcdx = 1;
+ plane[1].dcdx = 1 << 8;
plane[1].dcdy = 0;
- plane[1].c = bbox.x1+1;
+ plane[1].c = (bbox.x1+1) << 8;
plane[1].eo = 0;
plane[2].dcdx = 0;
- plane[2].dcdy = 1;
- plane[2].c = 1-bbox.y0;
- plane[2].eo = 1;
+ plane[2].dcdy = 1 << 8;
+ plane[2].c = (1-bbox.y0) << 8;
+ plane[2].eo = 1 << 8;
plane[3].dcdx = 0;
- plane[3].dcdy = -1;
- plane[3].c = bbox.y1+1;
+ plane[3].dcdy = -1 << 8;
+ plane[3].c = (bbox.y1+1) << 8;
plane[3].eo = 0;
}
#if defined(PIPE_ARCH_SSE)
#include <emmintrin.h>
+#elif defined(_ARCH_PWR8) && defined(PIPE_ARCH_LITTLE_ENDIAN)
+#include <altivec.h>
+#include "util/u_pwr8.h"
#endif
static inline int
struct fixed_position {
int32_t x[4];
int32_t y[4];
- int64_t area;
int32_t dx01;
int32_t dy01;
int32_t dx20;
int32_t dy20;
+ int64_t area;
};
plane = GET_PLANES(tri);
#if defined(PIPE_ARCH_SSE)
- if (setup->fb.width <= MAX_FIXED_LENGTH32 &&
- setup->fb.height <= MAX_FIXED_LENGTH32 &&
- (bbox.x1 - bbox.x0) <= MAX_FIXED_LENGTH32 &&
- (bbox.y1 - bbox.y0) <= MAX_FIXED_LENGTH32) {
+ if (1) {
__m128i vertx, verty;
__m128i shufx, shufy;
- __m128i dcdx, dcdy, c;
- __m128i unused;
+ __m128i dcdx, dcdy;
+ __m128i cdx02, cdx13, cdy02, cdy13, c02, c13;
+ __m128i c01, c23, unused;
__m128i dcdx_neg_mask;
__m128i dcdy_neg_mask;
__m128i dcdx_zero_mask;
- __m128i top_left_flag;
- __m128i c_inc_mask, c_inc;
+ __m128i top_left_flag, c_dec;
__m128i eo, p0, p1, p2;
__m128i zero = _mm_setzero_si128();
- PIPE_ALIGN_VAR(16) int32_t temp_vec[4];
- vertx = _mm_loadu_si128((__m128i *)position->x); /* vertex x coords */
- verty = _mm_loadu_si128((__m128i *)position->y); /* vertex y coords */
+ vertx = _mm_load_si128((__m128i *)position->x); /* vertex x coords */
+ verty = _mm_load_si128((__m128i *)position->y); /* vertex y coords */
shufx = _mm_shuffle_epi32(vertx, _MM_SHUFFLE(3,0,2,1));
shufy = _mm_shuffle_epi32(verty, _MM_SHUFFLE(3,0,2,1));
top_left_flag = _mm_set1_epi32((setup->bottom_edge_rule == 0) ? ~0 : 0);
- c_inc_mask = _mm_or_si128(dcdx_neg_mask,
- _mm_and_si128(dcdx_zero_mask,
- _mm_xor_si128(dcdy_neg_mask,
- top_left_flag)));
-
- c_inc = _mm_srli_epi32(c_inc_mask, 31);
+ c_dec = _mm_or_si128(dcdx_neg_mask,
+ _mm_and_si128(dcdx_zero_mask,
+ _mm_xor_si128(dcdy_neg_mask,
+ top_left_flag)));
- c = _mm_sub_epi32(mm_mullo_epi32(dcdx, vertx),
- mm_mullo_epi32(dcdy, verty));
-
- c = _mm_add_epi32(c, c_inc);
+ /*
+ * 64 bit arithmetic.
+ * Note we need _signed_ mul (_mm_mul_epi32) which we emulate.
+ */
+ cdx02 = mm_mullohi_epi32(dcdx, vertx, &cdx13);
+ cdy02 = mm_mullohi_epi32(dcdy, verty, &cdy13);
+ c02 = _mm_sub_epi64(cdx02, cdy02);
+ c13 = _mm_sub_epi64(cdx13, cdy13);
+ c02 = _mm_sub_epi64(c02, _mm_shuffle_epi32(c_dec,
+ _MM_SHUFFLE(2,2,0,0)));
+ c13 = _mm_sub_epi64(c13, _mm_shuffle_epi32(c_dec,
+ _MM_SHUFFLE(3,3,1,1)));
+
+ /*
+ * Useful for very small fbs/tris (or fewer subpixel bits) only:
+ * c = _mm_sub_epi32(mm_mullo_epi32(dcdx, vertx),
+ * mm_mullo_epi32(dcdy, verty));
+ *
+ * c = _mm_sub_epi32(c, c_dec);
+ */
/* Scale up to match c:
*/
dcdx = _mm_slli_epi32(dcdx, FIXED_ORDER);
dcdy = _mm_slli_epi32(dcdy, FIXED_ORDER);
- /* Calculate trivial reject values:
+ /*
+ * Calculate trivial reject values:
+ * Note eo cannot overflow even if dcdx/dcdy would already have
+ * 31 bits (which they shouldn't have). This is because eo
+ * is never negative (albeit if we rely on that need to be careful...)
*/
eo = _mm_sub_epi32(_mm_andnot_si128(dcdy_neg_mask, dcdy),
_mm_and_si128(dcdx_neg_mask, dcdx));
/* ei = _mm_sub_epi32(_mm_sub_epi32(dcdy, dcdx), eo); */
+ /*
+ * Pointless transpose which gets undone immediately in
+ * rasterization.
+ * It is actually difficult to do away with it - would essentially
+ * need GET_PLANES_DX, GET_PLANES_DY etc., but the calculations
+ * for this then would need to depend on the number of planes.
+ * The transpose is quite special here due to c being 64bit...
+ * The store has to be unaligned (unless we'd make the plane size
+ * a multiple of 128), and of course storing eo separately...
+ */
+ c01 = _mm_unpacklo_epi64(c02, c13);
+ c23 = _mm_unpackhi_epi64(c02, c13);
+ transpose2_64_2_32(&c01, &c23, &dcdx, &dcdy,
+ &p0, &p1, &p2, &unused);
+ _mm_storeu_si128((__m128i *)&plane[0], p0);
+ plane[0].eo = (uint32_t)_mm_cvtsi128_si32(eo);
+ _mm_storeu_si128((__m128i *)&plane[1], p1);
+ eo = _mm_shuffle_epi32(eo, _MM_SHUFFLE(3,2,0,1));
+ plane[1].eo = (uint32_t)_mm_cvtsi128_si32(eo);
+ _mm_storeu_si128((__m128i *)&plane[2], p2);
+ eo = _mm_shuffle_epi32(eo, _MM_SHUFFLE(0,0,0,2));
+ plane[2].eo = (uint32_t)_mm_cvtsi128_si32(eo);
+ } else
+#elif defined(_ARCH_PWR8) && defined(PIPE_ARCH_LITTLE_ENDIAN)
+ /*
+ * XXX this code is effectively disabled for all practical purposes,
+ * as the allowed fb size is tiny if FIXED_ORDER is 8.
+ */
+ if (setup->fb.width <= MAX_FIXED_LENGTH32 &&
+ setup->fb.height <= MAX_FIXED_LENGTH32 &&
+ (bbox.x1 - bbox.x0) <= MAX_FIXED_LENGTH32 &&
+ (bbox.y1 - bbox.y0) <= MAX_FIXED_LENGTH32) {
+ unsigned int bottom_edge;
+ __m128i vertx, verty;
+ __m128i shufx, shufy;
+ __m128i dcdx, dcdy, c;
+ __m128i unused;
+ __m128i dcdx_neg_mask;
+ __m128i dcdy_neg_mask;
+ __m128i dcdx_zero_mask;
+ __m128i top_left_flag;
+ __m128i c_inc_mask, c_inc;
+ __m128i eo, p0, p1, p2;
+ __m128i_union vshuf_mask;
+ __m128i zero = vec_splats((unsigned char) 0);
+ PIPE_ALIGN_VAR(16) int32_t temp_vec[4];
+
+#ifdef PIPE_ARCH_LITTLE_ENDIAN
+ vshuf_mask.i[0] = 0x07060504;
+ vshuf_mask.i[1] = 0x0B0A0908;
+ vshuf_mask.i[2] = 0x03020100;
+ vshuf_mask.i[3] = 0x0F0E0D0C;
+#else
+ vshuf_mask.i[0] = 0x00010203;
+ vshuf_mask.i[1] = 0x0C0D0E0F;
+ vshuf_mask.i[2] = 0x04050607;
+ vshuf_mask.i[3] = 0x08090A0B;
+#endif
+
+ /* vertex x coords */
+ vertx = vec_load_si128((const uint32_t *) position->x);
+ /* vertex y coords */
+ verty = vec_load_si128((const uint32_t *) position->y);
+
+ shufx = vec_perm (vertx, vertx, vshuf_mask.m128i);
+ shufy = vec_perm (verty, verty, vshuf_mask.m128i);
+
+ dcdx = vec_sub_epi32(verty, shufy);
+ dcdy = vec_sub_epi32(vertx, shufx);
+
+ dcdx_neg_mask = vec_srai_epi32(dcdx, 31);
+ dcdx_zero_mask = vec_cmpeq_epi32(dcdx, zero);
+ dcdy_neg_mask = vec_srai_epi32(dcdy, 31);
+
+ bottom_edge = (setup->bottom_edge_rule == 0) ? ~0 : 0;
+ top_left_flag = (__m128i) vec_splats(bottom_edge);
+
+ c_inc_mask = vec_or(dcdx_neg_mask,
+ vec_and(dcdx_zero_mask,
+ vec_xor(dcdy_neg_mask,
+ top_left_flag)));
+
+ c_inc = vec_srli_epi32(c_inc_mask, 31);
+
+ c = vec_sub_epi32(vec_mullo_epi32(dcdx, vertx),
+ vec_mullo_epi32(dcdy, verty));
+
+ c = vec_add_epi32(c, c_inc);
+
+ /* Scale up to match c:
+ */
+ dcdx = vec_slli_epi32(dcdx, FIXED_ORDER);
+ dcdy = vec_slli_epi32(dcdy, FIXED_ORDER);
+
+ /* Calculate trivial reject values:
+ */
+ eo = vec_sub_epi32(vec_andc(dcdy_neg_mask, dcdy),
+ vec_and(dcdx_neg_mask, dcdx));
+
+ /* ei = _mm_sub_epi32(_mm_sub_epi32(dcdy, dcdx), eo); */
+
/* Pointless transpose which gets undone immediately in
* rasterization:
*/
transpose4_epi32(&c, &dcdx, &dcdy, &eo,
&p0, &p1, &p2, &unused);
-#define STORE_PLANE(plane, vec) do { \
- _mm_store_si128((__m128i *)&temp_vec, vec); \
- plane.c = (int64_t)temp_vec[0]; \
- plane.dcdx = temp_vec[1]; \
- plane.dcdy = temp_vec[2]; \
- plane.eo = temp_vec[3]; \
+#define STORE_PLANE(plane, vec) do { \
+ vec_store_si128((uint32_t *)&temp_vec, vec); \
+ plane.c = (int64_t)temp_vec[0]; \
+ plane.dcdx = temp_vec[1]; \
+ plane.dcdy = temp_vec[2]; \
+ plane.eo = temp_vec[3]; \
} while(0)
STORE_PLANE(plane[0], p0);
plane[2].dcdx = position->dy20;
for (i = 0; i < 3; i++) {
- /* half-edge constants, will be interated over the whole render
+ /* half-edge constants, will be iterated over the whole render
* target.
*/
plane[i].c = IMUL64(plane[i].dcdx, position->x[i]) -
- IMUL64(plane[i].dcdy, position->y[i]);
+ IMUL64(plane[i].dcdy, position->y[i]);
/* correct for top-left vs. bottom-left fill convention.
- */
+ */
if (plane[i].dcdx < 0) {
/* both fill conventions want this - adjust for left edges */
- plane[i].c++;
+ plane[i].c++;
}
else if (plane[i].dcdx == 0) {
if (setup->bottom_edge_rule == 0){
}
if (0) {
- debug_printf("p0: %"PRIx64"/%08x/%08x/%"PRIx64"\n",
+ debug_printf("p0: %"PRIx64"/%08x/%08x/%08x\n",
plane[0].c,
plane[0].dcdx,
plane[0].dcdy,
plane[0].eo);
-
- debug_printf("p1: %"PRIx64"/%08x/%08x/%"PRIx64"\n",
+
+ debug_printf("p1: %"PRIx64"/%08x/%08x/%08x\n",
plane[1].c,
plane[1].dcdx,
plane[1].dcdy,
plane[1].eo);
-
- debug_printf("p2: %"PRIx64"/%08x/%08x/%"PRIx64"\n",
+
+ debug_printf("p2: %"PRIx64"/%08x/%08x/%08x\n",
plane[2].c,
plane[2].dcdx,
plane[2].dcdy,
if (nr_planes == 7) {
const struct u_rect *scissor = &setup->scissors[viewport_index];
- plane[3].dcdx = -1;
+ plane[3].dcdx = -1 << 8;
plane[3].dcdy = 0;
- plane[3].c = 1-scissor->x0;
- plane[3].eo = 1;
+ plane[3].c = (1-scissor->x0) << 8;
+ plane[3].eo = 1 << 8;
- plane[4].dcdx = 1;
+ plane[4].dcdx = 1 << 8;
plane[4].dcdy = 0;
- plane[4].c = scissor->x1+1;
+ plane[4].c = (scissor->x1+1) << 8;
plane[4].eo = 0;
plane[5].dcdx = 0;
- plane[5].dcdy = 1;
- plane[5].c = 1-scissor->y0;
- plane[5].eo = 1;
+ plane[5].dcdy = 1 << 8;
+ plane[5].c = (1-scissor->y0) << 8;
+ plane[5].eo = 1 << 8;
plane[6].dcdx = 0;
- plane[6].dcdy = -1;
- plane[6].c = scissor->y1+1;
+ plane[6].dcdy = -1 << 8;
+ plane[6].c = (scissor->y1+1) << 8;
plane[6].eo = 0;
}
static inline uint32_t
floor_pot(uint32_t n)
{
-#if defined(PIPE_CC_GCC) && defined(PIPE_ARCH_X86)
+#if defined(PIPE_CC_GCC) && (defined(PIPE_ARCH_X86) || defined(PIPE_ARCH_X86_64))
if (n == 0)
return 0;
ei[i] = (plane[i].dcdy -
plane[i].dcdx -
- plane[i].eo) << TILE_ORDER;
+ (int64_t)plane[i].eo) << TILE_ORDER;
- eo[i] = plane[i].eo << TILE_ORDER;
+ eo[i] = (int64_t)plane[i].eo << TILE_ORDER;
xstep[i] = -(((int64_t)plane[i].dcdx) << TILE_ORDER);
ystep[i] = ((int64_t)plane[i].dcdy) << TILE_ORDER;
}
/**
* Calculate fixed position data for a triangle
+ * It is unfortunate we need to do that here (as we need area
+ * calculated in fixed point), as there's quite some code duplication
+ * to what is done in the jit setup prog.
*/
static inline void
-calc_fixed_position( struct lp_setup_context *setup,
- struct fixed_position* position,
- const float (*v0)[4],
- const float (*v1)[4],
- const float (*v2)[4])
+calc_fixed_position(struct lp_setup_context *setup,
+ struct fixed_position* position,
+ const float (*v0)[4],
+ const float (*v1)[4],
+ const float (*v2)[4])
{
+ /*
+ * The rounding may not be quite the same with PIPE_ARCH_SSE
+ * (util_iround right now only does nearest/even on x87,
+ * otherwise nearest/away-from-zero).
+ * Both should be acceptable, I think.
+ */
+#if defined(PIPE_ARCH_SSE)
+ __m128d v0r, v1r, v2r;
+ __m128 vxy0xy2, vxy1xy0;
+ __m128i vxy0xy2i, vxy1xy0i;
+ __m128i dxdy0120, x0x2y0y2, x1x0y1y0, x0120, y0120;
+ __m128 pix_offset = _mm_set1_ps(setup->pixel_offset);
+ __m128 fixed_one = _mm_set1_ps((float)FIXED_ONE);
+ v0r = _mm_load_sd((const double *)v0[0]);
+ v1r = _mm_load_sd((const double *)v1[0]);
+ v2r = _mm_load_sd((const double *)v2[0]);
+ vxy0xy2 = _mm_castpd_ps(_mm_unpacklo_pd(v0r, v2r));
+ vxy1xy0 = _mm_castpd_ps(_mm_unpacklo_pd(v1r, v0r));
+ vxy0xy2 = _mm_sub_ps(vxy0xy2, pix_offset);
+ vxy1xy0 = _mm_sub_ps(vxy1xy0, pix_offset);
+ vxy0xy2 = _mm_mul_ps(vxy0xy2, fixed_one);
+ vxy1xy0 = _mm_mul_ps(vxy1xy0, fixed_one);
+ vxy0xy2i = _mm_cvtps_epi32(vxy0xy2);
+ vxy1xy0i = _mm_cvtps_epi32(vxy1xy0);
+ dxdy0120 = _mm_sub_epi32(vxy0xy2i, vxy1xy0i);
+ _mm_store_si128((__m128i *)&position->dx01, dxdy0120);
+ /*
+ * For the mul, would need some more shuffles, plus emulation
+ * for the signed mul (without sse41), so don't bother.
+ */
+ x0x2y0y2 = _mm_shuffle_epi32(vxy0xy2i, _MM_SHUFFLE(3,1,2,0));
+ x1x0y1y0 = _mm_shuffle_epi32(vxy1xy0i, _MM_SHUFFLE(3,1,2,0));
+ x0120 = _mm_unpacklo_epi32(x0x2y0y2, x1x0y1y0);
+ y0120 = _mm_unpackhi_epi32(x0x2y0y2, x1x0y1y0);
+ _mm_store_si128((__m128i *)&position->x[0], x0120);
+ _mm_store_si128((__m128i *)&position->y[0], y0120);
+
+#else
position->x[0] = subpixel_snap(v0[0][0] - setup->pixel_offset);
position->x[1] = subpixel_snap(v1[0][0] - setup->pixel_offset);
position->x[2] = subpixel_snap(v2[0][0] - setup->pixel_offset);
- position->x[3] = 0;
+ position->x[3] = 0; // should be unused
position->y[0] = subpixel_snap(v0[0][1] - setup->pixel_offset);
position->y[1] = subpixel_snap(v1[0][1] - setup->pixel_offset);
position->y[2] = subpixel_snap(v2[0][1] - setup->pixel_offset);
- position->y[3] = 0;
+ position->y[3] = 0; // should be unused
position->dx01 = position->x[0] - position->x[1];
position->dy01 = position->y[0] - position->y[1];
position->dx20 = position->x[2] - position->x[0];
position->dy20 = position->y[2] - position->y[0];
+#endif
position->area = IMUL64(position->dx01, position->dy20) -
IMUL64(position->dx20, position->dy01);
/**
* Draw triangle if it's CW, cull otherwise.
*/
-static void triangle_cw( struct lp_setup_context *setup,
- const float (*v0)[4],
- const float (*v1)[4],
- const float (*v2)[4] )
+static void triangle_cw(struct lp_setup_context *setup,
+ const float (*v0)[4],
+ const float (*v1)[4],
+ const float (*v2)[4])
{
- struct fixed_position position;
+ PIPE_ALIGN_VAR(16) struct fixed_position position;
calc_fixed_position(setup, &position, v0, v1, v2);
}
-static void triangle_ccw( struct lp_setup_context *setup,
- const float (*v0)[4],
- const float (*v1)[4],
- const float (*v2)[4])
+static void triangle_ccw(struct lp_setup_context *setup,
+ const float (*v0)[4],
+ const float (*v1)[4],
+ const float (*v2)[4])
{
- struct fixed_position position;
+ PIPE_ALIGN_VAR(16) struct fixed_position position;
calc_fixed_position(setup, &position, v0, v1, v2);
/**
* Draw triangle whether it's CW or CCW.
*/
-static void triangle_both( struct lp_setup_context *setup,
- const float (*v0)[4],
- const float (*v1)[4],
- const float (*v2)[4] )
+static void triangle_both(struct lp_setup_context *setup,
+ const float (*v0)[4],
+ const float (*v1)[4],
+ const float (*v2)[4])
{
- struct fixed_position position;
+ PIPE_ALIGN_VAR(16) struct fixed_position position;
struct llvmpipe_context *lp_context = (struct llvmpipe_context *)setup->pipe;
if (lp_context->active_statistics_queries &&
static void
compute_vertex_info(struct llvmpipe_context *llvmpipe)
{
- const struct lp_fragment_shader *lpfs = llvmpipe->fs;
+ const struct tgsi_shader_info *fsInfo = &llvmpipe->fs->info.base;
struct vertex_info *vinfo = &llvmpipe->vertex_info;
int vs_index;
uint i;
draw_prepare_shader_outputs(llvmpipe->draw);
- llvmpipe->color_slot[0] = 0;
- llvmpipe->color_slot[1] = 0;
- llvmpipe->bcolor_slot[0] = 0;
- llvmpipe->bcolor_slot[1] = 0;
- llvmpipe->viewport_index_slot = 0;
- llvmpipe->layer_slot = 0;
- llvmpipe->face_slot = 0;
- llvmpipe->psize_slot = 0;
+ /*
+ * Those can't actually be 0 (because pos is always at 0).
+ * But use ints anyway to avoid confusion (in vs outputs, they
+ * can very well be at pos 0).
+ */
+ llvmpipe->color_slot[0] = -1;
+ llvmpipe->color_slot[1] = -1;
+ llvmpipe->bcolor_slot[0] = -1;
+ llvmpipe->bcolor_slot[1] = -1;
+ llvmpipe->viewport_index_slot = -1;
+ llvmpipe->layer_slot = -1;
+ llvmpipe->face_slot = -1;
+ llvmpipe->psize_slot = -1;
/*
* Match FS inputs against VS outputs, emitting the necessary
vinfo->num_attribs = 0;
vs_index = draw_find_shader_output(llvmpipe->draw,
- TGSI_SEMANTIC_POSITION,
- 0);
+ TGSI_SEMANTIC_POSITION, 0);
- draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, vs_index);
+ draw_emit_vertex_attr(vinfo, EMIT_4F, vs_index);
- for (i = 0; i < lpfs->info.base.num_inputs; i++) {
+ for (i = 0; i < fsInfo->num_inputs; i++) {
/*
* Search for each input in current vs output:
*/
-
vs_index = draw_find_shader_output(llvmpipe->draw,
- lpfs->info.base.input_semantic_name[i],
- lpfs->info.base.input_semantic_index[i]);
+ fsInfo->input_semantic_name[i],
+ fsInfo->input_semantic_index[i]);
- if (lpfs->info.base.input_semantic_name[i] == TGSI_SEMANTIC_COLOR &&
- lpfs->info.base.input_semantic_index[i] < 2) {
- int idx = lpfs->info.base.input_semantic_index[i];
- llvmpipe->color_slot[idx] = vinfo->num_attribs;
+ if (fsInfo->input_semantic_name[i] == TGSI_SEMANTIC_COLOR &&
+ fsInfo->input_semantic_index[i] < 2) {
+ int idx = fsInfo->input_semantic_index[i];
+ llvmpipe->color_slot[idx] = (int)vinfo->num_attribs;
}
- if (lpfs->info.base.input_semantic_name[i] == TGSI_SEMANTIC_FACE) {
- llvmpipe->face_slot = vinfo->num_attribs;
- draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_CONSTANT, vs_index);
- } else if (lpfs->info.base.input_semantic_name[i] == TGSI_SEMANTIC_PRIMID) {
- draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_CONSTANT, vs_index);
+ if (fsInfo->input_semantic_name[i] == TGSI_SEMANTIC_FACE) {
+ llvmpipe->face_slot = (int)vinfo->num_attribs;
+ draw_emit_vertex_attr(vinfo, EMIT_4F, vs_index);
/*
* For vp index and layer, if the fs requires them but the vs doesn't
- * provide them, store the slot - we'll later replace the data directly
- * with zero (as required by ARB_fragment_layer_viewport). This is
- * because draw itself just redirects them to whatever was at output 0.
- * We'll also store the real vpindex/layer slot for setup use.
+ * provide them, draw (vbuf) will give us the required 0 (slot -1).
+ * (This means in this case we'll also use those slots in setup, which
+ * isn't necessary but they'll contain the correct (0) value.)
*/
- } else if (lpfs->info.base.input_semantic_name[i] ==
+ } else if (fsInfo->input_semantic_name[i] ==
TGSI_SEMANTIC_VIEWPORT_INDEX) {
- if (vs_index >= 0) {
- llvmpipe->viewport_index_slot = vinfo->num_attribs;
- }
- else {
- llvmpipe->fake_vpindex_slot = vinfo->num_attribs;
- }
- draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_CONSTANT, vs_index);
- } else if (lpfs->info.base.input_semantic_name[i] == TGSI_SEMANTIC_LAYER) {
- if (vs_index >= 0) {
- llvmpipe->layer_slot = vinfo->num_attribs;
- }
- else {
- llvmpipe->fake_layer_slot = vinfo->num_attribs;
- }
- draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_CONSTANT, vs_index);
+ llvmpipe->viewport_index_slot = (int)vinfo->num_attribs;
+ draw_emit_vertex_attr(vinfo, EMIT_4F, vs_index);
+ } else if (fsInfo->input_semantic_name[i] == TGSI_SEMANTIC_LAYER) {
+ llvmpipe->layer_slot = (int)vinfo->num_attribs;
+ draw_emit_vertex_attr(vinfo, EMIT_4F, vs_index);
} else {
/*
- * Emit the requested fs attribute for all but position.
+ * Note that we'd actually want to skip position (as we won't use
+ * the attribute in the fs) but can't. The reason is that we don't
+ * actually have a input/output map for setup (even though it looks
+ * like we do...). Could adjust for this though even without a map
+ * (in llvmpipe_create_fs_state()).
*/
- draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, vs_index);
+ draw_emit_vertex_attr(vinfo, EMIT_4F, vs_index);
}
}
TGSI_SEMANTIC_BCOLOR, i);
if (vs_index >= 0) {
- llvmpipe->bcolor_slot[i] = vinfo->num_attribs;
- draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, vs_index);
+ llvmpipe->bcolor_slot[i] = (int)vinfo->num_attribs;
+ draw_emit_vertex_attr(vinfo, EMIT_4F, vs_index);
}
}
TGSI_SEMANTIC_PSIZE, 0);
if (vs_index >= 0) {
- llvmpipe->psize_slot = vinfo->num_attribs;
- draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_CONSTANT, vs_index);
+ llvmpipe->psize_slot = (int)vinfo->num_attribs;
+ draw_emit_vertex_attr(vinfo, EMIT_4F, vs_index);
}
/* Figure out if we need viewport index (if it wasn't already in fs input) */
- if (llvmpipe->viewport_index_slot == 0) {
+ if (llvmpipe->viewport_index_slot < 0) {
vs_index = draw_find_shader_output(llvmpipe->draw,
TGSI_SEMANTIC_VIEWPORT_INDEX,
0);
if (vs_index >= 0) {
- llvmpipe->viewport_index_slot = vinfo->num_attribs;
- draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_CONSTANT, vs_index);
+ llvmpipe->viewport_index_slot =(int)vinfo->num_attribs;
+ draw_emit_vertex_attr(vinfo, EMIT_4F, vs_index);
}
}
/* Figure out if we need layer (if it wasn't already in fs input) */
- if (llvmpipe->layer_slot == 0) {
+ if (llvmpipe->layer_slot < 0) {
vs_index = draw_find_shader_output(llvmpipe->draw,
TGSI_SEMANTIC_LAYER,
0);
if (vs_index >= 0) {
- llvmpipe->layer_slot = vinfo->num_attribs;
- draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_CONSTANT, vs_index);
+ llvmpipe->layer_slot = (int)vinfo->num_attribs;
+ draw_emit_vertex_attr(vinfo, EMIT_4F, vs_index);
}
}
llvmpipe->dirty |= LP_NEW_SAMPLER_VIEW;
}
- if (llvmpipe->dirty & (LP_NEW_RASTERIZER |
- LP_NEW_FS |
+ if (llvmpipe->dirty & (LP_NEW_FS |
LP_NEW_VS))
- compute_vertex_info( llvmpipe );
+ compute_vertex_info(llvmpipe);
if (llvmpipe->dirty & (LP_NEW_FS |
LP_NEW_FRAMEBUFFER |
switch (shader->info.base.input_interpolate[i]) {
case TGSI_INTERPOLATE_CONSTANT:
- shader->inputs[i].interp = LP_INTERP_CONSTANT;
- break;
+ shader->inputs[i].interp = LP_INTERP_CONSTANT;
+ break;
case TGSI_INTERPOLATE_LINEAR:
- shader->inputs[i].interp = LP_INTERP_LINEAR;
- break;
+ shader->inputs[i].interp = LP_INTERP_LINEAR;
+ break;
case TGSI_INTERPOLATE_PERSPECTIVE:
- shader->inputs[i].interp = LP_INTERP_PERSPECTIVE;
- break;
+ shader->inputs[i].interp = LP_INTERP_PERSPECTIVE;
+ break;
case TGSI_INTERPOLATE_COLOR:
- shader->inputs[i].interp = LP_INTERP_COLOR;
- break;
+ shader->inputs[i].interp = LP_INTERP_COLOR;
+ break;
default:
- assert(0);
- break;
+ assert(0);
+ break;
}
switch (shader->info.base.input_semantic_name[i]) {
case TGSI_SEMANTIC_FACE:
- shader->inputs[i].interp = LP_INTERP_FACING;
- break;
+ shader->inputs[i].interp = LP_INTERP_FACING;
+ break;
case TGSI_SEMANTIC_POSITION:
- /* Position was already emitted above
- */
- shader->inputs[i].interp = LP_INTERP_POSITION;
- shader->inputs[i].src_index = 0;
- continue;
+ /* Position was already emitted above
+ */
+ shader->inputs[i].interp = LP_INTERP_POSITION;
+ shader->inputs[i].src_index = 0;
+ continue;
}
+ /* XXX this is a completely pointless index map... */
shader->inputs[i].src_index = i+1;
}
/* Potentially modify it according to twoside, etc:
*/
if (key->twoside) {
- if (vert_attr == key->color_slot && key->bcolor_slot > 0)
+ if (vert_attr == key->color_slot && key->bcolor_slot >= 0)
lp_twoside(gallivm, args, key, key->bcolor_slot, attribv);
- else if (vert_attr == key->spec_slot && key->bspec_slot > 0)
+ else if (vert_attr == key->spec_slot && key->bspec_slot >= 0)
lp_twoside(gallivm, args, key, key->bspec_slot, attribv);
}
}
*/
break;
- case LP_INTERP_ZERO:
- /*
- * The information we get from the output is bogus, replace it
- * with zero.
- */
- emit_constant_coef4(gallivm, args, slot+1, args->bld.zero);
- break;
case LP_INTERP_FACING:
emit_facing_coef(gallivm, args, slot+1);
break;
key->pad = 0;
memcpy(key->inputs, fs->inputs, key->num_inputs * sizeof key->inputs[0]);
for (i = 0; i < key->num_inputs; i++) {
- if (key->inputs[i].interp == LP_INTERP_CONSTANT) {
- if (key->inputs[i].src_index == lp->fake_vpindex_slot ||
- key->inputs[i].src_index == lp->fake_layer_slot) {
- key->inputs[i].interp = LP_INTERP_ZERO;
- }
- }
- else if (key->inputs[i].interp == LP_INTERP_COLOR) {
+ if (key->inputs[i].interp == LP_INTERP_COLOR) {
if (lp->rasterizer->flatshade)
key->inputs[i].interp = LP_INTERP_CONSTANT;
else
struct lp_setup_variant_key {
unsigned size:16;
unsigned num_inputs:8;
- unsigned color_slot:8;
- unsigned bcolor_slot:8;
- unsigned spec_slot:8;
- unsigned bspec_slot:8;
+ int color_slot:8;
+ int bcolor_slot:8;
+ int spec_slot:8;
+ int bspec_slot:8;
unsigned flatshade_first:1;
unsigned pixel_center_half:1;
unsigned twoside:1;
LLVMBuildStore(builder, res, res_ptr);
- LLVMBuildRetVoid(builder);;
+ LLVMBuildRetVoid(builder);
gallivm_verify_function(gallivm, func);
LLVMBuildStore(builder, dst[i], ptr);
}
- LLVMBuildRetVoid(builder);;
+ LLVMBuildRetVoid(builder);
gallivm_verify_function(gallivm, func);
SV_VERTEX_STRIDE,
SV_INVOCATION_INFO,
SV_THREAD_KILL,
+ SV_BASEVERTEX,
+ SV_BASEINSTANCE,
+ SV_DRAWID,
SV_UNDEFINED,
SV_LAST
};
Value *
BuildUtil::loadImm(Value *dst, double d)
{
- return mkOp1v(OP_MOV, TYPE_F64, dst ? dst : getScratch(), mkImm(d));
+ return mkOp1v(OP_MOV, TYPE_F64, dst ? dst : getScratch(8), mkImm(d));
}
Value *
return v;
} else {
- return up->getScratch();
+ return up->getScratch(eltSize);
}
}
inline LValue *
BuildUtil::mkLoadv(DataType ty, Symbol *mem, Value *ptr)
{
- LValue *dst = getScratch();
+ LValue *dst = getScratch(typeSizeof(ty));
mkLoad(ty, dst, mem, ptr);
return dst;
}
union {
struct {
uint32_t inputMask[4]; /* mask of attributes read (1 bit per scalar) */
+ bool usesDrawParameters;
} vp;
struct {
uint8_t inputPatchSize;
uint8_t clipDistances; /* number of clip distance outputs */
uint8_t cullDistances; /* number of cull distance outputs */
int8_t genUserClip; /* request user clip planes for ClipVertex */
+ uint8_t auxCBSlot; /* constant buffer index of UCP/draw data */
uint16_t ucpBase; /* base address for UCPs */
- uint8_t ucpCBSlot; /* constant buffer index of UCP data */
+ uint16_t drawInfoBase; /* base address for draw parameters */
uint8_t pointSize; /* output index for PointSize */
uint8_t instanceId; /* system value index of InstanceID */
uint8_t vertexId; /* system value index of VertexID */
emitCC (0x2f);
emitField(0x2d, 1, (insn->op == OP_NEG) || insn->src(0).mod.neg());
emitFMZ (0x2c, 1);
+ emitField(0x29, 1, insn->subOp);
emitRND (0x27, rnd, 0x2a);
emitField(0x0a, 2, util_logbase2(typeSizeof(insn->sType)));
emitField(0x08, 2, util_logbase2(typeSizeof(insn->dType)));
// for 8/16 source types, the byte/word is in subOp. word 1 is
// represented as 2.
- code[1] |= i->subOp << 0x17;
+ if (!isFloatType(i->sType))
+ code[1] |= i->subOp << 0x17;
+ else
+ code[1] |= i->subOp << 0x18;
if (sat)
code[0] |= 0x20;
x |= 2;
return x;
}
+ case TGSI_OPCODE_PK2H:
+ return 0x3;
+ case TGSI_OPCODE_UP2H:
+ return 0x1;
default:
break;
}
case TGSI_FILE_PREDICATE: return nv50_ir::FILE_PREDICATE;
case TGSI_FILE_IMMEDIATE: return nv50_ir::FILE_IMMEDIATE;
case TGSI_FILE_SYSTEM_VALUE: return nv50_ir::FILE_SYSTEM_VALUE;
- case TGSI_FILE_RESOURCE: return nv50_ir::FILE_MEMORY_GLOBAL;
+ //case TGSI_FILE_RESOURCE: return nv50_ir::FILE_MEMORY_GLOBAL;
case TGSI_FILE_SAMPLER:
case TGSI_FILE_NULL:
default:
case TGSI_SEMANTIC_TESSINNER: return nv50_ir::SV_TESS_INNER;
case TGSI_SEMANTIC_VERTICESIN: return nv50_ir::SV_VERTEX_COUNT;
case TGSI_SEMANTIC_HELPER_INVOCATION: return nv50_ir::SV_THREAD_KILL;
+ case TGSI_SEMANTIC_BASEVERTEX: return nv50_ir::SV_BASEVERTEX;
+ case TGSI_SEMANTIC_BASEINSTANCE: return nv50_ir::SV_BASEINSTANCE;
+ case TGSI_SEMANTIC_DRAWID: return nv50_ir::SV_DRAWID;
default:
assert(0);
return nv50_ir::SV_CLOCK;
case TGSI_OPCODE_ATOMUMAX:
case TGSI_OPCODE_UBFE:
case TGSI_OPCODE_UMSB:
+ case TGSI_OPCODE_UP2H:
return nv50_ir::TYPE_U32;
case TGSI_OPCODE_I2F:
case TGSI_OPCODE_I2D:
case TGSI_OPCODE_DSGE:
case TGSI_OPCODE_DSLT:
case TGSI_OPCODE_DSNE:
+ case TGSI_OPCODE_PK2H:
return nv50_ir::TYPE_U32;
case TGSI_OPCODE_I2F:
case TGSI_OPCODE_U2F:
case TGSI_OPCODE_D2F:
+ case TGSI_OPCODE_UP2H:
return nv50_ir::TYPE_F32;
case TGSI_OPCODE_I2D:
case TGSI_OPCODE_U2D:
clipVertexOutput = -1;
textureViews.resize(scan.file_max[TGSI_FILE_SAMPLER_VIEW] + 1);
- resources.resize(scan.file_max[TGSI_FILE_RESOURCE] + 1);
+ //resources.resize(scan.file_max[TGSI_FILE_RESOURCE] + 1);
info->immd.bufSize = 0;
case TGSI_SEMANTIC_SAMPLEPOS:
info->prop.fp.sampleInterp = 1;
break;
+ case TGSI_SEMANTIC_BASEVERTEX:
+ case TGSI_SEMANTIC_BASEINSTANCE:
+ case TGSI_SEMANTIC_DRAWID:
+ info->prop.vp.usesDrawParameters = true;
+ break;
default:
break;
}
}
}
break;
+/*
case TGSI_FILE_RESOURCE:
for (i = first; i <= last; ++i) {
resources[i].target = decl->Resource.Resource;
resources[i].slot = i;
}
break;
+*/
case TGSI_FILE_SAMPLER_VIEW:
for (i = first; i <= last; ++i)
textureViews[i].target = decl->SamplerView.Resource;
if (src.isIndirect(0))
mainTempsInLMem = true;
} else
+/*
if (src.getFile() == TGSI_FILE_RESOURCE) {
if (src.getIndex(0) == TGSI_RESOURCE_GLOBAL)
info->io.globalAccess |= (insn.getOpcode() == TGSI_OPCODE_LOAD) ?
0x1 : 0x2;
} else
+*/
if (src.getFile() == TGSI_FILE_OUTPUT) {
if (src.isIndirect(0)) {
// We don't know which one is accessed, just mark everything for
unsigned int r;
switch (getSrc(s).getFile()) {
+/*
case TGSI_FILE_RESOURCE:
r = getSrc(s).getIndex(0);
return translateTexture(code->resources.at(r).target);
+*/
case TGSI_FILE_SAMPLER_VIEW:
r = getSrc(s).getIndex(0);
return translateTexture(code->textureViews.at(r).target);
// don't load masked inputs, won't be assigned a slot
if (!ptr && !(info->in[idx].mask & (1 << swz)))
return loadImm(NULL, swz == TGSI_SWIZZLE_W ? 1.0f : 0.0f);
- if (!ptr && info->in[idx].sn == TGSI_SEMANTIC_FACE)
- return mkOp1v(OP_RDSV, TYPE_F32, getSSA(), mkSysVal(SV_FACE, 0));
return interpolate(src, c, shiftAddress(ptr));
} else
if (prog->getType() == Program::TYPE_GEOMETRY) {
const int idx = dst.getIndex(0);
const int idx2d = dst.is2D() ? dst.getIndex(1) : 0;
- if (dst.isMasked(c) || f == TGSI_FILE_RESOURCE)
+ if (dst.isMasked(c)/* || f == TGSI_FILE_RESOURCE*/)
return NULL;
if (dst.isIndirect(0) ||
FOR_EACH_DST_ENABLED_CHANNEL(0, c, tgsi)
mkCvt(OP_CVT, dstTy, dst0[c], srcTy, fetchSrc(0, c));
break;
+ case TGSI_OPCODE_PK2H:
+ val0 = getScratch();
+ val1 = getScratch();
+ mkCvt(OP_CVT, TYPE_F16, val0, TYPE_F32, fetchSrc(0, 0));
+ mkCvt(OP_CVT, TYPE_F16, val1, TYPE_F32, fetchSrc(0, 1));
+ FOR_EACH_DST_ENABLED_CHANNEL(0, c, tgsi)
+ mkOp3(OP_INSBF, TYPE_U32, dst0[c], val1, mkImm(0x1010), val0);
+ break;
+ case TGSI_OPCODE_UP2H:
+ src0 = fetchSrc(0, 0);
+ FOR_EACH_DST_ENABLED_CHANNEL(0, c, tgsi) {
+ geni = mkCvt(OP_CVT, TYPE_F32, dst0[c], TYPE_F16, src0);
+ geni->subOp = c & 1;
+ }
+ break;
case TGSI_OPCODE_EMIT:
/* export the saved viewport index */
if (viewport != NULL) {
for (c = 0; c < 4; ++c) {
for (i = 0; i < info->io.genUserClip; ++i) {
- Symbol *sym = mkSymbol(FILE_MEMORY_CONST, info->io.ucpCBSlot,
+ Symbol *sym = mkSymbol(FILE_MEMORY_CONST, info->io.auxCBSlot,
TYPE_F32, info->io.ucpBase + i * 16 + c * 4);
Value *ucp = mkLoadv(TYPE_F32, sym, NULL);
if (c == 0)
ld = bld.mkOp1(OP_PIXLD, TYPE_U32, i->getDef(0), bld.mkImm(0));
ld->subOp = NV50_IR_SUBOP_PIXLD_COVMASK;
break;
+ case SV_BASEVERTEX:
+ case SV_BASEINSTANCE:
+ case SV_DRAWID:
+ ld = bld.mkLoad(TYPE_U32, i->getDef(0),
+ bld.mkSymbol(FILE_MEMORY_CONST,
+ prog->driver->io.auxCBSlot,
+ TYPE_U32,
+ prog->driver->io.drawInfoBase +
+ 4 * (sv - SV_BASEVERTEX)),
+ NULL);
+ break;
default:
if (prog->getType() == Program::TYPE_TESSELLATION_EVAL && !i->perPatch)
vtx = bld.mkOp1v(OP_PFETCH, TYPE_U32, bld.getSSA(), bld.mkImm(0));
switch (i->op) {
case OP_MAD:
case OP_FMA: {
- i->op = OP_ADD;
+ ImmediateValue src0, src1 = *i->getSrc(0)->asImm();
- /* Move the immediate to the second arg, otherwise the ADD operation
- * won't be emittable
- */
- i->setSrc(1, i->getSrc(0));
+ // Move the immediate into position 1, where we know it might be
+ // emittable. However it might not be anyways, as there may be other
+ // restrictions, so move it into a separate LValue.
+ bld.setPosition(i, false);
+ i->op = OP_ADD;
+ i->setSrc(1, bld.mkMov(bld.getSSA(type), i->getSrc(0), type)->getDef(0));
i->setSrc(0, i->getSrc(2));
i->src(0).mod = i->src(2).mod;
i->setSrc(2, NULL);
- ImmediateValue src0;
if (i->src(0).getImmediate(src0))
- expr(i, src0, *i->getSrc(1)->asImm());
- if (i->saturate && !prog->getTarget()->isSatSupported(i)) {
- bld.setPosition(i, false);
- i->setSrc(1, bld.loadImm(NULL, res.data.u32));
- }
+ expr(i, src0, src1);
+ else
+ opnd(i, src1, 1);
break;
}
case OP_PFETCH:
arg = shift->getSrc(0);
offset = imm.reg.data.u32;
}
+ // We just AND'd the high bits away, which means this is effectively an
+ // unsigned value.
+ cvt->sType = TYPE_U32;
} else if (insn->op == OP_SHR &&
insn->sType == cvt->sType &&
insn->src(1).getImmediate(imm)) {
case SV_SAMPLE_INDEX: return 0;
case SV_SAMPLE_POS: return 0;
case SV_SAMPLE_MASK: return 0;
+ case SV_BASEVERTEX: return 0;
+ case SV_BASEINSTANCE: return 0;
+ case SV_DRAWID: return 0;
default:
return 0xffffffff;
}
info.bin.sourceRep = NV50_PROGRAM_IR_TGSI;
info.bin.source = tokens;
- info.io.ucpCBSlot = 15;
+ info.io.auxCBSlot = 15;
info.io.ucpBase = NV50_CB_AUX_UCP_OFFSET;
info.io.resInfoCBSlot = 15;
if (nv_dbg)
nouveau_mesa_debug = atoi(nv_dbg);
+ /* These must be set before any failure is possible, as the cleanup
+ * paths assume they're responsible for deleting them.
+ */
+ screen->drm = nouveau_drm(&dev->object);
+ screen->device = dev;
+
/*
* this is initialized to 1 in nouveau_drm_screen_create after screen
* is fully constructed and added to the global screen list.
data, size, &screen->channel);
if (ret)
return ret;
- screen->device = dev;
ret = nouveau_client_new(screen->device, &screen->client);
if (ret)
void
nouveau_screen_fini(struct nouveau_screen *screen)
{
+ int fd = screen->drm->fd;
+
nouveau_mm_destroy(screen->mm_GART);
nouveau_mm_destroy(screen->mm_VRAM);
nouveau_object_del(&screen->channel);
nouveau_device_del(&screen->device);
+ nouveau_drm_del(&screen->drm);
+ close(fd);
}
static void
struct nouveau_screen {
struct pipe_screen base;
+ struct nouveau_drm *drm;
struct nouveau_device *device;
struct nouveau_object *channel;
struct nouveau_client *client;
unsigned fence_seq, fw_sizes, last_frame_num, tmp_stride, ref_stride;
unsigned bsp_idx, vp_idx, ppp_idx;
+
+ /* End of the bsp bo where new data should be appended between one begin/end
+ * frame.
+ */
+ char *bsp_ptr;
};
struct comm {
enum pipe_video_profile profile,
unsigned chipset);
+void
+nouveau_vp3_bsp_begin(struct nouveau_vp3_decoder *dec);
+
+void
+nouveau_vp3_bsp_next(struct nouveau_vp3_decoder *dec, unsigned num_buffers,
+ const void *const *data, const unsigned *num_bytes);
+
uint32_t
-nouveau_vp3_bsp(struct nouveau_vp3_decoder *dec, union pipe_desc desc,
- struct nouveau_vp3_video_buffer *target,
- unsigned comm_seq, unsigned num_buffers,
- const void *const *data, const unsigned *num_bytes);
+nouveau_vp3_bsp_end(struct nouveau_vp3_decoder *dec, union pipe_desc desc);
void
nouveau_vp3_vp_caps(struct nouveau_vp3_decoder *dec, union pipe_desc desc,
return caps | 3;
}
+static inline struct strparm_bsp *strparm_bsp(struct nouveau_vp3_decoder *dec)
+{
+ unsigned comm_seq = dec->fence_seq;
+ struct nouveau_bo *bsp_bo = dec->bsp_bo[comm_seq % NOUVEAU_VP3_VIDEO_QDEPTH];
+ return (struct strparm_bsp *)(bsp_bo->map + 0x100);
+}
+
+void
+nouveau_vp3_bsp_begin(struct nouveau_vp3_decoder *dec)
+{
+ struct strparm_bsp *str_bsp = strparm_bsp(dec);
+
+ dec->bsp_ptr = (void *)str_bsp;
+ memset(str_bsp, 0, 0x80);
+ dec->bsp_ptr += 0x100;
+ /* Reserved for picparm_vp */
+ dec->bsp_ptr += 0x300;
+ /* Reserved for comm */
+#if !NOUVEAU_VP3_DEBUG_FENCE
+ memset(dec->bsp_ptr, 0, 0x200);
+#endif
+ dec->bsp_ptr += 0x200;
+}
+
+void
+nouveau_vp3_bsp_next(struct nouveau_vp3_decoder *dec, unsigned num_buffers,
+ const void *const *data, const unsigned *num_bytes)
+{
+#ifndef NDEBUG
+ unsigned comm_seq = dec->fence_seq;
+ struct nouveau_bo *bsp_bo = dec->bsp_bo[comm_seq % NOUVEAU_VP3_VIDEO_QDEPTH];
+#endif
+ struct strparm_bsp *str_bsp = strparm_bsp(dec);
+ int i;
+
+ for (i = 0; i < num_buffers; ++i) {
+ assert(bsp_bo->size >= str_bsp->w0[0] + num_bytes[i]);
+ memcpy(dec->bsp_ptr, data[i], num_bytes[i]);
+ dec->bsp_ptr += num_bytes[i];
+ str_bsp->w0[0] += num_bytes[i];
+ }
+}
+
uint32_t
-nouveau_vp3_bsp(struct nouveau_vp3_decoder *dec, union pipe_desc desc,
- struct nouveau_vp3_video_buffer *target,
- unsigned comm_seq, unsigned num_buffers,
- const void *const *data, const unsigned *num_bytes)
+nouveau_vp3_bsp_end(struct nouveau_vp3_decoder *dec, union pipe_desc desc)
{
enum pipe_video_format codec = u_reduce_video_profile(dec->base.profile);
+ unsigned comm_seq = dec->fence_seq;
struct nouveau_bo *bsp_bo = dec->bsp_bo[comm_seq % NOUVEAU_VP3_VIDEO_QDEPTH];
- char *bsp;
uint32_t endmarker, caps;
- struct strparm_bsp *str_bsp;
- int i;
-
- bsp = bsp_bo->map;
+ struct strparm_bsp *str_bsp = strparm_bsp(dec);
+ char *bsp = bsp_bo->map;
/*
* 0x000..0x100: picparm_bsp
* 0x200..0x500: picparm_vp
caps |= 1 << 17; // enable watchdog
caps |= 0 << 18; // do not report error to VP, so it can continue decoding what we have
caps |= 0 << 19; // if enabled, use crypto crap?
- bsp += 0x100;
- str_bsp = (struct strparm_bsp *)bsp;
- memset(str_bsp, 0, 0x80);
- str_bsp->w0[0] = 16;
+ str_bsp = strparm_bsp(dec);
str_bsp->w1[0] = 0x1;
- bsp += 0x100;
- /* Reserved for picparm_vp */
- bsp += 0x300;
- /* Reserved for comm */
-#if !NOUVEAU_VP3_DEBUG_FENCE
- memset(bsp, 0, 0x200);
-#endif
- bsp += 0x200;
- for (i = 0; i < num_buffers; ++i) {
- memcpy(bsp, data[i], num_bytes[i]);
- bsp += num_bytes[i];
- str_bsp->w0[0] += num_bytes[i];
- }
/* Append end sequence */
- *(uint32_t *)bsp = endmarker;
- bsp += 4;
- *(uint32_t *)bsp = 0x00000000;
- bsp += 4;
- *(uint32_t *)bsp = endmarker;
- bsp += 4;
- *(uint32_t *)bsp = 0x00000000;
+ *(uint32_t *)dec->bsp_ptr = endmarker;
+ dec->bsp_ptr += 4;
+ *(uint32_t *)dec->bsp_ptr = 0x00000000;
+ dec->bsp_ptr += 4;
+ *(uint32_t *)dec->bsp_ptr = endmarker;
+ dec->bsp_ptr += 4;
+ *(uint32_t *)dec->bsp_ptr = 0x00000000;
+ str_bsp->w0[0] += 16;
+
+ dec->bsp_ptr = NULL;
return caps;
}
#include "pipe/p_defines.h"
+#include <drm.h>
#include <nouveau.h>
#ifndef NV04_PFIFO_MAX_PACKET_LEN
return flags;
}
-extern struct pipe_screen *
+extern struct nouveau_screen *
nv30_screen_create(struct nouveau_device *);
-extern struct pipe_screen *
+extern struct nouveau_screen *
nv50_screen_create(struct nouveau_device *);
-extern struct pipe_screen *
+extern struct nouveau_screen *
nvc0_screen_create(struct nouveau_device *);
#endif
static const struct {
unsigned emit;
- unsigned interp;
unsigned vp30;
unsigned vp40;
unsigned ow40;
} vroute [] = {
- [TGSI_SEMANTIC_POSITION] = { EMIT_4F, INTERP_PERSPECTIVE, 0, 0, 0x00000000 },
- [TGSI_SEMANTIC_COLOR ] = { EMIT_4F, INTERP_LINEAR , 3, 1, 0x00000001 },
- [TGSI_SEMANTIC_BCOLOR ] = { EMIT_4F, INTERP_LINEAR , 1, 3, 0x00000004 },
- [TGSI_SEMANTIC_FOG ] = { EMIT_4F, INTERP_PERSPECTIVE, 5, 5, 0x00000010 },
- [TGSI_SEMANTIC_PSIZE ] = { EMIT_1F_PSIZE, INTERP_POS , 6, 6, 0x00000020 },
- [TGSI_SEMANTIC_TEXCOORD] = { EMIT_4F, INTERP_PERSPECTIVE, 8, 7, 0x00004000 },
+ [TGSI_SEMANTIC_POSITION] = { EMIT_4F, 0, 0, 0x00000000 },
+ [TGSI_SEMANTIC_COLOR ] = { EMIT_4F, 3, 1, 0x00000001 },
+ [TGSI_SEMANTIC_BCOLOR ] = { EMIT_4F, 1, 3, 0x00000004 },
+ [TGSI_SEMANTIC_FOG ] = { EMIT_4F, 5, 5, 0x00000010 },
+ [TGSI_SEMANTIC_PSIZE ] = { EMIT_1F_PSIZE, 6, 6, 0x00000020 },
+ [TGSI_SEMANTIC_TEXCOORD] = { EMIT_4F, 8, 7, 0x00004000 },
};
static bool
if (emit == EMIT_OMIT)
return false;
- draw_emit_vertex_attr(vinfo, emit, vroute[sem].interp, attrib);
+ draw_emit_vertex_attr(vinfo, emit, attrib);
format = draw_translate_vinfo_format(emit);
r->vtxfmt[attrib] = nv30_vtxfmt(&screen->base.base, format)->hw;
case PIPE_CAP_USER_VERTEX_BUFFERS:
case PIPE_CAP_COMPUTE:
case PIPE_CAP_DRAW_INDIRECT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT_PARAMS:
case PIPE_CAP_TGSI_FS_FINE_DERIVATIVE:
case PIPE_CAP_CONDITIONAL_RENDER_INVERTED:
case PIPE_CAP_SAMPLER_VIEW_TARGET:
case PIPE_CAP_SHAREABLE_SHADERS:
case PIPE_CAP_COPY_BETWEEN_COMPRESSED_AND_PLAIN_FORMATS:
case PIPE_CAP_CLEAR_TEXTURE:
+ case PIPE_CAP_DRAW_PARAMETERS:
+ case PIPE_CAP_TGSI_PACK_HALF_FLOAT:
+ case PIPE_CAP_TGSI_FS_POSITION_IS_SYSVAL:
+ case PIPE_CAP_TGSI_FS_FACE_IS_INTEGER_SYSVAL:
+ case PIPE_CAP_SHADER_BUFFER_OFFSET_ALIGNMENT:
+ case PIPE_CAP_INVALIDATE_BUFFER:
+ case PIPE_CAP_GENERATE_MIPMAP:
return 0;
case PIPE_CAP_VENDOR_ID:
case PIPE_SHADER_CAP_TGSI_DFRACEXP_DLDEXP_SUPPORTED:
case PIPE_SHADER_CAP_TGSI_FMA_SUPPORTED:
case PIPE_SHADER_CAP_TGSI_ANY_INOUT_DECL_RANGE:
+ case PIPE_SHADER_CAP_MAX_SHADER_BUFFERS:
return 0;
case PIPE_SHADER_CAP_MAX_UNROLL_ITERATIONS_HINT:
return 32;
case PIPE_SHADER_CAP_TGSI_DFRACEXP_DLDEXP_SUPPORTED:
case PIPE_SHADER_CAP_TGSI_FMA_SUPPORTED:
case PIPE_SHADER_CAP_TGSI_ANY_INOUT_DECL_RANGE:
+ case PIPE_SHADER_CAP_MAX_SHADER_BUFFERS:
return 0;
case PIPE_SHADER_CAP_MAX_UNROLL_ITERATIONS_HINT:
return 32;
#define FAIL_SCREEN_INIT(str, err) \
do { \
NOUVEAU_ERR(str, err); \
- nv30_screen_destroy(pscreen); \
- return NULL; \
+ screen->base.base.context_create = NULL; \
+ return &screen->base; \
} while(0)
-struct pipe_screen *
+struct nouveau_screen *
nv30_screen_create(struct nouveau_device *dev)
{
- struct nv30_screen *screen = CALLOC_STRUCT(nv30_screen);
+ struct nv30_screen *screen;
struct pipe_screen *pscreen;
struct nouveau_pushbuf *push;
struct nv04_fifo *fifo;
unsigned oclass = 0;
int ret, i;
- if (!screen)
- return NULL;
-
switch (dev->chipset & 0xf0) {
case 0x30:
if (RANKINE_0397_CHIPSET & (1 << (dev->chipset & 0x0f)))
if (!oclass) {
NOUVEAU_ERR("unknown 3d class for 0x%02x\n", dev->chipset);
- FREE(screen);
return NULL;
}
+ screen = CALLOC_STRUCT(nv30_screen);
+ if (!screen)
+ return NULL;
+
+ pscreen = &screen->base.base;
+ pscreen->destroy = nv30_screen_destroy;
+
/*
* Some modern apps try to use msaa without keeping in mind the
* restrictions on videomem of older cards. Resulting in dmesg saying:
if (screen->max_sample_count > 4)
screen->max_sample_count = 4;
- pscreen = &screen->base.base;
- pscreen->destroy = nv30_screen_destroy;
pscreen->get_param = nv30_screen_get_param;
pscreen->get_paramf = nv30_screen_get_paramf;
pscreen->get_shader_param = nv30_screen_get_shader_param;
nouveau_pushbuf_kick(push, push->channel);
nouveau_fence_new(&screen->base, &screen->base.fence.current, false);
- return pscreen;
+ return &screen->base;
}
struct nv50_constbuf constbuf[3][NV50_MAX_PIPE_CONSTBUFS];
uint16_t constbuf_dirty[3];
uint16_t constbuf_valid[3];
+ uint16_t constbuf_coherent[3];
struct pipe_vertex_buffer vtxbuf[PIPE_MAX_ATTRIBS];
unsigned num_vtxbufs;
+ uint32_t vtxbufs_coherent;
struct pipe_index_buffer idxbuf;
uint32_t vbo_fifo; /* bitmask of vertex elements to be pushed to FIFO */
uint32_t vbo_user; /* bitmask of vertex buffers pointing to user memory */
struct pipe_sampler_view *textures[3][PIPE_MAX_SAMPLERS];
unsigned num_textures[3];
+ uint32_t textures_coherent[3];
struct nv50_tsc_entry *samplers[3][PIPE_MAX_SAMPLERS];
unsigned num_samplers[3];
const struct pipe_resource *templ)
{
struct nouveau_device *dev = nouveau_screen(pscreen)->device;
+ struct nouveau_drm *drm = nouveau_screen(pscreen)->drm;
struct nv50_miptree *mt = CALLOC_STRUCT(nv50_miptree);
struct pipe_resource *pt = &mt->base.base;
- bool compressed = dev->drm_version >= 0x01000101;
+ bool compressed = drm->version >= 0x01000101;
int ret;
union nouveau_bo_config bo_config;
uint32_t bo_flags;
for (m = 0, i = 0; i < info->numInputs; ++i) {
switch (info->in[i].sn) {
case TGSI_SEMANTIC_POSITION:
- case TGSI_SEMANTIC_FACE:
continue;
default:
m += info->in[i].flat ? 0 : 1;
for (c = 0; c < 4; ++c)
if (info->in[i].mask & (1 << c))
info->in[i].slot[c] = nintp++;
- } else
- if (info->in[i].sn == TGSI_SEMANTIC_FACE) {
- info->in[i].slot[0] = 255;
} else {
unsigned j = info->in[i].flat ? m++ : n++;
info->bin.sourceRep = NV50_PROGRAM_IR_TGSI;
info->bin.source = (void *)prog->pipe.tokens;
- info->io.ucpCBSlot = 15;
+ info->io.auxCBSlot = 15;
info->io.ucpBase = NV50_CB_AUX_UCP_OFFSET;
info->io.genUserClip = prog->vp.clpd_nr;
nv50_hw_destroy_query(struct nv50_context *nv50, struct nv50_query *q)
{
struct nv50_hw_query *hq = nv50_hw_query(q);
+
+ if (hq->funcs && hq->funcs->destroy_query) {
+ hq->funcs->destroy_query(nv50, hq);
+ return;
+ }
+
nv50_hw_query_allocate(nv50, q, 0);
nouveau_fence_ref(NULL, &hq->fence);
FREE(hq);
unsigned i;
for (i = 0; i < hmq->num_queries; i++)
- hmq->queries[i]->funcs->destroy_query(nv50, hmq->queries[i]);
+ if (hmq->queries[i]->funcs->destroy_query)
+ hmq->queries[i]->funcs->destroy_query(nv50, hmq->queries[i]);
FREE(hmq);
}
nv50_hw_sm_destroy_query(struct nv50_context *nv50, struct nv50_hw_query *hq)
{
struct nv50_query *q = &hq->base;
- q->funcs->destroy_query(nv50, q);
+ nv50_hw_query_allocate(nv50, q, 0);
+ nouveau_fence_ref(NULL, &hq->fence);
+ FREE(hq);
}
static boolean
case PIPE_CAP_SHAREABLE_SHADERS:
case PIPE_CAP_CLEAR_TEXTURE:
case PIPE_CAP_COMPUTE:
+ case PIPE_CAP_TGSI_FS_FACE_IS_INTEGER_SYSVAL:
return 1;
case PIPE_CAP_SEAMLESS_CUBE_MAP:
return 1; /* class_3d >= NVA0_3D_CLASS; */
case PIPE_CAP_TEXTURE_GATHER_OFFSETS:
case PIPE_CAP_TGSI_VS_WINDOW_SPACE_POSITION:
case PIPE_CAP_DRAW_INDIRECT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT_PARAMS:
case PIPE_CAP_VERTEXID_NOBASE:
case PIPE_CAP_MULTISAMPLE_Z_RESOLVE: /* potentially supported on some hw */
case PIPE_CAP_RESOURCE_FROM_USER_MEMORY:
case PIPE_CAP_DEVICE_RESET_STATUS_QUERY:
case PIPE_CAP_MAX_SHADER_PATCH_VARYINGS:
+ case PIPE_CAP_DRAW_PARAMETERS:
+ case PIPE_CAP_TGSI_PACK_HALF_FLOAT:
+ case PIPE_CAP_TGSI_FS_POSITION_IS_SYSVAL:
+ case PIPE_CAP_SHADER_BUFFER_OFFSET_ALIGNMENT:
+ case PIPE_CAP_INVALIDATE_BUFFER:
+ case PIPE_CAP_GENERATE_MIPMAP:
return 0;
case PIPE_CAP_VENDOR_ID:
case PIPE_SHADER_CAP_TGSI_DFRACEXP_DLDEXP_SUPPORTED:
case PIPE_SHADER_CAP_TGSI_FMA_SUPPORTED:
case PIPE_SHADER_CAP_TGSI_ANY_INOUT_DECL_RANGE:
+ case PIPE_SHADER_CAP_MAX_SHADER_BUFFERS:
return 0;
case PIPE_SHADER_CAP_MAX_UNROLL_ITERATIONS_HINT:
return 32;
if (screen->blitter)
nv50_blitter_destroy(screen);
+ if (screen->pm.prog) {
+ screen->pm.prog->code = NULL; /* hardcoded, don't FREE */
+ nv50_program_destroy(NULL, screen->pm.prog);
+ FREE(screen->pm.prog);
+ }
nouveau_bo_ref(NULL, &screen->code);
nouveau_bo_ref(NULL, &screen->tls_bo);
}
BEGIN_NV04(push, NV50_3D(ZETA_COMP_ENABLE), 1);
- PUSH_DATA(push, screen->base.device->drm_version >= 0x01000101);
+ PUSH_DATA(push, screen->base.drm->version >= 0x01000101);
BEGIN_NV04(push, NV50_3D(RT_COMP_ENABLE(0)), 8);
for (i = 0; i < 8; ++i)
- PUSH_DATA(push, screen->base.device->drm_version >= 0x01000101);
+ PUSH_DATA(push, screen->base.drm->version >= 0x01000101);
BEGIN_NV04(push, NV50_3D(RT_CONTROL), 1);
PUSH_DATA (push, 1);
return 1;
}
-struct pipe_screen *
+struct nouveau_screen *
nv50_screen_create(struct nouveau_device *dev)
{
struct nv50_screen *screen;
if (!screen)
return NULL;
pscreen = &screen->base.base;
+ pscreen->destroy = nv50_screen_destroy;
ret = nouveau_screen_init(&screen->base, dev);
if (ret) {
chan = screen->base.channel;
- pscreen->destroy = nv50_screen_destroy;
pscreen->context_create = nv50_create;
pscreen->is_format_supported = nv50_screen_is_format_supported;
pscreen->get_param = nv50_screen_get_param;
nouveau_fence_new(&screen->base, &screen->base.fence.current, false);
- return pscreen;
+ return &screen->base;
fail:
- nv50_screen_destroy(pscreen);
- return NULL;
+ screen->base.base.context_create = NULL;
+ return &screen->base;
}
int
if (old)
nv50_screen_tic_unlock(nv50->screen, old);
+ if (views[i] && views[i]->texture) {
+ struct pipe_resource *res = views[i]->texture;
+ if (res->target == PIPE_BUFFER &&
+ (res->flags & PIPE_RESOURCE_FLAG_MAP_COHERENT))
+ nv50->textures_coherent[s] |= 1 << i;
+ else
+ nv50->textures_coherent[s] &= ~(1 << i);
+ } else {
+ nv50->textures_coherent[s] &= ~(1 << i);
+ }
+
pipe_sampler_view_reference(&nv50->textures[s][i], views[i]);
}
nv50->constbuf[s][i].u.data = cb->user_buffer;
nv50->constbuf[s][i].size = MIN2(cb->buffer_size, 0x10000);
nv50->constbuf_valid[s] |= 1 << i;
+ nv50->constbuf_coherent[s] &= ~(1 << i);
} else
if (res) {
nv50->constbuf[s][i].offset = cb->buffer_offset;
nv50->constbuf[s][i].size = MIN2(align(cb->buffer_size, 0x100), 0x10000);
nv50->constbuf_valid[s] |= 1 << i;
+ if (res->flags & PIPE_RESOURCE_FLAG_MAP_COHERENT)
+ nv50->constbuf_coherent[s] |= 1 << i;
+ else
+ nv50->constbuf_coherent[s] &= ~(1 << i);
} else {
nv50->constbuf_valid[s] &= ~(1 << i);
+ nv50->constbuf_coherent[s] &= ~(1 << i);
}
nv50->constbuf_dirty[s] |= 1 << i;
if (!vb) {
nv50->vbo_user &= ~(((1ull << count) - 1) << start_slot);
nv50->vbo_constant &= ~(((1ull << count) - 1) << start_slot);
+ nv50->vtxbufs_coherent &= ~(((1ull << count) - 1) << start_slot);
return;
}
nv50->vbo_constant |= 1 << dst_index;
else
nv50->vbo_constant &= ~(1 << dst_index);
+ nv50->vtxbufs_coherent &= ~(1 << dst_index);
} else {
nv50->vbo_user &= ~(1 << dst_index);
nv50->vbo_constant &= ~(1 << dst_index);
+
+ if (vb[i].buffer &&
+ vb[i].buffer->flags & PIPE_RESOURCE_FLAG_MAP_COHERENT)
+ nv50->vtxbufs_coherent |= (1 << dst_index);
+ else
+ nv50->vtxbufs_coherent &= ~(1 << dst_index);
}
}
}
}
so->element[i].state = nv50_format_table[fmt].vtx;
so->need_conversion = true;
+ pipe_debug_message(&nouveau_context(pipe)->debug, FALLBACK,
+ "Converting vertex element %d, no hw format %s",
+ i, util_format_name(ve->src_format));
}
so->element[i].state |= i;
BEGIN_NV04(push, NV50_3D(VERTEX_BEGIN_GL), 1);
PUSH_DATA (push, prim);
- PUSH_REFN(push, buf->bo, NOUVEAU_BO_RD | buf->domain);
nouveau_pushbuf_space(push, 8, 0, 1);
+ PUSH_REFN(push, buf->bo, NOUVEAU_BO_RD | buf->domain);
switch (index_size) {
case 4:
struct nv50_context *nv50 = nv50_context(pipe);
struct nouveau_pushbuf *push = nv50->base.pushbuf;
bool tex_dirty = false;
- int i, s;
+ int s;
/* NOTE: caller must ensure that (min_index + index_bias) is >= 0 */
nv50->vb_elt_first = info->min_index + info->index_bias;
push->kick_notify = nv50_draw_vbo_kick_notify;
- /* TODO: Instead of iterating over all the buffer resources looking for
- * coherent buffers, keep track of a context-wide count.
- */
for (s = 0; s < 3 && !nv50->cb_dirty; ++s) {
- uint32_t valid = nv50->constbuf_valid[s];
-
- while (valid && !nv50->cb_dirty) {
- const unsigned i = ffs(valid) - 1;
- struct pipe_resource *res;
-
- valid &= ~(1 << i);
- if (nv50->constbuf[s][i].user)
- continue;
-
- res = nv50->constbuf[s][i].u.buf;
- if (!res)
- continue;
-
- if (res->flags & PIPE_RESOURCE_FLAG_MAP_COHERENT)
- nv50->cb_dirty = true;
- }
+ if (nv50->constbuf_coherent[s])
+ nv50->cb_dirty = true;
}
/* If there are any coherent constbufs, flush the cache */
}
for (s = 0; s < 3 && !tex_dirty; ++s) {
- for (i = 0; i < nv50->num_textures[s] && !tex_dirty; ++i) {
- if (!nv50->textures[s][i] ||
- nv50->textures[s][i]->texture->target != PIPE_BUFFER)
- continue;
- if (nv50->textures[s][i]->texture->flags &
- PIPE_RESOURCE_FLAG_MAP_COHERENT)
- tex_dirty = true;
- }
+ if (nv50->textures_coherent[s])
+ tex_dirty = true;
}
+
if (tex_dirty) {
BEGIN_NV04(push, NV50_3D(TEX_CACHE_CTL), 1);
PUSH_DATA (push, 0x20);
PUSH_DATA (push, info->start_instance);
}
- for (i = 0; i < nv50->num_vtxbufs && !nv50->base.vbo_dirty; ++i) {
- if (!nv50->vtxbuf[i].buffer)
- continue;
- if (nv50->vtxbuf[i].buffer->flags & PIPE_RESOURCE_FLAG_MAP_COHERENT)
- nv50->base.vbo_dirty = true;
- }
+ nv50->base.vbo_dirty |= !!nv50->vtxbufs_coherent;
if (nv50->base.vbo_dirty) {
BEGIN_NV04(push, NV50_3D(VERTEX_ARRAY_FLUSH), 1);
#include "util/u_sampler.h"
#include "util/u_format.h"
+#include <nvif/class.h>
+
static void
nv98_decoder_decode_bitstream(struct pipe_video_codec *decoder,
struct pipe_video_buffer *video_target,
nv98_decoder_ppp(dec, desc, target, comm_seq);
}
+static const struct nouveau_mclass
+nv98_decoder_msvld[] = {
+ { G98_MSVLD, -1 },
+ { IGT21A_MSVLD, -1 },
+ { GT212_MSVLD, -1 },
+ {}
+};
+
+static const struct nouveau_mclass
+nv98_decoder_mspdec[] = {
+ { G98_MSPDEC, -1 },
+ { GT212_MSPDEC, -1 },
+ {}
+};
+
+static const struct nouveau_mclass
+nv98_decoder_msppp[] = {
+ { G98_MSPPP, -1 },
+ { GT212_MSPPP, -1 },
+ {}
+};
+
struct pipe_video_codec *
nv98_create_decoder(struct pipe_context *context,
const struct pipe_video_codec *templ)
}
push = dec->pushbuf;
- if (!ret)
- ret = nouveau_object_new(dec->channel[0], 0x390b1, 0x85b1, NULL, 0, &dec->bsp);
- if (!ret)
- ret = nouveau_object_new(dec->channel[1], 0x190b2, 0x85b2, NULL, 0, &dec->vp);
- if (!ret)
- ret = nouveau_object_new(dec->channel[2], 0x290b3, 0x85b3, NULL, 0, &dec->ppp);
+ if (!ret) {
+ ret = nouveau_object_mclass(dec->channel[0], nv98_decoder_msvld);
+ if (ret >= 0) {
+ ret = nouveau_object_new(dec->channel[0], 0xbeef85b1,
+ nv98_decoder_msvld[ret].oclass, NULL, 0,
+ &dec->bsp);
+ }
+ }
+
+ if (!ret) {
+ ret = nouveau_object_mclass(dec->channel[1], nv98_decoder_mspdec);
+ if (ret >= 0) {
+ ret = nouveau_object_new(dec->channel[1], 0xbeef85b2,
+ nv98_decoder_mspdec[ret].oclass, NULL, 0,
+ &dec->vp);
+ }
+ }
+
+ if (!ret) {
+ ret = nouveau_object_mclass(dec->channel[2], nv98_decoder_msppp);
+ if (ret >= 0) {
+ ret = nouveau_object_new(dec->channel[2], 0xbeef85b3,
+ nv98_decoder_msppp[ret].oclass, NULL, 0,
+ &dec->ppp);
+ }
+ }
+
if (ret)
goto fail;
bsp_size += (1 << 20) - 1;
bsp_size &= ~((1 << 20) - 1);
- ret = nouveau_bo_new(dec->bitplane_bo->device, NOUVEAU_BO_VRAM, 0, bsp_size, NULL, &tmp_bo);
+ ret = nouveau_bo_new(dec->client->device, NOUVEAU_BO_VRAM, 0, bsp_size, NULL, &tmp_bo);
if (ret) {
debug_printf("reallocating bsp %u -> %u failed with %i\n",
bsp_bo ? (unsigned)bsp_bo->size : 0, bsp_size, ret);
if (!inter_bo || bsp_bo->size * 4 > inter_bo->size) {
struct nouveau_bo *tmp_bo = NULL;
- ret = nouveau_bo_new(dec->bitplane_bo->device, NOUVEAU_BO_VRAM, 0, bsp_bo->size * 4, NULL, &tmp_bo);
+ ret = nouveau_bo_new(dec->client->device, NOUVEAU_BO_VRAM, 0, bsp_bo->size * 4, NULL, &tmp_bo);
if (ret) {
debug_printf("reallocating inter %u -> %u failed with %i\n",
inter_bo ? (unsigned)inter_bo->size : 0, (unsigned)bsp_bo->size * 4, ret);
return -1;
}
- caps = nouveau_vp3_bsp(dec, desc, target, comm_seq,
- num_buffers, data, num_bytes);
+ nouveau_vp3_bsp_begin(dec);
+ nouveau_vp3_bsp_next(dec, num_buffers, data, num_bytes);
+ caps = nouveau_vp3_bsp_end(dec, desc);
nouveau_vp3_vp_caps(dec, desc, target, comm_seq, vp_caps, is_ref, refs);
* Forcefully sets VERTEX_ID_BASE to the value of VB_ELEMENT_BASE.
*
* arg = mode
- * parm[0] = count
- * parm[1] = instance_count
- * parm[2] = start
- * parm[3] = index_bias
- * parm[4] = start_instance
+ * parm[0] = start_drawid
+ * parm[1] = numparams
+ * parm[2 + 5n + 0] = count
+ * parm[2 + 5n + 1] = instance_count
+ * parm[2 + 5n + 2] = start
+ * parm[2 + 5n + 3] = index_bias
+ * parm[2 + 5n + 4] = start_instance
+ *
+ * SCRATCH[0] = saved VB_ELEMENT_BASE
+ * SCRATCH[1] = saved VB_INSTANCE_BASE
*/
.section #mme9097_draw_elts_indirect
+ read $r6 0x50d /* VB_ELEMENT_BASE */
+ read $r7 0x50e /* VB_INSTANCE_BASE */
+ maddr 0x1d00
+ send $r6 /* SCRATCH[0] = VB_ELEMENT_BASE */
+ send $r7 /* SCRATCH[1] = VB_INSTANCE_BASE */
+ parm $r6 /* start_drawid */
+ parm $r7 /* numparams */
+dei_draw_again:
parm $r3 /* count */
parm $r2 /* instance_count */
parm $r4 maddr 0x5f7 /* INDEX_BATCH_FIRST, start */
parm $r4 send $r4 /* index_bias, send start */
+ maddr 0x18e3 /* CB_POS */
+ send 0x180 /* 256 + 128 */
braz $r2 #dei_end
- parm $r5 /* start_instance */
- read $r6 0x50d /* VB_ELEMENT_BASE */
- read $r7 0x50e /* VB_INSTANCE_BASE */
+ parm $r5 send $r4 /* start_instance, send index_bias */
+ send $r5 /* send start_instance */
+ send $r6 /* draw id */
maddr 0x150d /* VB_ELEMENT,INSTANCE_BASE */
send $r4
send $r5
maddr 0x446
send $r4
mov $r4 0x1
+ mov $r1 (extrinsrt $r1 0x0 0 1 26) /* clear INSTANCE_NEXT */
dei_again:
maddr 0x586 /* VERTEX_BEGIN_GL */
send $r1 /* mode */
maddrsend 0x585 /* VERTEX_END_GL */
branz $r2 #dei_again
mov $r1 (extrinsrt $r1 $r4 0 1 26) /* set INSTANCE_NEXT */
+dei_end:
+ mov $r7 (add $r7 -1)
+ branz $r7 #dei_draw_again
+ mov $r6 (add $r6 1)
+ read $r6 0xd00
+ read $r7 0xd01
maddr 0x150d /* VB_ELEMENT,INSTANCE_BASE */
send $r6
send $r7
exit maddr 0x446
send $r6
-dei_end:
- exit
- nop
/* NVC0_3D_MACRO_DRAW_ARRAYS_INDIRECT:
*
* NOTE: Saves and restores VB_INSTANCE_BASE.
*
* arg = mode
- * parm[0] = count
- * parm[1] = instance_count
- * parm[2] = start
- * parm[3] = start_instance
+ * parm[0] = start_drawid
+ * parm[1] = numparams
+ * parm[2 + 4n + 0] = count
+ * parm[2 + 4n + 1] = instance_count
+ * parm[2 + 4n + 2] = start
+ * parm[2 + 4n + 3] = start_instance
*/
.section #mme9097_draw_arrays_indirect
+ read $r5 0x50e /* VB_INSTANCE_BASE */
+ parm $r6 /* start_drawid */
+ parm $r7 /* numparams */
+dai_draw_again:
parm $r2 /* count */
parm $r3 /* instance_count */
parm $r4 maddr 0x35d /* VERTEX_BUFFER_FIRST, start */
- parm $r4 send $r4 /* start_instance */
braz $r3 #dai_end
- read $r6 0x50e /* VB_INSTANCE_BASE */
+ parm $r4 send $r4 /* start_instance */
+ maddr 0x18e3 /* CB_POS */
+ send 0x180 /* 256 + 128 */
+ send 0x0 /* send 0 as base_vertex */
+ send $r4 /* send start_instance */
+ send $r6 /* draw id */
maddr 0x50e /* VB_INSTANCE_BASE */
- mov $r5 0x1
send $r4
+ mov $r4 0x1
+ mov $r1 (extrinsrt $r1 0x0 0 1 26) /* clear INSTANCE_NEXT */
dai_again:
maddr 0x586 /* VERTEX_BEGIN_GL */
send $r1 /* mode */
maddr 0x35e /* VERTEX_BUFFER_COUNT */
send $r2
- mov $r3 (sub $r3 $r5)
+ mov $r3 (sub $r3 $r4)
maddrsend 0x585 /* VERTEX_END_GL */
branz $r3 #dai_again
- mov $r1 (extrinsrt $r1 $r5 0 1 26) /* set INSTANCE_NEXT */
+ mov $r1 (extrinsrt $r1 $r4 0 1 26) /* set INSTANCE_NEXT */
+dai_end:
+ mov $r7 (add $r7 -1)
+ branz $r7 #dai_draw_again
+ mov $r6 (add $r6 1)
exit maddr 0x50e /* VB_INSTANCE_BASE to restore */
+ send $r5
+
+/* NVC0_3D_MACRO_DRAW_ELEMENTS_INDIRECT_COUNT
+ *
+ * NOTE: Saves and restores VB_ELEMENT,INSTANCE_BASE.
+ * Forcefully sets VERTEX_ID_BASE to the value of VB_ELEMENT_BASE.
+ *
+ * arg = mode
+ * parm[0] = start_drawid
+ * parm[1] = numparams
+ * parm[2] = totaldraws
+ * parm[3 + 5n + 0] = count
+ * parm[3 + 5n + 1] = instance_count
+ * parm[3 + 5n + 2] = start
+ * parm[3 + 5n + 3] = index_bias
+ * parm[3 + 5n + 4] = start_instance
+ *
+ * SCRATCH[0] = saved VB_ELEMENT_BASE
+ * SCRATCH[1] = saved VB_INSTANCE_BASE
+ * SCRATCH[2] = draws left
+ */
+.section #mme9097_draw_elts_indirect_count
+ read $r6 0x50d /* VB_ELEMENT_BASE */
+ read $r7 0x50e /* VB_INSTANCE_BASE */
+ maddr 0x1d00
+ send $r6 /* SCRATCH[0] = VB_ELEMENT_BASE */
+ send $r7 /* SCRATCH[1] = VB_INSTANCE_BASE */
+ parm $r6 /* start_drawid */
+ parm $r7 /* numparams */
+ parm $r5 /* totaldraws */
+ mov $r5 (sub $r5 $r6) /* draws left */
+ braz $r5 #deic_runout
+ mov $r3 (extrinsrt 0x0 $r5 31 1 0) /* extract high bit */
+ branz $r3 #deic_runout
+ send $r5
+deic_draw_again:
+ parm $r3 /* count */
+ parm $r2 /* instance_count */
+ parm $r4 maddr 0x5f7 /* INDEX_BATCH_FIRST, start */
+ parm $r4 send $r4 /* index_bias, send start */
+ maddr 0x18e3 /* CB_POS */
+ send 0x180 /* 256 + 128 */
+ braz $r2 #deic_end
+ parm $r5 send $r4 /* start_instance, send index_bias */
+ send $r5 /* send start_instance */
+ send $r6 /* draw id */
+ maddr 0x150d /* VB_ELEMENT,INSTANCE_BASE */
+ send $r4
+ send $r5
+ maddr 0x446
+ send $r4
+ mov $r4 0x1
+ mov $r1 (extrinsrt $r1 0x0 0 1 26) /* clear INSTANCE_NEXT */
+deic_again:
+ maddr 0x586 /* VERTEX_BEGIN_GL */
+ send $r1 /* mode */
+ maddr 0x5f8 /* INDEX_BATCH_COUNT */
+ send $r3 /* count */
+ mov $r2 (sub $r2 $r4)
+ maddrsend 0x585 /* VERTEX_END_GL */
+ branz $r2 #deic_again
+ mov $r1 (extrinsrt $r1 $r4 0 1 26) /* set INSTANCE_NEXT */
+deic_end:
+ read $r5 0xd02
+ mov $r5 (add $r5 -1)
+ braz $r5 #deic_runout_check
+ mov $r7 (add $r7 -1)
+ maddr 0xd02
+ send $r5
+ branz $r7 #deic_draw_again
+ mov $r6 (add $r6 1)
+deic_restore:
+ read $r6 0xd00
+ read $r7 0xd01
+ maddr 0x150d /* VB_ELEMENT,INSTANCE_BASE */
send $r6
-dai_end:
- exit
- nop
+ send $r7
+ exit maddr 0x446
+ send $r6
+deic_runout:
+ parm $r2
+ parm $r2
+ parm $r2
+ parm $r2
+ parm $r2
+ mov $r7 (add $r7 -1)
+deic_runout_check:
+ branz annul $r7 #deic_runout
+ bra annul #deic_restore
+
+/* NVC0_3D_MACRO_DRAW_ARRAYS_INDIRECT_COUNT:
+ *
+ * NOTE: Saves and restores VB_INSTANCE_BASE.
+ *
+ * arg = mode
+ * parm[0] = start_drawid
+ * parm[1] = numparams
+ * parm[2] = totaldraws
+ * parm[3 + 4n + 0] = count
+ * parm[3 + 4n + 1] = instance_count
+ * parm[3 + 4n + 2] = start
+ * parm[3 + 4n + 3] = start_instance
+ *
+ * SCRATCH[0] = VB_INSTANCE_BASE
+ */
+.section #mme9097_draw_arrays_indirect_count
+ read $r5 0x50e /* VB_INSTANCE_BASE */
+ maddr 0xd00
+ parm $r6 send $r5 /* start_drawid, save VB_INSTANCE_BASE */
+ parm $r7 /* numparams */
+ parm $r5 /* totaldraws */
+ mov $r5 (sub $r5 $r6) /* draws left */
+ braz $r5 #daic_runout
+ mov $r3 (extrinsrt 0x0 $r5 31 1 0) /* extract high bit */
+ branz annul $r3 #daic_runout
+daic_draw_again:
+ parm $r2 /* count */
+ parm $r3 /* instance_count */
+ parm $r4 maddr 0x35d /* VERTEX_BUFFER_FIRST, start */
+ braz $r3 #daic_end
+ parm $r4 send $r4 /* start_instance */
+ maddr 0x18e3 /* CB_POS */
+ send 0x180 /* 256 + 128 */
+ send 0x0 /* send 0 as base_vertex */
+ send $r4 /* send start_instance */
+ send $r6 /* draw id */
+ maddr 0x50e /* VB_INSTANCE_BASE */
+ send $r4
+ mov $r4 0x1
+ mov $r1 (extrinsrt $r1 0x0 0 1 26) /* clear INSTANCE_NEXT */
+daic_again:
+ maddr 0x586 /* VERTEX_BEGIN_GL */
+ send $r1 /* mode */
+ maddr 0x35e /* VERTEX_BUFFER_COUNT */
+ send $r2
+ mov $r3 (sub $r3 $r4)
+ maddrsend 0x585 /* VERTEX_END_GL */
+ branz $r3 #daic_again
+ mov $r1 (extrinsrt $r1 $r4 0 1 26) /* set INSTANCE_NEXT */
+daic_end:
+ mov $r5 (add $r5 -1)
+ braz $r5 #daic_runout_check
+ mov $r7 (add $r7 -1)
+ branz $r7 #daic_draw_again
+ mov $r6 (add $r6 1)
+daic_restore:
+ read $r5 0xd00
+ exit maddr 0x50e /* VB_INSTANCE_BASE to restore */
+ send $r5
+daic_runout:
+ parm $r2
+ parm $r2
+ parm $r2
+ parm $r2
+ mov $r7 (add $r7 -1)
+daic_runout_check:
+ branz annul $r7 #daic_runout
+ bra annul #daic_restore
};
uint32_t mme9097_draw_elts_indirect[] = {
+ 0x01434615,
+/* 0x0007: dei_draw_again */
+ 0x01438715,
+ 0x07400021,
+ 0x00003041,
+ 0x00003841,
+ 0x00000601,
+/* 0x0018: dei_again */
+ 0x00000701,
0x00000301,
+/* 0x0020: dei_end */
0x00000201,
0x017dc451,
-/* 0x000e: dei_again */
0x00002431,
- 0x0005d007,
- 0x00000501,
-/* 0x001b: dei_end */
- 0x01434615,
- 0x01438715,
+ 0x0638c021,
+ 0x00600041,
+ 0x0004d007,
+ 0x00002531,
+ 0x00002841,
+ 0x00003041,
0x05434021,
0x00002041,
0x00002841,
0x01118021,
0x00002041,
0x00004411,
+ 0xd0400912,
0x01618021,
0x00000841,
0x017e0021,
0x01614071,
0xfffe9017,
0xd0410912,
+ 0xffffff11,
+ 0xfff9b817,
+ 0x00007611,
+ 0x03400615,
+ 0x03404715,
0x05434021,
0x00003041,
0x00003841,
0x011180a1,
0x00003041,
- 0x00000091,
- 0x00000011,
};
uint32_t mme9097_draw_arrays_indirect[] = {
+/* 0x0003: dai_draw_again */
+ 0x01438515,
+ 0x00000601,
+ 0x00000701,
0x00000201,
+/* 0x0011: dai_again */
0x00000301,
-/* 0x0009: dai_again */
0x00d74451,
+/* 0x0019: dai_end */
+ 0x0004d807,
0x00002431,
-/* 0x0013: dai_end */
- 0x0003d807,
- 0x01438615,
+ 0x0638c021,
+ 0x00600041,
+ 0x00000041,
+ 0x00002041,
+ 0x00003041,
0x01438021,
- 0x00004511,
0x00002041,
+ 0x00004411,
+ 0xd0400912,
0x01618021,
0x00000841,
0x00d78021,
0x00001041,
- 0x00055b10,
+ 0x00051b10,
0x01614071,
0xfffe9817,
- 0xd0414912,
+ 0xd0410912,
+ 0xffffff11,
+ 0xfffa7817,
+ 0x00007611,
0x014380a1,
+ 0x00002841,
+};
+
+uint32_t mme9097_draw_elts_indirect_count[] = {
+ 0x01434615,
+ 0x01438715,
+ 0x07400021,
+/* 0x000d: deic_draw_again */
0x00003041,
- 0x00000091,
- 0x00000011,
+ 0x00003841,
+ 0x00000601,
+ 0x00000701,
+/* 0x001e: deic_again */
+ 0x00000501,
+ 0x0005ad10,
+/* 0x0026: deic_end */
+ 0x000b2807,
+ 0x007f4312,
+/* 0x002e: deic_restore */
+ 0x000a9817,
+ 0x00002841,
+/* 0x0035: deic_runout */
+ 0x00000301,
+/* 0x003b: deic_runout_check */
+ 0x00000201,
+ 0x017dc451,
+ 0x00002431,
+ 0x0638c021,
+ 0x00600041,
+ 0x0004d007,
+ 0x00002531,
+ 0x00002841,
+ 0x00003041,
+ 0x05434021,
+ 0x00002041,
+ 0x00002841,
+ 0x01118021,
+ 0x00002041,
+ 0x00004411,
+ 0xd0400912,
+ 0x01618021,
+ 0x00000841,
+ 0x017e0021,
+ 0x00001841,
+ 0x00051210,
+ 0x01614071,
+ 0xfffe9017,
+ 0xd0410912,
+ 0x03408515,
+ 0xffffed11,
+ 0x0004e807,
+ 0xffffff11,
+ 0x03408021,
+ 0x00002841,
+ 0xfff87817,
+ 0x00007611,
+ 0x03400615,
+ 0x03404715,
+ 0x05434021,
+ 0x00003041,
+ 0x00003841,
+ 0x011180a1,
+ 0x00003041,
+ 0x00000201,
+ 0x00000201,
+ 0x00000201,
+ 0x00000201,
+ 0x00000201,
+ 0xffffff11,
+ 0xfffeb837,
+ 0xfffc8027,
+};
+
+uint32_t mme9097_draw_arrays_indirect_count[] = {
+ 0x01438515,
+ 0x03400021,
+/* 0x0009: daic_draw_again */
+ 0x00002e31,
+ 0x00000701,
+ 0x00000501,
+/* 0x0017: daic_again */
+ 0x0005ad10,
+ 0x00086807,
+/* 0x001f: daic_end */
+ 0x007f4312,
+ 0x0007d837,
+/* 0x0024: daic_restore */
+/* 0x0027: daic_runout */
+ 0x00000201,
+ 0x00000301,
+/* 0x002c: daic_runout_check */
+ 0x00d74451,
+ 0x0004d807,
+ 0x00002431,
+ 0x0638c021,
+ 0x00600041,
+ 0x00000041,
+ 0x00002041,
+ 0x00003041,
+ 0x01438021,
+ 0x00002041,
+ 0x00004411,
+ 0xd0400912,
+ 0x01618021,
+ 0x00000841,
+ 0x00d78021,
+ 0x00001041,
+ 0x00051b10,
+ 0x01614071,
+ 0xfffe9817,
+ 0xd0410912,
+ 0xffffed11,
+ 0x00032807,
+ 0xffffff11,
+ 0xfff9f817,
+ 0x00007611,
+ 0x03400515,
+ 0x014380a1,
+ 0x00002841,
+ 0x00000201,
+ 0x00000201,
+ 0x00000201,
+ 0x00000201,
+ 0xffffff11,
+ 0xfffef837,
+ 0xfffdc027,
};
int ret;
ret = !nvc0_compute_state_validate(nvc0);
- if (ret)
- goto out;
+ if (ret) {
+ NOUVEAU_ERR("Failed to launch grid !\n");
+ return;
+ }
nvc0_compute_upload_input(nvc0, input);
/* rebind all the 3D constant buffers
* (looks like binding a CB on COMPUTE clobbers 3D state) */
nvc0->dirty |= NVC0_NEW_CONSTBUF;
- for (s = 0; s < 6; s++) {
+ for (s = 0; s < 5; s++) {
for (i = 0; i < NVC0_MAX_PIPE_CONSTBUFS; i++)
if (nvc0->constbuf[s][i].u.buf)
nvc0->constbuf_dirty[s] |= 1 << i;
}
memset(nvc0->state.uniform_buffer_bound, 0,
sizeof(nvc0->state.uniform_buffer_bound));
-
-out:
- if (ret)
- NOUVEAU_ERR("Failed to launch grid !\n");
}
struct nvc0_constbuf constbuf[6][NVC0_MAX_PIPE_CONSTBUFS];
uint16_t constbuf_dirty[6];
uint16_t constbuf_valid[6];
+ uint16_t constbuf_coherent[6];
bool cb_dirty;
struct pipe_vertex_buffer vtxbuf[PIPE_MAX_ATTRIBS];
unsigned num_vtxbufs;
+ uint32_t vtxbufs_coherent;
struct pipe_index_buffer idxbuf;
uint32_t constant_vbos;
uint32_t vbo_user; /* bitmask of vertex buffers pointing to user memory */
struct pipe_sampler_view *textures[6][PIPE_MAX_SAMPLERS];
unsigned num_textures[6];
uint32_t textures_dirty[6];
+ uint32_t textures_coherent[6];
struct nv50_tsc_entry *samplers[6][PIPE_MAX_SAMPLERS];
unsigned num_samplers[6];
uint16_t samplers_dirty[6];
#define NVC0_3D_MACRO_DRAW_ELEMENTS_INDIRECT 0x00003840
+#define NVC0_3D_MACRO_DRAW_ARRAYS_INDIRECT_COUNT 0x00003848
+
+#define NVC0_3D_MACRO_DRAW_ELEMENTS_INDIRECT_COUNT 0x00003850
+
#endif /* __NVC0_MACROS_H__ */
const struct pipe_resource *templ)
{
struct nouveau_device *dev = nouveau_screen(pscreen)->device;
+ struct nouveau_drm *drm = nouveau_screen(pscreen)->drm;
struct nv50_miptree *mt = CALLOC_STRUCT(nv50_miptree);
struct pipe_resource *pt = &mt->base.base;
- bool compressed = dev->drm_version >= 0x01000101;
+ bool compressed = drm->version >= 0x01000101;
int ret;
union nouveau_bo_config bo_config;
uint32_t bo_flags;
case TGSI_SEMANTIC_INSTANCEID: return 0x2f8;
case TGSI_SEMANTIC_VERTEXID: return 0x2fc;
case TGSI_SEMANTIC_TEXCOORD: return 0x300 + si * 0x10;
- case TGSI_SEMANTIC_FACE: return 0x3fc;
default:
assert(!"invalid TGSI input semantic");
return ~0;
break;
case PIPE_PRIM_TRIANGLES:
tp->tp.tess_mode = NVC0_3D_TESS_MODE_PRIM_TRIANGLES;
- if (info->prop.tp.winding > 0)
- tp->tp.tess_mode |= NVC0_3D_TESS_MODE_CW;
break;
case PIPE_PRIM_QUADS:
tp->tp.tess_mode = NVC0_3D_TESS_MODE_PRIM_QUADS;
tp->tp.tess_mode = ~0;
return;
}
+
+ if (info->prop.tp.winding > 0)
+ tp->tp.tess_mode |= NVC0_3D_TESS_MODE_CW;
+
if (info->prop.tp.outputPrim != PIPE_PRIM_POINTS)
tp->tp.tess_mode |= NVC0_3D_TESS_MODE_CONNECTED;
info->bin.source = (void *)prog->pipe.tokens;
info->io.genUserClip = prog->vp.num_ucps;
+ info->io.auxCBSlot = 15;
info->io.ucpBase = 256;
- info->io.ucpCBSlot = 15;
+ info->io.drawInfoBase = 256 + 128;
if (prog->type == PIPE_SHADER_COMPUTE) {
if (chipset >= NVISA_GK104_CHIPSET) {
prog->num_barriers = info->numBarriers;
prog->vp.need_vertex_id = info->io.vertexId < PIPE_MAX_SHADER_INPUTS;
+ prog->vp.need_draw_parameters = info->prop.vp.usesDrawParameters;
if (info->io.edgeFlagOut < PIPE_MAX_ATTRIBS)
info->out[info->io.edgeFlagOut].mask = 0; /* for headergen */
uint8_t num_ucps; /* also set to max if ClipDistance is used */
uint8_t edgeflag; /* attribute index of edgeflag input */
bool need_vertex_id;
+ bool need_draw_parameters;
} vp;
struct {
uint8_t early_z;
count++;
#endif
- if (screen->base.device->drm_version >= 0x01000101) {
+ if (screen->base.drm->version >= 0x01000101) {
if (screen->compute) {
if (screen->base.class_3d == NVE4_3D_CLASS) {
count += 2;
nvc0_hw_destroy_query(struct nvc0_context *nvc0, struct nvc0_query *q)
{
struct nvc0_hw_query *hq = nvc0_hw_query(q);
+
+ if (hq->funcs && hq->funcs->destroy_query) {
+ hq->funcs->destroy_query(nvc0, hq);
+ return;
+ }
+
nvc0_hw_query_allocate(nvc0, q, 0);
nouveau_fence_ref(NULL, &hq->fence);
FREE(hq);
{
struct nvc0_hw_query *hq = nvc0_hw_query(q);
-#define NVC0_IB_ENTRY_1_NO_PREFETCH (1 << (31 - 8))
-
PUSH_REFN(push, hq->bo, NOUVEAU_BO_RD | NOUVEAU_BO_GART);
- nouveau_pushbuf_space(push, 0, 0, 1);
nouveau_pushbuf_data(push, hq->bo, hq->offset + result_offset, 4 |
NVC0_IB_ENTRY_1_NO_PREFETCH);
}
unsigned i;
for (i = 0; i < hmq->num_queries; i++)
- hmq->queries[i]->funcs->destroy_query(nvc0, hmq->queries[i]);
+ if (hmq->queries[i]->funcs->destroy_query)
+ hmq->queries[i]->funcs->destroy_query(nvc0, hmq->queries[i]);
FREE(hmq);
}
{
switch (hq->base.type - NVE4_HW_METRIC_QUERY(0)) {
case NVE4_HW_METRIC_QUERY_ACHIEVED_OCCUPANCY:
- return sm20_hw_metric_calc_result(hq, res64);
+ /* (active_warps / active_cycles) / max. number of warps on a MP */
+ if (res64[1])
+ return (res64[0] / (double)res64[1]) / 64;
+ break;
case NVE4_HW_METRIC_QUERY_BRANCH_EFFICIENCY:
return sm20_hw_metric_calc_result(hq, res64);
case NVE4_HW_METRIC_QUERY_INST_ISSUED:
uint16_t class_3d = screen->base.class_3d;
int count = 0;
- if (screen->base.device->drm_version >= 0x01000101) {
+ if (screen->base.drm->version >= 0x01000101) {
if (screen->compute) {
if (screen->base.class_3d == NVE4_3D_CLASS) {
count += NVE4_HW_METRIC_QUERY_COUNT;
nvc0_hw_sm_destroy_query(struct nvc0_context *nvc0, struct nvc0_hw_query *hq)
{
struct nvc0_query *q = &hq->base;
- q->funcs->destroy_query(nvc0, q);
+ nvc0_hw_query_allocate(nvc0, q, 0);
+ nouveau_fence_ref(NULL, &hq->fence);
+ FREE(hq);
}
static boolean
return true;
}
-/* Metric calculations:
- * sum(x) ... sum of x over all MPs
- * avg(x) ... average of x over all MPs
- *
- * IPC : sum(inst_executed) / clock
- * INST_REPLAY_OHEAD: (sum(inst_issued) - sum(inst_executed)) / sum(inst_issued)
- * MP_OCCUPANCY : avg((active_warps / 64) / active_cycles)
- * MP_EFFICIENCY : avg(active_cycles / clock)
- *
- * NOTE: Interpretation of IPC requires knowledge of MP count.
- */
static boolean
nvc0_hw_sm_get_query_result(struct nvc0_context *nvc0, struct nvc0_hw_query *hq,
boolean wait, union pipe_query_result *result)
struct nvc0_hw_query *hq;
unsigned space;
- if (nvc0->screen->base.device->drm_version < 0x01000101)
+ if (nvc0->screen->base.drm->version < 0x01000101)
return NULL;
if ((type < NVE4_HW_SM_QUERY(0) || type > NVE4_HW_SM_QUERY_LAST) &&
{
int count = 0;
- if (screen->base.device->drm_version >= 0x01000101) {
+ if (screen->base.drm->version >= 0x01000101) {
if (screen->compute) {
if (screen->base.class_3d == NVE4_3D_CLASS) {
count += NVE4_HW_SM_QUERY_COUNT;
#include <xf86drm.h>
#include <nouveau_drm.h>
+#include <nvif/class.h>
#include "util/u_format.h"
#include "util/u_format_s3tc.h"
#include "pipe/p_screen.h"
case PIPE_CAP_FORCE_PERSAMPLE_INTERP:
case PIPE_CAP_SHAREABLE_SHADERS:
case PIPE_CAP_CLEAR_TEXTURE:
+ case PIPE_CAP_DRAW_PARAMETERS:
+ case PIPE_CAP_TGSI_PACK_HALF_FLOAT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT_PARAMS:
+ case PIPE_CAP_TGSI_FS_FACE_IS_INTEGER_SYSVAL:
return 1;
case PIPE_CAP_SEAMLESS_CUBE_MAP_PER_TEXTURE:
return (class_3d >= NVE4_3D_CLASS) ? 1 : 0;
case PIPE_CAP_VERTEXID_NOBASE:
case PIPE_CAP_RESOURCE_FROM_USER_MEMORY:
case PIPE_CAP_DEVICE_RESET_STATUS_QUERY:
+ case PIPE_CAP_TGSI_FS_POSITION_IS_SYSVAL:
+ case PIPE_CAP_SHADER_BUFFER_OFFSET_ALIGNMENT:
+ case PIPE_CAP_INVALIDATE_BUFFER:
+ case PIPE_CAP_GENERATE_MIPMAP:
return 0;
case PIPE_CAP_VENDOR_ID:
case PIPE_SHADER_CAP_TGSI_DFRACEXP_DLDEXP_SUPPORTED:
case PIPE_SHADER_CAP_TGSI_FMA_SUPPORTED:
case PIPE_SHADER_CAP_TGSI_ANY_INOUT_DECL_RANGE:
+ case PIPE_SHADER_CAP_MAX_SHADER_BUFFERS:
return 0;
case PIPE_SHADER_CAP_MAX_TEXTURE_SAMPLERS:
return 16; /* would be 32 in linked (OpenGL-style) mode */
if (screen->pm.prog) {
screen->pm.prog->code = NULL; /* hardcoded, don't FREE */
nvc0_program_destroy(NULL, screen->pm.prog);
+ FREE(screen->pm.prog);
}
nouveau_bo_ref(NULL, &screen->text);
#define FAIL_SCREEN_INIT(str, err) \
do { \
NOUVEAU_ERR(str, err); \
- nvc0_screen_destroy(pscreen); \
- return NULL; \
+ goto fail; \
} while(0)
-struct pipe_screen *
+struct nouveau_screen *
nvc0_screen_create(struct nouveau_device *dev)
{
struct nvc0_screen *screen;
if (!screen)
return NULL;
pscreen = &screen->base.base;
+ pscreen->destroy = nvc0_screen_destroy;
ret = nouveau_screen_init(&screen->base, dev);
if (ret) {
screen->base.vidmem_bindings = 0;
}
- pscreen->destroy = nvc0_screen_destroy;
pscreen->context_create = nvc0_create;
pscreen->is_format_supported = nvc0_screen_is_format_supported;
pscreen->get_param = nvc0_screen_get_param;
screen->base.base.is_video_format_supported = nouveau_vp3_screen_video_supported;
flags = NOUVEAU_BO_GART | NOUVEAU_BO_MAP;
- if (dev->drm_version >= 0x01000202)
+ if (screen->base.drm->version >= 0x01000202)
flags |= NOUVEAU_BO_COHERENT;
ret = nouveau_bo_new(dev, flags, 0, 4096, NULL, &screen->fence.bo);
screen->base.fence.update = nvc0_screen_fence_update;
- ret = nouveau_object_new(chan,
- (dev->chipset < 0xe0) ? 0x1f906e : 0x906e, 0x906e,
- NULL, 0, &screen->nvsw);
+ ret = nouveau_object_new(chan, (dev->chipset < 0xe0) ? 0x1f906e : 0x906e,
+ NVIF_CLASS_SW_GF100, NULL, 0, &screen->nvsw);
if (ret)
FAIL_SCREEN_INIT("Error creating SW object: %d\n", ret);
+ BEGIN_NVC0(push, SUBC_SW(NV01_SUBCHAN_OBJECT), 1);
+ PUSH_DATA (push, screen->nvsw->handle);
switch (dev->chipset & ~0xf) {
case 0x110:
PUSH_DATA (push, 0x17);
}
- IMMED_NVC0(push, NVC0_3D(ZETA_COMP_ENABLE), dev->drm_version >= 0x01000101);
+ IMMED_NVC0(push, NVC0_3D(ZETA_COMP_ENABLE),
+ screen->base.drm->version >= 0x01000101);
BEGIN_NVC0(push, NVC0_3D(RT_COMP_ENABLE(0)), 8);
for (i = 0; i < 8; ++i)
- PUSH_DATA(push, dev->drm_version >= 0x01000101);
+ PUSH_DATA(push, screen->base.drm->version >= 0x01000101);
BEGIN_NVC0(push, NVC0_3D(RT_CONTROL), 1);
PUSH_DATA (push, 1);
PUSH_DATAh(push, screen->uniform_bo->offset + (5 << 16) + (6 << 9));
PUSH_DATA (push, screen->uniform_bo->offset + (5 << 16) + (6 << 9));
- if (dev->drm_version >= 0x01000101) {
+ if (screen->base.drm->version >= 0x01000101) {
ret = nouveau_getparam(dev, NOUVEAU_GETPARAM_GRAPH_UNITS, &value);
if (ret) {
NOUVEAU_ERR("NOUVEAU_GETPARAM_GRAPH_UNITS failed.\n");
MK_MACRO(NVC0_3D_MACRO_POLYGON_MODE_BACK, mme9097_poly_mode_back);
MK_MACRO(NVC0_3D_MACRO_DRAW_ARRAYS_INDIRECT, mme9097_draw_arrays_indirect);
MK_MACRO(NVC0_3D_MACRO_DRAW_ELEMENTS_INDIRECT, mme9097_draw_elts_indirect);
+ MK_MACRO(NVC0_3D_MACRO_DRAW_ARRAYS_INDIRECT_COUNT, mme9097_draw_arrays_indirect_count);
+ MK_MACRO(NVC0_3D_MACRO_DRAW_ELEMENTS_INDIRECT_COUNT, mme9097_draw_elts_indirect_count);
BEGIN_NVC0(push, NVC0_3D(RASTERIZE_ENABLE), 1);
PUSH_DATA (push, 1);
nouveau_fence_new(&screen->base, &screen->base.fence.current, false);
- return pscreen;
+ return &screen->base;
fail:
- nvc0_screen_destroy(pscreen);
- return NULL;
+ screen->base.base.context_create = NULL;
+ return &screen->base;
}
int
if (!targ->clean)
nvc0_hw_query_fifo_wait(push, nvc0_query(targ->pq));
+ nouveau_pushbuf_space(push, 0, 0, 1);
BEGIN_NVC0(push, NVC0_3D(TFB_BUFFER_ENABLE(b)), 5);
PUSH_DATA (push, 1);
PUSH_DATAh(push, buf->address + targ->pipe.buffer_offset);
continue;
nvc0->textures_dirty[s] |= 1 << i;
+ if (views[i] && views[i]->texture) {
+ struct pipe_resource *res = views[i]->texture;
+ if (res->target == PIPE_BUFFER &&
+ (res->flags & PIPE_RESOURCE_FLAG_MAP_COHERENT))
+ nvc0->textures_coherent[s] |= 1 << i;
+ else
+ nvc0->textures_coherent[s] &= ~(1 << i);
+ } else {
+ nvc0->textures_coherent[s] &= ~(1 << i);
+ }
+
if (old) {
nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_TEX(s, i));
nvc0_screen_tic_unlock(nvc0->screen, old);
continue;
nvc0->textures_dirty[s] |= 1 << i;
+ if (views[p] && views[p]->texture) {
+ struct pipe_resource *res = views[p]->texture;
+ if (res->target == PIPE_BUFFER &&
+ (res->flags & PIPE_RESOURCE_FLAG_MAP_COHERENT))
+ nvc0->textures_coherent[s] |= 1 << i;
+ else
+ nvc0->textures_coherent[s] &= ~(1 << i);
+ } else {
+ nvc0->textures_coherent[s] &= ~(1 << i);
+ }
+
if (nvc0->textures[s][i]) {
struct nv50_tic_entry *old = nv50_tic_entry(nvc0->textures[s][i]);
nouveau_bufctx_reset(bctx, bin + i);
nvc0->constbuf[s][i].u.data = cb->user_buffer;
nvc0->constbuf[s][i].size = MIN2(cb->buffer_size, 0x10000);
nvc0->constbuf_valid[s] |= 1 << i;
+ nvc0->constbuf_coherent[s] &= ~(1 << i);
} else
if (cb) {
nvc0->constbuf[s][i].offset = cb->buffer_offset;
nvc0->constbuf[s][i].size = MIN2(align(cb->buffer_size, 0x100), 0x10000);
nvc0->constbuf_valid[s] |= 1 << i;
+ if (res && res->flags & PIPE_RESOURCE_FLAG_MAP_COHERENT)
+ nvc0->constbuf_coherent[s] |= 1 << i;
+ else
+ nvc0->constbuf_coherent[s] &= ~(1 << i);
}
else {
nvc0->constbuf_valid[s] &= ~(1 << i);
+ nvc0->constbuf_coherent[s] &= ~(1 << i);
}
}
if (!vb) {
nvc0->vbo_user &= ~(((1ull << count) - 1) << start_slot);
nvc0->constant_vbos &= ~(((1ull << count) - 1) << start_slot);
+ nvc0->vtxbufs_coherent &= ~(((1ull << count) - 1) << start_slot);
return;
}
nvc0->constant_vbos |= 1 << dst_index;
else
nvc0->constant_vbos &= ~(1 << dst_index);
+ nvc0->vtxbufs_coherent &= ~(1 << dst_index);
} else {
nvc0->vbo_user &= ~(1 << dst_index);
nvc0->constant_vbos &= ~(1 << dst_index);
+
+ if (vb[i].buffer &&
+ vb[i].buffer->flags & PIPE_RESOURCE_FLAG_MAP_COHERENT)
+ nvc0->vtxbufs_coherent |= (1 << dst_index);
+ else
+ nvc0->vtxbufs_coherent &= ~(1 << dst_index);
}
}
}
nvc0->base.pipe.render_condition(&nvc0->base.pipe, nvc0->cond_query,
nvc0->cond_cond, nvc0->cond_mode);
+ nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_VTX_TMP);
nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_FB);
nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_TEX(4, 0));
nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_TEX(4, 1));
+ nouveau_scratch_done(&nvc0->base);
nvc0->dirty = blit->saved.dirty |
(NVC0_NEW_FRAMEBUFFER | NVC0_NEW_SCISSOR | NVC0_NEW_SAMPLE_MASK |
}
so->element[i].state = nvc0_format_table[fmt].vtx;
so->need_conversion = true;
+ pipe_debug_message(&nouveau_context(pipe)->debug, FALLBACK,
+ "Converting vertex element %d, no hw format %s",
+ i, util_format_name(ve->src_format));
}
size = util_format_get_blocksize(fmt);
}
while (num_instances--) {
- PUSH_SPACE(push, 8);
+ nouveau_pushbuf_space(push, 9, 0, 1);
BEGIN_NVC0(push, NVC0_3D(VERTEX_BEGIN_GL), 1);
PUSH_DATA (push, mode);
BEGIN_NVC0(push, NVC0_3D(DRAW_TFB_BASE), 1);
{
struct nouveau_pushbuf *push = nvc0->base.pushbuf;
struct nv04_resource *buf = nv04_resource(info->indirect);
- unsigned size;
- const uint32_t offset = buf->offset + info->indirect_offset;
+ struct nv04_resource *buf_count = nv04_resource(info->indirect_params);
+ unsigned size, macro, count = info->indirect_count, drawid = info->drawid;
+ uint32_t offset = buf->offset + info->indirect_offset;
/* must make FIFO wait for engines idle before continuing to process */
- if (buf->fence_wr && !nouveau_fence_signalled(buf->fence_wr))
+ if ((buf->fence_wr && !nouveau_fence_signalled(buf->fence_wr)) ||
+ (buf_count && buf_count->fence_wr &&
+ !nouveau_fence_signalled(buf_count->fence_wr))) {
IMMED_NVC0(push, SUBC_3D(NV10_SUBCHAN_REF_CNT), 0);
+ }
+
+ /* Queue things up to let the macros write params to the driver constbuf */
+ BEGIN_NVC0(push, NVC0_3D(CB_SIZE), 3);
+ PUSH_DATA (push, 512);
+ PUSH_DATAh(push, nvc0->screen->uniform_bo->offset + (5 << 16) + (0 << 9));
+ PUSH_DATA (push, nvc0->screen->uniform_bo->offset + (5 << 16) + (0 << 9));
- PUSH_SPACE(push, 8);
if (info->indexed) {
assert(nvc0->idxbuf.buffer);
assert(nouveau_resource_mapped_by_gpu(nvc0->idxbuf.buffer));
- size = 5 * 4;
- BEGIN_1IC0(push, NVC0_3D(MACRO_DRAW_ELEMENTS_INDIRECT), 1 + size / 4);
+ size = 5;
+ if (buf_count)
+ macro = NVC0_3D_MACRO_DRAW_ELEMENTS_INDIRECT_COUNT;
+ else
+ macro = NVC0_3D_MACRO_DRAW_ELEMENTS_INDIRECT;
} else {
if (nvc0->state.index_bias) {
/* index_bias is implied 0 if !info->indexed (really ?) */
IMMED_NVC0(push, NVC0_3D(VERTEX_ID_BASE), 0);
nvc0->state.index_bias = 0;
}
- size = 4 * 4;
- BEGIN_1IC0(push, NVC0_3D(MACRO_DRAW_ARRAYS_INDIRECT), 1 + size / 4);
- }
- PUSH_DATA(push, nvc0_prim_gl(info->mode));
-#define NVC0_IB_ENTRY_1_NO_PREFETCH (1 << (31 - 8))
- PUSH_REFN(push, buf->bo, NOUVEAU_BO_RD | buf->domain);
- nouveau_pushbuf_space(push, 0, 0, 1);
- nouveau_pushbuf_data(push,
- buf->bo, offset, NVC0_IB_ENTRY_1_NO_PREFETCH | size);
+ size = 4;
+ if (buf_count)
+ macro = NVC0_3D_MACRO_DRAW_ARRAYS_INDIRECT_COUNT;
+ else
+ macro = NVC0_3D_MACRO_DRAW_ARRAYS_INDIRECT;
+ }
+
+ /* If the stride is not the natural stride, we have to stick a separate
+ * push data reference for each draw. Otherwise it can all go in as one.
+ * Of course there is a maximum packet size, so we have to break things up
+ * along those borders as well.
+ */
+ while (count) {
+ unsigned draws = count, pushes, i;
+ if (info->indirect_stride == size * 4) {
+ draws = MIN2(draws, (NV04_PFIFO_MAX_PACKET_LEN - 4) / size);
+ pushes = 1;
+ } else {
+ draws = MIN2(draws, 32);
+ pushes = draws;
+ }
+
+ nouveau_pushbuf_space(push, 16, 0, pushes + !!buf_count);
+ PUSH_REFN(push, buf->bo, NOUVEAU_BO_RD | buf->domain);
+ if (buf_count)
+ PUSH_REFN(push, buf_count->bo, NOUVEAU_BO_RD | buf_count->domain);
+ PUSH_DATA(push,
+ NVC0_FIFO_PKHDR_1I(0, macro, 3 + !!buf_count + draws * size));
+ PUSH_DATA(push, nvc0_prim_gl(info->mode));
+ PUSH_DATA(push, drawid);
+ PUSH_DATA(push, draws);
+ if (buf_count) {
+ nouveau_pushbuf_data(push,
+ buf_count->bo,
+ buf_count->offset + info->indirect_params_offset,
+ NVC0_IB_ENTRY_1_NO_PREFETCH | 4);
+ }
+ if (pushes == 1) {
+ nouveau_pushbuf_data(push,
+ buf->bo, offset,
+ NVC0_IB_ENTRY_1_NO_PREFETCH | (size * 4 * draws));
+ offset += draws * info->indirect_stride;
+ } else {
+ for (i = 0; i < pushes; i++) {
+ nouveau_pushbuf_data(push,
+ buf->bo, offset,
+ NVC0_IB_ENTRY_1_NO_PREFETCH | (size * 4));
+ offset += info->indirect_stride;
+ }
+ }
+ count -= draws;
+ drawid += draws;
+ }
}
static inline void
{
struct nvc0_context *nvc0 = nvc0_context(pipe);
struct nouveau_pushbuf *push = nvc0->base.pushbuf;
- int i, s;
+ int s;
/* NOTE: caller must ensure that (min_index + index_bias) is >= 0 */
nvc0->vb_elt_first = info->min_index + info->index_bias;
/* 8 as minimum to avoid immediate double validation of new buffers */
nvc0_state_validate(nvc0, ~0, 8);
+ if (nvc0->vertprog->vp.need_draw_parameters) {
+ BEGIN_NVC0(push, NVC0_3D(CB_SIZE), 3);
+ PUSH_DATA (push, 512);
+ PUSH_DATAh(push, nvc0->screen->uniform_bo->offset + (5 << 16) + (0 << 9));
+ PUSH_DATA (push, nvc0->screen->uniform_bo->offset + (5 << 16) + (0 << 9));
+ if (!info->indirect) {
+ BEGIN_1IC0(push, NVC0_3D(CB_POS), 1 + 3);
+ PUSH_DATA (push, 256 + 128);
+ PUSH_DATA (push, info->index_bias);
+ PUSH_DATA (push, info->start_instance);
+ PUSH_DATA (push, info->drawid);
+ }
+ }
+
push->kick_notify = nvc0_draw_vbo_kick_notify;
- /* TODO: Instead of iterating over all the buffer resources looking for
- * coherent buffers, keep track of a context-wide count.
- */
for (s = 0; s < 5 && !nvc0->cb_dirty; ++s) {
- uint32_t valid = nvc0->constbuf_valid[s];
-
- while (valid && !nvc0->cb_dirty) {
- const unsigned i = ffs(valid) - 1;
- struct pipe_resource *res;
-
- valid &= ~(1 << i);
- if (nvc0->constbuf[s][i].user)
- continue;
-
- res = nvc0->constbuf[s][i].u.buf;
- if (!res)
- continue;
-
- if (res->flags & PIPE_RESOURCE_FLAG_MAP_COHERENT)
- nvc0->cb_dirty = true;
- }
+ if (nvc0->constbuf_coherent[s])
+ nvc0->cb_dirty = true;
}
if (nvc0->cb_dirty) {
}
for (s = 0; s < 5; ++s) {
+ if (!nvc0->textures_coherent[s])
+ continue;
+
for (int i = 0; i < nvc0->num_textures[s]; ++i) {
struct nv50_tic_entry *tic = nv50_tic_entry(nvc0->textures[s][i]);
- struct pipe_resource *res;
- if (!tic)
- continue;
- res = nvc0->textures[s][i]->texture;
- if (res->target != PIPE_BUFFER ||
- !(res->flags & PIPE_RESOURCE_FLAG_MAP_COHERENT))
+ if (!(nvc0->textures_coherent[s] & (1 << i)))
continue;
BEGIN_NVC0(push, NVC0_3D(TEX_CACHE_CTL), 1);
PUSH_DATA (push, info->start_instance);
}
- for (i = 0; i < nvc0->num_vtxbufs && !nvc0->base.vbo_dirty; ++i) {
- if (!nvc0->vtxbuf[i].buffer)
- continue;
- if (nvc0->vtxbuf[i].buffer->flags & PIPE_RESOURCE_FLAG_MAP_COHERENT)
- nvc0->base.vbo_dirty = true;
- }
+ nvc0->base.vbo_dirty |= !!nvc0->vtxbufs_coherent;
if (!nvc0->base.vbo_dirty && nvc0->idxbuf.buffer &&
nvc0->idxbuf.buffer->flags & PIPE_RESOURCE_FLAG_MAP_COHERENT)
#include "util/u_format.h"
static void
+nvc0_decoder_begin_frame(struct pipe_video_codec *decoder,
+ struct pipe_video_buffer *target,
+ struct pipe_picture_desc *picture)
+{
+ struct nouveau_vp3_decoder *dec = (struct nouveau_vp3_decoder *)decoder;
+ uint32_t comm_seq = ++dec->fence_seq;
+ unsigned ret = 0;
+
+ assert(dec);
+ assert(target);
+ assert(target->buffer_format == PIPE_FORMAT_NV12);
+
+ ret = nvc0_decoder_bsp_begin(dec, comm_seq);
+
+ assert(ret == 2);
+}
+
+static void
nvc0_decoder_decode_bitstream(struct pipe_video_codec *decoder,
struct pipe_video_buffer *video_target,
struct pipe_picture_desc *picture,
const unsigned *num_bytes)
{
struct nouveau_vp3_decoder *dec = (struct nouveau_vp3_decoder *)decoder;
+ uint32_t comm_seq = dec->fence_seq;
+ unsigned ret = 0;
+
+ assert(decoder);
+
+ ret = nvc0_decoder_bsp_next(dec, comm_seq, num_buffers, data, num_bytes);
+
+ assert(ret == 2);
+}
+
+static void
+nvc0_decoder_end_frame(struct pipe_video_codec *decoder,
+ struct pipe_video_buffer *video_target,
+ struct pipe_picture_desc *picture)
+{
+ struct nouveau_vp3_decoder *dec = (struct nouveau_vp3_decoder *)decoder;
struct nouveau_vp3_video_buffer *target = (struct nouveau_vp3_video_buffer *)video_target;
- uint32_t comm_seq = ++dec->fence_seq;
+ uint32_t comm_seq = dec->fence_seq;
union pipe_desc desc;
unsigned vp_caps, is_ref, ret;
desc.base = picture;
- assert(target->base.buffer_format == PIPE_FORMAT_NV12);
-
- ret = nvc0_decoder_bsp(dec, desc, target, comm_seq,
- num_buffers, data, num_bytes,
- &vp_caps, &is_ref, refs);
+ ret = nvc0_decoder_bsp_end(dec, desc, target, comm_seq, &vp_caps, &is_ref, refs);
/* did we decode bitstream correctly? */
assert(ret == 2);
PUSH_DATA (push[2], dec->ppp->handle);
dec->base.context = context;
+ dec->base.begin_frame = nvc0_decoder_begin_frame;
dec->base.decode_bitstream = nvc0_decoder_decode_bitstream;
+ dec->base.end_frame = nvc0_decoder_end_frame;
for (i = 0; i < NOUVEAU_VP3_VIDEO_QDEPTH && !ret; ++i)
ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM,
0, 1 << 20, &cfg, &dec->bsp_bo[i]);
- if (!ret)
+ if (!ret) {
+ /* total fudge factor... just has to be bigger for higher bitrates? */
+ unsigned inter_size = align(templ->width * templ->height * 2, 4 << 20);
ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM,
- 0x100, 4 << 20, &cfg, &dec->inter_bo[0]);
+ 0x100, inter_size, &cfg, &dec->inter_bo[0]);
+ }
if (!ret) {
ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM,
0x100, dec->inter_bo[0]->size, &cfg,
#include "util/u_video.h"
extern unsigned
-nvc0_decoder_bsp(struct nouveau_vp3_decoder *dec, union pipe_desc desc,
- struct nouveau_vp3_video_buffer *target,
- unsigned comm_seq, unsigned num_buffers,
- const void *const *data, const unsigned *num_bytes,
- unsigned *vp_caps, unsigned *is_ref,
- struct nouveau_vp3_video_buffer *refs[16]);
+nvc0_decoder_bsp_begin(struct nouveau_vp3_decoder *dec, unsigned comm_seq);
+
+extern unsigned
+nvc0_decoder_bsp_next(struct nouveau_vp3_decoder *dec,
+ unsigned comm_seq, unsigned num_buffers,
+ const void *const *data, const unsigned *num_bytes);
+
+extern unsigned
+nvc0_decoder_bsp_end(struct nouveau_vp3_decoder *dec, union pipe_desc desc,
+ struct nouveau_vp3_video_buffer *target,
+ unsigned comm_seq, unsigned *vp_caps, unsigned *is_ref,
+ struct nouveau_vp3_video_buffer *refs[16]);
extern void
nvc0_decoder_vp(struct nouveau_vp3_decoder *dec, union pipe_desc desc,
#endif
unsigned
-nvc0_decoder_bsp(struct nouveau_vp3_decoder *dec, union pipe_desc desc,
- struct nouveau_vp3_video_buffer *target,
- unsigned comm_seq, unsigned num_buffers,
- const void *const *data, const unsigned *num_bytes,
- unsigned *vp_caps, unsigned *is_ref,
- struct nouveau_vp3_video_buffer *refs[16])
+nvc0_decoder_bsp_begin(struct nouveau_vp3_decoder *dec, unsigned comm_seq)
{
- struct nouveau_pushbuf *push = dec->pushbuf[0];
- enum pipe_video_format codec = u_reduce_video_profile(dec->base.profile);
- uint32_t bsp_addr, comm_addr, inter_addr;
- uint32_t slice_size, bucket_size, ring_size, bsp_size;
- uint32_t caps, i;
- int ret;
struct nouveau_bo *bsp_bo = dec->bsp_bo[comm_seq % NOUVEAU_VP3_VIDEO_QDEPTH];
- struct nouveau_bo *inter_bo = dec->inter_bo[comm_seq & 1];
- unsigned fence_extra = 0;
- struct nouveau_pushbuf_refn bo_refs[] = {
- { bsp_bo, NOUVEAU_BO_RD | NOUVEAU_BO_VRAM },
- { inter_bo, NOUVEAU_BO_WR | NOUVEAU_BO_VRAM },
-#if NOUVEAU_VP3_DEBUG_FENCE
- { dec->fence_bo, NOUVEAU_BO_WR | NOUVEAU_BO_GART },
-#endif
- { dec->bitplane_bo, NOUVEAU_BO_RDWR | NOUVEAU_BO_VRAM },
- };
- int num_refs = ARRAY_SIZE(bo_refs);
+ unsigned ret = 0;
- if (!dec->bitplane_bo)
- num_refs--;
+ ret = nouveau_bo_map(bsp_bo, NOUVEAU_BO_WR, dec->client);
+ if (ret) {
+ debug_printf("map failed: %i %s\n", ret, strerror(-ret));
+ return -1;
+ }
-#if NOUVEAU_VP3_DEBUG_FENCE
- fence_extra = 4;
-#endif
+ nouveau_vp3_bsp_begin(dec);
+
+ return 2;
+}
- bsp_size = NOUVEAU_VP3_BSP_RESERVED_SIZE;
+unsigned
+nvc0_decoder_bsp_next(struct nouveau_vp3_decoder *dec,
+ unsigned comm_seq, unsigned num_buffers,
+ const void *const *data, const unsigned *num_bytes)
+{
+ struct nouveau_bo *bsp_bo = dec->bsp_bo[comm_seq % NOUVEAU_VP3_VIDEO_QDEPTH];
+ struct nouveau_bo *inter_bo = dec->inter_bo[comm_seq & 1];
+ uint32_t bsp_size = 0;
+ uint32_t i = 0;
+ unsigned ret = 0;
+
+ bsp_size = dec->bsp_ptr - (char *)bsp_bo->map;
for (i = 0; i < num_buffers; i++)
bsp_size += num_bytes[i];
bsp_size += 256; /* the 4 end markers */
bsp_size += (1 << 20) - 1;
bsp_size &= ~((1 << 20) - 1);
- ret = nouveau_bo_new(dec->bitplane_bo->device, NOUVEAU_BO_VRAM, 0, bsp_size, &cfg, &tmp_bo);
+ ret = nouveau_bo_new(dec->client->device, NOUVEAU_BO_VRAM, 0, bsp_size, &cfg, &tmp_bo);
if (ret) {
debug_printf("reallocating bsp %u -> %u failed with %i\n",
bsp_bo ? (unsigned)bsp_bo->size : 0, bsp_size, ret);
return -1;
}
+
+ ret = nouveau_bo_map(tmp_bo, NOUVEAU_BO_WR, dec->client);
+ if (ret) {
+ debug_printf("map failed: %i %s\n", ret, strerror(-ret));
+ return -1;
+ }
+
+ /* Preserve previous buffer. */
+ /* TODO: offload this copy to the GPU, as otherwise we're reading and
+ * writing to VRAM. */
+ memcpy(tmp_bo->map, bsp_bo->map, bsp_bo->size);
+
+ /* update position to current chunk */
+ dec->bsp_ptr = tmp_bo->map + (dec->bsp_ptr - (char *)bsp_bo->map);
+
nouveau_bo_ref(NULL, &bsp_bo);
- bo_refs[0].bo = dec->bsp_bo[comm_seq % NOUVEAU_VP3_VIDEO_QDEPTH] = bsp_bo = tmp_bo;
+ dec->bsp_bo[comm_seq % NOUVEAU_VP3_VIDEO_QDEPTH] = bsp_bo = tmp_bo;
}
if (!inter_bo || bsp_bo->size * 4 > inter_bo->size) {
cfg.nvc0.tile_mode = 0x10;
cfg.nvc0.memtype = 0xfe;
- ret = nouveau_bo_new(dec->bitplane_bo->device, NOUVEAU_BO_VRAM, 0, bsp_bo->size * 4, &cfg, &tmp_bo);
+ ret = nouveau_bo_new(dec->client->device, NOUVEAU_BO_VRAM, 0, bsp_bo->size * 4, &cfg, &tmp_bo);
if (ret) {
debug_printf("reallocating inter %u -> %u failed with %i\n",
inter_bo ? (unsigned)inter_bo->size : 0, (unsigned)bsp_bo->size * 4, ret);
return -1;
}
+
+ ret = nouveau_bo_map(tmp_bo, NOUVEAU_BO_WR, dec->client);
+ if (ret) {
+ debug_printf("map failed: %i %s\n", ret, strerror(-ret));
+ return -1;
+ }
+
nouveau_bo_ref(NULL, &inter_bo);
- bo_refs[1].bo = dec->inter_bo[comm_seq & 1] = inter_bo = tmp_bo;
+ dec->inter_bo[comm_seq & 1] = inter_bo = tmp_bo;
}
- ret = nouveau_bo_map(bsp_bo, NOUVEAU_BO_WR, dec->client);
- if (ret) {
- debug_printf("map failed: %i %s\n", ret, strerror(-ret));
- return -1;
- }
+ nouveau_vp3_bsp_next(dec, num_buffers, data, num_bytes);
+
+ return 2;
+}
+
+
+unsigned
+nvc0_decoder_bsp_end(struct nouveau_vp3_decoder *dec, union pipe_desc desc,
+ struct nouveau_vp3_video_buffer *target, unsigned comm_seq,
+ unsigned *vp_caps, unsigned *is_ref,
+ struct nouveau_vp3_video_buffer *refs[16])
+{
+ struct nouveau_pushbuf *push = dec->pushbuf[0];
+ enum pipe_video_format codec = u_reduce_video_profile(dec->base.profile);
+ uint32_t bsp_addr, comm_addr, inter_addr;
+ uint32_t slice_size, bucket_size, ring_size;
+ uint32_t caps;
+ struct nouveau_bo *bsp_bo = dec->bsp_bo[comm_seq % NOUVEAU_VP3_VIDEO_QDEPTH];
+ struct nouveau_bo *inter_bo = dec->inter_bo[comm_seq & 1];
+ unsigned fence_extra = 0;
+ struct nouveau_pushbuf_refn bo_refs[] = {
+ { bsp_bo, NOUVEAU_BO_RD | NOUVEAU_BO_VRAM },
+ { inter_bo, NOUVEAU_BO_WR | NOUVEAU_BO_VRAM },
+#if NOUVEAU_VP3_DEBUG_FENCE
+ { dec->fence_bo, NOUVEAU_BO_WR | NOUVEAU_BO_GART },
+#endif
+ { dec->bitplane_bo, NOUVEAU_BO_RDWR | NOUVEAU_BO_VRAM },
+ };
+ int num_refs = ARRAY_SIZE(bo_refs);
+
+ if (!dec->bitplane_bo)
+ num_refs--;
+
+#if NOUVEAU_VP3_DEBUG_FENCE
+ fence_extra = 4;
+#endif
- caps = nouveau_vp3_bsp(dec, desc, target, comm_seq,
- num_buffers, data, num_bytes);
+ caps = nouveau_vp3_bsp_end(dec, desc);
nouveau_vp3_vp_caps(dec, desc, target, comm_seq, vp_caps, is_ref, refs);
#define SUBC_SW(m) 7, (m)
#define NVC0_3D_SERIALIZE NV50_GRAPH_SERIALIZE
+#define NVC0_IB_ENTRY_1_NO_PREFETCH (1 << (31 - 8))
static inline uint32_t
NVC0_FIFO_PKHDR_SQ(int subc, int mthd, unsigned size)
case R500_INST_TYPE_OUT: str = "OUT"; break;
case R500_INST_TYPE_FC: str = "FC"; break;
case R500_INST_TYPE_TEX: str = "TEX"; break;
- };
+ }
fprintf(stderr,"%s %s %s %s %s ", str,
inst & R500_INST_TEX_SEM_WAIT ? "TEX_WAIT" : "",
inst & R500_INST_LAST ? "LAST" : "",
r300->context.create_video_codec = vl_create_decoder;
r300->context.create_video_buffer = vl_video_buffer_create;
- r300->uploader = u_upload_create(&r300->context, 256 * 1024, 4,
- PIPE_BIND_CUSTOM);
+ r300->uploader = u_upload_create(&r300->context, 256 * 1024,
+ PIPE_BIND_CUSTOM, PIPE_USAGE_STREAM);
r300->blitter = util_blitter_create(&r300->context);
if (r300->blitter == NULL)
CS_LOCALS(r300);
DBG(r300, DBG_DRAW, "r300: render_draw_elements (count: %d)\n", count);
- u_upload_data(r300->uploader, 0, count * 2, indices,
+ u_upload_data(r300->uploader, 0, count * 2, 4, indices,
&index_buffer_offset, &index_buffer);
if (!index_buffer) {
return;
switch (*index_size) {
case 1:
*out_buffer = NULL;
- u_upload_alloc(r300->uploader, 0, count * 2,
+ u_upload_alloc(r300->uploader, 0, count * 2, 4,
&out_offset, out_buffer, &ptr);
util_shorten_ubyte_elts_to_userptr(
case 2:
if (index_offset) {
*out_buffer = NULL;
- u_upload_alloc(r300->uploader, 0, count * 2,
+ u_upload_alloc(r300->uploader, 0, count * 2, 4,
&out_offset, out_buffer, &ptr);
util_rebuild_ushort_elts_to_userptr(&r300->context, ib,
case 4:
if (index_offset) {
*out_buffer = NULL;
- u_upload_alloc(r300->uploader, 0, count * 4,
+ u_upload_alloc(r300->uploader, 0, count * 4, 4,
&out_offset, out_buffer, &ptr);
util_rebuild_uint_elts_to_userptr(&r300->context, ib,
case PIPE_CAP_SAMPLE_SHADING:
case PIPE_CAP_TEXTURE_GATHER_OFFSETS:
case PIPE_CAP_DRAW_INDIRECT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT_PARAMS:
case PIPE_CAP_TGSI_FS_FINE_DERIVATIVE:
case PIPE_CAP_CONDITIONAL_RENDER_INVERTED:
case PIPE_CAP_SAMPLER_VIEW_TARGET:
case PIPE_CAP_SHAREABLE_SHADERS:
case PIPE_CAP_COPY_BETWEEN_COMPRESSED_AND_PLAIN_FORMATS:
case PIPE_CAP_CLEAR_TEXTURE:
+ case PIPE_CAP_DRAW_PARAMETERS:
+ case PIPE_CAP_TGSI_PACK_HALF_FLOAT:
+ case PIPE_CAP_TGSI_FS_POSITION_IS_SYSVAL:
+ case PIPE_CAP_TGSI_FS_FACE_IS_INTEGER_SYSVAL:
+ case PIPE_CAP_SHADER_BUFFER_OFFSET_ALIGNMENT:
+ case PIPE_CAP_INVALIDATE_BUFFER:
+ case PIPE_CAP_GENERATE_MIPMAP:
return 0;
/* SWTCL-only features. */
case PIPE_SHADER_CAP_TGSI_DROUND_SUPPORTED:
case PIPE_SHADER_CAP_TGSI_DFRACEXP_DLDEXP_SUPPORTED:
case PIPE_SHADER_CAP_TGSI_FMA_SUPPORTED:
+ case PIPE_SHADER_CAP_MAX_SHADER_BUFFERS:
return 0;
case PIPE_SHADER_CAP_MAX_UNROLL_ITERATIONS_HINT:
return 32;
case PIPE_SHADER_CAP_TGSI_DROUND_SUPPORTED:
case PIPE_SHADER_CAP_TGSI_DFRACEXP_DLDEXP_SUPPORTED:
case PIPE_SHADER_CAP_TGSI_FMA_SUPPORTED:
+ case PIPE_SHADER_CAP_MAX_SHADER_BUFFERS:
return 0;
case PIPE_SHADER_CAP_MAX_UNROLL_ITERATIONS_HINT:
return 32;
*index_buffer = NULL;
u_upload_data(r300->uploader,
- 0, count * index_size,
+ 0, count * index_size, 4,
ptr + (*start * index_size),
&index_offset,
index_buffer);
static void r300_draw_emit_attrib(struct r300_context* r300,
enum attrib_emit emit,
- enum interp_mode interp,
int index)
{
struct r300_vertex_shader* vs = r300->vs_state.state;
output = draw_find_shader_output(r300->draw,
info->output_semantic_name[index],
info->output_semantic_index[index]);
- draw_emit_vertex_attr(&r300->vertex_info, emit, interp, output);
+ draw_emit_vertex_attr(&r300->vertex_info, emit, output);
}
static void r300_draw_emit_all_attribs(struct r300_context* r300)
/* Position. */
if (vs_outputs->pos != ATTR_UNUSED) {
- r300_draw_emit_attrib(r300, EMIT_4F, INTERP_PERSPECTIVE,
- vs_outputs->pos);
+ r300_draw_emit_attrib(r300, EMIT_4F, vs_outputs->pos);
} else {
assert(0);
}
/* Point size. */
if (vs_outputs->psize != ATTR_UNUSED) {
- r300_draw_emit_attrib(r300, EMIT_1F_PSIZE, INTERP_POS,
- vs_outputs->psize);
+ r300_draw_emit_attrib(r300, EMIT_1F_PSIZE, vs_outputs->psize);
}
/* Colors. */
for (i = 0; i < ATTR_COLOR_COUNT; i++) {
if (vs_outputs->color[i] != ATTR_UNUSED) {
- r300_draw_emit_attrib(r300, EMIT_4F, INTERP_LINEAR,
- vs_outputs->color[i]);
+ r300_draw_emit_attrib(r300, EMIT_4F, vs_outputs->color[i]);
}
}
/* Back-face colors. */
for (i = 0; i < ATTR_COLOR_COUNT; i++) {
if (vs_outputs->bcolor[i] != ATTR_UNUSED) {
- r300_draw_emit_attrib(r300, EMIT_4F, INTERP_LINEAR,
- vs_outputs->bcolor[i]);
+ r300_draw_emit_attrib(r300, EMIT_4F, vs_outputs->bcolor[i]);
}
}
for (i = 0; i < ATTR_GENERIC_COUNT && gen_count < 8; i++) {
if (vs_outputs->generic[i] != ATTR_UNUSED &&
!(r300->sprite_coord_enable & (1 << i))) {
- r300_draw_emit_attrib(r300, EMIT_4F, INTERP_PERSPECTIVE,
- vs_outputs->generic[i]);
+ r300_draw_emit_attrib(r300, EMIT_4F, vs_outputs->generic[i]);
gen_count++;
}
}
/* Fog coordinates. */
if (gen_count < 8 && vs_outputs->fog != ATTR_UNUSED) {
- r300_draw_emit_attrib(r300, EMIT_4F, INTERP_PERSPECTIVE,
- vs_outputs->fog);
+ r300_draw_emit_attrib(r300, EMIT_4F, vs_outputs->fog);
gen_count++;
}
if (r300_fs(r300)->shader->inputs.wpos != ATTR_UNUSED && gen_count < 8) {
DBG(r300, DBG_SWTCL, "draw_emit_attrib: WPOS, index: %i\n",
vs_outputs->wpos);
- r300_draw_emit_attrib(r300, EMIT_4F, INTERP_PERSPECTIVE,
- vs_outputs->wpos);
+ r300_draw_emit_attrib(r300, EMIT_4F, vs_outputs->wpos);
}
}
ctx->screen->has_compressed_msaa_texturing);
bc->type = TGSI_PROCESSOR_COMPUTE;
bc->isa = ctx->isa;
- r600_llvm_compile(mod, ctx->b.family, bc, &use_kill, dump);
+ r600_llvm_compile(mod, ctx->b.family, bc, &use_kill, dump, &ctx->b.debug);
if (dump && !sb_disasm) {
r600_bytecode_disasm(bc);
if (!gs_ring_buffer) {
radeon_set_context_reg_flag(cs, reg_alu_constbuf_size + buffer_index * 4,
- ALIGN_DIVUP(cb->buffer_size >> 4, 16), pkt_flags);
+ DIV_ROUND_UP(cb->buffer_size, 256), pkt_flags);
radeon_set_context_reg_flag(cs, reg_alu_const_cache + buffer_index * 4, va >> 8,
pkt_flags);
}
struct r600_command_buffer *cb = &rctx->start_cs_cmd;
int tmp, i;
- r600_init_command_buffer(cb, 342);
+ r600_init_command_buffer(cb, 338);
/* This must be first. */
r600_store_value(cb, PKT3(PKT3_CONTEXT_CONTROL, 1, 0));
r600_store_context_reg(cb, R_028B98_VGT_STRMOUT_BUFFER_CONFIG, 0);
- r600_store_context_reg_seq(cb, R_028AB4_VGT_REUSE_OFF, 2);
- r600_store_value(cb, 0); /* R_028AB4_VGT_REUSE_OFF */
- r600_store_value(cb, 0); /* R_028AB8_VGT_VTX_CNT_EN */
-
r600_store_config_reg(cb, R_008A14_PA_CL_ENHANCE, (3 << 1) | 1);
r600_store_context_reg_seq(cb, CM_R_028BD4_PA_SC_CENTROID_PRIORITY_0, 2);
return;
}
- r600_init_command_buffer(cb, 342);
+ r600_init_command_buffer(cb, 338);
/* This must be first. */
r600_store_value(cb, PKT3(PKT3_CONTEXT_CONTROL, 1, 0));
r600_store_value(cb, 0); /* R_028A3C_VGT_GROUP_VECT_1_FMT_CNTL */
r600_store_value(cb, 0); /* R_028A40_VGT_GS_MODE */
- r600_store_context_reg_seq(cb, R_028AB4_VGT_REUSE_OFF, 2);
- r600_store_value(cb, 0); /* R_028AB4_VGT_REUSE_OFF */
- r600_store_value(cb, 0); /* R_028AB8_VGT_VTX_CNT_EN */
-
r600_store_config_reg(cb, R_008A14_PA_CL_ENHANCE, (3 << 1) | 1);
r600_store_context_reg(cb, R_0288F0_SQ_VTX_SEMANTIC_CLEAR, ~0);
r600_init_atom(rctx, &rctx->blend_color.atom, id++, r600_emit_blend_color, 6);
r600_init_atom(rctx, &rctx->blend_state.atom, id++, r600_emit_cso_state, 0);
r600_init_atom(rctx, &rctx->cb_misc_state.atom, id++, evergreen_emit_cb_misc_state, 4);
- r600_init_atom(rctx, &rctx->clip_misc_state.atom, id++, r600_emit_clip_misc_state, 6);
+ r600_init_atom(rctx, &rctx->clip_misc_state.atom, id++, r600_emit_clip_misc_state, 9);
r600_init_atom(rctx, &rctx->clip_state.atom, id++, evergreen_emit_clip_state, 26);
r600_init_atom(rctx, &rctx->db_misc_state.atom, id++, evergreen_emit_db_misc_state, 10);
r600_init_atom(rctx, &rctx->db_state.atom, id++, evergreen_emit_db_state, 14);
* That operand should be passed as a float value in the args array
* right after the coord vector. After packing it's not used anymore,
* that's why arg_count is not increased */
- coords[4] = lp_build_emit_fetch(bld_base, inst, 1, 0);
+ coords[4] = lp_build_emit_fetch(bld_base, inst, 1, TGSI_CHAN_X);
}
if ((inst->Texture.Texture == TGSI_TEXTURE_CUBE ||
enum radeon_family family,
struct r600_bytecode *bc,
boolean *use_kill,
- unsigned dump)
+ unsigned dump,
+ struct pipe_debug_callback *debug)
{
unsigned r;
struct radeon_shader_binary binary;
const char * gpu_family = r600_get_llvm_processor_name(family);
memset(&binary, 0, sizeof(struct radeon_shader_binary));
- r = radeon_llvm_compile(mod, &binary, gpu_family, dump, dump, NULL);
+ if (dump)
+ LLVMDumpModule(mod);
+ r = radeon_llvm_compile(mod, &binary, gpu_family, NULL, debug);
r = r600_create_shader(bc, &binary, use_kill);
#include "radeon/radeon_llvm.h"
#include <llvm-c/Core.h>
+struct pipe_debug_callback;
struct r600_bytecode;
struct r600_shader_ctx;
struct radeon_llvm_context;
enum radeon_family family,
struct r600_bytecode *bc,
boolean *use_kill,
- unsigned dump);
+ unsigned dump,
+ struct pipe_debug_callback *debug);
unsigned r600_create_shader(struct r600_bytecode *bc,
const struct radeon_shader_binary *binary,
case PIPE_CAP_TEXTURE_HALF_FLOAT_LINEAR:
case PIPE_CAP_TGSI_TXQS:
case PIPE_CAP_COPY_BETWEEN_COMPRESSED_AND_PLAIN_FORMATS:
+ case PIPE_CAP_INVALIDATE_BUFFER:
return 1;
case PIPE_CAP_DEVICE_RESET_STATUS_QUERY:
case PIPE_CAP_FORCE_PERSAMPLE_INTERP:
case PIPE_CAP_SHAREABLE_SHADERS:
case PIPE_CAP_CLEAR_TEXTURE:
+ case PIPE_CAP_DRAW_PARAMETERS:
+ case PIPE_CAP_TGSI_PACK_HALF_FLOAT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT_PARAMS:
+ case PIPE_CAP_TGSI_FS_POSITION_IS_SYSVAL:
+ case PIPE_CAP_TGSI_FS_FACE_IS_INTEGER_SYSVAL:
+ case PIPE_CAP_SHADER_BUFFER_OFFSET_ALIGNMENT:
+ case PIPE_CAP_GENERATE_MIPMAP:
return 0;
case PIPE_CAP_MAX_SHADER_PATCH_VARYINGS:
case PIPE_SHADER_CAP_TGSI_DROUND_SUPPORTED:
case PIPE_SHADER_CAP_TGSI_DFRACEXP_DLDEXP_SUPPORTED:
case PIPE_SHADER_CAP_TGSI_FMA_SUPPORTED:
+ case PIPE_SHADER_CAP_MAX_SHADER_BUFFERS:
return 0;
case PIPE_SHADER_CAP_MAX_UNROLL_ITERATIONS_HINT:
/* due to a bug in the shader compiler, some loops hang
unsigned clip_plane_enable; /* from rasterizer */
unsigned clip_dist_write; /* from vertex shader */
boolean clip_disable; /* from vertex shader */
+ boolean vs_out_viewport; /* from vertex shader */
};
struct r600_alphatest_state {
{
return value * (1 << frac_bits);
}
-#define ALIGN_DIVUP(x, y) (((x) + (y) - 1) / (y))
/* 12.4 fixed-point */
static inline unsigned r600_pack_float_12p4(float x)
struct r600_context *rctx = (struct r600_context *)ctx;
struct r600_pipe_shader_selector *sel = shader->selector;
int r;
- bool dump = r600_can_dump_shader(&rctx->screen->b, sel->tokens);
+ bool dump = r600_can_dump_shader(&rctx->screen->b,
+ tgsi_get_processor_type(sel->tokens));
unsigned use_sb = !(rctx->screen->b.debug_flags & DBG_NO_SB);
unsigned sb_disasm = use_sb || (rctx->screen->b.debug_flags & DBG_SB_DISASM);
unsigned export_shader;
static int tgsi_is_supported(struct r600_shader_ctx *ctx)
{
struct tgsi_full_instruction *i = &ctx->parse.FullToken.FullInstruction;
- int j;
+ unsigned j;
if (i->Instruction.NumDstRegs > 1 && i->Instruction.Opcode != TGSI_OPCODE_DFRACEXP) {
R600_ERR("too many dst (%d)\n", i->Instruction.NumDstRegs);
*/
static int evergreen_gpr_count(struct r600_shader_ctx *ctx)
{
- int i;
+ unsigned i;
int num_baryc;
struct tgsi_parse_context parse;
static int tgsi_split_gs_inputs(struct r600_shader_ctx *ctx)
{
struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
- int i;
+ unsigned i;
for (i = 0; i < inst->Instruction.NumSrcRegs; i++) {
struct tgsi_full_src_register *src = &inst->Src[i];
static int tgsi_split_lds_inputs(struct r600_shader_ctx *ctx)
{
struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
- int i;
+ unsigned i;
for (i = 0; i < inst->Instruction.NumSrcRegs; i++) {
struct tgsi_full_src_register *src = &inst->Src[i];
static int r600_emit_tess_factor(struct r600_shader_ctx *ctx)
{
- int i;
+ unsigned i;
int stride, outer_comps, inner_comps;
int tessinner_idx = -1, tessouter_idx = -1;
int r;
if (use_llvm) {
struct radeon_llvm_context radeon_llvm_ctx;
LLVMModuleRef mod;
- bool dump = r600_can_dump_shader(&rscreen->b, tokens);
+ bool dump = r600_can_dump_shader(&rscreen->b,
+ tgsi_get_processor_type(tokens));
boolean use_kill = false;
memset(&radeon_llvm_ctx, 0, sizeof(radeon_llvm_ctx));
ctx.shader->has_txq_cube_array_z_comp = radeon_llvm_ctx.has_txq_cube_array_z_comp;
ctx.shader->uses_tex_buffers = radeon_llvm_ctx.uses_tex_buffers;
- if (r600_llvm_compile(mod, rscreen->b.family, ctx.bc, &use_kill, dump)) {
+ if (r600_llvm_compile(mod, rscreen->b.family, ctx.bc, &use_kill,
+ dump, &rctx->b.debug)) {
radeon_llvm_dispose(&radeon_llvm_ctx);
use_llvm = 0;
fprintf(stderr, "R600 LLVM backend failed to compile "
memset(&alu, 0, sizeof(struct r600_bytecode_alu));
alu.op = ctx->inst_info->op;
for (j = 0; j < inst->Instruction.NumSrcRegs; j++) {
- r600_bytecode_src(&alu.src[j], &ctx->src[j], k * 2 + ((i == 3) ? 0 : 1));;
+ r600_bytecode_src(&alu.src[j], &ctx->src[j], k * 2 + ((i == 3) ? 0 : 1));
}
alu.dst.sel = t1;
alu.dst.chan = i;
{
int chan;
int sel;
- int i;
+ unsigned i;
if (ctx->bc->chip_class == CAYMAN) {
for (i = 0; i < 3; i++) {
struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
struct r600_bytecode_alu alu;
int r;
- int i;
+ unsigned i;
/* result.x = 2^floor(src); */
if (inst->Dst[0].Register.WriteMask & 1) {
struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
struct r600_bytecode_alu alu;
int r;
- int i;
+ unsigned i;
/* result.x = floor(log2(|src|)); */
if (inst->Dst[0].Register.WriteMask & 1) {
static int tgsi_endloop(struct r600_shader_ctx *ctx)
{
- int i;
+ unsigned i;
r600_bytecode_add_cfinst(ctx->bc, CF_OP_LOOP_END);
if (!gs_ring_buffer) {
radeon_set_context_reg(cs, reg_alu_constbuf_size + buffer_index * 4,
- ALIGN_DIVUP(cb->buffer_size >> 4, 16));
+ DIV_ROUND_UP(cb->buffer_size, 256));
radeon_set_context_reg(cs, reg_alu_const_cache + buffer_index * 4, offset >> 8);
}
tmpPtr[i] = util_cpu_to_le32(((uint32_t *)ptr)[i]);
}
- u_upload_data(rctx->b.uploader, 0, size, tmpPtr, &cb->buffer_offset, &cb->buffer);
+ u_upload_data(rctx->b.uploader, 0, size, 256, tmpPtr, &cb->buffer_offset, &cb->buffer);
free(tmpPtr);
} else {
- u_upload_data(rctx->b.uploader, 0, input->buffer_size, ptr, &cb->buffer_offset, &cb->buffer);
+ u_upload_data(rctx->b.uploader, 0, input->buffer_size, 256, ptr, &cb->buffer_offset, &cb->buffer);
}
/* account it in gtt */
rctx->b.gtt += input->buffer_size;
{
if (current->pa_cl_vs_out_cntl != rctx->clip_misc_state.pa_cl_vs_out_cntl ||
current->shader.clip_dist_write != rctx->clip_misc_state.clip_dist_write ||
- current->shader.vs_position_window_space != rctx->clip_misc_state.clip_disable) {
- rctx->clip_misc_state.pa_cl_vs_out_cntl = current->pa_cl_vs_out_cntl;
- rctx->clip_misc_state.clip_dist_write = current->shader.clip_dist_write;
- rctx->clip_misc_state.clip_disable = current->shader.vs_position_window_space;
- r600_mark_atom_dirty(rctx, &rctx->clip_misc_state.atom);
+ current->shader.vs_position_window_space != rctx->clip_misc_state.clip_disable ||
+ current->shader.vs_out_viewport != rctx->clip_misc_state.vs_out_viewport) {
+ rctx->clip_misc_state.pa_cl_vs_out_cntl = current->pa_cl_vs_out_cntl;
+ rctx->clip_misc_state.clip_dist_write = current->shader.clip_dist_write;
+ rctx->clip_misc_state.clip_disable = current->shader.vs_position_window_space;
+ rctx->clip_misc_state.vs_out_viewport = current->shader.vs_out_viewport;
+ r600_mark_atom_dirty(rctx, &rctx->clip_misc_state.atom);
}
}
radeon_set_context_reg(cs, R_02881C_PA_CL_VS_OUT_CNTL,
state->pa_cl_vs_out_cntl |
(state->clip_plane_enable & state->clip_dist_write));
+ /* reuse needs to be set off if we write oViewport */
+ if (rctx->b.chip_class >= EVERGREEN)
+ radeon_set_context_reg(cs, R_028AB4_VGT_REUSE_OFF,
+ S_028AB4_REUSE_OFF(state->vs_out_viewport));
}
static void r600_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *dinfo)
}
}
- u_upload_alloc(rctx->b.uploader, start, count * 2,
+ u_upload_alloc(rctx->b.uploader, start, count * 2, 256,
&out_offset, &out_buffer, &ptr);
util_shorten_ubyte_elts_to_userptr(
if (ib.user_buffer && (R600_BIG_ENDIAN || info.indirect ||
info.instance_count > 1 ||
info.count*ib.index_size > 20)) {
- u_upload_data(rctx->b.uploader, 0, info.count * ib.index_size,
+ u_upload_data(rctx->b.uploader, 0, info.count * ib.index_size, 256,
ib.user_buffer, &ib.offset, &ib.buffer);
ib.user_buffer = NULL;
}
if NEED_RADEON_LLVM
AM_CFLAGS += \
- $(LLVM_CFLAGS)
+ $(LLVM_CFLAGS) \
+ $(LIBELF_CFLAGS)
libradeon_la_SOURCES += \
$(LLVM_C_FILES)
libradeon_la_LIBADD = \
$(CLOCK_LIB) \
$(LLVM_LIBS) \
- $(ELF_LIB)
+ $(LIBELF_LIBS)
libradeon_la_LDFLAGS = \
$(LLVM_LDFLAGS)
FREE(rbuffer);
}
+static bool
+r600_do_invalidate_resource(struct r600_common_context *rctx,
+ struct r600_resource *rbuffer)
+{
+ /* In AMD_pinned_memory, the user pointer association only gets
+ * broken when the buffer is explicitly re-allocated.
+ */
+ if (rctx->ws->buffer_is_user_ptr(rbuffer->buf))
+ return false;
+
+ /* Check if mapping this buffer would cause waiting for the GPU. */
+ if (r600_rings_is_buffer_referenced(rctx, rbuffer->buf, RADEON_USAGE_READWRITE) ||
+ !rctx->ws->buffer_wait(rbuffer->buf, 0, RADEON_USAGE_READWRITE)) {
+ rctx->invalidate_buffer(&rctx->b, &rbuffer->b.b);
+ } else {
+ util_range_set_empty(&rbuffer->valid_buffer_range);
+ }
+
+ return true;
+}
+
+void r600_invalidate_resource(struct pipe_context *ctx,
+ struct pipe_resource *resource)
+{
+ struct r600_common_context *rctx = (struct r600_common_context*)ctx;
+ struct r600_resource *rbuffer = r600_resource(resource);
+
+ (void)r600_do_invalidate_resource(rctx, rbuffer);
+}
+
static void *r600_buffer_get_transfer(struct pipe_context *ctx,
struct pipe_resource *resource,
unsigned level,
!(usage & PIPE_TRANSFER_UNSYNCHRONIZED)) {
assert(usage & PIPE_TRANSFER_WRITE);
- /* Check if mapping this buffer would cause waiting for the GPU. */
- if (r600_rings_is_buffer_referenced(rctx, rbuffer->buf, RADEON_USAGE_READWRITE) ||
- !rctx->ws->buffer_wait(rbuffer->buf, 0, RADEON_USAGE_READWRITE)) {
- rctx->invalidate_buffer(&rctx->b, &rbuffer->b.b);
+ if (r600_do_invalidate_resource(rctx, rbuffer)) {
+ /* At this point, the buffer is always idle. */
+ usage |= PIPE_TRANSFER_UNSYNCHRONIZED;
}
- /* At this point, the buffer is always idle. */
- usage |= PIPE_TRANSFER_UNSYNCHRONIZED;
}
else if ((usage & PIPE_TRANSFER_DISCARD_RANGE) &&
!(usage & PIPE_TRANSFER_UNSYNCHRONIZED) &&
struct r600_resource *staging = NULL;
u_upload_alloc(rctx->uploader, 0, box->width + (box->x % R600_MAP_BUFFER_ALIGNMENT),
- &offset, (struct pipe_resource**)&staging, (void**)&data);
+ 256, &offset, (struct pipe_resource**)&staging, (void**)&data);
if (staging) {
data += box->x % R600_MAP_BUFFER_ALIGNMENT;
for (i = 0; i < query->num_counters; ++i) {
struct r600_pc_counter *counter = &query->counters[i];
- if (counter->base == ~0)
- continue;
-
for (j = 0; j < counter->dwords; ++j) {
uint32_t value = results[counter->base + j * counter->stride];
result->batch[i].u32 += value;
/* Upload vertices. The hw rectangle has only 3 vertices,
* I guess the 4th one is derived from the first 3.
* The vertex specification should match u_blitter's vertex element state. */
- u_upload_alloc(rctx->uploader, 0, sizeof(float) * 24, &offset, &buf, (void**)&vb);
+ u_upload_alloc(rctx->uploader, 0, sizeof(float) * 24, 256, &offset, &buf, (void**)&vb);
if (!buf)
return;
return PIPE_UNKNOWN_CONTEXT_RESET;
}
+static void r600_set_debug_callback(struct pipe_context *ctx,
+ const struct pipe_debug_callback *cb)
+{
+ struct r600_common_context *rctx = (struct r600_common_context *)ctx;
+
+ if (cb)
+ rctx->debug = *cb;
+ else
+ memset(&rctx->debug, 0, sizeof(rctx->debug));
+}
+
bool r600_common_context_init(struct r600_common_context *rctx,
struct r600_common_screen *rscreen)
{
else
rctx->max_db = 4;
+ rctx->b.invalidate_resource = r600_invalidate_resource;
rctx->b.transfer_map = u_transfer_map_vtbl;
rctx->b.transfer_flush_region = u_transfer_flush_region_vtbl;
rctx->b.transfer_unmap = u_transfer_unmap_vtbl;
rctx->b.transfer_inline_write = u_default_transfer_inline_write;
rctx->b.memory_barrier = r600_memory_barrier;
rctx->b.flush = r600_flush_from_st;
+ rctx->b.set_debug_callback = r600_set_debug_callback;
if (rscreen->info.drm_major == 2 && rscreen->info.drm_minor >= 43) {
rctx->b.get_device_reset_status = r600_get_reset_status;
if (!rctx->allocator_so_filled_size)
return false;
- rctx->uploader = u_upload_create(&rctx->b, 1024 * 1024, 256,
+ rctx->uploader = u_upload_create(&rctx->b, 1024 * 1024,
PIPE_BIND_INDEX_BUFFER |
- PIPE_BIND_CONSTANT_BUFFER);
+ PIPE_BIND_CONSTANT_BUFFER, PIPE_USAGE_STREAM);
if (!rctx->uploader)
return false;
}
bool r600_can_dump_shader(struct r600_common_screen *rscreen,
- const struct tgsi_token *tokens)
+ unsigned processor)
{
- /* Compute shader don't have tgsi_tokens */
- if (!tokens)
- return (rscreen->debug_flags & DBG_CS) != 0;
-
- switch (tgsi_get_processor_type(tokens)) {
+ switch (processor) {
case TGSI_PROCESSOR_VERTEX:
return (rscreen->debug_flags & DBG_VS) != 0;
case TGSI_PROCESSOR_TESS_CTRL:
* the GPU addresses are updated. */
struct list_head texture_buffers;
+ struct pipe_debug_callback debug;
+
/* Copy one resource to another using async DMA. */
void (*dma_copy)(struct pipe_context *ctx,
struct pipe_resource *dst,
r600_buffer_from_user_memory(struct pipe_screen *screen,
const struct pipe_resource *templ,
void *user_memory);
+void
+r600_invalidate_resource(struct pipe_context *ctx,
+ struct pipe_resource *resource);
/* r600_common_pipe.c */
void r600_draw_rectangle(struct blitter_context *blitter,
void r600_common_context_cleanup(struct r600_common_context *rctx);
void r600_context_add_resource_size(struct pipe_context *ctx, struct pipe_resource *r);
bool r600_can_dump_shader(struct r600_common_screen *rscreen,
- const struct tgsi_token *tokens);
+ unsigned processor);
void r600_screen_clear_buffer(struct r600_common_screen *rscreen, struct pipe_resource *dst,
unsigned offset, unsigned size, unsigned value,
bool is_framebuffer);
rctx->b.flush(&rctx->b, &query->fence, 0);
break;
case R600_QUERY_DRAW_CALLS:
- query->begin_result = rctx->num_draw_calls;
+ query->end_result = rctx->num_draw_calls;
break;
case R600_QUERY_REQUESTED_VRAM:
case R600_QUERY_REQUESTED_GTT:
query->begin_result = 0;
break;
case R600_QUERY_NUM_COMPILATIONS:
- query->begin_result = p_atomic_read(&rctx->screen->num_compilations);
+ query->end_result = p_atomic_read(&rctx->screen->num_compilations);
break;
case R600_QUERY_NUM_SHADERS_CREATED:
- query->begin_result = p_atomic_read(&rctx->screen->num_shaders_created);
+ query->end_result = p_atomic_read(&rctx->screen->num_shaders_created);
break;
default:
unreachable("r600_query_sw_end: bad query type");
* Authors: Tom Stellard <thomas.stellard@amd.com>
*
*/
+
#include "radeon_llvm_emit.h"
#include "radeon_elf_util.h"
#include "c11/threads.h"
#include "gallivm/lp_bld_misc.h"
+#include "util/u_debug.h"
#include "util/u_memory.h"
#include "pipe/p_shader_tokens.h"
+#include "pipe/p_state.h"
#include <llvm-c/Target.h>
#include <llvm-c/TargetMachine.h>
return target;
}
+struct radeon_llvm_diagnostics {
+ struct pipe_debug_callback *debug;
+ unsigned retval;
+};
+
static void radeonDiagnosticHandler(LLVMDiagnosticInfoRef di, void *context)
{
- if (LLVMGetDiagInfoSeverity(di) == LLVMDSError) {
- unsigned int *diagnosticflag = (unsigned int *)context;
- char *diaginfo_message = LLVMGetDiagInfoDescription(di);
+ struct radeon_llvm_diagnostics *diag = (struct radeon_llvm_diagnostics *)context;
+ LLVMDiagnosticSeverity severity = LLVMGetDiagInfoSeverity(di);
+ char *description = LLVMGetDiagInfoDescription(di);
+ const char *severity_str = NULL;
+
+ switch (severity) {
+ case LLVMDSError:
+ severity_str = "error";
+ break;
+ case LLVMDSWarning:
+ severity_str = "warning";
+ break;
+ case LLVMDSRemark:
+ severity_str = "remark";
+ break;
+ case LLVMDSNote:
+ severity_str = "note";
+ break;
+ default:
+ severity_str = "unknown";
+ }
- *diagnosticflag = 1;
- fprintf(stderr,"LLVM triggered Diagnostic Handler: %s\n", diaginfo_message);
- LLVMDisposeMessage(diaginfo_message);
+ pipe_debug_message(diag->debug, SHADER_INFO,
+ "LLVM diagnostic (%s): %s", severity_str, description);
+
+ if (severity == LLVMDSError) {
+ diag->retval = 1;
+ fprintf(stderr,"LLVM triggered Diagnostic Handler: %s\n", description);
}
+
+ LLVMDisposeMessage(description);
}
/**
* @returns 0 for success, 1 for failure
*/
unsigned radeon_llvm_compile(LLVMModuleRef M, struct radeon_shader_binary *binary,
- const char *gpu_family, bool dump_ir, bool dump_asm,
- LLVMTargetMachineRef tm)
+ const char *gpu_family,
+ LLVMTargetMachineRef tm,
+ struct pipe_debug_callback *debug)
{
-
+ struct radeon_llvm_diagnostics diag;
char cpu[CPU_STRING_LEN];
char fs[FS_STRING_LEN];
char *err;
bool dispose_tm = false;
LLVMContextRef llvm_ctx;
- unsigned rval = 0;
LLVMMemoryBufferRef out_buffer;
unsigned buffer_size;
const char *buffer_data;
char triple[TRIPLE_STRING_LEN];
LLVMBool mem_err;
+ diag.debug = debug;
+ diag.retval = 0;
+
if (!tm) {
strncpy(triple, "r600--", TRIPLE_STRING_LEN);
LLVMTargetRef target = radeon_llvm_get_r600_target(triple);
}
strncpy(cpu, gpu_family, CPU_STRING_LEN);
memset(fs, 0, sizeof(fs));
- if (dump_asm)
- strncpy(fs, "+DumpCode", FS_STRING_LEN);
+ strncpy(fs, "+DumpCode", FS_STRING_LEN);
tm = LLVMCreateTargetMachine(target, triple, cpu, fs,
LLVMCodeGenLevelDefault, LLVMRelocDefault,
LLVMCodeModelDefault);
dispose_tm = true;
}
- if (dump_ir)
- LLVMDumpModule(M);
+
/* Setup Diagnostic Handler*/
llvm_ctx = LLVMGetModuleContext(M);
- LLVMContextSetDiagnosticHandler(llvm_ctx, radeonDiagnosticHandler, &rval);
- rval = 0;
+ LLVMContextSetDiagnosticHandler(llvm_ctx, radeonDiagnosticHandler, &diag);
/* Compile IR*/
mem_err = LLVMTargetMachineEmitToMemoryBuffer(tm, M, LLVMObjectFile, &err,
/* Process Errors/Warnings */
if (mem_err) {
fprintf(stderr, "%s: %s", __FUNCTION__, err);
+ pipe_debug_message(debug, SHADER_INFO,
+ "LLVM emit error: %s", err);
FREE(err);
- LLVMDisposeTargetMachine(tm);
- return 1;
- }
-
- if (0 != rval) {
- fprintf(stderr, "%s: Processing Diag Flag\n", __FUNCTION__);
+ diag.retval = 1;
+ goto out;
}
/* Extract Shader Code*/
/* Clean up */
LLVMDisposeMemoryBuffer(out_buffer);
+out:
if (dispose_tm) {
LLVMDisposeTargetMachine(tm);
}
- return rval;
+ if (diag.retval != 0)
+ pipe_debug_message(debug, SHADER_INFO, "LLVM compile failed");
+ return diag.retval;
}
#include <llvm-c/TargetMachine.h>
#include <stdbool.h>
+struct pipe_debug_callback;
struct radeon_shader_binary;
void radeon_llvm_shader_type(LLVMValueRef F, unsigned type);
LLVMTargetRef radeon_llvm_get_r600_target(const char *triple);
unsigned radeon_llvm_compile(LLVMModuleRef M, struct radeon_shader_binary *binary,
- const char *gpu_family, bool dump_ir, bool dump_asm,
- LLVMTargetMachineRef tm);
+ const char *gpu_family,
+ LLVMTargetMachineRef tm,
+ struct pipe_debug_callback *debug);
#endif /* RADEON_LLVM_EMIT_H */
void *pointer, unsigned size);
/**
+ * Whether the buffer was created from a user pointer.
+ *
+ * \param buf A winsys buffer object
+ * \return whether \p buf was created via buffer_from_ptr
+ */
+ bool (*buffer_is_user_ptr)(struct pb_buffer *buf);
+
+ /**
* Get a winsys handle from a winsys buffer. The internal structure
* of the handle is platform-specific and only a winsys should access it.
*
(tile_split << 11) | (mt << 8) | (array_mode << 3) |
lbpe;
cs->buf[cs->cdw++] = y << 16; /* | x */
- cs->buf[cs->cdw++] = 0; /* z */;
+ cs->buf[cs->cdw++] = 0; /* z */
cs->buf[cs->cdw++] = addr & 0xfffffffc;
cs->buf[cs->cdw++] = addr >> 32;
cs->buf[cs->cdw++] = (pitch / bpe) - 1;
program->shader.binary.global_symbol_offsets[i];
unsigned scratch_bytes_needed;
- si_shader_binary_read_config(sctx->screen,
- &program->shader, offset);
- scratch_bytes_needed = program->shader.scratch_bytes_per_wave;
+ si_shader_binary_read_config(&program->shader.binary,
+ &program->shader.config, offset);
+ scratch_bytes_needed = program->shader.config.scratch_bytes_per_wave;
scratch_bytes = MAX2(scratch_bytes, scratch_bytes_needed);
}
* to the maximum bytes needed, so it can compute the stride
* correctly.
*/
- program->shader.scratch_bytes_per_wave = scratch_bytes;
+ program->shader.config.scratch_bytes_per_wave = scratch_bytes;
/* Patch the shader with the scratch buffer address. */
si_shader_apply_scratch_relocs(sctx,
for (i = 0; i < program->num_kernels; i++) {
LLVMModuleRef mod = radeon_llvm_get_kernel_module(program->llvm_ctx, i,
code, header->num_bytes);
- si_compile_llvm(sctx->screen, &program->kernels[i], sctx->tm,
- mod);
+ si_compile_llvm(sctx->screen, &program->kernels[i].binary,
+ &program->kernels[i].config, sctx->tm,
+ mod, &sctx->b.debug, TGSI_PROCESSOR_COMPUTE);
+ si_shader_dump(sctx->screen, &program->kernels[i],
+ &sctx->b.debug, TGSI_PROCESSOR_COMPUTE);
+ si_shader_binary_upload(sctx->screen, &program->kernels[i]);
LLVMDisposeModule(mod);
}
}
* the shader code to the GPU.
*/
init_scratch_buffer(sctx, program);
- si_shader_binary_read(sctx->screen, &program->shader);
+ si_shader_binary_read_config(&program->shader.binary,
+ &program->shader.config, 0);
+ si_shader_dump(sctx->screen, &program->shader, &sctx->b.debug,
+ TGSI_PROCESSOR_COMPUTE);
+ si_shader_binary_upload(sctx->screen, &program->shader);
#endif
program->input_buffer = si_resource_create_custom(sctx->b.b.screen,
#if HAVE_LLVM >= 0x0306
/* Read the config information */
- si_shader_binary_read_config(sctx->screen, shader, pc);
+ si_shader_binary_read_config(&shader->binary, &shader->config, pc);
#endif
/* Upload the kernel arguments */
memcpy(kernel_args + (num_work_size_bytes / 4), input, program->input_size);
- if (shader->scratch_bytes_per_wave > 0) {
+ if (shader->config.scratch_bytes_per_wave > 0) {
COMPUTE_DBG(sctx->screen, "Waves: %u; Scratch per wave: %u bytes; "
"Total Scratch: %u bytes\n", num_waves_for_scratch,
- shader->scratch_bytes_per_wave,
- shader->scratch_bytes_per_wave *
+ shader->config.scratch_bytes_per_wave,
+ shader->config.scratch_bytes_per_wave *
num_waves_for_scratch);
radeon_add_to_buffer_list(&sctx->b, &sctx->b.gfx,
si_pm4_set_reg(pm4, R_00B900_COMPUTE_USER_DATA_0 + 8, scratch_buffer_va);
si_pm4_set_reg(pm4, R_00B900_COMPUTE_USER_DATA_0 + 12,
S_008F04_BASE_ADDRESS_HI(scratch_buffer_va >> 32)
- | S_008F04_STRIDE(shader->scratch_bytes_per_wave / 64));
+ | S_008F04_STRIDE(shader->config.scratch_bytes_per_wave / 64));
si_pm4_set_reg(pm4, R_00B810_COMPUTE_START_X, 0);
si_pm4_set_reg(pm4, R_00B814_COMPUTE_START_Y, 0);
si_pm4_set_reg(pm4, R_00B830_COMPUTE_PGM_LO, shader_va >> 8);
si_pm4_set_reg(pm4, R_00B834_COMPUTE_PGM_HI, shader_va >> 40);
- si_pm4_set_reg(pm4, R_00B848_COMPUTE_PGM_RSRC1, shader->rsrc1);
+ si_pm4_set_reg(pm4, R_00B848_COMPUTE_PGM_RSRC1, shader->config.rsrc1);
- lds_blocks = shader->lds_size;
+ lds_blocks = shader->config.lds_size;
/* XXX: We are over allocating LDS. For SI, the shader reports LDS in
* blocks of 256 bytes, so if there are 4 bytes lds allocated in
* the shader and 4 bytes allocated by the state tracker, then
assert(lds_blocks <= 0xFF);
- shader->rsrc2 &= C_00B84C_LDS_SIZE;
- shader->rsrc2 |= S_00B84C_LDS_SIZE(lds_blocks);
+ shader->config.rsrc2 &= C_00B84C_LDS_SIZE;
+ shader->config.rsrc2 |= S_00B84C_LDS_SIZE(lds_blocks);
- si_pm4_set_reg(pm4, R_00B84C_COMPUTE_PGM_RSRC2, shader->rsrc2);
+ si_pm4_set_reg(pm4, R_00B84C_COMPUTE_PGM_RSRC2, shader->config.rsrc2);
si_pm4_set_reg(pm4, R_00B854_COMPUTE_RESOURCE_LIMITS, 0);
si_pm4_set_reg(pm4, R_00B858_COMPUTE_STATIC_THREAD_MGMT_SE0,
* COMPUTE_PGM_RSRC2.SCRATCH_EN is enabled.
*/
S_00B860_WAVES(num_waves_for_scratch)
- | S_00B860_WAVESIZE(shader->scratch_bytes_per_wave >> 10))
+ | S_00B860_WAVESIZE(shader->config.scratch_bytes_per_wave >> 10))
;
si_pm4_cmd_begin(pm4, PKT3_DISPATCH_DIRECT);
#include "si_shader.h"
#include "sid.h"
#include "sid_tables.h"
+#include "radeon/radeon_elf_util.h"
#include "ddebug/dd_util.h"
+#include "util/u_memory.h"
+DEBUG_GET_ONCE_OPTION(replace_shaders, "RADEON_REPLACE_SHADERS", NULL)
static void si_dump_shader(struct si_shader_ctx_state *state, const char *name,
FILE *f)
fprintf(f, "%s\n\n", state->current->binary.disasm_string);
}
+/**
+ * Shader compiles can be overridden with arbitrary ELF objects by setting
+ * the environment variable RADEON_REPLACE_SHADERS=num1:filename1[;num2:filename2]
+ */
+bool si_replace_shader(unsigned num, struct radeon_shader_binary *binary)
+{
+ const char *p = debug_get_option_replace_shaders();
+ const char *semicolon;
+ char *copy = NULL;
+ FILE *f;
+ long filesize, nread;
+ char *buf = NULL;
+ bool replaced = false;
+
+ if (!p)
+ return false;
+
+ while (*p) {
+ unsigned long i;
+ char *endp;
+ i = strtoul(p, &endp, 0);
+
+ p = endp;
+ if (*p != ':') {
+ fprintf(stderr, "RADEON_REPLACE_SHADERS formatted badly.\n");
+ exit(1);
+ }
+ ++p;
+
+ if (i == num)
+ break;
+
+ p = strchr(p, ';');
+ if (!p)
+ return false;
+ ++p;
+ }
+ if (!*p)
+ return false;
+
+ semicolon = strchr(p, ';');
+ if (semicolon) {
+ p = copy = strndup(p, semicolon - p);
+ if (!copy) {
+ fprintf(stderr, "out of memory\n");
+ return false;
+ }
+ }
+
+ fprintf(stderr, "radeonsi: replace shader %u by %s\n", num, p);
+
+ f = fopen(p, "r");
+ if (!f) {
+ perror("radeonsi: failed to open file");
+ goto out_free;
+ }
+
+ if (fseek(f, 0, SEEK_END) != 0)
+ goto file_error;
+
+ filesize = ftell(f);
+ if (filesize < 0)
+ goto file_error;
+
+ if (fseek(f, 0, SEEK_SET) != 0)
+ goto file_error;
+
+ buf = MALLOC(filesize);
+ if (!buf) {
+ fprintf(stderr, "out of memory\n");
+ goto out_close;
+ }
+
+ nread = fread(buf, 1, filesize, f);
+ if (nread != filesize)
+ goto file_error;
+
+ radeon_elf_read(buf, filesize, binary);
+ replaced = true;
+
+out_close:
+ fclose(f);
+out_free:
+ FREE(buf);
+ free(copy);
+ return replaced;
+
+file_error:
+ perror("radeonsi: reading shader");
+ goto out_close;
+}
+
/* Parsed IBs are difficult to read without colors. Use "less -R file" to
* read them, or use "aha -b -f file" to convert them to html.
*/
if (!desc->list_dirty)
return true;
- u_upload_alloc(sctx->b.uploader, 0, list_size,
+ u_upload_alloc(sctx->b.uploader, 0, list_size, 256,
&desc->buffer_offset,
(struct pipe_resource**)&desc->buffer, &ptr);
if (!desc->buffer)
* directly through a staging buffer and don't go through
* the fine-grained upload path.
*/
- u_upload_alloc(sctx->b.uploader, 0, count * 16, &desc->buffer_offset,
+ u_upload_alloc(sctx->b.uploader, 0, count * 16, 256, &desc->buffer_offset,
(struct pipe_resource**)&desc->buffer, (void**)&ptr);
if (!desc->buffer)
return false;
{
void *tmp;
- u_upload_alloc(sctx->b.uploader, 0, size, const_offset,
+ u_upload_alloc(sctx->b.uploader, 0, size, 256, const_offset,
(struct pipe_resource**)rbuffer, &tmp);
if (rbuffer)
util_memcpy_cpu_to_le32(tmp, ptr, size);
for (i = 0; i < SI_NUM_SHADERS; i++) {
si_init_buffer_resources(&sctx->const_buffers[i],
- SI_NUM_CONST_BUFFERS, SI_SGPR_CONST,
+ SI_NUM_CONST_BUFFERS, SI_SGPR_CONST_BUFFERS,
RADEON_USAGE_READ, RADEON_PRIO_CONST_BUFFER);
si_init_buffer_resources(&sctx->rw_buffers[i],
SI_NUM_RW_BUFFERS, SI_SGPR_RW_BUFFERS,
RADEON_USAGE_READWRITE, RADEON_PRIO_RINGS_STREAMOUT);
si_init_descriptors(&sctx->samplers[i].views.desc,
- SI_SGPR_RESOURCE, 8, SI_NUM_SAMPLER_VIEWS);
+ SI_SGPR_SAMPLER_VIEWS, 8, SI_NUM_SAMPLER_VIEWS);
si_init_descriptors(&sctx->samplers[i].states.desc,
- SI_SGPR_SAMPLER, 4, SI_NUM_SAMPLER_STATES);
+ SI_SGPR_SAMPLER_STATES, 4, SI_NUM_SAMPLER_STATES);
}
- si_init_descriptors(&sctx->vertex_buffers, SI_SGPR_VERTEX_BUFFER,
+ si_init_descriptors(&sctx->vertex_buffers, SI_SGPR_VERTEX_BUFFERS,
4, SI_NUM_VERTEX_BUFFERS);
/* Set pipe_context functions. */
dw = count + regs->num_prelude;
if (count >= regs->num_multi)
- count += regs->num_multi;
+ dw += regs->num_multi;
radeon_set_uconfig_reg_seq(cs, regs->select0, dw);
for (idx = 0; idx < regs->num_prelude; ++idx)
radeon_emit(cs, 0);
case PIPE_CAP_TGSI_TXQS:
case PIPE_CAP_FORCE_PERSAMPLE_INTERP:
case PIPE_CAP_COPY_BETWEEN_COMPRESSED_AND_PLAIN_FORMATS:
+ case PIPE_CAP_TGSI_FS_POSITION_IS_SYSVAL:
+ case PIPE_CAP_TGSI_FS_FACE_IS_INTEGER_SYSVAL:
+ case PIPE_CAP_INVALIDATE_BUFFER:
return 1;
case PIPE_CAP_RESOURCE_FROM_USER_MEMORY:
case PIPE_CAP_TEXTURE_GATHER_OFFSETS:
case PIPE_CAP_VERTEXID_NOBASE:
case PIPE_CAP_CLEAR_TEXTURE:
+ case PIPE_CAP_DRAW_PARAMETERS:
+ case PIPE_CAP_TGSI_PACK_HALF_FLOAT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT_PARAMS:
+ case PIPE_CAP_SHADER_BUFFER_OFFSET_ALIGNMENT:
+ case PIPE_CAP_GENERATE_MIPMAP:
return 0;
case PIPE_CAP_MAX_SHADER_PATCH_VARYINGS:
return 1;
case PIPE_SHADER_CAP_MAX_UNROLL_ITERATIONS_HINT:
return 32;
+ case PIPE_SHADER_CAP_MAX_SHADER_BUFFERS:
+ return 0;
}
return 0;
}
/* si_debug.c */
void si_init_debug_functions(struct si_context *sctx);
void si_check_vm_faults(struct si_context *sctx);
+bool si_replace_shader(unsigned num, struct radeon_shader_binary *binary);
/* si_dma.c */
void si_dma_copy(struct pipe_context *ctx,
int param_es2gs_offset;
LLVMTargetMachineRef tm;
LLVMValueRef const_md;
- LLVMValueRef const_resource[SI_NUM_CONST_BUFFERS];
+ LLVMValueRef const_buffers[SI_NUM_CONST_BUFFERS];
LLVMValueRef lds;
LLVMValueRef *constants[SI_NUM_CONST_BUFFERS];
- LLVMValueRef resources[SI_NUM_SAMPLER_VIEWS];
- LLVMValueRef samplers[SI_NUM_SAMPLER_STATES];
+ LLVMValueRef sampler_views[SI_NUM_SAMPLER_VIEWS];
+ LLVMValueRef sampler_states[SI_NUM_SAMPLER_STATES];
LLVMValueRef so_buffers[4];
LLVMValueRef esgs_ring;
LLVMValueRef gsvs_ring[4];
LLVMValueRef input;
/* Load the T list */
- t_list_ptr = LLVMGetParam(si_shader_ctx->radeon_bld.main_fn, SI_PARAM_VERTEX_BUFFER);
+ t_list_ptr = LLVMGetParam(si_shader_ctx->radeon_bld.main_fn, SI_PARAM_VERTEX_BUFFERS);
t_offset = lp_build_const_int32(gallivm, input_index);
unsigned chan;
- if (decl->Semantic.Name == TGSI_SEMANTIC_POSITION) {
- for (chan = 0; chan < TGSI_NUM_CHANNELS; chan++) {
- unsigned soa_index =
- radeon_llvm_reg_index_soa(input_index, chan);
- radeon_bld->inputs[soa_index] =
- LLVMGetParam(main_fn, SI_PARAM_POS_X_FLOAT + chan);
-
- if (chan == 3)
- /* RCP for fragcoord.w */
- radeon_bld->inputs[soa_index] =
- LLVMBuildFDiv(gallivm->builder,
- lp_build_const_float(gallivm, 1.0f),
- radeon_bld->inputs[soa_index],
- "");
- }
- return;
- }
-
- if (decl->Semantic.Name == TGSI_SEMANTIC_FACE) {
- radeon_bld->inputs[radeon_llvm_reg_index_soa(input_index, 0)] =
- LLVMGetParam(main_fn, SI_PARAM_FRONT_FACE);
- radeon_bld->inputs[radeon_llvm_reg_index_soa(input_index, 1)] =
- radeon_bld->inputs[radeon_llvm_reg_index_soa(input_index, 2)] =
- lp_build_const_float(gallivm, 0.0f);
- radeon_bld->inputs[radeon_llvm_reg_index_soa(input_index, 3)] =
- lp_build_const_float(gallivm, 1.0f);
-
- return;
- }
-
shader->ps_input_param_offset[input_index] = shader->nparam++;
attr_number = lp_build_const_int32(gallivm,
shader->ps_input_param_offset[input_index]);
face = LLVMGetParam(main_fn, SI_PARAM_FRONT_FACE);
- is_face_positive = LLVMBuildFCmp(gallivm->builder,
- LLVMRealOGT, face,
- lp_build_const_float(gallivm, 0.0f),
- "");
+ is_face_positive = LLVMBuildICmp(gallivm->builder, LLVMIntNE,
+ face, uint->zero, "");
args[2] = params;
args[3] = interp_param;
struct lp_build_context *uint_bld = &radeon_bld->soa.bld_base.uint_bld;
struct gallivm_state *gallivm = &radeon_bld->gallivm;
LLVMBuilderRef builder = gallivm->builder;
- LLVMValueRef desc = LLVMGetParam(si_shader_ctx->radeon_bld.main_fn, SI_PARAM_CONST);
+ LLVMValueRef desc = LLVMGetParam(si_shader_ctx->radeon_bld.main_fn, SI_PARAM_CONST_BUFFERS);
LLVMValueRef buf_index = lp_build_const_int32(gallivm, SI_DRIVER_STATE_CONST_BUF);
LLVMValueRef resource = build_indexed_load_const(si_shader_ctx, desc, buf_index);
assert(!"INVOCATIONID not implemented");
break;
+ case TGSI_SEMANTIC_POSITION:
+ {
+ LLVMValueRef pos[4] = {
+ LLVMGetParam(radeon_bld->main_fn, SI_PARAM_POS_X_FLOAT),
+ LLVMGetParam(radeon_bld->main_fn, SI_PARAM_POS_Y_FLOAT),
+ LLVMGetParam(radeon_bld->main_fn, SI_PARAM_POS_Z_FLOAT),
+ lp_build_emit_llvm_unary(&radeon_bld->soa.bld_base, TGSI_OPCODE_RCP,
+ LLVMGetParam(radeon_bld->main_fn,
+ SI_PARAM_POS_W_FLOAT)),
+ };
+ value = lp_build_gather_values(gallivm, pos, 4);
+ break;
+ }
+
+ case TGSI_SEMANTIC_FACE:
+ value = LLVMGetParam(radeon_bld->main_fn, SI_PARAM_FRONT_FACE);
+ break;
+
case TGSI_SEMANTIC_SAMPLEID:
value = get_sample_id(radeon_bld);
break;
}
if (reg->Register.Dimension && reg->Dimension.Indirect) {
- LLVMValueRef ptr = LLVMGetParam(si_shader_ctx->radeon_bld.main_fn, SI_PARAM_CONST);
+ LLVMValueRef ptr = LLVMGetParam(si_shader_ctx->radeon_bld.main_fn, SI_PARAM_CONST_BUFFERS);
LLVMValueRef index;
index = get_indirect_index(si_shader_ctx, ®->DimIndirect,
reg->Dimension.Index);
bufp = build_indexed_load_const(si_shader_ctx, ptr, index);
} else
- bufp = si_shader_ctx->const_resource[buf];
+ bufp = si_shader_ctx->const_buffers[buf];
addr = si_shader_ctx->radeon_bld.soa.addr[ireg->Index][ireg->Swizzle];
addr = LLVMBuildLoad(base->gallivm->builder, addr, "load addr reg");
addr2 = lp_build_add(&bld_base->uint_bld, addr2,
lp_build_const_int32(base->gallivm, idx * 4));
- result2 = buffer_load_const(base->gallivm->builder, si_shader_ctx->const_resource[buf],
+ result2 = buffer_load_const(base->gallivm->builder, si_shader_ctx->const_buffers[buf],
addr2, bld_base->base.elem_type);
result = radeon_llvm_emit_fetch_double(bld_base,
if (si_shader_ctx->type == TGSI_PROCESSOR_FRAGMENT) {
int cbuf = target - V_008DFC_SQ_EXP_MRT;
- if (cbuf >= 0 && cbuf < 8) {
+ if (cbuf >= 0 && cbuf < 8)
compressed = (si_shader_ctx->shader->key.ps.export_16bpc >> cbuf) & 0x1;
-
- if (compressed)
- si_shader_ctx->shader->spi_shader_col_format |=
- V_028714_SPI_SHADER_FP16_ABGR << (4 * cbuf);
- else
- si_shader_ctx->shader->spi_shader_col_format |=
- V_028714_SPI_SHADER_32_ABGR << (4 * cbuf);
-
- si_shader_ctx->shader->cb_shader_mask |= 0xf << (4 * cbuf);
- }
}
/* Set COMPR flag */
LLVMInt32TypeInContext(base->gallivm->context),
pack_args, 2,
LLVMReadNoneAttribute | LLVMNoUnwindAttribute);
- args[chan + 7] = args[chan + 5] =
+ args[chan + 5] =
LLVMBuildBitCast(base->gallivm->builder,
packed,
LLVMFloatTypeInContext(base->gallivm->context),
"");
+ args[chan + 7] = base->undef;
}
} else
memcpy(&args[5], values, sizeof(values[0]) * 4);
}
-/* Load from output pointers and initialize arguments for the shader export intrinsic */
-static void si_llvm_init_export_args_load(struct lp_build_tgsi_context *bld_base,
- LLVMValueRef *out_ptr,
- unsigned target,
- LLVMValueRef *args)
-{
- struct gallivm_state *gallivm = bld_base->base.gallivm;
- LLVMValueRef values[4];
- int i;
-
- for (i = 0; i < 4; i++)
- values[i] = LLVMBuildLoad(gallivm->builder, out_ptr[i], "");
-
- si_llvm_init_export_args(bld_base, values, target, args);
-}
-
static void si_alpha_test(struct lp_build_tgsi_context *bld_base,
- LLVMValueRef alpha_ptr)
+ LLVMValueRef alpha)
{
struct si_shader_context *si_shader_ctx = si_shader_context(bld_base);
struct gallivm_state *gallivm = bld_base->base.gallivm;
LLVMValueRef alpha_pass =
lp_build_cmp(&bld_base->base,
si_shader_ctx->shader->key.ps.alpha_func,
- LLVMBuildLoad(gallivm->builder, alpha_ptr, ""),
- alpha_ref);
+ alpha, alpha_ref);
LLVMValueRef arg =
lp_build_select(&bld_base->base,
alpha_pass,
LLVMVoidTypeInContext(gallivm->context),
NULL, 0, 0);
}
-
- si_shader_ctx->shader->db_shader_control |= S_02880C_KILL_ENABLE(1);
}
-static void si_scale_alpha_by_sample_mask(struct lp_build_tgsi_context *bld_base,
- LLVMValueRef alpha_ptr)
+static LLVMValueRef si_scale_alpha_by_sample_mask(struct lp_build_tgsi_context *bld_base,
+ LLVMValueRef alpha)
{
struct si_shader_context *si_shader_ctx = si_shader_context(bld_base);
struct gallivm_state *gallivm = bld_base->base.gallivm;
- LLVMValueRef coverage, alpha;
+ LLVMValueRef coverage;
/* alpha = alpha * popcount(coverage) / SI_NUM_SMOOTH_AA_SAMPLES */
coverage = LLVMGetParam(si_shader_ctx->radeon_bld.main_fn,
lp_build_const_float(gallivm,
1.0 / SI_NUM_SMOOTH_AA_SAMPLES), "");
- alpha = LLVMBuildLoad(gallivm->builder, alpha_ptr, "");
- alpha = LLVMBuildFMul(gallivm->builder, alpha, coverage, "");
- LLVMBuildStore(gallivm->builder, alpha, alpha_ptr);
+ return LLVMBuildFMul(gallivm->builder, alpha, coverage, "");
}
static void si_llvm_emit_clipvertex(struct lp_build_tgsi_context * bld_base,
unsigned chan;
unsigned const_chan;
LLVMValueRef base_elt;
- LLVMValueRef ptr = LLVMGetParam(si_shader_ctx->radeon_bld.main_fn, SI_PARAM_CONST);
+ LLVMValueRef ptr = LLVMGetParam(si_shader_ctx->radeon_bld.main_fn, SI_PARAM_CONST_BUFFERS);
LLVMValueRef constbuf_index = lp_build_const_int32(base->gallivm, SI_DRIVER_STATE_CONST_BUF);
LLVMValueRef const_resource = build_indexed_load_const(si_shader_ctx, ptr, constbuf_index);
FREE(outputs);
}
-static void si_llvm_emit_fs_epilogue(struct lp_build_tgsi_context * bld_base)
+static void si_export_mrt_z(struct lp_build_tgsi_context *bld_base,
+ LLVMValueRef depth, LLVMValueRef stencil,
+ LLVMValueRef samplemask)
{
- struct si_shader_context * si_shader_ctx = si_shader_context(bld_base);
- struct si_shader * shader = si_shader_ctx->shader;
- struct lp_build_context * base = &bld_base->base;
- struct lp_build_context * uint = &bld_base->uint_bld;
- struct tgsi_shader_info *info = &shader->selector->info;
- LLVMBuilderRef builder = base->gallivm->builder;
+ struct si_screen *sscreen = si_shader_context(bld_base)->screen;
+ struct lp_build_context *base = &bld_base->base;
+ struct lp_build_context *uint = &bld_base->uint_bld;
LLVMValueRef args[9];
- LLVMValueRef last_args[9] = { 0 };
- int depth_index = -1, stencil_index = -1, samplemask_index = -1;
- int i;
+ unsigned mask = 0;
- for (i = 0; i < info->num_outputs; i++) {
- unsigned semantic_name = info->output_semantic_name[i];
- unsigned semantic_index = info->output_semantic_index[i];
- unsigned target;
- LLVMValueRef alpha_ptr;
+ assert(depth || stencil || samplemask);
- /* Select the correct target */
- switch (semantic_name) {
- case TGSI_SEMANTIC_POSITION:
- depth_index = i;
- continue;
- case TGSI_SEMANTIC_STENCIL:
- stencil_index = i;
- continue;
- case TGSI_SEMANTIC_SAMPLEMASK:
- samplemask_index = i;
- continue;
- case TGSI_SEMANTIC_COLOR:
- target = V_008DFC_SQ_EXP_MRT + semantic_index;
- alpha_ptr = si_shader_ctx->radeon_bld.soa.outputs[i][3];
+ args[1] = uint->one; /* whether the EXEC mask is valid */
+ args[2] = uint->one; /* DONE bit */
- if (si_shader_ctx->shader->key.ps.clamp_color) {
- for (int j = 0; j < 4; j++) {
- LLVMValueRef ptr = si_shader_ctx->radeon_bld.soa.outputs[i][j];
- LLVMValueRef result = LLVMBuildLoad(builder, ptr, "");
+ /* Specify the target we are exporting */
+ args[3] = lp_build_const_int32(base->gallivm, V_008DFC_SQ_EXP_MRTZ);
- result = radeon_llvm_saturate(bld_base, result);
- LLVMBuildStore(builder, result, ptr);
- }
- }
+ args[4] = uint->zero; /* COMP flag */
+ args[5] = base->undef; /* R, depth */
+ args[6] = base->undef; /* G, stencil test value[0:7], stencil op value[8:15] */
+ args[7] = base->undef; /* B, sample mask */
+ args[8] = base->undef; /* A, alpha to mask */
- if (si_shader_ctx->shader->key.ps.alpha_to_one)
- LLVMBuildStore(base->gallivm->builder,
- base->one, alpha_ptr);
+ if (depth) {
+ args[5] = depth;
+ mask |= 0x1;
+ }
- if (semantic_index == 0 &&
- si_shader_ctx->shader->key.ps.alpha_func != PIPE_FUNC_ALWAYS)
- si_alpha_test(bld_base, alpha_ptr);
+ if (stencil) {
+ args[6] = stencil;
+ mask |= 0x2;
+ }
- if (si_shader_ctx->shader->key.ps.poly_line_smoothing)
- si_scale_alpha_by_sample_mask(bld_base, alpha_ptr);
+ if (samplemask) {
+ args[7] = samplemask;
+ mask |= 0x4;
+ }
- break;
- default:
- target = 0;
- fprintf(stderr,
- "Warning: SI unhandled fs output type:%d\n",
- semantic_name);
- }
+ /* SI (except OLAND) has a bug that it only looks
+ * at the X writemask component. */
+ if (sscreen->b.chip_class == SI &&
+ sscreen->b.family != CHIP_OLAND)
+ mask |= 0x1;
- si_llvm_init_export_args_load(bld_base,
- si_shader_ctx->radeon_bld.soa.outputs[i],
- target, args);
-
- if (semantic_name == TGSI_SEMANTIC_COLOR) {
- /* If there is an export instruction waiting to be emitted, do so now. */
- if (last_args[0]) {
- lp_build_intrinsic(base->gallivm->builder,
- "llvm.SI.export",
- LLVMVoidTypeInContext(base->gallivm->context),
- last_args, 9, 0);
- }
+ /* Specify which components to enable */
+ args[0] = lp_build_const_int32(base->gallivm, mask);
- /* This instruction will be emitted at the end of the shader. */
- memcpy(last_args, args, sizeof(args));
-
- /* Handle FS_COLOR0_WRITES_ALL_CBUFS. */
- if (shader->selector->info.properties[TGSI_PROPERTY_FS_COLOR0_WRITES_ALL_CBUFS] &&
- semantic_index == 0 &&
- si_shader_ctx->shader->key.ps.last_cbuf > 0) {
- for (int c = 1; c <= si_shader_ctx->shader->key.ps.last_cbuf; c++) {
- si_llvm_init_export_args_load(bld_base,
- si_shader_ctx->radeon_bld.soa.outputs[i],
- V_008DFC_SQ_EXP_MRT + c, args);
- lp_build_intrinsic(base->gallivm->builder,
- "llvm.SI.export",
- LLVMVoidTypeInContext(base->gallivm->context),
- args, 9, 0);
- }
- }
- } else {
- lp_build_intrinsic(base->gallivm->builder,
- "llvm.SI.export",
+ lp_build_intrinsic(base->gallivm->builder, "llvm.SI.export",
+ LLVMVoidTypeInContext(base->gallivm->context),
+ args, 9, 0);
+}
+
+static void si_export_mrt_color(struct lp_build_tgsi_context *bld_base,
+ LLVMValueRef *color, unsigned index,
+ bool is_last)
+{
+ struct si_shader_context *si_shader_ctx = si_shader_context(bld_base);
+ struct lp_build_context *base = &bld_base->base;
+ LLVMValueRef args[9];
+ int i;
+
+ /* Clamp color */
+ if (si_shader_ctx->shader->key.ps.clamp_color)
+ for (i = 0; i < 4; i++)
+ color[i] = radeon_llvm_saturate(bld_base, color[i]);
+
+ /* Alpha to one */
+ if (si_shader_ctx->shader->key.ps.alpha_to_one)
+ color[3] = base->one;
+
+ /* Alpha test */
+ if (index == 0 &&
+ si_shader_ctx->shader->key.ps.alpha_func != PIPE_FUNC_ALWAYS)
+ si_alpha_test(bld_base, color[3]);
+
+ /* Line & polygon smoothing */
+ if (si_shader_ctx->shader->key.ps.poly_line_smoothing)
+ color[3] = si_scale_alpha_by_sample_mask(bld_base, color[3]);
+
+ /* If last_cbuf > 0, FS_COLOR0_WRITES_ALL_CBUFS is true. */
+ if (index == 0 &&
+ si_shader_ctx->shader->key.ps.last_cbuf > 0) {
+ for (int c = 1; c <= si_shader_ctx->shader->key.ps.last_cbuf; c++) {
+ si_llvm_init_export_args(bld_base, color,
+ V_008DFC_SQ_EXP_MRT + c, args);
+ lp_build_intrinsic(base->gallivm->builder, "llvm.SI.export",
LLVMVoidTypeInContext(base->gallivm->context),
args, 9, 0);
}
}
- if (depth_index >= 0 || stencil_index >= 0 || samplemask_index >= 0) {
- LLVMValueRef out_ptr;
- unsigned mask = 0;
-
- /* Specify the target we are exporting */
- args[3] = lp_build_const_int32(base->gallivm, V_008DFC_SQ_EXP_MRTZ);
-
- args[5] = base->zero; /* R, depth */
- args[6] = base->zero; /* G, stencil test value[0:7], stencil op value[8:15] */
- args[7] = base->zero; /* B, sample mask */
- args[8] = base->zero; /* A, alpha to mask */
-
- if (depth_index >= 0) {
- out_ptr = si_shader_ctx->radeon_bld.soa.outputs[depth_index][2];
- args[5] = LLVMBuildLoad(base->gallivm->builder, out_ptr, "");
- mask |= 0x1;
- si_shader_ctx->shader->db_shader_control |= S_02880C_Z_EXPORT_ENABLE(1);
- }
-
- if (stencil_index >= 0) {
- out_ptr = si_shader_ctx->radeon_bld.soa.outputs[stencil_index][1];
- args[6] = LLVMBuildLoad(base->gallivm->builder, out_ptr, "");
- mask |= 0x2;
- si_shader_ctx->shader->db_shader_control |=
- S_02880C_STENCIL_TEST_VAL_EXPORT_ENABLE(1);
- }
-
- if (samplemask_index >= 0) {
- out_ptr = si_shader_ctx->radeon_bld.soa.outputs[samplemask_index][0];
- args[7] = LLVMBuildLoad(base->gallivm->builder, out_ptr, "");
- mask |= 0x4;
- si_shader_ctx->shader->db_shader_control |= S_02880C_MASK_EXPORT_ENABLE(1);
- }
-
- /* SI (except OLAND) has a bug that it only looks
- * at the X writemask component. */
- if (si_shader_ctx->screen->b.chip_class == SI &&
- si_shader_ctx->screen->b.family != CHIP_OLAND)
- mask |= 0x1;
+ /* Export */
+ si_llvm_init_export_args(bld_base, color, V_008DFC_SQ_EXP_MRT + index,
+ args);
+ if (is_last) {
+ args[1] = bld_base->uint_bld.one; /* whether the EXEC mask is valid */
+ args[2] = bld_base->uint_bld.one; /* DONE bit */
+ }
+ lp_build_intrinsic(base->gallivm->builder, "llvm.SI.export",
+ LLVMVoidTypeInContext(base->gallivm->context),
+ args, 9, 0);
+}
- if (samplemask_index >= 0)
- si_shader_ctx->shader->spi_shader_z_format = V_028710_SPI_SHADER_32_ABGR;
- else if (stencil_index >= 0)
- si_shader_ctx->shader->spi_shader_z_format = V_028710_SPI_SHADER_32_GR;
- else
- si_shader_ctx->shader->spi_shader_z_format = V_028710_SPI_SHADER_32_R;
+static void si_export_null(struct lp_build_tgsi_context *bld_base)
+{
+ struct lp_build_context *base = &bld_base->base;
+ struct lp_build_context *uint = &bld_base->uint_bld;
+ LLVMValueRef args[9];
- /* Specify which components to enable */
- args[0] = lp_build_const_int32(base->gallivm, mask);
+ args[0] = lp_build_const_int32(base->gallivm, 0x0); /* enabled channels */
+ args[1] = uint->one; /* whether the EXEC mask is valid */
+ args[2] = uint->one; /* DONE bit */
+ args[3] = lp_build_const_int32(base->gallivm, V_008DFC_SQ_EXP_NULL);
+ args[4] = uint->zero; /* COMPR flag (0 = 32-bit export) */
+ args[5] = uint->undef; /* R */
+ args[6] = uint->undef; /* G */
+ args[7] = uint->undef; /* B */
+ args[8] = uint->undef; /* A */
+
+ lp_build_intrinsic(base->gallivm->builder, "llvm.SI.export",
+ LLVMVoidTypeInContext(base->gallivm->context),
+ args, 9, 0);
+}
- args[1] =
- args[2] =
- args[4] = uint->zero;
+static void si_llvm_emit_fs_epilogue(struct lp_build_tgsi_context * bld_base)
+{
+ struct si_shader_context * si_shader_ctx = si_shader_context(bld_base);
+ struct si_shader * shader = si_shader_ctx->shader;
+ struct lp_build_context * base = &bld_base->base;
+ struct tgsi_shader_info *info = &shader->selector->info;
+ LLVMBuilderRef builder = base->gallivm->builder;
+ LLVMValueRef depth = NULL, stencil = NULL, samplemask = NULL;
+ int last_color_export = -1;
+ int i;
- if (last_args[0])
- lp_build_intrinsic(base->gallivm->builder,
- "llvm.SI.export",
- LLVMVoidTypeInContext(base->gallivm->context),
- args, 9, 0);
- else
- memcpy(last_args, args, sizeof(args));
+ /* If there are no outputs, add a dummy export. */
+ if (!info->num_outputs) {
+ si_export_null(bld_base);
+ return;
}
- if (!last_args[0]) {
- /* Specify which components to enable */
- last_args[0] = lp_build_const_int32(base->gallivm, 0x0);
+ /* Determine the last export. If MRTZ is present, it's always last.
+ * Otherwise, find the last color export.
+ */
+ if (!info->writes_z && !info->writes_stencil && !info->writes_samplemask)
+ for (i = 0; i < info->num_outputs; i++)
+ if (info->output_semantic_name[i] == TGSI_SEMANTIC_COLOR)
+ last_color_export = i;
- /* Specify the target we are exporting */
- last_args[3] = lp_build_const_int32(base->gallivm, V_008DFC_SQ_EXP_MRT);
+ for (i = 0; i < info->num_outputs; i++) {
+ unsigned semantic_name = info->output_semantic_name[i];
+ unsigned semantic_index = info->output_semantic_index[i];
+ unsigned j;
+ LLVMValueRef color[4] = {};
- /* Set COMPR flag to zero to export data as 32-bit */
- last_args[4] = uint->zero;
+ /* Select the correct target */
+ switch (semantic_name) {
+ case TGSI_SEMANTIC_POSITION:
+ depth = LLVMBuildLoad(builder,
+ si_shader_ctx->radeon_bld.soa.outputs[i][2], "");
+ break;
+ case TGSI_SEMANTIC_STENCIL:
+ stencil = LLVMBuildLoad(builder,
+ si_shader_ctx->radeon_bld.soa.outputs[i][1], "");
+ break;
+ case TGSI_SEMANTIC_SAMPLEMASK:
+ samplemask = LLVMBuildLoad(builder,
+ si_shader_ctx->radeon_bld.soa.outputs[i][0], "");
+ break;
+ case TGSI_SEMANTIC_COLOR:
+ for (j = 0; j < 4; j++)
+ color[j] = LLVMBuildLoad(builder,
+ si_shader_ctx->radeon_bld.soa.outputs[i][j], "");
- /* dummy bits */
- last_args[5]= uint->zero;
- last_args[6]= uint->zero;
- last_args[7]= uint->zero;
- last_args[8]= uint->zero;
+ si_export_mrt_color(bld_base, color, semantic_index,
+ last_color_export == i);
+ break;
+ default:
+ fprintf(stderr,
+ "Warning: SI unhandled fs output type:%d\n",
+ semantic_name);
+ }
}
- /* Specify whether the EXEC mask represents the valid mask */
- last_args[1] = uint->one;
-
- /* Specify that this is the last export */
- last_args[2] = lp_build_const_int32(base->gallivm, 1);
-
- lp_build_intrinsic(base->gallivm->builder,
- "llvm.SI.export",
- LLVMVoidTypeInContext(base->gallivm->context),
- last_args, 9, 0);
+ if (depth || stencil || samplemask)
+ si_export_mrt_z(bld_base, depth, stencil, samplemask);
}
static void build_tex_intrinsic(const struct lp_build_tgsi_action * action,
ind_index = get_indirect_index(si_shader_ctx, ®->Indirect, reg->Register.Index);
- *res_ptr = LLVMGetParam(si_shader_ctx->radeon_bld.main_fn, SI_PARAM_RESOURCE);
+ *res_ptr = LLVMGetParam(si_shader_ctx->radeon_bld.main_fn, SI_PARAM_SAMPLER_VIEWS);
*res_ptr = build_indexed_load_const(si_shader_ctx, *res_ptr, ind_index);
- *samp_ptr = LLVMGetParam(si_shader_ctx->radeon_bld.main_fn, SI_PARAM_SAMPLER);
+ *samp_ptr = LLVMGetParam(si_shader_ctx->radeon_bld.main_fn, SI_PARAM_SAMPLER_STATES);
*samp_ptr = build_indexed_load_const(si_shader_ctx, *samp_ptr, ind_index);
if (target == TGSI_TEXTURE_2D_MSAA ||
ind_index = LLVMBuildAdd(gallivm->builder, ind_index,
lp_build_const_int32(gallivm,
SI_FMASK_TEX_OFFSET), "");
- *fmask_ptr = LLVMGetParam(si_shader_ctx->radeon_bld.main_fn, SI_PARAM_RESOURCE);
+ *fmask_ptr = LLVMGetParam(si_shader_ctx->radeon_bld.main_fn, SI_PARAM_SAMPLER_VIEWS);
*fmask_ptr = build_indexed_load_const(si_shader_ctx, *fmask_ptr, ind_index);
}
} else {
- *res_ptr = si_shader_ctx->resources[sampler_index];
- *samp_ptr = si_shader_ctx->samplers[sampler_index];
- *fmask_ptr = si_shader_ctx->resources[SI_FMASK_TEX_OFFSET + sampler_index];
+ *res_ptr = si_shader_ctx->sampler_views[sampler_index];
+ *samp_ptr = si_shader_ctx->sampler_states[sampler_index];
+ *fmask_ptr = si_shader_ctx->sampler_views[SI_FMASK_TEX_OFFSET + sampler_index];
}
}
emit_data->dst_type = LLVMVectorType(bld_base->base.elem_type, 4);
emit_data->args[0] = res;
emit_data->args[1] = bld_base->uint_bld.zero;
- emit_data->args[2] = lp_build_emit_fetch(bld_base, emit_data->inst, 0, 0);
+ emit_data->args[2] = lp_build_emit_fetch(bld_base, emit_data->inst, 0, TGSI_CHAN_X);
emit_data->arg_count = 3;
return;
}
if (opcode == TGSI_OPCODE_TXB)
address[count++] = coords[3];
if (opcode == TGSI_OPCODE_TXB2)
- address[count++] = lp_build_emit_fetch(bld_base, inst, 1, 0);
+ address[count++] = lp_build_emit_fetch(bld_base, inst, 1, TGSI_CHAN_X);
/* Pack depth comparison value */
if (tgsi_is_shadow_target(target) && opcode != TGSI_OPCODE_LODQ) {
if (target == TGSI_TEXTURE_SHADOWCUBE_ARRAY) {
- address[count++] = lp_build_emit_fetch(bld_base, inst, 1, 0);
+ address[count++] = lp_build_emit_fetch(bld_base, inst, 1, TGSI_CHAN_X);
} else {
assert(ref_pos >= 0);
address[count++] = coords[ref_pos];
if (opcode == TGSI_OPCODE_TXL || opcode == TGSI_OPCODE_TXF)
address[count++] = coords[3];
else if (opcode == TGSI_OPCODE_TXL2)
- address[count++] = lp_build_emit_fetch(bld_base, inst, 1, 0);
+ address[count++] = lp_build_emit_fetch(bld_base, inst, 1, TGSI_CHAN_X);
if (count > 16) {
assert(!"Cannot handle more than 16 texture address parameters");
/* offset is in second src, first two channels */
emit_data->args[0] = lp_build_emit_fetch(bld_base,
emit_data->inst, 1,
- 0);
+ TGSI_CHAN_X);
emit_data->args[1] = lp_build_emit_fetch(bld_base,
emit_data->inst, 1,
- 1);
+ TGSI_CHAN_Y);
emit_data->arg_count = 2;
} else if (inst->Instruction.Opcode == TGSI_OPCODE_INTERP_SAMPLE) {
LLVMValueRef sample_position;
* and place into first two channels.
*/
sample_id = lp_build_emit_fetch(bld_base,
- emit_data->inst, 1, 0);
+ emit_data->inst, 1, TGSI_CHAN_X);
sample_id = LLVMBuildBitCast(gallivm->builder, sample_id,
LLVMInt32TypeInContext(gallivm->context),
"");
v16i8 = LLVMVectorType(i8, 16);
params[SI_PARAM_RW_BUFFERS] = const_array(v16i8, SI_NUM_RW_BUFFERS);
- params[SI_PARAM_CONST] = const_array(v16i8, SI_NUM_CONST_BUFFERS);
- params[SI_PARAM_SAMPLER] = const_array(v4i32, SI_NUM_SAMPLER_STATES);
- params[SI_PARAM_RESOURCE] = const_array(v8i32, SI_NUM_SAMPLER_VIEWS);
- last_array_pointer = SI_PARAM_RESOURCE;
+ params[SI_PARAM_CONST_BUFFERS] = const_array(v16i8, SI_NUM_CONST_BUFFERS);
+ params[SI_PARAM_SAMPLER_STATES] = const_array(v4i32, SI_NUM_SAMPLER_STATES);
+ params[SI_PARAM_SAMPLER_VIEWS] = const_array(v8i32, SI_NUM_SAMPLER_VIEWS);
+ last_array_pointer = SI_PARAM_SAMPLER_VIEWS;
switch (si_shader_ctx->type) {
case TGSI_PROCESSOR_VERTEX:
- params[SI_PARAM_VERTEX_BUFFER] = const_array(v16i8, SI_NUM_VERTEX_BUFFERS);
- last_array_pointer = SI_PARAM_VERTEX_BUFFER;
+ params[SI_PARAM_VERTEX_BUFFERS] = const_array(v16i8, SI_NUM_VERTEX_BUFFERS);
+ last_array_pointer = SI_PARAM_VERTEX_BUFFERS;
params[SI_PARAM_BASE_VERTEX] = i32;
params[SI_PARAM_START_INSTANCE] = i32;
num_params = SI_PARAM_START_INSTANCE+1;
num_params = SI_PARAM_LS_OUT_LAYOUT+1;
} else {
if (shader->is_gs_copy_shader) {
- last_array_pointer = SI_PARAM_CONST;
- num_params = SI_PARAM_CONST+1;
+ last_array_pointer = SI_PARAM_CONST_BUFFERS;
+ num_params = SI_PARAM_CONST_BUFFERS+1;
} else {
params[SI_PARAM_VS_STATE_BITS] = i32;
num_params = SI_PARAM_VS_STATE_BITS+1;
params[SI_PARAM_POS_Y_FLOAT] = f32;
params[SI_PARAM_POS_Z_FLOAT] = f32;
params[SI_PARAM_POS_W_FLOAT] = f32;
- params[SI_PARAM_FRONT_FACE] = f32;
+ params[SI_PARAM_FRONT_FACE] = i32;
params[SI_PARAM_ANCILLARY] = i32;
params[SI_PARAM_SAMPLE_COVERAGE] = f32;
params[SI_PARAM_POS_FIXED_PT] = f32;
struct gallivm_state * gallivm = bld_base->base.gallivm;
const struct tgsi_shader_info * info = bld_base->info;
unsigned buf;
- LLVMValueRef ptr = LLVMGetParam(si_shader_ctx->radeon_bld.main_fn, SI_PARAM_CONST);
+ LLVMValueRef ptr = LLVMGetParam(si_shader_ctx->radeon_bld.main_fn, SI_PARAM_CONST_BUFFERS);
for (buf = 0; buf < SI_NUM_CONST_BUFFERS; buf++) {
unsigned i, num_const = info->const_file_max[buf] + 1;
si_shader_ctx->constants[buf] = CALLOC(num_const * 4, sizeof(LLVMValueRef));
/* Load the resource descriptor */
- si_shader_ctx->const_resource[buf] =
+ si_shader_ctx->const_buffers[buf] =
build_indexed_load_const(si_shader_ctx, ptr, lp_build_const_int32(gallivm, buf));
/* Load the constants, we rely on the code sinking to do the rest */
for (i = 0; i < num_const * 4; ++i) {
si_shader_ctx->constants[buf][i] =
buffer_load_const(gallivm->builder,
- si_shader_ctx->const_resource[buf],
+ si_shader_ctx->const_buffers[buf],
lp_build_const_int32(gallivm, i * 4),
bld_base->base.elem_type);
}
if (num_samplers == 0)
return;
- res_ptr = LLVMGetParam(si_shader_ctx->radeon_bld.main_fn, SI_PARAM_RESOURCE);
- samp_ptr = LLVMGetParam(si_shader_ctx->radeon_bld.main_fn, SI_PARAM_SAMPLER);
+ res_ptr = LLVMGetParam(si_shader_ctx->radeon_bld.main_fn, SI_PARAM_SAMPLER_VIEWS);
+ samp_ptr = LLVMGetParam(si_shader_ctx->radeon_bld.main_fn, SI_PARAM_SAMPLER_STATES);
/* Load the resources and samplers, we rely on the code sinking to do the rest */
for (i = 0; i < num_samplers; ++i) {
/* Resource */
offset = lp_build_const_int32(gallivm, i);
- si_shader_ctx->resources[i] = build_indexed_load_const(si_shader_ctx, res_ptr, offset);
+ si_shader_ctx->sampler_views[i] = build_indexed_load_const(si_shader_ctx, res_ptr, offset);
/* Sampler */
offset = lp_build_const_int32(gallivm, i);
- si_shader_ctx->samplers[i] = build_indexed_load_const(si_shader_ctx, samp_ptr, offset);
+ si_shader_ctx->sampler_states[i] = build_indexed_load_const(si_shader_ctx, samp_ptr, offset);
/* FMASK resource */
if (info->is_msaa_sampler[i]) {
offset = lp_build_const_int32(gallivm, SI_FMASK_TEX_OFFSET + i);
- si_shader_ctx->resources[SI_FMASK_TEX_OFFSET + i] =
+ si_shader_ctx->sampler_views[SI_FMASK_TEX_OFFSET + i] =
build_indexed_load_const(si_shader_ctx, res_ptr, offset);
}
}
}
}
-void si_shader_binary_read_config(const struct si_screen *sscreen,
- struct si_shader *shader,
- unsigned symbol_offset)
+void si_shader_binary_read_config(struct radeon_shader_binary *binary,
+ struct si_shader_config *conf,
+ unsigned symbol_offset)
{
unsigned i;
const unsigned char *config =
- radeon_shader_binary_config_start(&shader->binary,
- symbol_offset);
+ radeon_shader_binary_config_start(binary, symbol_offset);
/* XXX: We may be able to emit some of these values directly rather than
* extracting fields to be emitted later.
*/
- for (i = 0; i < shader->binary.config_size_per_symbol; i+= 8) {
+ for (i = 0; i < binary->config_size_per_symbol; i+= 8) {
unsigned reg = util_le32_to_cpu(*(uint32_t*)(config + i));
unsigned value = util_le32_to_cpu(*(uint32_t*)(config + i + 4));
switch (reg) {
case R_00B128_SPI_SHADER_PGM_RSRC1_VS:
case R_00B228_SPI_SHADER_PGM_RSRC1_GS:
case R_00B848_COMPUTE_PGM_RSRC1:
- shader->num_sgprs = MAX2(shader->num_sgprs, (G_00B028_SGPRS(value) + 1) * 8);
- shader->num_vgprs = MAX2(shader->num_vgprs, (G_00B028_VGPRS(value) + 1) * 4);
- shader->float_mode = G_00B028_FLOAT_MODE(value);
- shader->rsrc1 = value;
+ conf->num_sgprs = MAX2(conf->num_sgprs, (G_00B028_SGPRS(value) + 1) * 8);
+ conf->num_vgprs = MAX2(conf->num_vgprs, (G_00B028_VGPRS(value) + 1) * 4);
+ conf->float_mode = G_00B028_FLOAT_MODE(value);
+ conf->rsrc1 = value;
break;
case R_00B02C_SPI_SHADER_PGM_RSRC2_PS:
- shader->lds_size = MAX2(shader->lds_size, G_00B02C_EXTRA_LDS_SIZE(value));
+ conf->lds_size = MAX2(conf->lds_size, G_00B02C_EXTRA_LDS_SIZE(value));
break;
case R_00B84C_COMPUTE_PGM_RSRC2:
- shader->lds_size = MAX2(shader->lds_size, G_00B84C_LDS_SIZE(value));
- shader->rsrc2 = value;
+ conf->lds_size = MAX2(conf->lds_size, G_00B84C_LDS_SIZE(value));
+ conf->rsrc2 = value;
break;
case R_0286CC_SPI_PS_INPUT_ENA:
- shader->spi_ps_input_ena = value;
+ conf->spi_ps_input_ena = value;
break;
case R_0286E8_SPI_TMPRING_SIZE:
case R_00B860_COMPUTE_TMPRING_SIZE:
/* WAVESIZE is in units of 256 dwords. */
- shader->scratch_bytes_per_wave =
+ conf->scratch_bytes_per_wave =
G_00B860_WAVESIZE(value) * 256 * 4 * 1;
break;
default:
uint32_t scratch_rsrc_dword0 = scratch_va;
uint32_t scratch_rsrc_dword1 =
S_008F04_BASE_ADDRESS_HI(scratch_va >> 32)
- | S_008F04_STRIDE(shader->scratch_bytes_per_wave / 64);
+ | S_008F04_STRIDE(shader->config.scratch_bytes_per_wave / 64);
for (i = 0 ; i < shader->binary.reloc_count; i++) {
const struct radeon_shader_reloc *reloc =
return 0;
}
-int si_shader_binary_read(struct si_screen *sscreen, struct si_shader *shader)
+static void si_shader_dump_disassembly(const struct radeon_shader_binary *binary,
+ struct pipe_debug_callback *debug)
{
- const struct radeon_shader_binary *binary = &shader->binary;
- unsigned i;
- int r;
- bool dump = r600_can_dump_shader(&sscreen->b,
- shader->selector ? shader->selector->tokens : NULL);
+ char *line, *p;
+ unsigned i, count;
+
+ if (binary->disasm_string) {
+ fprintf(stderr, "\nShader Disassembly:\n\n");
+ fprintf(stderr, "%s\n", binary->disasm_string);
+
+ if (debug && debug->debug_message) {
+ /* Very long debug messages are cut off, so send the
+ * disassembly one line at a time. This causes more
+ * overhead, but on the plus side it simplifies
+ * parsing of resulting logs.
+ */
+ pipe_debug_message(debug, SHADER_INFO,
+ "Shader Disassembly Begin");
- si_shader_binary_read_config(sscreen, shader, 0);
- r = si_shader_binary_upload(sscreen, shader);
- if (r)
- return r;
-
- if (dump) {
- if (!(sscreen->b.debug_flags & DBG_NO_ASM)) {
- if (binary->disasm_string) {
- fprintf(stderr, "\nShader Disassembly:\n\n");
- fprintf(stderr, "%s\n", binary->disasm_string);
- } else {
- fprintf(stderr, "SI CODE:\n");
- for (i = 0; i < binary->code_size; i+=4 ) {
- fprintf(stderr, "@0x%x: %02x%02x%02x%02x\n", i, binary->code[i + 3],
- binary->code[i + 2], binary->code[i + 1],
- binary->code[i]);
+ line = binary->disasm_string;
+ while (*line) {
+ p = strchrnul(line, '\n');
+ count = p - line;
+
+ if (count) {
+ pipe_debug_message(debug, SHADER_INFO,
+ "%.*s", count, line);
}
+
+ if (!*p)
+ break;
+ line = p + 1;
}
+
+ pipe_debug_message(debug, SHADER_INFO,
+ "Shader Disassembly End");
+ }
+ } else {
+ fprintf(stderr, "SI CODE:\n");
+ for (i = 0; i < binary->code_size; i += 4) {
+ fprintf(stderr, "@0x%x: %02x%02x%02x%02x\n", i,
+ binary->code[i + 3], binary->code[i + 2],
+ binary->code[i + 1], binary->code[i]);
}
+ }
+}
+static void si_shader_dump_stats(struct si_screen *sscreen,
+ struct si_shader_config *conf,
+ unsigned code_size,
+ struct pipe_debug_callback *debug,
+ unsigned processor)
+{
+ if (r600_can_dump_shader(&sscreen->b, processor)) {
fprintf(stderr, "*** SHADER STATS ***\n"
"SGPRS: %d\nVGPRS: %d\nCode Size: %d bytes\nLDS: %d blocks\n"
"Scratch: %d bytes per wave\n********************\n",
- shader->num_sgprs, shader->num_vgprs, binary->code_size,
- shader->lds_size, shader->scratch_bytes_per_wave);
+ conf->num_sgprs, conf->num_vgprs, code_size,
+ conf->lds_size, conf->scratch_bytes_per_wave);
}
- return 0;
+
+ pipe_debug_message(debug, SHADER_INFO,
+ "Shader Stats: SGPRS: %d VGPRS: %d Code Size: %d LDS: %d Scratch: %d",
+ conf->num_sgprs, conf->num_vgprs, code_size,
+ conf->lds_size, conf->scratch_bytes_per_wave);
+}
+
+void si_shader_dump(struct si_screen *sscreen, struct si_shader *shader,
+ struct pipe_debug_callback *debug, unsigned processor)
+{
+ if (r600_can_dump_shader(&sscreen->b, processor))
+ if (!(sscreen->b.debug_flags & DBG_NO_ASM))
+ si_shader_dump_disassembly(&shader->binary, debug);
+
+ si_shader_dump_stats(sscreen, &shader->config,
+ shader->binary.code_size, debug, processor);
}
-int si_compile_llvm(struct si_screen *sscreen, struct si_shader *shader,
- LLVMTargetMachineRef tm, LLVMModuleRef mod)
+int si_compile_llvm(struct si_screen *sscreen,
+ struct radeon_shader_binary *binary,
+ struct si_shader_config *conf,
+ LLVMTargetMachineRef tm,
+ LLVMModuleRef mod,
+ struct pipe_debug_callback *debug,
+ unsigned processor)
{
int r = 0;
- bool dump_asm = r600_can_dump_shader(&sscreen->b,
- shader->selector ? shader->selector->tokens : NULL);
- bool dump_ir = dump_asm && !(sscreen->b.debug_flags & DBG_NO_IR);
-
- r = radeon_llvm_compile(mod, &shader->binary,
- r600_get_llvm_processor_name(sscreen->b.family), dump_ir, dump_asm, tm);
- if (r)
- return r;
-
- r = si_shader_binary_read(sscreen, shader);
-
- FREE(shader->binary.config);
- FREE(shader->binary.rodata);
- FREE(shader->binary.global_symbol_offsets);
- if (shader->scratch_bytes_per_wave == 0) {
- FREE(shader->binary.code);
- FREE(shader->binary.relocs);
- memset(&shader->binary, 0,
- offsetof(struct radeon_shader_binary, disasm_string));
+ unsigned count = p_atomic_inc_return(&sscreen->b.num_compilations);
+
+ if (r600_can_dump_shader(&sscreen->b, processor)) {
+ fprintf(stderr, "radeonsi: Compiling shader %d\n", count);
+
+ if (!(sscreen->b.debug_flags & DBG_NO_IR))
+ LLVMDumpModule(mod);
}
+
+ if (!si_replace_shader(count, binary)) {
+ r = radeon_llvm_compile(mod, binary,
+ r600_get_llvm_processor_name(sscreen->b.family), tm,
+ debug);
+ if (r)
+ return r;
+ }
+
+ si_shader_binary_read_config(binary, conf, 0);
+
+ FREE(binary->config);
+ FREE(binary->global_symbol_offsets);
+ binary->config = NULL;
+ binary->global_symbol_offsets = NULL;
return r;
}
/* Generate code for the hardware VS shader stage to go with a geometry shader */
static int si_generate_gs_copy_shader(struct si_screen *sscreen,
struct si_shader_context *si_shader_ctx,
- struct si_shader *gs, bool dump)
+ struct si_shader *gs, bool dump,
+ struct pipe_debug_callback *debug)
{
struct gallivm_state *gallivm = &si_shader_ctx->radeon_bld.gallivm;
struct lp_build_tgsi_context *bld_base = &si_shader_ctx->radeon_bld.soa.bld_base;
if (dump)
fprintf(stderr, "Copy Vertex Shader for Geometry Shader:\n\n");
- r = si_compile_llvm(sscreen, si_shader_ctx->shader,
- si_shader_ctx->tm, bld_base->base.gallivm->module);
+ r = si_compile_llvm(sscreen, &si_shader_ctx->shader->binary,
+ &si_shader_ctx->shader->config, si_shader_ctx->tm,
+ bld_base->base.gallivm->module,
+ debug, TGSI_PROCESSOR_GEOMETRY);
+ if (!r) {
+ si_shader_dump(sscreen, si_shader_ctx->shader, debug,
+ TGSI_PROCESSOR_GEOMETRY);
+ r = si_shader_binary_upload(sscreen, si_shader_ctx->shader);
+ }
radeon_llvm_dispose(&si_shader_ctx->radeon_bld);
}
int si_shader_create(struct si_screen *sscreen, LLVMTargetMachineRef tm,
- struct si_shader *shader)
+ struct si_shader *shader,
+ struct pipe_debug_callback *debug)
{
struct si_shader_selector *sel = shader->selector;
struct tgsi_token *tokens = sel->tokens;
int r = 0;
bool poly_stipple = sel->type == PIPE_SHADER_FRAGMENT &&
shader->key.ps.poly_stipple;
- bool dump = r600_can_dump_shader(&sscreen->b, sel->tokens);
+ bool dump = r600_can_dump_shader(&sscreen->b, sel->info.processor);
if (poly_stipple) {
tokens = util_pstipple_create_fragment_shader(tokens, NULL,
- SI_POLY_STIPPLE_SAMPLER);
+ SI_POLY_STIPPLE_SAMPLER,
+ TGSI_FILE_SYSTEM_VALUE);
tgsi_scan_shader(tokens, &stipple_shader_info);
}
if (sel->type != PIPE_SHADER_COMPUTE)
shader->dx10_clamp_mode = true;
- if (sel->info.uses_kill)
- shader->db_shader_control |= S_02880C_KILL_ENABLE(1);
-
shader->uses_instanceid = sel->info.uses_instanceid;
bld_base->info = poly_stipple ? &stipple_shader_info : &sel->info;
bld_base->emit_fetch_funcs[TGSI_FILE_CONSTANT] = fetch_constant;
case TGSI_PROCESSOR_FRAGMENT:
si_shader_ctx.radeon_bld.load_input = declare_input_fs;
bld_base->emit_epilogue = si_llvm_emit_fs_epilogue;
-
- switch (sel->info.properties[TGSI_PROPERTY_FS_DEPTH_LAYOUT]) {
- case TGSI_FS_DEPTH_LAYOUT_GREATER:
- shader->db_shader_control |=
- S_02880C_CONSERVATIVE_Z_EXPORT(V_02880C_EXPORT_GREATER_THAN_Z);
- break;
- case TGSI_FS_DEPTH_LAYOUT_LESS:
- shader->db_shader_control |=
- S_02880C_CONSERVATIVE_Z_EXPORT(V_02880C_EXPORT_LESS_THAN_Z);
- break;
- }
break;
default:
assert(!"Unsupported shader type");
radeon_llvm_finalize_module(&si_shader_ctx.radeon_bld);
mod = bld_base->base.gallivm->module;
- r = si_compile_llvm(sscreen, shader, tm, mod);
+ r = si_compile_llvm(sscreen, &shader->binary, &shader->config, tm,
+ mod, debug, si_shader_ctx.type);
if (r) {
fprintf(stderr, "LLVM failed to compile shader\n");
goto out;
}
+ si_shader_dump(sscreen, shader, debug, si_shader_ctx.type);
+
+ r = si_shader_binary_upload(sscreen, shader);
+ if (r) {
+ fprintf(stderr, "LLVM failed to upload shader\n");
+ goto out;
+ }
+
radeon_llvm_dispose(&si_shader_ctx.radeon_bld);
if (si_shader_ctx.type == TGSI_PROCESSOR_GEOMETRY) {
shader->gs_copy_shader->key = shader->key;
si_shader_ctx.shader = shader->gs_copy_shader;
if ((r = si_generate_gs_copy_shader(sscreen, &si_shader_ctx,
- shader, dump))) {
+ shader, dump, debug))) {
free(shader->gs_copy_shader);
shader->gs_copy_shader = NULL;
goto out;
return r;
}
+void si_shader_destroy_binary(struct radeon_shader_binary *binary)
+{
+ FREE(binary->code);
+ FREE(binary->rodata);
+ FREE(binary->relocs);
+ FREE(binary->disasm_string);
+}
+
void si_shader_destroy(struct si_shader *shader)
{
if (shader->gs_copy_shader) {
r600_resource_reference(&shader->scratch_bo, NULL);
r600_resource_reference(&shader->bo, NULL);
-
- FREE(shader->binary.code);
- FREE(shader->binary.relocs);
- FREE(shader->binary.disasm_string);
+ si_shader_destroy_binary(&shader->binary);
}
struct radeon_shader_reloc;
#define SI_SGPR_RW_BUFFERS 0 /* rings (& stream-out, VS only) */
-#define SI_SGPR_CONST 2
-#define SI_SGPR_SAMPLER 4
-#define SI_SGPR_RESOURCE 6
-#define SI_SGPR_VERTEX_BUFFER 8 /* VS only */
+#define SI_SGPR_CONST_BUFFERS 2
+#define SI_SGPR_SAMPLER_STATES 4
+#define SI_SGPR_SAMPLER_VIEWS 6
+#define SI_SGPR_VERTEX_BUFFERS 8 /* VS only */
#define SI_SGPR_BASE_VERTEX 10 /* VS only */
#define SI_SGPR_START_INSTANCE 11 /* VS only */
#define SI_SGPR_VS_STATE_BITS 12 /* VS(VS) only */
/* LLVM function parameter indices */
#define SI_PARAM_RW_BUFFERS 0
-#define SI_PARAM_CONST 1
-#define SI_PARAM_SAMPLER 2
-#define SI_PARAM_RESOURCE 3
+#define SI_PARAM_CONST_BUFFERS 1
+#define SI_PARAM_SAMPLER_STATES 2
+#define SI_PARAM_SAMPLER_VIEWS 3
/* VS only parameters */
-#define SI_PARAM_VERTEX_BUFFER 4
+#define SI_PARAM_VERTEX_BUFFERS 4
#define SI_PARAM_BASE_VERTEX 5
#define SI_PARAM_START_INSTANCE 6
/* [0] = clamp vertex color */
bool forces_persample_interp_for_persp;
bool forces_persample_interp_for_linear;
+ /* GS parameters. */
unsigned esgs_itemsize;
unsigned gs_input_verts_per_prim;
unsigned gs_output_prim;
unsigned gsvs_vertex_size;
unsigned max_gsvs_emit_size;
+ /* PS parameters. */
+ unsigned db_shader_control;
+
/* masks of "get_unique_index" bits */
uint64_t outputs_written;
uint32_t patch_outputs_written;
} tes; /* tessellation evaluation shader */
};
+struct si_shader_config {
+ unsigned num_sgprs;
+ unsigned num_vgprs;
+ unsigned lds_size;
+ unsigned spi_ps_input_ena;
+ unsigned float_mode;
+ unsigned scratch_bytes_per_wave;
+ unsigned rsrc1;
+ unsigned rsrc2;
+};
+
struct si_shader {
struct si_shader_selector *selector;
struct si_shader *next_variant;
struct si_pm4_state *pm4;
struct r600_resource *bo;
struct r600_resource *scratch_bo;
- struct radeon_shader_binary binary;
- unsigned num_sgprs;
- unsigned num_vgprs;
- unsigned lds_size;
- unsigned spi_ps_input_ena;
- unsigned float_mode;
- unsigned scratch_bytes_per_wave;
- unsigned spi_shader_col_format;
- unsigned spi_shader_z_format;
- unsigned db_shader_control;
- unsigned cb_shader_mask;
union si_shader_key key;
+ struct radeon_shader_binary binary;
+ struct si_shader_config config;
unsigned nparam;
unsigned vs_output_param_offset[PIPE_MAX_SHADER_OUTPUTS];
unsigned nr_param_exports;
bool is_gs_copy_shader;
bool dx10_clamp_mode; /* convert NaNs to 0 */
-
- unsigned rsrc1;
- unsigned rsrc2;
};
static inline struct tgsi_shader_info *si_get_vs_info(struct si_context *sctx)
/* radeonsi_shader.c */
int si_shader_create(struct si_screen *sscreen, LLVMTargetMachineRef tm,
- struct si_shader *shader);
+ struct si_shader *shader,
+ struct pipe_debug_callback *debug);
void si_dump_shader_key(unsigned shader, union si_shader_key *key, FILE *f);
-int si_compile_llvm(struct si_screen *sscreen, struct si_shader *shader,
- LLVMTargetMachineRef tm, LLVMModuleRef mod);
+int si_compile_llvm(struct si_screen *sscreen,
+ struct radeon_shader_binary *binary,
+ struct si_shader_config *conf,
+ LLVMTargetMachineRef tm,
+ LLVMModuleRef mod,
+ struct pipe_debug_callback *debug,
+ unsigned processor);
void si_shader_destroy(struct si_shader *shader);
+void si_shader_destroy_binary(struct radeon_shader_binary *binary);
unsigned si_shader_io_get_unique_index(unsigned semantic_name, unsigned index);
int si_shader_binary_upload(struct si_screen *sscreen, struct si_shader *shader);
-int si_shader_binary_read(struct si_screen *sscreen, struct si_shader *shader);
+void si_shader_dump(struct si_screen *sscreen, struct si_shader *shader,
+ struct pipe_debug_callback *debug, unsigned processor);
void si_shader_apply_scratch_relocs(struct si_context *sctx,
struct si_shader *shader,
uint64_t scratch_va);
-void si_shader_binary_read_config(const struct si_screen *sscreen,
- struct si_shader *shader,
- unsigned symbol_offset);
+void si_shader_binary_read_config(struct radeon_shader_binary *binary,
+ struct si_shader_config *conf,
+ unsigned symbol_offset);
#endif
(clipdist_mask ? 0 :
sctx->queued.named.rasterizer->clip_plane_enable & SIX_BITS) |
S_028810_CLIP_DISABLE(window_space));
+
+ /* reuse needs to be set off if we write oViewport */
+ radeon_set_context_reg(cs, R_028AB4_VGT_REUSE_OFF,
+ S_028AB4_REUSE_OFF(info->writes_viewport_index));
}
static void si_set_scissor_states(struct pipe_context *ctx,
si_pm4_set_reg(pm4, R_028B28_VGT_STRMOUT_DRAW_OPAQUE_OFFSET, 0);
si_pm4_set_reg(pm4, R_028B98_VGT_STRMOUT_BUFFER_CONFIG, 0x0);
- si_pm4_set_reg(pm4, R_028AB4_VGT_REUSE_OFF, 0);
si_pm4_set_reg(pm4, R_028AB8_VGT_VTX_CNT_EN, 0x0);
if (sctx->b.chip_class < CIK)
si_pm4_set_reg(pm4, R_008A14_PA_CL_ENHANCE, S_008A14_NUM_CLIP_SEQ(3) |
perpatch_output_offset = output_patch0_offset + pervertex_output_patch_size;
lds_size = output_patch0_offset + output_patch_size * *num_patches;
- ls_rsrc2 = ls->current->rsrc2;
+ ls_rsrc2 = ls->current->config.rsrc2;
if (sctx->b.chip_class >= CIK) {
assert(lds_size <= 65536);
if (sctx->b.chip_class == CIK && sctx->b.family != CHIP_HAWAII)
radeon_set_sh_reg(cs, R_00B52C_SPI_SHADER_PGM_RSRC2_LS, ls_rsrc2);
radeon_set_sh_reg_seq(cs, R_00B528_SPI_SHADER_PGM_RSRC1_LS, 2);
- radeon_emit(cs, ls->current->rsrc1);
+ radeon_emit(cs, ls->current->config.rsrc1);
radeon_emit(cs, ls_rsrc2);
/* Compute userdata SGPRs. */
si_get_draw_start_count(sctx, info, &start, &count);
start_offset = start * ib.index_size;
- u_upload_alloc(sctx->b.uploader, start_offset, count * 2,
+ u_upload_alloc(sctx->b.uploader, start_offset, count * 2, 256,
&out_offset, &out_buffer, &ptr);
if (!out_buffer) {
pipe_resource_reference(&ib.buffer, NULL);
start_offset = start * ib.index_size;
u_upload_data(sctx->b.uploader, start_offset, count * ib.index_size,
- (char*)ib.user_buffer + start_offset,
+ 256, (char*)ib.user_buffer + start_offset,
&ib.offset, &ib.buffer);
if (!ib.buffer)
return;
vgpr_comp_cnt = shader->uses_instanceid ? 3 : 1;
num_user_sgprs = SI_LS_NUM_USER_SGPR;
- num_sgprs = shader->num_sgprs;
+ num_sgprs = shader->config.num_sgprs;
if (num_user_sgprs > num_sgprs) {
/* Last 2 reserved SGPRs are used for VCC */
num_sgprs = num_user_sgprs + 2;
si_pm4_set_reg(pm4, R_00B520_SPI_SHADER_PGM_LO_LS, va >> 8);
si_pm4_set_reg(pm4, R_00B524_SPI_SHADER_PGM_HI_LS, va >> 40);
- shader->rsrc1 = S_00B528_VGPRS((shader->num_vgprs - 1) / 4) |
+ shader->config.rsrc1 = S_00B528_VGPRS((shader->config.num_vgprs - 1) / 4) |
S_00B528_SGPRS((num_sgprs - 1) / 8) |
S_00B528_VGPR_COMP_CNT(vgpr_comp_cnt) |
S_00B528_DX10_CLAMP(shader->dx10_clamp_mode);
- shader->rsrc2 = S_00B52C_USER_SGPR(num_user_sgprs) |
- S_00B52C_SCRATCH_EN(shader->scratch_bytes_per_wave > 0);
+ shader->config.rsrc2 = S_00B52C_USER_SGPR(num_user_sgprs) |
+ S_00B52C_SCRATCH_EN(shader->config.scratch_bytes_per_wave > 0);
}
static void si_shader_hs(struct si_shader *shader)
si_pm4_add_bo(pm4, shader->bo, RADEON_USAGE_READ, RADEON_PRIO_USER_SHADER);
num_user_sgprs = SI_TCS_NUM_USER_SGPR;
- num_sgprs = shader->num_sgprs;
+ num_sgprs = shader->config.num_sgprs;
/* One SGPR after user SGPRs is pre-loaded with tessellation factor
* buffer offset. */
if ((num_user_sgprs + 1) > num_sgprs) {
si_pm4_set_reg(pm4, R_00B420_SPI_SHADER_PGM_LO_HS, va >> 8);
si_pm4_set_reg(pm4, R_00B424_SPI_SHADER_PGM_HI_HS, va >> 40);
si_pm4_set_reg(pm4, R_00B428_SPI_SHADER_PGM_RSRC1_HS,
- S_00B428_VGPRS((shader->num_vgprs - 1) / 4) |
+ S_00B428_VGPRS((shader->config.num_vgprs - 1) / 4) |
S_00B428_SGPRS((num_sgprs - 1) / 8) |
S_00B428_DX10_CLAMP(shader->dx10_clamp_mode));
si_pm4_set_reg(pm4, R_00B42C_SPI_SHADER_PGM_RSRC2_HS,
S_00B42C_USER_SGPR(num_user_sgprs) |
- S_00B42C_SCRATCH_EN(shader->scratch_bytes_per_wave > 0));
+ S_00B42C_SCRATCH_EN(shader->config.scratch_bytes_per_wave > 0));
}
static void si_shader_es(struct si_shader *shader)
} else
unreachable("invalid shader selector type");
- num_sgprs = shader->num_sgprs;
+ num_sgprs = shader->config.num_sgprs;
/* One SGPR after user SGPRs is pre-loaded with es2gs_offset */
if ((num_user_sgprs + 1) > num_sgprs) {
/* Last 2 reserved SGPRs are used for VCC */
si_pm4_set_reg(pm4, R_00B320_SPI_SHADER_PGM_LO_ES, va >> 8);
si_pm4_set_reg(pm4, R_00B324_SPI_SHADER_PGM_HI_ES, va >> 40);
si_pm4_set_reg(pm4, R_00B328_SPI_SHADER_PGM_RSRC1_ES,
- S_00B328_VGPRS((shader->num_vgprs - 1) / 4) |
+ S_00B328_VGPRS((shader->config.num_vgprs - 1) / 4) |
S_00B328_SGPRS((num_sgprs - 1) / 8) |
S_00B328_VGPR_COMP_CNT(vgpr_comp_cnt) |
S_00B328_DX10_CLAMP(shader->dx10_clamp_mode));
si_pm4_set_reg(pm4, R_00B32C_SPI_SHADER_PGM_RSRC2_ES,
S_00B32C_USER_SGPR(num_user_sgprs) |
- S_00B32C_SCRATCH_EN(shader->scratch_bytes_per_wave > 0));
+ S_00B32C_SCRATCH_EN(shader->config.scratch_bytes_per_wave > 0));
if (shader->selector->type == PIPE_SHADER_TESS_EVAL)
si_set_tesseval_regs(shader, pm4);
si_pm4_set_reg(pm4, R_00B224_SPI_SHADER_PGM_HI_GS, va >> 40);
num_user_sgprs = SI_GS_NUM_USER_SGPR;
- num_sgprs = shader->num_sgprs;
+ num_sgprs = shader->config.num_sgprs;
/* Two SGPRs after user SGPRs are pre-loaded with gs2vs_offset, gs_wave_id */
if ((num_user_sgprs + 2) > num_sgprs) {
/* Last 2 reserved SGPRs are used for VCC */
assert(num_sgprs <= 104);
si_pm4_set_reg(pm4, R_00B228_SPI_SHADER_PGM_RSRC1_GS,
- S_00B228_VGPRS((shader->num_vgprs - 1) / 4) |
+ S_00B228_VGPRS((shader->config.num_vgprs - 1) / 4) |
S_00B228_SGPRS((num_sgprs - 1) / 8) |
S_00B228_DX10_CLAMP(shader->dx10_clamp_mode));
si_pm4_set_reg(pm4, R_00B22C_SPI_SHADER_PGM_RSRC2_GS,
S_00B22C_USER_SGPR(num_user_sgprs) |
- S_00B22C_SCRATCH_EN(shader->scratch_bytes_per_wave > 0));
+ S_00B22C_SCRATCH_EN(shader->config.scratch_bytes_per_wave > 0));
}
static void si_shader_vs(struct si_shader *shader)
} else
unreachable("invalid shader selector type");
- num_sgprs = shader->num_sgprs;
+ num_sgprs = shader->config.num_sgprs;
if (num_user_sgprs > num_sgprs) {
/* Last 2 reserved SGPRs are used for VCC */
num_sgprs = num_user_sgprs + 2;
si_pm4_set_reg(pm4, R_00B120_SPI_SHADER_PGM_LO_VS, va >> 8);
si_pm4_set_reg(pm4, R_00B124_SPI_SHADER_PGM_HI_VS, va >> 40);
si_pm4_set_reg(pm4, R_00B128_SPI_SHADER_PGM_RSRC1_VS,
- S_00B128_VGPRS((shader->num_vgprs - 1) / 4) |
+ S_00B128_VGPRS((shader->config.num_vgprs - 1) / 4) |
S_00B128_SGPRS((num_sgprs - 1) / 8) |
S_00B128_VGPR_COMP_CNT(vgpr_comp_cnt) |
S_00B128_DX10_CLAMP(shader->dx10_clamp_mode));
S_00B12C_SO_BASE2_EN(!!shader->selector->so.stride[2]) |
S_00B12C_SO_BASE3_EN(!!shader->selector->so.stride[3]) |
S_00B12C_SO_EN(!!shader->selector->so.num_outputs) |
- S_00B12C_SCRATCH_EN(shader->scratch_bytes_per_wave > 0));
+ S_00B12C_SCRATCH_EN(shader->config.scratch_bytes_per_wave > 0));
if (window_space)
si_pm4_set_reg(pm4, R_028818_PA_CL_VTE_CNTL,
S_028818_VTX_XY_FMT(1) | S_028818_VTX_Z_FMT(1));
struct tgsi_shader_info *info = &shader->selector->info;
struct si_pm4_state *pm4;
unsigned i, spi_ps_in_control;
+ unsigned spi_shader_col_format = 0, cb_shader_mask = 0;
+ unsigned colors_written, export_16bpc;
unsigned num_sgprs, num_user_sgprs;
- unsigned spi_baryc_cntl = 0;
+ unsigned spi_baryc_cntl = S_0286E0_FRONT_FACE_ALL_BITS(1);
uint64_t va;
bool has_centroid;
if (!pm4)
return;
- for (i = 0; i < info->num_inputs; i++) {
- switch (info->input_semantic_name[i]) {
- case TGSI_SEMANTIC_POSITION:
- /* SPI_BARYC_CNTL.POS_FLOAT_LOCATION
- * Possible vaules:
- * 0 -> Position = pixel center (default)
- * 1 -> Position = pixel centroid
- * 2 -> Position = at sample position
- */
- switch (info->input_interpolate_loc[i]) {
- case TGSI_INTERPOLATE_LOC_CENTROID:
- spi_baryc_cntl |= S_0286E0_POS_FLOAT_LOCATION(1);
- break;
- case TGSI_INTERPOLATE_LOC_SAMPLE:
- spi_baryc_cntl |= S_0286E0_POS_FLOAT_LOCATION(2);
- break;
- }
+ /* SPI_BARYC_CNTL.POS_FLOAT_LOCATION
+ * Possible vaules:
+ * 0 -> Position = pixel center
+ * 1 -> Position = pixel centroid
+ * 2 -> Position = at sample position
+ *
+ * From GLSL 4.5 specification, section 7.1:
+ * "The variable gl_FragCoord is available as an input variable from
+ * within fragment shaders and it holds the window relative coordinates
+ * (x, y, z, 1/w) values for the fragment. If multi-sampling, this
+ * value can be for any location within the pixel, or one of the
+ * fragment samples. The use of centroid does not further restrict
+ * this value to be inside the current primitive."
+ *
+ * Meaning that centroid has no effect and we can return anything within
+ * the pixel. Thus, return the value at sample position, because that's
+ * the most accurate one shaders can get.
+ */
+ spi_baryc_cntl |= S_0286E0_POS_FLOAT_LOCATION(2);
- if (info->properties[TGSI_PROPERTY_FS_COORD_PIXEL_CENTER] ==
- TGSI_FS_COORD_PIXEL_CENTER_INTEGER)
- spi_baryc_cntl |= S_0286E0_POS_FLOAT_ULC(1);
- break;
- }
+ if (info->properties[TGSI_PROPERTY_FS_COORD_PIXEL_CENTER] ==
+ TGSI_FS_COORD_PIXEL_CENTER_INTEGER)
+ spi_baryc_cntl |= S_0286E0_POS_FLOAT_ULC(1);
+
+ /* Find out what SPI_SHADER_COL_FORMAT and CB_SHADER_MASK should be. */
+ colors_written = info->colors_written;
+ export_16bpc = shader->key.ps.export_16bpc;
+
+ if (info->colors_written == 0x1 &&
+ info->properties[TGSI_PROPERTY_FS_COLOR0_WRITES_ALL_CBUFS]) {
+ colors_written |= (1 << (shader->key.ps.last_cbuf + 1)) - 1;
+ }
+
+ while (colors_written) {
+ i = u_bit_scan(&colors_written);
+ if (export_16bpc & (1 << i))
+ spi_shader_col_format |= V_028714_SPI_SHADER_FP16_ABGR << (4 * i);
+ else
+ spi_shader_col_format |= V_028714_SPI_SHADER_32_ABGR << (4 * i);
+ cb_shader_mask |= 0xf << (4 * i);
}
- has_centroid = G_0286CC_PERSP_CENTROID_ENA(shader->spi_ps_input_ena) ||
- G_0286CC_LINEAR_CENTROID_ENA(shader->spi_ps_input_ena);
+ /* Set interpolation controls. */
+ has_centroid = G_0286CC_PERSP_CENTROID_ENA(shader->config.spi_ps_input_ena) ||
+ G_0286CC_LINEAR_CENTROID_ENA(shader->config.spi_ps_input_ena);
spi_ps_in_control = S_0286D8_NUM_INTERP(shader->nparam) |
S_0286D8_BC_OPTIMIZE_DISABLE(has_centroid);
+ /* Set registers. */
si_pm4_set_reg(pm4, R_0286E0_SPI_BARYC_CNTL, spi_baryc_cntl);
si_pm4_set_reg(pm4, R_0286D8_SPI_PS_IN_CONTROL, spi_ps_in_control);
- si_pm4_set_reg(pm4, R_028710_SPI_SHADER_Z_FORMAT, shader->spi_shader_z_format);
- si_pm4_set_reg(pm4, R_028714_SPI_SHADER_COL_FORMAT,
- shader->spi_shader_col_format);
- si_pm4_set_reg(pm4, R_02823C_CB_SHADER_MASK, shader->cb_shader_mask);
+ si_pm4_set_reg(pm4, R_028710_SPI_SHADER_Z_FORMAT,
+ info->writes_samplemask ? V_028710_SPI_SHADER_32_ABGR :
+ info->writes_stencil ? V_028710_SPI_SHADER_32_GR :
+ info->writes_z ? V_028710_SPI_SHADER_32_R :
+ V_028710_SPI_SHADER_ZERO);
+
+ si_pm4_set_reg(pm4, R_028714_SPI_SHADER_COL_FORMAT, spi_shader_col_format);
+ si_pm4_set_reg(pm4, R_02823C_CB_SHADER_MASK, cb_shader_mask);
va = shader->bo->gpu_address;
si_pm4_add_bo(pm4, shader->bo, RADEON_USAGE_READ, RADEON_PRIO_USER_SHADER);
si_pm4_set_reg(pm4, R_00B024_SPI_SHADER_PGM_HI_PS, va >> 40);
num_user_sgprs = SI_PS_NUM_USER_SGPR;
- num_sgprs = shader->num_sgprs;
+ num_sgprs = shader->config.num_sgprs;
/* One SGPR after user SGPRs is pre-loaded with {prim_mask, lds_offset} */
if ((num_user_sgprs + 1) > num_sgprs) {
/* Last 2 reserved SGPRs are used for VCC */
assert(num_sgprs <= 104);
si_pm4_set_reg(pm4, R_00B028_SPI_SHADER_PGM_RSRC1_PS,
- S_00B028_VGPRS((shader->num_vgprs - 1) / 4) |
+ S_00B028_VGPRS((shader->config.num_vgprs - 1) / 4) |
S_00B028_SGPRS((num_sgprs - 1) / 8) |
S_00B028_DX10_CLAMP(shader->dx10_clamp_mode));
si_pm4_set_reg(pm4, R_00B02C_SPI_SHADER_PGM_RSRC2_PS,
- S_00B02C_EXTRA_LDS_SIZE(shader->lds_size) |
+ S_00B02C_EXTRA_LDS_SIZE(shader->config.lds_size) |
S_00B02C_USER_SGPR(num_user_sgprs) |
- S_00B32C_SCRATCH_EN(shader->scratch_bytes_per_wave > 0));
+ S_00B32C_SCRATCH_EN(shader->config.scratch_bytes_per_wave > 0));
}
static void si_shader_init_pm4_state(struct si_shader *shader)
}
}
+static unsigned si_get_alpha_test_func(struct si_context *sctx)
+{
+ /* Alpha-test should be disabled if colorbuffer 0 is integer. */
+ if (sctx->queued.named.dsa &&
+ !sctx->framebuffer.cb0_is_integer)
+ return sctx->queued.named.dsa->alpha_func;
+
+ return PIPE_FUNC_ALWAYS;
+}
+
/* Compute the key for the hw shader variant */
static inline void si_shader_selector_key(struct pipe_context *ctx,
struct si_shader_selector *sel,
case PIPE_SHADER_FRAGMENT: {
struct si_state_rasterizer *rs = sctx->queued.named.rasterizer;
- if (sel->info.properties[TGSI_PROPERTY_FS_COLOR0_WRITES_ALL_CBUFS])
+ if (sel->info.properties[TGSI_PROPERTY_FS_COLOR0_WRITES_ALL_CBUFS] &&
+ sel->info.colors_written == 0x1)
key->ps.last_cbuf = MAX2(sctx->framebuffer.state.nr_cbufs, 1) - 1;
+
key->ps.export_16bpc = sctx->framebuffer.export_16bpc;
if (rs) {
key->ps.clamp_color = rs->clamp_fragment_color;
}
- key->ps.alpha_func = PIPE_FUNC_ALWAYS;
- /* Alpha-test should be disabled if colorbuffer 0 is integer. */
- if (sctx->queued.named.dsa &&
- !sctx->framebuffer.cb0_is_integer)
- key->ps.alpha_func = sctx->queued.named.dsa->alpha_func;
+ key->ps.alpha_func = si_get_alpha_test_func(sctx);
break;
}
default:
shader->selector = sel;
shader->key = key;
- r = si_shader_create(sctx->screen, sctx->tm, shader);
+ r = si_shader_create(sctx->screen, sctx->tm, shader, &sctx->b.debug);
if (unlikely(r)) {
R600_ERR("Failed to build shader variant (type=%u) %d\n",
sel->type, r);
sel->last_variant = shader;
}
state->current = shader;
- p_atomic_inc(&sctx->screen->b.num_compilations);
pipe_mutex_unlock(sel->mutex);
return 0;
}
break;
}
+ /* DB_SHADER_CONTROL */
+ sel->db_shader_control =
+ S_02880C_Z_EXPORT_ENABLE(sel->info.writes_z) |
+ S_02880C_STENCIL_TEST_VAL_EXPORT_ENABLE(sel->info.writes_stencil) |
+ S_02880C_MASK_EXPORT_ENABLE(sel->info.writes_samplemask) |
+ S_02880C_KILL_ENABLE(sel->info.uses_kill);
+
+ switch (sel->info.properties[TGSI_PROPERTY_FS_DEPTH_LAYOUT]) {
+ case TGSI_FS_DEPTH_LAYOUT_GREATER:
+ sel->db_shader_control |=
+ S_02880C_CONSERVATIVE_Z_EXPORT(V_02880C_EXPORT_GREATER_THAN_Z);
+ break;
+ case TGSI_FS_DEPTH_LAYOUT_LESS:
+ sel->db_shader_control |=
+ S_02880C_CONSERVATIVE_Z_EXPORT(V_02880C_EXPORT_LESS_THAN_Z);
+ break;
+ }
+
+ /* Pre-compilation. */
if (sscreen->b.debug_flags & DBG_PRECOMPILE) {
struct si_shader_ctx_state state = {sel};
unsigned index = psinfo->input_semantic_index[i];
unsigned interpolate = psinfo->input_interpolate[i];
unsigned param_offset = ps->ps_input_param_offset[i];
-
- if (name == TGSI_SEMANTIC_POSITION ||
- name == TGSI_SEMANTIC_FACE)
- /* Read from preloaded VGPRs, not parameters */
- continue;
-
bcolor:
tmp = 0;
if (!ps)
return;
- input_ena = ps->spi_ps_input_ena;
+ input_ena = ps->config.spi_ps_input_ena;
/* we need to enable at least one of them, otherwise we hang the GPU */
assert(G_0286CC_PERSP_SAMPLE_ENA(input_ena) ||
return 0;
/* This shader doesn't need a scratch buffer */
- if (shader->scratch_bytes_per_wave == 0)
+ if (shader->config.scratch_bytes_per_wave == 0)
return 0;
/* This shader is already configured to use the current
static unsigned si_get_scratch_buffer_bytes_per_wave(struct si_shader *shader)
{
- return shader ? shader->scratch_bytes_per_wave : 0;
+ return shader ? shader->config.scratch_bytes_per_wave : 0;
}
static unsigned si_get_max_scratch_bytes_per_wave(struct si_context *sctx)
si_get_max_scratch_bytes_per_wave(sctx);
unsigned scratch_needed_size = scratch_bytes_per_wave *
sctx->scratch_waves;
+ unsigned spi_tmpring_size;
int r;
if (scratch_needed_size > 0) {
assert((scratch_needed_size & ~0x3FF) == scratch_needed_size &&
"scratch size should already be aligned correctly.");
- sctx->spi_tmpring_size = S_0286E8_WAVES(sctx->scratch_waves) |
- S_0286E8_WAVESIZE(scratch_bytes_per_wave >> 10);
+ spi_tmpring_size = S_0286E8_WAVES(sctx->scratch_waves) |
+ S_0286E8_WAVESIZE(scratch_bytes_per_wave >> 10);
+ if (spi_tmpring_size != sctx->spi_tmpring_size) {
+ sctx->spi_tmpring_size = spi_tmpring_size;
+ sctx->emit_scratch_reloc = true;
+ }
return true;
}
si_update_vgt_shader_config(sctx);
if (sctx->ps_shader.cso) {
+ unsigned db_shader_control =
+ sctx->ps_shader.cso->db_shader_control |
+ S_02880C_KILL_ENABLE(si_get_alpha_test_func(sctx) != PIPE_FUNC_ALWAYS);
+
r = si_shader_select(ctx, &sctx->ps_shader);
if (r)
return false;
si_mark_atom_dirty(sctx, &sctx->spi_ps_input);
}
- if (sctx->ps_db_shader_control != sctx->ps_shader.current->db_shader_control) {
- sctx->ps_db_shader_control = sctx->ps_shader.current->db_shader_control;
+ if (sctx->ps_db_shader_control != db_shader_control) {
+ sctx->ps_db_shader_control = db_shader_control;
si_mark_atom_dirty(sctx, &sctx->db_render_state);
}
#include "draw/draw_vertex.h"
#include "sp_quad_pipe.h"
+#include "sp_setup.h"
/** Do polygon stipple in the draw module? */
unsigned const_buffer_size[PIPE_SHADER_TYPES][PIPE_MAX_CONSTANT_BUFFERS];
/** Vertex format */
+ struct sp_setup_info setup_info;
struct vertex_info vertex_info;
- struct vertex_info vertex_info_vbuf;
/** Which vertex shader output slot contains point size */
- int psize_slot;
+ int8_t psize_slot;
/** Which vertex shader output slot contains viewport index */
- int viewport_index_slot;
+ int8_t viewport_index_slot;
/** Which vertex shader output slot contains layer */
- int layer_slot;
+ int8_t layer_slot;
/** The reduced version of the primitive supplied by the state tracker */
unsigned reduced_api_prim;
{
struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr);
struct softpipe_context *softpipe = cvbr->softpipe;
- const unsigned stride = softpipe->vertex_info_vbuf.size * sizeof(float);
+ const unsigned stride = softpipe->vertex_info.size * sizeof(float);
const void *vertex_buffer = cvbr->vertex_buffer;
struct setup_context *setup = cvbr->setup;
const boolean flatshade_first = softpipe->rasterizer->flatshade_first;
struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr);
struct softpipe_context *softpipe = cvbr->softpipe;
struct setup_context *setup = cvbr->setup;
- const unsigned stride = softpipe->vertex_info_vbuf.size * sizeof(float);
+ const unsigned stride = softpipe->vertex_info.size * sizeof(float);
const void *vertex_buffer =
(void *) get_vert(cvbr->vertex_buffer, start, stride);
const boolean flatshade_first = softpipe->rasterizer->flatshade_first;
break;
case PIPE_QUERY_PIPELINE_STATISTICS:
memcpy(vresult, &sq->stats,
- sizeof(struct pipe_query_data_pipeline_statistics));;
+ sizeof(struct pipe_query_data_pipeline_statistics));
break;
case PIPE_QUERY_GPU_FINISHED:
vresult->b = TRUE;
case PIPE_CAP_SHAREABLE_SHADERS:
case PIPE_CAP_COPY_BETWEEN_COMPRESSED_AND_PLAIN_FORMATS:
case PIPE_CAP_CLEAR_TEXTURE:
+ case PIPE_CAP_DRAW_PARAMETERS:
+ case PIPE_CAP_TGSI_PACK_HALF_FLOAT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT_PARAMS:
+ case PIPE_CAP_TGSI_FS_POSITION_IS_SYSVAL:
+ case PIPE_CAP_TGSI_FS_FACE_IS_INTEGER_SYSVAL:
+ case PIPE_CAP_SHADER_BUFFER_OFFSET_ALIGNMENT:
+ case PIPE_CAP_INVALIDATE_BUFFER:
+ case PIPE_CAP_GENERATE_MIPMAP:
return 0;
}
/* should only get here on unhandled cases */
#include "sp_setup.h"
#include "sp_state.h"
#include "draw/draw_context.h"
-#include "draw/draw_vertex.h"
#include "pipe/p_shader_tokens.h"
#include "util/u_math.h"
#include "util/u_memory.h"
{
struct softpipe_context *softpipe = setup->softpipe;
const struct tgsi_shader_info *fsInfo = &setup->softpipe->fs_variant->info;
- const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe);
+ const struct sp_setup_info *sinfo = &softpipe->setup_info;
uint fragSlot;
float v[3];
+ assert(sinfo->valid);
+
/* z and w are done by linear interpolation:
*/
v[0] = setup->vmin[0][2];
/* setup interpolation for all the remaining attributes:
*/
for (fragSlot = 0; fragSlot < fsInfo->num_inputs; fragSlot++) {
- const uint vertSlot = vinfo->attrib[fragSlot].src_index;
+ const uint vertSlot = sinfo->attrib[fragSlot].src_index;
uint j;
- switch (vinfo->attrib[fragSlot].interp_mode) {
- case INTERP_CONSTANT:
- for (j = 0; j < TGSI_NUM_CHANNELS; j++)
+ switch (sinfo->attrib[fragSlot].interp) {
+ case SP_INTERP_CONSTANT:
+ for (j = 0; j < TGSI_NUM_CHANNELS; j++) {
const_coeff(setup, &setup->coef[fragSlot], vertSlot, j);
+ }
break;
- case INTERP_LINEAR:
+ case SP_INTERP_LINEAR:
for (j = 0; j < TGSI_NUM_CHANNELS; j++) {
tri_apply_cylindrical_wrap(setup->vmin[vertSlot][j],
setup->vmid[vertSlot][j],
tri_linear_coeff(setup, &setup->coef[fragSlot], j, v);
}
break;
- case INTERP_PERSPECTIVE:
+ case SP_INTERP_PERSPECTIVE:
for (j = 0; j < TGSI_NUM_CHANNELS; j++) {
tri_apply_cylindrical_wrap(setup->vmin[vertSlot][j],
setup->vmid[vertSlot][j],
tri_persp_coeff(setup, &setup->coef[fragSlot], j, v);
}
break;
- case INTERP_POS:
+ case SP_INTERP_POS:
setup_fragcoord_coeff(setup, fragSlot);
break;
default:
{
struct softpipe_context *softpipe = setup->softpipe;
const struct tgsi_shader_info *fsInfo = &setup->softpipe->fs_variant->info;
- const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe);
+ const struct sp_setup_info *sinfo = &softpipe->setup_info;
uint fragSlot;
float area;
float v[2];
+ assert(sinfo->valid);
+
/* use setup->vmin, vmax to point to vertices */
if (softpipe->rasterizer->flatshade_first)
setup->vprovoke = v0;
/* setup interpolation for all the remaining attributes:
*/
for (fragSlot = 0; fragSlot < fsInfo->num_inputs; fragSlot++) {
- const uint vertSlot = vinfo->attrib[fragSlot].src_index;
+ const uint vertSlot = sinfo->attrib[fragSlot].src_index;
uint j;
- switch (vinfo->attrib[fragSlot].interp_mode) {
- case INTERP_CONSTANT:
+ switch (sinfo->attrib[fragSlot].interp) {
+ case SP_INTERP_CONSTANT:
for (j = 0; j < TGSI_NUM_CHANNELS; j++)
const_coeff(setup, &setup->coef[fragSlot], vertSlot, j);
break;
- case INTERP_LINEAR:
+ case SP_INTERP_LINEAR:
for (j = 0; j < TGSI_NUM_CHANNELS; j++) {
line_apply_cylindrical_wrap(setup->vmin[vertSlot][j],
setup->vmax[vertSlot][j],
line_linear_coeff(setup, &setup->coef[fragSlot], j, v);
}
break;
- case INTERP_PERSPECTIVE:
+ case SP_INTERP_PERSPECTIVE:
for (j = 0; j < TGSI_NUM_CHANNELS; j++) {
line_apply_cylindrical_wrap(setup->vmin[vertSlot][j],
setup->vmax[vertSlot][j],
line_persp_coeff(setup, &setup->coef[fragSlot], j, v);
}
break;
- case INTERP_POS:
+ case SP_INTERP_POS:
setup_fragcoord_coeff(setup, fragSlot);
break;
default:
const boolean round = (boolean) setup->softpipe->rasterizer->point_smooth;
const float x = v0[0][0]; /* Note: data[0] is always position */
const float y = v0[0][1];
- const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe);
+ const struct sp_setup_info *sinfo = &softpipe->setup_info;
uint fragSlot;
uint layer = 0;
unsigned viewport_index = 0;
print_vertex(setup, v0);
#endif
+ assert(sinfo->valid);
+
if (setup->softpipe->no_rast || setup->softpipe->rasterizer->rasterizer_discard)
return;
const_coeff(setup, &setup->posCoef, 0, 3);
for (fragSlot = 0; fragSlot < fsInfo->num_inputs; fragSlot++) {
- const uint vertSlot = vinfo->attrib[fragSlot].src_index;
+ const uint vertSlot = sinfo->attrib[fragSlot].src_index;
uint j;
- switch (vinfo->attrib[fragSlot].interp_mode) {
- case INTERP_CONSTANT:
+ switch (sinfo->attrib[fragSlot].interp) {
+ case SP_INTERP_CONSTANT:
/* fall-through */
- case INTERP_LINEAR:
+ case SP_INTERP_LINEAR:
for (j = 0; j < TGSI_NUM_CHANNELS; j++)
const_coeff(setup, &setup->coef[fragSlot], vertSlot, j);
break;
- case INTERP_PERSPECTIVE:
+ case SP_INTERP_PERSPECTIVE:
for (j = 0; j < TGSI_NUM_CHANNELS; j++)
point_persp_coeff(setup, setup->vprovoke,
&setup->coef[fragSlot], vertSlot, j);
break;
- case INTERP_POS:
+ case SP_INTERP_POS:
setup_fragcoord_coeff(setup, fragSlot);
break;
default:
struct setup_context;
struct softpipe_context;
+/**
+ * Attribute interpolation mode
+ */
+enum sp_interp_mode {
+ SP_INTERP_POS, /**< special case for frag position */
+ SP_INTERP_CONSTANT,
+ SP_INTERP_LINEAR,
+ SP_INTERP_PERSPECTIVE
+};
+
+
+struct sp_setup_info {
+ unsigned valid;
+ struct {
+ unsigned interp:8; /**< SP_INTERP_X */
+ int src_index:8;
+ } attrib[PIPE_MAX_SHADER_OUTPUTS];
+};
+
void
-sp_setup_tri( struct setup_context *setup,
- const float (*v0)[4],
- const float (*v1)[4],
- const float (*v2)[4] );
+sp_setup_tri(struct setup_context *setup,
+ const float (*v0)[4],
+ const float (*v1)[4],
+ const float (*v2)[4]);
void
sp_setup_line(struct setup_context *setup,
struct vertex_info *
-softpipe_get_vertex_info(struct softpipe_context *softpipe);
-
-struct vertex_info *
softpipe_get_vbuf_vertex_info(struct softpipe_context *softpipe);
static void
invalidate_vertex_layout(struct softpipe_context *softpipe)
{
- softpipe->vertex_info.num_attribs = 0;
+ softpipe->setup_info.valid = 0;
}
* (simple float[][4]) used by the 'draw' module into vertices for
* rasterization.
*
- * This function validates the vertex layout and returns a pointer to a
- * vertex_info object.
+ * This function validates the vertex layout.
*/
-struct vertex_info *
-softpipe_get_vertex_info(struct softpipe_context *softpipe)
+static void
+softpipe_compute_vertex_info(struct softpipe_context *softpipe)
{
- struct vertex_info *vinfo = &softpipe->vertex_info;
- int vs_index;
+ struct sp_setup_info *sinfo = &softpipe->setup_info;
- if (vinfo->num_attribs == 0) {
- /* compute vertex layout now */
+ if (sinfo->valid == 0) {
const struct tgsi_shader_info *fsInfo = &softpipe->fs_variant->info;
- struct vertex_info *vinfo_vbuf = &softpipe->vertex_info_vbuf;
- const uint num = draw_num_shader_outputs(softpipe->draw);
+ struct vertex_info *vinfo = &softpipe->vertex_info;
uint i;
-
- /* Tell draw_vbuf to simply emit the whole post-xform vertex
- * as-is. No longer any need to try and emit draw vertex_header
- * info.
+ int vs_index;
+ /*
+ * This doesn't quite work right (wrt face injection, prim id,
+ * wide points) - hit a couple assertions, misrenderings plus
+ * memory corruption. Albeit could fix (the former two) by calling
+ * this "more often" (rasterizer changes etc.). (The latter would
+ * need to be included in draw_prepare_shader_outputs, but it looks
+ * like that would potentially allocate quite some unused additional
+ * vertex outputs.)
+ * draw_prepare_shader_outputs(softpipe->draw);
*/
- vinfo_vbuf->num_attribs = 0;
- for (i = 0; i < num; i++) {
- draw_emit_vertex_attr(vinfo_vbuf, EMIT_4F, INTERP_PERSPECTIVE, i);
- }
- draw_compute_vertex_size(vinfo_vbuf);
/*
- * Loop over fragment shader inputs, searching for the matching output
- * from the vertex shader.
+ * Those can't actually be 0 (because pos is always at 0).
+ * But use ints anyway to avoid confusion (in vs outputs, they
+ * can very well be at pos 0).
*/
+ softpipe->viewport_index_slot = -1;
+ softpipe->layer_slot = -1;
+ softpipe->psize_slot = -1;
+
vinfo->num_attribs = 0;
+
+ /*
+ * Put position always first (setup needs it there).
+ */
+ vs_index = draw_find_shader_output(softpipe->draw,
+ TGSI_SEMANTIC_POSITION, 0);
+
+ draw_emit_vertex_attr(vinfo, EMIT_4F, vs_index);
+
+ /*
+ * Match FS inputs against VS outputs, emitting the necessary
+ * attributes.
+ */
for (i = 0; i < fsInfo->num_inputs; i++) {
- int src;
- enum interp_mode interp = INTERP_LINEAR;
+ enum sp_interp_mode interp = SP_INTERP_LINEAR;
switch (fsInfo->input_interpolate[i]) {
case TGSI_INTERPOLATE_CONSTANT:
- interp = INTERP_CONSTANT;
+ interp = SP_INTERP_CONSTANT;
break;
case TGSI_INTERPOLATE_LINEAR:
- interp = INTERP_LINEAR;
+ interp = SP_INTERP_LINEAR;
break;
case TGSI_INTERPOLATE_PERSPECTIVE:
- interp = INTERP_PERSPECTIVE;
+ interp = SP_INTERP_PERSPECTIVE;
break;
case TGSI_INTERPOLATE_COLOR:
assert(fsInfo->input_semantic_name[i] == TGSI_SEMANTIC_COLOR);
switch (fsInfo->input_semantic_name[i]) {
case TGSI_SEMANTIC_POSITION:
- interp = INTERP_POS;
+ interp = SP_INTERP_POS;
break;
case TGSI_SEMANTIC_COLOR:
if (fsInfo->input_interpolate[i] == TGSI_INTERPOLATE_COLOR) {
if (softpipe->rasterizer->flatshade)
- interp = INTERP_CONSTANT;
+ interp = SP_INTERP_CONSTANT;
else
- interp = INTERP_PERSPECTIVE;
+ interp = SP_INTERP_PERSPECTIVE;
}
break;
}
- /* this includes texcoords and varying vars */
- src = draw_find_shader_output(softpipe->draw,
- fsInfo->input_semantic_name[i],
- fsInfo->input_semantic_index[i]);
- if (fsInfo->input_semantic_name[i] == TGSI_SEMANTIC_COLOR && src == -1)
- /* try and find a bcolor */
- src = draw_find_shader_output(softpipe->draw,
- TGSI_SEMANTIC_BCOLOR, fsInfo->input_semantic_index[i]);
+ /*
+ * Search for each input in current vs output:
+ */
+ vs_index = draw_find_shader_output(softpipe->draw,
+ fsInfo->input_semantic_name[i],
+ fsInfo->input_semantic_index[i]);
+
+ if (fsInfo->input_semantic_name[i] == TGSI_SEMANTIC_COLOR &&
+ vs_index == -1) {
+ /*
+ * try and find a bcolor.
+ * Note that if there's both front and back color, draw will
+ * have copied back to front color already.
+ */
+ vs_index = draw_find_shader_output(softpipe->draw,
+ TGSI_SEMANTIC_BCOLOR,
+ fsInfo->input_semantic_index[i]);
+ }
- draw_emit_vertex_attr(vinfo, EMIT_4F, interp, src);
+ sinfo->attrib[i].interp = interp;
+ /* extremely pointless index map */
+ sinfo->attrib[i].src_index = i + 1;
+ /*
+ * For vp index and layer, if the fs requires them but the vs doesn't
+ * provide them, draw (vbuf) will give us the required 0 (slot -1).
+ * (This means in this case we'll also use those slots in setup, which
+ * isn't necessary but they'll contain the correct (0) value.)
+ */
+ if (fsInfo->input_semantic_name[i] ==
+ TGSI_SEMANTIC_VIEWPORT_INDEX) {
+ softpipe->viewport_index_slot = (int)vinfo->num_attribs;
+ draw_emit_vertex_attr(vinfo, EMIT_4F, vs_index);
+ } else if (fsInfo->input_semantic_name[i] == TGSI_SEMANTIC_LAYER) {
+ softpipe->layer_slot = (int)vinfo->num_attribs;
+ draw_emit_vertex_attr(vinfo, EMIT_4F, vs_index);
+ /*
+ * Note that we'd actually want to skip position (as we won't use
+ * the attribute in the fs) but can't. The reason is that we don't
+ * actually have a input/output map for setup (even though it looks
+ * like we do...). Could adjust for this though even without a map.
+ */
+ } else {
+ /*
+ * Note that we'd actually want to skip position (as we won't use
+ * the attribute in the fs) but can't. The reason is that we don't
+ * actually have a input/output map for setup (even though it looks
+ * like we do...). Could adjust for this though even without a map.
+ */
+ draw_emit_vertex_attr(vinfo, EMIT_4F, vs_index);
+ }
}
- /* Figure out if we need pointsize as well. */
+ /* Figure out if we need pointsize as well.
+ */
vs_index = draw_find_shader_output(softpipe->draw,
TGSI_SEMANTIC_PSIZE, 0);
if (vs_index >= 0) {
- softpipe->psize_slot = vinfo->num_attribs;
- draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_CONSTANT, vs_index);
+ softpipe->psize_slot = (int)vinfo->num_attribs;
+ draw_emit_vertex_attr(vinfo, EMIT_4F, vs_index);
}
- /* Figure out if we need viewport index */
- vs_index = draw_find_shader_output(softpipe->draw,
- TGSI_SEMANTIC_VIEWPORT_INDEX,
- 0);
- if (vs_index >= 0) {
- softpipe->viewport_index_slot = vinfo->num_attribs;
- draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_CONSTANT, vs_index);
- } else {
- softpipe->viewport_index_slot = 0;
+ /* Figure out if we need viewport index (if it wasn't already in fs input) */
+ if (softpipe->viewport_index_slot < 0) {
+ vs_index = draw_find_shader_output(softpipe->draw,
+ TGSI_SEMANTIC_VIEWPORT_INDEX,
+ 0);
+ if (vs_index >= 0) {
+ softpipe->viewport_index_slot =(int)vinfo->num_attribs;
+ draw_emit_vertex_attr(vinfo, EMIT_4F, vs_index);
+ }
}
- /* Figure out if we need layer */
- vs_index = draw_find_shader_output(softpipe->draw,
- TGSI_SEMANTIC_LAYER,
- 0);
- if (vs_index >= 0) {
- softpipe->layer_slot = vinfo->num_attribs;
- draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_CONSTANT, vs_index);
- } else {
- softpipe->layer_slot = 0;
+ /* Figure out if we need layer (if it wasn't already in fs input) */
+ if (softpipe->layer_slot < 0) {
+ vs_index = draw_find_shader_output(softpipe->draw,
+ TGSI_SEMANTIC_LAYER,
+ 0);
+ if (vs_index >= 0) {
+ softpipe->layer_slot = (int)vinfo->num_attribs;
+ draw_emit_vertex_attr(vinfo, EMIT_4F, vs_index);
+ }
}
draw_compute_vertex_size(vinfo);
+ softpipe->setup_info.valid = 1;
}
-
- return vinfo;
+ return;
}
/**
* Called from vbuf module.
*
- * Note that there's actually two different vertex layouts in softpipe.
- *
- * The normal one is computed in softpipe_get_vertex_info() above and is
- * used by the point/line/tri "setup" code.
- *
- * The other one (this one) is only used by the vbuf module (which is
- * not normally used by default but used in testing). For the vbuf module,
- * we basically want to pass-through the draw module's vertex layout as-is.
- * When the softpipe vbuf code begins drawing, the normal vertex layout
- * will come into play again.
+ * This will trigger validation of the vertex layout (and also compute
+ * the required information for setup).
*/
struct vertex_info *
softpipe_get_vbuf_vertex_info(struct softpipe_context *softpipe)
{
- (void) softpipe_get_vertex_info(softpipe);
- return &softpipe->vertex_info_vbuf;
+ softpipe_compute_vertex_info(softpipe);
+ return &softpipe->vertex_info;
}
/* get new shader that implements polygon stippling */
var->tokens =
util_pstipple_create_fragment_shader(curfs->tokens,
- &var->stipple_sampler_unit, 0);
+ &var->stipple_sampler_unit, 0,
+ TGSI_FILE_INPUT);
}
else
#endif
mipSizes[0].height = height;
mipSizes[0].depth = 1;
- swc->commit(swc);;
+ swc->commit(swc);
return PIPE_OK;
}
swc->surface_relocation(swc, &cmd->sid, NULL, sid,
SVGA_RELOC_WRITE | SVGA_RELOC_INTERNAL);
- swc->commit(swc);;
+ swc->commit(swc);
return PIPE_OK;
}
pSuffix->flags = flags;
swc->commit(swc);
+ swc->hints |= SVGA_HINT_FLAG_CAN_PRE_FLUSH;
return PIPE_OK;
}
pSuffix->flags = flags;
swc->commit(swc);
+ swc->hints |= SVGA_HINT_FLAG_CAN_PRE_FLUSH;
return PIPE_OK;
}
*decls = declArray;
*ranges = rangeArray;
- swc->hints |= SVGA_HINT_FLAG_DRAW_EMITTED;
+ swc->hints |= SVGA_HINT_FLAG_CAN_PRE_FLUSH;
return PIPE_OK;
}
cmd->box = *box;
swc->commit(swc);
+ swc->hints |= SVGA_HINT_FLAG_CAN_PRE_FLUSH;
return PIPE_OK;
}
SVGA_RELOC_WRITE | SVGA_RELOC_INTERNAL);
swc->commit(swc);
+ swc->hints |= SVGA_HINT_FLAG_CAN_PRE_FLUSH;
return PIPE_OK;
}
cmd->image.mipmap = mipLevel;
swc->commit(swc);
+ swc->hints |= SVGA_HINT_FLAG_CAN_PRE_FLUSH;
return PIPE_OK;
}
SVGA_RELOC_READ | SVGA_RELOC_INTERNAL);
swc->commit(swc);
+ swc->hints |= SVGA_HINT_FLAG_CAN_PRE_FLUSH;
return PIPE_OK;
}
cmd->invertBox = invertBox;
swc->commit(swc);
+ swc->hints |= SVGA_HINT_FLAG_CAN_PRE_FLUSH;
return PIPE_OK;
}
const SVGA3dBox *box,
unsigned subResource);
+enum pipe_error
+SVGA3D_vgpu10_GenMips(struct svga_winsys_context *swc,
+ const SVGA3dShaderResourceViewId shaderResourceViewId,
+ struct svga_winsys_surface *view);
#endif /* __SVGA3D_H__ */
SVGA3D_COPY_BASIC_2(vertexCount, startVertexLocation);
- swc->hints |= SVGA_HINT_FLAG_DRAW_EMITTED;
+ swc->hints |= SVGA_HINT_FLAG_CAN_PRE_FLUSH;
swc->commit(swc);
return PIPE_OK;
}
SVGA3D_COPY_BASIC_3(indexCount, startIndexLocation,
baseVertexLocation);
- swc->hints |= SVGA_HINT_FLAG_DRAW_EMITTED;
+ swc->hints |= SVGA_HINT_FLAG_CAN_PRE_FLUSH;
swc->commit(swc);
return PIPE_OK;
}
SVGA3D_COPY_BASIC_4(vertexCountPerInstance, instanceCount,
startVertexLocation, startInstanceLocation);
- swc->hints |= SVGA_HINT_FLAG_DRAW_EMITTED;
+ swc->hints |= SVGA_HINT_FLAG_CAN_PRE_FLUSH;
swc->commit(swc);
return PIPE_OK;
}
startInstanceLocation);
- swc->hints |= SVGA_HINT_FLAG_DRAW_EMITTED;
+ swc->hints |= SVGA_HINT_FLAG_CAN_PRE_FLUSH;
swc->commit(swc);
return PIPE_OK;
}
{
SVGA3D_CREATE_COMMAND(DrawAuto, DRAW_AUTO);
- swc->hints |= SVGA_HINT_FLAG_DRAW_EMITTED;
+ swc->hints |= SVGA_HINT_FLAG_CAN_PRE_FLUSH;
swc->commit(swc);
return PIPE_OK;
}
swc->commit(swc);
return PIPE_OK;
}
+
+enum pipe_error
+SVGA3D_vgpu10_GenMips(struct svga_winsys_context *swc,
+ SVGA3dShaderResourceViewId shaderResourceViewId,
+ struct svga_winsys_surface *view)
+{
+ SVGA3dCmdDXGenMips *cmd;
+
+ cmd = SVGA3D_FIFOReserve(swc, SVGA_3D_CMD_DX_GENMIPS,
+ sizeof(SVGA3dCmdDXGenMips), 1);
+
+ if (!cmd)
+ return PIPE_ERROR_OUT_OF_MEMORY;
+
+ swc->surface_relocation(swc, &cmd->shaderResourceViewId, NULL, view,
+ SVGA_RELOC_WRITE);
+ cmd->shaderResourceViewId = shaderResourceViewId;
+
+ swc->commit(swc);
+ return PIPE_OK;
+}
#include "svga_winsys.h"
#define CONST0_UPLOAD_DEFAULT_SIZE 65536
-#define CONST0_UPLOAD_ALIGNMENT 256
DEBUG_GET_ONCE_BOOL_OPTION(no_swtnl, "SVGA_NO_SWTNL", FALSE)
DEBUG_GET_ONCE_BOOL_OPTION(force_swtnl, "SVGA_FORCE_SWTNL", FALSE);
svga->const0_upload = u_upload_create(&svga->pipe,
CONST0_UPLOAD_DEFAULT_SIZE,
- CONST0_UPLOAD_ALIGNMENT,
- PIPE_BIND_CONSTANT_BUFFER);
+ PIPE_BIND_CONSTANT_BUFFER,
+ PIPE_USAGE_STREAM);
if (!svga->const0_upload)
goto cleanup;
#define SVGA_QUERY_NUM_RESOURCES (PIPE_QUERY_DRIVER_SPECIFIC + 9)
#define SVGA_QUERY_NUM_STATE_OBJECTS (PIPE_QUERY_DRIVER_SPECIFIC + 10)
#define SVGA_QUERY_NUM_SURFACE_VIEWS (PIPE_QUERY_DRIVER_SPECIFIC + 11)
+#define SVGA_QUERY_NUM_GENERATE_MIPMAP (PIPE_QUERY_DRIVER_SPECIFIC + 12)
/*SVGA_QUERY_MAX has to be last because it is size of an array*/
-#define SVGA_QUERY_MAX (PIPE_QUERY_DRIVER_SPECIFIC + 12)
+#define SVGA_QUERY_MAX (PIPE_QUERY_DRIVER_SPECIFIC + 13)
/**
* Maximum supported number of constant buffers per shader
*/
#define SVGA_MAX_CONST_BUF_SIZE (4096 * 4 * sizeof(int))
+#define CONST0_UPLOAD_ALIGNMENT 256
+
struct draw_vertex_shader;
struct draw_fragment_shader;
struct svga_shader_variant;
struct svga_sampler_view *v;
unsigned min_lod;
unsigned max_lod;
- int dirty;
+ boolean dirty;
};
/* Updated by calling svga_update_state( SVGA_STATE_HW_DRAW )
SVGA3dElementLayoutId layout_id;
SVGA3dPrimitiveType topology;
+ /** Vertex buffer state */
+ SVGA3dVertexBuffer vbuffers[PIPE_MAX_ATTRIBS];
+ struct svga_winsys_surface *vbuffer_handles[PIPE_MAX_ATTRIBS];
+ unsigned num_vbuffers;
+
struct svga_winsys_surface *ib; /**< index buffer for drawing */
SVGA3dSurfaceFormat ib_format;
unsigned ib_offset;
uint64_t num_state_objects; /**< SVGA_QUERY_NUM_STATE_OBJECTS */
uint64_t num_surface_views; /**< SVGA_QUERY_NUM_SURFACE_VIEWS */
uint64_t num_bytes_uploaded; /**< SVGA_QUERY_NUM_BYTES_UPLOADED */
+ uint64_t num_generate_mipmap; /**< SVGA_QUERY_NUM_GENERATE_MIPMAP */
} hud;
/** The currently bound stream output targets */
buffers[i].offset = hwtnl->cmd.vbufs[i].buffer_offset;
}
if (vbuf_count > 0) {
- ret = SVGA3D_vgpu10_SetVertexBuffers(svga->swc, vbuf_count,
- 0, /* startBuffer */
- buffers, vb_handle);
- if (ret != PIPE_OK)
- return ret;
+ /* If we haven't yet emitted a drawing command or if any
+ * vertex buffer state is changing, issue that state now.
+ */
+ if (((hwtnl->cmd.swc->hints & SVGA_HINT_FLAG_CAN_PRE_FLUSH) == 0) ||
+ vbuf_count != svga->state.hw_draw.num_vbuffers ||
+ memcmp(buffers, svga->state.hw_draw.vbuffers,
+ vbuf_count * sizeof(buffers[0])) ||
+ memcmp(vb_handle, svga->state.hw_draw.vbuffer_handles,
+ vbuf_count * sizeof(vb_handle[0]))) {
+ ret = SVGA3D_vgpu10_SetVertexBuffers(svga->swc, vbuf_count,
+ 0, /* startBuffer */
+ buffers, vb_handle);
+ if (ret != PIPE_OK)
+ return ret;
+
+ svga->state.hw_draw.num_vbuffers = vbuf_count;
+ memcpy(svga->state.hw_draw.vbuffers, buffers,
+ vbuf_count * sizeof(buffers[0]));
+ memcpy(svga->state.hw_draw.vbuffer_handles, vb_handle,
+ vbuf_count * sizeof(vb_handle[0]));
+ }
}
}
{
/* Gallium format SVGA3D vertex format SVGA3D pixel format Flags */
{ PIPE_FORMAT_NONE, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
- { PIPE_FORMAT_B8G8R8A8_UNORM, SVGA3D_B8G8R8A8_UNORM, SVGA3D_B8G8R8A8_UNORM, 0 },
- { PIPE_FORMAT_B8G8R8X8_UNORM, SVGA3D_FORMAT_INVALID, SVGA3D_B8G8R8X8_UNORM, 0 },
+ { PIPE_FORMAT_B8G8R8A8_UNORM, SVGA3D_B8G8R8A8_UNORM, SVGA3D_B8G8R8A8_UNORM, TF_GEN_MIPS },
+ { PIPE_FORMAT_B8G8R8X8_UNORM, SVGA3D_FORMAT_INVALID, SVGA3D_B8G8R8X8_UNORM, TF_GEN_MIPS },
{ PIPE_FORMAT_A8R8G8B8_UNORM, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
{ PIPE_FORMAT_X8R8G8B8_UNORM, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
- { PIPE_FORMAT_B5G5R5A1_UNORM, SVGA3D_FORMAT_INVALID, SVGA3D_B5G5R5A1_UNORM, 0 },
+ { PIPE_FORMAT_B5G5R5A1_UNORM, SVGA3D_FORMAT_INVALID, SVGA3D_B5G5R5A1_UNORM, TF_GEN_MIPS },
{ PIPE_FORMAT_B4G4R4A4_UNORM, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
- { PIPE_FORMAT_B5G6R5_UNORM, SVGA3D_FORMAT_INVALID, SVGA3D_B5G6R5_UNORM, 0 },
- { PIPE_FORMAT_R10G10B10A2_UNORM, SVGA3D_R10G10B10A2_UNORM, SVGA3D_R10G10B10A2_UNORM, 0 },
+ { PIPE_FORMAT_B5G6R5_UNORM, SVGA3D_FORMAT_INVALID, SVGA3D_B5G6R5_UNORM, TF_GEN_MIPS },
+ { PIPE_FORMAT_R10G10B10A2_UNORM, SVGA3D_R10G10B10A2_UNORM, SVGA3D_R10G10B10A2_UNORM, TF_GEN_MIPS },
{ PIPE_FORMAT_L8_UNORM, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
- { PIPE_FORMAT_A8_UNORM, SVGA3D_FORMAT_INVALID, SVGA3D_A8_UNORM, 0 },
+ { PIPE_FORMAT_A8_UNORM, SVGA3D_FORMAT_INVALID, SVGA3D_A8_UNORM, TF_GEN_MIPS },
{ PIPE_FORMAT_I8_UNORM, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
{ PIPE_FORMAT_L8A8_UNORM, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
{ PIPE_FORMAT_L16_UNORM, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
{ PIPE_FORMAT_R64G64_FLOAT, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
{ PIPE_FORMAT_R64G64B64_FLOAT, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
{ PIPE_FORMAT_R64G64B64A64_FLOAT, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
- { PIPE_FORMAT_R32_FLOAT, SVGA3D_R32_FLOAT, SVGA3D_R32_FLOAT, 0 },
- { PIPE_FORMAT_R32G32_FLOAT, SVGA3D_R32G32_FLOAT, SVGA3D_R32G32_FLOAT, 0 },
- { PIPE_FORMAT_R32G32B32_FLOAT, SVGA3D_R32G32B32_FLOAT, SVGA3D_R32G32B32_FLOAT, 0 },
- { PIPE_FORMAT_R32G32B32A32_FLOAT, SVGA3D_R32G32B32A32_FLOAT, SVGA3D_R32G32B32A32_FLOAT, 0 },
+ { PIPE_FORMAT_R32_FLOAT, SVGA3D_R32_FLOAT, SVGA3D_R32_FLOAT, TF_GEN_MIPS },
+ { PIPE_FORMAT_R32G32_FLOAT, SVGA3D_R32G32_FLOAT, SVGA3D_R32G32_FLOAT, TF_GEN_MIPS },
+ { PIPE_FORMAT_R32G32B32_FLOAT, SVGA3D_R32G32B32_FLOAT, SVGA3D_R32G32B32_FLOAT, TF_GEN_MIPS },
+ { PIPE_FORMAT_R32G32B32A32_FLOAT, SVGA3D_R32G32B32A32_FLOAT, SVGA3D_R32G32B32A32_FLOAT, TF_GEN_MIPS },
{ PIPE_FORMAT_R32_UNORM, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
{ PIPE_FORMAT_R32G32_UNORM, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
{ PIPE_FORMAT_R32G32B32_UNORM, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
{ PIPE_FORMAT_R32G32_SSCALED, SVGA3D_R32G32_SINT, SVGA3D_FORMAT_INVALID, VF_I_TO_F_CAST },
{ PIPE_FORMAT_R32G32B32_SSCALED, SVGA3D_R32G32B32_SINT, SVGA3D_FORMAT_INVALID, VF_I_TO_F_CAST },
{ PIPE_FORMAT_R32G32B32A32_SSCALED, SVGA3D_R32G32B32A32_SINT, SVGA3D_FORMAT_INVALID, VF_I_TO_F_CAST },
- { PIPE_FORMAT_R16_UNORM, SVGA3D_R16_UNORM, SVGA3D_R16_UNORM, 0 },
- { PIPE_FORMAT_R16G16_UNORM, SVGA3D_R16G16_UNORM, SVGA3D_R16G16_UNORM, 0 },
+ { PIPE_FORMAT_R16_UNORM, SVGA3D_R16_UNORM, SVGA3D_R16_UNORM, TF_GEN_MIPS },
+ { PIPE_FORMAT_R16G16_UNORM, SVGA3D_R16G16_UNORM, SVGA3D_R16G16_UNORM, TF_GEN_MIPS },
{ PIPE_FORMAT_R16G16B16_UNORM, SVGA3D_R16G16B16A16_UNORM, SVGA3D_FORMAT_INVALID, VF_W_TO_1 },
- { PIPE_FORMAT_R16G16B16A16_UNORM, SVGA3D_R16G16B16A16_UNORM, SVGA3D_R16G16B16A16_UNORM, 0 },
+ { PIPE_FORMAT_R16G16B16A16_UNORM, SVGA3D_R16G16B16A16_UNORM, SVGA3D_R16G16B16A16_UNORM, TF_GEN_MIPS },
{ PIPE_FORMAT_R16_USCALED, SVGA3D_R16_UINT, SVGA3D_FORMAT_INVALID, VF_U_TO_F_CAST },
{ PIPE_FORMAT_R16G16_USCALED, SVGA3D_R16G16_UINT, SVGA3D_FORMAT_INVALID, VF_U_TO_F_CAST },
{ PIPE_FORMAT_R16G16B16_USCALED, SVGA3D_R16G16B16A16_UINT, SVGA3D_FORMAT_INVALID, VF_W_TO_1 | VF_U_TO_F_CAST },
{ PIPE_FORMAT_R16G16_SSCALED, SVGA3D_R16G16_SINT, SVGA3D_FORMAT_INVALID, VF_I_TO_F_CAST },
{ PIPE_FORMAT_R16G16B16_SSCALED, SVGA3D_R16G16B16A16_SINT, SVGA3D_FORMAT_INVALID, VF_W_TO_1 | VF_I_TO_F_CAST },
{ PIPE_FORMAT_R16G16B16A16_SSCALED, SVGA3D_R16G16B16A16_SINT, SVGA3D_FORMAT_INVALID, VF_I_TO_F_CAST },
- { PIPE_FORMAT_R8_UNORM, SVGA3D_R8_UNORM, SVGA3D_R8_UNORM, 0 },
- { PIPE_FORMAT_R8G8_UNORM, SVGA3D_R8G8_UNORM, SVGA3D_R8G8_UNORM, 0 },
+ { PIPE_FORMAT_R8_UNORM, SVGA3D_R8_UNORM, SVGA3D_R8_UNORM, TF_GEN_MIPS },
+ { PIPE_FORMAT_R8G8_UNORM, SVGA3D_R8G8_UNORM, SVGA3D_R8G8_UNORM, TF_GEN_MIPS },
{ PIPE_FORMAT_R8G8B8_UNORM, SVGA3D_R8G8B8A8_UNORM, SVGA3D_FORMAT_INVALID, VF_W_TO_1 },
- { PIPE_FORMAT_R8G8B8A8_UNORM, SVGA3D_R8G8B8A8_UNORM, SVGA3D_R8G8B8A8_UNORM, 0 },
+ { PIPE_FORMAT_R8G8B8A8_UNORM, SVGA3D_R8G8B8A8_UNORM, SVGA3D_R8G8B8A8_UNORM, TF_GEN_MIPS },
{ PIPE_FORMAT_X8B8G8R8_UNORM, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
{ PIPE_FORMAT_R8_USCALED, SVGA3D_R8_UINT, SVGA3D_FORMAT_INVALID, VF_U_TO_F_CAST },
{ PIPE_FORMAT_R8G8_USCALED, SVGA3D_R8G8_UINT, SVGA3D_FORMAT_INVALID, VF_U_TO_F_CAST },
{ PIPE_FORMAT_R32G32_FIXED, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
{ PIPE_FORMAT_R32G32B32_FIXED, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
{ PIPE_FORMAT_R32G32B32A32_FIXED, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
- { PIPE_FORMAT_R16_FLOAT, SVGA3D_R16_FLOAT, SVGA3D_R16_FLOAT, 0 },
- { PIPE_FORMAT_R16G16_FLOAT, SVGA3D_R16G16_FLOAT, SVGA3D_R16G16_FLOAT, 0 },
+ { PIPE_FORMAT_R16_FLOAT, SVGA3D_R16_FLOAT, SVGA3D_R16_FLOAT, TF_GEN_MIPS },
+ { PIPE_FORMAT_R16G16_FLOAT, SVGA3D_R16G16_FLOAT, SVGA3D_R16G16_FLOAT, TF_GEN_MIPS },
{ PIPE_FORMAT_R16G16B16_FLOAT, SVGA3D_R16G16B16A16_FLOAT, SVGA3D_FORMAT_INVALID, VF_W_TO_1 },
- { PIPE_FORMAT_R16G16B16A16_FLOAT, SVGA3D_R16G16B16A16_FLOAT, SVGA3D_R16G16B16A16_FLOAT, 0 },
+ { PIPE_FORMAT_R16G16B16A16_FLOAT, SVGA3D_R16G16B16A16_FLOAT, SVGA3D_R16G16B16A16_FLOAT, TF_GEN_MIPS },
{ PIPE_FORMAT_L8_SRGB, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
{ PIPE_FORMAT_L8A8_SRGB, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
{ PIPE_FORMAT_R8G8B8_SRGB, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
{ PIPE_FORMAT_A8B8G8R8_SRGB, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
{ PIPE_FORMAT_X8B8G8R8_SRGB, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
- { PIPE_FORMAT_B8G8R8A8_SRGB, SVGA3D_FORMAT_INVALID, SVGA3D_B8G8R8A8_UNORM_SRGB, 0 },
- { PIPE_FORMAT_B8G8R8X8_SRGB, SVGA3D_FORMAT_INVALID, SVGA3D_B8G8R8X8_UNORM_SRGB, 0 },
+ { PIPE_FORMAT_B8G8R8A8_SRGB, SVGA3D_FORMAT_INVALID, SVGA3D_B8G8R8A8_UNORM_SRGB, TF_GEN_MIPS },
+ { PIPE_FORMAT_B8G8R8X8_SRGB, SVGA3D_FORMAT_INVALID, SVGA3D_B8G8R8X8_UNORM_SRGB, TF_GEN_MIPS },
{ PIPE_FORMAT_A8R8G8B8_SRGB, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
{ PIPE_FORMAT_X8R8G8B8_SRGB, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
- { PIPE_FORMAT_R8G8B8A8_SRGB, SVGA3D_FORMAT_INVALID, SVGA3D_R8G8B8A8_UNORM_SRGB, 0 },
+ { PIPE_FORMAT_R8G8B8A8_SRGB, SVGA3D_FORMAT_INVALID, SVGA3D_R8G8B8A8_UNORM_SRGB, TF_GEN_MIPS },
{ PIPE_FORMAT_DXT1_RGB, SVGA3D_FORMAT_INVALID, SVGA3D_BC1_UNORM, 0 },
{ PIPE_FORMAT_DXT1_RGBA, SVGA3D_FORMAT_INVALID, SVGA3D_BC1_UNORM, 0 },
{ PIPE_FORMAT_DXT3_RGBA, SVGA3D_FORMAT_INVALID, SVGA3D_BC2_UNORM, 0 },
{ PIPE_FORMAT_A8B8G8R8_UNORM, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
{ PIPE_FORMAT_B5G5R5X1_UNORM, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
{ PIPE_FORMAT_R10G10B10A2_USCALED, SVGA3D_R10G10B10A2_UNORM, SVGA3D_FORMAT_INVALID, VF_PUINT_TO_USCALED },
- { PIPE_FORMAT_R11G11B10_FLOAT, SVGA3D_FORMAT_INVALID, SVGA3D_R11G11B10_FLOAT, 0 },
+ { PIPE_FORMAT_R11G11B10_FLOAT, SVGA3D_FORMAT_INVALID, SVGA3D_R11G11B10_FLOAT, TF_GEN_MIPS },
{ PIPE_FORMAT_R9G9B9E5_FLOAT, SVGA3D_FORMAT_INVALID, SVGA3D_R9G9B9E5_SHAREDEXP, 0 },
{ PIPE_FORMAT_Z32_FLOAT_S8X24_UINT, SVGA3D_FORMAT_INVALID, SVGA3D_D32_FLOAT_S8X24_UINT, 0 },
{ PIPE_FORMAT_R1_UNORM, SVGA3D_FORMAT_INVALID, SVGA3D_FORMAT_INVALID, 0 },
}
}
+boolean
+svga_format_support_gen_mips(enum pipe_format format)
+{
+ assert(format < Elements(format_conversion_table));
+ return ((format_conversion_table[format].flags & TF_GEN_MIPS) > 0);
+}
+
/**
* Given a texture format, return the expected data type returned from
#define VF_PUINT_TO_USCALED (1 << 6) /* 10_10_10_2 to uscaled */
#define VF_PUINT_TO_SSCALED (1 << 7) /* 10_10_10_2 to sscaled */
+/**
+ * Texture format flags.
+ */
+#define TF_GEN_MIPS (1 << 8) /* supports hw generate mipmap */
void
svga_translate_vertex_format_vgpu10(enum pipe_format format,
boolean
svga_format_is_integer(SVGA3dSurfaceFormat format);
+boolean
+svga_format_support_gen_mips(enum pipe_format format);
+
enum tgsi_return_type
svga_get_texture_datatype(enum pipe_format format);
case SVGA_QUERY_NUM_SURFACE_VIEWS:
case SVGA_QUERY_NUM_RESOURCES_MAPPED:
case SVGA_QUERY_NUM_BYTES_UPLOADED:
+ case SVGA_QUERY_NUM_GENERATE_MIPMAP:
break;
default:
assert(!"unexpected query type in svga_create_query()");
case SVGA_QUERY_NUM_SURFACE_VIEWS:
case SVGA_QUERY_NUM_RESOURCES_MAPPED:
case SVGA_QUERY_NUM_BYTES_UPLOADED:
+ case SVGA_QUERY_NUM_GENERATE_MIPMAP:
/* nothing */
break;
default:
case SVGA_QUERY_NUM_RESOURCES:
case SVGA_QUERY_NUM_STATE_OBJECTS:
case SVGA_QUERY_NUM_SURFACE_VIEWS:
+ case SVGA_QUERY_NUM_GENERATE_MIPMAP:
/* nothing */
break;
default:
case SVGA_QUERY_NUM_RESOURCES:
case SVGA_QUERY_NUM_STATE_OBJECTS:
case SVGA_QUERY_NUM_SURFACE_VIEWS:
+ case SVGA_QUERY_NUM_GENERATE_MIPMAP:
/* nothing */
break;
default:
case SVGA_QUERY_NUM_SURFACE_VIEWS:
vresult->u64 = svga->hud.num_surface_views;
break;
+ case SVGA_QUERY_NUM_GENERATE_MIPMAP:
+ vresult->u64 = svga->hud.num_generate_mipmap;
+ break;
default:
assert(!"unexpected query type in svga_get_query_result");
}
struct svga_context *svga = svga_context(pipe);
struct svga_rasterizer_state *raster = (struct svga_rasterizer_state *)state;
+ if (!raster ||
+ !svga->curr.rast ||
+ raster->templ.poly_stipple_enable !=
+ svga->curr.rast->templ.poly_stipple_enable) {
+ svga->dirty |= SVGA_NEW_STIPPLE;
+ }
+
svga->curr.rast = raster;
svga->dirty |= SVGA_NEW_RAST;
-
- if (raster && raster->templ.poly_stipple_enable) {
- svga->dirty |= SVGA_NEW_STIPPLE;
- }
}
static void
{
struct svga_context *svga = svga_context(pipe);
unsigned i;
+ boolean any_change = FALSE;
assert(shader < PIPE_SHADER_TYPES);
assert(start + num <= PIPE_MAX_SAMPLERS);
if (!svga_have_vgpu10(svga) && shader != PIPE_SHADER_FRAGMENT)
return;
- for (i = 0; i < num; i++)
+ for (i = 0; i < num; i++) {
+ if (svga->curr.sampler[shader][start + i] != samplers[i])
+ any_change = TRUE;
svga->curr.sampler[shader][start + i] = samplers[i];
+ }
+
+ if (!any_change) {
+ return;
+ }
/* find highest non-null sampler[] entry */
{
unsigned flag_1d = 0;
unsigned flag_srgb = 0;
uint i;
+ boolean any_change = FALSE;
assert(shader < PIPE_SHADER_TYPES);
assert(start + num <= Elements(svga->curr.sampler_views[shader]));
pipe_sampler_view_release(pipe, &svga->curr.sampler_views[shader][start + i]);
pipe_sampler_view_reference(&svga->curr.sampler_views[shader][start + i],
views[i]);
+ any_change = TRUE;
}
if (!views[i])
flag_1d |= 1 << (start + i);
}
+ if (!any_change) {
+ return;
+ }
+
/* find highest non-null sampler_views[] entry */
{
unsigned j = MAX2(svga->curr.num_sampler_views[shader], start + num);
svga->pipe.transfer_flush_region = u_transfer_flush_region_vtbl;
svga->pipe.transfer_unmap = u_transfer_unmap_vtbl;
svga->pipe.transfer_inline_write = u_transfer_inline_write_vtbl;
+
+ if (svga_have_vgpu10(svga)) {
+ svga->pipe.generate_mipmap = svga_texture_generate_mipmap;
+ } else {
+ svga->pipe.generate_mipmap = NULL;
+ }
}
void
struct svga_3d_update_gb_image *whole_update_cmd = NULL;
uint32 numBoxes = sbuf->map.num_ranges;
struct pipe_resource *dummy;
- unsigned int i;
+ unsigned i;
assert(numBoxes);
assert(sbuf->dma.updates == NULL);
pipe_resource_reference(&dummy, &sbuf->b.b);
SVGA_FIFOCommitAll(swc);
+ swc->hints |= SVGA_HINT_FLAG_CAN_PRE_FLUSH;
sbuf->dma.flags.discard = FALSE;
return PIPE_OK;
SVGA_FIFOCommitAll(swc);
+ swc->hints |= SVGA_HINT_FLAG_CAN_PRE_FLUSH;
sbuf->dma.flags.discard = FALSE;
return PIPE_OK;
return &tex->b.b;
}
+
+boolean
+svga_texture_generate_mipmap(struct pipe_context *pipe,
+ struct pipe_resource *pt,
+ enum pipe_format format,
+ unsigned base_level,
+ unsigned last_level,
+ unsigned first_layer,
+ unsigned last_layer)
+{
+ struct pipe_sampler_view templ, *psv;
+ struct svga_pipe_sampler_view *sv;
+ struct svga_context *svga = svga_context(pipe);
+ struct svga_texture *tex = svga_texture(pt);
+ enum pipe_error ret;
+
+ assert(svga_have_vgpu10(svga));
+
+ /* Only support 2D texture for now */
+ if (pt->target != PIPE_TEXTURE_2D)
+ return FALSE;
+
+ /* Fallback to the mipmap generation utility for those formats that
+ * do not support hw generate mipmap
+ */
+ if (!svga_format_support_gen_mips(format))
+ return FALSE;
+
+ /* Make sure the texture surface was created with
+ * SVGA3D_SURFACE_BIND_RENDER_TARGET
+ */
+ if (!tex->handle || !(tex->key.flags & SVGA3D_SURFACE_BIND_RENDER_TARGET))
+ return FALSE;
+
+ templ.format = format;
+ templ.u.tex.first_layer = first_layer;
+ templ.u.tex.last_layer = last_layer;
+ templ.u.tex.first_level = base_level;
+ templ.u.tex.last_level = last_level;
+
+ psv = pipe->create_sampler_view(pipe, pt, &templ);
+ if (psv == NULL)
+ return FALSE;
+
+ sv = svga_pipe_sampler_view(psv);
+ svga_validate_pipe_sampler_view(svga, sv);
+
+ ret = SVGA3D_vgpu10_GenMips(svga->swc, sv->id, tex->handle);
+ if (ret != PIPE_OK) {
+ svga_context_flush(svga, NULL);
+ ret = SVGA3D_vgpu10_GenMips(svga->swc, sv->id, tex->handle);
+ }
+ pipe_sampler_view_reference(&psv, NULL);
+
+ svga->hud.num_generate_mipmap++;
+
+ return TRUE;
+}
const struct pipe_resource *template,
struct winsys_handle *whandle);
-
+boolean
+svga_texture_generate_mipmap(struct pipe_context *pipe,
+ struct pipe_resource *pt,
+ enum pipe_format format,
+ unsigned base_level,
+ unsigned last_level,
+ unsigned first_layer,
+ unsigned last_layer);
#endif /* SVGA_TEXTURE_H */
struct pipe_context;
struct pipe_screen;
struct svga_context;
+struct svga_pipe_sampler_view;
struct svga_winsys_surface;
struct svga_surface;
enum SVGA3dSurfaceFormat;
svga_check_sampler_view_resource_collision(struct svga_context *svga,
struct svga_winsys_surface *res,
unsigned shader);
+
+enum pipe_error
+svga_validate_pipe_sampler_view(struct svga_context *svga,
+ struct svga_pipe_sampler_view *sv);
+
#endif
case PIPE_CAP_PRIMITIVE_RESTART:
return 1; /* may be a sw fallback, depending on restart index */
+ case PIPE_CAP_GENERATE_MIPMAP:
+ return sws->have_vgpu10;
+
/* Unsupported features */
case PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION:
case PIPE_CAP_TEXTURE_MIRROR_CLAMP:
case PIPE_CAP_TEXTURE_GATHER_OFFSETS:
case PIPE_CAP_TGSI_VS_WINDOW_SPACE_POSITION:
case PIPE_CAP_DRAW_INDIRECT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT_PARAMS:
case PIPE_CAP_TGSI_FS_FINE_DERIVATIVE:
case PIPE_CAP_CONDITIONAL_RENDER_INVERTED:
case PIPE_CAP_SAMPLER_VIEW_TARGET:
case PIPE_CAP_VERTEXID_NOBASE:
case PIPE_CAP_POLYGON_OFFSET_CLAMP:
case PIPE_CAP_MULTISAMPLE_Z_RESOLVE:
+ case PIPE_CAP_TGSI_PACK_HALF_FLOAT:
+ case PIPE_CAP_SHADER_BUFFER_OFFSET_ALIGNMENT:
+ case PIPE_CAP_INVALIDATE_BUFFER:
return 0;
case PIPE_CAP_MIN_MAP_BUFFER_ALIGNMENT:
return 64;
case PIPE_CAP_SHAREABLE_SHADERS:
case PIPE_CAP_COPY_BETWEEN_COMPRESSED_AND_PLAIN_FORMATS:
case PIPE_CAP_CLEAR_TEXTURE:
+ case PIPE_CAP_DRAW_PARAMETERS:
+ case PIPE_CAP_TGSI_FS_POSITION_IS_SYSVAL:
+ case PIPE_CAP_TGSI_FS_FACE_IS_INTEGER_SYSVAL:
return 0;
}
case PIPE_SHADER_CAP_TGSI_DFRACEXP_DLDEXP_SUPPORTED:
case PIPE_SHADER_CAP_TGSI_FMA_SUPPORTED:
case PIPE_SHADER_CAP_TGSI_ANY_INOUT_DECL_RANGE:
+ case PIPE_SHADER_CAP_MAX_SHADER_BUFFERS:
return 0;
case PIPE_SHADER_CAP_MAX_UNROLL_ITERATIONS_HINT:
return 32;
case PIPE_SHADER_CAP_TGSI_DFRACEXP_DLDEXP_SUPPORTED:
case PIPE_SHADER_CAP_TGSI_FMA_SUPPORTED:
case PIPE_SHADER_CAP_TGSI_ANY_INOUT_DECL_RANGE:
+ case PIPE_SHADER_CAP_MAX_SHADER_BUFFERS:
return 0;
case PIPE_SHADER_CAP_MAX_UNROLL_ITERATIONS_HINT:
return 32;
case PIPE_SHADER_CAP_TGSI_DFRACEXP_DLDEXP_SUPPORTED:
case PIPE_SHADER_CAP_TGSI_FMA_SUPPORTED:
case PIPE_SHADER_CAP_TGSI_ANY_INOUT_DECL_RANGE:
+ case PIPE_SHADER_CAP_MAX_SHADER_BUFFERS:
return 0;
case PIPE_SHADER_CAP_MAX_UNROLL_ITERATIONS_HINT:
return 32;
PIPE_DRIVER_QUERY_TYPE_UINT64),
QUERY("num-surface-views", SVGA_QUERY_NUM_SURFACE_VIEWS,
PIPE_DRIVER_QUERY_TYPE_UINT64),
+ QUERY("num-generate-mipmap", SVGA_QUERY_NUM_GENERATE_MIPMAP,
+ PIPE_DRIVER_QUERY_TYPE_UINT64),
};
#undef QUERY
const struct svga_tracked_state *atoms[],
unsigned *state)
{
+#ifdef DEBUG
boolean debug = TRUE;
+#else
+ boolean debug = FALSE;
+#endif
enum pipe_error ret = PIPE_OK;
unsigned i;
*/
new_buf_size = align(new_buf_size, 16);
- u_upload_alloc(svga->const0_upload, 0, new_buf_size, &offset,
+ u_upload_alloc(svga->const0_upload, 0, new_buf_size,
+ CONST0_UPLOAD_ALIGNMENT, &offset,
&dst_buffer, &dst_map);
if (!dst_map) {
if (src_map)
SVGA_NEW_TEXTURE_BINDING |
SVGA_NEW_NEED_SWTNL |
SVGA_NEW_RAST |
+ SVGA_NEW_STIPPLE |
SVGA_NEW_REDUCED_PRIMITIVE |
SVGA_NEW_SAMPLER |
SVGA_NEW_FRAME_BUFFER |
* Create a DX ShaderResourceSamplerView for the given pipe_sampler_view,
* if needed.
*/
-static enum pipe_error
+enum pipe_error
svga_validate_pipe_sampler_view(struct svga_context *svga,
struct svga_pipe_sampler_view *sv)
{
return;
}
+ /* SVGA_NEW_PRESCALE */
key->vs.need_prescale = svga->state.hw_clear.prescale.enabled &&
(svga->curr.gs == NULL);
+
+ /* SVGA_NEW_RAST */
key->vs.allow_psiz = svga->curr.rast->templ.point_size_per_vertex;
/* SVGA_NEW_FS */
struct draw_context *draw = svga->swtnl.draw;
struct vertex_info *vinfo = &svga_render->vertex_info;
SVGA3dVertexDecl vdecl[PIPE_MAX_ATTRIBS];
- const enum interp_mode colorInterp =
- svga->curr.rast->templ.flatshade ? INTERP_CONSTANT : INTERP_LINEAR;
struct svga_fragment_shader *fs = svga->curr.fs;
int offset = 0;
int nr_decls = 0;
/* always add position */
src = draw_find_shader_output(draw, TGSI_SEMANTIC_POSITION, 0);
- draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_LINEAR, src);
+ draw_emit_vertex_attr(vinfo, EMIT_4F, src);
vinfo->attrib[0].emit = EMIT_4F;
vdecl[0].array.offset = offset;
vdecl[0].identity.method = SVGA3D_DECLMETHOD_DEFAULT;
switch (sem_name) {
case TGSI_SEMANTIC_COLOR:
- draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src);
+ draw_emit_vertex_attr(vinfo, EMIT_4F, src);
vdecl[nr_decls].identity.usage = SVGA3D_DECLUSAGE_COLOR;
vdecl[nr_decls].identity.type = SVGA3D_DECLTYPE_FLOAT4;
offset += 16;
nr_decls++;
break;
case TGSI_SEMANTIC_GENERIC:
- draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src);
+ draw_emit_vertex_attr(vinfo, EMIT_4F, src);
vdecl[nr_decls].identity.usage = SVGA3D_DECLUSAGE_TEXCOORD;
vdecl[nr_decls].identity.type = SVGA3D_DECLTYPE_FLOAT4;
vdecl[nr_decls].identity.usageIndex =
nr_decls++;
break;
case TGSI_SEMANTIC_FOG:
- draw_emit_vertex_attr(vinfo, EMIT_1F, INTERP_PERSPECTIVE, src);
+ draw_emit_vertex_attr(vinfo, EMIT_1F, src);
vdecl[nr_decls].identity.usage = SVGA3D_DECLUSAGE_TEXCOORD;
vdecl[nr_decls].identity.type = SVGA3D_DECLTYPE_FLOAT1;
assert(vdecl[nr_decls].identity.usageIndex == 0);
static boolean
svga_arl_needs_adjustment( const struct svga_shader_emitter *emit )
{
- int i;
+ unsigned i;
for (i = 0; i < emit->num_arl_consts; ++i) {
if (emit->arl_consts[i].arl_num == emit->current_arl)
static int
svga_arl_adjustment( const struct svga_shader_emitter *emit )
{
- int i;
+ unsigned i;
for (i = 0; i < emit->num_arl_consts; ++i) {
if (emit->arl_consts[i].arl_num == emit->current_arl)
const struct src_register src1 =
translate_src_register(emit, &insn->Src[1] );
SVGA3dShaderDestToken temp = get_temp( emit );
- int i;
+ unsigned i;
/* For each enabled element, perform a RCP instruction. Note that
* RCP is scalar in SVGA3D:
const unsigned swizzleIn[4] = {swizzle_x, swizzle_y, swizzle_z, swizzle_w};
unsigned srcSwizzle[4];
unsigned srcWritemask = 0x0, zeroWritemask = 0x0, oneWritemask = 0x0;
- int i;
+ unsigned i;
/* build writemasks and srcSwizzle terms */
for (i = 0; i < 4; i++) {
struct src_register back[2];
SVGA3dShaderDestToken color[2];
int count = emit->internal_color_count;
- int i;
+ unsigned i;
SVGA3dShaderInstToken if_token;
if (count == 0)
pre_parse_add_indirect( struct svga_shader_emitter *emit,
int num, int current_arl)
{
- int i;
+ unsigned i;
assert(num < 0);
for (i = 0; i < emit->num_arl_consts; ++i) {
if (emit->unit == PIPE_SHADER_FRAGMENT && emit->key.fs.pstipple) {
unsigned unit;
- new_tokens = util_pstipple_create_fragment_shader(tokens, &unit, 0);
+ new_tokens = util_pstipple_create_fragment_shader(tokens, &unit, 0,
+ TGSI_FILE_INPUT);
if (new_tokens) {
/* Setup texture state for stipple */
emit->num_samplers = MAX2(emit->num_samplers, decl->Range.Last + 1);
return TRUE;
+#if 0
case TGSI_FILE_RESOURCE:
/*opcode0.opcodeType = VGPU10_OPCODE_DCL_RESOURCE;*/
/* XXX more, VGPU10_RETURN_TYPE_FLOAT */
assert(!"TGSI_FILE_RESOURCE not handled yet");
return FALSE;
+#endif
case TGSI_FILE_ADDRESS:
emit->num_address_regs = MAX2(emit->num_address_regs,
while (adjust_mask) {
unsigned index = u_bit_scan(&adjust_mask);
+
+ /* skip the instruction if this vertex attribute is not being used */
+ if (emit->info.input_usage_mask[index] == 0)
+ continue;
+
unsigned tmp = emit->vs.adjusted_input[index];
struct tgsi_full_src_register input_src =
make_src_reg(TGSI_FILE_INPUT, index);
tgsi_dump(tokens,0);
}
- new_tokens = util_pstipple_create_fragment_shader(tokens, &unit, 0);
+ new_tokens = util_pstipple_create_fragment_shader(tokens, &unit, 0,
+ TGSI_FILE_INPUT);
emit->fs.pstipple_sampler_unit = unit;
#define SVGA_QUERY_FLAG_SET (1 << 0)
#define SVGA_QUERY_FLAG_REF (1 << 1)
-#define SVGA_HINT_FLAG_DRAW_EMITTED (1 << 0)
+#define SVGA_HINT_FLAG_CAN_PRE_FLUSH (1 << 0) /* Can preemptively flush */
/** Opaque surface handle */
struct svga_winsys_surface;
}
+static inline boolean
+trace_context_generate_mipmap(struct pipe_context *_pipe,
+ struct pipe_resource *res,
+ enum pipe_format format,
+ unsigned base_level,
+ unsigned last_level,
+ unsigned first_layer,
+ unsigned last_layer)
+{
+ struct trace_context *tr_ctx = trace_context(_pipe);
+ struct pipe_context *pipe = tr_ctx->pipe;
+ boolean ret;
+
+ res = trace_resource_unwrap(tr_ctx, res);
+
+ trace_dump_call_begin("pipe_context", "generate_mipmap");
+
+ trace_dump_arg(ptr, pipe);
+ trace_dump_arg(ptr, res);
+
+ trace_dump_arg(format, format);
+ trace_dump_arg(uint, base_level);
+ trace_dump_arg(uint, last_level);
+ trace_dump_arg(uint, first_layer);
+ trace_dump_arg(uint, last_layer);
+
+ ret = pipe->generate_mipmap(pipe, res, format, base_level, last_level,
+ first_layer, last_layer);
+
+ trace_dump_ret(bool, ret);
+ trace_dump_call_end();
+
+ return ret;
+}
+
+
static inline void
trace_context_destroy(struct pipe_context *_pipe)
{
TR_CTX_INIT(clear_render_target);
TR_CTX_INIT(clear_depth_stencil);
TR_CTX_INIT(flush);
+ TR_CTX_INIT(generate_mipmap);
TR_CTX_INIT(texture_barrier);
TR_CTX_INIT(memory_barrier);
TR_CTX_INIT(set_tess_state);
vc4_program.c \
vc4_qir.c \
vc4_qir_lower_uniforms.c \
+ vc4_qir_schedule.c \
vc4_qir.h \
vc4_qpu.c \
vc4_qpu_defines.h \
bool old_msaa = vc4->msaa;
int old_tile_width = vc4->tile_width;
int old_tile_height = vc4->tile_height;
- bool msaa = (info->src.resource->nr_samples ||
- info->dst.resource->nr_samples);
+ bool msaa = (info->src.resource->nr_samples > 1 ||
+ info->dst.resource->nr_samples > 1);
int tile_width = msaa ? 32 : 64;
int tile_height = msaa ? 32 : 64;
pipe_surface_reference(&vc4->color_read, src_surf);
pipe_surface_reference(&vc4->color_write,
- dst_surf->texture->nr_samples ? NULL : dst_surf);
+ dst_surf->texture->nr_samples > 1 ?
+ NULL : dst_surf);
pipe_surface_reference(&vc4->msaa_color_write,
- dst_surf->texture->nr_samples ? dst_surf : NULL);
+ dst_surf->texture->nr_samples > 1 ?
+ dst_surf : NULL);
pipe_surface_reference(&vc4->zs_read, NULL);
pipe_surface_reference(&vc4->zs_write, NULL);
pipe_surface_reference(&vc4->msaa_zs_write, NULL);
cl_u8(&bcl, VC4_PACKET_FLUSH);
cl_end(&vc4->bcl, bcl);
- vc4->msaa = false;
if (cbuf && (vc4->resolve & PIPE_CLEAR_COLOR0)) {
pipe_surface_reference(&vc4->color_write,
- cbuf->texture->nr_samples ? NULL : cbuf);
+ cbuf->texture->nr_samples > 1 ?
+ NULL : cbuf);
pipe_surface_reference(&vc4->msaa_color_write,
- cbuf->texture->nr_samples ? cbuf : NULL);
-
- if (cbuf->texture->nr_samples)
- vc4->msaa = true;
+ cbuf->texture->nr_samples > 1 ?
+ cbuf : NULL);
if (!(vc4->cleared & PIPE_CLEAR_COLOR0)) {
pipe_surface_reference(&vc4->color_read, cbuf);
if (vc4->framebuffer.zsbuf &&
(vc4->resolve & (PIPE_CLEAR_DEPTH | PIPE_CLEAR_STENCIL))) {
pipe_surface_reference(&vc4->zs_write,
- zsbuf->texture->nr_samples ?
+ zsbuf->texture->nr_samples > 1 ?
NULL : zsbuf);
pipe_surface_reference(&vc4->msaa_zs_write,
- zsbuf->texture->nr_samples ?
+ zsbuf->texture->nr_samples > 1 ?
zsbuf : NULL);
- if (zsbuf->texture->nr_samples)
- vc4->msaa = true;
-
if (!(vc4->cleared & (PIPE_CLEAR_DEPTH | PIPE_CLEAR_STENCIL))) {
pipe_surface_reference(&vc4->zs_read, zsbuf);
} else {
if (!vc4->primconvert)
goto fail;
- vc4->uploader = u_upload_create(pctx, 16 * 1024, 4,
- PIPE_BIND_INDEX_BUFFER);
+ vc4->uploader = u_upload_create(pctx, 16 * 1024,
+ PIPE_BIND_INDEX_BUFFER,
+ PIPE_USAGE_STREAM);
vc4_debug |= saved_shaderdb_flag;
if (vc4->indexbuf.user_buffer) {
prsc = NULL;
u_upload_data(vc4->uploader, 0,
- info->count * index_size,
+ info->count * index_size, 4,
vc4->indexbuf.user_buffer,
&offset, &prsc);
} else {
#define DRM_VC4_CREATE_BO 0x03
#define DRM_VC4_MMAP_BO 0x04
#define DRM_VC4_CREATE_SHADER_BO 0x05
+#define DRM_VC4_GET_HANG_STATE 0x06
#define DRM_IOCTL_VC4_SUBMIT_CL DRM_IOWR( DRM_COMMAND_BASE + DRM_VC4_SUBMIT_CL, struct drm_vc4_submit_cl)
#define DRM_IOCTL_VC4_WAIT_SEQNO DRM_IOWR( DRM_COMMAND_BASE + DRM_VC4_WAIT_SEQNO, struct drm_vc4_wait_seqno)
#define DRM_IOCTL_VC4_CREATE_BO DRM_IOWR( DRM_COMMAND_BASE + DRM_VC4_CREATE_BO, struct drm_vc4_create_bo)
#define DRM_IOCTL_VC4_MMAP_BO DRM_IOWR( DRM_COMMAND_BASE + DRM_VC4_MMAP_BO, struct drm_vc4_mmap_bo)
#define DRM_IOCTL_VC4_CREATE_SHADER_BO DRM_IOWR( DRM_COMMAND_BASE + DRM_VC4_CREATE_SHADER_BO, struct drm_vc4_create_shader_bo)
+#define DRM_IOCTL_VC4_GET_HANG_STATE DRM_IOWR( DRM_COMMAND_BASE + DRM_VC4_GET_HANG_STATE, struct drm_vc4_get_hang_state)
struct drm_vc4_submit_rcl_surface {
uint32_t hindex; /* Handle index, or ~0 if not present. */
uint64_t offset;
};
+struct drm_vc4_get_hang_state_bo {
+ uint32_t handle;
+ uint32_t paddr;
+ uint32_t size;
+ uint32_t pad;
+};
+
+/**
+ * struct drm_vc4_hang_state - ioctl argument for collecting state
+ * from a GPU hang for analysis.
+*/
+struct drm_vc4_get_hang_state {
+ /** Pointer to array of struct drm_vc4_get_hang_state_bo. */
+ uint64_t bo;
+ /**
+ * On input, the size of the bo array. Output is the number
+ * of bos to be returned.
+ */
+ uint32_t bo_count;
+
+ uint32_t start_bin, start_render;
+
+ uint32_t ct0ca, ct0ea;
+ uint32_t ct1ca, ct1ea;
+ uint32_t ct0cs, ct1cs;
+ uint32_t ct0ra0, ct1ra0;
+
+ uint32_t bpca, bpcs;
+ uint32_t bpoa, bpos;
+
+ uint32_t vpmbase;
+
+ uint32_t dbge;
+ uint32_t fdbgo;
+ uint32_t fdbgb;
+ uint32_t fdbgr;
+ uint32_t fdbgs;
+ uint32_t errstat;
+
+ /* Pad that we may save more registers into in the future. */
+ uint32_t pad[16];
+};
+
#endif /* _UAPI_VC4_DRM_H_ */
submit_surf->hindex = vc4_gem_hindex(vc4, rsc->bo);
submit_surf->offset = surf->offset;
- if (psurf->texture->nr_samples == 0) {
+ if (psurf->texture->nr_samples <= 1) {
if (is_depth) {
submit_surf->bits =
VC4_SET_FIELD(VC4_LOADSTORE_TILE_BUFFER_ZS,
submit_surf->hindex = vc4_gem_hindex(vc4, rsc->bo);
submit_surf->offset = surf->offset;
- if (psurf->texture->nr_samples == 0) {
+ if (psurf->texture->nr_samples <= 1) {
submit_surf->bits =
VC4_SET_FIELD(vc4_rt_format_is_565(surf->base.format) ?
VC4_RENDER_CONFIG_FORMAT_BGR565 :
#else
ret = vc4_simulator_flush(vc4, &submit);
#endif
- if (ret) {
- fprintf(stderr, "VC4 submit failed\n");
- abort();
+ static bool warned = false;
+ if (ret && !warned) {
+ fprintf(stderr, "Draw call returned %s. "
+ "Expect corruption.\n", strerror(errno));
+ warned = true;
}
}
void
vc4_nir_lower_blend(struct vc4_compile *c)
{
- nir_foreach_overload(c->s, overload) {
- if (overload->impl) {
- nir_foreach_block(overload->impl,
+ nir_foreach_function(c->s, function) {
+ if (function->impl) {
+ nir_foreach_block(function->impl,
vc4_nir_lower_blend_block, c);
- nir_metadata_preserve(overload->impl,
+ nir_metadata_preserve(function->impl,
nir_metadata_block_index |
nir_metadata_dominance);
}
void
vc4_nir_lower_io(struct vc4_compile *c)
{
- nir_foreach_overload(c->s, overload) {
- if (overload->impl)
- vc4_nir_lower_io_impl(c, overload->impl);
+ nir_foreach_function(c->s, function) {
+ if (function->impl)
+ vc4_nir_lower_io_impl(c, function->impl);
}
}
void
vc4_nir_lower_txf_ms(struct vc4_compile *c)
{
- nir_foreach_overload(c->s, overload) {
- if (overload->impl)
- vc4_nir_lower_txf_ms_impl(c, overload->impl);
+ nir_foreach_function(c->s, function) {
+ if (function->impl)
+ vc4_nir_lower_txf_ms_impl(c, function->impl);
}
}
list_for_each_entry(struct qinst, inst, &c->instructions, link) {
switch (inst->op) {
- case QOP_SEL_X_Y_ZS:
- case QOP_SEL_X_Y_ZC:
- case QOP_SEL_X_Y_NS:
- case QOP_SEL_X_Y_NC:
- case QOP_SEL_X_Y_CS:
- case QOP_SEL_X_Y_CC:
- if (is_zero(c, inst->src[1])) {
- /* Replace references to a 0 uniform value
- * with the SEL_X_0 equivalent.
- */
- dump_from(c, inst);
- inst->op -= (QOP_SEL_X_Y_ZS - QOP_SEL_X_0_ZS);
- inst->src[1] = c->undef;
- progress = true;
- dump_to(c, inst);
- break;
- }
-
- if (is_zero(c, inst->src[0])) {
- /* Replace references to a 0 uniform value
- * with the SEL_X_0 equivalent, flipping the
- * condition being evaluated since the operand
- * order is flipped.
- */
- dump_from(c, inst);
- inst->op -= QOP_SEL_X_Y_ZS;
- inst->op ^= 1;
- inst->op += QOP_SEL_X_0_ZS;
- inst->src[0] = inst->src[1];
- inst->src[1] = c->undef;
- progress = true;
- dump_to(c, inst);
- break;
- }
-
- break;
-
case QOP_FMIN:
if (is_1f(c, inst->src[1]) &&
inst->src[0].pack >= QPU_UNPACK_8D_REP &&
range->dst_offset = c->next_ubo_dst_offset;
c->next_ubo_dst_offset += range->size;
c->num_ubo_ranges++;
- };
+ }
offset -= range->src_offset;
return r;
};
-static struct qreg
-get_swizzled_channel(struct vc4_compile *c,
- struct qreg *srcs, int swiz)
-{
- switch (swiz) {
- default:
- case UTIL_FORMAT_SWIZZLE_NONE:
- fprintf(stderr, "warning: unknown swizzle\n");
- /* FALLTHROUGH */
- case UTIL_FORMAT_SWIZZLE_0:
- return qir_uniform_f(c, 0.0);
- case UTIL_FORMAT_SWIZZLE_1:
- return qir_uniform_f(c, 1.0);
- case UTIL_FORMAT_SWIZZLE_X:
- case UTIL_FORMAT_SWIZZLE_Y:
- case UTIL_FORMAT_SWIZZLE_Z:
- case UTIL_FORMAT_SWIZZLE_W:
- return srcs[swiz];
- }
-}
-
static inline struct qreg
qir_SAT(struct vc4_compile *c, struct qreg val)
{
qir_uniform_f(c, 2.4));
qir_SF(c, qir_FSUB(c, srgb, qir_uniform_f(c, 0.04045)));
- return qir_SEL_X_Y_NS(c, low, high);
+ return qir_SEL(c, QPU_COND_NS, low, high);
}
static struct qreg
struct qreg tex = qir_TEX_RESULT(c);
c->num_texture_samples++;
- struct qreg texture_output[4];
+ struct qreg *dest = ntq_get_dest(c, &instr->dest);
enum pipe_format format = c->key->tex[unit].format;
if (util_format_is_depth_or_stencil(format)) {
struct qreg scaled = ntq_scale_depth_texture(c, tex);
for (int i = 0; i < 4; i++)
- texture_output[i] = scaled;
+ dest[i] = scaled;
} else {
- struct qreg tex_result_unpacked[4];
for (int i = 0; i < 4; i++)
- tex_result_unpacked[i] = qir_UNPACK_8_F(c, tex, i);
-
- const uint8_t *format_swiz =
- vc4_get_format_swizzle(c->key->tex[unit].format);
- for (int i = 0; i < 4; i++) {
- texture_output[i] =
- get_swizzled_channel(c, tex_result_unpacked,
- format_swiz[i]);
- }
+ dest[i] = qir_UNPACK_8_F(c, tex, i);
}
- struct qreg *dest = ntq_get_dest(c, &instr->dest);
for (int i = 0; i < 4; i++) {
- dest[i] = get_swizzled_channel(c, texture_output,
- c->key->tex[unit].swizzle[i]);
+ if (c->tex_srgb_decode[unit] & (1 << i))
+ dest[i] = qir_srgb_decode(c, dest[i]);
}
}
enum pipe_format format = c->key->tex[unit].format;
- struct qreg unpacked[4];
+ struct qreg *dest = ntq_get_dest(c, &instr->dest);
if (util_format_is_depth_or_stencil(format)) {
struct qreg normalized = ntq_scale_depth_texture(c, tex);
struct qreg depth_output;
- struct qreg one = qir_uniform_f(c, 1.0f);
+ struct qreg u0 = qir_uniform_f(c, 0.0f);
+ struct qreg u1 = qir_uniform_f(c, 1.0f);
if (c->key->tex[unit].compare_mode) {
if (has_proj)
compare = qir_FMUL(c, compare, proj);
depth_output = qir_uniform_f(c, 0.0f);
break;
case PIPE_FUNC_ALWAYS:
- depth_output = one;
+ depth_output = u1;
break;
case PIPE_FUNC_EQUAL:
qir_SF(c, qir_FSUB(c, compare, normalized));
- depth_output = qir_SEL_X_0_ZS(c, one);
+ depth_output = qir_SEL(c, QPU_COND_ZS, u1, u0);
break;
case PIPE_FUNC_NOTEQUAL:
qir_SF(c, qir_FSUB(c, compare, normalized));
- depth_output = qir_SEL_X_0_ZC(c, one);
+ depth_output = qir_SEL(c, QPU_COND_ZC, u1, u0);
break;
case PIPE_FUNC_GREATER:
qir_SF(c, qir_FSUB(c, compare, normalized));
- depth_output = qir_SEL_X_0_NC(c, one);
+ depth_output = qir_SEL(c, QPU_COND_NC, u1, u0);
break;
case PIPE_FUNC_GEQUAL:
qir_SF(c, qir_FSUB(c, normalized, compare));
- depth_output = qir_SEL_X_0_NS(c, one);
+ depth_output = qir_SEL(c, QPU_COND_NS, u1, u0);
break;
case PIPE_FUNC_LESS:
qir_SF(c, qir_FSUB(c, compare, normalized));
- depth_output = qir_SEL_X_0_NS(c, one);
+ depth_output = qir_SEL(c, QPU_COND_NS, u1, u0);
break;
case PIPE_FUNC_LEQUAL:
qir_SF(c, qir_FSUB(c, normalized, compare));
- depth_output = qir_SEL_X_0_NC(c, one);
+ depth_output = qir_SEL(c, QPU_COND_NC, u1, u0);
break;
}
} else {
}
for (int i = 0; i < 4; i++)
- unpacked[i] = depth_output;
+ dest[i] = depth_output;
} else {
for (int i = 0; i < 4; i++)
- unpacked[i] = qir_UNPACK_8_F(c, tex, i);
- }
-
- const uint8_t *format_swiz = vc4_get_format_swizzle(format);
- struct qreg texture_output[4];
- for (int i = 0; i < 4; i++) {
- texture_output[i] = get_swizzled_channel(c, unpacked,
- format_swiz[i]);
+ dest[i] = qir_UNPACK_8_F(c, tex, i);
}
- if (util_format_is_srgb(format)) {
- for (int i = 0; i < 3; i++)
- texture_output[i] = qir_srgb_decode(c,
- texture_output[i]);
- }
-
- struct qreg *dest = ntq_get_dest(c, &instr->dest);
for (int i = 0; i < 4; i++) {
- dest[i] = get_swizzled_channel(c, texture_output,
- c->key->tex[unit].swizzle[i]);
+ if (c->tex_srgb_decode[unit] & (1 << i))
+ dest[i] = qir_srgb_decode(c, dest[i]);
}
}
struct qreg trunc = qir_ITOF(c, qir_FTOI(c, src));
struct qreg diff = qir_FSUB(c, src, trunc);
qir_SF(c, diff);
- return qir_SEL_X_Y_NS(c,
- qir_FADD(c, diff, qir_uniform_f(c, 1.0)),
- diff);
+ return qir_SEL(c, QPU_COND_NS,
+ qir_FADD(c, diff, qir_uniform_f(c, 1.0)), diff);
}
/**
*/
qir_SF(c, qir_FSUB(c, src, trunc));
- return qir_SEL_X_Y_NS(c,
- qir_FSUB(c, trunc, qir_uniform_f(c, 1.0)),
- trunc);
+ return qir_SEL(c, QPU_COND_NS,
+ qir_FSUB(c, trunc, qir_uniform_f(c, 1.0)), trunc);
}
/**
*/
qir_SF(c, qir_FSUB(c, trunc, src));
- return qir_SEL_X_Y_NS(c,
- qir_FADD(c, trunc, qir_uniform_f(c, 1.0)),
- trunc);
+ return qir_SEL(c, QPU_COND_NS,
+ qir_FADD(c, trunc, qir_uniform_f(c, 1.0)), trunc);
}
static struct qreg
static struct qreg
ntq_fsign(struct vc4_compile *c, struct qreg src)
{
+ struct qreg t = qir_get_temp(c);
+
qir_SF(c, src);
- return qir_SEL_X_Y_NC(c,
- qir_SEL_X_0_ZC(c, qir_uniform_f(c, 1.0)),
- qir_uniform_f(c, -1.0));
+ qir_MOV_dest(c, t, qir_uniform_f(c, 0.0));
+ qir_MOV_dest(c, t, qir_uniform_f(c, 1.0))->cond = QPU_COND_ZC;
+ qir_MOV_dest(c, t, qir_uniform_f(c, -1.0))->cond = QPU_COND_NS;
+ return t;
}
static void
return qir_UNPACK_8_I(c, base, offset_bit / 8);
}
+/**
+ * If compare_instr is a valid comparison instruction, emits the
+ * compare_instr's comparison and returns the sel_instr's return value based
+ * on the compare_instr's result.
+ */
+static bool
+ntq_emit_comparison(struct vc4_compile *c, struct qreg *dest,
+ nir_alu_instr *compare_instr,
+ nir_alu_instr *sel_instr)
+{
+ enum qpu_cond cond;
+
+ switch (compare_instr->op) {
+ case nir_op_feq:
+ case nir_op_ieq:
+ case nir_op_seq:
+ cond = QPU_COND_ZS;
+ break;
+ case nir_op_fne:
+ case nir_op_ine:
+ case nir_op_sne:
+ cond = QPU_COND_ZC;
+ break;
+ case nir_op_fge:
+ case nir_op_ige:
+ case nir_op_uge:
+ case nir_op_sge:
+ cond = QPU_COND_NC;
+ break;
+ case nir_op_flt:
+ case nir_op_ilt:
+ case nir_op_slt:
+ cond = QPU_COND_NS;
+ break;
+ default:
+ return false;
+ }
+
+ struct qreg src0 = ntq_get_alu_src(c, compare_instr, 0);
+ struct qreg src1 = ntq_get_alu_src(c, compare_instr, 1);
+
+ if (nir_op_infos[compare_instr->op].input_types[0] == nir_type_float)
+ qir_SF(c, qir_FSUB(c, src0, src1));
+ else
+ qir_SF(c, qir_SUB(c, src0, src1));
+
+ switch (sel_instr->op) {
+ case nir_op_seq:
+ case nir_op_sne:
+ case nir_op_sge:
+ case nir_op_slt:
+ *dest = qir_SEL(c, cond,
+ qir_uniform_f(c, 1.0), qir_uniform_f(c, 0.0));
+ break;
+
+ case nir_op_bcsel:
+ *dest = qir_SEL(c, cond,
+ ntq_get_alu_src(c, sel_instr, 1),
+ ntq_get_alu_src(c, sel_instr, 2));
+ break;
+
+ default:
+ *dest = qir_SEL(c, cond,
+ qir_uniform_ui(c, ~0), qir_uniform_ui(c, 0));
+ break;
+ }
+
+ return true;
+}
+
+/**
+ * Attempts to fold a comparison generating a boolean result into the
+ * condition code for selecting between two values, instead of comparing the
+ * boolean result against 0 to generate the condition code.
+ */
+static struct qreg ntq_emit_bcsel(struct vc4_compile *c, nir_alu_instr *instr,
+ struct qreg *src)
+{
+ if (!instr->src[0].src.is_ssa)
+ goto out;
+ nir_alu_instr *compare =
+ nir_instr_as_alu(instr->src[0].src.ssa->parent_instr);
+ if (!compare)
+ goto out;
+
+ struct qreg dest;
+ if (ntq_emit_comparison(c, &dest, compare, instr))
+ return dest;
+
+out:
+ qir_SF(c, src[0]);
+ return qir_SEL(c, QPU_COND_NS, src[1], src[2]);
+}
+
static void
ntq_emit_alu(struct vc4_compile *c, nir_alu_instr *instr)
{
case nir_op_i2b:
case nir_op_f2b:
qir_SF(c, src[0]);
- *dest = qir_SEL_X_0_ZC(c, qir_uniform_ui(c, ~0));
+ *dest = qir_SEL(c, QPU_COND_ZC,
+ qir_uniform_ui(c, ~0),
+ qir_uniform_ui(c, 0));
break;
case nir_op_iadd:
break;
case nir_op_seq:
- qir_SF(c, qir_FSUB(c, src[0], src[1]));
- *dest = qir_SEL_X_0_ZS(c, qir_uniform_f(c, 1.0));
- break;
case nir_op_sne:
- qir_SF(c, qir_FSUB(c, src[0], src[1]));
- *dest = qir_SEL_X_0_ZC(c, qir_uniform_f(c, 1.0));
- break;
case nir_op_sge:
- qir_SF(c, qir_FSUB(c, src[0], src[1]));
- *dest = qir_SEL_X_0_NC(c, qir_uniform_f(c, 1.0));
- break;
case nir_op_slt:
- qir_SF(c, qir_FSUB(c, src[0], src[1]));
- *dest = qir_SEL_X_0_NS(c, qir_uniform_f(c, 1.0));
- break;
case nir_op_feq:
- qir_SF(c, qir_FSUB(c, src[0], src[1]));
- *dest = qir_SEL_X_0_ZS(c, qir_uniform_ui(c, ~0));
- break;
case nir_op_fne:
- qir_SF(c, qir_FSUB(c, src[0], src[1]));
- *dest = qir_SEL_X_0_ZC(c, qir_uniform_ui(c, ~0));
- break;
case nir_op_fge:
- qir_SF(c, qir_FSUB(c, src[0], src[1]));
- *dest = qir_SEL_X_0_NC(c, qir_uniform_ui(c, ~0));
- break;
case nir_op_flt:
- qir_SF(c, qir_FSUB(c, src[0], src[1]));
- *dest = qir_SEL_X_0_NS(c, qir_uniform_ui(c, ~0));
- break;
case nir_op_ieq:
- qir_SF(c, qir_SUB(c, src[0], src[1]));
- *dest = qir_SEL_X_0_ZS(c, qir_uniform_ui(c, ~0));
- break;
case nir_op_ine:
- qir_SF(c, qir_SUB(c, src[0], src[1]));
- *dest = qir_SEL_X_0_ZC(c, qir_uniform_ui(c, ~0));
- break;
case nir_op_ige:
- qir_SF(c, qir_SUB(c, src[0], src[1]));
- *dest = qir_SEL_X_0_NC(c, qir_uniform_ui(c, ~0));
- break;
case nir_op_uge:
- qir_SF(c, qir_SUB(c, src[0], src[1]));
- *dest = qir_SEL_X_0_CC(c, qir_uniform_ui(c, ~0));
- break;
case nir_op_ilt:
- qir_SF(c, qir_SUB(c, src[0], src[1]));
- *dest = qir_SEL_X_0_NS(c, qir_uniform_ui(c, ~0));
+ if (!ntq_emit_comparison(c, dest, instr, instr)) {
+ fprintf(stderr, "Bad comparison instruction\n");
+ }
break;
case nir_op_bcsel:
- qir_SF(c, src[0]);
- *dest = qir_SEL_X_Y_NS(c, src[1], src[2]);
+ *dest = ntq_emit_bcsel(c, instr, src);
break;
case nir_op_fcsel:
qir_SF(c, src[0]);
- *dest = qir_SEL_X_Y_ZC(c, src[1], src[2]);
+ *dest = qir_SEL(c, QPU_COND_ZC, src[1], src[2]);
break;
case nir_op_frcp:
ntq_setup_registers(c, &c->s->registers);
/* Find the main function and emit the body. */
- nir_foreach_overload(c->s, overload) {
- assert(strcmp(overload->function->name, "main") == 0);
- assert(overload->impl);
- ntq_emit_impl(c, overload->impl);
+ nir_foreach_function(c->s, function) {
+ assert(strcmp(function->name, "main") == 0);
+ assert(function->impl);
+ ntq_emit_impl(c, function->impl);
}
}
count_nir_instrs(nir_shader *nir)
{
int count = 0;
- nir_foreach_overload(nir, overload) {
- if (!overload->impl)
+ nir_foreach_function(nir, function) {
+ if (!function->impl)
continue;
- nir_foreach_block(overload->impl, count_nir_instrs_in_block, &count);
+ nir_foreach_block(function->impl, count_nir_instrs_in_block, &count);
}
return count;
}
if (stage == QSTAGE_FRAG)
vc4_nir_lower_blend(c);
+ struct nir_lower_tex_options tex_options = {
+ /* We would need to implement txs, but we don't want the
+ * int/float conversions
+ */
+ .lower_rect = false,
+
+ /* We want to use this, but we don't want to newton-raphson
+ * its rcp.
+ */
+ .lower_txp = false,
+
+ /* Apply swizzles to all samplers. */
+ .swizzle_result = ~0,
+ };
+
+ /* Lower the format swizzle and ARB_texture_swizzle-style swizzle.
+ * The format swizzling applies before sRGB decode, and
+ * ARB_texture_swizzle is the last thing before returning the sample.
+ */
+ for (int i = 0; i < ARRAY_SIZE(key->tex); i++) {
+ enum pipe_format format = c->key->tex[i].format;
+
+ if (!format)
+ continue;
+
+ const uint8_t *format_swizzle = vc4_get_format_swizzle(format);
+
+ for (int j = 0; j < 4; j++) {
+ uint8_t arb_swiz = c->key->tex[i].swizzle[j];
+
+ if (arb_swiz <= 3) {
+ tex_options.swizzles[i][j] =
+ format_swizzle[arb_swiz];
+ } else {
+ tex_options.swizzles[i][j] = arb_swiz;
+ }
+
+ /* If ARB_texture_swizzle is reading from the R, G, or
+ * B channels of an sRGB texture, then we need to
+ * apply sRGB decode to this channel at sample time.
+ */
+ if (arb_swiz < 3 && util_format_is_srgb(format)) {
+ c->tex_srgb_decode[i] |= (1 << j);
+ }
+
+ }
+ }
+
+ nir_lower_tex(c->s, &tex_options);
+
if (c->fs_key && c->fs_key->light_twoside)
nir_lower_two_sided_color(c->s);
qir_optimize(c);
qir_lower_uniforms(c);
+ qir_schedule_instructions(c);
+
if (vc4_debug & VC4_DEBUG_QIR) {
fprintf(stderr, "%s prog %d/%d QIR:\n",
qir_get_stage_name(c->stage),
c->program_id, c->variant_id);
qir_dump(c);
}
+
qir_reorder_uniforms(c);
vc4_generate_code(vc4, c);
key->tex[i].swizzle[2] = sampler->swizzle_b;
key->tex[i].swizzle[3] = sampler->swizzle_a;
- if (sampler->texture->nr_samples) {
+ if (sampler->texture->nr_samples > 1) {
key->tex[i].msaa_width = sampler->texture->width0;
key->tex[i].msaa_height = sampler->texture->height0;
} else if (sampler){
[QOP_XOR] = { "xor", 1, 2 },
[QOP_NOT] = { "not", 1, 1 },
- [QOP_SEL_X_0_NS] = { "fsel_x_0_ns", 1, 1, false, true },
- [QOP_SEL_X_0_NC] = { "fsel_x_0_nc", 1, 1, false, true },
- [QOP_SEL_X_0_ZS] = { "fsel_x_0_zs", 1, 1, false, true },
- [QOP_SEL_X_0_ZC] = { "fsel_x_0_zc", 1, 1, false, true },
- [QOP_SEL_X_0_CS] = { "fsel_x_0_cs", 1, 1, false, true },
- [QOP_SEL_X_0_CC] = { "fsel_x_0_cc", 1, 1, false, true },
- [QOP_SEL_X_Y_NS] = { "fsel_x_y_ns", 1, 2, false, true },
- [QOP_SEL_X_Y_NC] = { "fsel_x_y_nc", 1, 2, false, true },
- [QOP_SEL_X_Y_ZS] = { "fsel_x_y_zs", 1, 2, false, true },
- [QOP_SEL_X_Y_ZC] = { "fsel_x_y_zc", 1, 2, false, true },
- [QOP_SEL_X_Y_CS] = { "fsel_x_y_cs", 1, 2, false, true },
- [QOP_SEL_X_Y_CC] = { "fsel_x_y_cc", 1, 2, false, true },
-
[QOP_RCP] = { "rcp", 1, 1, false, true },
[QOP_RSQ] = { "rsq", 1, 1, false, true },
[QOP_EXP2] = { "exp2", 1, 2, false, true },
bool
qir_depends_on_flags(struct qinst *inst)
{
- switch (inst->op) {
- case QOP_SEL_X_0_NS:
- case QOP_SEL_X_0_NC:
- case QOP_SEL_X_0_ZS:
- case QOP_SEL_X_0_ZC:
- case QOP_SEL_X_0_CS:
- case QOP_SEL_X_0_CC:
- case QOP_SEL_X_Y_NS:
- case QOP_SEL_X_Y_NC:
- case QOP_SEL_X_Y_ZS:
- case QOP_SEL_X_Y_ZC:
- case QOP_SEL_X_Y_CS:
- case QOP_SEL_X_Y_CC:
- return true;
- default:
- return false;
- }
+ return (inst->cond != QPU_COND_ALWAYS &&
+ inst->cond != QPU_COND_NEVER);
}
bool
void
qir_dump_inst(struct vc4_compile *c, struct qinst *inst)
{
- fprintf(stderr, "%s%s ",
+ static const char *conditions[] = {
+ [QPU_COND_ALWAYS] = "",
+ [QPU_COND_NEVER] = ".never",
+ [QPU_COND_ZS] = ".zs",
+ [QPU_COND_ZC] = ".zc",
+ [QPU_COND_NS] = ".ns",
+ [QPU_COND_NC] = ".nc",
+ [QPU_COND_CS] = ".cs",
+ [QPU_COND_CC] = ".cc",
+ };
+ fprintf(stderr, "%s%s%s ",
qir_get_op_name(inst->op),
+ conditions[inst->cond],
inst->sf ? ".sf" : "");
qir_print_reg(c, inst->dst, true);
inst->src = calloc(2, sizeof(inst->src[0]));
inst->src[0] = src0;
inst->src[1] = src1;
+ inst->cond = QPU_COND_ALWAYS;
return inst;
}
if (!list_empty(&c->instructions))
last_inst = (struct qinst *)c->instructions.prev;
- if (!last_inst ||
- last_inst->dst.file != src.file ||
- last_inst->dst.index != src.index ||
+ if (src.file != QFILE_TEMP ||
+ !c->defs[src.index] ||
+ last_inst != c->defs[src.index] ||
qir_is_multi_instruction(last_inst)) {
src = qir_MOV(c, src);
last_inst = (struct qinst *)c->instructions.prev;
QOP_XOR,
QOP_NOT,
- /* Note: Orderings of these compares must be the same as in
- * qpu_defines.h. Selects the src[0] if the ns flag bit is set,
- * otherwise 0. */
- QOP_SEL_X_0_ZS,
- QOP_SEL_X_0_ZC,
- QOP_SEL_X_0_NS,
- QOP_SEL_X_0_NC,
- QOP_SEL_X_0_CS,
- QOP_SEL_X_0_CC,
- /* Selects the src[0] if the ns flag bit is set, otherwise src[1]. */
- QOP_SEL_X_Y_ZS,
- QOP_SEL_X_Y_ZC,
- QOP_SEL_X_Y_NS,
- QOP_SEL_X_Y_NC,
- QOP_SEL_X_Y_CS,
- QOP_SEL_X_Y_CC,
-
QOP_FTOI,
QOP_ITOF,
QOP_RCP,
struct qreg dst;
struct qreg *src;
bool sf;
+ uint8_t cond;
};
enum qstage {
uint8_t vattr_sizes[8];
+ /* Bitfield for whether a given channel of a sampler needs sRGB
+ * decode.
+ */
+ uint8_t tex_srgb_decode[VC4_MAX_TEXTURE_SAMPLERS];
+
/**
* Array of the VARYING_SLOT_* of all FS QFILE_VARY reads.
*
struct qreg qir_uniform(struct vc4_compile *c,
enum quniform_contents contents,
uint32_t data);
+void qir_schedule_instructions(struct vc4_compile *c);
void qir_reorder_uniforms(struct vc4_compile *c);
void qir_emit(struct vc4_compile *c, struct qinst *inst);
-static inline void qir_emit_nodef(struct vc4_compile *c, struct qinst *inst)
+static inline struct qinst *
+qir_emit_nodef(struct vc4_compile *c, struct qinst *inst)
{
list_addtail(&inst->link, &c->instructions);
+ return inst;
}
struct qreg qir_get_temp(struct vc4_compile *c);
qir_emit(c, qir_inst(QOP_##name, t, a, c->undef)); \
return t; \
} \
-static inline void \
+static inline struct qinst * \
qir_##name##_dest(struct vc4_compile *c, struct qreg dest, \
struct qreg a) \
{ \
- qir_emit_nodef(c, qir_inst(QOP_##name, dest, a, c->undef)); \
+ return qir_emit_nodef(c, qir_inst(QOP_##name, dest, a, \
+ c->undef)); \
}
#define QIR_ALU2(name) \
QIR_ALU2(V8ADDS)
QIR_ALU2(V8SUBS)
QIR_ALU2(MUL24)
-QIR_ALU1(SEL_X_0_ZS)
-QIR_ALU1(SEL_X_0_ZC)
-QIR_ALU1(SEL_X_0_NS)
-QIR_ALU1(SEL_X_0_NC)
-QIR_ALU1(SEL_X_0_CS)
-QIR_ALU1(SEL_X_0_CC)
-QIR_ALU2(SEL_X_Y_ZS)
-QIR_ALU2(SEL_X_Y_ZC)
-QIR_ALU2(SEL_X_Y_NS)
-QIR_ALU2(SEL_X_Y_NC)
-QIR_ALU2(SEL_X_Y_CS)
-QIR_ALU2(SEL_X_Y_CC)
QIR_ALU2(FMIN)
QIR_ALU2(FMAX)
QIR_ALU2(FMINABS)
QIR_NODST_1(MS_MASK)
static inline struct qreg
+qir_SEL(struct vc4_compile *c, uint8_t cond, struct qreg src0, struct qreg src1)
+{
+ struct qreg t = qir_get_temp(c);
+ struct qinst *a = qir_MOV_dest(c, t, src0);
+ struct qinst *b = qir_MOV_dest(c, t, src1);
+ a->cond = cond;
+ b->cond = cond ^ 1;
+ return t;
+}
+
+static inline struct qreg
qir_UNPACK_8_F(struct vc4_compile *c, struct qreg src, int i)
{
struct qreg t = qir_FMOV(c, src);
--- /dev/null
+/*
+ * Copyright © 2010 Intel Corporation
+ * Copyright © 2014-2015 Broadcom
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+/**
+ * @file vc4_qir_schedule.c
+ *
+ * The basic model of the list scheduler is to take a basic block, compute a
+ * DAG of the dependencies from the bottom up, and make a list of the DAG
+ * heads. Heuristically pick a DAG head and schedule (remove) it, then put
+ * all the parents that are now DAG heads into the list of things to
+ * schedule.
+ *
+ * The goal of scheduling here, before register allocation and conversion to
+ * QPU instructions, is to reduce register pressure by reordering instructions
+ * to consume values when possible.
+ */
+
+#include "vc4_qir.h"
+
+static bool debug;
+
+struct schedule_node {
+ struct list_head link;
+ struct qinst *inst;
+
+ struct schedule_node **children;
+ uint32_t child_count;
+ uint32_t child_array_size;
+ uint32_t parent_count;
+
+ /* Length of the longest (latency) chain from a DAG head to the this
+ * instruction.
+ */
+ uint32_t delay;
+
+ /* Longest time + latency_between(parent, this) of any parent of this
+ * node.
+ */
+ uint32_t unblocked_time;
+};
+
+struct schedule_state {
+ /* List of struct schedule_node *. This starts out with all
+ * instructions, and after dependency updates it's trimmed to be just
+ * the DAG heads.
+ */
+ struct list_head worklist;
+
+ uint32_t time;
+
+ uint32_t *temp_writes;
+
+ BITSET_WORD *temp_live;
+};
+
+/* When walking the instructions in reverse, we need to swap before/after in
+ * add_dep().
+ */
+enum direction { F, R };
+
+/**
+ * Marks a dependency between two intructions, that @after must appear after
+ * @before.
+ *
+ * Our dependencies are tracked as a DAG. Since we're scheduling bottom-up,
+ * the latest instructions with nothing left to schedule are the DAG heads,
+ * and their inputs are their children.
+ */
+static void
+add_dep(enum direction dir,
+ struct schedule_node *before,
+ struct schedule_node *after)
+{
+ if (!before || !after)
+ return;
+
+ assert(before != after);
+
+ if (dir == R) {
+ struct schedule_node *t = before;
+ before = after;
+ after = t;
+ }
+
+ for (int i = 0; i < after->child_count; i++) {
+ if (after->children[i] == after)
+ return;
+ }
+
+ if (after->child_array_size <= after->child_count) {
+ after->child_array_size = MAX2(after->child_array_size * 2, 16);
+ after->children = reralloc(after, after->children,
+ struct schedule_node *,
+ after->child_array_size);
+ }
+
+ after->children[after->child_count] = before;
+ after->child_count++;
+ before->parent_count++;
+}
+
+static void
+add_write_dep(enum direction dir,
+ struct schedule_node **before,
+ struct schedule_node *after)
+{
+ add_dep(dir, *before, after);
+ *before = after;
+}
+
+struct schedule_setup_state {
+ struct schedule_node **last_temp_write;
+ struct schedule_node *last_sf;
+ struct schedule_node *last_vary_read;
+ struct schedule_node *last_vpm_read;
+ struct schedule_node *last_vpm_write;
+ struct schedule_node *last_tex_coord;
+ struct schedule_node *last_tex_result;
+ struct schedule_node *last_tlb;
+ enum direction dir;
+
+ /**
+ * Texture FIFO tracking. This is done top-to-bottom, and is used to
+ * track the QOP_TEX_RESULTs and add dependencies on previous ones
+ * when trying to submit texture coords with TFREQ full or new texture
+ * fetches with TXRCV full.
+ */
+ struct {
+ struct schedule_node *node;
+ int coords;
+ } tex_fifo[8];
+ int tfreq_count; /**< Number of texture coords outstanding. */
+ int tfrcv_count; /**< Number of texture results outstanding. */
+ int tex_fifo_pos;
+};
+
+static void
+block_until_tex_result(struct schedule_setup_state *state, struct schedule_node *n)
+{
+ add_dep(state->dir, state->tex_fifo[0].node, n);
+
+ state->tfreq_count -= state->tex_fifo[0].coords;
+ state->tfrcv_count--;
+
+ memmove(&state->tex_fifo[0],
+ &state->tex_fifo[1],
+ state->tex_fifo_pos * sizeof(state->tex_fifo[0]));
+ state->tex_fifo_pos--;
+}
+
+/**
+ * Common code for dependencies that need to be tracked both forward and
+ * backward.
+ *
+ * This is for things like "all VPM reads have to happen in order."
+ */
+static void
+calculate_deps(struct schedule_setup_state *state, struct schedule_node *n)
+{
+ struct qinst *inst = n->inst;
+ enum direction dir = state->dir;
+
+
+ /* Add deps for temp registers and varyings accesses. Note that we
+ * ignore uniforms accesses, because qir_reorder_uniforms() happens
+ * after this.
+ */
+ for (int i = 0; i < qir_get_op_nsrc(inst->op); i++) {
+ switch (inst->src[i].file) {
+ case QFILE_TEMP:
+ add_dep(dir,
+ state->last_temp_write[inst->src[i].index], n);
+ break;
+
+ case QFILE_VARY:
+ add_write_dep(dir, &state->last_vary_read, n);
+ break;
+
+ case QFILE_VPM:
+ add_write_dep(dir, &state->last_vpm_read, n);
+ break;
+
+ default:
+ break;
+ }
+ }
+
+ switch (inst->op) {
+ case QOP_VARY_ADD_C:
+ add_dep(dir, state->last_vary_read, n);
+ break;
+
+ case QOP_TEX_S:
+ case QOP_TEX_T:
+ case QOP_TEX_R:
+ case QOP_TEX_B:
+ case QOP_TEX_DIRECT:
+ /* Texturing setup gets scheduled in order, because
+ * the uniforms referenced by them have to land in a
+ * specific order.
+ */
+ add_write_dep(dir, &state->last_tex_coord, n);
+ break;
+
+ case QOP_TEX_RESULT:
+ /* Results have to be fetched in order. */
+ add_write_dep(dir, &state->last_tex_result, n);
+ break;
+
+ case QOP_TLB_COLOR_WRITE:
+ case QOP_TLB_COLOR_READ:
+ case QOP_TLB_Z_WRITE:
+ case QOP_TLB_STENCIL_SETUP:
+ case QOP_MS_MASK:
+ add_write_dep(dir, &state->last_tlb, n);
+ break;
+
+ case QOP_TLB_DISCARD_SETUP:
+ add_write_dep(dir, &state->last_sf, n);
+ add_write_dep(dir, &state->last_tlb, n);
+ break;
+
+ default:
+ break;
+ }
+
+ if (inst->dst.file == QFILE_VPM)
+ add_write_dep(dir, &state->last_vpm_write, n);
+ else if (inst->dst.file == QFILE_TEMP)
+ add_write_dep(dir, &state->last_temp_write[inst->dst.index], n);
+
+ if (qir_depends_on_flags(inst))
+ add_dep(dir, state->last_sf, n);
+
+ if (inst->sf)
+ add_write_dep(dir, &state->last_sf, n);
+}
+
+static void
+calculate_forward_deps(struct vc4_compile *c, void *mem_ctx,
+ struct list_head *schedule_list)
+{
+ struct schedule_setup_state state;
+
+ memset(&state, 0, sizeof(state));
+ state.last_temp_write = rzalloc_array(mem_ctx, struct schedule_node *,
+ c->num_temps);
+ state.dir = F;
+
+ list_for_each_entry(struct schedule_node, n, schedule_list, link) {
+ struct qinst *inst = n->inst;
+
+ calculate_deps(&state, n);
+
+ switch (inst->op) {
+ case QOP_TEX_S:
+ case QOP_TEX_T:
+ case QOP_TEX_R:
+ case QOP_TEX_B:
+ case QOP_TEX_DIRECT:
+ /* If the texture coordinate fifo is full,
+ * block this on the last QOP_TEX_RESULT.
+ */
+ if (state.tfreq_count == 8) {
+ block_until_tex_result(&state, n);
+ }
+
+ /* If the texture result fifo is full, block
+ * adding any more to it until the last
+ * QOP_TEX_RESULT.
+ */
+ if (inst->op == QOP_TEX_S ||
+ inst->op == QOP_TEX_DIRECT) {
+ if (state.tfrcv_count == 4)
+ block_until_tex_result(&state, n);
+ state.tfrcv_count++;
+ }
+
+ state.tex_fifo[state.tex_fifo_pos].coords++;
+ state.tfreq_count++;
+ break;
+
+ case QOP_TEX_RESULT:
+ /* Results have to be fetched after the
+ * coordinate setup. Note that we're assuming
+ * here that our input shader has the texture
+ * coord setup and result fetch in order,
+ * which is true initially but not of our
+ * instruction stream after this pass.
+ */
+ add_dep(state.dir, state.last_tex_coord, n);
+
+ state.tex_fifo[state.tex_fifo_pos].node = n;
+
+ state.tex_fifo_pos++;
+ memset(&state.tex_fifo[state.tex_fifo_pos], 0,
+ sizeof(state.tex_fifo[0]));
+ break;
+ default:
+ assert(!qir_is_tex(inst));
+ break;
+ }
+ }
+}
+
+static void
+calculate_reverse_deps(struct vc4_compile *c, void *mem_ctx,
+ struct list_head *schedule_list)
+{
+ struct schedule_setup_state state;
+
+ memset(&state, 0, sizeof(state));
+ state.dir = R;
+ state.last_temp_write = rzalloc_array(mem_ctx, struct schedule_node *,
+ c->num_temps);
+
+ list_for_each_entry_rev(struct schedule_node, n, schedule_list, link) {
+ calculate_deps(&state, n);
+ }
+}
+
+static int
+get_register_pressure_cost(struct schedule_state *state, struct qinst *inst)
+{
+ int cost = 0;
+
+ if (inst->dst.file == QFILE_TEMP &&
+ state->temp_writes[inst->dst.index] == 1)
+ cost--;
+
+ for (int i = 0; i < qir_get_op_nsrc(inst->op); i++) {
+ if (inst->src[i].file == QFILE_TEMP &&
+ !BITSET_TEST(state->temp_live, inst->src[i].index)) {
+ cost++;
+ }
+ }
+
+ return cost;
+}
+
+static bool
+locks_scoreboard(struct qinst *inst)
+{
+ switch (inst->op) {
+ case QOP_TLB_Z_WRITE:
+ case QOP_TLB_COLOR_WRITE:
+ case QOP_TLB_COLOR_WRITE_MS:
+ case QOP_TLB_COLOR_READ:
+ return true;
+ default:
+ return false;
+ }
+}
+
+static struct schedule_node *
+choose_instruction(struct schedule_state *state)
+{
+ struct schedule_node *chosen = NULL;
+
+ list_for_each_entry(struct schedule_node, n, &state->worklist, link) {
+ if (!chosen) {
+ chosen = n;
+ continue;
+ }
+
+ /* Prefer scheduling things that lock the scoreboard, so that
+ * they appear late in the program and we get more parallelism
+ * between shaders on multiple QPUs hitting the same fragment.
+ */
+ if (locks_scoreboard(n->inst) &&
+ !locks_scoreboard(chosen->inst)) {
+ chosen = n;
+ continue;
+ } else if (!locks_scoreboard(n->inst) &&
+ locks_scoreboard(chosen->inst)) {
+ continue;
+ }
+
+ /* If we would block on the previously chosen node, but would
+ * block less on this one, then then prefer it.
+ */
+ if (chosen->unblocked_time > state->time &&
+ n->unblocked_time < chosen->unblocked_time) {
+ chosen = n;
+ continue;
+ } else if (n->unblocked_time > state->time &&
+ n->unblocked_time > chosen->unblocked_time) {
+ continue;
+ }
+
+ /* If we can definitely reduce register pressure, do so
+ * immediately.
+ */
+ int register_pressure_cost =
+ get_register_pressure_cost(state, n->inst);
+ int chosen_register_pressure_cost =
+ get_register_pressure_cost(state, chosen->inst);
+
+ if (register_pressure_cost < chosen_register_pressure_cost) {
+ chosen = n;
+ continue;
+ } else if (register_pressure_cost >
+ chosen_register_pressure_cost) {
+ continue;
+ }
+
+ /* Otherwise, prefer instructions with the deepest chain to
+ * the end of the program. This avoids the problem of
+ * "everything generates a temp, nothing finishes freeing one,
+ * guess I'll just keep emitting varying mul/adds".
+ */
+ if (n->delay > chosen->delay) {
+ chosen = n;
+ continue;
+ } else if (n->delay < chosen->delay) {
+ continue;
+ }
+ }
+
+ return chosen;
+}
+
+static void
+dump_state(struct vc4_compile *c, struct schedule_state *state)
+{
+ uint32_t i = 0;
+ list_for_each_entry(struct schedule_node, n, &state->worklist, link) {
+ fprintf(stderr, "%3d: ", i++);
+ qir_dump_inst(c, n->inst);
+ fprintf(stderr, " (%d cost)\n",
+ get_register_pressure_cost(state, n->inst));
+
+ for (int i = 0; i < n->child_count; i++) {
+ struct schedule_node *child = n->children[i];
+ fprintf(stderr, " - ");
+ qir_dump_inst(c, child->inst);
+ fprintf(stderr, " (%d parents)\n", child->parent_count);
+ }
+ }
+}
+
+/* Estimate of how many instructions we should schedule between operations.
+ *
+ * These aren't in real cycle counts, because we're just estimating cycle
+ * times anyway. QIR instructions will get paired up when turned into QPU
+ * instructions, or extra NOP delays will have to be added due to register
+ * allocation choices.
+ */
+static uint32_t
+latency_between(struct schedule_node *before, struct schedule_node *after)
+{
+ if ((before->inst->op == QOP_TEX_S ||
+ before->inst->op == QOP_TEX_DIRECT) &&
+ after->inst->op == QOP_TEX_RESULT)
+ return 100;
+
+ return 1;
+}
+
+/** Recursive computation of the delay member of a node. */
+static void
+compute_delay(struct schedule_node *n)
+{
+ if (!n->child_count) {
+ /* The color read needs to be scheduled late, to avoid locking
+ * the scoreboard early. This is our best tool for
+ * encouraging that. The other scoreboard locking ops will
+ * have this happen by default, since they are generally the
+ * DAG heads or close to them.
+ */
+ if (n->inst->op == QOP_TLB_COLOR_READ)
+ n->delay = 1000;
+ else
+ n->delay = 1;
+ } else {
+ for (int i = 0; i < n->child_count; i++) {
+ if (!n->children[i]->delay)
+ compute_delay(n->children[i]);
+ n->delay = MAX2(n->delay,
+ n->children[i]->delay +
+ latency_between(n, n->children[i]));
+ }
+ }
+}
+
+static void
+schedule_instructions(struct vc4_compile *c, struct schedule_state *state)
+{
+ if (debug) {
+ fprintf(stderr, "initial deps:\n");
+ dump_state(c, state);
+ }
+
+ /* Remove non-DAG heads from the list. */
+ list_for_each_entry_safe(struct schedule_node, n,
+ &state->worklist, link) {
+ if (n->parent_count != 0)
+ list_del(&n->link);
+ }
+
+ state->time = 0;
+ while (!list_empty(&state->worklist)) {
+ struct schedule_node *chosen = choose_instruction(state);
+ struct qinst *inst = chosen->inst;
+
+ if (debug) {
+ fprintf(stderr, "current list:\n");
+ dump_state(c, state);
+ fprintf(stderr, "chose: ");
+ qir_dump_inst(c, inst);
+ fprintf(stderr, " (%d cost)\n",
+ get_register_pressure_cost(state, inst));
+ }
+
+ state->time = MAX2(state->time, chosen->unblocked_time);
+
+ /* Schedule this instruction back onto the QIR list. */
+ list_del(&chosen->link);
+ list_add(&inst->link, &c->instructions);
+
+ /* Now that we've scheduled a new instruction, some of its
+ * children can be promoted to the list of instructions ready to
+ * be scheduled. Update the children's unblocked time for this
+ * DAG edge as we do so.
+ */
+ for (int i = chosen->child_count - 1; i >= 0; i--) {
+ struct schedule_node *child = chosen->children[i];
+
+ child->unblocked_time = MAX2(child->unblocked_time,
+ state->time +
+ latency_between(chosen,
+ child));
+ child->parent_count--;
+ if (child->parent_count == 0)
+ list_add(&child->link, &state->worklist);
+ }
+
+ /* Update our tracking of register pressure. */
+ for (int i = 0; i < qir_get_op_nsrc(inst->op); i++) {
+ if (inst->src[i].file == QFILE_TEMP)
+ BITSET_SET(state->temp_live, inst->src[i].index);
+ }
+ if (inst->dst.file == QFILE_TEMP) {
+ state->temp_writes[inst->dst.index]--;
+ if (state->temp_writes[inst->dst.index] == 0)
+ BITSET_CLEAR(state->temp_live, inst->dst.index);
+ }
+
+ state->time++;
+ }
+}
+
+void
+qir_schedule_instructions(struct vc4_compile *c)
+{
+ void *mem_ctx = ralloc_context(NULL);
+ struct schedule_state state = { 0 };
+
+ if (debug) {
+ fprintf(stderr, "Pre-schedule instructions\n");
+ qir_dump(c);
+ }
+
+ state.temp_writes = rzalloc_array(mem_ctx, uint32_t, c->num_temps);
+ state.temp_live = rzalloc_array(mem_ctx, BITSET_WORD,
+ BITSET_WORDS(c->num_temps));
+ list_inithead(&state.worklist);
+
+ /* Wrap each instruction in a scheduler structure. */
+ list_for_each_entry_safe(struct qinst, inst, &c->instructions, link) {
+ struct schedule_node *n = rzalloc(mem_ctx, struct schedule_node);
+
+ n->inst = inst;
+ list_del(&inst->link);
+ list_addtail(&n->link, &state.worklist);
+
+ if (inst->dst.file == QFILE_TEMP)
+ state.temp_writes[inst->dst.index]++;
+ }
+
+ /* Dependencies tracked top-to-bottom. */
+ calculate_forward_deps(c, mem_ctx, &state.worklist);
+ /* Dependencies tracked bottom-to-top. */
+ calculate_reverse_deps(c, mem_ctx, &state.worklist);
+
+ list_for_each_entry(struct schedule_node, n, &state.worklist, link)
+ compute_delay(n);
+
+ schedule_instructions(c, &state);
+
+ if (debug) {
+ fprintf(stderr, "Post-schedule instructions\n");
+ qir_dump(c);
+ }
+
+ ralloc_free(mem_ctx);
+}
*last_inst(c) = qpu_set_cond_add(*last_inst(c), cond);
}
+static void
+set_last_cond_mul(struct vc4_compile *c, uint32_t cond)
+{
+ *last_inst(c) = qpu_set_cond_mul(*last_inst(c), cond);
+}
+
/**
* Some special registers can be read from either file, which lets us resolve
* raddr conflicts without extra MOVs.
break;
}
- switch (qinst->op) {
- case QOP_SEL_X_0_ZS:
- case QOP_SEL_X_0_ZC:
- case QOP_SEL_X_0_NS:
- case QOP_SEL_X_0_NC:
- case QOP_SEL_X_0_CS:
- case QOP_SEL_X_0_CC:
- queue(c, qpu_a_MOV(dst, src[0]) | unpack);
- set_last_cond_add(c, qinst->op - QOP_SEL_X_0_ZS +
- QPU_COND_ZS);
-
- queue(c, qpu_a_XOR(dst, qpu_r0(), qpu_r0()));
- set_last_cond_add(c, ((qinst->op - QOP_SEL_X_0_ZS) ^
- 1) + QPU_COND_ZS);
- break;
-
- case QOP_SEL_X_Y_ZS:
- case QOP_SEL_X_Y_ZC:
- case QOP_SEL_X_Y_NS:
- case QOP_SEL_X_Y_NC:
- case QOP_SEL_X_Y_CS:
- case QOP_SEL_X_Y_CC:
- queue(c, qpu_a_MOV(dst, src[0]));
- if (qinst->src[0].pack)
- *(last_inst(c)) |= unpack;
- set_last_cond_add(c, qinst->op - QOP_SEL_X_Y_ZS +
- QPU_COND_ZS);
-
- queue(c, qpu_a_MOV(dst, src[1]));
- if (qinst->src[1].pack)
- *(last_inst(c)) |= unpack;
- set_last_cond_add(c, ((qinst->op - QOP_SEL_X_Y_ZS) ^
- 1) + QPU_COND_ZS);
-
- break;
+ bool handled_qinst_cond = true;
+ switch (qinst->op) {
case QOP_RCP:
case QOP_RSQ:
case QOP_EXP2:
queue(c, qpu_m_alu2(translate[qinst->op].op,
dst,
src[0], src[1]) | unpack);
+ set_last_cond_mul(c, qinst->cond);
} else {
queue(c, qpu_a_alu2(translate[qinst->op].op,
dst,
src[0], src[1]) | unpack);
+ set_last_cond_add(c, qinst->cond);
}
+ handled_qinst_cond = true;
set_last_dst_pack(c, qinst);
break;
}
+ assert(qinst->cond == QPU_COND_ALWAYS ||
+ handled_qinst_cond);
+
if (qinst->sf) {
assert(!qir_is_multi_instruction(qinst));
*last_inst(c) |= QPU_SF;
uint32_t child_array_size;
uint32_t parent_count;
- /* Longest cycles + n->latency of any parent of this node. */
+ /* Longest cycles + instruction_latency() of any parent of this node. */
uint32_t unblocked_time;
/**
}
} else if (is_tmu_write(waddr)) {
add_write_dep(state, &state->last_tmu_write, n);
- } else if (qpu_waddr_is_tlb(waddr)) {
+ } else if (qpu_waddr_is_tlb(waddr) ||
+ waddr == QPU_W_MS_FLAGS) {
add_write_dep(state, &state->last_tlb, n);
} else {
switch (waddr) {
}
}
+static uint32_t waddr_latency(uint32_t waddr, uint64_t after)
+{
+ if (waddr < 32)
+ return 2;
+
+ /* Apply some huge latency between texture fetch requests and getting
+ * their results back.
+ */
+ if (waddr == QPU_W_TMU0_S) {
+ if (QPU_GET_FIELD(after, QPU_SIG) == QPU_SIG_LOAD_TMU0)
+ return 100;
+ }
+ if (waddr == QPU_W_TMU1_S) {
+ if (QPU_GET_FIELD(after, QPU_SIG) == QPU_SIG_LOAD_TMU1)
+ return 100;
+ }
+
+ switch(waddr) {
+ case QPU_W_SFU_RECIP:
+ case QPU_W_SFU_RECIPSQRT:
+ case QPU_W_SFU_EXP:
+ case QPU_W_SFU_LOG:
+ return 3;
+ default:
+ return 1;
+ }
+}
+
+static uint32_t
+instruction_latency(struct schedule_node *before, struct schedule_node *after)
+{
+ uint64_t before_inst = before->inst->inst;
+ uint64_t after_inst = after->inst->inst;
+
+ return MAX2(waddr_latency(QPU_GET_FIELD(before_inst, QPU_WADDR_ADD),
+ after_inst),
+ waddr_latency(QPU_GET_FIELD(before_inst, QPU_WADDR_MUL),
+ after_inst));
+}
+
/** Recursive computation of the delay member of a node. */
static void
compute_delay(struct schedule_node *n)
if (!n->children[i].node->delay)
compute_delay(n->children[i].node);
n->delay = MAX2(n->delay,
- n->children[i].node->delay + n->latency);
+ n->children[i].node->delay +
+ instruction_latency(n, n->children[i].node));
}
}
}
* immediately after (or paired with!) the thing reading the
* destination.
*/
- int latency_from_previous = war_only ? 0 : node->latency;
+ uint32_t latency = 0;
+ if (!war_only) {
+ latency = instruction_latency(node,
+ node->children[i].node);
+ }
+
child->unblocked_time = MAX2(child->unblocked_time,
- time + latency_from_previous);
+ time + latency);
child->parent_count--;
if (child->parent_count == 0)
list_add(&child->link, schedule_list);
return time;
}
-static uint32_t waddr_latency(uint32_t waddr)
-{
- if (waddr < 32)
- return 2;
-
- /* Some huge number, really. */
- if (waddr >= QPU_W_TMU0_S && waddr <= QPU_W_TMU1_B)
- return 100;
-
- switch(waddr) {
- case QPU_W_SFU_RECIP:
- case QPU_W_SFU_RECIPSQRT:
- case QPU_W_SFU_EXP:
- case QPU_W_SFU_LOG:
- return 3;
- default:
- return 1;
- }
-}
-
-static uint32_t
-instruction_latency(uint64_t inst)
-{
- return MAX2(waddr_latency(QPU_GET_FIELD(inst, QPU_WADDR_ADD)),
- waddr_latency(QPU_GET_FIELD(inst, QPU_WADDR_MUL)));
-}
-
uint32_t
qpu_schedule_instructions(struct vc4_compile *c)
{
struct schedule_node *n = rzalloc(mem_ctx, struct schedule_node);
n->inst = inst;
- n->latency = instruction_latency(inst->inst);
if (reads_uniform(inst->inst)) {
n->uniform = next_uniform++;
/* If the resource is multisampled, we need to resolve to single
* sample. This seems like it should be handled at a higher layer.
*/
- if (prsc->nr_samples) {
+ if (prsc->nr_samples > 1) {
trans->ss_resource = vc4_get_temp_resource(pctx, prsc, box);
if (!trans->ss_resource)
goto fail;
if (!rsc->tiled) {
slice->tiling = VC4_TILING_FORMAT_LINEAR;
- if (prsc->nr_samples) {
+ if (prsc->nr_samples > 1) {
/* MSAA (4x) surfaces are stored as raw tile buffer contents. */
level_width = align(level_width, 32);
level_height = align(level_height, 32);
prsc->screen = pscreen;
rsc->base.vtbl = &vc4_resource_vtbl;
- if (prsc->nr_samples == 0)
+ if (prsc->nr_samples <= 1)
rsc->cpp = util_format_get_blocksize(tmpl->format);
else
rsc->cpp = sizeof(uint32_t);
uint8_t format = vc4_get_tex_format(prsc->format);
if (!rsc->tiled) {
- if (prsc->nr_samples) {
+ if (prsc->nr_samples > 1) {
return ~0;
} else {
assert(format == VC4_TEXTURE_TYPE_RGBA8888);
* communicate metadata about tiling currently.
*/
if (tmpl->target == PIPE_BUFFER ||
- tmpl->nr_samples ||
+ tmpl->nr_samples > 1 ||
(tmpl->bind & (PIPE_BIND_SCANOUT |
PIPE_BIND_LINEAR |
PIPE_BIND_SHARED |
if (!psurf)
return;
- if (psurf->texture->nr_samples)
+ if (psurf->texture->nr_samples > 1)
vc4_dump_surface_msaa(psurf);
else
vc4_dump_surface_non_msaa(psurf);
void *data;
struct pipe_resource *shadow_rsc = NULL;
- u_upload_alloc(vc4->uploader, 0, count * 2,
+ u_upload_alloc(vc4->uploader, 0, count * 2, 4,
shadow_offset, &shadow_rsc, &data);
uint16_t *dst = data;
"Flush after each draw call" },
{ "always_sync", VC4_DEBUG_ALWAYS_SYNC,
"Wait for finish after each flush" },
+#if USE_VC4_SIMULATOR
+ { "dump", VC4_DEBUG_DUMP,
+ "Write a GPU command stream trace file" },
+#endif
{ NULL }
};
case PIPE_CAP_TWO_SIDED_STENCIL:
case PIPE_CAP_USER_INDEX_BUFFERS:
case PIPE_CAP_TEXTURE_MULTISAMPLE:
+ case PIPE_CAP_TEXTURE_SWIZZLE:
return 1;
/* lying for GL 2.0 */
case PIPE_CAP_TEXTURE_BUFFER_OBJECTS:
case PIPE_CAP_CUBE_MAP_ARRAY:
case PIPE_CAP_TEXTURE_MIRROR_CLAMP:
- case PIPE_CAP_TEXTURE_SWIZZLE:
case PIPE_CAP_VERTEX_ELEMENT_INSTANCE_DIVISOR:
case PIPE_CAP_MIXED_COLORBUFFER_FORMATS:
case PIPE_CAP_SEAMLESS_CUBE_MAP:
case PIPE_CAP_MAX_TEXEL_OFFSET:
case PIPE_CAP_MAX_VERTEX_STREAMS:
case PIPE_CAP_DRAW_INDIRECT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT_PARAMS:
case PIPE_CAP_TGSI_FS_FINE_DERIVATIVE:
case PIPE_CAP_CONDITIONAL_RENDER_INVERTED:
case PIPE_CAP_SAMPLER_VIEW_TARGET:
case PIPE_CAP_MULTISAMPLE_Z_RESOLVE:
case PIPE_CAP_RESOURCE_FROM_USER_MEMORY:
case PIPE_CAP_DEVICE_RESET_STATUS_QUERY:
- case PIPE_CAP_MAX_SHADER_PATCH_VARYINGS:
- case PIPE_CAP_TEXTURE_FLOAT_LINEAR:
- case PIPE_CAP_TEXTURE_HALF_FLOAT_LINEAR:
- case PIPE_CAP_DEPTH_BOUNDS_TEST:
- case PIPE_CAP_TGSI_TXQS:
- case PIPE_CAP_FORCE_PERSAMPLE_INTERP:
- case PIPE_CAP_SHAREABLE_SHADERS:
- case PIPE_CAP_COPY_BETWEEN_COMPRESSED_AND_PLAIN_FORMATS:
- case PIPE_CAP_CLEAR_TEXTURE:
+ case PIPE_CAP_MAX_SHADER_PATCH_VARYINGS:
+ case PIPE_CAP_TEXTURE_FLOAT_LINEAR:
+ case PIPE_CAP_TEXTURE_HALF_FLOAT_LINEAR:
+ case PIPE_CAP_DEPTH_BOUNDS_TEST:
+ case PIPE_CAP_TGSI_TXQS:
+ case PIPE_CAP_FORCE_PERSAMPLE_INTERP:
+ case PIPE_CAP_SHAREABLE_SHADERS:
+ case PIPE_CAP_COPY_BETWEEN_COMPRESSED_AND_PLAIN_FORMATS:
+ case PIPE_CAP_CLEAR_TEXTURE:
+ case PIPE_CAP_DRAW_PARAMETERS:
+ case PIPE_CAP_TGSI_PACK_HALF_FLOAT:
+ case PIPE_CAP_TGSI_FS_POSITION_IS_SYSVAL:
+ case PIPE_CAP_TGSI_FS_FACE_IS_INTEGER_SYSVAL:
+ case PIPE_CAP_SHADER_BUFFER_OFFSET_ALIGNMENT:
+ case PIPE_CAP_INVALIDATE_BUFFER:
+ case PIPE_CAP_GENERATE_MIPMAP:
return 0;
/* Stream output. */
return PIPE_SHADER_IR_TGSI;
case PIPE_SHADER_CAP_MAX_UNROLL_ITERATIONS_HINT:
return 32;
+ case PIPE_SHADER_CAP_MAX_SHADER_BUFFERS:
+ return 0;
default:
fprintf(stderr, "unknown shader param %d\n", param);
return 0;
#define VC4_DEBUG_ALWAYS_FLUSH 0x0080
#define VC4_DEBUG_ALWAYS_SYNC 0x0100
#define VC4_DEBUG_NIR 0x0200
+#define VC4_DEBUG_DUMP 0x0400
#define VC4_MAX_MIP_LEVELS 12
#define VC4_MAX_TEXTURE_SAMPLERS 16
return 0;
}
+static void
+vc4_dump_to_file(struct vc4_exec_info *exec)
+{
+ static int dumpno = 0;
+ struct drm_vc4_get_hang_state *state;
+ struct drm_vc4_get_hang_state_bo *bo_state;
+ unsigned int dump_version = 0;
+
+ if (!(vc4_debug & VC4_DEBUG_DUMP))
+ return;
+
+ state = calloc(1, sizeof(*state));
+
+ int unref_count = 0;
+ list_for_each_entry_safe(struct drm_vc4_bo, bo, &exec->unref_list,
+ unref_head) {
+ unref_count++;
+ }
+
+ /* Add one more for the overflow area that isn't wrapped in a BO. */
+ state->bo_count = exec->bo_count + unref_count + 1;
+ bo_state = calloc(state->bo_count, sizeof(*bo_state));
+
+ char *filename = NULL;
+ asprintf(&filename, "vc4-dri-%d.dump", dumpno++);
+ FILE *f = fopen(filename, "w+");
+ if (!f) {
+ fprintf(stderr, "Couldn't open %s: %s", filename,
+ strerror(errno));
+ return;
+ }
+
+ fwrite(&dump_version, sizeof(dump_version), 1, f);
+
+ state->ct0ca = exec->ct0ca;
+ state->ct0ea = exec->ct0ea;
+ state->ct1ca = exec->ct1ca;
+ state->ct1ea = exec->ct1ea;
+ state->start_bin = exec->ct0ca;
+ state->start_render = exec->ct1ca;
+ fwrite(state, sizeof(*state), 1, f);
+
+ int i;
+ for (i = 0; i < exec->bo_count; i++) {
+ struct drm_gem_cma_object *cma_bo = exec->bo[i];
+ bo_state[i].handle = i; /* Not used by the parser. */
+ bo_state[i].paddr = cma_bo->paddr;
+ bo_state[i].size = cma_bo->base.size;
+ }
+
+ list_for_each_entry_safe(struct drm_vc4_bo, bo, &exec->unref_list,
+ unref_head) {
+ struct drm_gem_cma_object *cma_bo = &bo->base;
+ bo_state[i].handle = 0;
+ bo_state[i].paddr = cma_bo->paddr;
+ bo_state[i].size = cma_bo->base.size;
+ i++;
+ }
+
+ /* Add the static overflow memory area. */
+ bo_state[i].handle = exec->bo_count;
+ bo_state[i].paddr = 0;
+ bo_state[i].size = OVERFLOW_SIZE;
+ i++;
+
+ fwrite(bo_state, sizeof(*bo_state), state->bo_count, f);
+
+ for (int i = 0; i < exec->bo_count; i++) {
+ struct drm_gem_cma_object *cma_bo = exec->bo[i];
+ fwrite(cma_bo->vaddr, cma_bo->base.size, 1, f);
+ }
+
+ list_for_each_entry_safe(struct drm_vc4_bo, bo, &exec->unref_list,
+ unref_head) {
+ struct drm_gem_cma_object *cma_bo = &bo->base;
+ fwrite(cma_bo->vaddr, cma_bo->base.size, 1, f);
+ }
+
+ void *overflow = calloc(1, OVERFLOW_SIZE);
+ fwrite(overflow, 1, OVERFLOW_SIZE, f);
+ free(overflow);
+
+ free(state);
+ free(bo_state);
+ fclose(f);
+}
+
int
vc4_simulator_flush(struct vc4_context *vc4, struct drm_vc4_submit_cl *args)
{
exec.ct1ea - exec.ct1ca, true);
}
+ vc4_dump_to_file(&exec);
+
if (exec.ct0ca != exec.ct0ea) {
int bfc = simpenrose_do_binning(exec.ct0ca, exec.ct0ea);
if (bfc != 1) {
#define roundup(x, y) align(x, y)
#define round_up(x, y) align(x, y)
#define max(x, y) MAX2(x, y)
-#define min(x, y) MiN2(x, y)
+#define min(x, y) MIN2(x, y)
#define BUG_ON(condition) assert(!(condition))
static inline int
vc4->msaa = false;
if (cso->cbufs[0])
- vc4->msaa = cso->cbufs[0]->texture->nr_samples != 0;
+ vc4->msaa = cso->cbufs[0]->texture->nr_samples > 1;
else if (cso->zsbuf)
- vc4->msaa = cso->zsbuf->texture->nr_samples != 0;
+ vc4->msaa = cso->zsbuf->texture->nr_samples > 1;
if (vc4->msaa) {
vc4->tile_width = 32;
ib.offset = vctx->index_buffer.offset + info.start * ib.index_size;
if (ib.user_buffer) {
- u_upload_data(vctx->uploader, 0, info.count * ib.index_size,
+ u_upload_data(vctx->uploader, 0, info.count * ib.index_size, 256,
ib.user_buffer, &ib.offset, &ib.buffer);
ib.user_buffer = NULL;
}
16, UTIL_SLAB_SINGLETHREADED);
vctx->primconvert = util_primconvert_create(&vctx->base, rs->caps.caps.v1.prim_mask);
- vctx->uploader = u_upload_create(&vctx->base, 1024 * 1024, 256,
- PIPE_BIND_INDEX_BUFFER);
+ vctx->uploader = u_upload_create(&vctx->base, 1024 * 1024,
+ PIPE_BIND_INDEX_BUFFER, PIPE_USAGE_STREAM);
if (!vctx->uploader)
goto fail;
case PIPE_CAP_TGSI_VS_WINDOW_SPACE_POSITION:
case PIPE_CAP_MAX_VERTEX_STREAMS:
case PIPE_CAP_DRAW_INDIRECT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT:
+ case PIPE_CAP_MULTI_DRAW_INDIRECT_PARAMS:
case PIPE_CAP_TGSI_FS_FINE_DERIVATIVE:
case PIPE_CAP_CONDITIONAL_RENDER_INVERTED:
case PIPE_CAP_MAX_VERTEX_ATTRIB_STRIDE:
case PIPE_CAP_FORCE_PERSAMPLE_INTERP:
case PIPE_CAP_SHAREABLE_SHADERS:
case PIPE_CAP_CLEAR_TEXTURE:
+ case PIPE_CAP_DRAW_PARAMETERS:
+ case PIPE_CAP_TGSI_PACK_HALF_FLOAT:
+ case PIPE_CAP_TGSI_FS_POSITION_IS_SYSVAL:
+ case PIPE_CAP_TGSI_FS_FACE_IS_INTEGER_SYSVAL:
+ case PIPE_CAP_SHADER_BUFFER_OFFSET_ALIGNMENT:
+ case PIPE_CAP_INVALIDATE_BUFFER:
+ case PIPE_CAP_GENERATE_MIPMAP:
return 0;
case PIPE_CAP_VENDOR_ID:
return 0x1af4;
struct pipe_resource *resource);
/**
- * Invalidate the contents of the resource.
+ * Invalidate the contents of the resource. This is used to
*
- * This is used to implement EGL's semantic of undefined depth/stencil
+ * (1) implement EGL's semantic of undefined depth/stencil
* contenst after a swapbuffers. This allows a tiled renderer (for
* example) to not store the depth buffer.
+ *
+ * (2) implement GL's InvalidateBufferData. For backwards compatibility,
+ * you must only rely on the usability for this purpose when
+ * PIPE_CAP_INVALIDATE_BUFFER is enabled.
*/
void (*invalidate_resource)(struct pipe_context *ctx,
struct pipe_resource *resource);
*/
void (*dump_debug_state)(struct pipe_context *ctx, FILE *stream,
unsigned flags);
+
+ /**
+ * Generate mipmap.
+ * \return TRUE if mipmap generation succeeds, FALSE otherwise
+ */
+ boolean (*generate_mipmap)(struct pipe_context *ctx,
+ struct pipe_resource *resource,
+ enum pipe_format format,
+ unsigned base_level,
+ unsigned last_level,
+ unsigned first_layer,
+ unsigned last_layer);
};
PIPE_CAP_SHAREABLE_SHADERS,
PIPE_CAP_COPY_BETWEEN_COMPRESSED_AND_PLAIN_FORMATS,
PIPE_CAP_CLEAR_TEXTURE,
+ PIPE_CAP_DRAW_PARAMETERS,
+ PIPE_CAP_TGSI_PACK_HALF_FLOAT,
+ PIPE_CAP_MULTI_DRAW_INDIRECT,
+ PIPE_CAP_MULTI_DRAW_INDIRECT_PARAMS,
+ PIPE_CAP_TGSI_FS_POSITION_IS_SYSVAL,
+ PIPE_CAP_TGSI_FS_FACE_IS_INTEGER_SYSVAL,
+ PIPE_CAP_SHADER_BUFFER_OFFSET_ALIGNMENT,
+ PIPE_CAP_INVALIDATE_BUFFER,
+ PIPE_CAP_GENERATE_MIPMAP,
};
#define PIPE_QUIRK_TEXTURE_BORDER_COLOR_SWIZZLE_NV50 (1 << 0)
PIPE_SHADER_CAP_TGSI_FMA_SUPPORTED,
PIPE_SHADER_CAP_TGSI_ANY_INOUT_DECL_RANGE,
PIPE_SHADER_CAP_MAX_UNROLL_ITERATIONS_HINT,
+ PIPE_SHADER_CAP_MAX_SHADER_BUFFERS,
};
/**
TGSI_FILE_IMMEDIATE =7,
TGSI_FILE_PREDICATE =8,
TGSI_FILE_SYSTEM_VALUE =9,
- TGSI_FILE_RESOURCE =10,
+ TGSI_FILE_IMAGE =10,
TGSI_FILE_SAMPLER_VIEW =11,
+ TGSI_FILE_BUFFER =12,
TGSI_FILE_COUNT /**< how many TGSI_FILE_ types */
};
unsigned Invariant : 1; /**< invariant optimization? */
unsigned Local : 1; /**< optimize as subroutine local variable? */
unsigned Array : 1; /**< extra array info? */
- unsigned Padding : 6;
+ unsigned Atomic : 1; /**< atomic only? for TGSI_FILE_BUFFER */
+ unsigned Padding : 5;
};
struct tgsi_declaration_range
#define TGSI_SEMANTIC_TESSINNER 33 /**< inner tessellation levels */
#define TGSI_SEMANTIC_VERTICESIN 34 /**< number of input vertices */
#define TGSI_SEMANTIC_HELPER_INVOCATION 35 /**< current invocation is helper */
-#define TGSI_SEMANTIC_COUNT 36 /**< number of semantic values */
+#define TGSI_SEMANTIC_BASEINSTANCE 36
+#define TGSI_SEMANTIC_DRAWID 37
+#define TGSI_SEMANTIC_COUNT 38 /**< number of semantic values */
struct tgsi_declaration_semantic
{
unsigned Padding : 8;
};
-struct tgsi_declaration_resource {
+struct tgsi_declaration_image {
unsigned Resource : 8; /**< one of TGSI_TEXTURE_ */
unsigned Raw : 1;
unsigned Writable : 1;
- unsigned Padding : 22;
+ unsigned Format : 10; /**< one of PIPE_FORMAT_ */
+ unsigned Padding : 12;
};
enum tgsi_return_type {
#define TGSI_OPCODE_ENDSUB 102
#define TGSI_OPCODE_TXQ_LZ 103 /* TXQ for mipmap level 0 */
#define TGSI_OPCODE_TXQS 104
+#define TGSI_OPCODE_RESQ 105
/* gap */
#define TGSI_OPCODE_NOP 107
unsigned Predicate : 1; /* BOOL */
unsigned Label : 1;
unsigned Texture : 1;
- unsigned Padding : 2;
+ unsigned Memory : 1;
+ unsigned Padding : 1;
};
/*
unsigned Padding : 6;
};
+#define TGSI_MEMORY_COHERENT (1 << 0)
+#define TGSI_MEMORY_RESTRICT (1 << 1)
+#define TGSI_MEMORY_VOLATILE (1 << 2)
+
+/**
+ * Specifies the type of memory access to do for the LOAD/STORE instruction.
+ */
+struct tgsi_instruction_memory
+{
+ unsigned Qualifier : 3; /* TGSI_MEMORY_ */
+ unsigned Padding : 29;
+};
+
#ifdef __cplusplus
}
unsigned start_instance; /**< first instance id */
unsigned instance_count; /**< number of instances */
+ unsigned drawid; /**< id of this draw in a multidraw */
+
unsigned vertices_per_patch; /**< the number of vertices per patch */
/**
*/
struct pipe_stream_output_target *count_from_stream_output;
- /* Indirect parameters resource: If not NULL, most values are taken
+ /* Indirect draw parameters resource: If not NULL, most values are taken
* from this buffer instead, which is laid out as follows:
*
* if indexed is TRUE:
*/
struct pipe_resource *indirect;
unsigned indirect_offset; /**< must be 4 byte aligned */
+ unsigned indirect_stride; /**< must be 4 byte aligned */
+ unsigned indirect_count; /**< number of indirect draws */
+
+ /* Indirect draw count resource: If not NULL, contains a 32-bit value which
+ * is to be used as the real indirect_count. In that case indirect_count
+ * becomes the maximum possible value.
+ */
+ struct pipe_resource *indirect_params;
+ unsigned indirect_params_offset; /**< must be 4 byte aligned */
};
This->driver_caps.user_cbufs = GET_PCAP(USER_CONSTANT_BUFFERS);
if (!This->driver_caps.user_vbufs)
- This->vertex_uploader = u_upload_create(This->pipe, 65536, 4, PIPE_BIND_VERTEX_BUFFER);
+ This->vertex_uploader = u_upload_create(This->pipe, 65536,
+ PIPE_BIND_VERTEX_BUFFER, PIPE_USAGE_STREAM);
if (!This->driver_caps.user_ibufs)
- This->index_uploader = u_upload_create(This->pipe, 128 * 1024, 4, PIPE_BIND_INDEX_BUFFER);
+ This->index_uploader = u_upload_create(This->pipe, 128 * 1024,
+ PIPE_BIND_INDEX_BUFFER, PIPE_USAGE_STREAM);
if (!This->driver_caps.user_cbufs) {
- unsigned alignment = GET_PCAP(CONSTANT_BUFFER_OFFSET_ALIGNMENT);
-
+ This->constbuf_alignment = GET_PCAP(CONSTANT_BUFFER_OFFSET_ALIGNMENT);
This->constbuf_uploader = u_upload_create(This->pipe, This->vs_const_size,
- alignment, PIPE_BIND_CONSTANT_BUFFER);
+ PIPE_BIND_CONSTANT_BUFFER, PIPE_USAGE_STREAM);
}
This->driver_caps.window_space_position_support = GET_PCAP(TGSI_VS_WINDOW_SPACE_POSITION);
u_upload_data(This->vertex_uploader,
0,
(info.max_index + 1) * VertexStreamZeroStride, /* XXX */
+ 4,
vtxbuf.user_buffer,
&vtxbuf.buffer_offset,
&vtxbuf.buffer);
base,
(info.max_index -
info.min_index + 1) * VertexStreamZeroStride, /* XXX */
+ 4,
(const uint8_t *)vbuf.user_buffer + base,
&vbuf.buffer_offset,
&vbuf.buffer);
u_upload_data(This->index_uploader,
0,
info.count * ibuf.index_size,
+ 4,
ibuf.user_buffer,
&ibuf.offset,
&ibuf.buffer);
struct u_upload_mgr *vertex_uploader;
struct u_upload_mgr *index_uploader;
struct u_upload_mgr *constbuf_uploader;
+ unsigned constbuf_alignment;
struct nine_range_pool range_pool;
/* Fog.
*/
if (key->fog_mode) {
- struct ureg_src vPos = ureg_DECL_fs_input(ureg, TGSI_SEMANTIC_POSITION, 0, TGSI_INTERPOLATE_LINEAR);
+ struct ureg_src vPos;
+ if (device->screen->get_param(device->screen,
+ PIPE_CAP_TGSI_FS_POSITION_IS_SYSVAL)) {
+ vPos = ureg_DECL_system_value(ureg, TGSI_SEMANTIC_POSITION, 0);
+ } else {
+ vPos = ureg_DECL_fs_input(ureg, TGSI_SEMANTIC_POSITION, 0,
+ TGSI_INTERPOLATE_LINEAR);
+ }
+
struct ureg_dst rFog = ureg_writemask(ps.rTmp, TGSI_WRITEMASK_X);
if (key->fog_mode == D3DFOG_EXP) {
ureg_MUL(ureg, rFog, _ZZZZ(vPos), _ZZZZ(_CONST(22)));
u_upload_data(device->constbuf_uploader,
0,
cb.buffer_size,
+ device->constbuf_alignment,
cb.user_buffer,
&cb.buffer_offset,
&cb.buffer);
u_upload_data(device->constbuf_uploader,
0,
cb.buffer_size,
+ device->constbuf_alignment,
cb.user_buffer,
&cb.buffer_offset,
&cb.buffer);
boolean lower_preds;
boolean want_texcoord;
boolean shift_wpos;
+ boolean wpos_is_sysval;
+ boolean face_is_sysval_integer;
unsigned texcoord_sn;
struct sm1_instruction insn; /* current instruction */
case D3DSPR_MISCTYPE:
switch (param->idx) {
case D3DSMO_POSITION:
- if (ureg_src_is_undef(tx->regs.vPos))
- tx->regs.vPos = ureg_DECL_fs_input(ureg,
- TGSI_SEMANTIC_POSITION, 0,
- TGSI_INTERPOLATE_LINEAR);
+ if (ureg_src_is_undef(tx->regs.vPos)) {
+ if (tx->wpos_is_sysval) {
+ tx->regs.vPos =
+ ureg_DECL_system_value(ureg, TGSI_SEMANTIC_POSITION, 0);
+ } else {
+ tx->regs.vPos =
+ ureg_DECL_fs_input(ureg, TGSI_SEMANTIC_POSITION, 0,
+ TGSI_INTERPOLATE_LINEAR);
+ }
+ }
if (tx->shift_wpos) {
/* TODO: do this only once */
struct ureg_dst wpos = tx_scratch(tx);
break;
case D3DSMO_FACE:
if (ureg_src_is_undef(tx->regs.vFace)) {
- tx->regs.vFace = ureg_DECL_fs_input(ureg,
- TGSI_SEMANTIC_FACE, 0,
- TGSI_INTERPOLATE_CONSTANT);
+ if (tx->face_is_sysval_integer) {
+ tmp = tx_scratch(tx);
+ tx->regs.vFace =
+ ureg_DECL_system_value(ureg, TGSI_SEMANTIC_FACE, 0);
+
+ /* convert bool to float */
+ ureg_UCMP(ureg, tmp, ureg_scalar(tx->regs.vFace, TGSI_SWIZZLE_X),
+ ureg_imm1f(ureg, 1), ureg_imm1f(ureg, -1));
+ tx->regs.vFace = ureg_src(tmp);
+ } else {
+ tx->regs.vFace = ureg_DECL_fs_input(ureg,
+ TGSI_SEMANTIC_FACE, 0,
+ TGSI_INTERPOLATE_CONSTANT);
+ }
tx->regs.vFace = ureg_scalar(tx->regs.vFace, TGSI_SWIZZLE_X);
}
src = tx->regs.vFace;
return;
}
- if (tx->info->fog_mode != D3DFOG_NONE)
- depth = ureg_scalar(ureg_DECL_fs_input(ureg, TGSI_SEMANTIC_POSITION, 0,
- TGSI_INTERPOLATE_LINEAR),
- TGSI_SWIZZLE_Z);
+ if (tx->info->fog_mode != D3DFOG_NONE) {
+ if (tx->wpos_is_sysval) {
+ depth = ureg_DECL_system_value(ureg, TGSI_SEMANTIC_POSITION, 0);
+ } else {
+ depth = ureg_DECL_fs_input(ureg, TGSI_SEMANTIC_POSITION, 0,
+ TGSI_INTERPOLATE_LINEAR);
+ }
+ depth = ureg_scalar(depth, TGSI_SWIZZLE_Z);
+ }
nine_info_mark_const_f_used(tx->info, 33);
fog_color = NINE_CONSTANT_SRC(32);
tx->shift_wpos = !GET_CAP(TGSI_FS_COORD_PIXEL_CENTER_INTEGER);
tx->texcoord_sn = tx->want_texcoord ?
TGSI_SEMANTIC_TEXCOORD : TGSI_SEMANTIC_GENERIC;
+ tx->wpos_is_sysval = GET_CAP(TGSI_FS_POSITION_IS_SYSVAL);
+ tx->face_is_sysval_integer = GET_CAP(TGSI_FS_FACE_IS_INTEGER_SYSVAL);
if (IS_VS) {
tx->num_constf_allowed = NINE_MAX_CONST_F;
u_upload_data(device->constbuf_uploader,
0,
cb.buffer_size,
+ device->constbuf_alignment,
cb.user_buffer,
&cb.buffer_offset,
&cb.buffer);
u_upload_data(device->constbuf_uploader,
0,
cb.buffer_size,
+ device->constbuf_alignment,
cb.user_buffer,
&cb.buffer_offset,
&cb.buffer);
case D3DSWAPEFFECT_FLIP:
UNTESTED(4);
case D3DSWAPEFFECT_DISCARD:
- /* rotate the queue */;
+ /* rotate the queue */
pipe_resource_reference(&res, This->buffers[0]->base.resource);
for (i = 1; i <= This->params.BackBufferCount; i++) {
NineSurface9_SetResourceResize(This->buffers[i - 1],
priv->state = OMX_StateInvalid;
tsem_up(priv->messageSem);
- /* wait for thread to exit */;
+ /* wait for thread to exit */
pthread_join(priv->messageHandlerThread, NULL);
return omx_base_component_Destructor(comp);
{
struct encode_task *i, *next;
+ if (!head)
+ return;
+
LIST_FOR_EACH_ENTRY_SAFE(i, next, head, list) {
pipe_resource_reference(&i->bitstream, NULL);
i->buf->destroy(i->buf);
OSMesaCreateContextExt(GLenum format, GLint depthBits, GLint stencilBits,
GLint accumBits, OSMesaContext sharelist)
{
+ int attribs[100], n = 0;
+
+ attribs[n++] = OSMESA_FORMAT;
+ attribs[n++] = format;
+ attribs[n++] = OSMESA_DEPTH_BITS;
+ attribs[n++] = depthBits;
+ attribs[n++] = OSMESA_STENCIL_BITS;
+ attribs[n++] = stencilBits;
+ attribs[n++] = OSMESA_ACCUM_BITS;
+ attribs[n++] = accumBits;
+ attribs[n++] = 0;
+
+ return OSMesaCreateContextAttribs(attribs, sharelist);
+}
+
+
+/**
+ * New in Mesa 11.2
+ *
+ * Create context with attribute list.
+ */
+GLAPI OSMesaContext GLAPIENTRY
+OSMesaCreateContextAttribs(const int *attribList, OSMesaContext sharelist)
+{
OSMesaContext osmesa;
struct st_context_iface *st_shared;
enum st_context_error st_error = 0;
struct st_context_attribs attribs;
struct st_api *stapi = get_st_api();
+ GLenum format = GL_RGBA;
+ int depthBits = 0, stencilBits = 0, accumBits = 0;
+ int profile = OSMESA_COMPAT_PROFILE, version_major = 1, version_minor = 0;
+ int i;
if (sharelist) {
st_shared = sharelist->stctx;
st_shared = NULL;
}
+ for (i = 0; attribList[i]; i += 2) {
+ switch (attribList[i]) {
+ case OSMESA_FORMAT:
+ format = attribList[i+1];
+ switch (format) {
+ case OSMESA_COLOR_INDEX:
+ case OSMESA_RGBA:
+ case OSMESA_BGRA:
+ case OSMESA_ARGB:
+ case OSMESA_RGB:
+ case OSMESA_BGR:
+ case OSMESA_RGB_565:
+ /* legal */
+ break;
+ default:
+ return NULL;
+ }
+ break;
+ case OSMESA_DEPTH_BITS:
+ depthBits = attribList[i+1];
+ if (depthBits < 0)
+ return NULL;
+ break;
+ case OSMESA_STENCIL_BITS:
+ stencilBits = attribList[i+1];
+ if (stencilBits < 0)
+ return NULL;
+ break;
+ case OSMESA_ACCUM_BITS:
+ accumBits = attribList[i+1];
+ if (accumBits < 0)
+ return NULL;
+ break;
+ case OSMESA_PROFILE:
+ profile = attribList[i+1];
+ if (profile != OSMESA_CORE_PROFILE &&
+ profile != OSMESA_COMPAT_PROFILE)
+ return NULL;
+ break;
+ case OSMESA_CONTEXT_MAJOR_VERSION:
+ version_major = attribList[i+1];
+ if (version_major < 1)
+ return NULL;
+ break;
+ case OSMESA_CONTEXT_MINOR_VERSION:
+ version_minor = attribList[i+1];
+ if (version_minor < 0)
+ return NULL;
+ break;
+ case 0:
+ /* end of list */
+ break;
+ default:
+ fprintf(stderr, "Bad attribute in OSMesaCreateContextAttribs()\n");
+ return NULL;
+ }
+ }
+
osmesa = (OSMesaContext) CALLOC_STRUCT(osmesa_context);
if (!osmesa)
return NULL;
/*
* Create the rendering context
*/
- attribs.profile = ST_PROFILE_DEFAULT;
- attribs.major = 2;
- attribs.minor = 1;
+ memset(&attribs, 0, sizeof(attribs));
+ attribs.profile = (profile == OSMESA_CORE_PROFILE)
+ ? ST_PROFILE_OPENGL_CORE : ST_PROFILE_DEFAULT;
+ attribs.major = version_major;
+ attribs.minor = version_minor;
attribs.flags = 0; /* ST_CONTEXT_FLAG_x */
attribs.options.force_glsl_extensions_warn = FALSE;
attribs.options.disable_blend_func_extended = FALSE;
}
+
/**
* Destroy an Off-Screen Mesa rendering context.
*
static struct name_function functions[] = {
{ "OSMesaCreateContext", (OSMESAproc) OSMesaCreateContext },
{ "OSMesaCreateContextExt", (OSMESAproc) OSMesaCreateContextExt },
+ { "OSMesaCreateContextAttribs", (OSMESAproc) OSMesaCreateContextAttribs },
{ "OSMesaDestroyContext", (OSMESAproc) OSMesaDestroyContext },
{ "OSMesaMakeCurrent", (OSMESAproc) OSMesaMakeCurrent },
{ "OSMesaGetCurrentContext", (OSMESAproc) OSMesaGetCurrentContext },
unsigned int size, unsigned int num_elements, void *data,
VABufferID *buf_id)
{
+ vlVaDriver *drv;
vlVaBuffer *buf;
if (!ctx)
if (data)
memcpy(buf->data, data, size * num_elements);
- *buf_id = handle_table_add(VL_VA_DRIVER(ctx)->htab, buf);
+ drv = VL_VA_DRIVER(ctx);
+ pipe_mutex_lock(drv->mutex);
+ *buf_id = handle_table_add(drv->htab, buf);
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_SUCCESS;
}
vlVaBufferSetNumElements(VADriverContextP ctx, VABufferID buf_id,
unsigned int num_elements)
{
+ vlVaDriver *drv;
vlVaBuffer *buf;
if (!ctx)
return VA_STATUS_ERROR_INVALID_CONTEXT;
- buf = handle_table_get(VL_VA_DRIVER(ctx)->htab, buf_id);
+ drv = VL_VA_DRIVER(ctx);
+ pipe_mutex_lock(drv->mutex);
+ buf = handle_table_get(drv->htab, buf_id);
+ pipe_mutex_unlock(drv->mutex);
if (!buf)
return VA_STATUS_ERROR_INVALID_BUFFER;
if (!pbuff)
return VA_STATUS_ERROR_INVALID_PARAMETER;
+ pipe_mutex_lock(drv->mutex);
buf = handle_table_get(drv->htab, buf_id);
- if (!buf)
- return VA_STATUS_ERROR_INVALID_BUFFER;
-
- if (buf->export_refcount > 0)
+ if (!buf || buf->export_refcount > 0) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_INVALID_BUFFER;
+ }
if (buf->derived_surface.resource) {
*pbuff = pipe_buffer_map(drv->pipe, buf->derived_surface.resource,
PIPE_TRANSFER_WRITE,
&buf->derived_surface.transfer);
+ pipe_mutex_unlock(drv->mutex);
if (!buf->derived_surface.transfer || !*pbuff)
return VA_STATUS_ERROR_INVALID_BUFFER;
} else {
+ pipe_mutex_unlock(drv->mutex);
*pbuff = buf->data;
}
if (!drv)
return VA_STATUS_ERROR_INVALID_CONTEXT;
+ pipe_mutex_lock(drv->mutex);
buf = handle_table_get(drv->htab, buf_id);
- if (!buf)
- return VA_STATUS_ERROR_INVALID_BUFFER;
-
- if (buf->export_refcount > 0)
+ if (!buf || buf->export_refcount > 0) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_INVALID_BUFFER;
+ }
if (buf->derived_surface.resource) {
- if (!buf->derived_surface.transfer)
+ if (!buf->derived_surface.transfer) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_INVALID_BUFFER;
+ }
pipe_buffer_unmap(drv->pipe, buf->derived_surface.transfer);
buf->derived_surface.transfer = NULL;
}
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_SUCCESS;
}
VAStatus
vlVaDestroyBuffer(VADriverContextP ctx, VABufferID buf_id)
{
+ vlVaDriver *drv;
vlVaBuffer *buf;
if (!ctx)
return VA_STATUS_ERROR_INVALID_CONTEXT;
- buf = handle_table_get(VL_VA_DRIVER(ctx)->htab, buf_id);
- if (!buf)
+ drv = VL_VA_DRIVER(ctx);
+ pipe_mutex_lock(drv->mutex);
+ buf = handle_table_get(drv->htab, buf_id);
+ if (!buf) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_INVALID_BUFFER;
+ }
if (buf->derived_surface.resource) {
- if (buf->export_refcount > 0)
+ if (buf->export_refcount > 0) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_INVALID_BUFFER;
+ }
pipe_resource_reference(&buf->derived_surface.resource, NULL);
}
FREE(buf->data);
FREE(buf);
handle_table_remove(VL_VA_DRIVER(ctx)->htab, buf_id);
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_SUCCESS;
}
vlVaBufferInfo(VADriverContextP ctx, VABufferID buf_id, VABufferType *type,
unsigned int *size, unsigned int *num_elements)
{
+ vlVaDriver *drv;
vlVaBuffer *buf;
if (!ctx)
return VA_STATUS_ERROR_INVALID_CONTEXT;
- buf = handle_table_get(VL_VA_DRIVER(ctx)->htab, buf_id);
+ drv = VL_VA_DRIVER(ctx);
+ pipe_mutex_lock(drv->mutex);
+ buf = handle_table_get(drv->htab, buf_id);
+ pipe_mutex_unlock(drv->mutex);
if (!buf)
return VA_STATUS_ERROR_INVALID_BUFFER;
vlVaAcquireBufferHandle(VADriverContextP ctx, VABufferID buf_id,
VABufferInfo *out_buf_info)
{
+ vlVaDriver *drv;
uint32_t i;
uint32_t mem_type;
vlVaBuffer *buf ;
if (!ctx)
return VA_STATUS_ERROR_INVALID_CONTEXT;
+ drv = VL_VA_DRIVER(ctx);
+ screen = VL_VA_PSCREEN(ctx);
+ pipe_mutex_lock(drv->mutex);
buf = handle_table_get(VL_VA_DRIVER(ctx)->htab, buf_id);
+ pipe_mutex_unlock(drv->mutex);
if (!buf)
return VA_STATUS_ERROR_INVALID_BUFFER;
if (!buf->derived_surface.resource)
return VA_STATUS_ERROR_INVALID_BUFFER;
- screen = VL_VA_PSCREEN(ctx);
-
- if (buf->derived_surface.fence) {
- screen->fence_finish(screen, buf->derived_surface.fence, PIPE_TIMEOUT_INFINITE);
- screen->fence_reference(screen, &buf->derived_surface.fence, NULL);
- }
-
if (buf->export_refcount > 0) {
if (buf->export_state.mem_type != mem_type)
return VA_STATUS_ERROR_INVALID_PARAMETER;
case VA_SURFACE_ATTRIB_MEM_TYPE_DRM_PRIME: {
struct winsys_handle whandle;
+ pipe_mutex_lock(drv->mutex);
+ drv->pipe->flush(drv->pipe, NULL, 0);
+ pipe_mutex_unlock(drv->mutex);
+
memset(&whandle, 0, sizeof(whandle));
whandle.type = DRM_API_HANDLE_TYPE_FD;
VAStatus
vlVaReleaseBufferHandle(VADriverContextP ctx, VABufferID buf_id)
{
+ vlVaDriver *drv;
vlVaBuffer *buf;
if (!ctx)
return VA_STATUS_ERROR_INVALID_CONTEXT;
- buf = handle_table_get(VL_VA_DRIVER(ctx)->htab, buf_id);
+ drv = VL_VA_DRIVER(ctx);
+ pipe_mutex_lock(drv->mutex);
+ buf = handle_table_get(drv->htab, buf_id);
+ pipe_mutex_unlock(drv->mutex);
if (!buf)
return VA_STATUS_ERROR_INVALID_BUFFER;
vl_csc_get_matrix(VL_CSC_COLOR_STANDARD_BT_601, NULL, true, &drv->csc);
vl_compositor_set_csc_matrix(&drv->cstate, (const vl_csc_matrix *)&drv->csc);
+ pipe_mutex_init(drv->mutex);
ctx->pDriverData = (void *)drv;
ctx->version_major = 0;
}
context->desc.base.profile = config_id;
+ pipe_mutex_lock(drv->mutex);
*context_id = handle_table_add(drv->htab, context);
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_SUCCESS;
}
return VA_STATUS_ERROR_INVALID_CONTEXT;
drv = VL_VA_DRIVER(ctx);
+ pipe_mutex_lock(drv->mutex);
context = handle_table_get(drv->htab, context_id);
if (context->decoder) {
}
FREE(context);
handle_table_remove(drv->htab, context_id);
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_SUCCESS;
}
drv->pipe->destroy(drv->pipe);
drv->vscreen->destroy(drv->vscreen);
handle_table_destroy(drv->htab);
+ pipe_mutex_destroy(drv->mutex);
FREE(drv);
return VA_STATUS_SUCCESS;
#include "util/u_video.h"
#include "vl/vl_winsys.h"
+#include "vl/vl_video_buffer.h"
#include "va_private.h"
*width = p_surf->templat.width;
*height = p_surf->templat.height;
- if (component > 0) {
- if (p_surf->templat.chroma_format == PIPE_VIDEO_CHROMA_FORMAT_420) {
- *width /= 2;
- *height /= 2;
- } else if (p_surf->templat.chroma_format == PIPE_VIDEO_CHROMA_FORMAT_422)
- *width /= 2;
- }
- if (p_surf->templat.interlaced)
- *height /= 2;
+ vl_video_buffer_adjust_size(width, height, component,
+ p_surf->templat.chroma_format,
+ p_surf->templat.interlaced);
}
VAStatus
img = CALLOC(1, sizeof(VAImage));
if (!img)
return VA_STATUS_ERROR_ALLOCATION_FAILED;
+ pipe_mutex_lock(drv->mutex);
img->image_id = handle_table_add(drv->htab, img);
+ pipe_mutex_unlock(drv->mutex);
img->format = *format;
img->width = width;
return VA_STATUS_ERROR_ALLOCATION_FAILED;
}
+ pipe_mutex_lock(drv->mutex);
img->image_id = handle_table_add(drv->htab, img);
img_buf->type = VAImageBufferType;
- img_buf->size = image->data_size;
+ img_buf->size = img->data_size;
img_buf->num_elements = 1;
- img_buf->derived_surface.fence = surf->fence;
pipe_resource_reference(&img_buf->derived_surface.resource, surfaces[0]->texture);
img->buf = handle_table_add(VL_VA_DRIVER(ctx)->htab, img_buf);
+ pipe_mutex_unlock(drv->mutex);
*image = *img;
VAStatus
vlVaDestroyImage(VADriverContextP ctx, VAImageID image)
{
+ vlVaDriver *drv;
VAImage *vaimage;
if (!ctx)
return VA_STATUS_ERROR_INVALID_CONTEXT;
- vaimage = handle_table_get(VL_VA_DRIVER(ctx)->htab, image);
- if (!vaimage)
+ drv = VL_VA_DRIVER(ctx);
+ pipe_mutex_lock(drv->mutex);
+ vaimage = handle_table_get(drv->htab, image);
+ if (!vaimage) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_INVALID_IMAGE;
+ }
handle_table_remove(VL_VA_DRIVER(ctx)->htab, image);
+ pipe_mutex_unlock(drv->mutex);
FREE(vaimage);
return vlVaDestroyBuffer(ctx, vaimage->buf);
}
drv = VL_VA_DRIVER(ctx);
+ pipe_mutex_lock(drv->mutex);
surf = handle_table_get(drv->htab, surface);
- if (!surf || !surf->buffer)
+ if (!surf || !surf->buffer) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_INVALID_SURFACE;
+ }
vaimage = handle_table_get(drv->htab, image);
- if (!vaimage)
+ if (!vaimage) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_INVALID_IMAGE;
+ }
img_buf = handle_table_get(drv->htab, vaimage->buf);
- if (!img_buf)
+ if (!img_buf) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_INVALID_BUFFER;
+ }
format = VaFourccToPipeFormat(vaimage->format.fourcc);
- if (format == PIPE_FORMAT_NONE)
+ if (format == PIPE_FORMAT_NONE) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_OPERATION_FAILED;
+ }
if (format != surf->buffer->buffer_format) {
/* support NV12 to YV12 and IYUV conversion now only */
(format == PIPE_FORMAT_IYUV &&
surf->buffer->buffer_format == PIPE_FORMAT_NV12))
convert = true;
- else
+ else {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_OPERATION_FAILED;
+ }
}
views = surf->buffer->get_sampler_view_planes(surf->buffer);
- if (!views)
+ if (!views) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_OPERATION_FAILED;
+ }
for (i = 0; i < vaimage->num_planes; i++) {
data[i] = img_buf->data + vaimage->offsets[i];
uint8_t *map;
map = drv->pipe->transfer_map(drv->pipe, views[i]->texture, 0,
PIPE_TRANSFER_READ, &box, &transfer);
- if (!map)
+ if (!map) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_OPERATION_FAILED;
+ }
if (i == 1 && convert) {
u_copy_nv12_to_yv12(data, pitches, i, j,
pipe_transfer_unmap(drv->pipe, transfer);
}
}
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_SUCCESS;
}
return VA_STATUS_ERROR_INVALID_CONTEXT;
drv = VL_VA_DRIVER(ctx);
+ pipe_mutex_lock(drv->mutex);
surf = handle_table_get(drv->htab, surface);
- if (!surf || !surf->buffer)
+ if (!surf || !surf->buffer) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_INVALID_SURFACE;
+ }
vaimage = handle_table_get(drv->htab, image);
- if (!vaimage)
+ if (!vaimage) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_INVALID_IMAGE;
+ }
img_buf = handle_table_get(drv->htab, vaimage->buf);
- if (!img_buf)
+ if (!img_buf) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_INVALID_BUFFER;
+ }
if (img_buf->derived_surface.resource) {
/* Attempting to transfer derived image to surface */
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_UNIMPLEMENTED;
}
format = VaFourccToPipeFormat(vaimage->format.fourcc);
- if (format == PIPE_FORMAT_NONE)
+ if (format == PIPE_FORMAT_NONE) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_OPERATION_FAILED;
+ }
if (format != surf->buffer->buffer_format) {
struct pipe_video_buffer *tmp_buf;
if (!tmp_buf) {
surf->templat.buffer_format = old_surf_format;
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_ALLOCATION_FAILED;
}
}
views = surf->buffer->get_sampler_view_planes(surf->buffer);
- if (!views)
+ if (!views) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_OPERATION_FAILED;
+ }
for (i = 0; i < vaimage->num_planes; i++) {
data[i] = img_buf->data + vaimage->offsets[i];
pitches[i] * views[i]->texture->array_size, 0);
}
}
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_SUCCESS;
}
if (!drv)
return VA_STATUS_ERROR_INVALID_CONTEXT;
+ pipe_mutex_lock(drv->mutex);
context = handle_table_get(drv->htab, context_id);
- if (!context)
+ if (!context) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_INVALID_CONTEXT;
+ }
surf = handle_table_get(drv->htab, render_target);
+ pipe_mutex_unlock(drv->mutex);
if (!surf || !surf->buffer)
return VA_STATUS_ERROR_INVALID_SURFACE;
context->target = surf->buffer;
if (!context->decoder) {
+
/* VPP */
if (context->templat.profile == PIPE_VIDEO_PROFILE_UNKNOWN &&
- ((context->target->buffer_format != PIPE_FORMAT_B8G8R8A8_UNORM &&
- context->target->buffer_format != PIPE_FORMAT_R8G8B8A8_UNORM &&
- context->target->buffer_format != PIPE_FORMAT_B8G8R8X8_UNORM &&
- context->target->buffer_format != PIPE_FORMAT_R8G8B8X8_UNORM) ||
- context->target->interlaced))
+ context->target->buffer_format != PIPE_FORMAT_B8G8R8A8_UNORM &&
+ context->target->buffer_format != PIPE_FORMAT_R8G8B8A8_UNORM &&
+ context->target->buffer_format != PIPE_FORMAT_B8G8R8X8_UNORM &&
+ context->target->buffer_format != PIPE_FORMAT_R8G8B8X8_UNORM &&
+ context->target->buffer_format != PIPE_FORMAT_NV12)
return VA_STATUS_ERROR_UNIMPLEMENTED;
return VA_STATUS_SUCCESS;
handleSliceParameterBuffer(vlVaContext *context, vlVaBuffer *buf)
{
switch (u_reduce_video_profile(context->templat.profile)) {
+ case PIPE_VIDEO_FORMAT_MPEG12:
+ vlVaHandleSliceParameterBufferMPEG12(context, buf);
+ break;
+
+ case PIPE_VIDEO_FORMAT_VC1:
+ vlVaHandleSliceParameterBufferVC1(context, buf);
+ break;
+
case PIPE_VIDEO_FORMAT_MPEG4_AVC:
vlVaHandleSliceParameterBufferH264(context, buf);
break;
if (!drv)
return VA_STATUS_ERROR_INVALID_CONTEXT;
+ pipe_mutex_lock(drv->mutex);
context = handle_table_get(drv->htab, context_id);
- if (!context)
+ if (!context) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_INVALID_CONTEXT;
+ }
for (i = 0; i < num_buffers; ++i) {
vlVaBuffer *buf = handle_table_get(drv->htab, buffers[i]);
- if (!buf)
+ if (!buf) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_INVALID_BUFFER;
+ }
switch (buf->type) {
case VAPictureParameterBufferType:
break;
}
}
+ pipe_mutex_unlock(drv->mutex);
return vaStatus;
}
if (!drv)
return VA_STATUS_ERROR_INVALID_CONTEXT;
+ pipe_mutex_lock(drv->mutex);
context = handle_table_get(drv->htab, context_id);
+ pipe_mutex_unlock(drv->mutex);
if (!context)
return VA_STATUS_ERROR_INVALID_CONTEXT;
unsigned i;
assert(buf->size >= sizeof(VAPictureParameterBufferH264) && buf->num_elements == 1);
+ context->desc.h264.slice_count = 0;
/*CurrPic*/
context->desc.h264.field_order_cnt[0] = h264->CurrPic.TopFieldOrderCnt;
context->desc.h264.field_order_cnt[1] = h264->CurrPic.BottomFieldOrderCnt;
VASliceParameterBufferH264 *h264 = buf->data;
assert(buf->size >= sizeof(VASliceParameterBufferH264) && buf->num_elements == 1);
+ context->desc.h264.slice_count += buf->num_elements;
context->desc.h264.num_ref_idx_l0_active_minus1 =
h264->num_ref_idx_l0_active_minus1;
context->desc.h264.num_ref_idx_l1_active_minus1 =
for (i = 0 ; i < 15 ; i++) {
context->desc.h265.PicOrderCntVal[i] = hevc->ReferenceFrames[i].pic_order_cnt;
- unsigned int index = hevc->ReferenceFrames[i].picture_id & 0x7F;
-
- if (index == 0x7F)
- continue;
-
vlVaGetReferenceFrame(drv, hevc->ReferenceFrames[i].picture_id, &context->desc.h265.ref[i]);
if ((hevc->ReferenceFrames[i].flags & VA_PICTURE_HEVC_RPS_ST_CURR_BEFORE) && (iBefore < 8)) {
VAPictureParameterBufferMPEG2 *mpeg2 = buf->data;
assert(buf->size >= sizeof(VAPictureParameterBufferMPEG2) && buf->num_elements == 1);
+ context->desc.mpeg12.num_slices = 0;
/*horizontal_size;*/
/*vertical_size;*/
vlVaGetReferenceFrame(drv, mpeg2->forward_reference_picture, &context->desc.mpeg12.ref[0]);
context->desc.mpeg12.non_intra_matrix = NULL;
}
+void vlVaHandleSliceParameterBufferMPEG12(vlVaContext *context, vlVaBuffer *buf)
+{
+ assert(buf->size >= sizeof(VASliceParameterBufferMPEG2) && buf->num_elements == 1);
+ context->desc.mpeg12.num_slices += buf->num_elements;
+}
VAPictureParameterBufferVC1 * vc1 = buf->data;
assert(buf->size >= sizeof(VAPictureParameterBufferVC1) && buf->num_elements == 1);
+ context->desc.vc1.slice_count = 0;
vlVaGetReferenceFrame(drv, vc1->forward_reference_picture, &context->desc.vc1.ref[0]);
vlVaGetReferenceFrame(drv, vc1->backward_reference_picture, &context->desc.vc1.ref[1]);
context->desc.vc1.picture_type = vc1->picture_fields.bits.picture_type;
context->desc.vc1.deblockEnable = vc1->post_processing != 0;
context->desc.vc1.pquant = vc1->pic_quantizer_fields.bits.pic_quantizer_scale;
}
+
+void vlVaHandleSliceParameterBufferVC1(vlVaContext *context, vlVaBuffer *buf)
+{
+ assert(buf->size >= sizeof(VASliceParameterBufferVC1) && buf->num_elements == 1);
+ context->desc.vc1.slice_count += buf->num_elements;
+}
#include "util/u_handle_table.h"
+#include "vl/vl_defines.h"
+#include "vl/vl_video_buffer.h"
+
#include "va_private.h"
static const VARectangle *
return def;
}
-VAStatus
-vlVaHandleVAProcPipelineParameterBufferType(vlVaDriver *drv, vlVaContext *context, vlVaBuffer *buf)
+static VAStatus
+vlVaPostProcCompositor(vlVaDriver *drv, vlVaContext *context,
+ const VARectangle *src_region,
+ const VARectangle *dst_region,
+ struct pipe_video_buffer *src,
+ struct pipe_video_buffer *dst,
+ enum vl_compositor_deinterlace deinterlace)
{
- VARectangle def_src_region, def_dst_region;
- const VARectangle *src_region, *dst_region;
+ struct pipe_surface **surfaces;
struct u_rect src_rect;
struct u_rect dst_rect;
- vlVaSurface *src_surface;
- VAProcPipelineParameterBuffer *pipeline_param;
- struct pipe_surface **surfaces;
- struct pipe_screen *screen;
- struct pipe_surface *psurf;
-
- if (!drv || !context)
- return VA_STATUS_ERROR_INVALID_CONTEXT;
-
- if (!buf || !buf->data)
- return VA_STATUS_ERROR_INVALID_BUFFER;
-
- if (!context->target)
- return VA_STATUS_ERROR_INVALID_SURFACE;
-
- pipeline_param = (VAProcPipelineParameterBuffer *)buf->data;
-
- src_surface = handle_table_get(drv->htab, pipeline_param->surface);
- if (!src_surface || !src_surface->buffer)
- return VA_STATUS_ERROR_INVALID_SURFACE;
-
- surfaces = context->target->get_surfaces(context->target);
+ surfaces = dst->get_surfaces(dst);
if (!surfaces || !surfaces[0])
return VA_STATUS_ERROR_INVALID_SURFACE;
- screen = drv->pipe->screen;
-
- psurf = surfaces[0];
-
- src_region = vlVaRegionDefault(pipeline_param->surface_region, src_surface->buffer, &def_src_region);
- dst_region = vlVaRegionDefault(pipeline_param->output_region, context->target, &def_dst_region);
-
src_rect.x0 = src_region->x;
src_rect.y0 = src_region->y;
src_rect.x1 = src_region->x + src_region->width;
dst_rect.y1 = dst_region->y + dst_region->height;
vl_compositor_clear_layers(&drv->cstate);
- vl_compositor_set_buffer_layer(&drv->cstate, &drv->compositor, 0, src_surface->buffer, &src_rect, NULL, VL_COMPOSITOR_WEAVE);
+ vl_compositor_set_buffer_layer(&drv->cstate, &drv->compositor, 0, src,
+ &src_rect, NULL, deinterlace);
vl_compositor_set_layer_dst_area(&drv->cstate, 0, &dst_rect);
- vl_compositor_render(&drv->cstate, &drv->compositor, psurf, NULL, false);
+ vl_compositor_render(&drv->cstate, &drv->compositor, surfaces[0], NULL, false);
- screen->fence_reference(screen, &src_surface->fence, NULL);
- drv->pipe->flush(drv->pipe, &src_surface->fence, 0);
+ return VA_STATUS_SUCCESS;
+}
+
+static void vlVaGetBox(struct pipe_video_buffer *buf, unsigned idx,
+ struct pipe_box *box, const VARectangle *region)
+{
+ unsigned plane = buf->interlaced ? idx / 2: idx;
+ unsigned x, y, width, height;
+
+ x = abs(region->x);
+ y = abs(region->y);
+ width = region->width;
+ height = region->height;
+
+ vl_video_buffer_adjust_size(&x, &y, plane, buf->chroma_format,
+ buf->interlaced);
+ vl_video_buffer_adjust_size(&width, &height, plane, buf->chroma_format,
+ buf->interlaced);
+
+ box->x = region->x < 0 ? -x : x;
+ box->y = region->y < 0 ? -y : y;
+ box->width = width;
+ box->height = height;
+}
+
+static VAStatus vlVaPostProcBlit(vlVaDriver *drv, vlVaContext *context,
+ const VARectangle *src_region,
+ const VARectangle *dst_region,
+ struct pipe_video_buffer *src,
+ struct pipe_video_buffer *dst,
+ enum vl_compositor_deinterlace deinterlace)
+{
+ struct pipe_surface **src_surfaces;
+ struct pipe_surface **dst_surfaces;
+ unsigned i;
+
+ if (src->interlaced != dst->interlaced)
+ return VA_STATUS_ERROR_INVALID_SURFACE;
+
+ src_surfaces = src->get_surfaces(src);
+ if (!src_surfaces || !src_surfaces[0])
+ return VA_STATUS_ERROR_INVALID_SURFACE;
+
+ dst_surfaces = dst->get_surfaces(dst);
+ if (!dst_surfaces || !dst_surfaces[0])
+ return VA_STATUS_ERROR_INVALID_SURFACE;
+
+ for (i = 0; i < VL_MAX_SURFACES; ++i) {
+ struct pipe_surface *from = src_surfaces[i];
+ struct pipe_blit_info blit;
+
+ if (src->interlaced) {
+ /* Not 100% accurate, but close enough */
+ switch (deinterlace) {
+ case VL_COMPOSITOR_BOB_TOP:
+ from = src_surfaces[i & ~1];
+ break;
+ case VL_COMPOSITOR_BOB_BOTTOM:
+ from = src_surfaces[(i & ~1) + 1];
+ break;
+ default:
+ break;
+ }
+ }
+
+ if (!from || !dst_surfaces[i])
+ continue;
+
+ memset(&blit, 0, sizeof(blit));
+ blit.src.resource = from->texture;
+ blit.src.format = from->format;
+ blit.src.level = 0;
+ blit.src.box.z = from->u.tex.first_layer;
+ blit.src.box.depth = 1;
+ vlVaGetBox(src, i, &blit.src.box, src_region);
+
+ blit.dst.resource = dst_surfaces[i]->texture;
+ blit.dst.format = dst_surfaces[i]->format;
+ blit.dst.level = 0;
+ blit.dst.box.z = dst_surfaces[i]->u.tex.first_layer;
+ blit.dst.box.depth = 1;
+ vlVaGetBox(dst, i, &blit.dst.box, dst_region);
+
+ blit.mask = PIPE_MASK_RGBA;
+ blit.filter = PIPE_TEX_MIPFILTER_LINEAR;
+
+ drv->pipe->blit(drv->pipe, &blit);
+ }
+
+ // TODO: figure out why this is necessary for DMA-buf sharing
+ drv->pipe->flush(drv->pipe, NULL, 0);
return VA_STATUS_SUCCESS;
}
+VAStatus
+vlVaHandleVAProcPipelineParameterBufferType(vlVaDriver *drv, vlVaContext *context, vlVaBuffer *buf)
+{
+ enum vl_compositor_deinterlace deinterlace = VL_COMPOSITOR_WEAVE;
+ VARectangle def_src_region, def_dst_region;
+ const VARectangle *src_region, *dst_region;
+ VAProcPipelineParameterBuffer *param;
+ vlVaSurface *src_surface;
+ unsigned i;
+ if (!drv || !context)
+ return VA_STATUS_ERROR_INVALID_CONTEXT;
+
+ if (!buf || !buf->data)
+ return VA_STATUS_ERROR_INVALID_BUFFER;
+
+ if (!context->target)
+ return VA_STATUS_ERROR_INVALID_SURFACE;
+
+ param = buf->data;
+
+ src_surface = handle_table_get(drv->htab, param->surface);
+ if (!src_surface || !src_surface->buffer)
+ return VA_STATUS_ERROR_INVALID_SURFACE;
+
+ for (i = 0; i < param->num_filters; i++) {
+ vlVaBuffer *buf = handle_table_get(drv->htab, param->filters[i]);
+ VAProcFilterParameterBufferBase *filter;
+
+ if (!buf || buf->type != VAProcFilterParameterBufferType)
+ return VA_STATUS_ERROR_INVALID_BUFFER;
+
+ filter = buf->data;
+ switch (filter->type) {
+ case VAProcFilterDeinterlacing: {
+ VAProcFilterParameterBufferDeinterlacing *deint = buf->data;
+ switch (deint->algorithm) {
+ case VAProcDeinterlacingBob:
+ if (deint->flags & VA_DEINTERLACING_BOTTOM_FIELD)
+ deinterlace = VL_COMPOSITOR_BOB_BOTTOM;
+ else
+ deinterlace = VL_COMPOSITOR_BOB_TOP;
+ break;
+
+ case VAProcDeinterlacingWeave:
+ deinterlace = VL_COMPOSITOR_WEAVE;
+ break;
+
+ default:
+ return VA_STATUS_ERROR_UNIMPLEMENTED;
+ }
+
+ break;
+ }
+
+ default:
+ return VA_STATUS_ERROR_UNIMPLEMENTED;
+ }
+ }
+
+ src_region = vlVaRegionDefault(param->surface_region, src_surface->buffer, &def_src_region);
+ dst_region = vlVaRegionDefault(param->output_region, context->target, &def_dst_region);
+
+ if (context->target->buffer_format != PIPE_FORMAT_NV12)
+ return vlVaPostProcCompositor(drv, context, src_region, dst_region,
+ src_surface->buffer, context->target,
+ deinterlace);
+ else
+ return vlVaPostProcBlit(drv, context, src_region, dst_region,
+ src_surface->buffer, context->target,
+ deinterlace);
+}
vlVaCreateSubpicture(VADriverContextP ctx, VAImageID image,
VASubpictureID *subpicture)
{
+ vlVaDriver *drv;
vlVaSubpicture *sub;
VAImage *img;
if (!ctx)
return VA_STATUS_ERROR_INVALID_CONTEXT;
- img = handle_table_get(VL_VA_DRIVER(ctx)->htab, image);
- if (!img)
+ drv = VL_VA_DRIVER(ctx);
+ pipe_mutex_lock(drv->mutex);
+ img = handle_table_get(drv->htab, image);
+ if (!img) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_INVALID_IMAGE;
+ }
sub = CALLOC(1, sizeof(*sub));
- if (!sub)
+ if (!sub) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_ALLOCATION_FAILED;
+ }
sub->image = img;
*subpicture = handle_table_add(VL_VA_DRIVER(ctx)->htab, sub);
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_SUCCESS;
}
VAStatus
vlVaDestroySubpicture(VADriverContextP ctx, VASubpictureID subpicture)
{
+ vlVaDriver *drv;
vlVaSubpicture *sub;
if (!ctx)
return VA_STATUS_ERROR_INVALID_CONTEXT;
- sub = handle_table_get(VL_VA_DRIVER(ctx)->htab, subpicture);
- if (!sub)
+ drv = VL_VA_DRIVER(ctx);
+ pipe_mutex_lock(drv->mutex);
+
+ sub = handle_table_get(drv->htab, subpicture);
+ if (!sub) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_INVALID_SUBPICTURE;
+ }
FREE(sub);
- handle_table_remove(VL_VA_DRIVER(ctx)->htab, subpicture);
+ handle_table_remove(drv->htab, subpicture);
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_SUCCESS;
}
VAStatus
vlVaSubpictureImage(VADriverContextP ctx, VASubpictureID subpicture, VAImageID image)
{
+ vlVaDriver *drv;
vlVaSubpicture *sub;
VAImage *img;
if (!ctx)
return VA_STATUS_ERROR_INVALID_CONTEXT;
- img = handle_table_get(VL_VA_DRIVER(ctx)->htab, image);
- if (!img)
+ drv = VL_VA_DRIVER(ctx);
+ pipe_mutex_lock(drv->mutex);
+
+ img = handle_table_get(drv->htab, image);
+ if (!img) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_INVALID_IMAGE;
+ }
- sub = handle_table_get(VL_VA_DRIVER(ctx)->htab, subpicture);
+ sub = handle_table_get(drv->htab, subpicture);
+ pipe_mutex_unlock(drv->mutex);
if (!sub)
return VA_STATUS_ERROR_INVALID_SUBPICTURE;
if (!ctx)
return VA_STATUS_ERROR_INVALID_CONTEXT;
drv = VL_VA_DRIVER(ctx);
+ pipe_mutex_lock(drv->mutex);
sub = handle_table_get(drv->htab, subpicture);
- if (!sub)
+ if (!sub) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_INVALID_SUBPICTURE;
+ }
for (i = 0; i < num_surfaces; i++) {
surf = handle_table_get(drv->htab, target_surfaces[i]);
- if (!surf)
+ if (!surf) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_INVALID_SURFACE;
+ }
}
sub->src_rect = src_rect;
tex_temp.flags = 0;
if (!drv->pipe->screen->is_format_supported(
drv->pipe->screen, tex_temp.format, tex_temp.target,
- tex_temp.nr_samples, tex_temp.bind))
+ tex_temp.nr_samples, tex_temp.bind)) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_ALLOCATION_FAILED;
+ }
tex = drv->pipe->screen->resource_create(drv->pipe->screen, &tex_temp);
u_sampler_view_default_template(&sampler_templ, tex, tex->format);
sub->sampler = drv->pipe->create_sampler_view(drv->pipe, tex, &sampler_templ);
pipe_resource_reference(&tex, NULL);
- if (!sub->sampler)
+ if (!sub->sampler) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_ALLOCATION_FAILED;
+ }
for (i = 0; i < num_surfaces; i++) {
surf = handle_table_get(drv->htab, target_surfaces[i]);
util_dynarray_append(&surf->subpics, vlVaSubpicture *, sub);
}
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_SUCCESS;
}
if (!ctx)
return VA_STATUS_ERROR_INVALID_CONTEXT;
drv = VL_VA_DRIVER(ctx);
+ pipe_mutex_lock(drv->mutex);
sub = handle_table_get(drv->htab, subpicture);
- if (!sub)
+ if (!sub) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_INVALID_SUBPICTURE;
+ }
for (i = 0; i < num_surfaces; i++) {
surf = handle_table_get(drv->htab, target_surfaces[i]);
- if (!surf)
+ if (!surf) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_INVALID_SURFACE;
+ }
array = surf->subpics.data;
if (!array)
while (surf->subpics.size && util_dynarray_top(&surf->subpics, vlVaSubpicture *) == NULL)
(void)util_dynarray_pop(&surf->subpics, vlVaSubpicture *);
}
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_SUCCESS;
}
return VA_STATUS_ERROR_INVALID_CONTEXT;
drv = VL_VA_DRIVER(ctx);
+ pipe_mutex_lock(drv->mutex);
for (i = 0; i < num_surfaces; ++i) {
vlVaSurface *surf = handle_table_get(drv->htab, surface_list[i]);
if (surf->buffer)
surf->buffer->destroy(surf->buffer);
- if(surf->fence)
- drv->pipe->screen->fence_reference(drv->pipe->screen, &surf->fence, NULL);
util_dynarray_fini(&surf->subpics);
FREE(surf);
handle_table_remove(drv->htab, surface_list[i]);
}
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_SUCCESS;
}
return VA_STATUS_ERROR_INVALID_CONTEXT;
drv = VL_VA_DRIVER(ctx);
+ pipe_mutex_lock(drv->mutex);
surf = handle_table_get(drv->htab, surface_id);
- if (!surf)
+ if (!surf) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_INVALID_SURFACE;
+ }
screen = drv->pipe->screen;
vscreen = drv->vscreen;
- if(surf->fence) {
- screen->fence_finish(screen, surf->fence, PIPE_TIMEOUT_INFINITE);
- screen->fence_reference(screen, &surf->fence, NULL);
- }
-
tex = vscreen->texture_from_drawable(vscreen, draw);
- if (!tex)
+ if (!tex) {
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_INVALID_DISPLAY;
+ }
dirty_area = vscreen->get_dirty_area(vscreen);
surf_draw = drv->pipe->create_surface(drv->pipe, tex, &surf_templ);
if (!surf_draw) {
pipe_resource_reference(&tex, NULL);
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_ERROR_INVALID_DISPLAY;
}
vl_compositor_render(&drv->cstate, &drv->compositor, surf_draw, dirty_area, true);
status = vlVaPutSubpictures(surf, drv, surf_draw, dirty_area, &src_rect, &dst_rect);
- if (status)
+ if (status) {
+ pipe_mutex_unlock(drv->mutex);
return status;
+ }
screen->flush_frontbuffer(screen, tex, 0, 0,
vscreen->get_private(vscreen), NULL);
- screen->fence_reference(screen, &surf->fence, NULL);
- drv->pipe->flush(drv->pipe, &surf->fence, 0);
+ drv->pipe->flush(drv->pipe, NULL, 0);
pipe_resource_reference(&tex, NULL);
pipe_surface_reference(&surf_draw, NULL);
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_SUCCESS;
}
memset(surfaces, VA_INVALID_ID, num_surfaces * sizeof(VASurfaceID));
+ pipe_mutex_lock(drv->mutex);
for (i = 0; i < num_surfaces; i++) {
vlVaSurface *surf = CALLOC(1, sizeof(vlVaSurface));
if (!surf)
assert(0);
}
}
+ pipe_mutex_unlock(drv->mutex);
return VA_STATUS_SUCCESS;
no_res:
+ pipe_mutex_unlock(drv->mutex);
if (i)
vlVaDestroySurfaces(ctx, surfaces, i);
if (!num_filters || !filters)
return VA_STATUS_ERROR_INVALID_PARAMETER;
- filters[num++] = VAProcFilterNone;
+ filters[num++] = VAProcFilterDeinterlacing;
*num_filters = num;
switch (type) {
case VAProcFilterNone:
break;
+ case VAProcFilterDeinterlacing: {
+ VAProcFilterCapDeinterlacing *deint = filter_caps;
+
+ if (*num_filter_caps < 2) {
+ *num_filter_caps = 2;
+ return VA_STATUS_ERROR_MAX_NUM_EXCEEDED;
+ }
+
+ deint[i++].type = VAProcDeinterlacingBob;
+ deint[i++].type = VAProcDeinterlacingWeave;
+ break;
+ }
+
case VAProcFilterNoiseReduction:
- case VAProcFilterDeinterlacing:
case VAProcFilterSharpening:
case VAProcFilterColorBalance:
case VAProcFilterSkinToneEnhancement:
#include "vl/vl_csc.h"
#include "util/u_dynarray.h"
+#include "os/os_thread.h"
#define VL_VA_DRIVER(ctx) ((vlVaDriver *)ctx->pDriverData)
#define VL_VA_PSCREEN(ctx) (VL_VA_DRIVER(ctx)->vscreen->pscreen)
struct vl_compositor compositor;
struct vl_compositor_state cstate;
vl_csc_matrix csc;
+ pipe_mutex mutex;
} vlVaDriver;
typedef struct {
struct {
struct pipe_resource *resource;
struct pipe_transfer *transfer;
- struct pipe_fence_handle *fence;
} derived_surface;
unsigned int export_refcount;
VABufferInfo export_state;
typedef struct {
struct pipe_video_buffer templat, *buffer;
- struct pipe_fence_handle *fence;
struct util_dynarray subpics; /* vlVaSubpicture */
} vlVaSurface;
void vlVaGetReferenceFrame(vlVaDriver *drv, VASurfaceID surface_id, struct pipe_video_buffer **ref_frame);
void vlVaHandlePictureParameterBufferMPEG12(vlVaDriver *drv, vlVaContext *context, vlVaBuffer *buf);
void vlVaHandleIQMatrixBufferMPEG12(vlVaContext *context, vlVaBuffer *buf);
+void vlVaHandleSliceParameterBufferMPEG12(vlVaContext *context, vlVaBuffer *buf);
void vlVaHandlePictureParameterBufferH264(vlVaDriver *drv, vlVaContext *context, vlVaBuffer *buf);
void vlVaHandleIQMatrixBufferH264(vlVaContext *context, vlVaBuffer *buf);
void vlVaHandleSliceParameterBufferH264(vlVaContext *context, vlVaBuffer *buf);
void vlVaHandlePictureParameterBufferVC1(vlVaDriver *drv, vlVaContext *context, vlVaBuffer *buf);
+void vlVaHandleSliceParameterBufferVC1(vlVaContext *context, vlVaBuffer *buf);
void vlVaHandlePictureParameterBufferMPEG4(vlVaDriver *drv, vlVaContext *context, vlVaBuffer *buf);
void vlVaHandleIQMatrixBufferMPEG4(vlVaContext *context, vlVaBuffer *buf);
void vlVaHandleSliceParameterBufferMPEG4(vlVaContext *context, vlVaBuffer *buf);
default:
pipe_mutex_unlock(vmixer->device->mutex);
return VDP_STATUS_INVALID_VIDEO_MIXER_PICTURE_STRUCTURE;
- };
+ }
if (deinterlace != VL_COMPOSITOR_WEAVE && vmixer->deint.enabled &&
video_surface_past_count > 1 && video_surface_future_count > 0) {
*width = p_surf->templat.width;
*height = p_surf->templat.height;
- if (component > 0) {
- if (p_surf->templat.chroma_format == PIPE_VIDEO_CHROMA_FORMAT_420) {
- *width /= 2;
- *height /= 2;
- } else if (p_surf->templat.chroma_format == PIPE_VIDEO_CHROMA_FORMAT_422) {
- *width /= 2;
- }
- }
- if (p_surf->templat.interlaced)
- *height /= 2;
+ vl_video_buffer_adjust_size(width, height, component,
+ p_surf->templat.chroma_format,
+ p_surf->templat.interlaced);
}
/**
lib_LTLIBRARIES = lib@OPENCL_LIBNAME@.la
+AM_CPPFLAGS = \
+ $(LIBELF_CFLAGS)
+
lib@OPENCL_LIBNAME@_la_LDFLAGS = \
$(LLVM_LDFLAGS) \
-no-undefined \
$(top_builddir)/src/gallium/state_trackers/clover/libclover.la \
$(top_builddir)/src/gallium/auxiliary/libgallium.la \
$(top_builddir)/src/util/libmesautil.la \
- $(ELF_LIB) \
+ $(LIBELF_LIBS) \
$(DLOPEN_LIBS) \
-lclangCodeGen \
-lclangFrontendTool \
TARGET_CPPFLAGS =
TARGET_LIB_DEPS =
+include $(top_srcdir)/src/gallium/drivers/nouveau/Automake.inc
+
include $(top_srcdir)/src/gallium/drivers/r600/Automake.inc
include $(top_srcdir)/src/gallium/drivers/radeonsi/Automake.inc
pipe->launch_grid(pipe, block_layout, grid_layout, pc, input);
}
+static void test_default_init(void *p, int s, int x, int y)
+{
+ *(uint32_t *)p = 0xdeadbeef;
+}
+
+/* test_system_values */
+static void test_system_values_expect(void *p, int s, int x, int y)
+{
+ int id = x / 16, sv = (x % 16) / 4, c = x % 4;
+ int tid[] = { id % 20, (id % 240) / 20, id / 240, 0 };
+ int bsz[] = { 4, 3, 5, 1};
+ int gsz[] = { 5, 4, 1, 1};
+
+ switch (sv) {
+ case 0:
+ *(uint32_t *)p = tid[c] / bsz[c];
+ break;
+ case 1:
+ *(uint32_t *)p = bsz[c];
+ break;
+ case 2:
+ *(uint32_t *)p = gsz[c];
+ break;
+ case 3:
+ *(uint32_t *)p = tid[c] % bsz[c];
+ break;
+ }
+}
+
static void test_system_values(struct context *ctx)
{
const char *src = "COMP\n"
" STORE RES[0].xyzw, TEMP[0], SV[3]\n"
" RET\n"
"ENDSUB\n";
- void init(void *p, int s, int x, int y) {
- *(uint32_t *)p = 0xdeadbeef;
- }
- void expect(void *p, int s, int x, int y) {
- int id = x / 16, sv = (x % 16) / 4, c = x % 4;
- int tid[] = { id % 20, (id % 240) / 20, id / 240, 0 };
- int bsz[] = { 4, 3, 5, 1};
- int gsz[] = { 5, 4, 1, 1};
-
- switch (sv) {
- case 0:
- *(uint32_t *)p = tid[c] / bsz[c];
- break;
- case 1:
- *(uint32_t *)p = bsz[c];
- break;
- case 2:
- *(uint32_t *)p = gsz[c];
- break;
- case 3:
- *(uint32_t *)p = tid[c] % bsz[c];
- break;
- }
- }
printf("- %s\n", __func__);
init_prog(ctx, 0, 0, 0, src, NULL);
init_tex(ctx, 0, PIPE_BUFFER, true, PIPE_FORMAT_R32_FLOAT,
- 76800, 0, init);
+ 76800, 0, test_default_init);
init_compute_resources(ctx, (int []) { 0, -1 });
launch_grid(ctx, (uint []){4, 3, 5}, (uint []){5, 4, 1}, 0, NULL);
- check_tex(ctx, 0, expect, NULL);
+ check_tex(ctx, 0, test_system_values_expect, NULL);
destroy_compute_resources(ctx);
destroy_tex(ctx);
destroy_prog(ctx);
}
+/* test_resource_access */
+static void test_resource_access_init0(void *p, int s, int x, int y)
+{
+ *(float *)p = 8.0 - (float)x;
+}
+
+static void test_resource_access_expect(void *p, int s, int x, int y)
+{
+ *(float *)p = 8.0 - (float)((x + 4 * y) & 0x3f);
+}
+
static void test_resource_access(struct context *ctx)
{
const char *src = "COMP\n"
" STORE RES[1].xyzw, TEMP[1], TEMP[0]\n"
" RET\n"
" ENDSUB\n";
- void init0(void *p, int s, int x, int y) {
- *(float *)p = 8.0 - (float)x;
- }
- void init1(void *p, int s, int x, int y) {
- *(uint32_t *)p = 0xdeadbeef;
- }
- void expect(void *p, int s, int x, int y) {
- *(float *)p = 8.0 - (float)((x + 4*y) & 0x3f);
- }
printf("- %s\n", __func__);
init_prog(ctx, 0, 0, 0, src, NULL);
init_tex(ctx, 0, PIPE_BUFFER, true, PIPE_FORMAT_R32_FLOAT,
- 256, 0, init0);
+ 256, 0, test_resource_access_init0);
init_tex(ctx, 1, PIPE_TEXTURE_2D, true, PIPE_FORMAT_R32_FLOAT,
- 60, 12, init1);
+ 60, 12, test_default_init);
init_compute_resources(ctx, (int []) { 0, 1, -1 });
launch_grid(ctx, (uint []){1, 1, 1}, (uint []){15, 12, 1}, 0, NULL);
- check_tex(ctx, 1, expect, NULL);
+ check_tex(ctx, 1, test_resource_access_expect, NULL);
destroy_compute_resources(ctx);
destroy_tex(ctx);
destroy_prog(ctx);
}
+/* test_function_calls */
+static void test_function_calls_init(void *p, int s, int x, int y)
+{
+ *(uint32_t *)p = 15 * y + x;
+}
+
+static void test_function_calls_expect(void *p, int s, int x, int y)
+{
+ *(uint32_t *)p = (15 * y + x) < 4 ? 2 : 1 ;
+}
+
static void test_function_calls(struct context *ctx)
{
const char *src = "COMP\n"
"21: STORE RES[0].x, TEMP[2], TEMP[1].xxxx\n"
"22: RET\n"
"23: ENDSUB\n";
- void init(void *p, int s, int x, int y) {
- *(uint32_t *)p = 15 * y + x;
- }
- void expect(void *p, int s, int x, int y) {
- *(uint32_t *)p = (15 * y + x) < 4 ? 2 : 1 ;
- }
printf("- %s\n", __func__);
init_prog(ctx, 0, 0, 0, src, NULL);
init_tex(ctx, 0, PIPE_TEXTURE_2D, true, PIPE_FORMAT_R32_FLOAT,
- 15, 12, init);
+ 15, 12, test_function_calls_init);
init_compute_resources(ctx, (int []) { 0, -1 });
launch_grid(ctx, (uint []){3, 3, 3}, (uint []){5, 4, 1}, 15, NULL);
- check_tex(ctx, 0, expect, NULL);
+ check_tex(ctx, 0, test_function_calls_expect, NULL);
destroy_compute_resources(ctx);
destroy_tex(ctx);
destroy_prog(ctx);
}
+/* test_input_global */
+static void test_input_global_expect(void *p, int s, int x, int y)
+{
+ *(uint32_t *)p = 0xdeadbeef - (x == 0 ? 0x10001 + 2 * s : 0);
+}
+
static void test_input_global(struct context *ctx)
{
const char *src = "COMP\n"
" STORE RGLOBAL.x, TEMP[1].yyyy, TEMP[1]\n"
" RET\n"
" ENDSUB\n";
- void init(void *p, int s, int x, int y) {
- *(uint32_t *)p = 0xdeadbeef;
- }
- void expect(void *p, int s, int x, int y) {
- *(uint32_t *)p = 0xdeadbeef - (x == 0 ? 0x10001 + 2 * s : 0);
- }
uint32_t input[8] = { 0x10001, 0x10002, 0x10003, 0x10004,
0x10005, 0x10006, 0x10007, 0x10008 };
printf("- %s\n", __func__);
init_prog(ctx, 0, 0, 32, src, NULL);
- init_tex(ctx, 0, PIPE_BUFFER, true, PIPE_FORMAT_R32_FLOAT, 32, 0, init);
- init_tex(ctx, 1, PIPE_BUFFER, true, PIPE_FORMAT_R32_FLOAT, 32, 0, init);
- init_tex(ctx, 2, PIPE_BUFFER, true, PIPE_FORMAT_R32_FLOAT, 32, 0, init);
- init_tex(ctx, 3, PIPE_BUFFER, true, PIPE_FORMAT_R32_FLOAT, 32, 0, init);
+ init_tex(ctx, 0, PIPE_BUFFER, true, PIPE_FORMAT_R32_FLOAT, 32, 0,
+ test_default_init);
+ init_tex(ctx, 1, PIPE_BUFFER, true, PIPE_FORMAT_R32_FLOAT, 32, 0,
+ test_default_init);
+ init_tex(ctx, 2, PIPE_BUFFER, true, PIPE_FORMAT_R32_FLOAT, 32, 0,
+ test_default_init);
+ init_tex(ctx, 3, PIPE_BUFFER, true, PIPE_FORMAT_R32_FLOAT, 32, 0,
+ test_default_init);
init_globals(ctx, (int []){ 0, 1, 2, 3, -1 },
(uint32_t *[]){ &input[1], &input[3],
&input[5], &input[7] });
launch_grid(ctx, (uint []){4, 1, 1}, (uint []){1, 1, 1}, 0, input);
- check_tex(ctx, 0, expect, NULL);
- check_tex(ctx, 1, expect, NULL);
- check_tex(ctx, 2, expect, NULL);
- check_tex(ctx, 3, expect, NULL);
+ check_tex(ctx, 0, test_input_global_expect, NULL);
+ check_tex(ctx, 1, test_input_global_expect, NULL);
+ check_tex(ctx, 2, test_input_global_expect, NULL);
+ check_tex(ctx, 3, test_input_global_expect, NULL);
destroy_globals(ctx);
destroy_tex(ctx);
destroy_prog(ctx);
}
+/* test_private */
+static void test_private_expect(void *p, int s, int x, int y)
+{
+ *(uint32_t *)p = (x / 32) + x % 32;
+}
+
static void test_private(struct context *ctx)
{
const char *src = "COMP\n"
" ENDLOOP\n"
" RET\n"
" ENDSUB\n";
- void init(void *p, int s, int x, int y) {
- *(uint32_t *)p = 0xdeadbeef;
- }
- void expect(void *p, int s, int x, int y) {
- *(uint32_t *)p = (x / 32) + x % 32;
- }
printf("- %s\n", __func__);
init_prog(ctx, 0, 128, 0, src, NULL);
init_tex(ctx, 0, PIPE_BUFFER, true, PIPE_FORMAT_R32_FLOAT,
- 32768, 0, init);
+ 32768, 0, test_default_init);
init_compute_resources(ctx, (int []) { 0, -1 });
launch_grid(ctx, (uint []){16, 1, 1}, (uint []){16, 1, 1}, 0, NULL);
- check_tex(ctx, 0, expect, NULL);
+ check_tex(ctx, 0, test_private_expect, NULL);
destroy_compute_resources(ctx);
destroy_tex(ctx);
destroy_prog(ctx);
}
+/* test_local */
+static void test_local_expect(void *p, int s, int x, int y)
+{
+ *(uint32_t *)p = x & 0x20 ? 2 : 1;
+}
+
static void test_local(struct context *ctx)
{
const char *src = "COMP\n"
" STORE RES[0].x, TEMP[1], TEMP[0]\n"
" RET\n"
" ENDSUB\n";
- void init(void *p, int s, int x, int y) {
- *(uint32_t *)p = 0xdeadbeef;
- }
- void expect(void *p, int s, int x, int y) {
- *(uint32_t *)p = x & 0x20 ? 2 : 1;
- }
printf("- %s\n", __func__);
init_prog(ctx, 256, 0, 0, src, NULL);
init_tex(ctx, 0, PIPE_BUFFER, true, PIPE_FORMAT_R32_FLOAT,
- 4096, 0, init);
+ 4096, 0, test_default_init);
init_compute_resources(ctx, (int []) { 0, -1 });
launch_grid(ctx, (uint []){64, 1, 1}, (uint []){16, 1, 1}, 0, NULL);
- check_tex(ctx, 0, expect, NULL);
+ check_tex(ctx, 0, test_local_expect, NULL);
destroy_compute_resources(ctx);
destroy_tex(ctx);
destroy_prog(ctx);
}
+/* test_sample */
+static void test_sample_init(void *p, int s, int x, int y)
+{
+ *(float *)p = s ? 1 : x * y;
+}
+
+static void test_sample_expect(void *p, int s, int x, int y)
+{
+ switch (x % 4) {
+ case 0:
+ *(float *)p = x / 4 * y;
+ break;
+ case 1:
+ case 2:
+ *(float *)p = 0;
+ break;
+ case 3:
+ *(float *)p = 1;
+ break;
+ }
+}
+
static void test_sample(struct context *ctx)
{
const char *src = "COMP\n"
" STORE RES[0].xyzw, TEMP[0], TEMP[1]\n"
" RET\n"
" ENDSUB\n";
- void init(void *p, int s, int x, int y) {
- *(float *)p = s ? 1 : x * y;
- }
- void expect(void *p, int s, int x, int y) {
- switch (x % 4) {
- case 0:
- *(float *)p = x / 4 * y;
- break;
- case 1:
- case 2:
- *(float *)p = 0;
- break;
- case 3:
- *(float *)p = 1;
- break;
- }
- }
printf("- %s\n", __func__);
init_prog(ctx, 0, 0, 0, src, NULL);
init_tex(ctx, 0, PIPE_TEXTURE_2D, true, PIPE_FORMAT_R32_FLOAT,
- 128, 32, init);
+ 128, 32, test_sample_init);
init_tex(ctx, 1, PIPE_TEXTURE_2D, true, PIPE_FORMAT_R32_FLOAT,
- 512, 32, init);
+ 512, 32, test_sample_init);
init_compute_resources(ctx, (int []) { 1, -1 });
init_sampler_views(ctx, (int []) { 0, -1 });
init_sampler_states(ctx, 2);
launch_grid(ctx, (uint []){1, 1, 1}, (uint []){128, 32, 1}, 0, NULL);
- check_tex(ctx, 1, expect, NULL);
+ check_tex(ctx, 1, test_sample_expect, NULL);
destroy_sampler_states(ctx);
destroy_sampler_views(ctx);
destroy_compute_resources(ctx);
destroy_prog(ctx);
}
+/* test_many_kern */
+static void test_many_kern_expect(void *p, int s, int x, int y)
+{
+ *(uint32_t *)p = x;
+}
+
static void test_many_kern(struct context *ctx)
{
const char *src = "COMP\n"
" STORE RES[0].x, TEMP[0], IMM[0].wwww\n"
" RET\n"
" ENDSUB\n";
- void init(void *p, int s, int x, int y) {
- *(uint32_t *)p = 0xdeadbeef;
- }
- void expect(void *p, int s, int x, int y) {
- *(uint32_t *)p = x;
- }
printf("- %s\n", __func__);
init_prog(ctx, 0, 0, 0, src, NULL);
init_tex(ctx, 0, PIPE_BUFFER, true, PIPE_FORMAT_R32_FLOAT,
- 16, 0, init);
+ 16, 0, test_default_init);
init_compute_resources(ctx, (int []) { 0, -1 });
launch_grid(ctx, (uint []){1, 1, 1}, (uint []){1, 1, 1}, 0, NULL);
launch_grid(ctx, (uint []){1, 1, 1}, (uint []){1, 1, 1}, 5, NULL);
launch_grid(ctx, (uint []){1, 1, 1}, (uint []){1, 1, 1}, 10, NULL);
launch_grid(ctx, (uint []){1, 1, 1}, (uint []){1, 1, 1}, 15, NULL);
- check_tex(ctx, 0, expect, NULL);
+ check_tex(ctx, 0, test_many_kern_expect, NULL);
destroy_compute_resources(ctx);
destroy_tex(ctx);
destroy_prog(ctx);
}
+/* test_constant */
+static void test_constant_init(void *p, int s, int x, int y)
+{
+ *(float *)p = s ? 0xdeadbeef : 8.0 - (float)x;
+}
+
+static void test_constant_expect(void *p, int s, int x, int y)
+{
+ *(float *)p = 8.0 - (float)x;
+}
+
static void test_constant(struct context *ctx)
{
const char *src = "COMP\n"
" STORE RES[1].x, TEMP[0], TEMP[1]\n"
" RET\n"
" ENDSUB\n";
- void init(void *p, int s, int x, int y) {
- *(float *)p = s ? 0xdeadbeef : 8.0 - (float)x;
- }
- void expect(void *p, int s, int x, int y) {
- *(float *)p = 8.0 - (float)x;
- }
printf("- %s\n", __func__);
init_prog(ctx, 0, 0, 0, src, NULL);
init_tex(ctx, 0, PIPE_BUFFER, false, PIPE_FORMAT_R32_FLOAT,
- 256, 0, init);
+ 256, 0, test_constant_init);
init_tex(ctx, 1, PIPE_BUFFER, true, PIPE_FORMAT_R32_FLOAT,
- 256, 0, init);
+ 256, 0, test_constant_init);
init_compute_resources(ctx, (int []) { 0, 1, -1 });
launch_grid(ctx, (uint []){1, 1, 1}, (uint []){64, 1, 1}, 0, NULL);
- check_tex(ctx, 1, expect, NULL);
+ check_tex(ctx, 1, test_constant_expect, NULL);
destroy_compute_resources(ctx);
destroy_tex(ctx);
destroy_prog(ctx);
}
+/* test_resource_indirect */
+static void test_resource_indirect_init(void *p, int s, int x, int y)
+{
+ *(uint32_t *)p = s == 0 ? 0xdeadbeef :
+ s == 1 ? x % 2 :
+ s == 2 ? 2 * x :
+ 2 * x + 1;
+}
+
+static void test_resource_indirect_expect(void *p, int s, int x, int y)
+{
+ *(uint32_t *)p = 2 * x + (x % 2 ? 1 : 0);
+}
+
static void test_resource_indirect(struct context *ctx)
{
const char *src = "COMP\n"
" STORE RES[0].x, TEMP[0], TEMP[1]\n"
" RET\n"
" ENDSUB\n";
- void init(void *p, int s, int x, int y) {
- *(uint32_t *)p = s == 0 ? 0xdeadbeef :
- s == 1 ? x % 2 :
- s == 2 ? 2 * x :
- 2 * x + 1;
- }
- void expect(void *p, int s, int x, int y) {
- *(uint32_t *)p = 2 * x + (x % 2 ? 1 : 0);
- }
printf("- %s\n", __func__);
init_prog(ctx, 0, 0, 0, src, NULL);
init_tex(ctx, 0, PIPE_BUFFER, true, PIPE_FORMAT_R32_FLOAT,
- 256, 0, init);
+ 256, 0, test_resource_indirect_init);
init_tex(ctx, 1, PIPE_BUFFER, false, PIPE_FORMAT_R32_FLOAT,
- 256, 0, init);
+ 256, 0, test_resource_indirect_init);
init_tex(ctx, 2, PIPE_BUFFER, false, PIPE_FORMAT_R32_FLOAT,
- 256, 0, init);
+ 256, 0, test_resource_indirect_init);
init_tex(ctx, 3, PIPE_BUFFER, false, PIPE_FORMAT_R32_FLOAT,
- 256, 0, init);
+ 256, 0, test_resource_indirect_init);
init_compute_resources(ctx, (int []) { 0, 1, 2, 3, -1 });
launch_grid(ctx, (uint []){1, 1, 1}, (uint []){64, 1, 1}, 0, NULL);
- check_tex(ctx, 0, expect, NULL);
+ check_tex(ctx, 0, test_resource_indirect_expect, NULL);
destroy_compute_resources(ctx);
destroy_tex(ctx);
destroy_prog(ctx);
}
+/* test_surface_ld */
enum pipe_format surface_fmts[] = {
PIPE_FORMAT_B8G8R8A8_UNORM,
PIPE_FORMAT_B8G8R8X8_UNORM,
PIPE_FORMAT_R32G32B32A32_SINT
};
+static void test_surface_ld_init0f(void *p, int s, int x, int y)
+{
+ float v[] = { 1.0, -.75, .50, -.25 };
+ int i = 0;
+
+ util_format_write_4f(surface_fmts[i], v, 0, p, 0, 0, 0, 1, 1);
+}
+
+static void test_surface_ld_init0i(void *p, int s, int x, int y)
+{
+ int v[] = { 0xffffffff, 0xffff, 0xff, 0xf };
+ int i = 0;
+
+ util_format_write_4i(surface_fmts[i], v, 0, p, 0, 0, 0, 1, 1);
+}
+
+static void test_surface_ld_expectf(void *p, int s, int x, int y)
+{
+ float v[4], w[4];
+ int i = 0;
+
+ test_surface_ld_init0f(v, s, x / 4, y);
+ util_format_read_4f(surface_fmts[i], w, 0, v, 0, 0, 0, 1, 1);
+ *(float *)p = w[x % 4];
+}
+
+static void test_surface_ld_expecti(void *p, int s, int x, int y)
+{
+ int32_t v[4], w[4];
+ int i = 0;
+
+ test_surface_ld_init0i(v, s, x / 4, y);
+ util_format_read_4i(surface_fmts[i], w, 0, v, 0, 0, 0, 1, 1);
+ *(uint32_t *)p = w[x % 4];
+}
+
static void test_surface_ld(struct context *ctx)
{
const char *src = "COMP\n"
" RET\n"
" ENDSUB\n";
int i = 0;
- void init0f(void *p, int s, int x, int y) {
- float v[] = { 1.0, -.75, .50, -.25 };
- util_format_write_4f(surface_fmts[i], v, 0,
- p, 0, 0, 0, 1, 1);
- }
- void init0i(void *p, int s, int x, int y) {
- int v[] = { 0xffffffff, 0xffff, 0xff, 0xf };
- util_format_write_4i(surface_fmts[i], v, 0,
- p, 0, 0, 0, 1, 1);
- }
- void init1(void *p, int s, int x, int y) {
- *(uint32_t *)p = 0xdeadbeef;
- }
- void expectf(void *p, int s, int x, int y) {
- float v[4], w[4];
- init0f(v, s, x / 4, y);
- util_format_read_4f(surface_fmts[i], w, 0,
- v, 0, 0, 0, 1, 1);
- *(float *)p = w[x % 4];
- }
- void expecti(void *p, int s, int x, int y) {
- int32_t v[4], w[4];
- init0i(v, s, x / 4, y);
- util_format_read_4i(surface_fmts[i], w, 0,
- v, 0, 0, 0, 1, 1);
- *(uint32_t *)p = w[x % 4];
- }
printf("- %s\n", __func__);
}
init_tex(ctx, 0, PIPE_TEXTURE_2D, true, surface_fmts[i],
- 128, 32, (is_int ? init0i : init0f));
+ 128, 32, (is_int ? test_surface_ld_init0i : test_surface_ld_init0f));
init_tex(ctx, 1, PIPE_TEXTURE_2D, true, PIPE_FORMAT_R32_FLOAT,
- 512, 32, init1);
+ 512, 32, test_default_init);
init_compute_resources(ctx, (int []) { 0, 1, -1 });
init_sampler_states(ctx, 2);
launch_grid(ctx, (uint []){1, 1, 1}, (uint []){128, 32, 1}, 0,
NULL);
- check_tex(ctx, 1, (is_int ? expecti : expectf), NULL);
+ check_tex(ctx, 1, (is_int ? test_surface_ld_expecti : test_surface_ld_expectf), NULL);
destroy_sampler_states(ctx);
destroy_compute_resources(ctx);
destroy_tex(ctx);
destroy_prog(ctx);
}
+/* test_surface_st */
+static void test_surface_st_init0f(void *p, int s, int x, int y)
+{
+ float v[] = { 1.0, -.75, 0.5, -.25 };
+ *(float *)p = v[x % 4];
+}
+
+static void test_surface_st_init0i(void *p, int s, int x, int y)
+{
+ int v[] = { 0xffffffff, 0xffff, 0xff, 0xf };
+ *(int32_t *)p = v[x % 4];
+}
+
+static void test_surface_st_init1(void *p, int s, int x, int y)
+{
+ int i = 0;
+ memset(p, 1, util_format_get_blocksize(surface_fmts[i]));
+}
+
+static void test_surface_st_expectf(void *p, int s, int x, int y)
+{
+ float vf[4];
+ int i = 0, j;
+
+ for (j = 0; j < 4; j++)
+ test_surface_st_init0f(&vf[j], s, 4 * x + j, y);
+ util_format_write_4f(surface_fmts[i], vf, 0, p, 0, 0, 0, 1, 1);
+}
+
+static void test_surface_st_expects(void *p, int s, int x, int y)
+{
+ int32_t v[4];
+ int i = 0, j;
+
+ for (j = 0; j < 4; j++)
+ test_surface_st_init0i(&v[j], s, 4 * x + j, y);
+ util_format_write_4i(surface_fmts[i], v, 0, p, 0, 0, 0, 1, 1);
+}
+
+static void test_surface_st_expectu(void *p, int s, int x, int y)
+{
+ uint32_t v[4];
+ int i = 0, j;
+
+ for (j = 0; j < 4; j++)
+ test_surface_st_init0i(&v[j], s, 4 * x + j, y);
+ util_format_write_4ui(surface_fmts[i], v, 0, p, 0, 0, 0, 1, 1);
+}
+
+static bool test_surface_st_check(void *x, void *y, int sz)
+{
+ int i = 0, j;
+
+ if (util_format_is_float(surface_fmts[i])) {
+ return fabs(*(float *)x - *(float *)y) < 3.92156863e-3;
+
+ } else if ((sz % 4) == 0) {
+ for (j = 0; j < sz / 4; j++)
+ if (abs(((uint32_t *)x)[j] -
+ ((uint32_t *)y)[j]) > 1)
+ return false;
+ return true;
+ } else {
+ return !memcmp(x, y, sz);
+ }
+}
+
static void test_surface_st(struct context *ctx)
{
const char *src = "COMP\n"
" RET\n"
" ENDSUB\n";
int i = 0;
- void init0f(void *p, int s, int x, int y) {
- float v[] = { 1.0, -.75, 0.5, -.25 };
- *(float *)p = v[x % 4];
- }
- void init0i(void *p, int s, int x, int y) {
- int v[] = { 0xffffffff, 0xffff, 0xff, 0xf };
- *(int32_t *)p = v[x % 4];
- }
- void init1(void *p, int s, int x, int y) {
- memset(p, 1, util_format_get_blocksize(surface_fmts[i]));
- }
- void expectf(void *p, int s, int x, int y) {
- float vf[4];
- int j;
-
- for (j = 0; j < 4; j++)
- init0f(&vf[j], s, 4 * x + j, y);
- util_format_write_4f(surface_fmts[i], vf, 0,
- p, 0, 0, 0, 1, 1);
- }
- void expects(void *p, int s, int x, int y) {
- int32_t v[4];
- int j;
-
- for (j = 0; j < 4; j++)
- init0i(&v[j], s, 4 * x + j, y);
- util_format_write_4i(surface_fmts[i], v, 0,
- p, 0, 0, 0, 1, 1);
- }
- void expectu(void *p, int s, int x, int y) {
- uint32_t v[4];
- int j;
-
- for (j = 0; j < 4; j++)
- init0i(&v[j], s, 4 * x + j, y);
- util_format_write_4ui(surface_fmts[i], v, 0,
- p, 0, 0, 0, 1, 1);
- }
- bool check(void *x, void *y, int sz) {
- int j;
-
- if (util_format_is_float(surface_fmts[i])) {
- return fabs(*(float *)x - *(float *)y) < 3.92156863e-3;
-
- } else if ((sz % 4) == 0) {
- for (j = 0; j < sz / 4; j++)
- if (abs(((uint32_t *)x)[j] -
- ((uint32_t *)y)[j]) > 1)
- return false;
- return true;
- } else {
- return !memcmp(x, y, sz);
- }
- }
printf("- %s\n", __func__);
}
init_tex(ctx, 0, PIPE_TEXTURE_2D, true, PIPE_FORMAT_R32_FLOAT,
- 512, 32, (is_int ? init0i : init0f));
+ 512, 32, (is_int ? test_surface_st_init0i : test_surface_st_init0f));
init_tex(ctx, 1, PIPE_TEXTURE_2D, true, surface_fmts[i],
- 128, 32, init1);
+ 128, 32, test_surface_st_init1);
init_compute_resources(ctx, (int []) { 0, 1, -1 });
init_sampler_states(ctx, 2);
launch_grid(ctx, (uint []){1, 1, 1}, (uint []){128, 32, 1}, 0,
NULL);
- check_tex(ctx, 1, (is_int && is_signed ? expects :
- is_int && !is_signed ? expectu :
- expectf), check);
+ check_tex(ctx, 1, (is_int && is_signed ? test_surface_st_expects :
+ is_int && !is_signed ? test_surface_st_expectu :
+ test_surface_st_expectf), test_surface_st_check);
destroy_sampler_states(ctx);
destroy_compute_resources(ctx);
destroy_tex(ctx);
destroy_prog(ctx);
}
+/* test_barrier */
+static void test_barrier_expect(void *p, int s, int x, int y)
+{
+ *(uint32_t *)p = 31;
+}
+
static void test_barrier(struct context *ctx)
{
const char *src = "COMP\n"
" STORE RES[0].x, TEMP[1], TEMP[0]\n"
" RET\n"
" ENDSUB\n";
- void init(void *p, int s, int x, int y) {
- *(uint32_t *)p = 0xdeadbeef;
- }
- void expect(void *p, int s, int x, int y) {
- *(uint32_t *)p = 31;
- }
printf("- %s\n", __func__);
init_prog(ctx, 256, 0, 0, src, NULL);
init_tex(ctx, 0, PIPE_BUFFER, true, PIPE_FORMAT_R32_FLOAT,
- 4096, 0, init);
+ 4096, 0, test_default_init);
init_compute_resources(ctx, (int []) { 0, -1 });
launch_grid(ctx, (uint []){64, 1, 1}, (uint []){16, 1, 1}, 0, NULL);
- check_tex(ctx, 0, expect, NULL);
+ check_tex(ctx, 0, test_barrier_expect, NULL);
destroy_compute_resources(ctx);
destroy_tex(ctx);
destroy_prog(ctx);
}
+/* test_atom_ops */
+static void test_atom_ops_init(void *p, int s, int x, int y)
+{
+ *(uint32_t *)p = 0xbad;
+}
+
+static void test_atom_ops_expect(void *p, int s, int x, int y)
+{
+ switch (x) {
+ case 0:
+ *(uint32_t *)p = 0xce6c8eef;
+ break;
+ case 1:
+ *(uint32_t *)p = 0xdeadbeef;
+ break;
+ case 2:
+ *(uint32_t *)p = 0x11111111;
+ break;
+ case 3:
+ *(uint32_t *)p = 0x10011001;
+ break;
+ case 4:
+ *(uint32_t *)p = 0xdfbdbfff;
+ break;
+ case 5:
+ *(uint32_t *)p = 0x11111111;
+ break;
+ case 6:
+ *(uint32_t *)p = 0x11111111;
+ break;
+ case 7:
+ *(uint32_t *)p = 0xdeadbeef;
+ break;
+ case 8:
+ *(uint32_t *)p = 0xdeadbeef;
+ break;
+ case 9:
+ *(uint32_t *)p = 0x11111111;
+ break;
+ }
+}
+
static void test_atom_ops(struct context *ctx, bool global)
{
const char *src = "COMP\n"
" RET\n"
" ENDSUB\n";
- void init(void *p, int s, int x, int y) {
- *(uint32_t *)p = 0xbad;
- }
- void expect(void *p, int s, int x, int y) {
- switch (x) {
- case 0:
- *(uint32_t *)p = 0xce6c8eef;
- break;
- case 1:
- *(uint32_t *)p = 0xdeadbeef;
- break;
- case 2:
- *(uint32_t *)p = 0x11111111;
- break;
- case 3:
- *(uint32_t *)p = 0x10011001;
- break;
- case 4:
- *(uint32_t *)p = 0xdfbdbfff;
- break;
- case 5:
- *(uint32_t *)p = 0x11111111;
- break;
- case 6:
- *(uint32_t *)p = 0x11111111;
- break;
- case 7:
- *(uint32_t *)p = 0xdeadbeef;
- break;
- case 8:
- *(uint32_t *)p = 0xdeadbeef;
- break;
- case 9:
- *(uint32_t *)p = 0x11111111;
- break;
- }
- }
-
printf("- %s (%s)\n", __func__, global ? "global" : "local");
init_prog(ctx, 40, 0, 0, src,
(global ? "-DTARGET_GLOBAL" : "-DTARGET_LOCAL"));
init_tex(ctx, 0, PIPE_BUFFER, true, PIPE_FORMAT_R32_FLOAT,
- 40, 0, init);
+ 40, 0, test_atom_ops_init);
init_compute_resources(ctx, (int []) { 0, -1 });
launch_grid(ctx, (uint []){10, 1, 1}, (uint []){1, 1, 1}, 0, NULL);
- check_tex(ctx, 0, expect, NULL);
+ check_tex(ctx, 0, test_atom_ops_expect, NULL);
destroy_compute_resources(ctx);
destroy_tex(ctx);
destroy_prog(ctx);
}
+/* test_atom_race */
+static void test_atom_race_expect(void *p, int s, int x, int y)
+{
+ *(uint32_t *)p = x & 0x20 ? 0x11111111 : 0xffffffff;
+}
+
static void test_atom_race(struct context *ctx, bool global)
{
const char *src = "COMP\n"
" RET\n"
" ENDSUB\n";
- void init(void *p, int s, int x, int y) {
- *(uint32_t *)p = 0xdeadbeef;
- }
- void expect(void *p, int s, int x, int y) {
- *(uint32_t *)p = x & 0x20 ? 0x11111111 : 0xffffffff;
- }
-
printf("- %s (%s)\n", __func__, global ? "global" : "local");
init_prog(ctx, 256, 0, 0, src,
(global ? "-DTARGET_GLOBAL" : "-DTARGET_LOCAL"));
init_tex(ctx, 0, PIPE_BUFFER, true, PIPE_FORMAT_R32_FLOAT,
- 4096, 0, init);
+ 4096, 0, test_default_init);
init_compute_resources(ctx, (int []) { 0, -1 });
launch_grid(ctx, (uint []){64, 1, 1}, (uint []){16, 1, 1}, 0, NULL);
- check_tex(ctx, 0, expect, NULL);
+ check_tex(ctx, 0, test_atom_race_expect, NULL);
destroy_compute_resources(ctx);
destroy_tex(ctx);
destroy_prog(ctx);
return NULL;
}
+static bool amdgpu_bo_is_user_ptr(struct pb_buffer *buf)
+{
+ return ((struct amdgpu_winsys_bo*)buf)->user_ptr != NULL;
+}
+
static uint64_t amdgpu_bo_get_va(struct pb_buffer *buf)
{
return ((struct amdgpu_winsys_bo*)buf)->va;
ws->base.buffer_create = amdgpu_bo_create;
ws->base.buffer_from_handle = amdgpu_bo_from_handle;
ws->base.buffer_from_ptr = amdgpu_bo_from_ptr;
+ ws->base.buffer_is_user_ptr = amdgpu_bo_is_user_ptr;
ws->base.buffer_get_handle = amdgpu_bo_get_handle;
ws->base.buffer_get_virtual_address = amdgpu_bo_get_va;
ws->base.buffer_get_initial_domain = amdgpu_bo_get_initial_domain;
#include "nouveau/nouveau_winsys.h"
#include "nouveau/nouveau_screen.h"
+#include <nvif/class.h>
+#include <nvif/cl0080.h>
+
static struct util_hash_table *fd_tab = NULL;
pipe_static_mutex(nouveau_screen_mutex);
ret = --screen->refcount;
assert(ret >= 0);
if (ret == 0)
- util_hash_table_remove(fd_tab, intptr_to_pointer(screen->device->fd));
+ util_hash_table_remove(fd_tab, intptr_to_pointer(screen->drm->fd));
pipe_mutex_unlock(nouveau_screen_mutex);
return ret == 0;
}
PUBLIC struct pipe_screen *
nouveau_drm_screen_create(int fd)
{
+ struct nouveau_drm *drm = NULL;
struct nouveau_device *dev = NULL;
- struct pipe_screen *(*init)(struct nouveau_device *);
- struct nouveau_screen *screen;
- int ret, dupfd = -1;
+ struct nouveau_screen *(*init)(struct nouveau_device *);
+ struct nouveau_screen *screen = NULL;
+ int ret, dupfd;
pipe_mutex_lock(nouveau_screen_mutex);
if (!fd_tab) {
fd_tab = util_hash_table_create(hash_fd, compare_fd);
- if (!fd_tab)
- goto err;
+ if (!fd_tab) {
+ pipe_mutex_unlock(nouveau_screen_mutex);
+ return NULL;
+ }
}
screen = util_hash_table_get(fd_tab, intptr_to_pointer(fd));
* creation error.
*/
dupfd = dup(fd);
- ret = nouveau_device_wrap(dupfd, 1, &dev);
+
+ ret = nouveau_drm_new(dupfd, &drm);
+ if (ret)
+ goto err;
+
+ ret = nouveau_device_new(&drm->client, NV_DEVICE,
+ &(struct nv_device_v0) {
+ .device = ~0ULL,
+ }, sizeof(struct nv_device_v0), &dev);
if (ret)
goto err;
goto err;
}
- screen = (struct nouveau_screen*)init(dev);
- if (!screen)
+ screen = init(dev);
+ if (!screen || !screen->base.context_create)
goto err;
/* Use dupfd in hash table, to avoid errors if the original fd gets
return &screen->base;
err:
- if (dev)
+ if (screen) {
+ screen->base.destroy(&screen->base);
+ } else {
nouveau_device_del(&dev);
- else if (dupfd >= 0)
+ nouveau_drm_del(&drm);
close(dupfd);
+ }
pipe_mutex_unlock(nouveau_screen_mutex);
return NULL;
}
pipe_mutex_unlock(rws->bo_handles_mutex);
pb_reference(&b, &old_bo->base);
- return b;
+ return radeon_bo(b);
}
util_hash_table_set(rws->bo_vas, (void*)(uintptr_t)bo->va, bo);
else if (initial_domains & RADEON_DOMAIN_GTT)
rws->allocated_gtt += align(size, rws->size_align);
- return &bo->base;
+ return bo;
}
bool radeon_bo_can_reclaim(struct pb_buffer *_buf)
usage |= 1 << (flags + 3);
if (use_reusable_pool) {
- bo = pb_cache_reclaim_buffer(&ws->bo_cache, size, alignment, usage);
+ bo = radeon_bo(pb_cache_reclaim_buffer(&ws->bo_cache, size, alignment, usage));
if (bo)
- return bo;
+ return &bo->base;
}
bo = radeon_create_bo(ws, size, alignment, usage, domain, flags);
if (ws->info.r600_virtual_address) {
struct drm_radeon_gem_va va;
- bo->va = radeon_bomgr_find_va(rws, bo->base.size, 1 << 20);
+ bo->va = radeon_bomgr_find_va(ws, bo->base.size, 1 << 20);
va.handle = bo->handle;
va.operation = RADEON_VA_MAP;
if (ws->info.r600_virtual_address && !bo->va) {
struct drm_radeon_gem_va va;
- bo->va = radeon_bomgr_find_va(rws, bo->base.size, 1 << 20);
+ bo->va = radeon_bomgr_find_va(ws, bo->base.size, 1 << 20);
va.handle = bo->handle;
va.operation = RADEON_VA_MAP;
return TRUE;
}
+static bool radeon_winsys_bo_is_user_ptr(struct pb_buffer *buf)
+{
+ return ((struct radeon_bo*)buf)->user_ptr != NULL;
+}
+
static uint64_t radeon_winsys_bo_va(struct pb_buffer *buf)
{
return ((struct radeon_bo*)buf)->va;
ws->base.buffer_create = radeon_winsys_bo_create;
ws->base.buffer_from_handle = radeon_winsys_bo_from_handle;
ws->base.buffer_from_ptr = radeon_winsys_bo_from_ptr;
+ ws->base.buffer_is_user_ptr = radeon_winsys_bo_is_user_ptr;
ws->base.buffer_get_handle = radeon_winsys_bo_get_handle;
ws->base.buffer_get_virtual_address = radeon_winsys_bo_va;
ws->base.buffer_get_initial_domain = radeon_bo_get_initial_domain;
vswc->must_flush = FALSE;
debug_flush_flush(vswc->fctx);
#endif
- swc->hints &= ~SVGA_HINT_FLAG_DRAW_EMITTED;
+ swc->hints &= ~SVGA_HINT_FLAG_CAN_PRE_FLUSH;
vswc->preemptive_flush = FALSE;
vswc->seen_surfaces = 0;
vswc->seen_regions = 0;
if (vmw_swc_add_validate_buffer(vswc, reloc->buffer, flags)) {
vswc->seen_regions += reloc->buffer->size;
- if ((swc->hints & SVGA_HINT_FLAG_DRAW_EMITTED) &&
+ if ((swc->hints & SVGA_HINT_FLAG_CAN_PRE_FLUSH) &&
vswc->seen_regions >= VMW_GMR_POOL_SIZE/5)
vswc->preemptive_flush = TRUE;
}
if (vmw_swc_add_validate_buffer(vswc, pb_buffer, flags)) {
vswc->seen_mobs += pb_buffer->size;
- if ((swc->hints & SVGA_HINT_FLAG_DRAW_EMITTED) &&
+ if ((swc->hints & SVGA_HINT_FLAG_CAN_PRE_FLUSH) &&
vswc->seen_mobs >=
vswc->vws->ioctl.max_mob_memory / VMW_MAX_MOB_MEM_FACTOR)
vswc->preemptive_flush = TRUE;
++vswc->surface.staged;
vswc->seen_surfaces += vsurf->size;
- if ((swc->hints & SVGA_HINT_FLAG_DRAW_EMITTED) &&
+ if ((swc->hints & SVGA_HINT_FLAG_CAN_PRE_FLUSH) &&
vswc->seen_surfaces >=
vswc->vws->ioctl.max_surface_memory / VMW_MAX_SURF_MEM_FACTOR)
vswc->preemptive_flush = TRUE;
glsl_parser.h
glsl_parser.output
glsl_test
+spirv2nir
subtest-cr/
subtest-lf/
subtest-cr-lf/
tests/sampler-types-test \
tests/uniform-initializer-test
-noinst_PROGRAMS = glsl_compiler
+noinst_PROGRAMS = glsl_compiler spirv2nir
tests_blob_test_SOURCES = \
tests/blob_test.c
glsl_parser.h \
$(LIBGLSL_FILES) \
$(NIR_FILES) \
- $(NIR_GENERATED_FILES)
-
+ $(SPIRV_FILES) \
+ $(NIR_GENERATED_FILES) \
+ $(GLSL_TO_NIR_FILES)
libnir_la_SOURCES = \
$(NIR_FILES) \
+ $(SPIRV_FILES) \
$(NIR_GENERATED_FILES)
glsl_compiler_SOURCES = \
$(top_builddir)/src/util/libmesautil.la \
$(PTHREAD_LIBS)
+spirv2nir_SOURCES = \
+ standalone_scaffolding.cpp \
+ standalone_scaffolding.h \
+ nir/spirv2nir.c
+
+spirv2nir_LDADD = \
+ libglsl.la \
+ $(top_builddir)/src/libglsl_util.la \
+ $(PTHREAD_LIBS)
+
glsl_test_SOURCES = \
standalone_scaffolding.cpp \
test.cpp \
nir/nir_builder_opcodes.h: nir/nir_opcodes.py nir/nir_builder_opcodes_h.py
$(MKDIR_GEN)
- $(PYTHON_GEN) $(srcdir)/nir/nir_builder_opcodes_h.py > $@
+ $(PYTHON_GEN) $(srcdir)/nir/nir_builder_opcodes_h.py > $@ || ($(RM) $@; false)
nir/nir_constant_expressions.c: nir/nir_opcodes.py nir/nir_constant_expressions.py
$(MKDIR_GEN)
- $(PYTHON_GEN) $(srcdir)/nir/nir_constant_expressions.py > $@
+ $(PYTHON_GEN) $(srcdir)/nir/nir_constant_expressions.py > $@ || ($(RM) $@; false)
nir/nir_opcodes.h: nir/nir_opcodes.py nir/nir_opcodes_h.py
$(MKDIR_GEN)
- $(PYTHON_GEN) $(srcdir)/nir/nir_opcodes_h.py > $@
+ $(PYTHON_GEN) $(srcdir)/nir/nir_opcodes_h.py > $@ || ($(RM) $@; false)
nir/nir_opcodes.c: nir/nir_opcodes.py nir/nir_opcodes_c.py
$(MKDIR_GEN)
- $(PYTHON_GEN) $(srcdir)/nir/nir_opcodes_c.py > $@
+ $(PYTHON_GEN) $(srcdir)/nir/nir_opcodes_c.py > $@ || ($(RM) $@; false)
nir/nir_opt_algebraic.c: nir/nir_opt_algebraic.py nir/nir_algebraic.py
$(MKDIR_GEN)
- $(PYTHON_GEN) $(srcdir)/nir/nir_opt_algebraic.py > $@
+ $(PYTHON_GEN) $(srcdir)/nir/nir_opt_algebraic.py > $@ || ($(RM) $@; false)
nir_tests_control_flow_tests_SOURCES = \
nir/tests/control_flow_tests.cpp
nir/nir_opt_algebraic.c
NIR_FILES = \
- nir/glsl_to_nir.cpp \
- nir/glsl_to_nir.h \
nir/glsl_types.cpp \
nir/glsl_types.h \
nir/builtin_type_macros.h \
nir/nir_control_flow_private.h \
nir/nir_dominance.c \
nir/nir_from_ssa.c \
+ nir/nir_gather_info.c \
nir/nir_gs_count_vertices.c \
+ nir/nir_inline_functions.c \
nir/nir_intrinsics.c \
nir/nir_intrinsics.h \
nir/nir_instr_set.c \
nir/nir_lower_alu_to_scalar.c \
nir/nir_lower_atomics.c \
nir/nir_lower_clip.c \
+ nir/nir_lower_returns.c \
nir/nir_lower_global_vars_to_local.c \
nir/nir_lower_gs_intrinsics.c \
+ nir/nir_lower_indirect_derefs.c \
nir/nir_lower_load_const_to_scalar.c \
nir/nir_lower_locals_to_regs.c \
nir/nir_lower_idiv.c \
nir/nir_opt_peephole_select.c \
nir/nir_opt_remove_phis.c \
nir/nir_opt_undef.c \
+ nir/nir_phi_builder.c \
+ nir/nir_phi_builder.h \
nir/nir_print.c \
nir/nir_remove_dead_variables.c \
+ nir/nir_repair_ssa.c \
nir/nir_search.c \
nir/nir_search.h \
nir/nir_split_var_copies.c \
nir/shader_enums.h \
nir/shader_enums.c
+SPIRV_FILES = \
+ nir/spirv/nir_spirv.h \
+ nir/spirv/spirv_to_nir.c \
+ nir/spirv/vtn_alu.c \
+ nir/spirv/vtn_cfg.c \
+ nir/spirv/vtn_glsl450.c \
+ nir/spirv/vtn_private.h
+
# libglsl
LIBGLSL_FILES = \
s_expression.cpp \
s_expression.h
+# glsl to nir pass
+GLSL_TO_NIR_FILES = \
+ nir/glsl_to_nir.cpp \
+ nir/glsl_to_nir.h
+
# glsl_compiler
GLSL_COMPILER_CXX_FILES = \
# XXX: Remove this once we build NIR and NIR_FILES.
glsl_sources += [
'nir/glsl_types.cpp',
+ 'nir/shader_enums.c',
]
if env['msvc']:
struct _mesa_glsl_parse_state *state);
const char *name;
+ ast_type_qualifier *layout;
/* List of ast_declarator_list * */
exec_list declarations;
bool is_declaration;
return found;
}
+ virtual ir_visitor_status visit_enter(ir_expression *ir)
+ {
+ /* .length() doesn't actually read anything */
+ if (ir->operation == ir_unop_ssbo_unsized_array_length)
+ return visit_continue_with_parent;
+
+ return visit_continue;
+ }
+
private:
ir_variable *found;
};
case GLSL_TYPE_SAMPLER:
case GLSL_TYPE_IMAGE:
case GLSL_TYPE_INTERFACE:
+ case GLSL_TYPE_FUNCTION:
case GLSL_TYPE_ATOMIC_UINT:
case GLSL_TYPE_SUBROUTINE:
/* I assume a comparison of a struct containing a sampler just
_mesa_glsl_error(loc, state,
"misaligned atomic counter offset");
- var->data.atomic.offset = *offset;
+ var->data.offset = *offset;
*offset += var->type->atomic_size();
} else {
state->is_version(150, 0))
&& strcmp(var->name, "gl_FragCoord") == 0
&& earlier->type == var->type
- && earlier->data.mode == var->data.mode) {
+ && var->data.mode == ir_var_shader_in) {
/* Allow redeclaration of gl_FragCoord for ARB_fcc layout
* qualifiers.
*/
* The number of fields processed. A pointer to the array structure fields is
* stored in \c *fields_ret.
*/
-unsigned
+static unsigned
ast_process_struct_or_iface_block_members(exec_list *instructions,
struct _mesa_glsl_parse_state *state,
exec_list *declarations,
bool allow_reserved_names,
ir_variable_mode var_mode,
ast_type_qualifier *layout,
- unsigned block_stream)
+ unsigned block_stream,
+ unsigned expl_location)
{
unsigned decl_count = 0;
glsl_struct_field *const fields = ralloc_array(state, glsl_struct_field,
decl_count);
+ bool first_member = true;
+ bool first_member_has_explicit_location;
+
unsigned i = 0;
foreach_list_typed (ast_declarator_list, decl_list, link, declarations) {
const char *type_name;
"to struct or interface block members");
}
+ if (is_interface) {
+ if (!first_member) {
+ if (!layout->flags.q.explicit_location &&
+ ((first_member_has_explicit_location &&
+ !qual->flags.q.explicit_location) ||
+ (!first_member_has_explicit_location &&
+ qual->flags.q.explicit_location))) {
+ _mesa_glsl_error(&loc, state,
+ "when block-level location layout qualifier "
+ "is not supplied either all members must "
+ "have a location layout qualifier or all "
+ "members must not have a location layout "
+ "qualifier");
+ }
+ } else {
+ first_member = false;
+ first_member_has_explicit_location =
+ qual->flags.q.explicit_location;
+ }
+ }
+
if (qual->flags.q.std140 ||
qual->flags.q.std430 ||
qual->flags.q.packed ||
validate_array_dimensions(field_type, state, &loc);
fields[i].type = field_type;
fields[i].name = decl->identifier;
- fields[i].location = -1;
fields[i].interpolation =
interpret_interpolation_qualifier(qual, var_mode, state, &loc);
fields[i].centroid = qual->flags.q.centroid ? 1 : 0;
fields[i].patch = qual->flags.q.patch ? 1 : 0;
fields[i].precision = qual->precision;
+ if (qual->flags.q.explicit_location) {
+ unsigned qual_location;
+ if (process_qualifier_constant(state, &loc, "location",
+ qual->location, &qual_location)) {
+ fields[i].location = VARYING_SLOT_VAR0 + qual_location;
+ expl_location = fields[i].location +
+ fields[i].type->count_attribute_slots(false);
+ }
+ } else {
+ if (layout && layout->flags.q.explicit_location) {
+ fields[i].location = expl_location;
+ expl_location += fields[i].type->count_attribute_slots(false);
+ } else {
+ fields[i].location = -1;
+ }
+ }
+
/* Propogate row- / column-major information down the fields of the
* structure or interface block. Structures need this data because
* the structure may contain a structure that contains ... a matrix
state->struct_specifier_depth++;
+ unsigned expl_location = 0;
+ if (layout && layout->flags.q.explicit_location) {
+ if (!process_qualifier_constant(state, &loc, "location",
+ layout->location, &expl_location)) {
+ return NULL;
+ } else {
+ expl_location = VARYING_SLOT_VAR0 + expl_location;
+ }
+ }
+
glsl_struct_field *fields;
unsigned decl_count =
ast_process_struct_or_iface_block_members(instructions,
GLSL_MATRIX_LAYOUT_INHERITED,
false /* allow_reserved_names */,
ir_var_auto,
- NULL,
- 0 /* for interface only */);
+ layout,
+ 0, /* for interface only */
+ expl_location);
validate_identifier(this->name, loc, state);
return NULL;
}
+ unsigned expl_location = 0;
+ if (layout.flags.q.explicit_location) {
+ if (!process_qualifier_constant(state, &loc, "location",
+ layout.location, &expl_location)) {
+ return NULL;
+ } else {
+ expl_location = VARYING_SLOT_VAR0 + expl_location;
+ }
+ }
+
unsigned int num_variables =
ast_process_struct_or_iface_block_members(&declared_variables,
state,
redeclaring_per_vertex,
var_mode,
&this->layout,
- qual_stream);
+ qual_stream,
+ expl_location);
state->struct_specifier_depth--;
}
var->data.stream = qual_stream;
+ if (layout.flags.q.explicit_location) {
+ var->data.location = expl_location;
+ var->data.explicit_location = true;
+ }
state->symbols->add_variable(var);
instructions->push_tail(var);
var->data.sample = fields[i].sample;
var->data.patch = fields[i].patch;
var->data.stream = qual_stream;
+ var->data.location = fields[i].location;
+ if (fields[i].location != -1)
+ var->data.explicit_location = true;
var->init_interface_type(block_type);
if (var_mode == ir_var_shader_in || var_mode == ir_var_uniform)
}
static bool
+v140_or_es3(const _mesa_glsl_parse_state *state)
+{
+ return state->is_version(140, 300);
+}
+
+static bool
v400_fs_only(const _mesa_glsl_parse_state *state)
{
return state->is_version(400, 0) &&
ir_variable *in_var(const glsl_type *type, const char *name);
ir_variable *out_var(const glsl_type *type, const char *name);
ir_constant *imm(float f, unsigned vector_elements=1);
+ ir_constant *imm(bool b, unsigned vector_elements=1);
ir_constant *imm(int i, unsigned vector_elements=1);
ir_constant *imm(unsigned u, unsigned vector_elements=1);
ir_constant *imm(double d, unsigned vector_elements=1);
NULL);
add_function("inverse",
- _inverse_mat2(v120, glsl_type::mat2_type),
- _inverse_mat3(v120, glsl_type::mat3_type),
- _inverse_mat4(v120, glsl_type::mat4_type),
+ _inverse_mat2(v140_or_es3, glsl_type::mat2_type),
+ _inverse_mat3(v140_or_es3, glsl_type::mat3_type),
+ _inverse_mat4(v140_or_es3, glsl_type::mat4_type),
_inverse_mat2(fp64, glsl_type::dmat2_type),
_inverse_mat3(fp64, glsl_type::dmat3_type),
_inverse_mat4(fp64, glsl_type::dmat4_type),
}
ir_constant *
+builtin_builder::imm(bool b, unsigned vector_elements)
+{
+ return new(mem_ctx) ir_constant(b, vector_elements);
+}
+
+ir_constant *
builtin_builder::imm(float f, unsigned vector_elements)
{
return new(mem_ctx) ir_constant(f, vector_elements);
ir_function_signature *
builtin_builder::_any(const glsl_type *type)
{
- return unop(always_available, ir_unop_any, glsl_type::bool_type, type);
+ ir_variable *v = in_var(type, "v");
+ MAKE_SIG(glsl_type::bool_type, always_available, 1, v);
+
+ const unsigned vec_elem = v->type->vector_elements;
+ body.emit(ret(expr(ir_binop_any_nequal, v, imm(false, vec_elem))));
+
+ return sig;
}
ir_function_signature *
ir_variable *v = in_var(type, "v");
MAKE_SIG(glsl_type::bool_type, always_available, 1, v);
- switch (type->vector_elements) {
- case 2:
- body.emit(ret(logic_and(swizzle_x(v), swizzle_y(v))));
- break;
- case 3:
- body.emit(ret(logic_and(logic_and(swizzle_x(v), swizzle_y(v)),
- swizzle_z(v))));
- break;
- case 4:
- body.emit(ret(logic_and(logic_and(logic_and(swizzle_x(v), swizzle_y(v)),
- swizzle_z(v)),
- swizzle_w(v))));
- break;
- }
+ const unsigned vec_elem = v->type->vector_elements;
+ body.emit(ret(expr(ir_binop_all_equal, v, imm(true, vec_elem))));
return sig;
}
ir_function_signature *
builtin_builder::_bitfieldExtract(const glsl_type *type)
{
+ bool is_uint = type->base_type == GLSL_TYPE_UINT;
ir_variable *value = in_var(type, "value");
ir_variable *offset = in_var(glsl_type::int_type, "offset");
ir_variable *bits = in_var(glsl_type::int_type, "bits");
MAKE_SIG(type, gpu_shader5_or_es31, 3, value, offset, bits);
- body.emit(ret(expr(ir_triop_bitfield_extract, value, offset, bits)));
+ operand cast_offset = is_uint ? i2u(offset) : operand(offset);
+ operand cast_bits = is_uint ? i2u(bits) : operand(bits);
+
+ body.emit(ret(expr(ir_triop_bitfield_extract, value,
+ swizzle(cast_offset, SWIZZLE_XXXX, type->vector_elements),
+ swizzle(cast_bits, SWIZZLE_XXXX, type->vector_elements))));
return sig;
}
ir_function_signature *
builtin_builder::_bitfieldInsert(const glsl_type *type)
{
+ bool is_uint = type->base_type == GLSL_TYPE_UINT;
ir_variable *base = in_var(type, "base");
ir_variable *insert = in_var(type, "insert");
ir_variable *offset = in_var(glsl_type::int_type, "offset");
ir_variable *bits = in_var(glsl_type::int_type, "bits");
MAKE_SIG(type, gpu_shader5_or_es31, 4, base, insert, offset, bits);
- body.emit(ret(bitfield_insert(base, insert, offset, bits)));
+ operand cast_offset = is_uint ? i2u(offset) : operand(offset);
+ operand cast_bits = is_uint ? i2u(bits) : operand(bits);
+
+ body.emit(ret(bitfield_insert(base, insert,
+ swizzle(cast_offset, SWIZZLE_XXXX, type->vector_elements),
+ swizzle(cast_bits, SWIZZLE_XXXX, type->vector_elements))));
return sig;
}
add_system_value(SYSTEM_VALUE_INSTANCE_ID, int_t, "gl_InstanceIDARB");
if (state->ARB_draw_instanced_enable || state->is_version(140, 300))
add_system_value(SYSTEM_VALUE_INSTANCE_ID, int_t, "gl_InstanceID");
+ if (state->ARB_shader_draw_parameters_enable) {
+ add_system_value(SYSTEM_VALUE_BASE_VERTEX, int_t, "gl_BaseVertexARB");
+ add_system_value(SYSTEM_VALUE_BASE_INSTANCE, int_t, "gl_BaseInstanceARB");
+ add_system_value(SYSTEM_VALUE_DRAW_ID, int_t, "gl_DrawIDARB");
+ }
if (state->AMD_vertex_shader_layer_enable) {
var = add_output(VARYING_SLOT_LAYER, int_t, "gl_Layer");
var->data.interpolation = INTERP_QUALIFIER_FLAT;
{
ir_variable *var;
- add_input(VARYING_SLOT_POS, vec4_t, "gl_FragCoord");
- add_input(VARYING_SLOT_FACE, bool_t, "gl_FrontFacing");
+ if (this->state->ctx->Const.GLSLFragCoordIsSysVal)
+ add_system_value(SYSTEM_VALUE_FRAG_COORD, vec4_t, "gl_FragCoord");
+ else
+ add_input(VARYING_SLOT_POS, vec4_t, "gl_FragCoord");
+
+ if (this->state->ctx->Const.GLSLFrontFacingIsSysVal)
+ add_system_value(SYSTEM_VALUE_FRONT_FACE, bool_t, "gl_FrontFacing");
+ else
+ add_input(VARYING_SLOT_FACE, bool_t, "gl_FrontFacing");
+
if (state->is_version(120, 100))
add_input(VARYING_SLOT_PNTC, vec2_t, "gl_PointCoord");
if (extensions->ARB_shader_subroutine)
add_builtin_define(parser, "GL_ARB_shader_subroutine", 1);
+
+ if (extensions->ARB_shader_draw_parameters)
+ add_builtin_define(parser, "GL_ARB_shader_draw_parameters", 1);
}
}
$$->set_location_range(@1, @2);
$$->qualifier = $1;
$$->specifier = $2;
+ if ($$->specifier->structure != NULL &&
+ $$->specifier->structure->is_declaration) {
+ $$->specifier->structure->layout = &$$->qualifier;
+ }
}
;
this->extensions = &ctx->Extensions;
+ this->ARB_compute_shader_enable = true;
+
this->Const.MaxLights = ctx->Const.MaxLights;
this->Const.MaxClipPlanes = ctx->Const.MaxClipPlanes;
this->Const.MaxTextureUnits = ctx->Const.MaxTextureUnits;
}
-/**
- * Translate a gl_shader_stage to a short shader stage name for debug
- * printouts and error messages.
- */
-const char *
-_mesa_shader_stage_to_string(unsigned stage)
-{
- switch (stage) {
- case MESA_SHADER_VERTEX: return "vertex";
- case MESA_SHADER_FRAGMENT: return "fragment";
- case MESA_SHADER_GEOMETRY: return "geometry";
- case MESA_SHADER_COMPUTE: return "compute";
- case MESA_SHADER_TESS_CTRL: return "tess ctrl";
- case MESA_SHADER_TESS_EVAL: return "tess eval";
- }
-
- unreachable("Unknown shader stage.");
-}
-
-/**
- * Translate a gl_shader_stage to a shader stage abbreviation (VS, GS, FS)
- * for debug printouts and error messages.
- */
-const char *
-_mesa_shader_stage_to_abbrev(unsigned stage)
-{
- switch (stage) {
- case MESA_SHADER_VERTEX: return "VS";
- case MESA_SHADER_FRAGMENT: return "FS";
- case MESA_SHADER_GEOMETRY: return "GS";
- case MESA_SHADER_COMPUTE: return "CS";
- case MESA_SHADER_TESS_CTRL: return "TCS";
- case MESA_SHADER_TESS_EVAL: return "TES";
- }
-
- unreachable("Unknown shader stage.");
-}
-
/* This helper function will append the given message to the shader's
info log and report it via GL_ARB_debug_output. Per that extension,
'type' is one of the enum values classifying the message, and
EXT(ARB_shader_atomic_counters, true, false, ARB_shader_atomic_counters),
EXT(ARB_shader_bit_encoding, true, false, ARB_shader_bit_encoding),
EXT(ARB_shader_clock, true, false, ARB_shader_clock),
+ EXT(ARB_shader_draw_parameters, true, false, ARB_shader_draw_parameters),
EXT(ARB_shader_image_load_store, true, false, ARB_shader_image_load_store),
EXT(ARB_shader_image_size, true, false, ARB_shader_image_size),
EXT(ARB_shader_precision, true, false, ARB_shader_precision),
bool ARB_shader_bit_encoding_warn;
bool ARB_shader_clock_enable;
bool ARB_shader_clock_warn;
+ bool ARB_shader_draw_parameters_enable;
+ bool ARB_shader_draw_parameters_warn;
bool ARB_shader_image_load_store_enable;
bool ARB_shader_image_load_store_warn;
bool ARB_shader_image_size_enable;
extern "C" {
#endif
-/**
- * Get the textual name of the specified shader stage (which is a
- * gl_shader_stage).
- */
-extern const char *
-_mesa_shader_stage_to_string(unsigned stage);
-
-extern const char *
-_mesa_shader_stage_to_abbrev(unsigned stage);
-
extern int glcpp_preprocess(void *ctx, const char **shader, char **info_log,
const struct gl_extensions *extensions, struct gl_context *gl_ctx);
this->type = glsl_type::uvec2_type;
break;
- case ir_unop_any:
- this->type = glsl_type::bool_type;
- break;
-
case ir_unop_pack_snorm_2x16:
case ir_unop_pack_snorm_4x8:
case ir_unop_pack_unorm_2x16:
case ir_binop_borrow:
case ir_binop_lshift:
case ir_binop_rshift:
- case ir_binop_bfm:
case ir_binop_ldexp:
case ir_binop_interpolate_at_offset:
case ir_binop_interpolate_at_sample:
this->type = op0->type;
break;
- case ir_triop_bfi:
case ir_triop_csel:
this->type = op1->type;
break;
"bitcast_f2i",
"bitcast_u2f",
"bitcast_f2u",
- "any",
"trunc",
"ceil",
"floor",
"max",
"pow",
"packHalf2x16_split",
- "bfm",
"ubo_load",
"ldexp",
"vector_extract",
"fma",
"lrp",
"csel",
- "bfi",
"bitfield_extract",
"vector_insert",
"bitfield_insert",
this->data.mode = mode;
this->data.interpolation = INTERP_QUALIFIER_NONE;
this->data.max_array_access = 0;
- this->data.atomic.offset = 0;
+ this->data.offset = 0;
this->data.precision = GLSL_PRECISION_NONE;
this->data.image_read_only = false;
this->data.image_write_only = false;
/**
* Location an atomic counter is stored at.
*/
- struct {
- unsigned offset;
- } atomic;
+ unsigned offset;
/**
* Highest element accessed with a constant expression array index
ir_unop_bitcast_f2i, /**< Bit-identical float-to-int "conversion" */
ir_unop_bitcast_u2f, /**< Bit-identical uint-to-float "conversion" */
ir_unop_bitcast_f2u, /**< Bit-identical float-to-uint "conversion" */
- ir_unop_any,
/**
* \name Unary floating-point rounding operations.
/*@}*/
/**
- * \name First half of a lowered bitfieldInsert() operation.
- *
- * \see lower_instructions::bitfield_insert_to_bfm_bfi
- */
- /*@{*/
- ir_binop_bfm,
- /*@}*/
-
- /**
* Load a value the size of a given GLSL type from a uniform block.
*
* operand0 is the ir_constant uniform block index in the linked shader.
ir_triop_csel,
/*@}*/
- /**
- * \name Second half of a lowered bitfieldInsert() operation.
- *
- * \see lower_instructions::bitfield_insert_to_bfm_bfi
- */
- /*@{*/
- ir_triop_bfi,
- /*@}*/
-
ir_triop_bitfield_extract,
/**
{
return operation == ir_binop_all_equal ||
operation == ir_binop_any_nequal ||
- operation == ir_unop_any ||
operation == ir_binop_dot ||
+ operation == ir_binop_vector_extract ||
+ operation == ir_triop_vector_insert ||
operation == ir_quadop_vector;
}
return c;
}
+ case GLSL_TYPE_FUNCTION:
case GLSL_TYPE_SAMPLER:
case GLSL_TYPE_IMAGE:
case GLSL_TYPE_ATOMIC_UINT:
data.u[c] = bitcast_f2u(op[0]->value.f[c]);
}
break;
- case ir_unop_any:
- assert(op[0]->type->is_boolean());
- data.b[0] = false;
- for (unsigned c = 0; c < op[0]->type->components(); c++) {
- if (op[0]->value.b[c])
- data.b[0] = true;
- }
- break;
case ir_unop_d2f:
assert(op[0]->type->base_type == GLSL_TYPE_DOUBLE);
for (unsigned c = 0; c < op[0]->type->components(); c++) {
data.i[c] = -1;
else {
int count = 0;
- int top_bit = op[0]->type->base_type == GLSL_TYPE_UINT
- ? 0 : v & (1 << 31);
+ unsigned top_bit = op[0]->type->base_type == GLSL_TYPE_UINT
+ ? 0 : v & (1u << 31);
- while (((v & (1 << 31)) == top_bit) && count != 32) {
+ while (((v & (1u << 31)) == top_bit) && count != 32) {
count++;
v <<= 1;
}
break;
case ir_triop_bitfield_extract: {
- int offset = op[1]->value.i[0];
- int bits = op[2]->value.i[0];
-
for (unsigned c = 0; c < components; c++) {
+ int offset = op[1]->value.i[c];
+ int bits = op[2]->value.i[c];
+
if (bits == 0)
data.u[c] = 0;
else if (offset < 0 || bits < 0)
break;
}
- case ir_binop_bfm: {
- int bits = op[0]->value.i[0];
- int offset = op[1]->value.i[0];
-
- for (unsigned c = 0; c < components; c++) {
- if (bits == 0)
- data.u[c] = op[0]->value.u[c];
- else if (offset < 0 || bits < 0)
- data.u[c] = 0; /* Undefined for bitfieldInsert, per spec. */
- else if (offset + bits > 32)
- data.u[c] = 0; /* Undefined for bitfieldInsert, per spec. */
- else
- data.u[c] = ((1 << bits) - 1) << offset;
- }
- break;
- }
-
case ir_binop_ldexp:
for (unsigned c = 0; c < components; c++) {
if (op[0]->type->base_type == GLSL_TYPE_DOUBLE) {
}
case ir_quadop_bitfield_insert: {
- int offset = op[2]->value.i[0];
- int bits = op[3]->value.i[0];
-
for (unsigned c = 0; c < components; c++) {
+ int offset = op[2]->value.i[c];
+ int bits = op[3]->value.i[c];
+
if (bits == 0)
data.u[c] = op[0]->value.u[c];
else if (offset < 0 || bits < 0)
else if (offset + bits > 32)
data.u[c] = 0; /* Undefined, per spec. */
else {
- unsigned insert_mask = ((1 << bits) - 1) << offset;
+ unsigned insert_mask = ((1ull << bits) - 1) << offset;
unsigned insert = op[1]->value.u[c];
insert <<= offset;
ir_constant *
ir_dereference_variable::constant_expression_value(struct hash_table *variable_context)
{
- /* This may occur during compile and var->type is glsl_type::error_type */
- if (!var)
- return NULL;
+ assert(var);
/* Give priority to the context hashtable, if it exists */
if (variable_context) {
#define LOG_TO_LOG2 0x10
#define MOD_TO_FLOOR 0x20
#define INT_DIV_TO_MUL_RCP 0x40
-#define BITFIELD_INSERT_TO_BFM_BFI 0x80
-#define LDEXP_TO_ARITH 0x100
-#define CARRY_TO_ARITH 0x200
-#define BORROW_TO_ARITH 0x400
-#define SAT_TO_CLAMP 0x800
-#define DOPS_TO_DFRAC 0x1000
-#define DFREXP_DLDEXP_TO_ARITH 0x2000
+#define LDEXP_TO_ARITH 0x80
+#define CARRY_TO_ARITH 0x100
+#define BORROW_TO_ARITH 0x200
+#define SAT_TO_CLAMP 0x400
+#define DOPS_TO_DFRAC 0x800
+#define DFREXP_DLDEXP_TO_ARITH 0x1000
/**
* \see class lower_packing_builtins_visitor
var->data.mode == ir_var_system_value;
}
-static inline bool
-is_dual_slot(ir_variable *var)
-{
- const glsl_type *type = var->type->without_array();
- return type == glsl_type::dvec4_type || type == glsl_type::dvec3_type;
-}
-
static void
mark(struct gl_program *prog, ir_variable *var, int offset, int len,
- bool is_fragment_shader)
+ gl_shader_stage stage)
{
/* As of GLSL 1.20, varyings can only be floats, floating-point
* vectors or matrices, or arrays of them. For Mesa programs using
*/
for (int i = 0; i < len; i++) {
- bool dual_slot = is_dual_slot(var);
int idx = var->data.location + var->data.index + offset + i;
bool is_patch_generic = var->data.patch &&
idx != VARYING_SLOT_TESS_LEVEL_INNER &&
else
prog->InputsRead |= bitfield;
- if (dual_slot)
+ /* double inputs read is only for vertex inputs */
+ if (stage == MESA_SHADER_VERTEX &&
+ var->type->without_array()->is_dual_slot_double())
prog->DoubleInputsRead |= bitfield;
- if (is_fragment_shader) {
+
+ if (stage == MESA_SHADER_FRAGMENT) {
gl_fragment_program *fprog = (gl_fragment_program *) prog;
fprog->InterpQualifier[idx] =
(glsl_interp_qualifier) var->data.interpolation;
ir_set_program_inouts_visitor::mark_whole_variable(ir_variable *var)
{
const glsl_type *type = var->type;
+ bool vertex_input = false;
if (this->shader_stage == MESA_SHADER_GEOMETRY &&
var->data.mode == ir_var_shader_in && type->is_array()) {
type = type->fields.array;
type = type->fields.array;
}
- mark(this->prog, var, 0, type->count_attribute_slots(),
- this->shader_stage == MESA_SHADER_FRAGMENT);
+ if (this->shader_stage == MESA_SHADER_VERTEX &&
+ var->data.mode == ir_var_shader_in)
+ vertex_input = true;
+
+ mark(this->prog, var, 0, type->count_attribute_slots(vertex_input),
+ this->shader_stage);
}
/* Default handler: Mark all the locations in the variable as used. */
return false;
}
+ /* double element width for double types that takes two slots */
+ if (this->shader_stage != MESA_SHADER_VERTEX ||
+ var->data.mode != ir_var_shader_in) {
+ if (type->without_array()->is_dual_slot_double())
+ elem_width *= 2;
+ }
+
mark(this->prog, var, index_as_constant->value.u[0] * elem_width,
- elem_width, this->shader_stage == MESA_SHADER_FRAGMENT);
+ elem_width, this->shader_stage);
return true;
}
assert(ir->type->base_type == GLSL_TYPE_UINT);
break;
- case ir_unop_any:
- assert(ir->operands[0]->type->base_type == GLSL_TYPE_BOOL);
- assert(ir->type == glsl_type::bool_type);
- break;
-
case ir_unop_trunc:
case ir_unop_round_even:
case ir_unop_ceil:
assert(ir->operands[1]->type == glsl_type::float_type);
break;
- case ir_binop_bfm:
- assert(ir->type->is_integer());
- assert(ir->operands[0]->type->is_integer());
- assert(ir->operands[1]->type->is_integer());
- break;
-
case ir_binop_ubo_load:
assert(ir->operands[0]->type == glsl_type::uint_type);
assert(ir->type == ir->operands[2]->type);
break;
- case ir_triop_bfi:
- assert(ir->operands[0]->type->is_integer());
- assert(ir->operands[1]->type == ir->operands[2]->type);
- assert(ir->operands[1]->type == ir->type);
- break;
-
case ir_triop_bitfield_extract:
+ assert(ir->type->is_integer());
assert(ir->operands[0]->type == ir->type);
- assert(ir->operands[1]->type == glsl_type::int_type);
- assert(ir->operands[2]->type == glsl_type::int_type);
+ assert(ir->operands[1]->type == ir->type);
+ assert(ir->operands[2]->type == ir->type);
break;
case ir_triop_vector_insert:
break;
case ir_quadop_bitfield_insert:
+ assert(ir->type->is_integer());
assert(ir->operands[0]->type == ir->type);
assert(ir->operands[1]->type == ir->type);
- assert(ir->operands[2]->type == glsl_type::int_type);
- assert(ir->operands[3]->type == glsl_type::int_type);
+ assert(ir->operands[2]->type == ir->type);
+ assert(ir->operands[3]->type == ir->type);
break;
case ir_quadop_vector:
const active_atomic_counter *const first = (active_atomic_counter *) a;
const active_atomic_counter *const second = (active_atomic_counter *) b;
- return int(first->var->data.atomic.offset) - int(second->var->data.atomic.offset);
+ return int(first->var->data.offset) - int(second->var->data.offset);
}
bool
check_atomic_counters_overlap(const ir_variable *x, const ir_variable *y)
{
- return ((x->data.atomic.offset >= y->data.atomic.offset &&
- x->data.atomic.offset < y->data.atomic.offset + y->type->atomic_size()) ||
- (y->data.atomic.offset >= x->data.atomic.offset &&
- y->data.atomic.offset < x->data.atomic.offset + x->type->atomic_size()));
+ return ((x->data.offset >= y->data.offset &&
+ x->data.offset < y->data.offset + y->type->atomic_size()) ||
+ (y->data.offset >= x->data.offset &&
+ y->data.offset < x->data.offset + x->type->atomic_size()));
}
void
ir_variable *var = node->as_variable();
if (var && var->type->contains_atomic()) {
- int offset = var->data.atomic.offset;
+ int offset = var->data.offset;
unsigned uniform_loc = var->data.location;
process_atomic_variable(var->type, prog, &uniform_loc,
var, buffers, num_buffers, &offset, i);
linker_error(prog, "Atomic counter %s declared at offset %d "
"which is already in use.",
buffers[i].counters[j].var->name,
- buffers[i].counters[j].var->data.atomic.offset);
+ buffers[i].counters[j].var->data.offset);
}
}
}
var->data.binding = i;
storage->atomic_buffer_index = i;
- storage->offset = var->data.atomic.offset;
+ storage->offset = var->data.offset;
storage->array_stride = (var->type->is_array() ?
var->type->without_array()->atomic_size() : 0);
if (!var->type->is_matrix())
#include "glsl_symbol_table.h"
#include "linker.h"
#include "main/macros.h"
-#include "program/hash_table.h"
+#include "util/hash_table.h"
namespace {
/**
- * Information about a single interface block definition that we need to keep
- * track of in order to check linkage rules.
- *
- * Note: this class is expected to be short lived, so it doesn't make copies
- * of the strings it references; it simply borrows the pointers from the
- * ir_variable class.
- */
-struct interface_block_definition
-{
- /**
- * Extract an interface block definition from an ir_variable that
- * represents either the interface instance (for named interfaces), or a
- * member of the interface (for unnamed interfaces).
- */
- explicit interface_block_definition(ir_variable *var)
- : var(var),
- type(var->get_interface_type()),
- instance_name(NULL)
- {
- if (var->is_interface_instance()) {
- instance_name = var->name;
- }
- explicitly_declared = (var->data.how_declared != ir_var_declared_implicitly);
- }
- /**
- * Interface block ir_variable
- */
- ir_variable *var;
-
- /**
- * Interface block type
- */
- const glsl_type *type;
-
- /**
- * For a named interface block, the instance name. Otherwise NULL.
- */
- const char *instance_name;
-
- /**
- * True if this interface block was explicitly declared in the shader;
- * false if it was an implicitly declared built-in interface block.
- */
- bool explicitly_declared;
-};
-
-
-/**
* Check if two interfaces match, according to intrastage interface matching
* rules. If they do, and the first interface uses an unsized array, it will
* be updated to reflect the array size declared in the second interface.
*/
bool
-intrastage_match(interface_block_definition *a,
- const interface_block_definition *b,
- ir_variable_mode mode,
+intrastage_match(ir_variable *a,
+ ir_variable *b,
struct gl_shader_program *prog)
{
/* Types must match. */
- if (a->type != b->type) {
+ if (a->get_interface_type() != b->get_interface_type()) {
/* Exception: if both the interface blocks are implicitly declared,
* don't force their types to match. They might mismatch due to the two
* shaders using different GLSL versions, and that's ok.
*/
- if (a->explicitly_declared || b->explicitly_declared)
+ if (a->data.how_declared != ir_var_declared_implicitly ||
+ b->data.how_declared != ir_var_declared_implicitly)
return false;
}
/* Presence/absence of interface names must match. */
- if ((a->instance_name == NULL) != (b->instance_name == NULL))
+ if (a->is_interface_instance() != b->is_interface_instance())
return false;
/* For uniforms, instance names need not match. For shader ins/outs,
* it's not clear from the spec whether they need to match, but
* Mesa's implementation relies on them matching.
*/
- if (a->instance_name != NULL &&
- mode != ir_var_uniform && mode != ir_var_shader_storage &&
- strcmp(a->instance_name, b->instance_name) != 0) {
+ if (a->is_interface_instance() && b->data.mode != ir_var_uniform &&
+ b->data.mode != ir_var_shader_storage &&
+ strcmp(a->name, b->name) != 0) {
return false;
}
/* If a block is an array then it must match across the shader.
* Unsized arrays are also processed and matched agaist sized arrays.
*/
- if (b->var->type != a->var->type &&
- (b->instance_name != NULL || a->instance_name != NULL) &&
- !validate_intrastage_arrays(prog, b->var, a->var))
+ if (b->type != a->type &&
+ (b->is_interface_instance() || a->is_interface_instance()) &&
+ !validate_intrastage_arrays(prog, b, a))
return false;
return true;
* This is used for tessellation control and geometry shader consumers.
*/
bool
-interstage_match(const interface_block_definition *producer,
- const interface_block_definition *consumer,
+interstage_match(ir_variable *producer,
+ ir_variable *consumer,
bool extra_array_level)
{
/* Unsized arrays should not occur during interstage linking. They
* should have all been assigned a size by link_intrastage_shaders.
*/
- assert(!consumer->var->type->is_unsized_array());
- assert(!producer->var->type->is_unsized_array());
+ assert(!consumer->type->is_unsized_array());
+ assert(!producer->type->is_unsized_array());
/* Types must match. */
- if (consumer->type != producer->type) {
+ if (consumer->get_interface_type() != producer->get_interface_type()) {
/* Exception: if both the interface blocks are implicitly declared,
* don't force their types to match. They might mismatch due to the two
* shaders using different GLSL versions, and that's ok.
*/
- if (consumer->explicitly_declared || producer->explicitly_declared)
+ if (consumer->data.how_declared != ir_var_declared_implicitly ||
+ producer->data.how_declared != ir_var_declared_implicitly)
return false;
}
/* Ignore outermost array if geom shader */
const glsl_type *consumer_instance_type;
if (extra_array_level) {
- consumer_instance_type = consumer->var->type->fields.array;
+ consumer_instance_type = consumer->type->fields.array;
} else {
- consumer_instance_type = consumer->var->type;
+ consumer_instance_type = consumer->type;
}
/* If a block is an array then it must match across shaders.
* Since unsized arrays have been ruled out, we can check this by just
* making sure the types are equal.
*/
- if ((consumer->instance_name != NULL &&
+ if ((consumer->is_interface_instance() &&
consumer_instance_type->is_array()) ||
- (producer->instance_name != NULL &&
- producer->var->type->is_array())) {
- if (consumer_instance_type != producer->var->type)
+ (producer->is_interface_instance() &&
+ producer->type->is_array())) {
+ if (consumer_instance_type != producer->type)
return false;
}
public:
interface_block_definitions()
: mem_ctx(ralloc_context(NULL)),
- ht(hash_table_ctor(0, hash_table_string_hash,
- hash_table_string_compare))
+ ht(_mesa_hash_table_create(NULL, _mesa_key_hash_string,
+ _mesa_key_string_equal))
{
}
~interface_block_definitions()
{
- hash_table_dtor(ht);
ralloc_free(mem_ctx);
+ _mesa_hash_table_destroy(ht, NULL);
}
/**
- * Lookup the interface definition having the given block name. Return
- * NULL if none is found.
+ * Lookup the interface definition. Return NULL if none is found.
*/
- interface_block_definition *lookup(const char *block_name)
+ ir_variable *lookup(ir_variable *var)
{
- return (interface_block_definition *) hash_table_find(ht, block_name);
+ if (var->data.explicit_location &&
+ var->data.location >= VARYING_SLOT_VAR0) {
+ char location_str[11];
+ snprintf(location_str, 11, "%d", var->data.location);
+
+ const struct hash_entry *entry =
+ _mesa_hash_table_search(ht, location_str);
+ return entry ? (ir_variable *) entry->data : NULL;
+ } else {
+ const struct hash_entry *entry =
+ _mesa_hash_table_search(ht, var->get_interface_type()->name);
+ return entry ? (ir_variable *) entry->data : NULL;
+ }
}
/**
* Add a new interface definition.
*/
- void store(const interface_block_definition &def)
+ void store(ir_variable *var)
{
- interface_block_definition *hash_entry =
- rzalloc(mem_ctx, interface_block_definition);
- *hash_entry = def;
- hash_table_insert(ht, hash_entry, def.type->name);
+ if (var->data.explicit_location &&
+ var->data.location >= VARYING_SLOT_VAR0) {
+ /* If explicit location is given then lookup the variable by location.
+ * We turn the location into a string and use this as the hash key
+ * rather than the name. Note: We allocate enough space for a 32-bit
+ * unsigned location value which is overkill but future proof.
+ */
+ char location_str[11];
+ snprintf(location_str, 11, "%d", var->data.location);
+ _mesa_hash_table_insert(ht, ralloc_strdup(mem_ctx, location_str), var);
+ } else {
+ _mesa_hash_table_insert(ht, var->get_interface_type()->name, var);
+ }
}
private:
/**
* Hash table mapping interface block name to an \c
- * interface_block_definition struct. interface_block_definition structs
- * are allocated using \c mem_ctx.
+ * ir_variable.
*/
hash_table *ht;
};
continue;
}
- const interface_block_definition def(var);
- interface_block_definition *prev_def =
- definitions->lookup(iface_type->name);
-
+ ir_variable *prev_def = definitions->lookup(var);
if (prev_def == NULL) {
/* This is the first time we've seen the interface, so save
* it into the appropriate data structure.
*/
- definitions->store(def);
- } else if (!intrastage_match(prev_def, &def,
- (ir_variable_mode) var->data.mode,
- prog)) {
+ definitions->store(var);
+ } else if (!intrastage_match(prev_def, var, prog)) {
linker_error(prog, "definitions of interface block `%s' do not"
" match\n", iface_type->name);
return;
if (!var || !var->get_interface_type() || var->data.mode != ir_var_shader_in)
continue;
- definitions.store(interface_block_definition(var));
+ definitions.store(var);
}
/* Verify that the producer's output interfaces match. */
if (!var || !var->get_interface_type() || var->data.mode != ir_var_shader_out)
continue;
- interface_block_definition *consumer_def =
- definitions.lookup(var->get_interface_type()->name);
+ ir_variable *consumer_def = definitions.lookup(var);
/* The consumer doesn't use this output block. Ignore it. */
if (consumer_def == NULL)
continue;
- const interface_block_definition producer_def(var);
-
- if (!interstage_match(&producer_def, consumer_def, extra_array_level)) {
+ if (!interstage_match(var, consumer_def, extra_array_level)) {
linker_error(prog, "definitions of interface block `%s' do not "
"match\n", var->get_interface_type()->name);
return;
var->data.mode != ir_var_shader_storage))
continue;
- interface_block_definition *old_def =
- definitions.lookup(var->get_interface_type()->name);
- const interface_block_definition new_def(var);
+ ir_variable *old_def = definitions.lookup(var);
if (old_def == NULL) {
- definitions.store(new_def);
+ definitions.store(var);
} else {
/* Interstage uniform matching rules are the same as intrastage
* uniform matchin rules (for uniforms, it is as though all
* shaders are in the same shader stage).
*/
- if (!intrastage_match(old_def, &new_def,
- (ir_variable_mode) var->data.mode,
- prog)) {
+ if (!intrastage_match(old_def, var, prog)) {
linker_error(prog, "definitions of interface block `%s' do not "
"match\n", var->get_interface_type()->name);
return;
#include "link_uniform_block_active_visitor.h"
#include "program.h"
-link_uniform_block_active *
+static link_uniform_block_active *
process_block(void *mem_ctx, struct hash_table *ht, ir_variable *var)
{
const hash_entry *const existing_block =
* and not over complicating the code we will end up with a count of 8.
* Here each dimension has 2 different indices counted so we end up with 2*2*2
*/
-struct uniform_block_array_elements **
+static struct uniform_block_array_elements **
process_arrays(void *mem_ctx, ir_dereference_array *ir,
struct link_uniform_block_active *block)
{
/* This function resizes the array types of the block so that later we can use
* this new size to correctly calculate the offest for indirect indexing.
*/
-const glsl_type *
+static const glsl_type *
resize_block_array(const glsl_type *type,
struct uniform_block_array_elements *ub_array)
{
case GLSL_TYPE_IMAGE:
case GLSL_TYPE_ATOMIC_UINT:
case GLSL_TYPE_INTERFACE:
+ case GLSL_TYPE_FUNCTION:
case GLSL_TYPE_VOID:
case GLSL_TYPE_SUBROUTINE:
case GLSL_TYPE_ERROR:
*/
if (var->is_interface_instance()) {
ubo_byte_offset = 0;
+ process(var->get_interface_type(),
+ var->get_interface_type()->name);
} else {
const struct gl_uniform_block *const block =
&prog->BufferInterfaceBlocks[ubo_block_index];
&block->Uniforms[var->data.location];
ubo_byte_offset = ubo_var->Offset;
- }
-
- if (var->is_interface_instance())
- process(var->get_interface_type(),
- var->get_interface_type()->name);
- else
process(var);
+ }
} else {
/* Store any explicit location and reset data location so we can
* reuse this variable for storing the uniform slot number.
/**
+ * Get the varying type stripped of the outermost array if we're processing
+ * a stage whose varyings are arrays indexed by a vertex number (such as
+ * geometry shader inputs).
+ */
+static const glsl_type *
+get_varying_type(const ir_variable *var, gl_shader_stage stage)
+{
+ const glsl_type *type = var->type;
+
+ if (!var->data.patch &&
+ ((var->data.mode == ir_var_shader_out &&
+ stage == MESA_SHADER_TESS_CTRL) ||
+ (var->data.mode == ir_var_shader_in &&
+ (stage == MESA_SHADER_TESS_CTRL || stage == MESA_SHADER_TESS_EVAL ||
+ stage == MESA_SHADER_GEOMETRY)))) {
+ assert(type->is_array());
+ type = type->fields.array;
+ }
+
+ return type;
+}
+
+/**
* Validate the types and qualifiers of an output from one stage against the
* matching input to another stage.
*/
}
}
+/**
+ * Demote shader inputs and outputs that are not used in other stages, and
+ * remove them via dead code elimination.
+ */
+void
+remove_unused_shader_inputs_and_outputs(bool is_separate_shader_object,
+ gl_shader *sh,
+ enum ir_variable_mode mode)
+{
+ if (is_separate_shader_object)
+ return;
+
+ foreach_in_list(ir_instruction, node, sh->ir) {
+ ir_variable *const var = node->as_variable();
+
+ if ((var == NULL) || (var->data.mode != int(mode)))
+ continue;
+
+ /* A shader 'in' or 'out' variable is only really an input or output if
+ * its value is used by other shader stages. This will cause the
+ * variable to have a location assigned.
+ */
+ if (var->data.is_unmatched_generic_inout) {
+ assert(var->data.mode != ir_var_temporary);
+ var->data.mode = ir_var_auto;
+ }
+ }
+
+ /* Eliminate code that is now dead due to unused inputs/outputs being
+ * demoted.
+ */
+ while (do_dead_code(sh->ir, false))
+ ;
+
+}
/**
* Initialize this object based on a string that was passed to
this->matched_candidate->type->fields.array->matrix_columns;
const unsigned vector_elements =
this->matched_candidate->type->fields.array->vector_elements;
+ const unsigned dmul =
+ this->matched_candidate->type->fields.array->is_double() ? 2 : 1;
unsigned actual_array_size;
switch (this->lowered_builtin_array_variable) {
case clip_distance:
return false;
}
unsigned array_elem_size = this->lowered_builtin_array_variable ?
- 1 : vector_elements * matrix_cols;
+ 1 : vector_elements * matrix_cols * dmul;
fine_location += array_elem_size * this->array_subscript;
this->size = 1;
} else {
if (!this->is_varying()) {
return 0;
}
-
return (this->num_components() + this->location_frac + 3)/4;
}
gl_shader_stage consumer_stage);
~varying_matches();
void record(ir_variable *producer_var, ir_variable *consumer_var);
- unsigned assign_locations(uint64_t reserved_slots, bool separate_shader);
+ unsigned assign_locations(struct gl_shader_program *prog,
+ uint64_t reserved_slots, bool separate_shader);
void store_locations() const;
private:
this->matches[this->num_matches].packing_order
= this->compute_packing_order(var);
if (this->disable_varying_packing) {
- const struct glsl_type *type = var->type;
unsigned slots;
+ gl_shader_stage stage =
+ (producer_var != NULL) ? producer_stage : consumer_stage;
- /* Some shader stages have 2-dimensional varyings. Use the inner type. */
- if (!var->data.patch &&
- ((var == producer_var && producer_stage == MESA_SHADER_TESS_CTRL) ||
- (var == consumer_var && (consumer_stage == MESA_SHADER_TESS_CTRL ||
- consumer_stage == MESA_SHADER_TESS_EVAL ||
- consumer_stage == MESA_SHADER_GEOMETRY)))) {
- assert(type->is_array());
- type = type->fields.array;
- }
+ const glsl_type *type = get_varying_type(var, stage);
- if (type->is_array()) {
- slots = 1;
- while (type->is_array()) {
- slots *= type->length;
- type = type->fields.array;
- }
- slots *= type->matrix_columns;
- } else {
- slots = type->matrix_columns;
- }
- this->matches[this->num_matches].num_components = 4 * slots;
+ slots = type->count_attribute_slots(false);
+ this->matches[this->num_matches].num_components = slots * 4;
} else {
this->matches[this->num_matches].num_components
= var->type->component_slots();
* passed to varying_matches::record().
*/
unsigned
-varying_matches::assign_locations(uint64_t reserved_slots, bool separate_shader)
+varying_matches::assign_locations(struct gl_shader_program *prog,
+ uint64_t reserved_slots,
+ bool separate_shader)
{
/* We disable varying sorting for separate shader programs for the
* following reasons:
for (unsigned i = 0; i < this->num_matches; i++) {
unsigned *location = &generic_location;
- if ((this->matches[i].consumer_var &&
- this->matches[i].consumer_var->data.patch) ||
- (this->matches[i].producer_var &&
- this->matches[i].producer_var->data.patch))
+ const ir_variable *var;
+ const glsl_type *type;
+ bool is_vertex_input = false;
+ if (matches[i].consumer_var) {
+ var = matches[i].consumer_var;
+ type = get_varying_type(var, consumer_stage);
+ if (consumer_stage == MESA_SHADER_VERTEX)
+ is_vertex_input = true;
+ } else {
+ var = matches[i].producer_var;
+ type = get_varying_type(var, producer_stage);
+ }
+
+ if (var->data.patch)
location = &generic_patch_location;
/* Advance to the next slot if this varying has a different packing
!= this->matches[i].packing_class) {
*location = ALIGN(*location, 4);
}
- while ((*location < MAX_VARYING * 4u) &&
- (reserved_slots & (1u << *location / 4u))) {
- *location = ALIGN(*location + 1, 4);
+
+ unsigned num_elements = type->count_attribute_slots(is_vertex_input);
+ unsigned slot_end = this->disable_varying_packing ? 4 :
+ type->without_array()->vector_elements;
+ slot_end += *location - 1;
+
+ /* FIXME: We could be smarter in the below code and loop back over
+ * trying to fill any locations that we skipped because we couldn't pack
+ * the varying between an explicit location. For now just let the user
+ * hit the linking error if we run out of room and suggest they use
+ * explicit locations.
+ */
+ for (unsigned j = 0; j < num_elements; j++) {
+ while ((slot_end < MAX_VARYING * 4u) &&
+ ((reserved_slots & (UINT64_C(1) << *location / 4u) ||
+ (reserved_slots & (UINT64_C(1) << slot_end / 4u))))) {
+
+ *location = ALIGN(*location + 1, 4);
+ slot_end = *location;
+
+ /* reset the counter and try again */
+ j = 0;
+ }
+
+ /* Increase the slot to make sure there is enough room for next
+ * array element.
+ */
+ if (this->disable_varying_packing)
+ slot_end += 4;
+ else
+ slot_end += type->without_array()->vector_elements;
+ }
+
+ if (!var->data.patch && *location >= MAX_VARYING * 4u) {
+ linker_error(prog, "insufficient contiguous locations available for "
+ "%s it is possible an array or struct could not be "
+ "packed between varyings with explicit locations. Try "
+ "using an explicit location for arrays and structs.",
+ var->name);
}
this->matches[i].generic_location = *location;
void process(ir_variable *var)
{
+ /* All named varying interface blocks should be flattened by now */
+ assert(!var->is_interface_instance());
+
this->toplevel_var = var;
this->varying_floats = 0;
- if (var->is_interface_instance())
- program_resource_visitor::process(var->get_interface_type(),
- var->get_interface_type()->name);
- else
- program_resource_visitor::process(var);
+ program_resource_visitor::process(var);
}
private:
foreach_in_list(ir_instruction, node, stage->ir) {
ir_variable *const var = node->as_variable();
- if (var == NULL || var->data.mode != io_mode || !var->data.explicit_location)
+ if (var == NULL || var->data.mode != io_mode ||
+ !var->data.explicit_location ||
+ var->data.location < VARYING_SLOT_VAR0)
continue;
var_slot = var->data.location - VARYING_SLOT_VAR0;
- if (var_slot >= 0 && var_slot < MAX_VARYING)
- slots |= 1u << var_slot;
+
+ unsigned num_elements = get_varying_type(var, stage->Stage)
+ ->count_attribute_slots(stage->Stage == MESA_SHADER_VERTEX);
+ for (unsigned i = 0; i < num_elements; i++) {
+ if (var_slot >= 0 && var_slot < MAX_VARYING)
+ slots |= UINT64_C(1) << var_slot;
+ var_slot += 1;
+ }
}
return slots;
reserved_varying_slot(producer, ir_var_shader_out) |
reserved_varying_slot(consumer, ir_var_shader_in);
- const unsigned slots_used = matches.assign_locations(reserved_slots,
+ const unsigned slots_used = matches.assign_locations(prog, reserved_slots,
prog->SeparateShader);
matches.store_locations();
hash_table_dtor(consumer_inputs);
hash_table_dtor(consumer_interface_inputs);
- if (!disable_varying_packing) {
- if (producer) {
- lower_packed_varyings(mem_ctx, slots_used, ir_var_shader_out,
- 0, producer);
- }
- if (consumer) {
- lower_packed_varyings(mem_ctx, slots_used, ir_var_shader_in,
- consumer_vertices, consumer);
- }
- }
-
if (consumer && producer) {
foreach_in_list(ir_instruction, node, consumer->ir) {
ir_variable *const var = node->as_variable();
var->name,
_mesa_shader_stage_to_string(producer->Stage));
}
-
- /* An 'in' variable is only really a shader input if its
- * value is written by the previous stage.
- */
- var->data.mode = ir_var_auto;
}
}
+
+ /* Now that validation is done its safe to remove unused varyings. As
+ * we have both a producer and consumer its safe to remove unused
+ * varyings even if the program is a SSO because the stages are being
+ * linked together i.e. we have a multi-stage SSO.
+ */
+ remove_unused_shader_inputs_and_outputs(false, producer,
+ ir_var_shader_out);
+ remove_unused_shader_inputs_and_outputs(false, consumer,
+ ir_var_shader_in);
+ }
+
+ if (!disable_varying_packing) {
+ if (producer) {
+ lower_packed_varyings(mem_ctx, slots_used, ir_var_shader_out,
+ 0, producer);
+ }
+ if (consumer) {
+ lower_packed_varyings(mem_ctx, slots_used, ir_var_shader_in,
+ consumer_vertices, consumer);
+ }
}
return true;
if (var && var->data.mode == ir_var_shader_out &&
var_counts_against_varying_limit(producer->Stage, var)) {
- output_vectors += var->type->count_attribute_slots();
+ /* outputs for fragment shader can't be doubles */
+ output_vectors += var->type->count_attribute_slots(false);
}
}
if (var && var->data.mode == ir_var_shader_in &&
var_counts_against_varying_limit(consumer->Stage, var)) {
- input_vectors += var->type->count_attribute_slots();
+ /* vertex inputs aren't varying counted */
+ input_vectors += var->type->count_attribute_slots(false);
}
}
if (this->lowered_builtin_array_variable)
return this->size;
else
- return this->vector_elements * this->matrix_columns * this->size;
+ return this->vector_elements * this->matrix_columns * this->size *
+ (this->is_double() ? 2 : 1);
}
unsigned get_location() const {
}
private:
+
+ bool is_double() const
+ {
+ switch (this->type) {
+ case GL_DOUBLE:
+ case GL_DOUBLE_VEC2:
+ case GL_DOUBLE_VEC3:
+ case GL_DOUBLE_VEC4:
+ case GL_DOUBLE_MAT2:
+ case GL_DOUBLE_MAT2x3:
+ case GL_DOUBLE_MAT2x4:
+ case GL_DOUBLE_MAT3:
+ case GL_DOUBLE_MAT3x2:
+ case GL_DOUBLE_MAT3x4:
+ case GL_DOUBLE_MAT4:
+ case GL_DOUBLE_MAT4x2:
+ case GL_DOUBLE_MAT4x3:
+ return true;
+ default:
+ return false;
+ }
+ }
+
/**
* The name that was supplied to glTransformFeedbackVaryings. Used for
* error reporting and glGetTransformFeedbackVarying().
const void *mem_ctx, unsigned num_names,
char **varying_names, tfeedback_decl *decls);
+void
+remove_unused_shader_inputs_and_outputs(bool is_separate_shader_object,
+ gl_shader *sh,
+ enum ir_variable_mode mode);
+
bool
store_tfeedback_info(struct gl_context *ctx, struct gl_shader_program *prog,
unsigned num_tfeedback_decls,
}
if (var->type->contains_atomic() &&
- var->data.atomic.offset != existing->data.atomic.offset) {
+ var->data.offset != existing->data.offset) {
linker_error(prog, "offset specifications for %s "
"`%s' have differing values\n",
mode_string(var), var->name);
mode_string(var), var->name);
return;
}
+ if (existing->data.image_format != var->data.image_format) {
+ linker_error(prog, "declarations for %s `%s` have "
+ "mismatching image format qualifiers\n",
+ mode_string(var), var->name);
+ return;
+ }
} else
variables.add_variable(var);
}
return false;
}
- const unsigned slots = var->type->count_attribute_slots();
+ const unsigned slots = var->type->count_attribute_slots(target_index == MESA_SHADER_VERTEX ? true : false);
/* If the variable is not a built-in and has a location statically
* assigned in the shader (presumably via a layout qualifier), make sure
* issue (3) of the GL_ARB_vertex_attrib_64bit behavior, this
* is optional behavior, but it seems preferable.
*/
- const glsl_type *type = var->type->without_array();
- if (type == glsl_type::dvec3_type ||
- type == glsl_type::dvec4_type ||
- type == glsl_type::dmat2x3_type ||
- type == glsl_type::dmat2x4_type ||
- type == glsl_type::dmat3_type ||
- type == glsl_type::dmat3x4_type ||
- type == glsl_type::dmat4x3_type ||
- type == glsl_type::dmat4_type) {
+ if (var->type->without_array()->is_dual_slot_double())
double_storage_locations |= (use_mask << attr);
- }
}
continue;
}
/**
- * Demote shader inputs and outputs that are not used in other stages
- */
-void
-demote_shader_inputs_and_outputs(gl_shader *sh, enum ir_variable_mode mode)
-{
- foreach_in_list(ir_instruction, node, sh->ir) {
- ir_variable *const var = node->as_variable();
-
- if ((var == NULL) || (var->data.mode != int(mode)))
- continue;
-
- /* A shader 'in' or 'out' variable is only really an input or output if
- * its value is used by other shader stages. This will cause the variable
- * to have a location assigned.
- */
- if (var->data.is_unmatched_generic_inout) {
- assert(var->data.mode != ir_var_temporary);
- var->data.mode = ir_var_auto;
- }
- }
-}
-
-
-/**
* Store the gl_FragDepth layout in the gl_shader_program struct.
*/
static void
foreach_in_list(ir_instruction, node, sh->ir) {
ir_variable *var = node->as_variable();
if (var && var->data.mode == ir_var_shader_out)
- fragment_outputs += var->type->count_attribute_slots();
+ /* since there are no double fs outputs - pass false */
+ fragment_outputs += var->type->count_attribute_slots(false);
}
}
}
return;
}
+ unsigned entries_total = 0;
for (unsigned i = 0; i < MESA_SHADER_STAGES; i++) {
struct gl_shader *sh = prog->_LinkedShaders[i];
foreach_in_list(ir_instruction, node, sh->ir) {
ir_variable *var = node->as_variable();
- if (var && (var->data.mode == ir_var_uniform &&
- var->data.explicit_location)) {
+ if (!var || var->data.mode != ir_var_uniform)
+ continue;
+
+ entries_total += var->type->uniform_locations();
+
+ if (var->data.explicit_location) {
bool ret;
if (var->type->is_subroutine())
ret = reserve_subroutine_explicit_locations(prog, sh, var);
}
}
+ /* Verify that total amount of entries for explicit and implicit locations
+ * is less than MAX_UNIFORM_LOCATIONS.
+ */
+ if (entries_total >= ctx->Const.MaxUserAssignableUniformLocations) {
+ linker_error(prog, "count of uniform locations >= MAX_UNIFORM_LOCATIONS"
+ "(%u >= %u)", entries_total,
+ ctx->Const.MaxUserAssignableUniformLocations);
+ }
delete uniform_map;
}
return stages;
}
+/**
+ * Create gl_shader_variable from ir_variable class.
+ */
+static gl_shader_variable *
+create_shader_variable(struct gl_shader_program *shProg, const ir_variable *in)
+{
+ gl_shader_variable *out = ralloc(shProg, struct gl_shader_variable);
+ if (!out)
+ return NULL;
+
+ out->type = in->type;
+ out->name = ralloc_strdup(shProg, in->name);
+
+ if (!out->name)
+ return NULL;
+
+ out->location = in->data.location;
+ out->index = in->data.index;
+ out->patch = in->data.patch;
+ out->mode = in->data.mode;
+
+ return out;
+}
+
static bool
add_interface_variables(struct gl_shader_program *shProg,
exec_list *ir, GLenum programInterface)
if (strncmp(var->name, "gl_out_FragData", 15) == 0)
continue;
- if (!add_program_resource(shProg, programInterface, var,
- build_stageref(shProg, var->name,
- var->data.mode) | mask))
+ gl_shader_variable *sha_v = create_shader_variable(shProg, var);
+ if (!sha_v)
+ return false;
+
+ if (!add_program_resource(shProg, programInterface, sha_v,
+ build_stageref(shProg, sha_v->name,
+ sha_v->mode) | mask))
return false;
}
return true;
}
static bool
-add_packed_varyings(struct gl_shader_program *shProg, int stage)
+add_packed_varyings(struct gl_shader_program *shProg, int stage, GLenum type)
{
struct gl_shader *sh = shProg->_LinkedShaders[stage];
GLenum iface;
default:
unreachable("unexpected type");
}
- if (!add_program_resource(shProg, iface, var,
- build_stageref(shProg, var->name,
- var->data.mode)))
- return false;
+
+ if (type == iface) {
+ gl_shader_variable *sha_v = create_shader_variable(shProg, var);
+ if (!sha_v)
+ return false;
+ if (!add_program_resource(shProg, iface, sha_v,
+ build_stageref(shProg, sha_v->name,
+ sha_v->mode)))
+ return false;
+ }
}
}
return true;
ir_variable *var = node->as_variable();
if (var) {
assert(var->data.mode == ir_var_shader_out);
- if (!add_program_resource(shProg, GL_PROGRAM_OUTPUT, var,
+ gl_shader_variable *sha_v = create_shader_variable(shProg, var);
+ if (!sha_v)
+ return false;
+ if (!add_program_resource(shProg, GL_PROGRAM_OUTPUT, sha_v,
1 << MESA_SHADER_FRAGMENT))
return false;
}
/* Program interface needs to expose varyings in case of SSO. */
if (shProg->SeparateShader) {
- if (!add_packed_varyings(shProg, input_stage))
+ if (!add_packed_varyings(shProg, input_stage, GL_PROGRAM_INPUT))
return;
- if (!add_packed_varyings(shProg, output_stage))
+
+ if (!add_packed_varyings(shProg, output_stage, GL_PROGRAM_OUTPUT))
return;
}
unsigned num_blocks,
struct gl_uniform_block ***ubos,
unsigned *num_ubos,
+ unsigned **ubo_interface_block_indices,
struct gl_uniform_block ***ssbos,
- unsigned *num_ssbos)
+ unsigned *num_ssbos,
+ unsigned **ssbo_interface_block_indices)
{
unsigned num_ubo_blocks = 0;
unsigned num_ssbo_blocks = 0;
*ssbos = ralloc_array(mem_ctx, gl_uniform_block *, num_ssbo_blocks);
*num_ssbos = 0;
+ if (ubo_interface_block_indices)
+ *ubo_interface_block_indices =
+ ralloc_array(mem_ctx, unsigned, num_ubo_blocks);
+
+ if (ssbo_interface_block_indices)
+ *ssbo_interface_block_indices =
+ ralloc_array(mem_ctx, unsigned, num_ssbo_blocks);
+
for (unsigned i = 0; i < num_blocks; i++) {
if (blocks[i].IsShaderStorage) {
- (*ssbos)[(*num_ssbos)++] = &blocks[i];
+ (*ssbos)[*num_ssbos] = &blocks[i];
+ if (ssbo_interface_block_indices)
+ (*ssbo_interface_block_indices)[*num_ssbos] = i;
+ (*num_ssbos)++;
} else {
- (*ubos)[(*num_ubos)++] = &blocks[i];
+ (*ubos)[*num_ubos] = &blocks[i];
+ if (ubo_interface_block_indices)
+ (*ubo_interface_block_indices)[*num_ubos] = i;
+ (*num_ubos)++;
}
}
do_dead_builtin_varyings(ctx, sh, NULL,
num_tfeedback_decls, tfeedback_decls);
- if (!prog->SeparateShader) {
- demote_shader_inputs_and_outputs(sh, ir_var_shader_out);
- /* Eliminate code that is now dead due to unused outputs being
- * demoted.
- */
- while (do_dead_code(sh->ir, false))
- ;
- }
+ remove_unused_shader_inputs_and_outputs(prog->SeparateShader, sh,
+ ir_var_shader_out);
}
else if (first == MESA_SHADER_FRAGMENT) {
/* If the program only contains a fragment shader...
NULL /* tfeedback_decls */))
goto done;
} else {
- demote_shader_inputs_and_outputs(sh, ir_var_shader_in);
- /* Eliminate code that is now dead due to unused inputs being
- * demoted.
- */
- while (do_dead_code(sh->ir, false))
- ;
+ remove_unused_shader_inputs_and_outputs(false, sh,
+ ir_var_shader_in);
}
}
next == MESA_SHADER_FRAGMENT ? num_tfeedback_decls : 0,
tfeedback_decls);
- demote_shader_inputs_and_outputs(sh_i, ir_var_shader_out);
- demote_shader_inputs_and_outputs(sh_next, ir_var_shader_in);
-
- /* Eliminate code that is now dead due to unused outputs being demoted.
- */
- while (do_dead_code(sh_i->ir, false))
- ;
- while (do_dead_code(sh_next->ir, false))
- ;
-
/* This must be done after all dead varyings are eliminated. */
if (!check_against_output_limit(ctx, prog, sh_i))
goto done;
sh->NumBufferInterfaceBlocks,
&sh->UniformBlocks,
&sh->NumUniformBlocks,
+ NULL,
&sh->ShaderStorageBlocks,
- &sh->NumShaderStorageBlocks);
+ &sh->NumShaderStorageBlocks,
+ NULL);
}
}
prog->NumBufferInterfaceBlocks,
&prog->UniformBlocks,
&prog->NumUniformBlocks,
+ &prog->UboInterfaceBlockIndex,
&prog->ShaderStorageBlocks,
- &prog->NumShaderStorageBlocks);
+ &prog->NumShaderStorageBlocks,
+ &prog->SsboInterfaceBlockIndex);
/* FINISHME: Assign fragment shader output locations. */
* - MOD_TO_FLOOR
* - LDEXP_TO_ARITH
* - DFREXP_TO_ARITH
- * - BITFIELD_INSERT_TO_BFM_BFI
* - CARRY_TO_ARITH
* - BORROW_TO_ARITH
* - SAT_TO_CLAMP
* Converts ir_binop_ldexp, ir_unop_frexp_sig, and ir_unop_frexp_exp to
* arithmetic and bit ops for double arguments.
*
- * BITFIELD_INSERT_TO_BFM_BFI:
- * ---------------------------
- * Breaks ir_quadop_bitfield_insert into ir_binop_bfm (bitfield mask) and
- * ir_triop_bfi (bitfield insert).
- *
- * Many GPUs implement the bitfieldInsert() built-in from ARB_gpu_shader_5
- * with a pair of instructions.
- *
* CARRY_TO_ARITH:
* ---------------
* Converts ir_carry into (x + y) < x.
void exp_to_exp2(ir_expression *);
void pow_to_exp2(ir_expression *);
void log_to_log2(ir_expression *);
- void bitfield_insert_to_bfm_bfi(ir_expression *);
void ldexp_to_arith(ir_expression *);
void dldexp_to_arith(ir_expression *);
void dfrexp_sig_to_arith(ir_expression *);
}
void
-lower_instructions_visitor::bitfield_insert_to_bfm_bfi(ir_expression *ir)
-{
- /* Translates
- * ir_quadop_bitfield_insert base insert offset bits
- * into
- * ir_triop_bfi (ir_binop_bfm bits offset) insert base
- */
-
- ir_rvalue *base_expr = ir->operands[0];
-
- ir->operation = ir_triop_bfi;
- ir->operands[0] = new(ir) ir_expression(ir_binop_bfm,
- ir->type->get_base_type(),
- ir->operands[3],
- ir->operands[2]);
- /* ir->operands[1] is still the value to insert. */
- ir->operands[2] = base_expr;
- ir->operands[3] = NULL;
-
- this->progress = true;
-}
-
-void
lower_instructions_visitor::ldexp_to_arith(ir_expression *ir)
{
/* Translates
ir_constant *sign_mask = new(ir) ir_constant(0x80000000u, vec_elem);
- ir_constant *exp_shift = new(ir) ir_constant(23);
- ir_constant *exp_width = new(ir) ir_constant(8);
+ ir_constant *exp_shift = new(ir) ir_constant(23, vec_elem);
+ ir_constant *exp_width = new(ir) ir_constant(8, vec_elem);
/* Temporary variables */
ir_variable *x = new(ir) ir_variable(ir->type, "x", ir_var_temporary);
exp_shift_clone, exp_width);
ir->operands[1] = NULL;
- /* Don't generate new IR that would need to be lowered in an additional
- * pass.
- */
- if (lowering(BITFIELD_INSERT_TO_BFM_BFI))
- bitfield_insert_to_bfm_bfi(ir->operands[0]->as_expression());
-
this->progress = true;
}
ir_constant *sign_mask = new(ir) ir_constant(0x80000000u);
- ir_constant *exp_shift = new(ir) ir_constant(20);
- ir_constant *exp_width = new(ir) ir_constant(11);
+ ir_constant *exp_shift = new(ir) ir_constant(20, vec_elem);
+ ir_constant *exp_width = new(ir) ir_constant(11, vec_elem);
ir_constant *exp_bias = new(ir) ir_constant(1022, vec_elem);
/* Temporary variables */
exp_shift->clone(ir, NULL),
exp_width->clone(ir, NULL));
- if (lowering(BITFIELD_INSERT_TO_BFM_BFI))
- bitfield_insert_to_bfm_bfi(bfi);
-
i.insert_before(assign(unpacked, bfi, WRITEMASK_Y));
results[elem] = expr(ir_unop_pack_double_2x32, unpacked);
pow_to_exp2(ir);
break;
- case ir_quadop_bitfield_insert:
- if (lowering(BITFIELD_INSERT_TO_BFM_BFI))
- bitfield_insert_to_bfm_bfi(ir);
- break;
-
case ir_binop_ldexp:
if (lowering(LDEXP_TO_ARITH) && ir->type->is_float())
ldexp_to_arith(ir);
}
ir_rvalue *const val = new(this->mem_ctx) ir_dereference_variable(tmp_bvec);
- ir_expression *any = new(this->mem_ctx) ir_expression(ir_unop_any, val);
+ uint8_t vec_elems = val->type->vector_elements;
+ ir_expression *any =
+ new(this->mem_ctx) ir_expression(ir_binop_any_nequal, val,
+ new(this->mem_ctx) ir_constant(false,
+ vec_elems));
if (test_equal)
any = new(this->mem_ctx) ir_expression(ir_unop_logic_not, any);
if (op_mask & LOWER_PACK_USE_BFI) {
return bitfield_insert(bit_and(swizzle_x(u), constant(0xffffu)),
swizzle_y(u),
- constant(16),
- constant(16));
+ constant(16u),
+ constant(16u));
}
/* return (u.y << 16) | (u.x & 0xffff); */
return bitfield_insert(bitfield_insert(
bitfield_insert(
bit_and(swizzle_x(u), constant(0xffu)),
- swizzle_y(u), constant(8), constant(8)),
- swizzle_z(u), constant(16), constant(8)),
- swizzle_w(u), constant(24), constant(8));
+ swizzle_y(u), constant(8u), constant(8u)),
+ swizzle_z(u), constant(16u), constant(8u)),
+ swizzle_w(u), constant(24u), constant(8u));
}
/* uvec4 u = UVEC4_RVAL & 0xff */
if (op_mask & LOWER_PACK_USE_BFE) {
/* u4.y = bitfield_extract(u, 8, 8); */
- factory.emit(assign(u4, bitfield_extract(u, constant(8), constant(8)),
+ factory.emit(assign(u4, bitfield_extract(u, constant(8u), constant(8u)),
WRITEMASK_Y));
/* u4.z = bitfield_extract(u, 16, 8); */
- factory.emit(assign(u4, bitfield_extract(u, constant(16), constant(8)),
+ factory.emit(assign(u4, bitfield_extract(u, constant(16u), constant(8u)),
WRITEMASK_Z));
} else {
/* u4.y = (u >> 8u) & 0xffu; */
class nir_visitor : public ir_visitor
{
public:
- nir_visitor(nir_shader *shader);
+ nir_visitor(nir_shader *shader, gl_shader *sh);
~nir_visitor();
virtual void visit(ir_variable *);
virtual void visit(ir_dereference_array *);
virtual void visit(ir_barrier *);
- void create_function(ir_function *ir);
+ void create_function(ir_function_signature *ir);
private:
- void create_overload(ir_function_signature *ir, nir_function *function);
void add_instr(nir_instr *instr, unsigned num_components);
nir_ssa_def *evaluate_rvalue(ir_rvalue *ir);
bool supports_ints;
+ struct gl_shader *sh;
+
nir_shader *shader;
nir_function_impl *impl;
nir_builder b;
nir_shader *shader = nir_shader_create(NULL, stage, options);
- nir_visitor v1(shader);
+ nir_visitor v1(shader, sh);
nir_function_visitor v2(&v1);
v2.run(sh->ir);
visit_exec_list(sh->ir, &v1);
- nir_lower_outputs_to_temporaries(shader);
+ nir_function *main = NULL;
+ nir_foreach_function(shader, func) {
+ if (strcmp(func->name, "main") == 0) {
+ main = func;
+ break;
+ }
+ }
+ assert(main);
+
+ nir_lower_outputs_to_temporaries(shader, main);
shader->info.name = ralloc_asprintf(shader, "GLSL%d", shader_prog->Name);
if (shader_prog->Label)
return shader;
}
-nir_visitor::nir_visitor(nir_shader *shader)
+nir_visitor::nir_visitor(nir_shader *shader, gl_shader *sh)
{
this->supports_ints = shader->options->native_integers;
this->shader = shader;
+ this->sh = sh;
this->is_global = true;
this->var_table = _mesa_hash_table_create(NULL, _mesa_hash_pointer,
_mesa_key_pointer_equal);
var->data.explicit_index = ir->data.explicit_index;
var->data.explicit_binding = ir->data.explicit_binding;
var->data.has_initializer = ir->data.has_initializer;
- var->data.is_unmatched_generic_inout = ir->data.is_unmatched_generic_inout;
var->data.location_frac = ir->data.location_frac;
var->data.from_named_ifc_block_array = ir->data.from_named_ifc_block_array;
var->data.from_named_ifc_block_nonarray = ir->data.from_named_ifc_block_nonarray;
}
var->data.index = ir->data.index;
+ var->data.descriptor_set = 0;
var->data.binding = ir->data.binding;
- var->data.atomic.offset = ir->data.atomic.offset;
+ var->data.offset = ir->data.offset;
var->data.image.read_only = ir->data.image_read_only;
var->data.image.write_only = ir->data.image_write_only;
var->data.image.coherent = ir->data.image_coherent;
ir_visitor_status
nir_function_visitor::visit_enter(ir_function *ir)
{
- visitor->create_function(ir);
- return visit_continue_with_parent;
-}
-
-
-void
-nir_visitor::create_function(ir_function *ir)
-{
- nir_function *func = nir_function_create(this->shader, ir->name);
foreach_in_list(ir_function_signature, sig, &ir->signatures) {
- create_overload(sig, func);
+ visitor->create_function(sig);
}
+ return visit_continue_with_parent;
}
-
-
void
-nir_visitor::create_overload(ir_function_signature *ir, nir_function *function)
+nir_visitor::create_function(ir_function_signature *ir)
{
if (ir->is_intrinsic)
return;
- nir_function_overload *overload = nir_function_overload_create(function);
+ nir_function *func = nir_function_create(shader, ir->function_name());
unsigned num_params = ir->parameters.length();
- overload->num_params = num_params;
- overload->params = ralloc_array(shader, nir_parameter, num_params);
+ func->num_params = num_params;
+ func->params = ralloc_array(shader, nir_parameter, num_params);
unsigned i = 0;
foreach_in_list(ir_variable, param, &ir->parameters) {
switch (param->data.mode) {
case ir_var_function_in:
- overload->params[i].param_type = nir_parameter_in;
+ func->params[i].param_type = nir_parameter_in;
break;
case ir_var_function_out:
- overload->params[i].param_type = nir_parameter_out;
+ func->params[i].param_type = nir_parameter_out;
break;
case ir_var_function_inout:
- overload->params[i].param_type = nir_parameter_inout;
+ func->params[i].param_type = nir_parameter_inout;
break;
default:
unreachable("not reached");
}
- overload->params[i].type = param->type;
+ func->params[i].type = param->type;
i++;
}
- overload->return_type = ir->return_type;
+ func->return_type = ir->return_type;
- _mesa_hash_table_insert(this->overload_table, ir, overload);
+ _mesa_hash_table_insert(this->overload_table, ir, func);
}
void
_mesa_hash_table_search(this->overload_table, ir);
assert(entry);
- nir_function_overload *overload = (nir_function_overload *) entry->data;
+ nir_function *func = (nir_function *) entry->data;
if (ir->is_defined) {
- nir_function_impl *impl = nir_function_impl_create(overload);
+ nir_function_impl *impl = nir_function_impl_create(func);
this->impl = impl;
- unsigned num_params = overload->num_params;
+ unsigned num_params = func->num_params;
impl->num_params = num_params;
impl->params = ralloc_array(this->shader, nir_variable *, num_params);
unsigned i = 0;
i++;
}
- if (overload->return_type == glsl_type::void_type) {
+ if (func->return_type == glsl_type::void_type) {
impl->return_var = NULL;
} else {
impl->return_var = ralloc(this->shader, nir_variable);
impl->return_var->name = ralloc_strdup(impl->return_var,
"return_var");
- impl->return_var->type = overload->return_type;
+ impl->return_var->type = func->return_type;
}
this->is_global = false;
this->is_global = true;
} else {
- overload->impl = NULL;
+ func->impl = NULL;
}
}
nir_intrinsic_instr *store_instr =
nir_intrinsic_instr_create(shader, nir_intrinsic_store_var);
store_instr->num_components = ir->return_deref->type->vector_elements;
+ store_instr->const_index[0] = (1 << store_instr->num_components) - 1;
store_instr->variables[0] =
evaluate_deref(&store_instr->instr, ir->return_deref);
struct hash_entry *entry =
_mesa_hash_table_search(this->overload_table, ir->callee);
assert(entry);
- nir_function_overload *callee = (nir_function_overload *) entry->data;
+ nir_function *callee = (nir_function *) entry->data;
nir_call_instr *instr = nir_call_instr_create(this->shader, callee);
nir_ssa_def *src = evaluate_rvalue(ir->rhs);
if (ir->write_mask != (1 << num_components) - 1 && ir->write_mask != 0) {
- /*
- * We have no good way to update only part of a variable, so just load
- * the LHS and do a vec operation to combine the old with the new, and
- * then store it
- * back into the LHS. Copy propagation should get rid of the mess.
+ /* GLSL IR will give us the input to the write-masked assignment in a
+ * single packed vector. So, for example, if the writemask is xzw, then
+ * we have to swizzle x -> x, y -> z, and z -> w and get the y component
+ * from the load.
*/
-
- nir_intrinsic_instr *load =
- nir_intrinsic_instr_create(this->shader, nir_intrinsic_load_var);
- load->num_components = ir->lhs->type->vector_elements;
- nir_ssa_dest_init(&load->instr, &load->dest, num_components, NULL);
- load->variables[0] = lhs_deref;
- ralloc_steal(load, load->variables[0]);
- nir_builder_instr_insert(&b, &load->instr);
-
- nir_ssa_def *srcs[4];
-
+ unsigned swiz[4];
unsigned component = 0;
- for (unsigned i = 0; i < ir->lhs->type->vector_elements; i++) {
- if (ir->write_mask & (1 << i)) {
- /* GLSL IR will give us the input to the write-masked assignment
- * in a single packed vector. So, for example, if the
- * writemask is xzw, then we have to swizzle x -> x, y -> z,
- * and z -> w and get the y component from the load.
- */
- srcs[i] = nir_channel(&b, src, component++);
- } else {
- srcs[i] = nir_channel(&b, &load->dest.ssa, i);
- }
+ for (unsigned i = 0; i < 4; i++) {
+ swiz[i] = ir->write_mask & (1 << i) ? component++ : 0;
}
-
- src = nir_vec(&b, srcs, ir->lhs->type->vector_elements);
+ src = nir_swizzle(&b, src, swiz, num_components, !supports_ints);
}
nir_intrinsic_instr *store =
nir_intrinsic_instr_create(this->shader, nir_intrinsic_store_var);
store->num_components = ir->lhs->type->vector_elements;
+ store->const_index[0] = ir->write_mask;
nir_deref *store_deref = nir_copy_deref(store, &lhs_deref->deref);
store->variables[0] = nir_deref_as_var(store_deref);
store->src[0] = nir_src_for_ssa(src);
/* no-op */
result = nir_imov(&b, srcs[0]);
break;
- case ir_unop_any:
- switch (ir->operands[0]->type->vector_elements) {
- case 2:
- result = supports_ints ? nir_bany2(&b, srcs[0])
- : nir_fany2(&b, srcs[0]);
- break;
- case 3:
- result = supports_ints ? nir_bany3(&b, srcs[0])
- : nir_fany3(&b, srcs[0]);
- break;
- case 4:
- result = supports_ints ? nir_bany4(&b, srcs[0])
- : nir_fany4(&b, srcs[0]);
- break;
- default:
- unreachable("not reached");
- }
- break;
case ir_unop_trunc: result = nir_ftrunc(&b, srcs[0]); break;
case ir_unop_ceil: result = nir_fceil(&b, srcs[0]); break;
case ir_unop_floor: result = nir_ffloor(&b, srcs[0]); break;
case ir_binop_pack_half_2x16_split:
result = nir_pack_half_2x16_split(&b, srcs[0], srcs[1]);
break;
- case ir_binop_bfm: result = nir_bfm(&b, srcs[0], srcs[1]); break;
case ir_binop_ldexp: result = nir_ldexp(&b, srcs[0], srcs[1]); break;
case ir_triop_fma:
result = nir_ffma(&b, srcs[0], srcs[1], srcs[2]);
else
result = nir_fcsel(&b, srcs[0], srcs[1], srcs[2]);
break;
- case ir_triop_bfi:
- result = nir_bfi(&b, srcs[0], srcs[1], srcs[2]);
- break;
case ir_triop_bitfield_extract:
result = (out_type == GLSL_TYPE_INT) ?
nir_ibitfield_extract(&b, srcs[0], srcs[1], srcs[2]) :
hash_table *glsl_type::array_types = NULL;
hash_table *glsl_type::record_types = NULL;
hash_table *glsl_type::interface_types = NULL;
+hash_table *glsl_type::function_types = NULL;
hash_table *glsl_type::subroutine_types = NULL;
void *glsl_type::mem_ctx = NULL;
mtx_unlock(&glsl_type::mutex);
}
+glsl_type::glsl_type(const glsl_type *return_type,
+ const glsl_function_param *params, unsigned num_params) :
+ gl_type(0),
+ base_type(GLSL_TYPE_FUNCTION),
+ sampler_dimensionality(0), sampler_shadow(0), sampler_array(0),
+ sampler_type(0), interface_packing(0),
+ vector_elements(0), matrix_columns(0),
+ length(num_params)
+{
+ unsigned int i;
+
+ mtx_lock(&glsl_type::mutex);
+
+ init_ralloc_type_ctx();
+
+ this->fields.parameters = rzalloc_array(this->mem_ctx,
+ glsl_function_param, num_params + 1);
+
+ /* We store the return type as the first parameter */
+ this->fields.parameters[0].type = return_type;
+ this->fields.parameters[0].in = false;
+ this->fields.parameters[0].out = true;
+
+ /* We store the i'th parameter in slot i+1 */
+ for (i = 0; i < length; i++) {
+ this->fields.parameters[i + 1].type = params[i].type;
+ this->fields.parameters[i + 1].in = params[i].in;
+ this->fields.parameters[i + 1].out = params[i].out;
+ }
+
+ mtx_unlock(&glsl_type::mutex);
+}
+
glsl_type::glsl_type(const char *subroutine_name) :
gl_type(0),
base_type(GLSL_TYPE_SUBROUTINE),
}
const glsl_type *
+glsl_type::get_image_instance(enum glsl_sampler_dim dim,
+ bool array, glsl_base_type type)
+{
+ switch (type) {
+ case GLSL_TYPE_FLOAT:
+ switch (dim) {
+ case GLSL_SAMPLER_DIM_1D:
+ return (array ? image1DArray_type : image1D_type);
+ case GLSL_SAMPLER_DIM_2D:
+ return (array ? image2DArray_type : image2D_type);
+ case GLSL_SAMPLER_DIM_3D:
+ return image3D_type;
+ case GLSL_SAMPLER_DIM_CUBE:
+ return (array ? imageCubeArray_type : imageCube_type);
+ case GLSL_SAMPLER_DIM_RECT:
+ if (array)
+ return error_type;
+ else
+ return image2DRect_type;
+ case GLSL_SAMPLER_DIM_BUF:
+ if (array)
+ return error_type;
+ else
+ return imageBuffer_type;
+ case GLSL_SAMPLER_DIM_MS:
+ return (array ? image2DMSArray_type : image2DMS_type);
+ case GLSL_SAMPLER_DIM_EXTERNAL:
+ return error_type;
+ }
+ case GLSL_TYPE_INT:
+ switch (dim) {
+ case GLSL_SAMPLER_DIM_1D:
+ return (array ? iimage1DArray_type : iimage1D_type);
+ case GLSL_SAMPLER_DIM_2D:
+ return (array ? iimage2DArray_type : iimage2D_type);
+ case GLSL_SAMPLER_DIM_3D:
+ if (array)
+ return error_type;
+ return iimage3D_type;
+ case GLSL_SAMPLER_DIM_CUBE:
+ return (array ? iimageCubeArray_type : iimageCube_type);
+ case GLSL_SAMPLER_DIM_RECT:
+ if (array)
+ return error_type;
+ return iimage2DRect_type;
+ case GLSL_SAMPLER_DIM_BUF:
+ if (array)
+ return error_type;
+ return iimageBuffer_type;
+ case GLSL_SAMPLER_DIM_MS:
+ return (array ? iimage2DMSArray_type : iimage2DMS_type);
+ case GLSL_SAMPLER_DIM_EXTERNAL:
+ return error_type;
+ }
+ case GLSL_TYPE_UINT:
+ switch (dim) {
+ case GLSL_SAMPLER_DIM_1D:
+ return (array ? uimage1DArray_type : uimage1D_type);
+ case GLSL_SAMPLER_DIM_2D:
+ return (array ? uimage2DArray_type : uimage2D_type);
+ case GLSL_SAMPLER_DIM_3D:
+ if (array)
+ return error_type;
+ return uimage3D_type;
+ case GLSL_SAMPLER_DIM_CUBE:
+ return (array ? uimageCubeArray_type : uimageCube_type);
+ case GLSL_SAMPLER_DIM_RECT:
+ if (array)
+ return error_type;
+ return uimage2DRect_type;
+ case GLSL_SAMPLER_DIM_BUF:
+ if (array)
+ return error_type;
+ return uimageBuffer_type;
+ case GLSL_SAMPLER_DIM_MS:
+ return (array ? uimage2DMSArray_type : uimage2DMS_type);
+ case GLSL_SAMPLER_DIM_EXTERNAL:
+ return error_type;
+ }
+ default:
+ return error_type;
+ }
+
+ unreachable("switch statement above should be complete");
+}
+
+const glsl_type *
glsl_type::get_array_instance(const glsl_type *base, unsigned array_size)
{
/* Generate a name using the base type pointer in the key. This is
}
+static bool
+function_key_compare(const void *a, const void *b)
+{
+ const glsl_type *const key1 = (glsl_type *) a;
+ const glsl_type *const key2 = (glsl_type *) b;
+
+ if (key1->length != key2->length)
+ return 1;
+
+ return memcmp(key1->fields.parameters, key2->fields.parameters,
+ (key1->length + 1) * sizeof(*key1->fields.parameters)) == 0;
+}
+
+
+static uint32_t
+function_key_hash(const void *a)
+{
+ const glsl_type *const key = (glsl_type *) a;
+ char hash_key[128];
+ unsigned size = 0;
+
+ size = snprintf(hash_key, sizeof(hash_key), "%08x", key->length);
+
+ for (unsigned i = 0; i < key->length; i++) {
+ if (size >= sizeof(hash_key))
+ break;
+
+ size += snprintf(& hash_key[size], sizeof(hash_key) - size,
+ "%p", (void *) key->fields.structure[i].type);
+ }
+
+ return _mesa_hash_string(hash_key);
+}
+
+const glsl_type *
+glsl_type::get_function_instance(const glsl_type *return_type,
+ const glsl_function_param *params,
+ unsigned num_params)
+{
+ const glsl_type key(return_type, params, num_params);
+
+ mtx_lock(&glsl_type::mutex);
+
+ if (function_types == NULL) {
+ function_types = _mesa_hash_table_create(NULL, function_key_hash,
+ function_key_compare);
+ }
+
+ struct hash_entry *entry = _mesa_hash_table_search(function_types, &key);
+ if (entry == NULL) {
+ mtx_unlock(&glsl_type::mutex);
+ const glsl_type *t = new glsl_type(return_type, params, num_params);
+ mtx_lock(&glsl_type::mutex);
+
+ entry = _mesa_hash_table_insert(function_types, t, (void *) t);
+ }
+
+ const glsl_type *t = (const glsl_type *)entry->data;
+
+ assert(t->base_type == GLSL_TYPE_FUNCTION);
+ assert(t->length == num_params);
+
+ mtx_unlock(&glsl_type::mutex);
+
+ return t;
+}
+
+
const glsl_type *
glsl_type::get_mul_type(const glsl_type *type_a, const glsl_type *type_b)
{
return 1;
case GLSL_TYPE_SUBROUTINE:
return 1;
+
+ case GLSL_TYPE_FUNCTION:
case GLSL_TYPE_SAMPLER:
case GLSL_TYPE_ATOMIC_UINT:
case GLSL_TYPE_VOID:
}
unsigned
-glsl_type::count_attribute_slots() const
+glsl_type::count_attribute_slots(bool vertex_input_slots) const
{
/* From page 31 (page 37 of the PDF) of the GLSL 1.50 spec:
*
* allows varying structs, the number of varying slots taken up by a
* varying struct is simply equal to the sum of the number of slots taken
* up by each element.
+ *
+ * Doubles are counted different depending on whether they are vertex
+ * inputs or everything else. Vertex inputs from ARB_vertex_attrib_64bit
+ * take one location no matter what size they are, otherwise dvec3/4
+ * take two locations.
*/
switch (this->base_type) {
case GLSL_TYPE_UINT:
case GLSL_TYPE_INT:
case GLSL_TYPE_FLOAT:
case GLSL_TYPE_BOOL:
- case GLSL_TYPE_DOUBLE:
return this->matrix_columns;
-
+ case GLSL_TYPE_DOUBLE:
+ if (this->vector_elements > 2 && !vertex_input_slots)
+ return this->matrix_columns * 2;
+ else
+ return this->matrix_columns;
case GLSL_TYPE_STRUCT:
case GLSL_TYPE_INTERFACE: {
unsigned size = 0;
for (unsigned i = 0; i < this->length; i++)
- size += this->fields.structure[i].type->count_attribute_slots();
+ size += this->fields.structure[i].type->count_attribute_slots(vertex_input_slots);
return size;
}
case GLSL_TYPE_ARRAY:
- return this->length * this->fields.array->count_attribute_slots();
+ return this->length * this->fields.array->count_attribute_slots(vertex_input_slots);
+ case GLSL_TYPE_FUNCTION:
case GLSL_TYPE_SAMPLER:
case GLSL_TYPE_IMAGE:
case GLSL_TYPE_ATOMIC_UINT:
GLSL_TYPE_IMAGE,
GLSL_TYPE_ATOMIC_UINT,
GLSL_TYPE_STRUCT,
+ GLSL_TYPE_FUNCTION,
GLSL_TYPE_INTERFACE,
GLSL_TYPE_ARRAY,
GLSL_TYPE_VOID,
*/
union {
const struct glsl_type *array; /**< Type of array elements. */
- const struct glsl_type *parameters; /**< Parameters to function. */
+ struct glsl_function_param *parameters; /**< Parameters to function. */
struct glsl_struct_field *structure; /**< List of struct fields. */
} fields;
bool array,
glsl_base_type type);
+ static const glsl_type *get_image_instance(enum glsl_sampler_dim dim,
+ bool array, glsl_base_type type);
/**
* Get the instance of an array type
static const glsl_type *get_subroutine_instance(const char *subroutine_name);
/**
+ * Get the instance of a function type
+ */
+ static const glsl_type *get_function_instance(const struct glsl_type *return_type,
+ const glsl_function_param *parameters,
+ unsigned num_params);
+
+ /**
* Get the type resulting from a multiplication of \p type_a * \p type_b
*/
static const glsl_type *get_mul_type(const glsl_type *type_a,
* varying slots the type will use up in the absence of varying packing
* (and thus, it can be used to measure the number of varying slots used by
* the varyings that are generated by lower_packed_varyings).
+ *
+ * For vertex shader attributes - doubles only take one slot.
+ * For inter-shader varyings - dvec3/dvec4 take two slots.
*/
- unsigned count_attribute_slots() const;
+ unsigned count_attribute_slots(bool vertex_input_slots) const;
/**
* Alignment in bytes of the start of this type in a std140 uniform
}
/**
+ * Query whether a double takes two slots.
+ */
+ bool is_dual_slot_double() const
+ {
+ return base_type == GLSL_TYPE_DOUBLE && vector_elements > 2;
+ }
+
+ /**
* Query whether or not a type is a non-array boolean type
*/
bool is_boolean() const
glsl_type(const glsl_struct_field *fields, unsigned num_fields,
enum glsl_interface_packing packing, const char *name);
+ /** Constructor for interface types */
+ glsl_type(const glsl_type *return_type,
+ const glsl_function_param *params, unsigned num_params);
+
/** Constructor for array types */
glsl_type(const glsl_type *array, unsigned length);
/** Hash table containing the known subroutine types. */
static struct hash_table *subroutine_types;
+ /** Hash table containing the known function types. */
+ static struct hash_table *function_types;
+
static bool record_key_compare(const void *a, const void *b);
static unsigned record_key_hash(const void *key);
/*@}*/
};
+#undef DECL_TYPE
+#undef STRUCT_TYPE
+#endif /* __cplusplus */
+
struct glsl_struct_field {
const struct glsl_type *type;
const char *name;
unsigned image_volatile:1;
unsigned image_restrict:1;
+#ifdef __cplusplus
glsl_struct_field(const struct glsl_type *_type, const char *_name)
: type(_type), name(_name), location(-1), interpolation(0), centroid(0),
sample(0), matrix_layout(GLSL_MATRIX_LAYOUT_INHERITED), patch(0),
{
/* empty */
}
+#endif
+};
+
+struct glsl_function_param {
+ const struct glsl_type *type;
+
+ bool in;
+ bool out;
};
static inline unsigned int
return (a + align - 1) / align * align;
}
-#undef DECL_TYPE
-#undef STRUCT_TYPE
-#endif /* __cplusplus */
-
#endif /* GLSL_TYPES_H */
exec_list_make_empty(&shader->uniforms);
exec_list_make_empty(&shader->inputs);
exec_list_make_empty(&shader->outputs);
+ exec_list_make_empty(&shader->shared);
shader->options = options;
memset(&shader->info, 0, sizeof(shader->info));
shader->num_inputs = 0;
shader->num_outputs = 0;
shader->num_uniforms = 0;
+ shader->num_shared = 0;
shader->stage = stage;
exec_list_push_tail(&shader->uniforms, &var->node);
break;
+ case nir_var_shared:
+ assert(shader->stage == MESA_SHADER_COMPUTE);
+ exec_list_push_tail(&shader->shared, &var->node);
+ break;
+
case nir_var_system_value:
exec_list_push_tail(&shader->system_values, &var->node);
break;
nir_local_variable_create(nir_function_impl *impl,
const struct glsl_type *type, const char *name)
{
- nir_variable *var = rzalloc(impl->overload->function->shader, nir_variable);
+ nir_variable *var = rzalloc(impl->function->shader, nir_variable);
var->name = ralloc_strdup(var, name);
var->type = type;
var->data.mode = nir_var_local;
nir_function *func = ralloc(shader, nir_function);
exec_list_push_tail(&shader->functions, &func->node);
- exec_list_make_empty(&func->overload_list);
+
func->name = ralloc_strdup(func, name);
func->shader = shader;
+ func->num_params = 0;
+ func->params = NULL;
+ func->return_type = glsl_void_type();
+ func->impl = NULL;
return func;
}
-nir_function_overload *
-nir_function_overload_create(nir_function *func)
-{
- void *mem_ctx = ralloc_parent(func);
-
- nir_function_overload *overload = ralloc(mem_ctx, nir_function_overload);
-
- overload->num_params = 0;
- overload->params = NULL;
- overload->return_type = glsl_void_type();
- overload->impl = NULL;
-
- exec_list_push_tail(&func->overload_list, &overload->node);
- overload->function = func;
-
- return overload;
-}
-
void nir_src_copy(nir_src *dest, const nir_src *src, void *mem_ctx)
{
dest->is_ssa = src->is_ssa;
}
nir_function_impl *
-nir_function_impl_create(nir_function_overload *overload)
+nir_function_impl_create_bare(nir_shader *shader)
{
- assert(overload->impl == NULL);
-
- void *mem_ctx = ralloc_parent(overload);
-
- nir_function_impl *impl = ralloc(mem_ctx, nir_function_impl);
+ nir_function_impl *impl = ralloc(shader, nir_function_impl);
- overload->impl = impl;
- impl->overload = overload;
+ impl->function = NULL;
cf_init(&impl->cf_node, nir_cf_node_function);
impl->valid_metadata = nir_metadata_none;
/* create start & end blocks */
- nir_block *start_block = nir_block_create(mem_ctx);
- nir_block *end_block = nir_block_create(mem_ctx);
+ nir_block *start_block = nir_block_create(shader);
+ nir_block *end_block = nir_block_create(shader);
start_block->cf_node.parent = &impl->cf_node;
end_block->cf_node.parent = &impl->cf_node;
impl->end_block = end_block;
return impl;
}
+nir_function_impl *
+nir_function_impl_create(nir_function *function)
+{
+ assert(function->impl == NULL);
+
+ nir_function_impl *impl = nir_function_impl_create_bare(function->shader);
+
+ function->impl = impl;
+ impl->function = function;
+
+ impl->num_params = function->num_params;
+ impl->params = ralloc_array(function->shader,
+ nir_variable *, impl->num_params);
+
+ return impl;
+}
+
nir_block *
nir_block_create(nir_shader *shader)
{
}
nir_call_instr *
-nir_call_instr_create(nir_shader *shader, nir_function_overload *callee)
+nir_call_instr_create(nir_shader *shader, nir_function *callee)
{
nir_call_instr *instr = ralloc(shader, nir_call_instr);
instr_init(&instr->instr, nir_instr_type_call);
for (unsigned i = 0; i < num_srcs; i++)
src_init(&instr->src[i].src);
+ instr->texture_index = 0;
+ instr->texture_array_size = 0;
+ instr->texture = NULL;
instr->sampler_index = 0;
- instr->sampler_array_size = 0;
instr->sampler = NULL;
return instr;
return nir_cf_node_as_function(node);
}
+/* Reduces a cursor by trying to convert everything to after and trying to
+ * go up to block granularity when possible.
+ */
+static nir_cursor
+reduce_cursor(nir_cursor cursor)
+{
+ switch (cursor.option) {
+ case nir_cursor_before_block:
+ if (exec_list_is_empty(&cursor.block->instr_list)) {
+ /* Empty block. After is as good as before. */
+ cursor.option = nir_cursor_after_block;
+ } else {
+ /* Try to switch to after the previous block if there is one.
+ * (This isn't likely, but it can happen.)
+ */
+ nir_cf_node *prev_node = nir_cf_node_prev(&cursor.block->cf_node);
+ if (prev_node && prev_node->type == nir_cf_node_block) {
+ cursor.block = nir_cf_node_as_block(prev_node);
+ cursor.option = nir_cursor_after_block;
+ }
+ }
+ return cursor;
+
+ case nir_cursor_after_block:
+ return cursor;
+
+ case nir_cursor_before_instr: {
+ nir_instr *prev_instr = nir_instr_prev(cursor.instr);
+ if (prev_instr) {
+ /* Before this instruction is after the previous */
+ cursor.instr = prev_instr;
+ cursor.option = nir_cursor_after_instr;
+ } else {
+ /* No previous instruction. Switch to before block */
+ cursor.block = cursor.instr->block;
+ cursor.option = nir_cursor_before_block;
+ }
+ return reduce_cursor(cursor);
+ }
+
+ case nir_cursor_after_instr:
+ if (nir_instr_next(cursor.instr) == NULL) {
+ /* This is the last instruction, switch to after block */
+ cursor.option = nir_cursor_after_block;
+ cursor.block = cursor.instr->block;
+ }
+ return cursor;
+
+ default:
+ unreachable("Inavlid cursor option");
+ }
+}
+
+bool
+nir_cursors_equal(nir_cursor a, nir_cursor b)
+{
+ /* Reduced cursors should be unique */
+ a = reduce_cursor(a);
+ b = reduce_cursor(b);
+
+ return a.block == b.block && a.option == b.option;
+}
+
static bool
add_use_cb(nir_src *src, void *state)
{
if (!visit_src(&instr->src[i].src, cb, state))
return false;
+ if (instr->texture != NULL)
+ if (!visit_deref_src(instr->texture, cb, state))
+ return false;
+
if (instr->sampler != NULL)
if (!visit_deref_src(instr->sampler, cb, state))
return false;
return nir_intrinsic_load_vertex_id;
case SYSTEM_VALUE_INSTANCE_ID:
return nir_intrinsic_load_instance_id;
+ case SYSTEM_VALUE_DRAW_ID:
+ return nir_intrinsic_load_draw_id;
+ case SYSTEM_VALUE_BASE_INSTANCE:
+ return nir_intrinsic_load_base_instance;
case SYSTEM_VALUE_VERTEX_ID_ZERO_BASE:
return nir_intrinsic_load_vertex_id_zero_base;
case SYSTEM_VALUE_BASE_VERTEX:
return SYSTEM_VALUE_VERTEX_ID;
case nir_intrinsic_load_instance_id:
return SYSTEM_VALUE_INSTANCE_ID;
+ case nir_intrinsic_load_draw_id:
+ return SYSTEM_VALUE_DRAW_ID;
+ case nir_intrinsic_load_base_instance:
+ return SYSTEM_VALUE_BASE_INSTANCE;
case nir_intrinsic_load_vertex_id_zero_base:
return SYSTEM_VALUE_VERTEX_ID_ZERO_BASE;
case nir_intrinsic_load_base_vertex:
return exec_node_data(out_type, parent, field); \
}
-struct nir_function_overload;
struct nir_function;
struct nir_shader;
struct nir_instr;
nir_var_local,
nir_var_uniform,
nir_var_shader_storage,
+ nir_var_shared,
nir_var_system_value
} nir_variable_mode;
unsigned has_initializer:1;
/**
- * Is this variable a generic output or input that has not yet been matched
- * up to a variable in another stage of the pipeline?
- *
- * This is used by the linker as scratch storage while assigning locations
- * to generic inputs and outputs.
- */
- unsigned is_unmatched_generic_inout:1;
-
- /**
* If non-zero, then this variable may be packed along with other variables
* into a single varying slot, so this offset should be applied when
* accessing components. For example, an offset of 1 means that the x
int index;
/**
+ * Descriptor set binding for sampler or UBO.
+ */
+ int descriptor_set;
+
+ /**
* Initial binding point for a sampler or UBO.
*
* For array types, this represents the binding point for the first element.
/**
* Location an atomic counter is stored at.
*/
- struct {
- unsigned offset;
- } atomic;
+ unsigned offset;
/**
* ARB_shader_image_load_store qualifiers.
#define nir_foreach_variable(var, var_list) \
foreach_list_typed(nir_variable, var, node, var_list)
+/**
+ * Returns the bits in the inputs_read, outputs_written, or
+ * system_values_read bitfield corresponding to this variable.
+ */
+static inline uint64_t
+nir_variable_get_io_mask(nir_variable *var, gl_shader_stage stage)
+{
+ assert(var->data.mode == nir_var_shader_in ||
+ var->data.mode == nir_var_shader_out ||
+ var->data.mode == nir_var_system_value);
+ assert(var->data.location >= 0);
+
+ const struct glsl_type *var_type = var->type;
+ if (stage == MESA_SHADER_GEOMETRY && var->data.mode == nir_var_shader_in) {
+ /* Most geometry shader inputs are per-vertex arrays */
+ if (var->data.location >= VARYING_SLOT_VAR0)
+ assert(glsl_type_is_array(var_type));
+
+ if (glsl_type_is_array(var_type))
+ var_type = glsl_get_array_element(var_type);
+ }
+
+ bool is_vertex_input = (var->data.mode == nir_var_shader_in &&
+ stage == MESA_SHADER_VERTEX);
+ unsigned slots = glsl_count_attribute_slots(var_type, is_vertex_input);
+ return ((1ull << slots) - 1) << var->data.location;
+}
+
typedef struct {
struct exec_node node;
bool is_ssa;
} nir_src;
-#define NIR_SRC_INIT (nir_src) { { NULL } }
+#ifdef __cplusplus
+# define NIR_SRC_INIT nir_src()
+#else
+# define NIR_SRC_INIT (nir_src) { { NULL } }
+#endif
#define nir_foreach_use(reg_or_ssa_def, src) \
list_for_each_entry(nir_src, src, &(reg_or_ssa_def)->uses, use_link)
bool is_ssa;
} nir_dest;
-#define NIR_DEST_INIT (nir_dest) { { { NULL } } }
+#ifdef __cplusplus
+# define NIR_DEST_INIT nir_dest()
+#else
+# define NIR_DEST_INIT (nir_dest) { { { NULL } } }
+#endif
#define nir_foreach_def(reg, dest) \
list_for_each_entry(nir_dest, dest, &(reg)->defs, reg.def_link)
nir_deref_var **params;
nir_deref_var *return_deref;
- struct nir_function_overload *callee;
+ struct nir_function *callee;
} nir_call_instr;
#define INTRINSIC(name, num_srcs, src_components, has_dest, dest_components, \
nir_tex_src_ms_index, /* MSAA sample index */
nir_tex_src_ddx,
nir_tex_src_ddy,
+ nir_tex_src_texture_offset, /* < dynamically uniform indirect offset */
nir_tex_src_sampler_offset, /* < dynamically uniform indirect offset */
nir_num_tex_src_types
} nir_tex_src_type;
/* gather component selector */
unsigned component : 2;
+ /** The texture index
+ *
+ * If this texture instruction has a nir_tex_src_texture_offset source,
+ * then the texture index is given by texture_index + texture_offset.
+ */
+ unsigned texture_index;
+
+ /** The size of the texture array or 0 if it's not an array */
+ unsigned texture_array_size;
+
+ /** The texture deref
+ *
+ * If both this and `sampler` are both NULL, use texture_index instead.
+ * If `texture` is NULL, but `sampler` is non-NULL, then the texture is
+ * implied from the sampler.
+ */
+ nir_deref_var *texture;
+
/** The sampler index
*
* If this texture instruction has a nir_tex_src_sampler_offset source,
*/
unsigned sampler_index;
- /** The size of the sampler array or 0 if it's not an array */
- unsigned sampler_array_size;
-
- nir_deref_var *sampler; /* if this is NULL, use sampler_index instead */
+ /** The sampler deref
+ *
+ * If this is null, use sampler_index instead.
+ */
+ nir_deref_var *sampler;
} nir_tex_instr;
static inline unsigned
typedef struct {
nir_cf_node cf_node;
- /** pointer to the overload of which this is an implementation */
- struct nir_function_overload *overload;
+ /** pointer to the function of which this is an implementation */
+ struct nir_function *function;
struct exec_list body; /** < list of nir_cf_node */
const struct glsl_type *type;
} nir_parameter;
-typedef struct nir_function_overload {
+typedef struct nir_function {
struct exec_node node;
+ const char *name;
+ struct nir_shader *shader;
+
unsigned num_params;
nir_parameter *params;
const struct glsl_type *return_type;
- nir_function_impl *impl; /** < NULL if the overload is only declared yet */
-
- /** pointer to the function of which this is an overload */
- struct nir_function *function;
-} nir_function_overload;
-
-typedef struct nir_function {
- struct exec_node node;
-
- struct exec_list overload_list; /** < list of nir_function_overload */
- const char *name;
- struct nir_shader *shader;
+ /** The implementation of this function.
+ *
+ * If the function is only declared and not implemented, this is NULL.
+ */
+ nir_function_impl *impl;
} nir_function;
-#define nir_function_first_overload(func) \
- exec_node_data(nir_function_overload, \
- exec_list_get_head(&(func)->overload_list), node)
-
typedef struct nir_shader_compiler_options {
+ bool lower_fdiv;
bool lower_ffma;
bool lower_flrp;
bool lower_fpow;
bool lower_fsat;
bool lower_fsqrt;
+ bool lower_fmod;
+ bool lower_bitfield_extract;
+ bool lower_bitfield_insert;
+ bool lower_uadd_carry;
+ bool lower_usub_borrow;
/** lowers fneg and ineg to fsub and isub. */
bool lower_negate;
/** lowers fsub and isub to fadd+fneg and iadd+ineg. */
* are simulated by floats.)
*/
bool native_integers;
+
+ /* Indicates that the driver only has zero-based vertex id */
+ bool vertex_id_zero_based;
} nir_shader_compiler_options;
typedef struct nir_shader_info {
/** list of outputs (nir_variable) */
struct exec_list outputs;
+ /** list of shared compute variables (nir_variable) */
+ struct exec_list shared;
+
/** Set of driver-specific options for the shader.
*
* The memory for the options is expected to be kept in a single static
* the highest index a load_input_*, load_uniform_*, etc. intrinsic can
* access plus one
*/
- unsigned num_inputs, num_uniforms, num_outputs;
+ unsigned num_inputs, num_uniforms, num_outputs, num_shared;
/** The shader stage, such as MESA_SHADER_VERTEX. */
gl_shader_stage stage;
} nir_shader;
-#define nir_foreach_overload(shader, overload) \
- foreach_list_typed(nir_function, func, node, &(shader)->functions) \
- foreach_list_typed(nir_function_overload, overload, node, \
- &(func)->overload_list)
+#define nir_foreach_function(shader, func) \
+ foreach_list_typed(nir_function, func, node, &(shader)->functions)
nir_shader *nir_shader_create(void *mem_ctx,
gl_shader_stage stage,
/** creates a function and adds it to the shader's list of functions */
nir_function *nir_function_create(nir_shader *shader, const char *name);
-/** creates a null function returning null */
-nir_function_overload *nir_function_overload_create(nir_function *func);
-
-nir_function_impl *nir_function_impl_create(nir_function_overload *func);
+nir_function_impl *nir_function_impl_create(nir_function *func);
+/** creates a function_impl that isn't tied to any particular function */
+nir_function_impl *nir_function_impl_create_bare(nir_shader *shader);
nir_block *nir_block_create(nir_shader *shader);
nir_if *nir_if_create(nir_shader *shader);
nir_intrinsic_op op);
nir_call_instr *nir_call_instr_create(nir_shader *shader,
- nir_function_overload *callee);
+ nir_function *callee);
nir_tex_instr *nir_tex_instr_create(nir_shader *shader, unsigned num_srcs);
};
} nir_cursor;
+static inline nir_block *
+nir_cursor_current_block(nir_cursor cursor)
+{
+ if (cursor.option == nir_cursor_before_instr ||
+ cursor.option == nir_cursor_after_instr) {
+ return cursor.instr->block;
+ } else {
+ return cursor.block;
+ }
+}
+
+bool nir_cursors_equal(nir_cursor a, nir_cursor b);
+
static inline nir_cursor
nir_before_block(nir_block *block)
{
}
static inline nir_cursor
+nir_after_cf_node_and_phis(nir_cf_node *node)
+{
+ if (node->type == nir_cf_node_block)
+ return nir_after_block(nir_cf_node_as_block(node));
+
+ nir_block *block = nir_cf_node_as_block(nir_cf_node_next(node));
+ assert(block->cf_node.type == nir_cf_node_block);
+
+ nir_foreach_instr(block, instr) {
+ if (instr->type != nir_instr_type_phi)
+ return nir_before_instr(instr);
+ }
+ return nir_after_block(block);
+}
+
+static inline nir_cursor
nir_before_cf_list(struct exec_list *cf_list)
{
nir_cf_node *first_node = exec_node_data(nir_cf_node,
void nir_print_shader(nir_shader *shader, FILE *fp);
void nir_print_instr(const nir_instr *instr, FILE *fp);
-nir_shader * nir_shader_clone(void *mem_ctx, const nir_shader *s);
+nir_shader *nir_shader_clone(void *mem_ctx, const nir_shader *s);
+nir_function_impl *nir_function_impl_clone(const nir_function_impl *impl);
+nir_constant *nir_constant_clone(const nir_constant *c, nir_variable *var);
#ifdef DEBUG
void nir_validate_shader(nir_shader *shader);
void nir_metadata_set_validation_flag(nir_shader *shader);
void nir_metadata_check_validation_flag(nir_shader *shader);
+
+#include "util/debug.h"
+static inline bool
+should_clone_nir(void)
+{
+ static int should_clone = -1;
+ if (should_clone < 0)
+ should_clone = env_var_as_boolean("NIR_TEST_CLONE", false);
+
+ return should_clone;
+}
#else
static inline void nir_validate_shader(nir_shader *shader) { (void) shader; }
static inline void nir_metadata_set_validation_flag(nir_shader *shader) { (void) shader; }
static inline void nir_metadata_check_validation_flag(nir_shader *shader) { (void) shader; }
+static inline bool should_clone_nir(void) { return false; }
#endif /* DEBUG */
+#define _PASS(nir, do_pass) do { \
+ do_pass \
+ nir_validate_shader(nir); \
+ if (should_clone_nir()) { \
+ nir_shader *clone = nir_shader_clone(ralloc_parent(nir), nir); \
+ ralloc_free(nir); \
+ nir = clone; \
+ } \
+} while (0)
+
+#define NIR_PASS(progress, nir, pass, ...) _PASS(nir, \
+ nir_metadata_set_validation_flag(nir); \
+ if (pass(nir, ##__VA_ARGS__)) { \
+ progress = true; \
+ nir_metadata_check_validation_flag(nir); \
+ } \
+)
+
+#define NIR_PASS_V(nir, pass, ...) _PASS(nir, \
+ pass(nir, ##__VA_ARGS__); \
+)
+
void nir_calc_dominance_impl(nir_function_impl *impl);
void nir_calc_dominance(nir_shader *shader);
bool nir_split_var_copies(nir_shader *shader);
+bool nir_lower_returns_impl(nir_function_impl *impl);
+bool nir_lower_returns(nir_shader *shader);
+
+bool nir_inline_functions(nir_shader *shader);
+
void nir_lower_var_copy_instr(nir_intrinsic_instr *copy, void *mem_ctx);
void nir_lower_var_copies(nir_shader *shader);
bool nir_lower_global_vars_to_local(nir_shader *shader);
+bool nir_lower_indirect_derefs(nir_shader *shader, uint32_t mode_mask);
+
bool nir_lower_locals_to_regs(nir_shader *shader);
-void nir_lower_outputs_to_temporaries(nir_shader *shader);
+void nir_lower_outputs_to_temporaries(nir_shader *shader,
+ nir_function *entrypoint);
+
+void nir_shader_gather_info(nir_shader *shader, nir_function_impl *entrypoint);
void nir_assign_var_locations(struct exec_list *var_list,
unsigned *size,
void nir_lower_vars_to_ssa(nir_shader *shader);
-bool nir_remove_dead_variables(nir_shader *shader);
+bool nir_remove_dead_variables(nir_shader *shader, nir_variable_mode mode);
void nir_move_vec_src_uses_to_dest(nir_shader *shader);
bool nir_lower_vec_to_movs(nir_shader *shader);
void nir_convert_to_ssa_impl(nir_function_impl *impl);
void nir_convert_to_ssa(nir_shader *shader);
+bool nir_repair_ssa_impl(nir_function_impl *impl);
+bool nir_repair_ssa(nir_shader *shader);
+
/* If phi_webs_only is true, only convert SSA values involved in phi nodes to
* registers. If false, convert all values (even those not involved in a phi
* node) to registers.
if isinstance(self.value, (bool)):
return 'NIR_TRUE' if self.value else 'NIR_FALSE'
if isinstance(self.value, (int, long)):
- return hex(struct.unpack('I', struct.pack('i', self.value))[0])
+ return hex(struct.unpack('I', struct.pack('i' if self.value < 0 else 'I', self.value))[0])
elif isinstance(self.value, float):
return hex(struct.unpack('I', struct.pack('f', self.value))[0])
else:
condition_flags[${index}] = ${condition};
% endfor
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- progress |= ${pass_name}_impl(overload->impl, condition_flags);
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ progress |= ${pass_name}_impl(function->impl, condition_flags);
}
return progress;
{
memset(build, 0, sizeof(*build));
build->impl = impl;
- build->shader = impl->overload->function->shader;
+ build->shader = impl->function->shader;
+}
+
+static inline void
+nir_builder_init_simple_shader(nir_builder *build, void *mem_ctx,
+ gl_shader_stage stage,
+ const nir_shader_compiler_options *options)
+{
+ build->shader = nir_shader_create(mem_ctx, stage, options);
+ nir_function *func = nir_function_create(build->shader, "main");
+ build->impl = nir_function_impl_create(func);
+ build->cursor = nir_after_cf_list(&build->impl->body);
}
static inline void
}
static inline nir_ssa_def *
+nir_ssa_undef(nir_builder *build, unsigned num_components)
+{
+ nir_ssa_undef_instr *undef =
+ nir_ssa_undef_instr_create(build->shader, num_components);
+ if (!undef)
+ return NULL;
+
+ nir_instr_insert(nir_before_block(nir_start_block(build->impl)),
+ &undef->instr);
+
+ return &undef->def;
+}
+
+static inline nir_ssa_def *
nir_build_imm(nir_builder *build, unsigned num_components, nir_const_value value)
{
nir_load_const_instr *load_const =
nir_imov_alu(build, alu_src, num_components);
}
+/* Selects the right fdot given the number of components in each source. */
+static inline nir_ssa_def *
+nir_fdot(nir_builder *build, nir_ssa_def *src0, nir_ssa_def *src1)
+{
+ assert(src0->num_components == src1->num_components);
+ switch (src0->num_components) {
+ case 1: return nir_fmul(build, src0, src1);
+ case 2: return nir_fdot2(build, src0, src1);
+ case 3: return nir_fdot3(build, src0, src1);
+ case 4: return nir_fdot4(build, src0, src1);
+ default:
+ unreachable("bad component size");
+ }
+
+ return NULL;
+}
+
static inline nir_ssa_def *
nir_channel(nir_builder *b, nir_ssa_def *def, unsigned c)
{
}
static inline void
-nir_store_var(nir_builder *build, nir_variable *var, nir_ssa_def *value)
+nir_store_var(nir_builder *build, nir_variable *var, nir_ssa_def *value,
+ unsigned writemask)
{
const unsigned num_components = glsl_get_vector_elements(var->type);
nir_intrinsic_instr *store =
nir_intrinsic_instr_create(build->shader, nir_intrinsic_store_var);
store->num_components = num_components;
+ store->const_index[0] = writemask;
store->variables[0] = nir_deref_var_create(store, var);
store->src[0] = nir_src_for_ssa(value);
nir_builder_instr_insert(build, &store->instr);
}
+static inline void
+nir_store_deref_var(nir_builder *build, nir_deref_var *deref,
+ nir_ssa_def *value, unsigned writemask)
+{
+ const unsigned num_components =
+ glsl_get_vector_elements(nir_deref_tail(&deref->deref)->type);
+
+ nir_intrinsic_instr *store =
+ nir_intrinsic_instr_create(build->shader, nir_intrinsic_store_var);
+ store->num_components = num_components;
+ store->const_index[0] = writemask & ((1 << num_components) - 1);
+ store->variables[0] = nir_deref_as_var(nir_copy_deref(store, &deref->deref));
+ store->src[0] = nir_src_for_ssa(value);
+ nir_builder_instr_insert(build, &store->instr);
+}
+
+static inline void
+nir_copy_deref_var(nir_builder *build, nir_deref_var *dest, nir_deref_var *src)
+{
+ assert(nir_deref_tail(&dest->deref)->type ==
+ nir_deref_tail(&src->deref)->type);
+
+ nir_intrinsic_instr *copy =
+ nir_intrinsic_instr_create(build->shader, nir_intrinsic_copy_var);
+ copy->variables[0] = nir_deref_as_var(nir_copy_deref(copy, &dest->deref));
+ copy->variables[1] = nir_deref_as_var(nir_copy_deref(copy, &src->deref));
+ nir_builder_instr_insert(build, ©->instr);
+}
+
+static inline void
+nir_copy_var(nir_builder *build, nir_variable *dest, nir_variable *src)
+{
+ nir_intrinsic_instr *copy =
+ nir_intrinsic_instr_create(build->shader, nir_intrinsic_copy_var);
+ copy->variables[0] = nir_deref_var_create(copy, dest);
+ copy->variables[1] = nir_deref_var_create(copy, src);
+ nir_builder_instr_insert(build, ©->instr);
+}
+
+static inline nir_ssa_def *
+nir_load_system_value(nir_builder *build, nir_intrinsic_op op, int index)
+{
+ nir_intrinsic_instr *load = nir_intrinsic_instr_create(build->shader, op);
+ load->num_components = nir_intrinsic_infos[op].dest_components;
+ load->const_index[0] = index;
+ nir_ssa_dest_init(&load->instr, &load->dest,
+ nir_intrinsic_infos[op].dest_components, NULL);
+ nir_builder_instr_insert(build, &load->instr);
+ return &load->dest.ssa;
+}
+
+static inline void
+nir_jump(nir_builder *build, nir_jump_type jump_type)
+{
+ nir_jump_instr *jump = nir_jump_instr_create(build->shader, jump_type);
+ nir_builder_instr_insert(build, &jump->instr);
+}
+
#endif /* NIR_BUILDER_H */
*/
typedef struct {
+ /* True if we are cloning an entire shader. */
+ bool global_clone;
+
/* maps orig ptr -> cloned ptr: */
- struct hash_table *ptr_table;
+ struct hash_table *remap_table;
/* List of phi sources. */
struct list_head phi_srcs;
} clone_state;
static void
-init_clone_state(clone_state *state)
+init_clone_state(clone_state *state, bool global)
{
- state->ptr_table = _mesa_hash_table_create(NULL, _mesa_hash_pointer,
- _mesa_key_pointer_equal);
+ state->global_clone = global;
+ state->remap_table = _mesa_hash_table_create(NULL, _mesa_hash_pointer,
+ _mesa_key_pointer_equal);
list_inithead(&state->phi_srcs);
}
static void
free_clone_state(clone_state *state)
{
- _mesa_hash_table_destroy(state->ptr_table, NULL);
+ _mesa_hash_table_destroy(state->remap_table, NULL);
}
-static void *
-lookup_ptr(clone_state *state, const void *ptr)
+static inline void *
+_lookup_ptr(clone_state *state, const void *ptr, bool global)
{
struct hash_entry *entry;
if (!ptr)
return NULL;
- entry = _mesa_hash_table_search(state->ptr_table, ptr);
+ if (!state->global_clone && global)
+ return (void *)ptr;
+
+ entry = _mesa_hash_table_search(state->remap_table, ptr);
assert(entry && "Failed to find pointer!");
if (!entry)
return NULL;
}
static void
-store_ptr(clone_state *state, void *nptr, const void *ptr)
+add_remap(clone_state *state, void *nptr, const void *ptr)
+{
+ _mesa_hash_table_insert(state->remap_table, ptr, nptr);
+}
+
+static void *
+remap_local(clone_state *state, const void *ptr)
+{
+ return _lookup_ptr(state, ptr, false);
+}
+
+static void *
+remap_global(clone_state *state, const void *ptr)
+{
+ return _lookup_ptr(state, ptr, true);
+}
+
+static nir_register *
+remap_reg(clone_state *state, const nir_register *reg)
+{
+ return _lookup_ptr(state, reg, reg->is_global);
+}
+
+static nir_variable *
+remap_var(clone_state *state, const nir_variable *var)
{
- _mesa_hash_table_insert(state->ptr_table, ptr, nptr);
+ return _lookup_ptr(state, var, var->data.mode != nir_var_local);
}
-static nir_constant *
-clone_constant(clone_state *state, const nir_constant *c, nir_variable *nvar)
+nir_constant *
+nir_constant_clone(const nir_constant *c, nir_variable *nvar)
{
nir_constant *nc = ralloc(nvar, nir_constant);
nc->num_elements = c->num_elements;
nc->elements = ralloc_array(nvar, nir_constant *, c->num_elements);
for (unsigned i = 0; i < c->num_elements; i++) {
- nc->elements[i] = clone_constant(state, c->elements[i], nvar);
+ nc->elements[i] = nir_constant_clone(c->elements[i], nvar);
}
return nc;
clone_variable(clone_state *state, const nir_variable *var)
{
nir_variable *nvar = rzalloc(state->ns, nir_variable);
- store_ptr(state, nvar, var);
+ add_remap(state, nvar, var);
nvar->type = var->type;
nvar->name = ralloc_strdup(nvar, var->name);
var->num_state_slots * sizeof(nir_state_slot));
if (var->constant_initializer) {
nvar->constant_initializer =
- clone_constant(state, var->constant_initializer, nvar);
+ nir_constant_clone(var->constant_initializer, nvar);
}
nvar->interface_type = var->interface_type;
clone_register(clone_state *state, const nir_register *reg)
{
nir_register *nreg = rzalloc(state->ns, nir_register);
- store_ptr(state, nreg, reg);
+ add_remap(state, nreg, reg);
nreg->num_components = reg->num_components;
nreg->num_array_elems = reg->num_array_elems;
{
nsrc->is_ssa = src->is_ssa;
if (src->is_ssa) {
- nsrc->ssa = lookup_ptr(state, src->ssa);
+ nsrc->ssa = remap_local(state, src->ssa);
} else {
- nsrc->reg.reg = lookup_ptr(state, src->reg.reg);
+ nsrc->reg.reg = remap_reg(state, src->reg.reg);
if (src->reg.indirect) {
nsrc->reg.indirect = ralloc(ninstr_or_if, nir_src);
__clone_src(state, ninstr_or_if, nsrc->reg.indirect, src->reg.indirect);
ndst->is_ssa = dst->is_ssa;
if (dst->is_ssa) {
nir_ssa_dest_init(ninstr, ndst, dst->ssa.num_components, dst->ssa.name);
- store_ptr(state, &ndst->ssa, &dst->ssa);
+ add_remap(state, &ndst->ssa, &dst->ssa);
} else {
- ndst->reg.reg = lookup_ptr(state, dst->reg.reg);
+ ndst->reg.reg = remap_reg(state, dst->reg.reg);
if (dst->reg.indirect) {
ndst->reg.indirect = ralloc(ninstr, nir_src);
__clone_src(state, ninstr, ndst->reg.indirect, dst->reg.indirect);
clone_deref_var(clone_state *state, const nir_deref_var *dvar,
nir_instr *ninstr)
{
- nir_variable *nvar = lookup_ptr(state, dvar->var);
+ nir_variable *nvar = remap_var(state, dvar->var);
nir_deref_var *ndvar = nir_deref_var_create(ninstr, nvar);
if (dvar->deref.child)
memcpy(&nlc->value, &lc->value, sizeof(nlc->value));
- store_ptr(state, &nlc->def, &lc->def);
+ add_remap(state, &nlc->def, &lc->def);
return nlc;
}
nir_ssa_undef_instr *nsa =
nir_ssa_undef_instr_create(state->ns, sa->def.num_components);
- store_ptr(state, &nsa->def, &sa->def);
+ add_remap(state, &nsa->def, &sa->def);
return nsa;
}
ntex->is_new_style_shadow = tex->is_new_style_shadow;
memcpy(ntex->const_offset, tex->const_offset, sizeof(ntex->const_offset));
ntex->component = tex->component;
+ ntex->texture_index = tex->texture_index;
+ ntex->texture_array_size = tex->texture_array_size;
+ if (tex->texture)
+ ntex->texture = clone_deref_var(state, tex->texture, &ntex->instr);
ntex->sampler_index = tex->sampler_index;
- ntex->sampler_array_size = tex->sampler_array_size;
if (tex->sampler)
ntex->sampler = clone_deref_var(state, tex->sampler, &ntex->instr);
static nir_call_instr *
clone_call(clone_state *state, const nir_call_instr *call)
{
- nir_function_overload *ncallee = lookup_ptr(state, call->callee);
+ nir_function *ncallee = remap_global(state, call->callee);
nir_call_instr *ncall = nir_call_instr_create(state->ns, ncallee);
for (unsigned i = 0; i < ncall->num_params; i++)
assert(exec_list_is_empty(&nblk->instr_list));
/* We need this for phi sources */
- store_ptr(state, nblk, blk);
+ add_remap(state, nblk, blk);
nir_foreach_instr(blk, instr) {
if (instr->type == nir_instr_type_phi) {
}
static nir_function_impl *
-clone_function_impl(clone_state *state, const nir_function_impl *fi,
- nir_function_overload *nfo)
+clone_function_impl(clone_state *state, const nir_function_impl *fi)
{
- nir_function_impl *nfi = nir_function_impl_create(nfo);
+ nir_function_impl *nfi = nir_function_impl_create_bare(state->ns);
clone_var_list(state, &nfi->locals, &fi->locals);
clone_reg_list(state, &nfi->registers, &fi->registers);
nfi->num_params = fi->num_params;
nfi->params = ralloc_array(state->ns, nir_variable *, fi->num_params);
for (unsigned i = 0; i < fi->num_params; i++) {
- nfi->params[i] = lookup_ptr(state, fi->params[i]);
+ nfi->params[i] = remap_local(state, fi->params[i]);
}
- nfi->return_var = lookup_ptr(state, fi->return_var);
+ nfi->return_var = remap_local(state, fi->return_var);
assert(list_empty(&state->phi_srcs));
* add it to the phi_srcs list and we fix it up here.
*/
list_for_each_entry_safe(nir_phi_src, src, &state->phi_srcs, src.use_link) {
- src->pred = lookup_ptr(state, src->pred);
+ src->pred = remap_local(state, src->pred);
assert(src->src.is_ssa);
- src->src.ssa = lookup_ptr(state, src->src.ssa);
+ src->src.ssa = remap_local(state, src->src.ssa);
/* Remove from this list and place in the uses of the SSA def */
list_del(&src->src.use_link);
return nfi;
}
-static nir_function_overload *
-clone_function_overload(clone_state *state, const nir_function_overload *fo,
- nir_function *nfxn)
+nir_function_impl *
+nir_function_impl_clone(const nir_function_impl *fi)
{
- nir_function_overload *nfo = nir_function_overload_create(nfxn);
-
- /* Needed for call instructions */
- store_ptr(state, nfo, fo);
+ clone_state state;
+ init_clone_state(&state, false);
- nfo->num_params = fo->num_params;
- nfo->params = ralloc_array(state->ns, nir_parameter, fo->num_params);
- memcpy(nfo->params, fo->params, sizeof(nir_parameter) * fo->num_params);
+ /* We use the same shader */
+ state.ns = fi->function->shader;
- nfo->return_type = fo->return_type;
+ nir_function_impl *nfi = clone_function_impl(&state, fi);
- /* At first glance, it looks like we should clone the function_impl here.
- * However, call instructions need to be able to reference at least the
- * overload and those will get processed as we clone the function_impl's.
- * We stop here and do function_impls as a second pass.
- */
+ free_clone_state(&state);
- return nfo;
+ return nfi;
}
static nir_function *
assert(ns == state->ns);
nir_function *nfxn = nir_function_create(ns, fxn->name);
- foreach_list_typed(nir_function_overload, fo, node, &fxn->overload_list)
- clone_function_overload(state, fo, nfxn);
+ /* Needed for call instructions */
+ add_remap(state, nfxn, fxn);
+
+ nfxn->num_params = fxn->num_params;
+ nfxn->params = ralloc_array(state->ns, nir_parameter, fxn->num_params);
+ memcpy(nfxn->params, fxn->params, sizeof(nir_parameter) * fxn->num_params);
+
+ nfxn->return_type = fxn->return_type;
+
+ /* At first glance, it looks like we should clone the function_impl here.
+ * However, call instructions need to be able to reference at least the
+ * function and those will get processed as we clone the function_impl's.
+ * We stop here and do function_impls as a second pass.
+ */
return nfxn;
}
nir_shader_clone(void *mem_ctx, const nir_shader *s)
{
clone_state state;
- init_clone_state(&state);
+ init_clone_state(&state, true);
nir_shader *ns = nir_shader_create(mem_ctx, s->stage, s->options);
state.ns = ns;
clone_var_list(&state, &ns->uniforms, &s->uniforms);
clone_var_list(&state, &ns->inputs, &s->inputs);
clone_var_list(&state, &ns->outputs, &s->outputs);
+ clone_var_list(&state, &ns->shared, &s->shared);
clone_var_list(&state, &ns->globals, &s->globals);
clone_var_list(&state, &ns->system_values, &s->system_values);
- /* Go through and clone functions and overloads */
+ /* Go through and clone functions */
foreach_list_typed(nir_function, fxn, node, &s->functions)
clone_function(&state, fxn, ns);
- /* Only after all overloads are cloned can we clone the actual function
+ /* Only after all functions are cloned can we clone the actual function
* implementations. This is because nir_call_instr's need to reference the
- * overloads of other functions and we don't know what order the functions
+ * functions of other functions and we don't know what order the functions
* will have in the list.
*/
- nir_foreach_overload(s, fo) {
- nir_function_overload *nfo = lookup_ptr(&state, fo);
- clone_function_impl(&state, fo->impl, nfo);
+ nir_foreach_function(s, fxn) {
+ nir_function *nfxn = remap_global(&state, fxn);
+ nfxn->impl = clone_function_impl(&state, fxn->impl);
+ nfxn->impl->function = nfxn;
}
clone_reg_list(&state, &ns->registers, &s->registers);
ns->num_inputs = s->num_inputs;
ns->num_uniforms = s->num_uniforms;
ns->num_outputs = s->num_outputs;
+ ns->num_shared = s->num_shared;
free_clone_state(&state);
nir_block *next_block = nir_cf_node_as_block(next);
link_blocks(block, next_block, NULL);
- } else {
- assert(parent->type == nir_cf_node_loop);
+ } else if (parent->type == nir_cf_node_loop) {
nir_loop *loop = nir_cf_node_as_loop(parent);
nir_cf_node *head = nir_loop_first_cf_node(loop);
link_blocks(block, head_block, NULL);
insert_phi_undef(head_block, block);
+ } else {
+ assert(parent->type == nir_cf_node_function);
+ nir_function_impl *impl = nir_cf_node_as_function(parent);
+ link_blocks(block, impl->end_block, NULL);
}
} else {
nir_cf_node *next = nir_cf_node_next(&block->cf_node);
{
nir_block *block_begin, *block_end, *block_before, *block_after;
+ if (nir_cursors_equal(begin, end)) {
+ exec_list_make_empty(&extracted->list);
+ extracted->impl = NULL; /* we shouldn't need this */
+ return;
+ }
+
/* In the case where begin points to an instruction in some basic block and
* end points to the end of the same basic block, we rely on the fact that
* splitting on an instruction moves earlier instructions into a new basic
{
nir_block *before, *after;
+ if (exec_list_is_empty(&cf_list->list))
+ return;
+
split_block_cursor(cursor, &before, &after);
foreach_list_typed_safe(nir_cf_node, node, node, &cf_list->list) {
}
}
- assert(new_idom);
if (block->imm_dom != new_idom) {
block->imm_dom = new_idom;
state->progress = true;
struct set_entry *entry;
set_foreach(block->predecessors, entry) {
nir_block *runner = (nir_block *) entry->key;
+
+ /* Skip unreachable predecessors */
+ if (runner->imm_dom == NULL)
+ continue;
+
while (runner != block->imm_dom) {
_mesa_set_add(runner->dom_frontier, block);
runner = runner->imm_dom;
void
nir_calc_dominance(nir_shader *shader)
{
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- nir_calc_dominance_impl(overload->impl);
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ nir_calc_dominance_impl(function->impl);
}
}
void
nir_dump_dom_tree_impl(nir_function_impl *impl, FILE *fp)
{
- fprintf(fp, "digraph doms_%s {\n", impl->overload->function->name);
+ fprintf(fp, "digraph doms_%s {\n", impl->function->name);
nir_foreach_block(impl, dump_block_dom, fp);
fprintf(fp, "}\n\n");
}
void
nir_dump_dom_tree(nir_shader *shader, FILE *fp)
{
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- nir_dump_dom_tree_impl(overload->impl, fp);
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ nir_dump_dom_tree_impl(function->impl, fp);
}
}
void
nir_dump_dom_frontier(nir_shader *shader, FILE *fp)
{
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- nir_dump_dom_frontier_impl(overload->impl, fp);
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ nir_dump_dom_frontier_impl(function->impl, fp);
}
}
void
nir_dump_cfg_impl(nir_function_impl *impl, FILE *fp)
{
- fprintf(fp, "digraph cfg_%s {\n", impl->overload->function->name);
+ fprintf(fp, "digraph cfg_%s {\n", impl->function->name);
nir_foreach_block(impl, dump_block_succs, fp);
fprintf(fp, "}\n\n");
}
void
nir_dump_cfg(nir_shader *shader, FILE *fp)
{
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- nir_dump_cfg_impl(overload->impl, fp);
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ nir_dump_cfg_impl(function->impl, fp);
}
}
void
nir_convert_from_ssa(nir_shader *shader, bool phi_webs_only)
{
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- nir_convert_from_ssa_impl(overload->impl, phi_webs_only);
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ nir_convert_from_ssa_impl(function->impl, phi_webs_only);
}
}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include "nir.h"
+
+static void
+gather_intrinsic_info(nir_intrinsic_instr *instr, nir_shader *shader)
+{
+ switch (instr->intrinsic) {
+ case nir_intrinsic_discard:
+ assert(shader->stage == MESA_SHADER_FRAGMENT);
+ shader->info.fs.uses_discard = true;
+ break;
+
+ case nir_intrinsic_load_front_face:
+ case nir_intrinsic_load_vertex_id:
+ case nir_intrinsic_load_vertex_id_zero_base:
+ case nir_intrinsic_load_base_vertex:
+ case nir_intrinsic_load_instance_id:
+ case nir_intrinsic_load_sample_id:
+ case nir_intrinsic_load_sample_pos:
+ case nir_intrinsic_load_sample_mask_in:
+ case nir_intrinsic_load_primitive_id:
+ case nir_intrinsic_load_invocation_id:
+ case nir_intrinsic_load_local_invocation_id:
+ case nir_intrinsic_load_work_group_id:
+ case nir_intrinsic_load_num_work_groups:
+ shader->info.system_values_read |=
+ (1 << nir_system_value_from_intrinsic(instr->intrinsic));
+ break;
+
+ case nir_intrinsic_end_primitive:
+ case nir_intrinsic_end_primitive_with_counter:
+ assert(shader->stage == MESA_SHADER_GEOMETRY);
+ shader->info.gs.uses_end_primitive = 1;
+ break;
+
+ default:
+ break;
+ }
+}
+
+static void
+gather_tex_info(nir_tex_instr *instr, nir_shader *shader)
+{
+ if (instr->op == nir_texop_tg4)
+ shader->info.uses_texture_gather = true;
+}
+
+static bool
+gather_info_block(nir_block *block, void *shader)
+{
+ nir_foreach_instr(block, instr) {
+ switch (instr->type) {
+ case nir_instr_type_intrinsic:
+ gather_intrinsic_info(nir_instr_as_intrinsic(instr), shader);
+ break;
+ case nir_instr_type_tex:
+ gather_tex_info(nir_instr_as_tex(instr), shader);
+ break;
+ case nir_instr_type_call:
+ assert(!"nir_shader_gather_info only works if functions are inlined");
+ break;
+ default:
+ break;
+ }
+ }
+
+ return true;
+}
+
+void
+nir_shader_gather_info(nir_shader *shader, nir_function_impl *entrypoint)
+{
+ shader->info.inputs_read = 0;
+ foreach_list_typed(nir_variable, var, node, &shader->inputs)
+ shader->info.inputs_read |= nir_variable_get_io_mask(var, shader->stage);
+
+ /* TODO: Some day we may need to add stream support to NIR */
+ shader->info.outputs_written = 0;
+ foreach_list_typed(nir_variable, var, node, &shader->outputs)
+ shader->info.outputs_written |= nir_variable_get_io_mask(var, shader->stage);
+
+ shader->info.system_values_read = 0;
+ foreach_list_typed(nir_variable, var, node, &shader->system_values)
+ shader->info.system_values_read |= nir_variable_get_io_mask(var, shader->stage);
+
+ nir_foreach_block(entrypoint, gather_info_block, shader);
+}
{
int count = -1;
- nir_foreach_overload(shader, overload) {
- if (!overload->impl)
+ nir_foreach_function(shader, function) {
+ if (!function->impl)
continue;
/* set_vertex_count intrinsics only appear in predecessors of the
* end block. So we don't need to walk all of them.
*/
struct set_entry *entry;
- set_foreach(overload->impl->end_block->predecessors, entry) {
+ set_foreach(function->impl->end_block->predecessors, entry) {
nir_block *block = (nir_block *) entry->key;
nir_foreach_instr_reverse(block, instr) {
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include "nir.h"
+#include "nir_builder.h"
+#include "nir_control_flow.h"
+
+struct inline_functions_state {
+ struct set *inlined;
+ nir_builder builder;
+ bool progress;
+};
+
+static bool inline_function_impl(nir_function_impl *impl, struct set *inlined);
+
+static bool
+inline_functions_block(nir_block *block, void *void_state)
+{
+ struct inline_functions_state *state = void_state;
+
+ nir_builder *b = &state->builder;
+
+ /* This is tricky. We're iterating over instructions in a block but, as
+ * we go, the block and its instruction list are being split into
+ * pieces. However, this *should* be safe since foreach_safe always
+ * stashes the next thing in the iteration. That next thing will
+ * properly get moved to the next block when it gets split, and we
+ * continue iterating there.
+ */
+ nir_foreach_instr_safe(block, instr) {
+ if (instr->type != nir_instr_type_call)
+ continue;
+
+ state->progress = true;
+
+ nir_call_instr *call = nir_instr_as_call(instr);
+ assert(call->callee->impl);
+
+ inline_function_impl(call->callee->impl, state->inlined);
+
+ nir_function_impl *callee_copy =
+ nir_function_impl_clone(call->callee->impl);
+
+ exec_list_append(&b->impl->locals, &callee_copy->locals);
+ exec_list_append(&b->impl->registers, &callee_copy->registers);
+
+ b->cursor = nir_before_instr(&call->instr);
+
+ /* Add copies of all in parameters */
+ assert(call->num_params == callee_copy->num_params);
+ for (unsigned i = 0; i < callee_copy->num_params; i++) {
+ /* Only in or inout parameters */
+ if (call->callee->params[i].param_type == nir_parameter_out)
+ continue;
+
+ nir_copy_deref_var(b, nir_deref_var_create(b->shader,
+ callee_copy->params[i]),
+ call->params[i]);
+ }
+
+ /* Pluck the body out of the function and place it here */
+ nir_cf_list body;
+ nir_cf_list_extract(&body, &callee_copy->body);
+ nir_cf_reinsert(&body, b->cursor);
+
+ b->cursor = nir_before_instr(&call->instr);
+
+ /* Add copies of all out parameters and the return */
+ assert(call->num_params == callee_copy->num_params);
+ for (unsigned i = 0; i < callee_copy->num_params; i++) {
+ /* Only out or inout parameters */
+ if (call->callee->params[i].param_type == nir_parameter_in)
+ continue;
+
+ nir_copy_deref_var(b, call->params[i],
+ nir_deref_var_create(b->shader,
+ callee_copy->params[i]));
+ }
+ if (!glsl_type_is_void(call->callee->return_type)) {
+ nir_copy_deref_var(b, call->return_deref,
+ nir_deref_var_create(b->shader,
+ callee_copy->return_var));
+ }
+
+ nir_instr_remove(&call->instr);
+ }
+
+ return true;
+}
+
+static bool
+inline_function_impl(nir_function_impl *impl, struct set *inlined)
+{
+ if (_mesa_set_search(inlined, impl))
+ return false; /* Already inlined */
+
+ struct inline_functions_state state;
+
+ state.inlined = inlined;
+ state.progress = false;
+ nir_builder_init(&state.builder, impl);
+
+ nir_foreach_block(impl, inline_functions_block, &state);
+
+ if (state.progress) {
+ /* SSA and register indices are completely messed up now */
+ nir_index_ssa_defs(impl);
+ nir_index_local_regs(impl);
+
+ nir_metadata_preserve(impl, nir_metadata_none);
+ }
+
+ _mesa_set_add(inlined, impl);
+
+ return state.progress;
+}
+
+bool
+nir_inline_functions(nir_shader *shader)
+{
+ struct set *inlined = _mesa_set_create(NULL, _mesa_hash_pointer,
+ _mesa_key_pointer_equal);
+ bool progress = false;
+
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ progress = inline_function_impl(function->impl, inlined) || progress;
+ }
+
+ _mesa_set_destroy(inlined, NULL);
+
+ return progress;
+}
hash = HASH(hash, instr->const_offset);
unsigned component = instr->component;
hash = HASH(hash, component);
+ hash = HASH(hash, instr->texture_index);
+ hash = HASH(hash, instr->texture_array_size);
hash = HASH(hash, instr->sampler_index);
- hash = HASH(hash, instr->sampler_array_size);
assert(!instr->sampler);
memcmp(tex1->const_offset, tex2->const_offset,
sizeof(tex1->const_offset)) != 0 ||
tex1->component != tex2->component ||
- tex1->sampler_index != tex2->sampler_index ||
- tex1->sampler_array_size != tex2->sampler_array_size) {
+ tex1->texture_index != tex2->texture_index ||
+ tex1->texture_array_size != tex2->texture_array_size ||
+ tex1->sampler_index != tex2->sampler_index) {
return false;
}
/* Don't support un-lowered sampler derefs currently. */
- assert(!tex1->sampler && !tex2->sampler);
+ assert(!tex1->texture && !tex1->sampler &&
+ !tex2->texture && !tex2->sampler);
return true;
}
nir_tex_instr *tex = nir_instr_as_tex(instr);
/* Don't support un-lowered sampler derefs currently. */
- if (tex->sampler)
+ if (tex->texture || tex->sampler)
return false;
return true;
INTRINSIC(load_var, 0, ARR(), true, 0, 1, 0, NIR_INTRINSIC_CAN_ELIMINATE)
-INTRINSIC(store_var, 1, ARR(0), false, 0, 1, 0, 0)
+INTRINSIC(store_var, 1, ARR(0), false, 0, 1, 1, 0)
INTRINSIC(copy_var, 0, ARR(), false, 0, 2, 0, 0)
/*
NIR_INTRINSIC_CAN_ELIMINATE | NIR_INTRINSIC_CAN_REORDER)
/*
+ * Vulkan descriptor set intrinsic
+ *
+ * The Vulkan API uses a different binding model from GL. In the Vulkan
+ * API, all external resources are represented by a tripple:
+ *
+ * (descriptor set, binding, array index)
+ *
+ * where the array index is the only thing allowed to be indirect. The
+ * vulkan_surface_index intrinsic takes the descriptor set and binding as
+ * its first two indices and the array index as its source. The third
+ * index is a nir_variable_mode in case that's useful to the backend.
+ *
+ * The intended usage is that the shader will call vulkan_surface_index to
+ * get an index and then pass that as the buffer index ubo/ssbo calls.
+ */
+INTRINSIC(vulkan_resource_index, 1, ARR(1), true, 1, 0, 3,
+ NIR_INTRINSIC_CAN_ELIMINATE | NIR_INTRINSIC_CAN_REORDER)
+
+/*
+ * variable atomic intrinsics
+ *
+ * All of these variable atomic memory operations read a value from memory,
+ * compute a new value using one of the operations below, write the new value
+ * to memory, and return the original value read.
+ *
+ * All operations take 1 source except CompSwap that takes 2. These sources
+ * represent:
+ *
+ * 0: The data parameter to the atomic function (i.e. the value to add
+ * in shared_atomic_add, etc).
+ * 1: For CompSwap only: the second data parameter.
+ *
+ * All operations take 1 variable deref.
+ */
+INTRINSIC(var_atomic_add, 1, ARR(1), true, 1, 1, 0, 0)
+INTRINSIC(var_atomic_imin, 1, ARR(1), true, 1, 1, 0, 0)
+INTRINSIC(var_atomic_umin, 1, ARR(1), true, 1, 1, 0, 0)
+INTRINSIC(var_atomic_imax, 1, ARR(1), true, 1, 1, 0, 0)
+INTRINSIC(var_atomic_umax, 1, ARR(1), true, 1, 1, 0, 0)
+INTRINSIC(var_atomic_and, 1, ARR(1), true, 1, 1, 0, 0)
+INTRINSIC(var_atomic_or, 1, ARR(1), true, 1, 1, 0, 0)
+INTRINSIC(var_atomic_xor, 1, ARR(1), true, 1, 1, 0, 0)
+INTRINSIC(var_atomic_exchange, 1, ARR(1), true, 1, 1, 0, 0)
+INTRINSIC(var_atomic_comp_swap, 2, ARR(1, 1), true, 1, 1, 0, 0)
+
+/*
* SSBO atomic intrinsics
*
* All of the SSBO atomic memory operations read a value from memory,
SYSTEM_VALUE(vertex_id_zero_base, 1, 0)
SYSTEM_VALUE(base_vertex, 1, 0)
SYSTEM_VALUE(instance_id, 1, 0)
+SYSTEM_VALUE(base_instance, 1, 0)
+SYSTEM_VALUE(draw_id, 1, 0)
SYSTEM_VALUE(sample_id, 1, 0)
SYSTEM_VALUE(sample_pos, 2, 0)
SYSTEM_VALUE(sample_mask_in, 1, 0)
LOAD(per_vertex_output, 2, 1, NIR_INTRINSIC_CAN_ELIMINATE)
/* src[] = { offset }. const_index[] = { base } */
LOAD(shared, 1, 1, NIR_INTRINSIC_CAN_ELIMINATE)
+/* src[] = { offset }. const_index[] = { base, size } */
+LOAD(push_constant, 1, 2, NIR_INTRINSIC_CAN_ELIMINATE | NIR_INTRINSIC_CAN_REORDER)
/*
* Stores work the same way as loads, except now the first source is the value
#define STORE(name, srcs, indices, flags) \
INTRINSIC(store_##name, srcs, ARR(0, 1, 1, 1), false, 0, 0, indices, flags)
-/* src[] = { value, offset }. const_index[] = { base } */
-STORE(output, 2, 1, 0)
-/* src[] = { value, vertex, offset }. const_index[] = { base } */
-STORE(per_vertex_output, 3, 1, 0)
+/* src[] = { value, offset }. const_index[] = { base, write_mask } */
+STORE(output, 2, 2, 0)
+/* src[] = { value, vertex, offset }. const_index[] = { base, write_mask } */
+STORE(per_vertex_output, 3, 2, 0)
/* src[] = { value, block_index, offset }. const_index[] = { write_mask } */
STORE(ssbo, 3, 1, 0)
/* src[] = { value, offset }. const_index[] = { base, write_mask } */
-STORE(shared, 2, 1, 0)
+STORE(shared, 2, 2, 0)
LAST_INTRINSIC(store_shared)
LOWER_REDUCTION(nir_op_bany_inequal, nir_op_ine, nir_op_ior);
LOWER_REDUCTION(nir_op_fall_equal, nir_op_seq, nir_op_fand);
LOWER_REDUCTION(nir_op_fany_nequal, nir_op_sne, nir_op_for);
- LOWER_REDUCTION(nir_op_ball, nir_op_imov, nir_op_iand);
- LOWER_REDUCTION(nir_op_bany, nir_op_imov, nir_op_ior);
- LOWER_REDUCTION(nir_op_fall, nir_op_fmov, nir_op_fand);
- LOWER_REDUCTION(nir_op_fany, nir_op_fmov, nir_op_for);
default:
break;
void
nir_lower_alu_to_scalar(nir_shader *shader)
{
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- nir_lower_alu_to_scalar_impl(overload->impl);
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ nir_lower_alu_to_scalar_impl(function->impl);
}
}
}
if (instr->variables[0]->var->data.mode != nir_var_uniform &&
- instr->variables[0]->var->data.mode != nir_var_shader_storage)
+ instr->variables[0]->var->data.mode != nir_var_shader_storage &&
+ instr->variables[0]->var->data.mode != nir_var_shared)
return; /* atomics passed as function arguments can't be lowered */
void *mem_ctx = ralloc_parent(instr);
state->shader_program->UniformStorage[uniform_loc].opaque[state->shader->stage].index;
nir_load_const_instr *offset_const = nir_load_const_instr_create(mem_ctx, 1);
- offset_const->value.u[0] = instr->variables[0]->var->data.atomic.offset;
+ offset_const->value.u[0] = instr->variables[0]->var->data.offset;
nir_instr_insert_before(&instr->instr, &offset_const->instr);
.shader_program = shader_program,
};
- nir_foreach_overload(shader, overload) {
- if (overload->impl) {
- nir_foreach_block(overload->impl, lower_block, (void *) &state);
- nir_metadata_preserve(overload->impl, nir_metadata_block_index |
+ nir_foreach_function(shader, function) {
+ if (function->impl) {
+ nir_foreach_block(function->impl, lower_block, (void *) &state);
+ nir_metadata_preserve(function->impl, nir_metadata_block_index |
nir_metadata_dominance);
}
}
store = nir_intrinsic_instr_create(b->shader, nir_intrinsic_store_output);
store->num_components = 4;
store->const_index[0] = out->data.driver_location;
+ store->const_index[1] = 0xf; /* wrmask */
store->src[0].ssa = nir_vec4(b, val[0], val[1], val[2], val[3]);
store->src[0].is_ssa = true;
store->src[1] = nir_src_for_ssa(nir_imm_int(b, 0));
.drvloc = drvloc,
};
- nir_foreach_overload(shader, overload) {
- if (overload->impl) {
- nir_foreach_block_reverse(overload->impl,
+ nir_foreach_function(shader, function) {
+ if (function->impl) {
+ nir_foreach_block_reverse(function->impl,
find_output_in_block, &state);
}
}
for (int plane = 0; plane < MAX_CLIP_PLANES; plane++) {
if (ucp_enables & (1 << plane)) {
- nir_intrinsic_instr *ucp;
-
- /* insert intrinsic to fetch ucp[plane]: */
- ucp = nir_intrinsic_instr_create(b.shader,
- nir_intrinsic_load_user_clip_plane);
- ucp->num_components = 4;
- ucp->const_index[0] = plane;
- nir_ssa_dest_init(&ucp->instr, &ucp->dest, 4, NULL);
- nir_builder_instr_insert(&b, &ucp->instr);
+ nir_ssa_def *ucp =
+ nir_load_system_value(&b, nir_intrinsic_load_user_clip_plane, plane);
/* calculate clipdist[plane] - dot(ucp, cv): */
- clipdist[plane] = nir_fdot4(&b, &ucp->dest.ssa, cv);
+ clipdist[plane] = nir_fdot4(&b, ucp, cv);
}
else {
/* 0.0 == don't-clip == disabled: */
out[1] =
create_clipdist_var(shader, ++maxloc, true, VARYING_SLOT_CLIP_DIST1);
- nir_foreach_overload(shader, overload) {
- if (!strcmp(overload->function->name, "main"))
- lower_clip_vs(overload->impl, ucp_enables, cv, out);
+ nir_foreach_function(shader, function) {
+ if (!strcmp(function->name, "main"))
+ lower_clip_vs(function->impl, ucp_enables, cv, out);
}
}
create_clipdist_var(shader, ++maxloc, false,
VARYING_SLOT_CLIP_DIST1);
- nir_foreach_overload(shader, overload) {
- if (!strcmp(overload->function->name, "main"))
- lower_clip_fs(overload->impl, ucp_enables, in);
+ nir_foreach_function(shader, function) {
+ if (!strcmp(function->name, "main"))
+ lower_clip_fs(function->impl, ucp_enables, in);
}
}
state.var_func_table = _mesa_hash_table_create(NULL, _mesa_hash_pointer,
_mesa_key_pointer_equal);
- nir_foreach_overload(shader, overload) {
- if (overload->impl) {
- state.impl = overload->impl;
- nir_foreach_block(overload->impl, mark_global_var_uses_block, &state);
+ nir_foreach_function(shader, function) {
+ if (function->impl) {
+ state.impl = function->impl;
+ nir_foreach_block(function->impl, mark_global_var_uses_block, &state);
}
}
/* Increment the vertex count by 1 */
nir_store_var(b, state->vertex_count_var,
- nir_iadd(b, count, nir_imm_int(b, 1)));
+ nir_iadd(b, count, nir_imm_int(b, 1)),
+ 0x1); /* .x */
nir_instr_remove(&intrin->instr);
exec_list_push_tail(&shader->globals, &var->node);
state.vertex_count_var = var;
- nir_foreach_overload(shader, overload) {
- if (overload->impl) {
+ nir_foreach_function(shader, function) {
+ if (function->impl) {
nir_builder b;
- nir_builder_init(&b, overload->impl);
+ nir_builder_init(&b, function->impl);
state.builder = &b;
- nir_foreach_block(overload->impl, rewrite_intrinsics, &state);
+ nir_foreach_block(function->impl, rewrite_intrinsics, &state);
/* This only works because we have a single main() function. */
- append_set_vertex_count(overload->impl->end_block, &state);
+ append_set_vertex_count(function->impl->end_block, &state);
- nir_metadata_preserve(overload->impl, 0);
+ nir_metadata_preserve(function->impl, 0);
}
}
void
nir_lower_idiv(nir_shader *shader)
{
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- convert_impl(overload->impl);
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ convert_impl(function->impl);
}
}
--- /dev/null
+/*
+ * Copyright © 2016 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include "nir.h"
+#include "nir_builder.h"
+
+static void
+emit_load_store(nir_builder *b, nir_intrinsic_instr *orig_instr,
+ nir_deref_var *deref, nir_deref *tail,
+ nir_ssa_def **dest, nir_ssa_def *src);
+
+static void
+emit_indirect_load_store(nir_builder *b, nir_intrinsic_instr *orig_instr,
+ nir_deref_var *deref, nir_deref *arr_parent,
+ int start, int end,
+ nir_ssa_def **dest, nir_ssa_def *src)
+{
+ assert(arr_parent->child &&
+ arr_parent->child->deref_type == nir_deref_type_array);
+ nir_deref_array *arr = nir_deref_as_array(arr_parent->child);
+ assert(arr->deref_array_type == nir_deref_array_type_indirect);
+ assert(arr->indirect.is_ssa);
+
+ assert(start < end);
+ if (start == end - 1) {
+ /* Base case. Just emit the load/store op */
+ nir_deref_array direct = *arr;
+ direct.deref_array_type = nir_deref_array_type_direct;
+ direct.base_offset += start;
+ direct.indirect = NIR_SRC_INIT;
+
+ arr_parent->child = &direct.deref;
+ emit_load_store(b, orig_instr, deref, &arr->deref, dest, src);
+ arr_parent->child = &arr->deref;
+ } else {
+ int mid = start + (end - start) / 2;
+
+ nir_ssa_def *then_dest, *else_dest;
+
+ nir_if *if_stmt = nir_if_create(b->shader);
+ if_stmt->condition = nir_src_for_ssa(nir_ilt(b, arr->indirect.ssa,
+ nir_imm_int(b, mid)));
+ nir_cf_node_insert(b->cursor, &if_stmt->cf_node);
+
+ b->cursor = nir_after_cf_list(&if_stmt->then_list);
+ emit_indirect_load_store(b, orig_instr, deref, arr_parent,
+ start, mid, &then_dest, src);
+
+ b->cursor = nir_after_cf_list(&if_stmt->else_list);
+ emit_indirect_load_store(b, orig_instr, deref, arr_parent,
+ mid, end, &else_dest, src);
+
+ b->cursor = nir_after_cf_node(&if_stmt->cf_node);
+
+ if (src == NULL) {
+ /* We're a load. We need to insert a phi node */
+ nir_phi_instr *phi = nir_phi_instr_create(b->shader);
+ nir_ssa_dest_init(&phi->instr, &phi->dest,
+ then_dest->num_components, NULL);
+
+ nir_phi_src *src0 = ralloc(phi, nir_phi_src);
+ src0->pred = nir_cf_node_as_block(nir_if_last_then_node(if_stmt));
+ src0->src = nir_src_for_ssa(then_dest);
+ exec_list_push_tail(&phi->srcs, &src0->node);
+
+ nir_phi_src *src1 = ralloc(phi, nir_phi_src);
+ src1->pred = nir_cf_node_as_block(nir_if_last_else_node(if_stmt));
+ src1->src = nir_src_for_ssa(else_dest);
+ exec_list_push_tail(&phi->srcs, &src1->node);
+
+ nir_builder_instr_insert(b, &phi->instr);
+ *dest = &phi->dest.ssa;
+ }
+ }
+}
+
+static void
+emit_load_store(nir_builder *b, nir_intrinsic_instr *orig_instr,
+ nir_deref_var *deref, nir_deref *tail,
+ nir_ssa_def **dest, nir_ssa_def *src)
+{
+ for (; tail->child; tail = tail->child) {
+ if (tail->child->deref_type != nir_deref_type_array)
+ continue;
+
+ nir_deref_array *arr = nir_deref_as_array(tail->child);
+ if (arr->deref_array_type != nir_deref_array_type_indirect)
+ continue;
+
+ int length = glsl_get_length(tail->type);
+
+ emit_indirect_load_store(b, orig_instr, deref, tail, -arr->base_offset,
+ length - arr->base_offset, dest, src);
+ return;
+ }
+
+ assert(tail && tail->child == NULL);
+
+ /* We reached the end of the deref chain. Emit the instruction */
+
+ if (src == NULL) {
+ /* This is a load instruction */
+ nir_intrinsic_instr *load =
+ nir_intrinsic_instr_create(b->shader, nir_intrinsic_load_var);
+ load->num_components = orig_instr->num_components;
+ load->variables[0] =
+ nir_deref_as_var(nir_copy_deref(load, &deref->deref));
+ nir_ssa_dest_init(&load->instr, &load->dest,
+ load->num_components, NULL);
+ nir_builder_instr_insert(b, &load->instr);
+ *dest = &load->dest.ssa;
+ } else {
+ /* This is a store instruction */
+ nir_intrinsic_instr *store =
+ nir_intrinsic_instr_create(b->shader, nir_intrinsic_store_var);
+ store->num_components = orig_instr->num_components;
+ store->const_index[0] = orig_instr->const_index[0]; /* writemask */
+ store->variables[0] =
+ nir_deref_as_var(nir_copy_deref(store, &deref->deref));
+ store->src[0] = nir_src_for_ssa(src);
+ nir_builder_instr_insert(b, &store->instr);
+ }
+}
+
+static bool
+deref_has_indirect(nir_deref_var *deref)
+{
+ for (nir_deref *tail = deref->deref.child; tail; tail = tail->child) {
+ if (tail->deref_type != nir_deref_type_array)
+ continue;
+
+ nir_deref_array *arr = nir_deref_as_array(tail);
+ if (arr->deref_array_type == nir_deref_array_type_indirect)
+ return true;
+ }
+
+ return false;
+}
+
+struct lower_indirect_state {
+ nir_builder builder;
+ uint32_t mode_mask;
+ bool progress;
+};
+
+static bool
+lower_indirect_block(nir_block *block, void *void_state)
+{
+ struct lower_indirect_state *state = void_state;
+
+ nir_foreach_instr_safe(block, instr) {
+ if (instr->type != nir_instr_type_intrinsic)
+ continue;
+
+ nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);
+ if (intrin->intrinsic != nir_intrinsic_load_var &&
+ intrin->intrinsic != nir_intrinsic_store_var)
+ continue;
+
+ if (!deref_has_indirect(intrin->variables[0]))
+ continue;
+
+ /* Only lower variables whose mode is in the mask */
+ if (!(state->mode_mask & (1 << intrin->variables[0]->var->data.mode)))
+ continue;
+
+ state->builder.cursor = nir_before_instr(&intrin->instr);
+
+ if (intrin->intrinsic == nir_intrinsic_load_var) {
+ nir_ssa_def *result;
+ emit_load_store(&state->builder, intrin, intrin->variables[0],
+ &intrin->variables[0]->deref, &result, NULL);
+ nir_ssa_def_rewrite_uses(&intrin->dest.ssa, nir_src_for_ssa(result));
+ } else {
+ assert(intrin->src[0].is_ssa);
+ emit_load_store(&state->builder, intrin, intrin->variables[0],
+ &intrin->variables[0]->deref, NULL, intrin->src[0].ssa);
+ }
+ nir_instr_remove(&intrin->instr);
+ state->progress = true;
+ }
+
+ return true;
+}
+
+static bool
+lower_indirects_impl(nir_function_impl *impl, uint32_t mode_mask)
+{
+ struct lower_indirect_state state;
+
+ state.progress = false;
+ state.mode_mask = mode_mask;
+ nir_builder_init(&state.builder, impl);
+
+ nir_foreach_block(impl, lower_indirect_block, &state);
+
+ if (state.progress)
+ nir_metadata_preserve(impl, nir_metadata_none);
+
+ return state.progress;
+}
+
+/** Lowers indirect variable loads/stores to direct loads/stores.
+ *
+ * The pass works by replacing any indirect load or store with an if-ladder
+ * that does a binary search on the array index.
+ */
+bool
+nir_lower_indirect_derefs(nir_shader *shader, uint32_t mode_mask)
+{
+ bool progress = false;
+
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ progress = lower_indirects_impl(function->impl, mode_mask) || progress;
+ }
+
+ return progress;
+}
case nir_var_uniform:
op = nir_intrinsic_load_uniform;
break;
+ case nir_var_shared:
+ op = nir_intrinsic_load_shared;
+ break;
default:
unreachable("Unknown variable mode");
}
return op;
}
+static nir_intrinsic_op
+store_op(struct lower_io_state *state,
+ nir_variable_mode mode, bool per_vertex)
+{
+ nir_intrinsic_op op;
+ switch (mode) {
+ case nir_var_shader_in:
+ case nir_var_shader_out:
+ op = per_vertex ? nir_intrinsic_store_per_vertex_output :
+ nir_intrinsic_store_output;
+ break;
+ case nir_var_shared:
+ op = nir_intrinsic_store_shared;
+ break;
+ default:
+ unreachable("Unknown variable mode");
+ }
+ return op;
+}
+
+static nir_intrinsic_op
+atomic_op(nir_intrinsic_op opcode)
+{
+ switch (opcode) {
+#define OP(O) case nir_intrinsic_var_##O: return nir_intrinsic_shared_##O;
+ OP(atomic_exchange)
+ OP(atomic_comp_swap)
+ OP(atomic_add)
+ OP(atomic_imin)
+ OP(atomic_umin)
+ OP(atomic_imax)
+ OP(atomic_umax)
+ OP(atomic_and)
+ OP(atomic_or)
+ OP(atomic_xor)
+#undef OP
+ default:
+ unreachable("Invalid atomic");
+ }
+}
+
static bool
nir_lower_io_block(nir_block *block, void *void_state)
{
nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);
- if (intrin->intrinsic != nir_intrinsic_load_var &&
- intrin->intrinsic != nir_intrinsic_store_var)
+ switch (intrin->intrinsic) {
+ case nir_intrinsic_load_var:
+ case nir_intrinsic_store_var:
+ case nir_intrinsic_var_atomic_add:
+ case nir_intrinsic_var_atomic_imin:
+ case nir_intrinsic_var_atomic_umin:
+ case nir_intrinsic_var_atomic_imax:
+ case nir_intrinsic_var_atomic_umax:
+ case nir_intrinsic_var_atomic_and:
+ case nir_intrinsic_var_atomic_or:
+ case nir_intrinsic_var_atomic_xor:
+ case nir_intrinsic_var_atomic_exchange:
+ case nir_intrinsic_var_atomic_comp_swap:
+ /* We can lower the io for this nir instrinsic */
+ break;
+ default:
+ /* We can't lower the io for this nir instrinsic, so skip it */
continue;
+ }
nir_variable_mode mode = intrin->variables[0]->var->data.mode;
if (mode != nir_var_shader_in &&
mode != nir_var_shader_out &&
+ mode != nir_var_shared &&
mode != nir_var_uniform)
continue;
}
case nir_intrinsic_store_var: {
- assert(mode == nir_var_shader_out);
+ assert(mode == nir_var_shader_out || mode == nir_var_shared);
nir_ssa_def *offset;
nir_ssa_def *vertex_index;
per_vertex ? &vertex_index : NULL,
state->type_size);
- nir_intrinsic_op store_op =
- per_vertex ? nir_intrinsic_store_per_vertex_output :
- nir_intrinsic_store_output;
-
- nir_intrinsic_instr *store = nir_intrinsic_instr_create(state->mem_ctx,
- store_op);
+ nir_intrinsic_instr *store =
+ nir_intrinsic_instr_create(state->mem_ctx,
+ store_op(state, mode, per_vertex));
store->num_components = intrin->num_components;
nir_src_copy(&store->src[0], &intrin->src[0], store);
store->const_index[0] =
intrin->variables[0]->var->data.driver_location;
+ /* Copy the writemask */
+ store->const_index[1] = intrin->const_index[0];
+
if (per_vertex)
store->src[1] = nir_src_for_ssa(vertex_index);
break;
}
+ case nir_intrinsic_var_atomic_add:
+ case nir_intrinsic_var_atomic_imin:
+ case nir_intrinsic_var_atomic_umin:
+ case nir_intrinsic_var_atomic_imax:
+ case nir_intrinsic_var_atomic_umax:
+ case nir_intrinsic_var_atomic_and:
+ case nir_intrinsic_var_atomic_or:
+ case nir_intrinsic_var_atomic_xor:
+ case nir_intrinsic_var_atomic_exchange:
+ case nir_intrinsic_var_atomic_comp_swap: {
+ assert(mode == nir_var_shared);
+
+ nir_ssa_def *offset;
+
+ offset = get_io_offset(b, intrin->variables[0],
+ NULL, state->type_size);
+
+ nir_intrinsic_instr *atomic =
+ nir_intrinsic_instr_create(state->mem_ctx,
+ atomic_op(intrin->intrinsic));
+
+ atomic->src[0] = nir_src_for_ssa(offset);
+
+ atomic->const_index[0] =
+ intrin->variables[0]->var->data.driver_location;
+
+ nir_src_copy(&atomic->src[1], &intrin->src[0], atomic);
+
+ if (intrin->intrinsic == nir_intrinsic_var_atomic_comp_swap)
+ nir_src_copy(&atomic->src[2], &intrin->src[1], atomic);
+
+ if (intrin->dest.is_ssa) {
+ nir_ssa_dest_init(&atomic->instr, &atomic->dest,
+ intrin->dest.ssa.num_components, NULL);
+ nir_ssa_def_rewrite_uses(&intrin->dest.ssa,
+ nir_src_for_ssa(&atomic->dest.ssa));
+ } else {
+ nir_dest_copy(&atomic->dest, &intrin->dest, state->mem_ctx);
+ }
+
+ nir_instr_insert_before(&intrin->instr, &atomic->instr);
+ nir_instr_remove(&intrin->instr);
+ break;
+ }
+
default:
break;
}
nir_lower_io(nir_shader *shader, nir_variable_mode mode,
int (*type_size)(const struct glsl_type *))
{
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- nir_lower_io_impl(overload->impl, mode, type_size);
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ nir_lower_io_impl(function->impl, mode, type_size);
}
}
case nir_intrinsic_load_output:
case nir_intrinsic_load_uniform:
return &instr->src[0];
+ case nir_intrinsic_load_ubo:
+ case nir_intrinsic_load_ssbo:
case nir_intrinsic_load_per_vertex_input:
case nir_intrinsic_load_per_vertex_output:
case nir_intrinsic_store_output:
return &instr->src[1];
+ case nir_intrinsic_store_ssbo:
case nir_intrinsic_store_per_vertex_output:
return &instr->src[2];
default:
void
nir_lower_load_const_to_scalar(nir_shader *shader)
{
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- nir_lower_load_const_to_scalar_impl(overload->impl);
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ nir_lower_load_const_to_scalar_impl(function->impl);
}
}
nir_alu_instr *mov = nir_alu_instr_create(state->shader, nir_op_imov);
nir_src_copy(&mov->src[0].src, &intrin->src[0], mov);
- mov->dest.write_mask = (1 << intrin->num_components) - 1;
+ mov->dest.write_mask = intrin->const_index[0];
mov->dest.dest.is_ssa = false;
mov->dest.dest.reg.reg = reg_src.reg.reg;
mov->dest.dest.reg.base_offset = reg_src.reg.base_offset;
{
struct locals_to_regs_state state;
- state.shader = impl->overload->function->shader;
+ state.shader = impl->function->shader;
state.impl = impl;
state.progress = false;
state.regs_table = _mesa_hash_table_create(NULL, hash_deref, derefs_equal);
{
bool progress = false;
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- progress = nir_lower_locals_to_regs_impl(overload->impl) || progress;
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ progress = nir_lower_locals_to_regs_impl(function->impl) || progress;
}
return progress;
}
void
-nir_lower_outputs_to_temporaries(nir_shader *shader)
+nir_lower_outputs_to_temporaries(nir_shader *shader, nir_function *entrypoint)
{
struct lower_outputs_state state;
/* Reparent the name to the new variable */
ralloc_steal(output, output->name);
+ /* Reparent the constant initializer (if any) */
+ ralloc_steal(output, output->constant_initializer);
+
/* Give the output a new name with @out-temp appended */
temp->name = ralloc_asprintf(var, "%s@out-temp", output->name);
temp->data.mode = nir_var_global;
exec_list_push_tail(&shader->outputs, &output->node);
}
- nir_foreach_overload(shader, overload) {
- if (overload->impl == NULL)
+ nir_foreach_function(shader, function) {
+ if (function->impl == NULL)
continue;
if (shader->stage == MESA_SHADER_GEOMETRY) {
/* For geometry shaders, we have to emit the output copies right
* before each EmitVertex call.
*/
- nir_foreach_block(overload->impl, emit_output_copies_block, &state);
- } else if (strcmp(overload->function->name, "main") == 0) {
+ nir_foreach_block(function->impl, emit_output_copies_block, &state);
+ } else if (function == entrypoint) {
/* For all other shader types, we need to do the copies right before
* the jumps to the end block.
*/
struct set_entry *block_entry;
- set_foreach(overload->impl->end_block->predecessors, block_entry) {
+ set_foreach(function->impl->end_block->predecessors, block_entry) {
struct nir_block *block = (void *)block_entry->key;
emit_output_copies(nir_after_block_before_jump(block), &state);
}
}
- nir_metadata_preserve(overload->impl, nir_metadata_block_index |
+ nir_metadata_preserve(function->impl, nir_metadata_block_index |
nir_metadata_dominance);
}
void
nir_lower_phis_to_scalar(nir_shader *shader)
{
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- lower_phis_to_scalar_impl(overload->impl);
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ lower_phis_to_scalar_impl(function->impl);
}
}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include "nir.h"
+#include "nir_builder.h"
+#include "nir_control_flow.h"
+
+struct lower_returns_state {
+ nir_builder builder;
+ struct exec_list *cf_list;
+ nir_loop *loop;
+ nir_variable *return_flag;
+};
+
+static bool lower_returns_in_cf_list(struct exec_list *cf_list,
+ struct lower_returns_state *state);
+
+static void
+predicate_following(nir_cf_node *node, struct lower_returns_state *state)
+{
+ nir_builder *b = &state->builder;
+ b->cursor = nir_after_cf_node_and_phis(node);
+
+ if (nir_cursors_equal(b->cursor, nir_after_cf_list(state->cf_list)))
+ return; /* Nothing to predicate */
+
+ assert(state->return_flag);
+
+ nir_if *if_stmt = nir_if_create(b->shader);
+ if_stmt->condition = nir_src_for_ssa(nir_load_var(b, state->return_flag));
+ nir_cf_node_insert(b->cursor, &if_stmt->cf_node);
+
+ if (state->loop) {
+ /* If we're inside of a loop, then all we need to do is insert a
+ * conditional break.
+ */
+ nir_jump_instr *brk =
+ nir_jump_instr_create(state->builder.shader, nir_jump_break);
+ nir_instr_insert(nir_before_cf_list(&if_stmt->then_list), &brk->instr);
+ } else {
+ /* Otherwise, we need to actually move everything into the else case
+ * of the if statement.
+ */
+ nir_cf_list list;
+ nir_cf_extract(&list, nir_after_cf_node(&if_stmt->cf_node),
+ nir_after_cf_list(state->cf_list));
+ assert(!exec_list_is_empty(&list.list));
+ nir_cf_reinsert(&list, nir_before_cf_list(&if_stmt->else_list));
+ }
+}
+
+static bool
+lower_returns_in_loop(nir_loop *loop, struct lower_returns_state *state)
+{
+ nir_loop *parent = state->loop;
+ state->loop = loop;
+ bool progress = lower_returns_in_cf_list(&loop->body, state);
+ state->loop = parent;
+
+ /* If the recursive call made progress, then there were returns inside
+ * of the loop. These would have been lowered to breaks with the return
+ * flag set to true. We need to predicate everything following the loop
+ * on the return flag.
+ */
+ if (progress)
+ predicate_following(&loop->cf_node, state);
+
+ return progress;
+}
+
+static bool
+lower_returns_in_if(nir_if *if_stmt, struct lower_returns_state *state)
+{
+ bool progress;
+
+ progress = lower_returns_in_cf_list(&if_stmt->then_list, state);
+ progress = lower_returns_in_cf_list(&if_stmt->else_list, state) || progress;
+
+ /* If either of the recursive calls made progress, then there were
+ * returns inside of the body of the if. If we're in a loop, then these
+ * were lowered to breaks which automatically skip to the end of the
+ * loop so we don't have to do anything. If we're not in a loop, then
+ * all we know is that the return flag is set appropreately and that the
+ * recursive calls ensured that nothing gets executed *inside* the if
+ * after a return. In order to ensure nothing outside gets executed
+ * after a return, we need to predicate everything following on the
+ * return flag.
+ */
+ if (progress && !state->loop)
+ predicate_following(&if_stmt->cf_node, state);
+
+ return progress;
+}
+
+static bool
+lower_returns_in_block(nir_block *block, struct lower_returns_state *state)
+{
+ if (block->predecessors->entries == 0 &&
+ block != nir_start_block(state->builder.impl)) {
+ /* This block is unreachable. Delete it and everything after it. */
+ nir_cf_list list;
+ nir_cf_extract(&list, nir_before_cf_node(&block->cf_node),
+ nir_after_cf_list(state->cf_list));
+
+ if (exec_list_is_empty(&list.list)) {
+ /* There's nothing here, which also means there's nothing in this
+ * block so we have nothing to do.
+ */
+ return false;
+ } else {
+ nir_cf_delete(&list);
+ return true;
+ }
+ }
+
+ nir_instr *last_instr = nir_block_last_instr(block);
+ if (last_instr == NULL)
+ return false;
+
+ if (last_instr->type != nir_instr_type_jump)
+ return false;
+
+ nir_jump_instr *jump = nir_instr_as_jump(last_instr);
+ if (jump->type != nir_jump_return)
+ return false;
+
+ nir_instr_remove(&jump->instr);
+
+ nir_builder *b = &state->builder;
+ b->cursor = nir_after_block(block);
+
+ /* Set the return flag */
+ if (state->return_flag == NULL) {
+ state->return_flag =
+ nir_local_variable_create(b->impl, glsl_bool_type(), "return");
+
+ /* Set a default value of false */
+ state->return_flag->constant_initializer =
+ rzalloc(state->return_flag, nir_constant);
+ }
+ nir_store_var(b, state->return_flag, nir_imm_int(b, NIR_TRUE), 1);
+
+ if (state->loop) {
+ /* We're in a loop; we need to break out of it. */
+ nir_jump(b, nir_jump_break);
+ } else {
+ /* Not in a loop; we'll deal with predicating later*/
+ assert(nir_cf_node_next(&block->cf_node) == NULL);
+ }
+
+ return true;
+}
+
+static bool
+lower_returns_in_cf_list(struct exec_list *cf_list,
+ struct lower_returns_state *state)
+{
+ bool progress = false;
+
+ struct exec_list *parent_list = state->cf_list;
+ state->cf_list = cf_list;
+
+ /* We iterate over the list backwards because any given lower call may
+ * take everything following the given CF node and predicate it. In
+ * order to avoid recursion/iteration problems, we want everything after
+ * a given node to already be lowered before this happens.
+ */
+ foreach_list_typed_reverse_safe(nir_cf_node, node, node, cf_list) {
+ switch (node->type) {
+ case nir_cf_node_block:
+ if (lower_returns_in_block(nir_cf_node_as_block(node), state))
+ progress = true;
+ break;
+
+ case nir_cf_node_if:
+ if (lower_returns_in_if(nir_cf_node_as_if(node), state))
+ progress = true;
+ break;
+
+ case nir_cf_node_loop:
+ if (lower_returns_in_loop(nir_cf_node_as_loop(node), state))
+ progress = true;
+ break;
+
+ default:
+ unreachable("Invalid inner CF node type");
+ }
+ }
+
+ state->cf_list = parent_list;
+
+ return progress;
+}
+
+bool
+nir_lower_returns_impl(nir_function_impl *impl)
+{
+ struct lower_returns_state state;
+
+ state.cf_list = &impl->body;
+ state.loop = NULL;
+ state.return_flag = NULL;
+ nir_builder_init(&state.builder, impl);
+
+ bool progress = lower_returns_in_cf_list(&impl->body, &state);
+
+ if (progress) {
+ nir_metadata_preserve(impl, nir_metadata_none);
+ nir_repair_ssa_impl(impl);
+ }
+
+ return progress;
+}
+
+bool
+nir_lower_returns(nir_shader *shader)
+{
+ bool progress = false;
+
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ progress = nir_lower_returns_impl(function->impl) || progress;
+ }
+
+ return progress;
+}
if (instr->sampler == NULL)
return;
+ /* GLSL only has combined textures/samplers */
+ assert(instr->texture == NULL);
+
instr->sampler_index = 0;
unsigned location = instr->sampler->var->data.location;
unsigned array_elements = 1;
if (indirect) {
/* First, we have to resize the array of texture sources */
nir_tex_src *new_srcs = rzalloc_array(instr, nir_tex_src,
- instr->num_srcs + 1);
+ instr->num_srcs + 2);
for (unsigned i = 0; i < instr->num_srcs; i++) {
new_srcs[i].src_type = instr->src[i].src_type;
/* Now we can go ahead and move the source over to being a
* first-class texture source.
*/
+ instr->src[instr->num_srcs].src_type = nir_tex_src_texture_offset;
+ instr->num_srcs++;
+ nir_instr_rewrite_src(&instr->instr,
+ &instr->src[instr->num_srcs - 1].src,
+ nir_src_for_ssa(indirect));
+
instr->src[instr->num_srcs].src_type = nir_tex_src_sampler_offset;
instr->num_srcs++;
nir_instr_rewrite_src(&instr->instr,
&instr->src[instr->num_srcs - 1].src,
nir_src_for_ssa(indirect));
- instr->sampler_array_size = array_elements;
+ instr->texture_array_size = array_elements;
}
if (location > shader_program->NumUniformStorage - 1 ||
shader_program->UniformStorage[location].opaque[stage].index;
instr->sampler = NULL;
+
+ instr->texture_index = instr->sampler_index;
}
typedef struct {
nir_lower_samplers(nir_shader *shader,
const struct gl_shader_program *shader_program)
{
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- lower_impl(overload->impl, shader_program, shader->stage);
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ lower_impl(function->impl, shader_program, shader->stage);
}
}
--- /dev/null
+/*
+ * Copyright (C) 2005-2007 Brian Paul All Rights Reserved.
+ * Copyright (C) 2008 VMware, Inc. All Rights Reserved.
+ * Copyright © 2014 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include "nir.h"
+#include "../program.h"
+#include "program/hash_table.h"
+#include "ir_uniform.h"
+
+extern "C" {
+#include "main/compiler.h"
+#include "main/mtypes.h"
+#include "program/prog_parameter.h"
+#include "program/program.h"
+}
+
+static void
+add_indirect_to_tex(nir_tex_instr *instr, nir_src indirect)
+{
+ /* First, we have to resize the array of texture sources */
+ nir_tex_src *new_srcs = rzalloc_array(instr, nir_tex_src,
+ instr->num_srcs + 1);
+
+ for (unsigned i = 0; i < instr->num_srcs; i++) {
+ new_srcs[i].src_type = instr->src[i].src_type;
+ nir_instr_move_src(&instr->instr, &new_srcs[i].src, &instr->src[i].src);
+ }
+
+ ralloc_free(instr->src);
+ instr->src = new_srcs;
+
+ /* Now we can go ahead and move the source over to being a
+ * first-class texture source.
+ */
+ instr->src[instr->num_srcs].src_type = nir_tex_src_sampler_offset;
+ instr->num_srcs++;
+ nir_instr_rewrite_src(&instr->instr, &instr->src[instr->num_srcs - 1].src,
+ indirect);
+}
+
+static unsigned
+get_sampler_index(const struct gl_shader_program *shader_program,
+ gl_shader_stage stage, const char *name)
+{
+ unsigned location;
+ if (!shader_program->UniformHash->get(location, name)) {
+ assert(!"failed to find sampler");
+ return 0;
+ }
+
+ if (!shader_program->UniformStorage[location].sampler[stage].active) {
+ assert(!"cannot return a sampler");
+ return 0;
+ }
+
+ return shader_program->UniformStorage[location].sampler[stage].index;
+}
+
+static void
+lower_sampler(nir_tex_instr *instr, const struct gl_shader_program *shader_program,
+ gl_shader_stage stage, void *mem_ctx)
+{
+ if (instr->sampler == NULL)
+ return;
+
+ /* Get the name and the offset */
+ instr->sampler_index = 0;
+ char *name = ralloc_strdup(mem_ctx, instr->sampler->var->name);
+
+ for (nir_deref *deref = &instr->sampler->deref;
+ deref->child; deref = deref->child) {
+ switch (deref->child->deref_type) {
+ case nir_deref_type_array: {
+ nir_deref_array *deref_array = nir_deref_as_array(deref->child);
+
+ assert(deref_array->deref_array_type != nir_deref_array_type_wildcard);
+
+ if (deref_array->deref.child) {
+ ralloc_asprintf_append(&name, "[%u]",
+ deref_array->deref_array_type == nir_deref_array_type_direct ?
+ deref_array->base_offset : 0);
+ } else {
+ assert(deref->child->type->base_type == GLSL_TYPE_SAMPLER);
+ instr->sampler_index = deref_array->base_offset;
+ }
+
+ /* XXX: We're assuming here that the indirect is the last array
+ * thing we have. This should be ok for now as we don't support
+ * arrays_of_arrays yet.
+ */
+ if (deref_array->deref_array_type == nir_deref_array_type_indirect) {
+ /* First, we have to resize the array of texture sources */
+ nir_tex_src *new_srcs = rzalloc_array(instr, nir_tex_src,
+ instr->num_srcs + 1);
+
+ for (unsigned i = 0; i < instr->num_srcs; i++) {
+ new_srcs[i].src_type = instr->src[i].src_type;
+ nir_instr_move_src(&instr->instr, &new_srcs[i].src,
+ &instr->src[i].src);
+ }
+
+ ralloc_free(instr->src);
+ instr->src = new_srcs;
+
+ /* Now we can go ahead and move the source over to being a
+ * first-class texture source.
+ */
+ instr->src[instr->num_srcs].src_type = nir_tex_src_sampler_offset;
+ instr->num_srcs++;
+ nir_instr_move_src(&instr->instr,
+ &instr->src[instr->num_srcs - 1].src,
+ &deref_array->indirect);
+
+ instr->sampler_array_size = glsl_get_length(deref->type);
+ }
+ break;
+ }
+
+ case nir_deref_type_struct: {
+ nir_deref_struct *deref_struct = nir_deref_as_struct(deref->child);
+ const char *field = glsl_get_struct_elem_name(deref->type,
+ deref_struct->index);
+ ralloc_asprintf_append(&name, ".%s", field);
+ break;
+ }
+
+ default:
+ unreachable("Invalid deref type");
+ break;
+ }
+ }
+
+ instr->sampler_index += get_sampler_index(shader_program, stage, name);
+
+ instr->sampler = NULL;
+}
+
+typedef struct {
+ void *mem_ctx;
+ const struct gl_shader_program *shader_program;
+ gl_shader_stage stage;
+} lower_state;
+
+static bool
+lower_block_cb(nir_block *block, void *_state)
+{
+ lower_state *state = (lower_state *) _state;
+
+ nir_foreach_instr(block, instr) {
+ if (instr->type == nir_instr_type_tex) {
+ nir_tex_instr *tex_instr = nir_instr_as_tex(instr);
+ lower_sampler(tex_instr, state->shader_program, state->stage,
+ state->mem_ctx);
+ }
+ }
+
+ return true;
+}
+
+static void
+lower_impl(nir_function_impl *impl, const struct gl_shader_program *shader_program,
+ gl_shader_stage stage)
+{
+ lower_state state;
+
+ state.mem_ctx = ralloc_parent(impl);
+ state.shader_program = shader_program;
+ state.stage = stage;
+
+ nir_foreach_block(impl, lower_block_cb, &state);
+}
+
+extern "C" void
+nir_lower_samplers(nir_shader *shader,
+ const struct gl_shader_program *shader_program)
+{
+ nir_foreach_overload(shader, overload) {
+ if (overload->impl)
+ lower_impl(overload->impl, shader_program, shader->stage);
+ }
+}
+
+static bool
+lower_samplers_for_vk_block(nir_block *block, void *data)
+{
+ nir_foreach_instr(block, instr) {
+ if (instr->type != nir_instr_type_tex)
+ continue;
+
+ nir_tex_instr *tex = nir_instr_as_tex(instr);
+
+ assert(tex->sampler);
+
+ tex->sampler_set = tex->sampler->var->data.descriptor_set;
+ tex->sampler_index = tex->sampler->var->data.binding;
+
+ if (tex->sampler->deref.child) {
+ assert(tex->sampler->deref.child->deref_type == nir_deref_type_array);
+ nir_deref_array *arr = nir_deref_as_array(tex->sampler->deref.child);
+
+ /* Only one-level arrays are allowed in vulkan */
+ assert(arr->deref.child == NULL);
+
+ tex->sampler_index += arr->base_offset;
+ if (arr->deref_array_type == nir_deref_array_type_indirect) {
+ add_indirect_to_tex(tex, arr->indirect);
+ nir_instr_rewrite_src(instr, &arr->indirect, NIR_SRC_INIT);
+
+ tex->sampler_array_size = glsl_get_length(tex->sampler->deref.type);
+ }
+ }
+
+ tex->sampler = NULL;
+ }
+
+ return true;
+}
+
+extern "C" void
+nir_lower_samplers_for_vk(nir_shader *shader)
+{
+ nir_foreach_overload(shader, overload) {
+ if (overload->impl) {
+ nir_foreach_block(overload->impl, lower_samplers_for_vk_block, NULL);
+ }
+ }
+}
*/
#include "nir.h"
-#include "main/mtypes.h"
+#include "nir_builder.h"
-static bool
-convert_instr(nir_intrinsic_instr *instr)
-{
- if (instr->intrinsic != nir_intrinsic_load_var)
- return false;
-
- nir_variable *var = instr->variables[0]->var;
- if (var->data.mode != nir_var_system_value)
- return false;
-
- void *mem_ctx = ralloc_parent(instr);
-
- nir_intrinsic_op op = nir_intrinsic_from_system_value(var->data.location);
- nir_intrinsic_instr *new_instr = nir_intrinsic_instr_create(mem_ctx, op);
-
- if (instr->dest.is_ssa) {
- nir_ssa_dest_init(&new_instr->instr, &new_instr->dest,
- instr->dest.ssa.num_components, NULL);
- nir_ssa_def_rewrite_uses(&instr->dest.ssa,
- nir_src_for_ssa(&new_instr->dest.ssa));
- } else {
- nir_dest_copy(&new_instr->dest, &instr->dest, mem_ctx);
- }
-
- nir_instr_insert_before(&instr->instr, &new_instr->instr);
- nir_instr_remove(&instr->instr);
-
- return true;
-}
+struct lower_system_values_state {
+ nir_builder builder;
+ bool progress;
+};
static bool
-convert_block(nir_block *block, void *state)
+convert_block(nir_block *block, void *void_state)
{
- bool *progress = state;
+ struct lower_system_values_state *state = void_state;
+
+ nir_builder *b = &state->builder;
nir_foreach_instr_safe(block, instr) {
- if (instr->type == nir_instr_type_intrinsic)
- *progress = convert_instr(nir_instr_as_intrinsic(instr)) || *progress;
+ if (instr->type != nir_instr_type_intrinsic)
+ continue;
+
+ nir_intrinsic_instr *load_var = nir_instr_as_intrinsic(instr);
+
+ if (load_var->intrinsic != nir_intrinsic_load_var)
+ continue;
+
+ nir_variable *var = load_var->variables[0]->var;
+ if (var->data.mode != nir_var_system_value)
+ continue;
+
+ b->cursor = nir_after_instr(&load_var->instr);
+
+ nir_ssa_def *sysval;
+ switch (var->data.location) {
+ case SYSTEM_VALUE_GLOBAL_INVOCATION_ID: {
+ /* From the GLSL man page for gl_GlobalInvocationID:
+ *
+ * "The value of gl_GlobalInvocationID is equal to
+ * gl_WorkGroupID * gl_WorkGroupSize + gl_LocalInvocationID"
+ */
+
+ nir_const_value local_size;
+ local_size.u[0] = b->shader->info.cs.local_size[0];
+ local_size.u[1] = b->shader->info.cs.local_size[1];
+ local_size.u[2] = b->shader->info.cs.local_size[2];
+
+ nir_ssa_def *group_id =
+ nir_load_system_value(b, nir_intrinsic_load_work_group_id, 0);
+ nir_ssa_def *local_id =
+ nir_load_system_value(b, nir_intrinsic_load_local_invocation_id, 0);
+
+ sysval = nir_iadd(b, nir_imul(b, group_id,
+ nir_build_imm(b, 3, local_size)),
+ local_id);
+ break;
+ }
+
+ case SYSTEM_VALUE_LOCAL_INVOCATION_INDEX: {
+ /* From the GLSL man page for gl_LocalInvocationIndex:
+ *
+ * ?The value of gl_LocalInvocationIndex is equal to
+ * gl_LocalInvocationID.z * gl_WorkGroupSize.x *
+ * gl_WorkGroupSize.y + gl_LocalInvocationID.y *
+ * gl_WorkGroupSize.x + gl_LocalInvocationID.x"
+ */
+ nir_ssa_def *local_id =
+ nir_load_system_value(b, nir_intrinsic_load_invocation_id, 0);
+
+ unsigned stride_y = b->shader->info.cs.local_size[0];
+ unsigned stride_z = b->shader->info.cs.local_size[0] *
+ b->shader->info.cs.local_size[1];
+
+ sysval = nir_iadd(b, nir_imul(b, nir_channel(b, local_id, 2),
+ nir_imm_int(b, stride_z)),
+ nir_iadd(b, nir_imul(b, nir_channel(b, local_id, 1),
+ nir_imm_int(b, stride_y)),
+ nir_channel(b, local_id, 0)));
+ break;
+ }
+
+ case SYSTEM_VALUE_VERTEX_ID:
+ if (b->shader->options->vertex_id_zero_based) {
+ sysval = nir_iadd(b,
+ nir_load_system_value(b, nir_intrinsic_load_vertex_id_zero_base, 0),
+ nir_load_system_value(b, nir_intrinsic_load_base_vertex, 0));
+ } else {
+ sysval = nir_load_system_value(b, nir_intrinsic_load_vertex_id, 0);
+ }
+ break;
+
+ default: {
+ nir_intrinsic_op sysval_op =
+ nir_intrinsic_from_system_value(var->data.location);
+ sysval = nir_load_system_value(b, sysval_op, 0);
+ break;
+ } /* default */
+ }
+
+ nir_ssa_def_rewrite_uses(&load_var->dest.ssa, nir_src_for_ssa(sysval));
+ nir_instr_remove(&load_var->instr);
+
+ state->progress = true;
}
return true;
static bool
convert_impl(nir_function_impl *impl)
{
- bool progress = false;
+ struct lower_system_values_state state;
+
+ state.progress = false;
+ nir_builder_init(&state.builder, impl);
- nir_foreach_block(impl, convert_block, &progress);
+ nir_foreach_block(impl, convert_block, &state);
nir_metadata_preserve(impl, nir_metadata_block_index |
nir_metadata_dominance);
- return progress;
+ return state.progress;
}
bool
{
bool progress = false;
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- progress = convert_impl(overload->impl) || progress;
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ progress = convert_impl(function->impl) || progress;
}
exec_list_make_empty(&shader->system_values);
state.options = options;
state.progress = false;
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- nir_lower_tex_impl(overload->impl, &state);
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ nir_lower_tex_impl(function->impl, &state);
}
return state.progress;
void
nir_lower_to_source_mods(nir_shader *shader)
{
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- nir_lower_to_source_mods_impl(overload->impl);
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ nir_lower_to_source_mods_impl(function->impl);
}
}
if (setup_inputs(&state) != 0)
return;
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- nir_lower_two_sided_color_impl(overload->impl, &state);
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ nir_lower_two_sided_color_impl(function->impl, &state);
}
}
nir_intrinsic_instr *store =
nir_intrinsic_instr_create(mem_ctx, nir_intrinsic_store_var);
store->num_components = num_components;
+ store->const_index[0] = (1 << num_components) - 1;
store->variables[0] = nir_deref_as_var(nir_copy_deref(store, &dest_head->deref));
store->src[0].is_ssa = true;
void
nir_lower_var_copies(nir_shader *shader)
{
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- lower_var_copies_impl(overload->impl);
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ lower_var_copies_impl(function->impl);
}
}
*/
#include "nir.h"
+#include "nir_builder.h"
+#include "nir_phi_builder.h"
#include "nir_vla.h"
struct set *stores;
struct set *copies;
- nir_ssa_def **def_stack;
- nir_ssa_def **def_stack_tail;
+ struct nir_phi_builder_value *pb_value;
struct deref_node *wildcard;
struct deref_node *indirect;
*/
bool add_to_direct_deref_nodes;
- /* A hash table mapping phi nodes to deref_state data */
- struct hash_table *phi_table;
+ struct nir_phi_builder *phi_builder;
};
static struct deref_node *
return true;
}
-/** Pushes an SSA def onto the def stack for the given node
- *
- * Each node is potentially associated with a stack of SSA definitions.
- * This stack is used for determining what SSA definition reaches a given
- * point in the program for variable renaming. The stack is always kept in
- * dominance-order with at most one SSA def per block. If the SSA
- * definition on the top of the stack is in the same block as the one being
- * pushed, the top element is replaced.
- */
-static void
-def_stack_push(struct deref_node *node, nir_ssa_def *def,
- struct lower_variables_state *state)
-{
- if (node->def_stack == NULL) {
- node->def_stack = ralloc_array(state->dead_ctx, nir_ssa_def *,
- state->impl->num_blocks);
- node->def_stack_tail = node->def_stack - 1;
- }
-
- if (node->def_stack_tail >= node->def_stack) {
- nir_ssa_def *top_def = *node->def_stack_tail;
-
- if (def->parent_instr->block == top_def->parent_instr->block) {
- /* They're in the same block, just replace the top */
- *node->def_stack_tail = def;
- return;
- }
- }
-
- *(++node->def_stack_tail) = def;
-}
-
-/* Pop the top of the def stack if it's in the given block */
-static void
-def_stack_pop_if_in_block(struct deref_node *node, nir_block *block)
-{
- /* If we're popping, then we have presumably pushed at some time in the
- * past so this should exist.
- */
- assert(node->def_stack != NULL);
-
- /* The stack is already empty. Do nothing. */
- if (node->def_stack_tail < node->def_stack)
- return;
-
- nir_ssa_def *def = *node->def_stack_tail;
- if (def->parent_instr->block == block)
- node->def_stack_tail--;
-}
-
-/** Retrieves the SSA definition on the top of the stack for the given
- * node, if one exists. If the stack is empty, then we return the constant
- * initializer (if it exists) or an SSA undef.
- */
-static nir_ssa_def *
-get_ssa_def_for_block(struct deref_node *node, nir_block *block,
- struct lower_variables_state *state)
-{
- /* If we have something on the stack, go ahead and return it. We're
- * assuming that the top of the stack dominates the given block.
- */
- if (node->def_stack && node->def_stack_tail >= node->def_stack)
- return *node->def_stack_tail;
-
- /* If we got here then we don't have a definition that dominates the
- * given block. This means that we need to add an undef and use that.
- */
- nir_ssa_undef_instr *undef =
- nir_ssa_undef_instr_create(state->shader,
- glsl_get_vector_elements(node->type));
- nir_instr_insert_before_cf_list(&state->impl->body, &undef->instr);
- def_stack_push(node, &undef->def, state);
- return &undef->def;
-}
-
-/* Given a block and one of its predecessors, this function fills in the
- * souces of the phi nodes to take SSA defs from the given predecessor.
- * This function must be called exactly once per block/predecessor pair.
- */
-static void
-add_phi_sources(nir_block *block, nir_block *pred,
- struct lower_variables_state *state)
-{
- nir_foreach_instr(block, instr) {
- if (instr->type != nir_instr_type_phi)
- break;
-
- nir_phi_instr *phi = nir_instr_as_phi(instr);
-
- struct hash_entry *entry =
- _mesa_hash_table_search(state->phi_table, phi);
- if (!entry)
- continue;
-
- struct deref_node *node = entry->data;
-
- nir_phi_src *src = ralloc(phi, nir_phi_src);
- src->pred = pred;
- src->src.parent_instr = &phi->instr;
- src->src.is_ssa = true;
- src->src.ssa = get_ssa_def_for_block(node, pred, state);
-
- list_addtail(&src->src.use_link, &src->src.ssa->uses);
-
- exec_list_push_tail(&phi->srcs, &src->node);
- }
-}
-
/* Performs variable renaming by doing a DFS of the dominance tree
*
* This algorithm is very similar to the one outlined in "Efficiently
static bool
rename_variables_block(nir_block *block, struct lower_variables_state *state)
{
- nir_foreach_instr_safe(block, instr) {
- if (instr->type == nir_instr_type_phi) {
- nir_phi_instr *phi = nir_instr_as_phi(instr);
-
- struct hash_entry *entry =
- _mesa_hash_table_search(state->phi_table, phi);
-
- /* This can happen if we already have phi nodes in the program
- * that were not created in this pass.
- */
- if (!entry)
- continue;
-
- struct deref_node *node = entry->data;
-
- def_stack_push(node, &phi->dest.ssa, state);
- } else if (instr->type == nir_instr_type_intrinsic) {
- nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);
-
- switch (intrin->intrinsic) {
- case nir_intrinsic_load_var: {
- struct deref_node *node =
- get_deref_node(intrin->variables[0], state);
-
- if (node == NULL) {
- /* If we hit this path then we are referencing an invalid
- * value. Most likely, we unrolled something and are
- * reading past the end of some array. In any case, this
- * should result in an undefined value.
- */
- nir_ssa_undef_instr *undef =
- nir_ssa_undef_instr_create(state->shader,
- intrin->num_components);
-
- nir_instr_insert_before(&intrin->instr, &undef->instr);
- nir_instr_remove(&intrin->instr);
-
- nir_ssa_def_rewrite_uses(&intrin->dest.ssa,
- nir_src_for_ssa(&undef->def));
- continue;
- }
+ nir_builder b;
+ nir_builder_init(&b, state->impl);
- if (!node->lower_to_ssa)
- continue;
-
- nir_alu_instr *mov = nir_alu_instr_create(state->shader,
- nir_op_imov);
- mov->src[0].src.is_ssa = true;
- mov->src[0].src.ssa = get_ssa_def_for_block(node, block, state);
- for (unsigned i = intrin->num_components; i < 4; i++)
- mov->src[0].swizzle[i] = 0;
+ nir_foreach_instr_safe(block, instr) {
+ if (instr->type != nir_instr_type_intrinsic)
+ continue;
- assert(intrin->dest.is_ssa);
+ nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);
- mov->dest.write_mask = (1 << intrin->num_components) - 1;
- nir_ssa_dest_init(&mov->instr, &mov->dest.dest,
- intrin->num_components, NULL);
+ switch (intrin->intrinsic) {
+ case nir_intrinsic_load_var: {
+ struct deref_node *node =
+ get_deref_node(intrin->variables[0], state);
+
+ if (node == NULL) {
+ /* If we hit this path then we are referencing an invalid
+ * value. Most likely, we unrolled something and are
+ * reading past the end of some array. In any case, this
+ * should result in an undefined value.
+ */
+ nir_ssa_undef_instr *undef =
+ nir_ssa_undef_instr_create(state->shader,
+ intrin->num_components);
- nir_instr_insert_before(&intrin->instr, &mov->instr);
+ nir_instr_insert_before(&intrin->instr, &undef->instr);
nir_instr_remove(&intrin->instr);
nir_ssa_def_rewrite_uses(&intrin->dest.ssa,
- nir_src_for_ssa(&mov->dest.dest.ssa));
- break;
+ nir_src_for_ssa(&undef->def));
+ continue;
}
- case nir_intrinsic_store_var: {
- struct deref_node *node =
- get_deref_node(intrin->variables[0], state);
-
- if (node == NULL) {
- /* Probably an out-of-bounds array store. That should be a
- * no-op. */
- nir_instr_remove(&intrin->instr);
- continue;
- }
-
- if (!node->lower_to_ssa)
- continue;
-
- assert(intrin->num_components ==
- glsl_get_vector_elements(node->type));
-
- assert(intrin->src[0].is_ssa);
-
- nir_alu_instr *mov = nir_alu_instr_create(state->shader,
- nir_op_imov);
- mov->src[0].src.is_ssa = true;
- mov->src[0].src.ssa = intrin->src[0].ssa;
- for (unsigned i = intrin->num_components; i < 4; i++)
- mov->src[0].swizzle[i] = 0;
+ if (!node->lower_to_ssa)
+ continue;
- mov->dest.write_mask = (1 << intrin->num_components) - 1;
- nir_ssa_dest_init(&mov->instr, &mov->dest.dest,
- intrin->num_components, NULL);
+ nir_alu_instr *mov = nir_alu_instr_create(state->shader,
+ nir_op_imov);
+ mov->src[0].src = nir_src_for_ssa(
+ nir_phi_builder_value_get_block_def(node->pb_value, block));
+ for (unsigned i = intrin->num_components; i < 4; i++)
+ mov->src[0].swizzle[i] = 0;
- nir_instr_insert_before(&intrin->instr, &mov->instr);
+ assert(intrin->dest.is_ssa);
- def_stack_push(node, &mov->dest.dest.ssa, state);
+ mov->dest.write_mask = (1 << intrin->num_components) - 1;
+ nir_ssa_dest_init(&mov->instr, &mov->dest.dest,
+ intrin->num_components, NULL);
- /* We'll wait to remove the instruction until the next pass
- * where we pop the node we just pushed back off the stack.
- */
- break;
- }
+ nir_instr_insert_before(&intrin->instr, &mov->instr);
+ nir_instr_remove(&intrin->instr);
- default:
- break;
- }
+ nir_ssa_def_rewrite_uses(&intrin->dest.ssa,
+ nir_src_for_ssa(&mov->dest.dest.ssa));
+ break;
}
- }
- if (block->successors[0])
- add_phi_sources(block->successors[0], block, state);
- if (block->successors[1])
- add_phi_sources(block->successors[1], block, state);
+ case nir_intrinsic_store_var: {
+ struct deref_node *node =
+ get_deref_node(intrin->variables[0], state);
- for (unsigned i = 0; i < block->num_dom_children; ++i)
- rename_variables_block(block->dom_children[i], state);
-
- /* Now we iterate over the instructions and pop off any SSA defs that we
- * pushed in the first loop.
- */
- nir_foreach_instr_safe(block, instr) {
- if (instr->type == nir_instr_type_phi) {
- nir_phi_instr *phi = nir_instr_as_phi(instr);
-
- struct hash_entry *entry =
- _mesa_hash_table_search(state->phi_table, phi);
-
- /* This can happen if we already have phi nodes in the program
- * that were not created in this pass.
- */
- if (!entry)
- continue;
-
- struct deref_node *node = entry->data;
-
- def_stack_pop_if_in_block(node, block);
- } else if (instr->type == nir_instr_type_intrinsic) {
- nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);
-
- if (intrin->intrinsic != nir_intrinsic_store_var)
- continue;
-
- struct deref_node *node = get_deref_node(intrin->variables[0], state);
- if (!node)
+ if (node == NULL) {
+ /* Probably an out-of-bounds array store. That should be a
+ * no-op. */
+ nir_instr_remove(&intrin->instr);
continue;
+ }
if (!node->lower_to_ssa)
continue;
- def_stack_pop_if_in_block(node, block);
- nir_instr_remove(&intrin->instr);
- }
- }
+ assert(intrin->num_components ==
+ glsl_get_vector_elements(node->type));
- return true;
-}
+ assert(intrin->src[0].is_ssa);
-/* Inserts phi nodes for all variables marked lower_to_ssa
- *
- * This is the same algorithm as presented in "Efficiently Computing Static
- * Single Assignment Form and the Control Dependence Graph" by Cytron et.
- * al.
- */
-static void
-insert_phi_nodes(struct lower_variables_state *state)
-{
- NIR_VLA_ZERO(unsigned, work, state->impl->num_blocks);
- NIR_VLA_ZERO(unsigned, has_already, state->impl->num_blocks);
-
- /*
- * Since the work flags already prevent us from inserting a node that has
- * ever been inserted into W, we don't need to use a set to represent W.
- * Also, since no block can ever be inserted into W more than once, we know
- * that the maximum size of W is the number of basic blocks in the
- * function. So all we need to handle W is an array and a pointer to the
- * next element to be inserted and the next element to be removed.
- */
- NIR_VLA(nir_block *, W, state->impl->num_blocks);
-
- unsigned w_start, w_end;
- unsigned iter_count = 0;
-
- foreach_list_typed(struct deref_node, node, direct_derefs_link,
- &state->direct_deref_nodes) {
- if (node->stores == NULL)
- continue;
+ nir_ssa_def *new_def;
+ b.cursor = nir_before_instr(&intrin->instr);
- if (!node->lower_to_ssa)
- continue;
-
- w_start = w_end = 0;
- iter_count++;
-
- struct set_entry *store_entry;
- set_foreach(node->stores, store_entry) {
- nir_intrinsic_instr *store = (nir_intrinsic_instr *)store_entry->key;
- if (work[store->instr.block->index] < iter_count)
- W[w_end++] = store->instr.block;
- work[store->instr.block->index] = iter_count;
- }
-
- while (w_start != w_end) {
- nir_block *cur = W[w_start++];
- struct set_entry *dom_entry;
- set_foreach(cur->dom_frontier, dom_entry) {
- nir_block *next = (nir_block *) dom_entry->key;
-
- /*
- * If there's more than one return statement, then the end block
- * can be a join point for some definitions. However, there are
- * no instructions in the end block, so nothing would use those
- * phi nodes. Of course, we couldn't place those phi nodes
- * anyways due to the restriction of having no instructions in the
- * end block...
+ if (intrin->const_index[0] == (1 << intrin->num_components) - 1) {
+ /* Whole variable store - just copy the source. Note that
+ * intrin->num_components and intrin->src[0].ssa->num_components
+ * may differ.
*/
- if (next == state->impl->end_block)
- continue;
-
- if (has_already[next->index] < iter_count) {
- nir_phi_instr *phi = nir_phi_instr_create(state->shader);
- nir_ssa_dest_init(&phi->instr, &phi->dest,
- glsl_get_vector_elements(node->type), NULL);
- nir_instr_insert_before_block(next, &phi->instr);
+ unsigned swiz[4];
+ for (unsigned i = 0; i < 4; i++)
+ swiz[i] = i < intrin->num_components ? i : 0;
- _mesa_hash_table_insert(state->phi_table, phi, node);
-
- has_already[next->index] = iter_count;
- if (work[next->index] < iter_count) {
- work[next->index] = iter_count;
- W[w_end++] = next;
+ new_def = nir_swizzle(&b, intrin->src[0].ssa, swiz,
+ intrin->num_components, false);
+ } else {
+ nir_ssa_def *old_def =
+ nir_phi_builder_value_get_block_def(node->pb_value, block);
+ /* For writemasked store_var intrinsics, we combine the newly
+ * written values with the existing contents of unwritten
+ * channels, creating a new SSA value for the whole vector.
+ */
+ nir_ssa_def *srcs[4];
+ for (unsigned i = 0; i < intrin->num_components; i++) {
+ if (intrin->const_index[0] & (1 << i)) {
+ srcs[i] = nir_channel(&b, intrin->src[0].ssa, i);
+ } else {
+ srcs[i] = nir_channel(&b, old_def, i);
}
}
+ new_def = nir_vec(&b, srcs, intrin->num_components);
}
+
+ assert(new_def->num_components == intrin->num_components);
+
+ nir_phi_builder_value_set_block_def(node->pb_value, block, new_def);
+ nir_instr_remove(&intrin->instr);
+ break;
+ }
+
+ default:
+ break;
}
}
-}
+ for (unsigned i = 0; i < block->num_dom_children; ++i)
+ rename_variables_block(block->dom_children[i], state);
+
+ return true;
+}
/** Implements a pass to lower variable uses to SSA values
*
{
struct lower_variables_state state;
- state.shader = impl->overload->function->shader;
+ state.shader = impl->function->shader;
state.dead_ctx = ralloc_context(state.shader);
state.impl = impl;
_mesa_hash_pointer,
_mesa_key_pointer_equal);
exec_list_make_empty(&state.direct_deref_nodes);
- state.phi_table = _mesa_hash_table_create(state.dead_ctx,
- _mesa_hash_pointer,
- _mesa_key_pointer_equal);
/* Build the initial deref structures and direct_deref_nodes table */
state.add_to_direct_deref_nodes = true;
node->lower_to_ssa = true;
progress = true;
- if (deref->var->constant_initializer) {
- nir_load_const_instr *load =
- nir_deref_get_const_initializer_load(state.shader, deref);
- nir_ssa_def_init(&load->instr, &load->def,
- glsl_get_vector_elements(node->type), NULL);
- nir_instr_insert_before_cf_list(&impl->body, &load->instr);
- def_stack_push(node, &load->def, &state);
- }
-
foreach_deref_node_match(deref, lower_copies_to_load_store, &state);
}
*/
nir_foreach_block(impl, register_variable_uses_block, &state);
- insert_phi_nodes(&state);
+ state.phi_builder = nir_phi_builder_create(state.impl);
+
+ NIR_VLA(BITSET_WORD, store_blocks, BITSET_WORDS(state.impl->num_blocks));
+ foreach_list_typed(struct deref_node, node, direct_derefs_link,
+ &state.direct_deref_nodes) {
+ if (!node->lower_to_ssa)
+ continue;
+
+ memset(store_blocks, 0,
+ BITSET_WORDS(state.impl->num_blocks) * sizeof(*store_blocks));
+
+ if (node->stores) {
+ struct set_entry *store_entry;
+ set_foreach(node->stores, store_entry) {
+ nir_intrinsic_instr *store =
+ (nir_intrinsic_instr *)store_entry->key;
+ BITSET_SET(store_blocks, store->instr.block->index);
+ }
+ }
+
+ if (node->deref->var->constant_initializer)
+ BITSET_SET(store_blocks, 0);
+
+ node->pb_value =
+ nir_phi_builder_add_value(state.phi_builder,
+ glsl_get_vector_elements(node->type),
+ store_blocks);
+
+ if (node->deref->var->constant_initializer) {
+ nir_load_const_instr *load =
+ nir_deref_get_const_initializer_load(state.shader, node->deref);
+ nir_instr_insert_before_cf_list(&impl->body, &load->instr);
+ nir_phi_builder_value_set_block_def(node->pb_value,
+ nir_start_block(impl), &load->def);
+ }
+ }
+
rename_variables_block(nir_start_block(impl), &state);
+ nir_phi_builder_finish(state.phi_builder);
+
nir_metadata_preserve(impl, nir_metadata_block_index |
nir_metadata_dominance);
void
nir_lower_vars_to_ssa(nir_shader *shader)
{
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- nir_lower_vars_to_ssa_impl(overload->impl);
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ nir_lower_vars_to_ssa_impl(function->impl);
}
}
{
struct vec_to_movs_state *state = void_state;
nir_function_impl *impl = state->impl;
- nir_shader *shader = impl->overload->function->shader;
+ nir_shader *shader = impl->function->shader;
nir_foreach_instr_safe(block, instr) {
if (instr->type != nir_instr_type_alu)
{
bool progress = false;
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- progress = nir_lower_vec_to_movs_impl(overload->impl) || progress;
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ progress = nir_lower_vec_to_movs_impl(function->impl) || progress;
}
return progress;
void
nir_metadata_set_validation_flag(nir_shader *shader)
{
- nir_foreach_overload(shader, overload) {
- if (overload->impl) {
- overload->impl->valid_metadata |= nir_metadata_not_properly_reset;
+ nir_foreach_function(shader, function) {
+ if (function->impl) {
+ function->impl->valid_metadata |= nir_metadata_not_properly_reset;
}
}
}
void
nir_metadata_check_validation_flag(nir_shader *shader)
{
- nir_foreach_overload(shader, overload) {
- if (overload->impl) {
- assert(!(overload->impl->valid_metadata &
+ nir_foreach_function(shader, function) {
+ if (function->impl) {
+ assert(!(function->impl->valid_metadata &
nir_metadata_not_properly_reset));
}
}
void
nir_move_vec_src_uses_to_dest(nir_shader *shader)
{
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- nir_move_vec_src_uses_to_dest_impl(shader, overload->impl);
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ nir_move_vec_src_uses_to_dest_impl(shader, function->impl);
}
}
{
bool progress = false;
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- progress = normalize_cubemap_coords_impl(overload->impl) || progress;
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ progress = normalize_cubemap_coords_impl(function->impl) || progress;
}
return progress;
unop_convert("b2i", tbool, tint, "src0 ? 1 : 0") # Boolean-to-int conversion
unop_convert("u2f", tuint, tfloat, "src0") # Unsigned-to-float conversion.
-unop_reduce("bany", 1, tbool, tbool, "{src}", "{src0} || {src1}", "{src}")
-unop_reduce("ball", 1, tbool, tbool, "{src}", "{src0} && {src1}", "{src}")
-unop_reduce("fany", 1, tfloat, tfloat, "{src} != 0.0f", "{src0} || {src1}",
- "{src} ? 1.0f : 0.0f")
-unop_reduce("fall", 1, tfloat, tfloat, "{src} != 0.0f", "{src0} && {src1}",
- "{src} ? 1.0f : 0.0f")
-
# Unary floating-point rounding operations.
unop("ffract", tfloat, "src0 - floorf(src0)")
unop("fround_even", tfloat, "_mesa_roundevenf(src0)")
+unop("fquantize2f16", tfloat, "_mesa_half_to_float(_mesa_float_to_half(src0))")
# Trigonometric operations.
# returns a boolean representing the carry resulting from the addition of
# the two unsigned arguments.
-binop_convert("uadd_carry", tbool, tuint, commutative, "src0 + src1 < src0")
+binop_convert("uadd_carry", tuint, tuint, commutative, "src0 + src1 < src0")
# returns a boolean representing the borrow resulting from the subtraction
# of the two unsigned arguments.
-binop_convert("usub_borrow", tbool, tuint, "", "src1 < src0")
+binop_convert("usub_borrow", tuint, tuint, "", "src0 < src1")
-binop("fmod", tfloat, "", "src0 - src1 * floorf(src0 / src1)")
binop("umod", tuint, "", "src1 == 0 ? 0 : src0 % src1")
+# For signed integers, there are several different possible definitions of
+# "modulus" or "remainder". We follow the conventions used by LLVM and
+# SPIR-V. The irem opcode implements the standard C/C++ signed "%"
+# operation while the imod opcode implements the more mathematical
+# "modulus" operation. For details on the difference, see
+#
+# http://mathforum.org/library/drmath/view/52343.html
+
+binop("irem", tint, "", "src1 == 0 ? 0 : src0 % src1")
+binop("imod", tint, "",
+ "src1 == 0 ? 0 : ((src0 % src1 == 0 || (src0 >= 0) == (src1 >= 0)) ?"
+ " src0 % src1 : src0 % src1 + src1)")
+binop("fmod", tfloat, "", "src0 - src1 * floorf(src0 / src1)")
+binop("frem", tfloat, "", "src0 - src1 * truncf(src0 / src1)")
+
#
# Comparisons
#
binop_horiz("pack_half_2x16_split", 1, tuint, 1, tfloat, 1, tfloat,
"pack_half_1x16(src0.x) | (pack_half_1x16(src1.x) << 16)")
+# bfm implements the behavior of the first operation of the SM5 "bfi" assembly
+# and that of the "bfi1" i965 instruction. That is, it has undefined behavior
+# if either of its arguments are 32.
binop_convert("bfm", tuint, tint, "", """
-int offset = src0, bits = src1;
-if (offset < 0 || bits < 0 || offset + bits > 32)
- dst = 0; /* undefined per the spec */
+int bits = src0, offset = src1;
+if (offset < 0 || bits < 0 || offset > 31 || bits > 31 || offset + bits > 32)
+ dst = 0; /* undefined */
else
- dst = ((1 << bits)- 1) << offset;
+ dst = ((1u << bits) - 1) << offset;
""")
opcode("ldexp", 0, tfloat, [0, 0], [tfloat, tint], "", """
opcode("bcsel", 0, tuint, [0, 0, 0],
[tbool, tuint, tuint], "", "src0 ? src1 : src2")
+# SM5 bfi assembly
triop("bfi", tuint, """
unsigned mask = src0, insert = src1, base = src2;
if (mask == 0) {
}
""")
+# SM5 ubfe/ibfe assembly
+opcode("ubfe", 0, tuint,
+ [0, 0, 0], [tuint, tint, tint], "", """
+unsigned base = src0;
+int offset = src1, bits = src2;
+if (bits == 0) {
+ dst = 0;
+} else if (bits < 0 || offset < 0) {
+ dst = 0; /* undefined */
+} else if (offset + bits < 32) {
+ dst = (base << (32 - bits - offset)) >> (32 - bits);
+} else {
+ dst = base >> offset;
+}
+""")
+opcode("ibfe", 0, tint,
+ [0, 0, 0], [tint, tint, tint], "", """
+int base = src0;
+int offset = src1, bits = src2;
+if (bits == 0) {
+ dst = 0;
+} else if (bits < 0 || offset < 0) {
+ dst = 0; /* undefined */
+} else if (offset + bits < 32) {
+ dst = (base << (32 - bits - offset)) >> (32 - bits);
+} else {
+ dst = base >> offset;
+}
+""")
+
+# GLSL bitfieldExtract()
opcode("ubitfield_extract", 0, tuint,
- [0, 1, 1], [tuint, tint, tint], "", """
+ [0, 0, 0], [tuint, tint, tint], "", """
unsigned base = src0;
-int offset = src1.x, bits = src2.x;
+int offset = src1, bits = src2;
if (bits == 0) {
dst = 0;
} else if (bits < 0 || offset < 0 || offset + bits > 32) {
dst = 0; /* undefined per the spec */
} else {
- dst = (base >> offset) & ((1 << bits) - 1);
+ dst = (base >> offset) & ((1ull << bits) - 1);
}
""")
opcode("ibitfield_extract", 0, tint,
- [0, 1, 1], [tint, tint, tint], "", """
+ [0, 0, 0], [tint, tint, tint], "", """
int base = src0;
-int offset = src1.x, bits = src2.x;
+int offset = src1, bits = src2;
if (bits == 0) {
dst = 0;
} else if (offset < 0 || bits < 0 || offset + bits > 32) {
[tuint, tuint, tuint, tuint],
"", const_expr)
-opcode("bitfield_insert", 0, tuint, [0, 0, 1, 1],
+opcode("bitfield_insert", 0, tuint, [0, 0, 0, 0],
[tuint, tuint, tint, tint], "", """
unsigned base = src0, insert = src1;
-int offset = src2.x, bits = src3.x;
+int offset = src2, bits = src3;
if (bits == 0) {
dst = 0;
} else if (offset < 0 || bits < 0 || bits + offset > 32) {
dst = 0;
} else {
- unsigned mask = ((1 << bits) - 1) << offset;
+ unsigned mask = ((1ull << bits) - 1) << offset;
dst = (base & ~mask) | ((insert << bits) & mask);
}
""")
#! /usr/bin/env python
+# -*- encoding: utf-8 -*-
#
# Copyright (C) 2014 Intel Corporation
#
(('iadd', ('imul', a, b), ('imul', a, c)), ('imul', a, ('iadd', b, c))),
(('fadd', ('fneg', a), a), 0.0),
(('iadd', ('ineg', a), a), 0),
+ (('iadd', ('ineg', a), ('iadd', a, b)), b),
+ (('iadd', a, ('iadd', ('ineg', a), b)), b),
+ (('fadd', ('fneg', a), ('fadd', a, b)), b),
+ (('fadd', a, ('fadd', ('fneg', a), b)), b),
(('fmul', a, 0.0), 0.0),
(('imul', a, 0), 0),
(('umul_unorm_4x8', a, 0), 0),
(('imul', a, 1), a),
(('fmul', a, -1.0), ('fneg', a)),
(('imul', a, -1), ('ineg', a)),
+ (('fdiv', a, b), ('fmul', a, ('frcp', b)), 'options->lower_fdiv'),
(('ffma', 0.0, a, b), b),
(('ffma', a, 0.0, b), b),
(('ffma', a, b, 0.0), ('fmul', a, b)),
(('fmul', ('fexp2', a), ('fexp2', b)), ('fexp2', ('fadd', a, b))),
# Division and reciprocal
(('fdiv', 1.0, a), ('frcp', a)),
+ (('fdiv', a, b), ('fmul', a, ('frcp', b)), 'options->lower_fdiv'),
(('frcp', ('frcp', a)), a),
(('frcp', ('fsqrt', a)), ('frsq', a)),
(('fsqrt', a), ('frcp', ('frsq', a)), 'options->lower_fsqrt'),
(('iadd', a, ('isub', 0, b)), ('isub', a, b)),
(('fabs', ('fsub', 0.0, a)), ('fabs', a)),
(('iabs', ('isub', 0, a)), ('iabs', a)),
+
+ # Misc. lowering
+ (('fmod', a, b), ('fsub', a, ('fmul', b, ('ffloor', ('fdiv', a, b)))), 'options->lower_fmod'),
+ (('frem', a, b), ('fsub', a, ('fmul', b, ('ftrunc', ('fdiv', a, b)))), 'options->lower_fmod'),
+ (('uadd_carry', a, b), ('b2i', ('ult', ('iadd', a, b), a)), 'options->lower_uadd_carry'),
+ (('usub_borrow', a, b), ('b2i', ('ult', a, b)), 'options->lower_usub_borrow'),
+
+ (('bitfield_insert', 'base', 'insert', 'offset', 'bits'),
+ ('bcsel', ('ilt', 31, 'bits'), 'insert',
+ ('bfi', ('bfm', 'bits', 'offset'), 'insert', 'base')),
+ 'options->lower_bitfield_insert'),
+
+ (('ibitfield_extract', 'value', 'offset', 'bits'),
+ ('bcsel', ('ilt', 31, 'bits'), 'value',
+ ('ibfe', 'value', 'offset', 'bits')),
+ 'options->lower_bitfield_extract'),
+
+ (('ubitfield_extract', 'value', 'offset', 'bits'),
+ ('bcsel', ('ult', 31, 'bits'), 'value',
+ ('ubfe', 'value', 'offset', 'bits')),
+ 'options->lower_bitfield_extract'),
]
# Add optimizations to handle the case where the result of a ternary is
('bcsel', 'a', (op, 'd', 'b'), (op, 'd', 'c'))),
]
+def ldexp_to_arith(x, exp):
+ """
+ Translates
+ ldexp x exp
+ into
+
+ extracted_biased_exp = rshift(bitcast_f2i(abs(x)), exp_shift);
+ resulting_biased_exp = extracted_biased_exp + exp;
+
+ if (resulting_biased_exp < 1) {
+ return copysign(0.0, x);
+ }
+
+ return bitcast_u2f((bitcast_f2u(x) & sign_mantissa_mask) |
+ lshift(i2u(resulting_biased_exp), exp_shift));
+
+ which we can't actually implement as such, since NIR doesn't have
+ vectorized if-statements. We actually implement it without branches
+ using conditional-select:
+
+ extracted_biased_exp = rshift(bitcast_f2i(abs(x)), exp_shift);
+ resulting_biased_exp = extracted_biased_exp + exp;
+
+ is_not_zero_or_underflow = gequal(resulting_biased_exp, 1);
+ x = csel(is_not_zero_or_underflow, x, copysign(0.0f, x));
+ resulting_biased_exp = csel(is_not_zero_or_underflow,
+ resulting_biased_exp, 0);
+
+ return bitcast_u2f((bitcast_f2u(x) & sign_mantissa_mask) |
+ lshift(i2u(resulting_biased_exp), exp_shift));
+ """
+
+ sign_mask = 0x80000000
+ exp_shift = 23
+ exp_width = 8
+
+ # Extract the biased exponent from <x>.
+ extracted_biased_exp = ('ushr', ('fabs', x), exp_shift)
+ resulting_biased_exp = ('iadd', extracted_biased_exp, exp)
+
+ # Test if result is ±0.0, subnormal, or underflow by checking if the
+ # resulting biased exponent would be less than 0x1. If so, the result is
+ # 0.0 with the sign of x. (Actually, invert the conditions so that
+ # immediate values are the second arguments, which is better for i965)
+ zero_sign_x = ('iand', x, sign_mask)
+
+ is_not_zero_or_underflow = ('ige', resulting_biased_exp, 0x1)
+
+ # We could test for overflows by checking if the resulting biased exponent
+ # would be greater than 0xFE. Turns out we don't need to because the GLSL
+ # spec says:
+ #
+ # "If this product is too large to be represented in the
+ # floating-point type, the result is undefined."
+
+ return ('bitfield_insert',
+ ('bcsel', is_not_zero_or_underflow, x, zero_sign_x),
+ ('bcsel', is_not_zero_or_underflow, resulting_biased_exp, 0),
+ exp_shift, exp_width)
+
+optimizations += [(('ldexp', 'x', 'exp'), ldexp_to_arith('x', 'exp'))]
+
# This section contains "late" optimizations that should be run after the
# regular optimizations have finished. Optimizations should go here if
# they help code generation but do not necessarily produce code that is
{
bool progress = false;
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- progress |= nir_opt_constant_folding_impl(overload->impl);
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ progress |= nir_opt_constant_folding_impl(function->impl);
}
return progress;
{
bool progress = false;
- nir_foreach_overload(shader, overload) {
- if (overload->impl && nir_copy_prop_impl(overload->impl))
+ nir_foreach_function(shader, function) {
+ if (function->impl && nir_copy_prop_impl(function->impl))
progress = true;
}
{
bool progress = false;
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- progress |= nir_opt_cse_impl(overload->impl);
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ progress |= nir_opt_cse_impl(function->impl);
}
return progress;
nir_opt_dce(nir_shader *shader)
{
bool progress = false;
- nir_foreach_overload(shader, overload) {
- if (overload->impl && nir_opt_dce_impl(overload->impl))
+ nir_foreach_function(shader, function) {
+ if (function->impl && nir_opt_dce_impl(function->impl))
progress = true;
}
{
bool progress = false;
- nir_foreach_overload(shader, overload)
- if (overload->impl)
- progress |= opt_dead_cf_impl(overload->impl);
+ nir_foreach_function(shader, function)
+ if (function->impl)
+ progress |= opt_dead_cf_impl(function->impl);
return progress;
}
void
nir_opt_gcm(nir_shader *shader)
{
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- opt_gcm_impl(overload->impl);
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ opt_gcm_impl(function->impl);
}
}
{
bool progress = false;
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- progress |= nir_opt_peephole_select_impl(overload->impl);
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ progress |= nir_opt_peephole_select_impl(function->impl);
}
return progress;
{
bool progress = false;
- nir_foreach_overload(shader, overload)
- if (overload->impl)
- progress = remove_phis_impl(overload->impl) || progress;
+ nir_foreach_function(shader, function)
+ if (function->impl)
+ progress = remove_phis_impl(function->impl) || progress;
return progress;
}
{
bool progress = false;
- nir_foreach_overload(shader, overload) {
- if (overload->impl) {
- nir_foreach_block(overload->impl, opt_undef_block, &progress);
+ nir_foreach_function(shader, function) {
+ if (function->impl) {
+ nir_foreach_block(function->impl, opt_undef_block, &progress);
if (progress)
- nir_metadata_preserve(overload->impl,
+ nir_metadata_preserve(function->impl,
nir_metadata_block_index |
nir_metadata_dominance);
}
--- /dev/null
+/*
+ * Copyright © 2016 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include "nir_phi_builder.h"
+#include "nir/nir_vla.h"
+
+struct nir_phi_builder {
+ nir_shader *shader;
+ nir_function_impl *impl;
+
+ /* Copied from the impl for easy access */
+ unsigned num_blocks;
+
+ /* Array of all blocks indexed by block->index. */
+ nir_block **blocks;
+
+ /* Hold on to the values so we can easily iterate over them. */
+ struct exec_list values;
+
+ /* Worklist for phi adding */
+ unsigned iter_count;
+ unsigned *work;
+ nir_block **W;
+};
+
+#define NEEDS_PHI ((nir_ssa_def *)(intptr_t)-1)
+
+struct nir_phi_builder_value {
+ struct exec_node node;
+
+ struct nir_phi_builder *builder;
+
+ /* Needed so we can create phis and undefs */
+ unsigned num_components;
+
+ /* The list of phi nodes associated with this value. Phi nodes are not
+ * added directly. Instead, they are created, the instr->block pointer
+ * set, and then added to this list. Later, in phi_builder_finish, we
+ * set up their sources and add them to the top of their respective
+ * blocks.
+ */
+ struct exec_list phis;
+
+ /* Array of SSA defs, indexed by block. If a phi needs to be inserted
+ * in a given block, it will have the magic value NEEDS_PHI.
+ */
+ nir_ssa_def *defs[0];
+};
+
+static bool
+fill_block_array(nir_block *block, void *void_data)
+{
+ nir_block **blocks = void_data;
+ blocks[block->index] = block;
+ return true;
+}
+
+struct nir_phi_builder *
+nir_phi_builder_create(nir_function_impl *impl)
+{
+ struct nir_phi_builder *pb = ralloc(NULL, struct nir_phi_builder);
+
+ pb->shader = impl->function->shader;
+ pb->impl = impl;
+
+ assert(impl->valid_metadata & (nir_metadata_block_index |
+ nir_metadata_dominance));
+
+ pb->num_blocks = impl->num_blocks;
+ pb->blocks = ralloc_array(pb, nir_block *, pb->num_blocks);
+ nir_foreach_block(impl, fill_block_array, pb->blocks);
+
+ exec_list_make_empty(&pb->values);
+
+ pb->iter_count = 0;
+ pb->work = rzalloc_array(pb, unsigned, pb->num_blocks);
+ pb->W = ralloc_array(pb, nir_block *, pb->num_blocks);
+
+ return pb;
+}
+
+struct nir_phi_builder_value *
+nir_phi_builder_add_value(struct nir_phi_builder *pb, unsigned num_components,
+ const BITSET_WORD *defs)
+{
+ struct nir_phi_builder_value *val;
+ unsigned i, w_start = 0, w_end = 0;
+
+ val = rzalloc_size(pb, sizeof(*val) + sizeof(val->defs[0]) * pb->num_blocks);
+ val->builder = pb;
+ val->num_components = num_components;
+ exec_list_make_empty(&val->phis);
+ exec_list_push_tail(&pb->values, &val->node);
+
+ pb->iter_count++;
+
+ BITSET_WORD tmp;
+ BITSET_FOREACH_SET(i, tmp, defs, pb->num_blocks) {
+ if (pb->work[i] < pb->iter_count)
+ pb->W[w_end++] = pb->blocks[i];
+ pb->work[i] = pb->iter_count;
+ }
+
+ while (w_start != w_end) {
+ nir_block *cur = pb->W[w_start++];
+ struct set_entry *dom_entry;
+ set_foreach(cur->dom_frontier, dom_entry) {
+ nir_block *next = (nir_block *) dom_entry->key;
+
+ /*
+ * If there's more than one return statement, then the end block
+ * can be a join point for some definitions. However, there are
+ * no instructions in the end block, so nothing would use those
+ * phi nodes. Of course, we couldn't place those phi nodes
+ * anyways due to the restriction of having no instructions in the
+ * end block...
+ */
+ if (next == pb->impl->end_block)
+ continue;
+
+ if (val->defs[next->index] == NULL) {
+ val->defs[next->index] = NEEDS_PHI;
+
+ if (pb->work[next->index] < pb->iter_count) {
+ pb->work[next->index] = pb->iter_count;
+ pb->W[w_end++] = next;
+ }
+ }
+ }
+ }
+
+ return val;
+}
+
+void
+nir_phi_builder_value_set_block_def(struct nir_phi_builder_value *val,
+ nir_block *block, nir_ssa_def *def)
+{
+ val->defs[block->index] = def;
+}
+
+nir_ssa_def *
+nir_phi_builder_value_get_block_def(struct nir_phi_builder_value *val,
+ nir_block *block)
+{
+ if (val->defs[block->index] == NULL) {
+ if (block->imm_dom) {
+ /* Grab it from our immediate dominator. We'll stash it here for
+ * easy access later.
+ */
+ val->defs[block->index] =
+ nir_phi_builder_value_get_block_def(val, block->imm_dom);
+ return val->defs[block->index];
+ } else {
+ /* No immediate dominator means that this block is either the
+ * start block or unreachable. In either case, the value is
+ * undefined so we need an SSA undef.
+ */
+ nir_ssa_undef_instr *undef =
+ nir_ssa_undef_instr_create(val->builder->shader,
+ val->num_components);
+ nir_instr_insert(nir_before_cf_list(&val->builder->impl->body),
+ &undef->instr);
+ val->defs[block->index] = &undef->def;
+ return &undef->def;
+ }
+ } else if (val->defs[block->index] == NEEDS_PHI) {
+ /* If we need a phi instruction, go ahead and create one but don't
+ * add it to the program yet. Later, we'll go through and set up phi
+ * sources and add the instructions will be added at that time.
+ */
+ nir_phi_instr *phi = nir_phi_instr_create(val->builder->shader);
+ nir_ssa_dest_init(&phi->instr, &phi->dest, val->num_components, NULL);
+ phi->instr.block = block;
+ exec_list_push_tail(&val->phis, &phi->instr.node);
+ val->defs[block->index] = &phi->dest.ssa;
+ return &phi->dest.ssa;
+ } else {
+ return val->defs[block->index];
+ }
+}
+
+static int
+compare_blocks(const void *_a, const void *_b)
+{
+ nir_block * const * a = _a;
+ nir_block * const * b = _b;
+
+ return (*a)->index - (*b)->index;
+}
+
+void
+nir_phi_builder_finish(struct nir_phi_builder *pb)
+{
+ const unsigned num_blocks = pb->num_blocks;
+ NIR_VLA(nir_block *, preds, num_blocks);
+
+ foreach_list_typed(struct nir_phi_builder_value, val, node, &pb->values) {
+ /* We can't iterate over the list of phis normally because we are
+ * removing them as we go and, in some cases, adding new phis as we
+ * build the source lists of others.
+ */
+ while (!exec_list_is_empty(&val->phis)) {
+ struct exec_node *head = exec_list_get_head(&val->phis);
+ nir_phi_instr *phi = exec_node_data(nir_phi_instr, head, instr.node);
+ assert(phi->instr.type == nir_instr_type_phi);
+
+ exec_node_remove(&phi->instr.node);
+
+ /* Construct an array of predecessors. We sort it to ensure
+ * determinism in the phi insertion algorithm.
+ *
+ * XXX: Calling qsort this many times seems expensive.
+ */
+ int num_preds = 0;
+ struct set_entry *entry;
+ set_foreach(phi->instr.block->predecessors, entry)
+ preds[num_preds++] = (nir_block *)entry->key;
+ qsort(preds, num_preds, sizeof(*preds), compare_blocks);
+
+ for (unsigned i = 0; i < num_preds; i++) {
+ nir_phi_src *src = ralloc(phi, nir_phi_src);
+ src->pred = preds[i];
+ src->src = nir_src_for_ssa(
+ nir_phi_builder_value_get_block_def(val, preds[i]));
+ exec_list_push_tail(&phi->srcs, &src->node);
+ }
+
+ nir_instr_insert(nir_before_block(phi->instr.block), &phi->instr);
+ }
+ }
+
+ ralloc_free(pb);
+}
--- /dev/null
+/*
+ * Copyright © 2016 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#pragma once
+
+#include "nir.h"
+
+struct nir_phi_builder;
+struct nir_phi_builder_value;
+
+/* Create a new phi builder.
+ *
+ * While this is fairly cheap, it does allocate some memory and walk the list
+ * of blocks so it's recommended that you only call it once and use it to
+ * build phis for several values.
+ */
+struct nir_phi_builder *nir_phi_builder_create(nir_function_impl *impl);
+
+/* Register a value with the builder.
+ *
+ * The 'defs' parameter specifies a bitset of blocks in which the given value
+ * is defined. This is used to determine where to place the phi nodes.
+ */
+struct nir_phi_builder_value *
+nir_phi_builder_add_value(struct nir_phi_builder *pb, unsigned num_components,
+ const BITSET_WORD *defs);
+
+/* Register a definition for the given value and block.
+ *
+ * It is safe to call this function as many times as you wish for any given
+ * block/value pair. However, it always replaces whatever was there
+ * previously even if that definition is from a phi node. The phi builder
+ * always uses the latest information it has, so you must be careful about the
+ * order in which you register definitions. The final value at the end of the
+ * block must be the last value registered.
+ */
+void
+nir_phi_builder_value_set_block_def(struct nir_phi_builder_value *val,
+ nir_block *block, nir_ssa_def *def);
+
+/* Get the definition for the given value in the given block.
+ *
+ * This definition will always be the latest definition known for the given
+ * block. If no definition is immediately available, it will crawl up the
+ * dominance tree and insert phi nodes as needed until it finds one. In the
+ * case that no suitable definition is found, it will return the result of a
+ * nir_ssa_undef_instr with the correct number of components.
+ *
+ * Because this function only uses the latest available information for any
+ * given block, you must have already finished registering definitions for any
+ * blocks that dominate the current block in order to get the correct result.
+ */
+nir_ssa_def *
+nir_phi_builder_value_get_block_def(struct nir_phi_builder_value *val,
+ nir_block *block);
+
+/* Finish building phi nodes and free the builder.
+ *
+ * This function does far more than just free memory. Prior to calling
+ * nir_phi_builder_finish, no phi nodes have actually been inserted in the
+ * program. This function is what finishes setting up phi node sources and
+ * adds the phi nodes to the program.
+ */
+void nir_phi_builder_finish(struct nir_phi_builder *pb);
}
}
+static const char *
+get_var_name(nir_variable *var, print_state *state)
+{
+ if (state->ht == NULL)
+ return var->name;
+
+ assert(state->syms);
+
+ struct hash_entry *entry = _mesa_hash_table_search(state->ht, var);
+ if (entry)
+ return entry->data;
+
+ char *name;
+ if (var->name == NULL) {
+ name = ralloc_asprintf(state->syms, "@%u", state->index++);
+ } else {
+ struct set_entry *set_entry = _mesa_set_search(state->syms, var->name);
+ if (set_entry != NULL) {
+ /* we have a collision with another name, append an @ + a unique
+ * index */
+ name = ralloc_asprintf(state->syms, "%s@%u", var->name,
+ state->index++);
+ } else {
+ /* Mark this one as seen */
+ _mesa_set_add(state->syms, var->name);
+ name = var->name;
+ }
+ }
+
+ _mesa_hash_table_insert(state->ht, var, name);
+
+ return name;
+}
+
+static void
+print_constant(nir_constant *c, const struct glsl_type *type, print_state *state)
+{
+ FILE *fp = state->fp;
+ unsigned total_elems = glsl_get_components(type);
+ unsigned i;
+
+ switch (glsl_get_base_type(type)) {
+ case GLSL_TYPE_UINT:
+ case GLSL_TYPE_INT:
+ case GLSL_TYPE_BOOL:
+ for (i = 0; i < total_elems; i++) {
+ if (i > 0) fprintf(fp, ", ");
+ fprintf(fp, "0x%08x", c->value.u[i]);
+ }
+ break;
+
+ case GLSL_TYPE_FLOAT:
+ for (i = 0; i < total_elems; i++) {
+ if (i > 0) fprintf(fp, ", ");
+ fprintf(fp, "%f", c->value.f[i]);
+ }
+ break;
+
+ case GLSL_TYPE_STRUCT:
+ for (i = 0; i < c->num_elements; i++) {
+ if (i > 0) fprintf(fp, ", ");
+ fprintf(fp, "{ ");
+ print_constant(c->elements[i], glsl_get_struct_field(type, i), state);
+ fprintf(fp, " }");
+ }
+ break;
+
+ case GLSL_TYPE_ARRAY:
+ for (i = 0; i < c->num_elements; i++) {
+ if (i > 0) fprintf(fp, ", ");
+ fprintf(fp, "{ ");
+ print_constant(c->elements[i], glsl_get_array_element(type), state);
+ fprintf(fp, " }");
+ }
+ break;
+
+ default:
+ unreachable("not reached");
+ }
+}
+
static void
print_var_decl(nir_variable *var, print_state *state)
{
const char *const patch = (var->data.patch) ? "patch " : "";
const char *const inv = (var->data.invariant) ? "invariant " : "";
const char *const mode[] = { "shader_in ", "shader_out ", "", "",
- "uniform ", "shader_storage", "system " };
+ "uniform ", "shader_storage ", "shared ",
+ "system "};
fprintf(fp, "%s%s%s%s%s%s ",
cent, samp, patch, inv, mode[var->data.mode],
glsl_print_type(var->type, fp);
- struct set_entry *entry = NULL;
- if (state->syms)
- entry = _mesa_set_search(state->syms, var->name);
-
- char *name;
-
- if (entry != NULL) {
- /* we have a collision with another name, append an @ + a unique index */
- name = ralloc_asprintf(state->syms, "%s@%u", var->name, state->index++);
- } else {
- name = var->name;
- }
-
- fprintf(fp, " %s", name);
+ fprintf(fp, " %s", get_var_name(var, state));
if (var->data.mode == nir_var_shader_in ||
var->data.mode == nir_var_shader_out ||
fprintf(fp, " (%s, %u)", loc, var->data.driver_location);
}
- fprintf(fp, "\n");
-
- if (state->syms) {
- _mesa_set_add(state->syms, name);
- _mesa_hash_table_insert(state->ht, var, name);
+ if (var->constant_initializer) {
+ fprintf(fp, " = { ");
+ print_constant(var->constant_initializer, var->type, state);
+ fprintf(fp, " }");
}
+
+ fprintf(fp, "\n");
}
static void
print_var(nir_variable *var, print_state *state)
{
FILE *fp = state->fp;
- const char *name;
- if (state->ht) {
- struct hash_entry *entry = _mesa_hash_table_search(state->ht, var);
-
- assert(entry != NULL);
- name = entry->data;
- } else {
- name = var->name;
- }
-
- fprintf(fp, "%s", name);
+ fprintf(fp, "%s", get_var_name(var, state));
}
static void
case nir_tex_src_ddy:
fprintf(fp, "(ddy)");
break;
+ case nir_tex_src_texture_offset:
+ fprintf(fp, "(texture_offset)");
+ break;
case nir_tex_src_sampler_offset:
fprintf(fp, "(sampler_offset)");
break;
fprintf(fp, "%u (gather_component), ", instr->component);
}
+ if (instr->texture) {
+ assert(instr->sampler);
+ fprintf(fp, " (texture)");
+ }
if (instr->sampler) {
print_deref(instr->sampler, state);
+ fprintf(fp, " (sampler)");
} else {
- fprintf(fp, "%u", instr->sampler_index);
+ assert(instr->texture == NULL);
+ fprintf(fp, "%u (texture) %u (sampler)",
+ instr->texture_index, instr->sampler_index);
}
-
- fprintf(fp, " (sampler)");
}
static void
{
FILE *fp = state->fp;
- fprintf(fp, "call %s ", instr->callee->function->name);
+ fprintf(fp, "call %s ", instr->callee->name);
for (unsigned i = 0; i < instr->num_params; i++) {
if (i != 0)
{
FILE *fp = state->fp;
- fprintf(fp, "\nimpl %s ", impl->overload->function->name);
+ fprintf(fp, "\nimpl %s ", impl->function->name);
for (unsigned i = 0; i < impl->num_params; i++) {
if (i != 0)
}
static void
-print_function_overload(nir_function_overload *overload,
- print_state *state)
+print_function(nir_function *function, print_state *state)
{
FILE *fp = state->fp;
- fprintf(fp, "decl_overload %s ", overload->function->name);
+ fprintf(fp, "decl_function %s ", function->name);
- for (unsigned i = 0; i < overload->num_params; i++) {
+ for (unsigned i = 0; i < function->num_params; i++) {
if (i != 0)
fprintf(fp, ", ");
- switch (overload->params[i].param_type) {
+ switch (function->params[i].param_type) {
case nir_parameter_in:
fprintf(fp, "in ");
break;
unreachable("Invalid parameter type");
}
- glsl_print_type(overload->params[i].type, fp);
+ glsl_print_type(function->params[i].type, fp);
}
- if (overload->return_type != NULL) {
- if (overload->num_params != 0)
+ if (function->return_type != NULL) {
+ if (function->num_params != 0)
fprintf(fp, ", ");
fprintf(fp, "returning ");
- glsl_print_type(overload->return_type, fp);
+ glsl_print_type(function->return_type, fp);
}
fprintf(fp, "\n");
- if (overload->impl != NULL) {
- print_function_impl(overload->impl, state);
+ if (function->impl != NULL) {
+ print_function_impl(function->impl, state);
return;
}
}
static void
-print_function(nir_function *func, print_state *state)
-{
- foreach_list_typed(nir_function_overload, overload, node, &func->overload_list) {
- print_function_overload(overload, state);
- }
-}
-
-static void
init_print_state(print_state *state, nir_shader *shader, FILE *fp)
{
state->fp = fp;
fprintf(fp, "inputs: %u\n", shader->num_inputs);
fprintf(fp, "outputs: %u\n", shader->num_outputs);
fprintf(fp, "uniforms: %u\n", shader->num_uniforms);
+ fprintf(fp, "shared: %u\n", shader->num_shared);
nir_foreach_variable(var, &shader->uniforms) {
print_var_decl(var, &state);
print_var_decl(var, &state);
}
+ nir_foreach_variable(var, &shader->shared) {
+ print_var_decl(var, &state);
+ }
+
nir_foreach_variable(var, &shader->globals) {
print_var_decl(var, &state);
}
static void
add_var_use_shader(nir_shader *shader, struct set *live)
{
- nir_foreach_overload(shader, overload) {
- if (overload->impl) {
- nir_foreach_block(overload->impl, add_var_use_block, live);
+ nir_foreach_function(shader, function) {
+ if (function->impl) {
+ nir_foreach_block(function->impl, add_var_use_block, live);
}
}
}
}
bool
-nir_remove_dead_variables(nir_shader *shader)
+nir_remove_dead_variables(nir_shader *shader, nir_variable_mode mode)
{
bool progress = false;
struct set *live =
add_var_use_shader(shader, live);
- progress = remove_dead_vars(&shader->globals, live) || progress;
+ if (mode == nir_var_uniform || mode == nir_var_all)
+ progress = remove_dead_vars(&shader->uniforms, live) || progress;
- nir_foreach_overload(shader, overload) {
- if (overload->impl) {
- if (remove_dead_vars(&overload->impl->locals, live)) {
- nir_metadata_preserve(overload->impl, nir_metadata_block_index |
- nir_metadata_dominance |
- nir_metadata_live_ssa_defs);
- progress = true;
+ if (mode == nir_var_shader_in || mode == nir_var_all)
+ progress = remove_dead_vars(&shader->inputs, live) || progress;
+
+ if (mode == nir_var_shader_out || mode == nir_var_all)
+ progress = remove_dead_vars(&shader->outputs, live) || progress;
+
+ if (mode == nir_var_global || mode == nir_var_all)
+ progress = remove_dead_vars(&shader->globals, live) || progress;
+
+ if (mode == nir_var_system_value || mode == nir_var_all)
+ progress = remove_dead_vars(&shader->system_values, live) || progress;
+
+ if (mode == nir_var_local || mode == nir_var_all) {
+ nir_foreach_function(shader, function) {
+ if (function->impl) {
+ if (remove_dead_vars(&function->impl->locals, live)) {
+ nir_metadata_preserve(function->impl, nir_metadata_block_index |
+ nir_metadata_dominance |
+ nir_metadata_live_ssa_defs);
+ progress = true;
+ }
}
}
}
--- /dev/null
+/*
+ * Copyright © 2016 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include "nir.h"
+#include "nir_phi_builder.h"
+
+struct repair_ssa_state {
+ nir_function_impl *impl;
+
+ BITSET_WORD *def_set;
+ struct nir_phi_builder *phi_builder;
+
+ bool progress;
+};
+
+/* Get ready to build a phi and return the builder */
+static struct nir_phi_builder *
+prep_build_phi(struct repair_ssa_state *state)
+{
+ const unsigned num_words = BITSET_WORDS(state->impl->num_blocks);
+
+ /* We create the phi builder on-demand. */
+ if (state->phi_builder == NULL) {
+ state->phi_builder = nir_phi_builder_create(state->impl);
+ state->def_set = ralloc_array(NULL, BITSET_WORD, num_words);
+ }
+
+ /* We're going to build a phi. That's progress. */
+ state->progress = true;
+
+ /* Set the defs set to empty */
+ memset(state->def_set, 0, num_words * sizeof(*state->def_set));
+
+ return state->phi_builder;
+}
+
+static nir_block *
+get_src_block(nir_src *src)
+{
+ if (src->parent_instr->type == nir_instr_type_phi) {
+ return exec_node_data(nir_phi_src, src, src)->pred;
+ } else {
+ return src->parent_instr->block;
+ }
+}
+
+static bool
+repair_ssa_def(nir_ssa_def *def, void *void_state)
+{
+ struct repair_ssa_state *state = void_state;
+
+ bool is_valid = true;
+ nir_foreach_use(def, src) {
+ if (!nir_block_dominates(def->parent_instr->block, get_src_block(src))) {
+ is_valid = false;
+ break;
+ }
+ }
+
+ if (is_valid)
+ return true;
+
+ struct nir_phi_builder *pb = prep_build_phi(state);
+
+ BITSET_SET(state->def_set, def->parent_instr->block->index);
+
+ struct nir_phi_builder_value *val =
+ nir_phi_builder_add_value(pb, def->num_components, state->def_set);
+
+ nir_phi_builder_value_set_block_def(val, def->parent_instr->block, def);
+
+ nir_foreach_use_safe(def, src) {
+ nir_block *src_block = get_src_block(src);
+ if (!nir_block_dominates(def->parent_instr->block, src_block)) {
+ nir_instr_rewrite_src(src->parent_instr, src, nir_src_for_ssa(
+ nir_phi_builder_value_get_block_def(val, src_block)));
+ }
+ }
+
+ return true;
+}
+
+static bool
+repair_ssa_block(nir_block *block, void *state)
+{
+ nir_foreach_instr_safe(block, instr) {
+ nir_foreach_ssa_def(instr, repair_ssa_def, state);
+ }
+
+ return true;
+}
+
+bool
+nir_repair_ssa_impl(nir_function_impl *impl)
+{
+ struct repair_ssa_state state;
+
+ state.impl = impl;
+ state.phi_builder = NULL;
+ state.progress = false;
+
+ nir_metadata_require(impl, nir_metadata_block_index |
+ nir_metadata_dominance);
+
+ nir_foreach_block(impl, repair_ssa_block, &state);
+
+ if (state.progress)
+ nir_metadata_preserve(impl, nir_metadata_block_index |
+ nir_metadata_dominance);
+
+ if (state.phi_builder) {
+ nir_phi_builder_finish(state.phi_builder);
+ ralloc_free(state.def_set);
+ }
+
+ return state.progress;
+}
+
+/** This pass can be used to repair SSA form in a shader.
+ *
+ * Sometimes a transformation (such as return lowering) will have to make
+ * changes to a shader which, while still correct, break some of NIR's SSA
+ * invariants. This pass will insert ssa_undefs and phi nodes as needed to
+ * get the shader back into SSA that the validator will like.
+ */
+bool
+nir_repair_ssa(nir_shader *shader)
+{
+ bool progress = false;
+
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ progress = nir_repair_ssa_impl(function->impl) || progress;
+ }
+
+ return progress;
+}
{
bool progress = false;
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- progress = split_var_copies_impl(overload->impl) || progress;
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ progress = split_var_copies_impl(function->impl) || progress;
}
return progress;
sweep_function(nir_shader *nir, nir_function *f)
{
ralloc_steal(nir, f);
+ ralloc_steal(nir, f->params);
- foreach_list_typed(nir_function_overload, overload, node, &f->overload_list) {
- ralloc_steal(nir, overload);
- ralloc_steal(nir, overload->params);
- if (overload->impl)
- sweep_impl(nir, overload->impl);
- }
+ if (f->impl)
+ sweep_impl(nir, f->impl);
}
void
steal_list(nir, nir_variable, &nir->uniforms);
steal_list(nir, nir_variable, &nir->inputs);
steal_list(nir, nir_variable, &nir->outputs);
+ steal_list(nir, nir_variable, &nir->shared);
steal_list(nir, nir_variable, &nir->globals);
steal_list(nir, nir_variable, &nir->system_values);
steal_list(nir, nir_register, &nir->registers);
void
nir_convert_to_ssa(nir_shader *shader)
{
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- nir_convert_to_ssa_impl(overload->impl);
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ nir_convert_to_ssa_impl(function->impl);
}
}
return type->fields.structure[index].type;
}
+const glsl_type *
+glsl_get_function_return_type(const glsl_type *type)
+{
+ return type->fields.parameters[0].type;
+}
+
+const glsl_function_param *
+glsl_get_function_param(const glsl_type *type, unsigned index)
+{
+ return &type->fields.parameters[index + 1];
+}
+
const struct glsl_type *
glsl_get_column_type(const struct glsl_type *type)
{
return type->arrays_of_arrays_size();
}
+unsigned
+glsl_count_attribute_slots(const struct glsl_type *type,
+ bool vertex_input_slots)
+{
+ return type->count_attribute_slots(vertex_input_slots);
+}
+
const char *
glsl_get_struct_elem_name(const struct glsl_type *type, unsigned index)
{
return type->fields.structure[index].name;
}
+glsl_sampler_dim
+glsl_get_sampler_dim(const struct glsl_type *type)
+{
+ assert(glsl_type_is_sampler(type) || glsl_type_is_image(type));
+ return (glsl_sampler_dim)type->sampler_dimensionality;
+}
+
+glsl_base_type
+glsl_get_sampler_result_type(const struct glsl_type *type)
+{
+ assert(glsl_type_is_sampler(type) || glsl_type_is_image(type));
+ return (glsl_base_type)type->sampler_type;
+}
+
unsigned
glsl_get_record_location_offset(const struct glsl_type *type,
unsigned length)
}
bool
+glsl_type_is_error(const glsl_type *type)
+{
+ return type->is_error();
+}
+
+bool
glsl_type_is_vector(const struct glsl_type *type)
{
return type->is_vector();
return type->is_matrix();
}
+bool
+glsl_type_is_array(const struct glsl_type *type)
+{
+ return type->is_array();
+}
+
+bool
+glsl_type_is_struct(const struct glsl_type *type)
+{
+ return type->is_record() || type->is_interface();
+}
+
+bool
+glsl_type_is_sampler(const struct glsl_type *type)
+{
+ return type->is_sampler();
+}
+
+bool
+glsl_type_is_image(const struct glsl_type *type)
+{
+ return type->is_image();
+}
+
+bool
+glsl_sampler_type_is_shadow(const struct glsl_type *type)
+{
+ assert(glsl_type_is_sampler(type));
+ return type->sampler_shadow;
+}
+
+bool
+glsl_sampler_type_is_array(const struct glsl_type *type)
+{
+ assert(glsl_type_is_sampler(type) || glsl_type_is_image(type));
+ return type->sampler_array;
+}
+
const glsl_type *
glsl_void_type(void)
{
}
const glsl_type *
+glsl_vec_type(unsigned n)
+{
+ return glsl_type::vec(n);
+}
+
+const glsl_type *
glsl_vec4_type(void)
{
return glsl_type::vec4_type;
}
const glsl_type *
+glsl_int_type(void)
+{
+ return glsl_type::int_type;
+}
+
+const glsl_type *
glsl_uint_type(void)
{
return glsl_type::uint_type;
}
const glsl_type *
+glsl_bool_type(void)
+{
+ return glsl_type::bool_type;
+}
+
+const glsl_type *
+glsl_scalar_type(enum glsl_base_type base_type)
+{
+ return glsl_type::get_instance(base_type, 1, 1);
+}
+
+const glsl_type *
+glsl_vector_type(enum glsl_base_type base_type, unsigned components)
+{
+ assert(components > 1 && components <= 4);
+ return glsl_type::get_instance(base_type, components, 1);
+}
+
+const glsl_type *
+glsl_matrix_type(enum glsl_base_type base_type, unsigned rows, unsigned columns)
+{
+ assert(rows > 1 && rows <= 4 && columns >= 1 && columns <= 4);
+ return glsl_type::get_instance(base_type, rows, columns);
+}
+
+const glsl_type *
glsl_array_type(const glsl_type *base, unsigned elements)
{
return glsl_type::get_array_instance(base, elements);
}
+
+const glsl_type *
+glsl_struct_type(const glsl_struct_field *fields,
+ unsigned num_fields, const char *name)
+{
+ return glsl_type::get_record_instance(fields, num_fields, name);
+}
+
+const struct glsl_type *
+glsl_sampler_type(enum glsl_sampler_dim dim, bool is_shadow, bool is_array,
+ enum glsl_base_type base_type)
+{
+ return glsl_type::get_sampler_instance(dim, is_shadow, is_array, base_type);
+}
+
+const struct glsl_type *
+glsl_image_type(enum glsl_sampler_dim dim, bool is_array,
+ enum glsl_base_type base_type)
+{
+ return glsl_type::get_image_instance(dim, is_array, base_type);
+}
+
+const glsl_type *
+glsl_function_type(const glsl_type *return_type,
+ const glsl_function_param *params, unsigned num_params)
+{
+ return glsl_type::get_function_instance(return_type, params, num_params);
+}
+
+const glsl_type *
+glsl_transposed_type(const struct glsl_type *type)
+{
+ return glsl_type::get_instance(type->base_type, type->matrix_columns,
+ type->vector_elements);
+}
const struct glsl_type *glsl_get_column_type(const struct glsl_type *type);
+const struct glsl_type *
+glsl_get_function_return_type(const struct glsl_type *type);
+
+const struct glsl_function_param *
+glsl_get_function_param(const struct glsl_type *type, unsigned index);
+
enum glsl_base_type glsl_get_base_type(const struct glsl_type *type);
unsigned glsl_get_vector_elements(const struct glsl_type *type);
unsigned glsl_get_aoa_size(const struct glsl_type *type);
+unsigned glsl_count_attribute_slots(const struct glsl_type *type,
+ bool vertex_input_slots);
+
const char *glsl_get_struct_elem_name(const struct glsl_type *type,
unsigned index);
+enum glsl_sampler_dim glsl_get_sampler_dim(const struct glsl_type *type);
+enum glsl_base_type glsl_get_sampler_result_type(const struct glsl_type *type);
+
unsigned glsl_get_record_location_offset(const struct glsl_type *type,
unsigned length);
bool glsl_type_is_void(const struct glsl_type *type);
+bool glsl_type_is_error(const struct glsl_type *type);
bool glsl_type_is_vector(const struct glsl_type *type);
bool glsl_type_is_scalar(const struct glsl_type *type);
bool glsl_type_is_vector_or_scalar(const struct glsl_type *type);
bool glsl_type_is_matrix(const struct glsl_type *type);
+bool glsl_type_is_array(const struct glsl_type *type);
+bool glsl_type_is_struct(const struct glsl_type *type);
+bool glsl_type_is_sampler(const struct glsl_type *type);
+bool glsl_type_is_image(const struct glsl_type *type);
+bool glsl_sampler_type_is_shadow(const struct glsl_type *type);
+bool glsl_sampler_type_is_array(const struct glsl_type *type);
const struct glsl_type *glsl_void_type(void);
const struct glsl_type *glsl_float_type(void);
+const struct glsl_type *glsl_vec_type(unsigned n);
const struct glsl_type *glsl_vec4_type(void);
+const struct glsl_type *glsl_int_type(void);
const struct glsl_type *glsl_uint_type(void);
+const struct glsl_type *glsl_bool_type(void);
+
+const struct glsl_type *glsl_scalar_type(enum glsl_base_type base_type);
+const struct glsl_type *glsl_vector_type(enum glsl_base_type base_type,
+ unsigned components);
+const struct glsl_type *glsl_matrix_type(enum glsl_base_type base_type,
+ unsigned rows, unsigned columns);
const struct glsl_type *glsl_array_type(const struct glsl_type *base,
unsigned elements);
+const struct glsl_type *glsl_struct_type(const struct glsl_struct_field *fields,
+ unsigned num_fields, const char *name);
+const struct glsl_type *glsl_sampler_type(enum glsl_sampler_dim dim,
+ bool is_shadow, bool is_array,
+ enum glsl_base_type base_type);
+const struct glsl_type *glsl_image_type(enum glsl_sampler_dim dim,
+ bool is_array,
+ enum glsl_base_type base_type);
+const struct glsl_type * glsl_function_type(const struct glsl_type *return_type,
+ const struct glsl_function_param *params,
+ unsigned num_params);
+
+const struct glsl_type *glsl_transposed_type(const struct glsl_type *type);
#ifdef __cplusplus
}
assert(instr->variables[0]->var->data.mode != nir_var_shader_in &&
instr->variables[0]->var->data.mode != nir_var_uniform &&
instr->variables[0]->var->data.mode != nir_var_shader_storage);
+ assert((instr->const_index[0] & ~((1 << instr->num_components) - 1)) == 0);
break;
}
case nir_intrinsic_copy_var:
static void
validate_call_instr(nir_call_instr *instr, validate_state *state)
{
- if (instr->return_deref == NULL)
+ if (instr->return_deref == NULL) {
assert(glsl_type_is_void(instr->callee->return_type));
- else
+ } else {
assert(instr->return_deref->deref.type == instr->callee->return_type);
+ validate_deref_var(instr, instr->return_deref, state);
+ }
assert(instr->num_params == instr->callee->num_params);
assert(instr->callee->params[i].type == instr->params[i]->deref.type);
validate_deref_var(instr, instr->params[i], state);
}
-
- validate_deref_var(instr, instr->return_deref, state);
}
static void
static void
validate_function_impl(nir_function_impl *impl, validate_state *state)
{
- assert(impl->overload->impl == impl);
+ assert(impl->function->impl == impl);
assert(impl->cf_node.parent == NULL);
- assert(impl->num_params == impl->overload->num_params);
+ assert(impl->num_params == impl->function->num_params);
for (unsigned i = 0; i < impl->num_params; i++)
- assert(impl->params[i]->type == impl->overload->params[i].type);
+ assert(impl->params[i]->type == impl->function->params[i].type);
- if (glsl_type_is_void(impl->overload->return_type))
+ if (glsl_type_is_void(impl->function->return_type))
assert(impl->return_var == NULL);
else
- assert(impl->return_var->type == impl->overload->return_type);
+ assert(impl->return_var->type == impl->function->return_type);
assert(exec_list_is_empty(&impl->end_block->instr_list));
assert(impl->end_block->successors[0] == NULL);
}
static void
-validate_function_overload(nir_function_overload *overload,
- validate_state *state)
-{
- if (overload->impl != NULL)
- validate_function_impl(overload->impl, state);
-}
-
-static void
validate_function(nir_function *func, validate_state *state)
{
- exec_list_validate(&func->overload_list);
- foreach_list_typed(nir_function_overload, overload, node, &func->overload_list) {
- assert(overload->function == func);
- validate_function_overload(overload, state);
+ if (func->impl != NULL) {
+ assert(func->impl->function == func);
+ validate_function_impl(func->impl, state);
}
}
validate_var_decl(var, true, &state);
}
+ exec_list_validate(&shader->shared);
+ nir_foreach_variable(var, &shader->shared) {
+ validate_var_decl(var, true, &state);
+ }
+
exec_list_validate(&shader->globals);
nir_foreach_variable(var, &shader->globals) {
validate_var_decl(var, true, &state);
return NAME(stage);
}
+/**
+ * Translate a gl_shader_stage to a short shader stage name for debug
+ * printouts and error messages.
+ */
+const char * _mesa_shader_stage_to_string(unsigned stage)
+{
+ switch (stage) {
+ case MESA_SHADER_VERTEX: return "vertex";
+ case MESA_SHADER_FRAGMENT: return "fragment";
+ case MESA_SHADER_GEOMETRY: return "geometry";
+ case MESA_SHADER_COMPUTE: return "compute";
+ case MESA_SHADER_TESS_CTRL: return "tess ctrl";
+ case MESA_SHADER_TESS_EVAL: return "tess eval";
+ }
+
+ unreachable("Unknown shader stage.");
+}
+
+/**
+ * Translate a gl_shader_stage to a shader stage abbreviation (VS, GS, FS)
+ * for debug printouts and error messages.
+ */
+const char * _mesa_shader_stage_to_abbrev(unsigned stage)
+{
+ switch (stage) {
+ case MESA_SHADER_VERTEX: return "VS";
+ case MESA_SHADER_FRAGMENT: return "FS";
+ case MESA_SHADER_GEOMETRY: return "GS";
+ case MESA_SHADER_COMPUTE: return "CS";
+ case MESA_SHADER_TESS_CTRL: return "TCS";
+ case MESA_SHADER_TESS_EVAL: return "TES";
+ }
+
+ unreachable("Unknown shader stage.");
+}
+
const char * gl_vert_attrib_name(gl_vert_attrib attrib)
{
static const char *names[] = {
ENUM(SYSTEM_VALUE_TESS_LEVEL_OUTER),
ENUM(SYSTEM_VALUE_TESS_LEVEL_INNER),
ENUM(SYSTEM_VALUE_LOCAL_INVOCATION_ID),
+ ENUM(SYSTEM_VALUE_LOCAL_INVOCATION_INDEX),
+ ENUM(SYSTEM_VALUE_GLOBAL_INVOCATION_ID),
ENUM(SYSTEM_VALUE_WORK_GROUP_ID),
ENUM(SYSTEM_VALUE_NUM_WORK_GROUPS),
ENUM(SYSTEM_VALUE_VERTEX_CNT),
#ifndef SHADER_ENUMS_H
#define SHADER_ENUMS_H
+#ifdef __cplusplus
+extern "C" {
+#endif
+
/**
* Shader stages. Note that these will become 5 with tessellation.
*
const char * gl_shader_stage_name(gl_shader_stage stage);
+/**
+ * Translate a gl_shader_stage to a short shader stage name for debug
+ * printouts and error messages.
+ */
+const char * _mesa_shader_stage_to_string(unsigned stage);
+
+/**
+ * Translate a gl_shader_stage to a shader stage abbreviation (VS, GS, FS)
+ * for debug printouts and error messages.
+ */
+const char * _mesa_shader_stage_to_abbrev(unsigned stage);
+
#define MESA_SHADER_STAGES (MESA_SHADER_COMPUTE + 1)
* \sa SYSTEM_VALUE_VERTEX_ID, SYSTEM_VALUE_VERTEX_ID_ZERO_BASE
*/
SYSTEM_VALUE_BASE_VERTEX,
+
+ /**
+ * Value of \c baseinstance passed to instanced draw entry points
+ *
+ * \sa SYSTEM_VALUE_INSTANCE_ID
+ */
+ SYSTEM_VALUE_BASE_INSTANCE,
+
+ /**
+ * From _ARB_shader_draw_parameters:
+ *
+ * "Additionally, this extension adds a further built-in variable,
+ * gl_DrawID to the shading language. This variable contains the index
+ * of the draw currently being processed by a Multi* variant of a
+ * drawing command (such as MultiDrawElements or
+ * MultiDrawArraysIndirect)."
+ *
+ * If GL_ARB_multi_draw_indirect is not supported, this is always 0.
+ */
+ SYSTEM_VALUE_DRAW_ID,
/*@}*/
/**
* \name Fragment shader system values
*/
/*@{*/
- SYSTEM_VALUE_FRONT_FACE, /**< (not done yet) */
+ SYSTEM_VALUE_FRAG_COORD,
+ SYSTEM_VALUE_FRONT_FACE,
SYSTEM_VALUE_SAMPLE_ID,
SYSTEM_VALUE_SAMPLE_POS,
SYSTEM_VALUE_SAMPLE_MASK_IN,
*/
/*@{*/
SYSTEM_VALUE_LOCAL_INVOCATION_ID,
+ SYSTEM_VALUE_LOCAL_INVOCATION_INDEX,
+ SYSTEM_VALUE_GLOBAL_INVOCATION_ID,
SYSTEM_VALUE_WORK_GROUP_ID,
SYSTEM_VALUE_NUM_WORK_GROUPS,
/*@}*/
FRAG_DEPTH_LAYOUT_UNCHANGED
};
+#ifdef __cplusplus
+} /* extern "C" */
+#endif
+
#endif /* SHADER_ENUMS_H */
--- /dev/null
+/*
+** Copyright (c) 2014-2015 The Khronos Group Inc.
+**
+** Permission is hereby granted, free of charge, to any person obtaining a copy
+** of this software and/or associated documentation files (the "Materials"),
+** to deal in the Materials without restriction, including without limitation
+** the rights to use, copy, modify, merge, publish, distribute, sublicense,
+** and/or sell copies of the Materials, and to permit persons to whom the
+** Materials are furnished to do so, subject to the following conditions:
+**
+** The above copyright notice and this permission notice shall be included in
+** all copies or substantial portions of the Materials.
+**
+** MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS KHRONOS
+** STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS SPECIFICATIONS AND
+** HEADER INFORMATION ARE LOCATED AT https://www.khronos.org/registry/
+**
+** THE MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+** OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+** FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+** THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+** LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+** FROM,OUT OF OR IN CONNECTION WITH THE MATERIALS OR THE USE OR OTHER DEALINGS
+** IN THE MATERIALS.
+*/
+
+#ifndef GLSLstd450_H
+#define GLSLstd450_H
+
+const int GLSLstd450Version = 99;
+const int GLSLstd450Revision = 3;
+
+enum GLSLstd450 {
+ GLSLstd450Bad = 0, // Don't use
+
+ GLSLstd450Round = 1,
+ GLSLstd450RoundEven = 2,
+ GLSLstd450Trunc = 3,
+ GLSLstd450FAbs = 4,
+ GLSLstd450SAbs = 5,
+ GLSLstd450FSign = 6,
+ GLSLstd450SSign = 7,
+ GLSLstd450Floor = 8,
+ GLSLstd450Ceil = 9,
+ GLSLstd450Fract = 10,
+
+ GLSLstd450Radians = 11,
+ GLSLstd450Degrees = 12,
+ GLSLstd450Sin = 13,
+ GLSLstd450Cos = 14,
+ GLSLstd450Tan = 15,
+ GLSLstd450Asin = 16,
+ GLSLstd450Acos = 17,
+ GLSLstd450Atan = 18,
+ GLSLstd450Sinh = 19,
+ GLSLstd450Cosh = 20,
+ GLSLstd450Tanh = 21,
+ GLSLstd450Asinh = 22,
+ GLSLstd450Acosh = 23,
+ GLSLstd450Atanh = 24,
+ GLSLstd450Atan2 = 25,
+
+ GLSLstd450Pow = 26,
+ GLSLstd450Exp = 27,
+ GLSLstd450Log = 28,
+ GLSLstd450Exp2 = 29,
+ GLSLstd450Log2 = 30,
+ GLSLstd450Sqrt = 31,
+ GLSLstd450InverseSqrt = 32,
+
+ GLSLstd450Determinant = 33,
+ GLSLstd450MatrixInverse = 34,
+
+ GLSLstd450Modf = 35, // second operand needs an OpVariable to write to
+ GLSLstd450ModfStruct = 36, // no OpVariable operand
+ GLSLstd450FMin = 37,
+ GLSLstd450UMin = 38,
+ GLSLstd450SMin = 39,
+ GLSLstd450FMax = 40,
+ GLSLstd450UMax = 41,
+ GLSLstd450SMax = 42,
+ GLSLstd450FClamp = 43,
+ GLSLstd450UClamp = 44,
+ GLSLstd450SClamp = 45,
+ GLSLstd450FMix = 46,
+ GLSLstd450IMix = 47,
+ GLSLstd450Step = 48,
+ GLSLstd450SmoothStep = 49,
+
+ GLSLstd450Fma = 50,
+ GLSLstd450Frexp = 51, // second operand needs an OpVariable to write to
+ GLSLstd450FrexpStruct = 52, // no OpVariable operand
+ GLSLstd450Ldexp = 53,
+
+ GLSLstd450PackSnorm4x8 = 54,
+ GLSLstd450PackUnorm4x8 = 55,
+ GLSLstd450PackSnorm2x16 = 56,
+ GLSLstd450PackUnorm2x16 = 57,
+ GLSLstd450PackHalf2x16 = 58,
+ GLSLstd450PackDouble2x32 = 59,
+ GLSLstd450UnpackSnorm2x16 = 60,
+ GLSLstd450UnpackUnorm2x16 = 61,
+ GLSLstd450UnpackHalf2x16 = 62,
+ GLSLstd450UnpackSnorm4x8 = 63,
+ GLSLstd450UnpackUnorm4x8 = 64,
+ GLSLstd450UnpackDouble2x32 = 65,
+
+ GLSLstd450Length = 66,
+ GLSLstd450Distance = 67,
+ GLSLstd450Cross = 68,
+ GLSLstd450Normalize = 69,
+ GLSLstd450FaceForward = 70,
+ GLSLstd450Reflect = 71,
+ GLSLstd450Refract = 72,
+
+ GLSLstd450FindILsb = 73,
+ GLSLstd450FindSMsb = 74,
+ GLSLstd450FindUMsb = 75,
+
+ GLSLstd450InterpolateAtCentroid = 76,
+ GLSLstd450InterpolateAtSample = 77,
+ GLSLstd450InterpolateAtOffset = 78,
+
+ GLSLstd450Count
+};
+
+#endif // #ifndef GLSLstd450_H
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ * Authors:
+ * Jason Ekstrand (jason@jlekstrand.net)
+ *
+ */
+
+#pragma once
+
+#ifndef _NIR_SPIRV_H_
+#define _NIR_SPIRV_H_
+
+#include "nir.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+struct nir_spirv_specialization {
+ uint32_t id;
+ uint32_t data;
+};
+
+nir_function *spirv_to_nir(const uint32_t *words, size_t word_count,
+ struct nir_spirv_specialization *specializations,
+ unsigned num_specializations,
+ gl_shader_stage stage, const char *entry_point_name,
+ const nir_shader_compiler_options *options);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _NIR_SPIRV_H_ */
--- /dev/null
+/*
+** Copyright (c) 2014-2015 The Khronos Group Inc.
+**
+** Permission is hereby granted, free of charge, to any person obtaining a copy
+** of this software and/or associated documentation files (the "Materials"),
+** to deal in the Materials without restriction, including without limitation
+** the rights to use, copy, modify, merge, publish, distribute, sublicense,
+** and/or sell copies of the Materials, and to permit persons to whom the
+** Materials are furnished to do so, subject to the following conditions:
+**
+** The above copyright notice and this permission notice shall be included in
+** all copies or substantial portions of the Materials.
+**
+** MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS KHRONOS
+** STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS SPECIFICATIONS AND
+** HEADER INFORMATION ARE LOCATED AT https://www.khronos.org/registry/
+**
+** THE MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+** OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+** FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+** THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+** LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+** FROM,OUT OF OR IN CONNECTION WITH THE MATERIALS OR THE USE OR OTHER DEALINGS
+** IN THE MATERIALS.
+*/
+
+/*
+** This header is automatically generated by the same tool that creates
+** the Binary Section of the SPIR-V specification.
+*/
+
+/*
+** Enumeration tokens for SPIR-V, in various styles:
+** C, C++, C++11, JSON, Lua, Python
+**
+** - C will have tokens with a "Spv" prefix, e.g.: SpvSourceLanguageGLSL
+** - C++ will have tokens in the "spv" name space, e.g.: spv::SourceLanguageGLSL
+** - C++11 will use enum classes in the spv namespace, e.g.: spv::SourceLanguage::GLSL
+** - Lua will use tables, e.g.: spv.SourceLanguage.GLSL
+** - Python will use dictionaries, e.g.: spv['SourceLanguage']['GLSL']
+**
+** Some tokens act like mask values, which can be OR'd together,
+** while others are mutually exclusive. The mask-like ones have
+** "Mask" in their name, and a parallel enum that has the shift
+** amount (1 << x) for each corresponding enumerant.
+*/
+
+#ifndef spirv_H
+#define spirv_H
+
+typedef unsigned int SpvId;
+
+#define SPV_VERSION 0x10000
+#define SPV_REVISION 2
+
+static const unsigned int SpvMagicNumber = 0x07230203;
+static const unsigned int SpvVersion = 0x00010000;
+static const unsigned int SpvRevision = 2;
+static const unsigned int SpvOpCodeMask = 0xffff;
+static const unsigned int SpvWordCountShift = 16;
+
+typedef enum SpvSourceLanguage_ {
+ SpvSourceLanguageUnknown = 0,
+ SpvSourceLanguageESSL = 1,
+ SpvSourceLanguageGLSL = 2,
+ SpvSourceLanguageOpenCL_C = 3,
+ SpvSourceLanguageOpenCL_CPP = 4,
+} SpvSourceLanguage;
+
+typedef enum SpvExecutionModel_ {
+ SpvExecutionModelVertex = 0,
+ SpvExecutionModelTessellationControl = 1,
+ SpvExecutionModelTessellationEvaluation = 2,
+ SpvExecutionModelGeometry = 3,
+ SpvExecutionModelFragment = 4,
+ SpvExecutionModelGLCompute = 5,
+ SpvExecutionModelKernel = 6,
+} SpvExecutionModel;
+
+typedef enum SpvAddressingModel_ {
+ SpvAddressingModelLogical = 0,
+ SpvAddressingModelPhysical32 = 1,
+ SpvAddressingModelPhysical64 = 2,
+} SpvAddressingModel;
+
+typedef enum SpvMemoryModel_ {
+ SpvMemoryModelSimple = 0,
+ SpvMemoryModelGLSL450 = 1,
+ SpvMemoryModelOpenCL = 2,
+} SpvMemoryModel;
+
+typedef enum SpvExecutionMode_ {
+ SpvExecutionModeInvocations = 0,
+ SpvExecutionModeSpacingEqual = 1,
+ SpvExecutionModeSpacingFractionalEven = 2,
+ SpvExecutionModeSpacingFractionalOdd = 3,
+ SpvExecutionModeVertexOrderCw = 4,
+ SpvExecutionModeVertexOrderCcw = 5,
+ SpvExecutionModePixelCenterInteger = 6,
+ SpvExecutionModeOriginUpperLeft = 7,
+ SpvExecutionModeOriginLowerLeft = 8,
+ SpvExecutionModeEarlyFragmentTests = 9,
+ SpvExecutionModePointMode = 10,
+ SpvExecutionModeXfb = 11,
+ SpvExecutionModeDepthReplacing = 12,
+ SpvExecutionModeDepthGreater = 14,
+ SpvExecutionModeDepthLess = 15,
+ SpvExecutionModeDepthUnchanged = 16,
+ SpvExecutionModeLocalSize = 17,
+ SpvExecutionModeLocalSizeHint = 18,
+ SpvExecutionModeInputPoints = 19,
+ SpvExecutionModeInputLines = 20,
+ SpvExecutionModeInputLinesAdjacency = 21,
+ SpvExecutionModeTriangles = 22,
+ SpvExecutionModeInputTrianglesAdjacency = 23,
+ SpvExecutionModeQuads = 24,
+ SpvExecutionModeIsolines = 25,
+ SpvExecutionModeOutputVertices = 26,
+ SpvExecutionModeOutputPoints = 27,
+ SpvExecutionModeOutputLineStrip = 28,
+ SpvExecutionModeOutputTriangleStrip = 29,
+ SpvExecutionModeVecTypeHint = 30,
+ SpvExecutionModeContractionOff = 31,
+} SpvExecutionMode;
+
+typedef enum SpvStorageClass_ {
+ SpvStorageClassUniformConstant = 0,
+ SpvStorageClassInput = 1,
+ SpvStorageClassUniform = 2,
+ SpvStorageClassOutput = 3,
+ SpvStorageClassWorkgroup = 4,
+ SpvStorageClassCrossWorkgroup = 5,
+ SpvStorageClassPrivate = 6,
+ SpvStorageClassFunction = 7,
+ SpvStorageClassGeneric = 8,
+ SpvStorageClassPushConstant = 9,
+ SpvStorageClassAtomicCounter = 10,
+ SpvStorageClassImage = 11,
+} SpvStorageClass;
+
+typedef enum SpvDim_ {
+ SpvDim1D = 0,
+ SpvDim2D = 1,
+ SpvDim3D = 2,
+ SpvDimCube = 3,
+ SpvDimRect = 4,
+ SpvDimBuffer = 5,
+ SpvDimSubpassData = 6,
+} SpvDim;
+
+typedef enum SpvSamplerAddressingMode_ {
+ SpvSamplerAddressingModeNone = 0,
+ SpvSamplerAddressingModeClampToEdge = 1,
+ SpvSamplerAddressingModeClamp = 2,
+ SpvSamplerAddressingModeRepeat = 3,
+ SpvSamplerAddressingModeRepeatMirrored = 4,
+} SpvSamplerAddressingMode;
+
+typedef enum SpvSamplerFilterMode_ {
+ SpvSamplerFilterModeNearest = 0,
+ SpvSamplerFilterModeLinear = 1,
+} SpvSamplerFilterMode;
+
+typedef enum SpvImageFormat_ {
+ SpvImageFormatUnknown = 0,
+ SpvImageFormatRgba32f = 1,
+ SpvImageFormatRgba16f = 2,
+ SpvImageFormatR32f = 3,
+ SpvImageFormatRgba8 = 4,
+ SpvImageFormatRgba8Snorm = 5,
+ SpvImageFormatRg32f = 6,
+ SpvImageFormatRg16f = 7,
+ SpvImageFormatR11fG11fB10f = 8,
+ SpvImageFormatR16f = 9,
+ SpvImageFormatRgba16 = 10,
+ SpvImageFormatRgb10A2 = 11,
+ SpvImageFormatRg16 = 12,
+ SpvImageFormatRg8 = 13,
+ SpvImageFormatR16 = 14,
+ SpvImageFormatR8 = 15,
+ SpvImageFormatRgba16Snorm = 16,
+ SpvImageFormatRg16Snorm = 17,
+ SpvImageFormatRg8Snorm = 18,
+ SpvImageFormatR16Snorm = 19,
+ SpvImageFormatR8Snorm = 20,
+ SpvImageFormatRgba32i = 21,
+ SpvImageFormatRgba16i = 22,
+ SpvImageFormatRgba8i = 23,
+ SpvImageFormatR32i = 24,
+ SpvImageFormatRg32i = 25,
+ SpvImageFormatRg16i = 26,
+ SpvImageFormatRg8i = 27,
+ SpvImageFormatR16i = 28,
+ SpvImageFormatR8i = 29,
+ SpvImageFormatRgba32ui = 30,
+ SpvImageFormatRgba16ui = 31,
+ SpvImageFormatRgba8ui = 32,
+ SpvImageFormatR32ui = 33,
+ SpvImageFormatRgb10a2ui = 34,
+ SpvImageFormatRg32ui = 35,
+ SpvImageFormatRg16ui = 36,
+ SpvImageFormatRg8ui = 37,
+ SpvImageFormatR16ui = 38,
+ SpvImageFormatR8ui = 39,
+} SpvImageFormat;
+
+typedef enum SpvImageChannelOrder_ {
+ SpvImageChannelOrderR = 0,
+ SpvImageChannelOrderA = 1,
+ SpvImageChannelOrderRG = 2,
+ SpvImageChannelOrderRA = 3,
+ SpvImageChannelOrderRGB = 4,
+ SpvImageChannelOrderRGBA = 5,
+ SpvImageChannelOrderBGRA = 6,
+ SpvImageChannelOrderARGB = 7,
+ SpvImageChannelOrderIntensity = 8,
+ SpvImageChannelOrderLuminance = 9,
+ SpvImageChannelOrderRx = 10,
+ SpvImageChannelOrderRGx = 11,
+ SpvImageChannelOrderRGBx = 12,
+ SpvImageChannelOrderDepth = 13,
+ SpvImageChannelOrderDepthStencil = 14,
+ SpvImageChannelOrdersRGB = 15,
+ SpvImageChannelOrdersRGBx = 16,
+ SpvImageChannelOrdersRGBA = 17,
+ SpvImageChannelOrdersBGRA = 18,
+} SpvImageChannelOrder;
+
+typedef enum SpvImageChannelDataType_ {
+ SpvImageChannelDataTypeSnormInt8 = 0,
+ SpvImageChannelDataTypeSnormInt16 = 1,
+ SpvImageChannelDataTypeUnormInt8 = 2,
+ SpvImageChannelDataTypeUnormInt16 = 3,
+ SpvImageChannelDataTypeUnormShort565 = 4,
+ SpvImageChannelDataTypeUnormShort555 = 5,
+ SpvImageChannelDataTypeUnormInt101010 = 6,
+ SpvImageChannelDataTypeSignedInt8 = 7,
+ SpvImageChannelDataTypeSignedInt16 = 8,
+ SpvImageChannelDataTypeSignedInt32 = 9,
+ SpvImageChannelDataTypeUnsignedInt8 = 10,
+ SpvImageChannelDataTypeUnsignedInt16 = 11,
+ SpvImageChannelDataTypeUnsignedInt32 = 12,
+ SpvImageChannelDataTypeHalfFloat = 13,
+ SpvImageChannelDataTypeFloat = 14,
+ SpvImageChannelDataTypeUnormInt24 = 15,
+ SpvImageChannelDataTypeUnormInt101010_2 = 16,
+} SpvImageChannelDataType;
+
+typedef enum SpvImageOperandsShift_ {
+ SpvImageOperandsBiasShift = 0,
+ SpvImageOperandsLodShift = 1,
+ SpvImageOperandsGradShift = 2,
+ SpvImageOperandsConstOffsetShift = 3,
+ SpvImageOperandsOffsetShift = 4,
+ SpvImageOperandsConstOffsetsShift = 5,
+ SpvImageOperandsSampleShift = 6,
+ SpvImageOperandsMinLodShift = 7,
+} SpvImageOperandsShift;
+
+typedef enum SpvImageOperandsMask_ {
+ SpvImageOperandsMaskNone = 0,
+ SpvImageOperandsBiasMask = 0x00000001,
+ SpvImageOperandsLodMask = 0x00000002,
+ SpvImageOperandsGradMask = 0x00000004,
+ SpvImageOperandsConstOffsetMask = 0x00000008,
+ SpvImageOperandsOffsetMask = 0x00000010,
+ SpvImageOperandsConstOffsetsMask = 0x00000020,
+ SpvImageOperandsSampleMask = 0x00000040,
+ SpvImageOperandsMinLodMask = 0x00000080,
+} SpvImageOperandsMask;
+
+typedef enum SpvFPFastMathModeShift_ {
+ SpvFPFastMathModeNotNaNShift = 0,
+ SpvFPFastMathModeNotInfShift = 1,
+ SpvFPFastMathModeNSZShift = 2,
+ SpvFPFastMathModeAllowRecipShift = 3,
+ SpvFPFastMathModeFastShift = 4,
+} SpvFPFastMathModeShift;
+
+typedef enum SpvFPFastMathModeMask_ {
+ SpvFPFastMathModeMaskNone = 0,
+ SpvFPFastMathModeNotNaNMask = 0x00000001,
+ SpvFPFastMathModeNotInfMask = 0x00000002,
+ SpvFPFastMathModeNSZMask = 0x00000004,
+ SpvFPFastMathModeAllowRecipMask = 0x00000008,
+ SpvFPFastMathModeFastMask = 0x00000010,
+} SpvFPFastMathModeMask;
+
+typedef enum SpvFPRoundingMode_ {
+ SpvFPRoundingModeRTE = 0,
+ SpvFPRoundingModeRTZ = 1,
+ SpvFPRoundingModeRTP = 2,
+ SpvFPRoundingModeRTN = 3,
+} SpvFPRoundingMode;
+
+typedef enum SpvLinkageType_ {
+ SpvLinkageTypeExport = 0,
+ SpvLinkageTypeImport = 1,
+} SpvLinkageType;
+
+typedef enum SpvAccessQualifier_ {
+ SpvAccessQualifierReadOnly = 0,
+ SpvAccessQualifierWriteOnly = 1,
+ SpvAccessQualifierReadWrite = 2,
+} SpvAccessQualifier;
+
+typedef enum SpvFunctionParameterAttribute_ {
+ SpvFunctionParameterAttributeZext = 0,
+ SpvFunctionParameterAttributeSext = 1,
+ SpvFunctionParameterAttributeByVal = 2,
+ SpvFunctionParameterAttributeSret = 3,
+ SpvFunctionParameterAttributeNoAlias = 4,
+ SpvFunctionParameterAttributeNoCapture = 5,
+ SpvFunctionParameterAttributeNoWrite = 6,
+ SpvFunctionParameterAttributeNoReadWrite = 7,
+} SpvFunctionParameterAttribute;
+
+typedef enum SpvDecoration_ {
+ SpvDecorationRelaxedPrecision = 0,
+ SpvDecorationSpecId = 1,
+ SpvDecorationBlock = 2,
+ SpvDecorationBufferBlock = 3,
+ SpvDecorationRowMajor = 4,
+ SpvDecorationColMajor = 5,
+ SpvDecorationArrayStride = 6,
+ SpvDecorationMatrixStride = 7,
+ SpvDecorationGLSLShared = 8,
+ SpvDecorationGLSLPacked = 9,
+ SpvDecorationCPacked = 10,
+ SpvDecorationBuiltIn = 11,
+ SpvDecorationNoPerspective = 13,
+ SpvDecorationFlat = 14,
+ SpvDecorationPatch = 15,
+ SpvDecorationCentroid = 16,
+ SpvDecorationSample = 17,
+ SpvDecorationInvariant = 18,
+ SpvDecorationRestrict = 19,
+ SpvDecorationAliased = 20,
+ SpvDecorationVolatile = 21,
+ SpvDecorationConstant = 22,
+ SpvDecorationCoherent = 23,
+ SpvDecorationNonWritable = 24,
+ SpvDecorationNonReadable = 25,
+ SpvDecorationUniform = 26,
+ SpvDecorationSaturatedConversion = 28,
+ SpvDecorationStream = 29,
+ SpvDecorationLocation = 30,
+ SpvDecorationComponent = 31,
+ SpvDecorationIndex = 32,
+ SpvDecorationBinding = 33,
+ SpvDecorationDescriptorSet = 34,
+ SpvDecorationOffset = 35,
+ SpvDecorationXfbBuffer = 36,
+ SpvDecorationXfbStride = 37,
+ SpvDecorationFuncParamAttr = 38,
+ SpvDecorationFPRoundingMode = 39,
+ SpvDecorationFPFastMathMode = 40,
+ SpvDecorationLinkageAttributes = 41,
+ SpvDecorationNoContraction = 42,
+ SpvDecorationInputAttachmentIndex = 43,
+ SpvDecorationAlignment = 44,
+} SpvDecoration;
+
+typedef enum SpvBuiltIn_ {
+ SpvBuiltInPosition = 0,
+ SpvBuiltInPointSize = 1,
+ SpvBuiltInClipDistance = 3,
+ SpvBuiltInCullDistance = 4,
+ SpvBuiltInVertexId = 5,
+ SpvBuiltInInstanceId = 6,
+ SpvBuiltInPrimitiveId = 7,
+ SpvBuiltInInvocationId = 8,
+ SpvBuiltInLayer = 9,
+ SpvBuiltInViewportIndex = 10,
+ SpvBuiltInTessLevelOuter = 11,
+ SpvBuiltInTessLevelInner = 12,
+ SpvBuiltInTessCoord = 13,
+ SpvBuiltInPatchVertices = 14,
+ SpvBuiltInFragCoord = 15,
+ SpvBuiltInPointCoord = 16,
+ SpvBuiltInFrontFacing = 17,
+ SpvBuiltInSampleId = 18,
+ SpvBuiltInSamplePosition = 19,
+ SpvBuiltInSampleMask = 20,
+ SpvBuiltInFragDepth = 22,
+ SpvBuiltInHelperInvocation = 23,
+ SpvBuiltInNumWorkgroups = 24,
+ SpvBuiltInWorkgroupSize = 25,
+ SpvBuiltInWorkgroupId = 26,
+ SpvBuiltInLocalInvocationId = 27,
+ SpvBuiltInGlobalInvocationId = 28,
+ SpvBuiltInLocalInvocationIndex = 29,
+ SpvBuiltInWorkDim = 30,
+ SpvBuiltInGlobalSize = 31,
+ SpvBuiltInEnqueuedWorkgroupSize = 32,
+ SpvBuiltInGlobalOffset = 33,
+ SpvBuiltInGlobalLinearId = 34,
+ SpvBuiltInSubgroupSize = 36,
+ SpvBuiltInSubgroupMaxSize = 37,
+ SpvBuiltInNumSubgroups = 38,
+ SpvBuiltInNumEnqueuedSubgroups = 39,
+ SpvBuiltInSubgroupId = 40,
+ SpvBuiltInSubgroupLocalInvocationId = 41,
+ SpvBuiltInVertexIndex = 42,
+ SpvBuiltInInstanceIndex = 43,
+} SpvBuiltIn;
+
+typedef enum SpvSelectionControlShift_ {
+ SpvSelectionControlFlattenShift = 0,
+ SpvSelectionControlDontFlattenShift = 1,
+} SpvSelectionControlShift;
+
+typedef enum SpvSelectionControlMask_ {
+ SpvSelectionControlMaskNone = 0,
+ SpvSelectionControlFlattenMask = 0x00000001,
+ SpvSelectionControlDontFlattenMask = 0x00000002,
+} SpvSelectionControlMask;
+
+typedef enum SpvLoopControlShift_ {
+ SpvLoopControlUnrollShift = 0,
+ SpvLoopControlDontUnrollShift = 1,
+} SpvLoopControlShift;
+
+typedef enum SpvLoopControlMask_ {
+ SpvLoopControlMaskNone = 0,
+ SpvLoopControlUnrollMask = 0x00000001,
+ SpvLoopControlDontUnrollMask = 0x00000002,
+} SpvLoopControlMask;
+
+typedef enum SpvFunctionControlShift_ {
+ SpvFunctionControlInlineShift = 0,
+ SpvFunctionControlDontInlineShift = 1,
+ SpvFunctionControlPureShift = 2,
+ SpvFunctionControlConstShift = 3,
+} SpvFunctionControlShift;
+
+typedef enum SpvFunctionControlMask_ {
+ SpvFunctionControlMaskNone = 0,
+ SpvFunctionControlInlineMask = 0x00000001,
+ SpvFunctionControlDontInlineMask = 0x00000002,
+ SpvFunctionControlPureMask = 0x00000004,
+ SpvFunctionControlConstMask = 0x00000008,
+} SpvFunctionControlMask;
+
+typedef enum SpvMemorySemanticsShift_ {
+ SpvMemorySemanticsAcquireShift = 1,
+ SpvMemorySemanticsReleaseShift = 2,
+ SpvMemorySemanticsAcquireReleaseShift = 3,
+ SpvMemorySemanticsSequentiallyConsistentShift = 4,
+ SpvMemorySemanticsUniformMemoryShift = 6,
+ SpvMemorySemanticsSubgroupMemoryShift = 7,
+ SpvMemorySemanticsWorkgroupMemoryShift = 8,
+ SpvMemorySemanticsCrossWorkgroupMemoryShift = 9,
+ SpvMemorySemanticsAtomicCounterMemoryShift = 10,
+ SpvMemorySemanticsImageMemoryShift = 11,
+} SpvMemorySemanticsShift;
+
+typedef enum SpvMemorySemanticsMask_ {
+ SpvMemorySemanticsMaskNone = 0,
+ SpvMemorySemanticsAcquireMask = 0x00000002,
+ SpvMemorySemanticsReleaseMask = 0x00000004,
+ SpvMemorySemanticsAcquireReleaseMask = 0x00000008,
+ SpvMemorySemanticsSequentiallyConsistentMask = 0x00000010,
+ SpvMemorySemanticsUniformMemoryMask = 0x00000040,
+ SpvMemorySemanticsSubgroupMemoryMask = 0x00000080,
+ SpvMemorySemanticsWorkgroupMemoryMask = 0x00000100,
+ SpvMemorySemanticsCrossWorkgroupMemoryMask = 0x00000200,
+ SpvMemorySemanticsAtomicCounterMemoryMask = 0x00000400,
+ SpvMemorySemanticsImageMemoryMask = 0x00000800,
+} SpvMemorySemanticsMask;
+
+typedef enum SpvMemoryAccessShift_ {
+ SpvMemoryAccessVolatileShift = 0,
+ SpvMemoryAccessAlignedShift = 1,
+ SpvMemoryAccessNontemporalShift = 2,
+} SpvMemoryAccessShift;
+
+typedef enum SpvMemoryAccessMask_ {
+ SpvMemoryAccessMaskNone = 0,
+ SpvMemoryAccessVolatileMask = 0x00000001,
+ SpvMemoryAccessAlignedMask = 0x00000002,
+ SpvMemoryAccessNontemporalMask = 0x00000004,
+} SpvMemoryAccessMask;
+
+typedef enum SpvScope_ {
+ SpvScopeCrossDevice = 0,
+ SpvScopeDevice = 1,
+ SpvScopeWorkgroup = 2,
+ SpvScopeSubgroup = 3,
+ SpvScopeInvocation = 4,
+} SpvScope;
+
+typedef enum SpvGroupOperation_ {
+ SpvGroupOperationReduce = 0,
+ SpvGroupOperationInclusiveScan = 1,
+ SpvGroupOperationExclusiveScan = 2,
+} SpvGroupOperation;
+
+typedef enum SpvKernelEnqueueFlags_ {
+ SpvKernelEnqueueFlagsNoWait = 0,
+ SpvKernelEnqueueFlagsWaitKernel = 1,
+ SpvKernelEnqueueFlagsWaitWorkGroup = 2,
+} SpvKernelEnqueueFlags;
+
+typedef enum SpvKernelProfilingInfoShift_ {
+ SpvKernelProfilingInfoCmdExecTimeShift = 0,
+} SpvKernelProfilingInfoShift;
+
+typedef enum SpvKernelProfilingInfoMask_ {
+ SpvKernelProfilingInfoMaskNone = 0,
+ SpvKernelProfilingInfoCmdExecTimeMask = 0x00000001,
+} SpvKernelProfilingInfoMask;
+
+typedef enum SpvCapability_ {
+ SpvCapabilityMatrix = 0,
+ SpvCapabilityShader = 1,
+ SpvCapabilityGeometry = 2,
+ SpvCapabilityTessellation = 3,
+ SpvCapabilityAddresses = 4,
+ SpvCapabilityLinkage = 5,
+ SpvCapabilityKernel = 6,
+ SpvCapabilityVector16 = 7,
+ SpvCapabilityFloat16Buffer = 8,
+ SpvCapabilityFloat16 = 9,
+ SpvCapabilityFloat64 = 10,
+ SpvCapabilityInt64 = 11,
+ SpvCapabilityInt64Atomics = 12,
+ SpvCapabilityImageBasic = 13,
+ SpvCapabilityImageReadWrite = 14,
+ SpvCapabilityImageMipmap = 15,
+ SpvCapabilityPipes = 17,
+ SpvCapabilityGroups = 18,
+ SpvCapabilityDeviceEnqueue = 19,
+ SpvCapabilityLiteralSampler = 20,
+ SpvCapabilityAtomicStorage = 21,
+ SpvCapabilityInt16 = 22,
+ SpvCapabilityTessellationPointSize = 23,
+ SpvCapabilityGeometryPointSize = 24,
+ SpvCapabilityImageGatherExtended = 25,
+ SpvCapabilityStorageImageMultisample = 27,
+ SpvCapabilityUniformBufferArrayDynamicIndexing = 28,
+ SpvCapabilitySampledImageArrayDynamicIndexing = 29,
+ SpvCapabilityStorageBufferArrayDynamicIndexing = 30,
+ SpvCapabilityStorageImageArrayDynamicIndexing = 31,
+ SpvCapabilityClipDistance = 32,
+ SpvCapabilityCullDistance = 33,
+ SpvCapabilityImageCubeArray = 34,
+ SpvCapabilitySampleRateShading = 35,
+ SpvCapabilityImageRect = 36,
+ SpvCapabilitySampledRect = 37,
+ SpvCapabilityGenericPointer = 38,
+ SpvCapabilityInt8 = 39,
+ SpvCapabilityInputAttachment = 40,
+ SpvCapabilitySparseResidency = 41,
+ SpvCapabilityMinLod = 42,
+ SpvCapabilitySampled1D = 43,
+ SpvCapabilityImage1D = 44,
+ SpvCapabilitySampledCubeArray = 45,
+ SpvCapabilitySampledBuffer = 46,
+ SpvCapabilityImageBuffer = 47,
+ SpvCapabilityImageMSArray = 48,
+ SpvCapabilityStorageImageExtendedFormats = 49,
+ SpvCapabilityImageQuery = 50,
+ SpvCapabilityDerivativeControl = 51,
+ SpvCapabilityInterpolationFunction = 52,
+ SpvCapabilityTransformFeedback = 53,
+ SpvCapabilityGeometryStreams = 54,
+ SpvCapabilityStorageImageReadWithoutFormat = 55,
+ SpvCapabilityStorageImageWriteWithoutFormat = 56,
+ SpvCapabilityMultiViewport = 57,
+} SpvCapability;
+
+typedef enum SpvOp_ {
+ SpvOpNop = 0,
+ SpvOpUndef = 1,
+ SpvOpSourceContinued = 2,
+ SpvOpSource = 3,
+ SpvOpSourceExtension = 4,
+ SpvOpName = 5,
+ SpvOpMemberName = 6,
+ SpvOpString = 7,
+ SpvOpLine = 8,
+ SpvOpExtension = 10,
+ SpvOpExtInstImport = 11,
+ SpvOpExtInst = 12,
+ SpvOpMemoryModel = 14,
+ SpvOpEntryPoint = 15,
+ SpvOpExecutionMode = 16,
+ SpvOpCapability = 17,
+ SpvOpTypeVoid = 19,
+ SpvOpTypeBool = 20,
+ SpvOpTypeInt = 21,
+ SpvOpTypeFloat = 22,
+ SpvOpTypeVector = 23,
+ SpvOpTypeMatrix = 24,
+ SpvOpTypeImage = 25,
+ SpvOpTypeSampler = 26,
+ SpvOpTypeSampledImage = 27,
+ SpvOpTypeArray = 28,
+ SpvOpTypeRuntimeArray = 29,
+ SpvOpTypeStruct = 30,
+ SpvOpTypeOpaque = 31,
+ SpvOpTypePointer = 32,
+ SpvOpTypeFunction = 33,
+ SpvOpTypeEvent = 34,
+ SpvOpTypeDeviceEvent = 35,
+ SpvOpTypeReserveId = 36,
+ SpvOpTypeQueue = 37,
+ SpvOpTypePipe = 38,
+ SpvOpTypeForwardPointer = 39,
+ SpvOpConstantTrue = 41,
+ SpvOpConstantFalse = 42,
+ SpvOpConstant = 43,
+ SpvOpConstantComposite = 44,
+ SpvOpConstantSampler = 45,
+ SpvOpConstantNull = 46,
+ SpvOpSpecConstantTrue = 48,
+ SpvOpSpecConstantFalse = 49,
+ SpvOpSpecConstant = 50,
+ SpvOpSpecConstantComposite = 51,
+ SpvOpSpecConstantOp = 52,
+ SpvOpFunction = 54,
+ SpvOpFunctionParameter = 55,
+ SpvOpFunctionEnd = 56,
+ SpvOpFunctionCall = 57,
+ SpvOpVariable = 59,
+ SpvOpImageTexelPointer = 60,
+ SpvOpLoad = 61,
+ SpvOpStore = 62,
+ SpvOpCopyMemory = 63,
+ SpvOpCopyMemorySized = 64,
+ SpvOpAccessChain = 65,
+ SpvOpInBoundsAccessChain = 66,
+ SpvOpPtrAccessChain = 67,
+ SpvOpArrayLength = 68,
+ SpvOpGenericPtrMemSemantics = 69,
+ SpvOpInBoundsPtrAccessChain = 70,
+ SpvOpDecorate = 71,
+ SpvOpMemberDecorate = 72,
+ SpvOpDecorationGroup = 73,
+ SpvOpGroupDecorate = 74,
+ SpvOpGroupMemberDecorate = 75,
+ SpvOpVectorExtractDynamic = 77,
+ SpvOpVectorInsertDynamic = 78,
+ SpvOpVectorShuffle = 79,
+ SpvOpCompositeConstruct = 80,
+ SpvOpCompositeExtract = 81,
+ SpvOpCompositeInsert = 82,
+ SpvOpCopyObject = 83,
+ SpvOpTranspose = 84,
+ SpvOpSampledImage = 86,
+ SpvOpImageSampleImplicitLod = 87,
+ SpvOpImageSampleExplicitLod = 88,
+ SpvOpImageSampleDrefImplicitLod = 89,
+ SpvOpImageSampleDrefExplicitLod = 90,
+ SpvOpImageSampleProjImplicitLod = 91,
+ SpvOpImageSampleProjExplicitLod = 92,
+ SpvOpImageSampleProjDrefImplicitLod = 93,
+ SpvOpImageSampleProjDrefExplicitLod = 94,
+ SpvOpImageFetch = 95,
+ SpvOpImageGather = 96,
+ SpvOpImageDrefGather = 97,
+ SpvOpImageRead = 98,
+ SpvOpImageWrite = 99,
+ SpvOpImage = 100,
+ SpvOpImageQueryFormat = 101,
+ SpvOpImageQueryOrder = 102,
+ SpvOpImageQuerySizeLod = 103,
+ SpvOpImageQuerySize = 104,
+ SpvOpImageQueryLod = 105,
+ SpvOpImageQueryLevels = 106,
+ SpvOpImageQuerySamples = 107,
+ SpvOpConvertFToU = 109,
+ SpvOpConvertFToS = 110,
+ SpvOpConvertSToF = 111,
+ SpvOpConvertUToF = 112,
+ SpvOpUConvert = 113,
+ SpvOpSConvert = 114,
+ SpvOpFConvert = 115,
+ SpvOpQuantizeToF16 = 116,
+ SpvOpConvertPtrToU = 117,
+ SpvOpSatConvertSToU = 118,
+ SpvOpSatConvertUToS = 119,
+ SpvOpConvertUToPtr = 120,
+ SpvOpPtrCastToGeneric = 121,
+ SpvOpGenericCastToPtr = 122,
+ SpvOpGenericCastToPtrExplicit = 123,
+ SpvOpBitcast = 124,
+ SpvOpSNegate = 126,
+ SpvOpFNegate = 127,
+ SpvOpIAdd = 128,
+ SpvOpFAdd = 129,
+ SpvOpISub = 130,
+ SpvOpFSub = 131,
+ SpvOpIMul = 132,
+ SpvOpFMul = 133,
+ SpvOpUDiv = 134,
+ SpvOpSDiv = 135,
+ SpvOpFDiv = 136,
+ SpvOpUMod = 137,
+ SpvOpSRem = 138,
+ SpvOpSMod = 139,
+ SpvOpFRem = 140,
+ SpvOpFMod = 141,
+ SpvOpVectorTimesScalar = 142,
+ SpvOpMatrixTimesScalar = 143,
+ SpvOpVectorTimesMatrix = 144,
+ SpvOpMatrixTimesVector = 145,
+ SpvOpMatrixTimesMatrix = 146,
+ SpvOpOuterProduct = 147,
+ SpvOpDot = 148,
+ SpvOpIAddCarry = 149,
+ SpvOpISubBorrow = 150,
+ SpvOpUMulExtended = 151,
+ SpvOpSMulExtended = 152,
+ SpvOpAny = 154,
+ SpvOpAll = 155,
+ SpvOpIsNan = 156,
+ SpvOpIsInf = 157,
+ SpvOpIsFinite = 158,
+ SpvOpIsNormal = 159,
+ SpvOpSignBitSet = 160,
+ SpvOpLessOrGreater = 161,
+ SpvOpOrdered = 162,
+ SpvOpUnordered = 163,
+ SpvOpLogicalEqual = 164,
+ SpvOpLogicalNotEqual = 165,
+ SpvOpLogicalOr = 166,
+ SpvOpLogicalAnd = 167,
+ SpvOpLogicalNot = 168,
+ SpvOpSelect = 169,
+ SpvOpIEqual = 170,
+ SpvOpINotEqual = 171,
+ SpvOpUGreaterThan = 172,
+ SpvOpSGreaterThan = 173,
+ SpvOpUGreaterThanEqual = 174,
+ SpvOpSGreaterThanEqual = 175,
+ SpvOpULessThan = 176,
+ SpvOpSLessThan = 177,
+ SpvOpULessThanEqual = 178,
+ SpvOpSLessThanEqual = 179,
+ SpvOpFOrdEqual = 180,
+ SpvOpFUnordEqual = 181,
+ SpvOpFOrdNotEqual = 182,
+ SpvOpFUnordNotEqual = 183,
+ SpvOpFOrdLessThan = 184,
+ SpvOpFUnordLessThan = 185,
+ SpvOpFOrdGreaterThan = 186,
+ SpvOpFUnordGreaterThan = 187,
+ SpvOpFOrdLessThanEqual = 188,
+ SpvOpFUnordLessThanEqual = 189,
+ SpvOpFOrdGreaterThanEqual = 190,
+ SpvOpFUnordGreaterThanEqual = 191,
+ SpvOpShiftRightLogical = 194,
+ SpvOpShiftRightArithmetic = 195,
+ SpvOpShiftLeftLogical = 196,
+ SpvOpBitwiseOr = 197,
+ SpvOpBitwiseXor = 198,
+ SpvOpBitwiseAnd = 199,
+ SpvOpNot = 200,
+ SpvOpBitFieldInsert = 201,
+ SpvOpBitFieldSExtract = 202,
+ SpvOpBitFieldUExtract = 203,
+ SpvOpBitReverse = 204,
+ SpvOpBitCount = 205,
+ SpvOpDPdx = 207,
+ SpvOpDPdy = 208,
+ SpvOpFwidth = 209,
+ SpvOpDPdxFine = 210,
+ SpvOpDPdyFine = 211,
+ SpvOpFwidthFine = 212,
+ SpvOpDPdxCoarse = 213,
+ SpvOpDPdyCoarse = 214,
+ SpvOpFwidthCoarse = 215,
+ SpvOpEmitVertex = 218,
+ SpvOpEndPrimitive = 219,
+ SpvOpEmitStreamVertex = 220,
+ SpvOpEndStreamPrimitive = 221,
+ SpvOpControlBarrier = 224,
+ SpvOpMemoryBarrier = 225,
+ SpvOpAtomicLoad = 227,
+ SpvOpAtomicStore = 228,
+ SpvOpAtomicExchange = 229,
+ SpvOpAtomicCompareExchange = 230,
+ SpvOpAtomicCompareExchangeWeak = 231,
+ SpvOpAtomicIIncrement = 232,
+ SpvOpAtomicIDecrement = 233,
+ SpvOpAtomicIAdd = 234,
+ SpvOpAtomicISub = 235,
+ SpvOpAtomicSMin = 236,
+ SpvOpAtomicUMin = 237,
+ SpvOpAtomicSMax = 238,
+ SpvOpAtomicUMax = 239,
+ SpvOpAtomicAnd = 240,
+ SpvOpAtomicOr = 241,
+ SpvOpAtomicXor = 242,
+ SpvOpPhi = 245,
+ SpvOpLoopMerge = 246,
+ SpvOpSelectionMerge = 247,
+ SpvOpLabel = 248,
+ SpvOpBranch = 249,
+ SpvOpBranchConditional = 250,
+ SpvOpSwitch = 251,
+ SpvOpKill = 252,
+ SpvOpReturn = 253,
+ SpvOpReturnValue = 254,
+ SpvOpUnreachable = 255,
+ SpvOpLifetimeStart = 256,
+ SpvOpLifetimeStop = 257,
+ SpvOpGroupAsyncCopy = 259,
+ SpvOpGroupWaitEvents = 260,
+ SpvOpGroupAll = 261,
+ SpvOpGroupAny = 262,
+ SpvOpGroupBroadcast = 263,
+ SpvOpGroupIAdd = 264,
+ SpvOpGroupFAdd = 265,
+ SpvOpGroupFMin = 266,
+ SpvOpGroupUMin = 267,
+ SpvOpGroupSMin = 268,
+ SpvOpGroupFMax = 269,
+ SpvOpGroupUMax = 270,
+ SpvOpGroupSMax = 271,
+ SpvOpReadPipe = 274,
+ SpvOpWritePipe = 275,
+ SpvOpReservedReadPipe = 276,
+ SpvOpReservedWritePipe = 277,
+ SpvOpReserveReadPipePackets = 278,
+ SpvOpReserveWritePipePackets = 279,
+ SpvOpCommitReadPipe = 280,
+ SpvOpCommitWritePipe = 281,
+ SpvOpIsValidReserveId = 282,
+ SpvOpGetNumPipePackets = 283,
+ SpvOpGetMaxPipePackets = 284,
+ SpvOpGroupReserveReadPipePackets = 285,
+ SpvOpGroupReserveWritePipePackets = 286,
+ SpvOpGroupCommitReadPipe = 287,
+ SpvOpGroupCommitWritePipe = 288,
+ SpvOpEnqueueMarker = 291,
+ SpvOpEnqueueKernel = 292,
+ SpvOpGetKernelNDrangeSubGroupCount = 293,
+ SpvOpGetKernelNDrangeMaxSubGroupSize = 294,
+ SpvOpGetKernelWorkGroupSize = 295,
+ SpvOpGetKernelPreferredWorkGroupSizeMultiple = 296,
+ SpvOpRetainEvent = 297,
+ SpvOpReleaseEvent = 298,
+ SpvOpCreateUserEvent = 299,
+ SpvOpIsValidEvent = 300,
+ SpvOpSetUserEventStatus = 301,
+ SpvOpCaptureEventProfilingInfo = 302,
+ SpvOpGetDefaultQueue = 303,
+ SpvOpBuildNDRange = 304,
+ SpvOpImageSparseSampleImplicitLod = 305,
+ SpvOpImageSparseSampleExplicitLod = 306,
+ SpvOpImageSparseSampleDrefImplicitLod = 307,
+ SpvOpImageSparseSampleDrefExplicitLod = 308,
+ SpvOpImageSparseSampleProjImplicitLod = 309,
+ SpvOpImageSparseSampleProjExplicitLod = 310,
+ SpvOpImageSparseSampleProjDrefImplicitLod = 311,
+ SpvOpImageSparseSampleProjDrefExplicitLod = 312,
+ SpvOpImageSparseFetch = 313,
+ SpvOpImageSparseGather = 314,
+ SpvOpImageSparseDrefGather = 315,
+ SpvOpImageSparseTexelsResident = 316,
+ SpvOpNoLine = 317,
+ SpvOpAtomicFlagTestAndSet = 318,
+ SpvOpAtomicFlagClear = 319,
+} SpvOp;
+
+#endif // #ifndef spirv_H
+
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ * Authors:
+ * Jason Ekstrand (jason@jlekstrand.net)
+ *
+ */
+
+#include "vtn_private.h"
+#include "nir/nir_vla.h"
+#include "nir/nir_control_flow.h"
+#include "nir/nir_constant_expressions.h"
+
+static struct vtn_ssa_value *
+vtn_undef_ssa_value(struct vtn_builder *b, const struct glsl_type *type)
+{
+ struct vtn_ssa_value *val = rzalloc(b, struct vtn_ssa_value);
+ val->type = type;
+
+ if (glsl_type_is_vector_or_scalar(type)) {
+ unsigned num_components = glsl_get_vector_elements(val->type);
+ nir_ssa_undef_instr *undef =
+ nir_ssa_undef_instr_create(b->shader, num_components);
+
+ nir_instr_insert_before_cf_list(&b->impl->body, &undef->instr);
+ val->def = &undef->def;
+ } else {
+ unsigned elems = glsl_get_length(val->type);
+ val->elems = ralloc_array(b, struct vtn_ssa_value *, elems);
+ if (glsl_type_is_matrix(type)) {
+ const struct glsl_type *elem_type =
+ glsl_vector_type(glsl_get_base_type(type),
+ glsl_get_vector_elements(type));
+
+ for (unsigned i = 0; i < elems; i++)
+ val->elems[i] = vtn_undef_ssa_value(b, elem_type);
+ } else if (glsl_type_is_array(type)) {
+ const struct glsl_type *elem_type = glsl_get_array_element(type);
+ for (unsigned i = 0; i < elems; i++)
+ val->elems[i] = vtn_undef_ssa_value(b, elem_type);
+ } else {
+ for (unsigned i = 0; i < elems; i++) {
+ const struct glsl_type *elem_type = glsl_get_struct_field(type, i);
+ val->elems[i] = vtn_undef_ssa_value(b, elem_type);
+ }
+ }
+ }
+
+ return val;
+}
+
+static struct vtn_ssa_value *
+vtn_const_ssa_value(struct vtn_builder *b, nir_constant *constant,
+ const struct glsl_type *type)
+{
+ struct hash_entry *entry = _mesa_hash_table_search(b->const_table, constant);
+
+ if (entry)
+ return entry->data;
+
+ struct vtn_ssa_value *val = rzalloc(b, struct vtn_ssa_value);
+ val->type = type;
+
+ switch (glsl_get_base_type(type)) {
+ case GLSL_TYPE_INT:
+ case GLSL_TYPE_UINT:
+ case GLSL_TYPE_BOOL:
+ case GLSL_TYPE_FLOAT:
+ case GLSL_TYPE_DOUBLE:
+ if (glsl_type_is_vector_or_scalar(type)) {
+ unsigned num_components = glsl_get_vector_elements(val->type);
+ nir_load_const_instr *load =
+ nir_load_const_instr_create(b->shader, num_components);
+
+ for (unsigned i = 0; i < num_components; i++)
+ load->value.u[i] = constant->value.u[i];
+
+ nir_instr_insert_before_cf_list(&b->impl->body, &load->instr);
+ val->def = &load->def;
+ } else {
+ assert(glsl_type_is_matrix(type));
+ unsigned rows = glsl_get_vector_elements(val->type);
+ unsigned columns = glsl_get_matrix_columns(val->type);
+ val->elems = ralloc_array(b, struct vtn_ssa_value *, columns);
+
+ for (unsigned i = 0; i < columns; i++) {
+ struct vtn_ssa_value *col_val = rzalloc(b, struct vtn_ssa_value);
+ col_val->type = glsl_get_column_type(val->type);
+ nir_load_const_instr *load =
+ nir_load_const_instr_create(b->shader, rows);
+
+ for (unsigned j = 0; j < rows; j++)
+ load->value.u[j] = constant->value.u[rows * i + j];
+
+ nir_instr_insert_before_cf_list(&b->impl->body, &load->instr);
+ col_val->def = &load->def;
+
+ val->elems[i] = col_val;
+ }
+ }
+ break;
+
+ case GLSL_TYPE_ARRAY: {
+ unsigned elems = glsl_get_length(val->type);
+ val->elems = ralloc_array(b, struct vtn_ssa_value *, elems);
+ const struct glsl_type *elem_type = glsl_get_array_element(val->type);
+ for (unsigned i = 0; i < elems; i++)
+ val->elems[i] = vtn_const_ssa_value(b, constant->elements[i],
+ elem_type);
+ break;
+ }
+
+ case GLSL_TYPE_STRUCT: {
+ unsigned elems = glsl_get_length(val->type);
+ val->elems = ralloc_array(b, struct vtn_ssa_value *, elems);
+ for (unsigned i = 0; i < elems; i++) {
+ const struct glsl_type *elem_type =
+ glsl_get_struct_field(val->type, i);
+ val->elems[i] = vtn_const_ssa_value(b, constant->elements[i],
+ elem_type);
+ }
+ break;
+ }
+
+ default:
+ unreachable("bad constant type");
+ }
+
+ return val;
+}
+
+struct vtn_ssa_value *
+vtn_ssa_value(struct vtn_builder *b, uint32_t value_id)
+{
+ struct vtn_value *val = vtn_untyped_value(b, value_id);
+ switch (val->value_type) {
+ case vtn_value_type_undef:
+ return vtn_undef_ssa_value(b, val->type->type);
+
+ case vtn_value_type_constant:
+ return vtn_const_ssa_value(b, val->constant, val->const_type);
+
+ case vtn_value_type_ssa:
+ return val->ssa;
+
+ case vtn_value_type_deref:
+ /* This is needed for function parameters */
+ return vtn_variable_load(b, val->deref, val->deref_type);
+
+ default:
+ unreachable("Invalid type for an SSA value");
+ }
+}
+
+static char *
+vtn_string_literal(struct vtn_builder *b, const uint32_t *words,
+ unsigned word_count, unsigned *words_used)
+{
+ char *dup = ralloc_strndup(b, (char *)words, word_count * sizeof(*words));
+ if (words_used) {
+ /* Ammount of space taken by the string (including the null) */
+ unsigned len = strlen(dup) + 1;
+ *words_used = DIV_ROUND_UP(len, sizeof(*words));
+ }
+ return dup;
+}
+
+const uint32_t *
+vtn_foreach_instruction(struct vtn_builder *b, const uint32_t *start,
+ const uint32_t *end, vtn_instruction_handler handler)
+{
+ b->file = NULL;
+ b->line = -1;
+ b->col = -1;
+
+ const uint32_t *w = start;
+ while (w < end) {
+ SpvOp opcode = w[0] & SpvOpCodeMask;
+ unsigned count = w[0] >> SpvWordCountShift;
+ assert(count >= 1 && w + count <= end);
+
+ switch (opcode) {
+ case SpvOpNop:
+ break; /* Do nothing */
+
+ case SpvOpLine:
+ b->file = vtn_value(b, w[1], vtn_value_type_string)->str;
+ b->line = w[2];
+ b->col = w[3];
+ break;
+
+ case SpvOpNoLine:
+ b->file = NULL;
+ b->line = -1;
+ b->col = -1;
+ break;
+
+ default:
+ if (!handler(b, opcode, w, count))
+ return w;
+ break;
+ }
+
+ w += count;
+ }
+ assert(w == end);
+ return w;
+}
+
+static void
+vtn_handle_extension(struct vtn_builder *b, SpvOp opcode,
+ const uint32_t *w, unsigned count)
+{
+ switch (opcode) {
+ case SpvOpExtInstImport: {
+ struct vtn_value *val = vtn_push_value(b, w[1], vtn_value_type_extension);
+ if (strcmp((const char *)&w[2], "GLSL.std.450") == 0) {
+ val->ext_handler = vtn_handle_glsl450_instruction;
+ } else {
+ assert(!"Unsupported extension");
+ }
+ break;
+ }
+
+ case SpvOpExtInst: {
+ struct vtn_value *val = vtn_value(b, w[3], vtn_value_type_extension);
+ bool handled = val->ext_handler(b, w[4], w, count);
+ (void)handled;
+ assert(handled);
+ break;
+ }
+
+ default:
+ unreachable("Unhandled opcode");
+ }
+}
+
+static void
+_foreach_decoration_helper(struct vtn_builder *b,
+ struct vtn_value *base_value,
+ int parent_member,
+ struct vtn_value *value,
+ vtn_decoration_foreach_cb cb, void *data)
+{
+ for (struct vtn_decoration *dec = value->decoration; dec; dec = dec->next) {
+ int member;
+ if (dec->scope == VTN_DEC_DECORATION) {
+ member = parent_member;
+ } else if (dec->scope >= VTN_DEC_STRUCT_MEMBER0) {
+ assert(parent_member == -1);
+ member = dec->scope - VTN_DEC_STRUCT_MEMBER0;
+ } else {
+ /* Not a decoration */
+ continue;
+ }
+
+ if (dec->group) {
+ assert(dec->group->value_type == vtn_value_type_decoration_group);
+ _foreach_decoration_helper(b, base_value, member, dec->group,
+ cb, data);
+ } else {
+ cb(b, base_value, member, dec, data);
+ }
+ }
+}
+
+/** Iterates (recursively if needed) over all of the decorations on a value
+ *
+ * This function iterates over all of the decorations applied to a given
+ * value. If it encounters a decoration group, it recurses into the group
+ * and iterates over all of those decorations as well.
+ */
+void
+vtn_foreach_decoration(struct vtn_builder *b, struct vtn_value *value,
+ vtn_decoration_foreach_cb cb, void *data)
+{
+ _foreach_decoration_helper(b, value, -1, value, cb, data);
+}
+
+void
+vtn_foreach_execution_mode(struct vtn_builder *b, struct vtn_value *value,
+ vtn_execution_mode_foreach_cb cb, void *data)
+{
+ for (struct vtn_decoration *dec = value->decoration; dec; dec = dec->next) {
+ if (dec->scope != VTN_DEC_EXECUTION_MODE)
+ continue;
+
+ assert(dec->group == NULL);
+ cb(b, value, dec, data);
+ }
+}
+
+static void
+vtn_handle_decoration(struct vtn_builder *b, SpvOp opcode,
+ const uint32_t *w, unsigned count)
+{
+ const uint32_t *w_end = w + count;
+ const uint32_t target = w[1];
+ w += 2;
+
+ switch (opcode) {
+ case SpvOpDecorationGroup:
+ vtn_push_value(b, target, vtn_value_type_decoration_group);
+ break;
+
+ case SpvOpDecorate:
+ case SpvOpMemberDecorate:
+ case SpvOpExecutionMode: {
+ struct vtn_value *val = &b->values[target];
+
+ struct vtn_decoration *dec = rzalloc(b, struct vtn_decoration);
+ switch (opcode) {
+ case SpvOpDecorate:
+ dec->scope = VTN_DEC_DECORATION;
+ break;
+ case SpvOpMemberDecorate:
+ dec->scope = VTN_DEC_STRUCT_MEMBER0 + *(w++);
+ break;
+ case SpvOpExecutionMode:
+ dec->scope = VTN_DEC_EXECUTION_MODE;
+ break;
+ default:
+ unreachable("Invalid decoration opcode");
+ }
+ dec->decoration = *(w++);
+ dec->literals = w;
+
+ /* Link into the list */
+ dec->next = val->decoration;
+ val->decoration = dec;
+ break;
+ }
+
+ case SpvOpGroupMemberDecorate:
+ case SpvOpGroupDecorate: {
+ struct vtn_value *group =
+ vtn_value(b, target, vtn_value_type_decoration_group);
+
+ for (; w < w_end; w++) {
+ struct vtn_value *val = vtn_untyped_value(b, *w);
+ struct vtn_decoration *dec = rzalloc(b, struct vtn_decoration);
+
+ dec->group = group;
+ if (opcode == SpvOpGroupDecorate) {
+ dec->scope = VTN_DEC_DECORATION;
+ } else {
+ dec->scope = VTN_DEC_STRUCT_MEMBER0 + *(w++);
+ }
+
+ /* Link into the list */
+ dec->next = val->decoration;
+ val->decoration = dec;
+ }
+ break;
+ }
+
+ default:
+ unreachable("Unhandled opcode");
+ }
+}
+
+struct member_decoration_ctx {
+ struct glsl_struct_field *fields;
+ struct vtn_type *type;
+};
+
+/* does a shallow copy of a vtn_type */
+
+static struct vtn_type *
+vtn_type_copy(struct vtn_builder *b, struct vtn_type *src)
+{
+ struct vtn_type *dest = ralloc(b, struct vtn_type);
+ dest->type = src->type;
+ dest->is_builtin = src->is_builtin;
+ if (src->is_builtin)
+ dest->builtin = src->builtin;
+
+ if (!glsl_type_is_vector_or_scalar(src->type)) {
+ switch (glsl_get_base_type(src->type)) {
+ case GLSL_TYPE_ARRAY:
+ dest->array_element = src->array_element;
+ dest->stride = src->stride;
+ break;
+
+ case GLSL_TYPE_INT:
+ case GLSL_TYPE_UINT:
+ case GLSL_TYPE_BOOL:
+ case GLSL_TYPE_FLOAT:
+ case GLSL_TYPE_DOUBLE:
+ /* matrices */
+ dest->row_major = src->row_major;
+ dest->stride = src->stride;
+ break;
+
+ case GLSL_TYPE_STRUCT: {
+ unsigned elems = glsl_get_length(src->type);
+
+ dest->members = ralloc_array(b, struct vtn_type *, elems);
+ memcpy(dest->members, src->members, elems * sizeof(struct vtn_type *));
+
+ dest->offsets = ralloc_array(b, unsigned, elems);
+ memcpy(dest->offsets, src->offsets, elems * sizeof(unsigned));
+ break;
+ }
+
+ default:
+ unreachable("unhandled type");
+ }
+ }
+
+ return dest;
+}
+
+static struct vtn_type *
+mutable_matrix_member(struct vtn_builder *b, struct vtn_type *type, int member)
+{
+ type->members[member] = vtn_type_copy(b, type->members[member]);
+ type = type->members[member];
+
+ /* We may have an array of matrices.... Oh, joy! */
+ while (glsl_type_is_array(type->type)) {
+ type->array_element = vtn_type_copy(b, type->array_element);
+ type = type->array_element;
+ }
+
+ assert(glsl_type_is_matrix(type->type));
+
+ return type;
+}
+
+static void
+struct_member_decoration_cb(struct vtn_builder *b,
+ struct vtn_value *val, int member,
+ const struct vtn_decoration *dec, void *void_ctx)
+{
+ struct member_decoration_ctx *ctx = void_ctx;
+
+ if (member < 0)
+ return;
+
+ switch (dec->decoration) {
+ case SpvDecorationRelaxedPrecision:
+ break; /* FIXME: Do nothing with this for now. */
+ case SpvDecorationNoPerspective:
+ ctx->fields[member].interpolation = INTERP_QUALIFIER_NOPERSPECTIVE;
+ break;
+ case SpvDecorationFlat:
+ ctx->fields[member].interpolation = INTERP_QUALIFIER_FLAT;
+ break;
+ case SpvDecorationCentroid:
+ ctx->fields[member].centroid = true;
+ break;
+ case SpvDecorationSample:
+ ctx->fields[member].sample = true;
+ break;
+ case SpvDecorationLocation:
+ ctx->fields[member].location = dec->literals[0];
+ break;
+ case SpvDecorationBuiltIn:
+ ctx->type->members[member] = vtn_type_copy(b, ctx->type->members[member]);
+ ctx->type->members[member]->is_builtin = true;
+ ctx->type->members[member]->builtin = dec->literals[0];
+ ctx->type->builtin_block = true;
+ break;
+ case SpvDecorationOffset:
+ ctx->type->offsets[member] = dec->literals[0];
+ break;
+ case SpvDecorationMatrixStride:
+ mutable_matrix_member(b, ctx->type, member)->stride = dec->literals[0];
+ break;
+ case SpvDecorationColMajor:
+ break; /* Nothing to do here. Column-major is the default. */
+ case SpvDecorationRowMajor:
+ mutable_matrix_member(b, ctx->type, member)->row_major = true;
+ break;
+ default:
+ unreachable("Unhandled member decoration");
+ }
+}
+
+static void
+type_decoration_cb(struct vtn_builder *b,
+ struct vtn_value *val, int member,
+ const struct vtn_decoration *dec, void *ctx)
+{
+ struct vtn_type *type = val->type;
+
+ if (member != -1)
+ return;
+
+ switch (dec->decoration) {
+ case SpvDecorationArrayStride:
+ type->stride = dec->literals[0];
+ break;
+ case SpvDecorationBlock:
+ type->block = true;
+ break;
+ case SpvDecorationBufferBlock:
+ type->buffer_block = true;
+ break;
+ case SpvDecorationGLSLShared:
+ case SpvDecorationGLSLPacked:
+ /* Ignore these, since we get explicit offsets anyways */
+ break;
+
+ case SpvDecorationStream:
+ assert(dec->literals[0] == 0);
+ break;
+
+ default:
+ unreachable("Unhandled type decoration");
+ }
+}
+
+static unsigned
+translate_image_format(SpvImageFormat format)
+{
+ switch (format) {
+ case SpvImageFormatUnknown: return 0; /* GL_NONE */
+ case SpvImageFormatRgba32f: return 0x8814; /* GL_RGBA32F */
+ case SpvImageFormatRgba16f: return 0x881A; /* GL_RGBA16F */
+ case SpvImageFormatR32f: return 0x822E; /* GL_R32F */
+ case SpvImageFormatRgba8: return 0x8058; /* GL_RGBA8 */
+ case SpvImageFormatRgba8Snorm: return 0x8F97; /* GL_RGBA8_SNORM */
+ case SpvImageFormatRg32f: return 0x8230; /* GL_RG32F */
+ case SpvImageFormatRg16f: return 0x822F; /* GL_RG16F */
+ case SpvImageFormatR11fG11fB10f: return 0x8C3A; /* GL_R11F_G11F_B10F */
+ case SpvImageFormatR16f: return 0x822D; /* GL_R16F */
+ case SpvImageFormatRgba16: return 0x805B; /* GL_RGBA16 */
+ case SpvImageFormatRgb10A2: return 0x8059; /* GL_RGB10_A2 */
+ case SpvImageFormatRg16: return 0x822C; /* GL_RG16 */
+ case SpvImageFormatRg8: return 0x822B; /* GL_RG8 */
+ case SpvImageFormatR16: return 0x822A; /* GL_R16 */
+ case SpvImageFormatR8: return 0x8229; /* GL_R8 */
+ case SpvImageFormatRgba16Snorm: return 0x8F9B; /* GL_RGBA16_SNORM */
+ case SpvImageFormatRg16Snorm: return 0x8F99; /* GL_RG16_SNORM */
+ case SpvImageFormatRg8Snorm: return 0x8F95; /* GL_RG8_SNORM */
+ case SpvImageFormatR16Snorm: return 0x8F98; /* GL_R16_SNORM */
+ case SpvImageFormatR8Snorm: return 0x8F94; /* GL_R8_SNORM */
+ case SpvImageFormatRgba32i: return 0x8D82; /* GL_RGBA32I */
+ case SpvImageFormatRgba16i: return 0x8D88; /* GL_RGBA16I */
+ case SpvImageFormatRgba8i: return 0x8D8E; /* GL_RGBA8I */
+ case SpvImageFormatR32i: return 0x8235; /* GL_R32I */
+ case SpvImageFormatRg32i: return 0x823B; /* GL_RG32I */
+ case SpvImageFormatRg16i: return 0x8239; /* GL_RG16I */
+ case SpvImageFormatRg8i: return 0x8237; /* GL_RG8I */
+ case SpvImageFormatR16i: return 0x8233; /* GL_R16I */
+ case SpvImageFormatR8i: return 0x8231; /* GL_R8I */
+ case SpvImageFormatRgba32ui: return 0x8D70; /* GL_RGBA32UI */
+ case SpvImageFormatRgba16ui: return 0x8D76; /* GL_RGBA16UI */
+ case SpvImageFormatRgba8ui: return 0x8D7C; /* GL_RGBA8UI */
+ case SpvImageFormatR32ui: return 0x8236; /* GL_R32UI */
+ case SpvImageFormatRgb10a2ui: return 0x906F; /* GL_RGB10_A2UI */
+ case SpvImageFormatRg32ui: return 0x823C; /* GL_RG32UI */
+ case SpvImageFormatRg16ui: return 0x823A; /* GL_RG16UI */
+ case SpvImageFormatRg8ui: return 0x8238; /* GL_RG8UI */
+ case SpvImageFormatR16ui: return 0x823A; /* GL_RG16UI */
+ case SpvImageFormatR8ui: return 0x8232; /* GL_R8UI */
+ default:
+ assert(!"Invalid image format");
+ return 0;
+ }
+}
+
+static void
+vtn_handle_type(struct vtn_builder *b, SpvOp opcode,
+ const uint32_t *w, unsigned count)
+{
+ struct vtn_value *val = vtn_push_value(b, w[1], vtn_value_type_type);
+
+ val->type = rzalloc(b, struct vtn_type);
+ val->type->is_builtin = false;
+
+ switch (opcode) {
+ case SpvOpTypeVoid:
+ val->type->type = glsl_void_type();
+ break;
+ case SpvOpTypeBool:
+ val->type->type = glsl_bool_type();
+ break;
+ case SpvOpTypeInt:
+ val->type->type = glsl_int_type();
+ break;
+ case SpvOpTypeFloat:
+ val->type->type = glsl_float_type();
+ break;
+
+ case SpvOpTypeVector: {
+ struct vtn_type *base = vtn_value(b, w[2], vtn_value_type_type)->type;
+ unsigned elems = w[3];
+
+ assert(glsl_type_is_scalar(base->type));
+ val->type->type = glsl_vector_type(glsl_get_base_type(base->type), elems);
+
+ /* Vectors implicitly have sizeof(base_type) stride. For now, this
+ * is always 4 bytes. This will have to change if we want to start
+ * supporting doubles or half-floats.
+ */
+ val->type->stride = 4;
+ val->type->array_element = base;
+ break;
+ }
+
+ case SpvOpTypeMatrix: {
+ struct vtn_type *base = vtn_value(b, w[2], vtn_value_type_type)->type;
+ unsigned columns = w[3];
+
+ assert(glsl_type_is_vector(base->type));
+ val->type->type = glsl_matrix_type(glsl_get_base_type(base->type),
+ glsl_get_vector_elements(base->type),
+ columns);
+ assert(!glsl_type_is_error(val->type->type));
+ val->type->array_element = base;
+ val->type->row_major = false;
+ val->type->stride = 0;
+ break;
+ }
+
+ case SpvOpTypeRuntimeArray:
+ case SpvOpTypeArray: {
+ struct vtn_type *array_element =
+ vtn_value(b, w[2], vtn_value_type_type)->type;
+
+ unsigned length;
+ if (opcode == SpvOpTypeRuntimeArray) {
+ /* A length of 0 is used to denote unsized arrays */
+ length = 0;
+ } else {
+ length =
+ vtn_value(b, w[3], vtn_value_type_constant)->constant->value.u[0];
+ }
+
+ val->type->type = glsl_array_type(array_element->type, length);
+ val->type->array_element = array_element;
+ val->type->stride = 0;
+ break;
+ }
+
+ case SpvOpTypeStruct: {
+ unsigned num_fields = count - 2;
+ val->type->members = ralloc_array(b, struct vtn_type *, num_fields);
+ val->type->offsets = ralloc_array(b, unsigned, num_fields);
+
+ NIR_VLA(struct glsl_struct_field, fields, count);
+ for (unsigned i = 0; i < num_fields; i++) {
+ val->type->members[i] =
+ vtn_value(b, w[i + 2], vtn_value_type_type)->type;
+ fields[i] = (struct glsl_struct_field) {
+ .type = val->type->members[i]->type,
+ .name = ralloc_asprintf(b, "field%d", i),
+ .location = -1,
+ };
+ }
+
+ struct member_decoration_ctx ctx = {
+ .fields = fields,
+ .type = val->type
+ };
+
+ vtn_foreach_decoration(b, val, struct_member_decoration_cb, &ctx);
+
+ const char *name = val->name ? val->name : "struct";
+
+ val->type->type = glsl_struct_type(fields, num_fields, name);
+ break;
+ }
+
+ case SpvOpTypeFunction: {
+ const struct glsl_type *return_type =
+ vtn_value(b, w[2], vtn_value_type_type)->type->type;
+ NIR_VLA(struct glsl_function_param, params, count - 3);
+ for (unsigned i = 0; i < count - 3; i++) {
+ params[i].type = vtn_value(b, w[i + 3], vtn_value_type_type)->type->type;
+
+ /* FIXME: */
+ params[i].in = true;
+ params[i].out = true;
+ }
+ val->type->type = glsl_function_type(return_type, params, count - 3);
+ break;
+ }
+
+ case SpvOpTypePointer:
+ /* FIXME: For now, we'll just do the really lame thing and return
+ * the same type. The validator should ensure that the proper number
+ * of dereferences happen
+ */
+ val->type = vtn_value(b, w[3], vtn_value_type_type)->type;
+ break;
+
+ case SpvOpTypeImage: {
+ const struct glsl_type *sampled_type =
+ vtn_value(b, w[2], vtn_value_type_type)->type->type;
+
+ assert(glsl_type_is_vector_or_scalar(sampled_type));
+
+ enum glsl_sampler_dim dim;
+ switch ((SpvDim)w[3]) {
+ case SpvDim1D: dim = GLSL_SAMPLER_DIM_1D; break;
+ case SpvDim2D: dim = GLSL_SAMPLER_DIM_2D; break;
+ case SpvDim3D: dim = GLSL_SAMPLER_DIM_3D; break;
+ case SpvDimCube: dim = GLSL_SAMPLER_DIM_CUBE; break;
+ case SpvDimRect: dim = GLSL_SAMPLER_DIM_RECT; break;
+ case SpvDimBuffer: dim = GLSL_SAMPLER_DIM_BUF; break;
+ default:
+ unreachable("Invalid SPIR-V Sampler dimension");
+ }
+
+ bool is_shadow = w[4];
+ bool is_array = w[5];
+ bool multisampled = w[6];
+ unsigned sampled = w[7];
+ SpvImageFormat format = w[8];
+
+ if (count > 9)
+ val->type->access_qualifier = w[9];
+ else
+ val->type->access_qualifier = SpvAccessQualifierReadWrite;
+
+ assert(!multisampled && "FIXME: Handl multi-sampled textures");
+
+ val->type->image_format = translate_image_format(format);
+
+ if (sampled == 1) {
+ val->type->type = glsl_sampler_type(dim, is_shadow, is_array,
+ glsl_get_base_type(sampled_type));
+ } else if (sampled == 2) {
+ assert(format);
+ assert(!is_shadow);
+ val->type->type = glsl_image_type(dim, is_array,
+ glsl_get_base_type(sampled_type));
+ } else {
+ assert(!"We need to know if the image will be sampled");
+ }
+ break;
+ }
+
+ case SpvOpTypeSampledImage:
+ val->type = vtn_value(b, w[2], vtn_value_type_type)->type;
+ break;
+
+ case SpvOpTypeSampler:
+ /* The actual sampler type here doesn't really matter. It gets
+ * thrown away the moment you combine it with an image. What really
+ * matters is that it's a sampler type as opposed to an integer type
+ * so the backend knows what to do.
+ *
+ * TODO: Eventually we should consider adding a "bare sampler" type
+ * to glsl_types.
+ */
+ val->type->type = glsl_sampler_type(GLSL_SAMPLER_DIM_2D, false, false,
+ GLSL_TYPE_FLOAT);
+ break;
+
+ case SpvOpTypeOpaque:
+ case SpvOpTypeEvent:
+ case SpvOpTypeDeviceEvent:
+ case SpvOpTypeReserveId:
+ case SpvOpTypeQueue:
+ case SpvOpTypePipe:
+ default:
+ unreachable("Unhandled opcode");
+ }
+
+ vtn_foreach_decoration(b, val, type_decoration_cb, NULL);
+}
+
+static nir_constant *
+vtn_null_constant(struct vtn_builder *b, const struct glsl_type *type)
+{
+ nir_constant *c = rzalloc(b, nir_constant);
+
+ switch (glsl_get_base_type(type)) {
+ case GLSL_TYPE_INT:
+ case GLSL_TYPE_UINT:
+ case GLSL_TYPE_BOOL:
+ case GLSL_TYPE_FLOAT:
+ case GLSL_TYPE_DOUBLE:
+ /* Nothing to do here. It's already initialized to zero */
+ break;
+
+ case GLSL_TYPE_ARRAY:
+ assert(glsl_get_length(type) > 0);
+ c->num_elements = glsl_get_length(type);
+ c->elements = ralloc_array(b, nir_constant *, c->num_elements);
+
+ c->elements[0] = vtn_null_constant(b, glsl_get_array_element(type));
+ for (unsigned i = 1; i < c->num_elements; i++)
+ c->elements[i] = c->elements[0];
+ break;
+
+ case GLSL_TYPE_STRUCT:
+ c->num_elements = glsl_get_length(type);
+ c->elements = ralloc_array(b, nir_constant *, c->num_elements);
+
+ for (unsigned i = 0; i < c->num_elements; i++) {
+ c->elements[i] = vtn_null_constant(b, glsl_get_struct_field(type, i));
+ }
+ break;
+
+ default:
+ unreachable("Invalid type for null constant");
+ }
+
+ return c;
+}
+
+static void
+spec_constant_deocoration_cb(struct vtn_builder *b, struct vtn_value *v,
+ int member, const struct vtn_decoration *dec,
+ void *data)
+{
+ assert(member == -1);
+ if (dec->decoration != SpvDecorationSpecId)
+ return;
+
+ uint32_t *const_value = data;
+
+ for (unsigned i = 0; i < b->num_specializations; i++) {
+ if (b->specializations[i].id == dec->literals[0]) {
+ *const_value = b->specializations[i].data;
+ return;
+ }
+ }
+}
+
+static uint32_t
+get_specialization(struct vtn_builder *b, struct vtn_value *val,
+ uint32_t const_value)
+{
+ vtn_foreach_decoration(b, val, spec_constant_deocoration_cb, &const_value);
+ return const_value;
+}
+
+static void
+vtn_handle_constant(struct vtn_builder *b, SpvOp opcode,
+ const uint32_t *w, unsigned count)
+{
+ struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_constant);
+ val->const_type = vtn_value(b, w[1], vtn_value_type_type)->type->type;
+ val->constant = rzalloc(b, nir_constant);
+ switch (opcode) {
+ case SpvOpConstantTrue:
+ assert(val->const_type == glsl_bool_type());
+ val->constant->value.u[0] = NIR_TRUE;
+ break;
+ case SpvOpConstantFalse:
+ assert(val->const_type == glsl_bool_type());
+ val->constant->value.u[0] = NIR_FALSE;
+ break;
+
+ case SpvOpSpecConstantTrue:
+ case SpvOpSpecConstantFalse: {
+ assert(val->const_type == glsl_bool_type());
+ uint32_t int_val =
+ get_specialization(b, val, (opcode == SpvOpSpecConstantTrue));
+ val->constant->value.u[0] = int_val ? NIR_TRUE : NIR_FALSE;
+ break;
+ }
+
+ case SpvOpConstant:
+ assert(glsl_type_is_scalar(val->const_type));
+ val->constant->value.u[0] = w[3];
+ break;
+ case SpvOpSpecConstant:
+ assert(glsl_type_is_scalar(val->const_type));
+ val->constant->value.u[0] = get_specialization(b, val, w[3]);
+ break;
+ case SpvOpSpecConstantComposite:
+ case SpvOpConstantComposite: {
+ unsigned elem_count = count - 3;
+ nir_constant **elems = ralloc_array(b, nir_constant *, elem_count);
+ for (unsigned i = 0; i < elem_count; i++)
+ elems[i] = vtn_value(b, w[i + 3], vtn_value_type_constant)->constant;
+
+ switch (glsl_get_base_type(val->const_type)) {
+ case GLSL_TYPE_UINT:
+ case GLSL_TYPE_INT:
+ case GLSL_TYPE_FLOAT:
+ case GLSL_TYPE_BOOL:
+ if (glsl_type_is_matrix(val->const_type)) {
+ unsigned rows = glsl_get_vector_elements(val->const_type);
+ assert(glsl_get_matrix_columns(val->const_type) == elem_count);
+ for (unsigned i = 0; i < elem_count; i++)
+ for (unsigned j = 0; j < rows; j++)
+ val->constant->value.u[rows * i + j] = elems[i]->value.u[j];
+ } else {
+ assert(glsl_type_is_vector(val->const_type));
+ assert(glsl_get_vector_elements(val->const_type) == elem_count);
+ for (unsigned i = 0; i < elem_count; i++)
+ val->constant->value.u[i] = elems[i]->value.u[0];
+ }
+ ralloc_free(elems);
+ break;
+
+ case GLSL_TYPE_STRUCT:
+ case GLSL_TYPE_ARRAY:
+ ralloc_steal(val->constant, elems);
+ val->constant->num_elements = elem_count;
+ val->constant->elements = elems;
+ break;
+
+ default:
+ unreachable("Unsupported type for constants");
+ }
+ break;
+ }
+
+ case SpvOpSpecConstantOp: {
+ SpvOp opcode = get_specialization(b, val, w[3]);
+ switch (opcode) {
+ case SpvOpVectorShuffle: {
+ struct vtn_value *v0 = vtn_value(b, w[4], vtn_value_type_constant);
+ struct vtn_value *v1 = vtn_value(b, w[5], vtn_value_type_constant);
+ unsigned len0 = glsl_get_vector_elements(v0->const_type);
+ unsigned len1 = glsl_get_vector_elements(v1->const_type);
+
+ uint32_t u[8];
+ for (unsigned i = 0; i < len0; i++)
+ u[i] = v0->constant->value.u[i];
+ for (unsigned i = 0; i < len1; i++)
+ u[len0 + i] = v1->constant->value.u[i];
+
+ for (unsigned i = 0; i < count - 6; i++) {
+ uint32_t comp = w[i + 6];
+ if (comp == (uint32_t)-1) {
+ val->constant->value.u[i] = 0xdeadbeef;
+ } else {
+ val->constant->value.u[i] = u[comp];
+ }
+ }
+ return;
+ }
+
+ case SpvOpCompositeExtract:
+ case SpvOpCompositeInsert: {
+ struct vtn_value *comp;
+ unsigned deref_start;
+ struct nir_constant **c;
+ if (opcode == SpvOpCompositeExtract) {
+ comp = vtn_value(b, w[4], vtn_value_type_constant);
+ deref_start = 5;
+ c = &comp->constant;
+ } else {
+ comp = vtn_value(b, w[5], vtn_value_type_constant);
+ deref_start = 6;
+ val->constant = nir_constant_clone(comp->constant,
+ (nir_variable *)b);
+ c = &val->constant;
+ }
+
+ int elem = -1;
+ const struct glsl_type *type = comp->const_type;
+ for (unsigned i = deref_start; i < count; i++) {
+ switch (glsl_get_base_type(type)) {
+ case GLSL_TYPE_UINT:
+ case GLSL_TYPE_INT:
+ case GLSL_TYPE_FLOAT:
+ case GLSL_TYPE_BOOL:
+ /* If we hit this granularity, we're picking off an element */
+ if (elem < 0)
+ elem = 0;
+
+ if (glsl_type_is_matrix(type)) {
+ elem += w[i] * glsl_get_vector_elements(type);
+ type = glsl_get_column_type(type);
+ } else {
+ assert(glsl_type_is_vector(type));
+ elem += w[i];
+ type = glsl_scalar_type(glsl_get_base_type(type));
+ }
+ continue;
+
+ case GLSL_TYPE_ARRAY:
+ c = &(*c)->elements[w[i]];
+ type = glsl_get_array_element(type);
+ continue;
+
+ case GLSL_TYPE_STRUCT:
+ c = &(*c)->elements[w[i]];
+ type = glsl_get_struct_field(type, w[i]);
+ continue;
+
+ default:
+ unreachable("Invalid constant type");
+ }
+ }
+
+ if (opcode == SpvOpCompositeExtract) {
+ if (elem == -1) {
+ val->constant = *c;
+ } else {
+ unsigned num_components = glsl_get_vector_elements(type);
+ for (unsigned i = 0; i < num_components; i++)
+ val->constant->value.u[i] = (*c)->value.u[elem + i];
+ }
+ } else {
+ struct vtn_value *insert =
+ vtn_value(b, w[4], vtn_value_type_constant);
+ assert(insert->const_type == type);
+ if (elem == -1) {
+ *c = insert->constant;
+ } else {
+ unsigned num_components = glsl_get_vector_elements(type);
+ for (unsigned i = 0; i < num_components; i++)
+ (*c)->value.u[elem + i] = insert->constant->value.u[i];
+ }
+ }
+ return;
+ }
+
+ default: {
+ bool swap;
+ nir_op op = vtn_nir_alu_op_for_spirv_opcode(opcode, &swap);
+
+ unsigned num_components = glsl_get_vector_elements(val->const_type);
+
+ nir_const_value src[3];
+ assert(count <= 7);
+ for (unsigned i = 0; i < count - 4; i++) {
+ nir_constant *c =
+ vtn_value(b, w[4 + i], vtn_value_type_constant)->constant;
+
+ unsigned j = swap ? 1 - i : i;
+ for (unsigned k = 0; k < num_components; k++)
+ src[j].u[k] = c->value.u[k];
+ }
+
+ nir_const_value res = nir_eval_const_opcode(op, num_components, src);
+
+ for (unsigned k = 0; k < num_components; k++)
+ val->constant->value.u[k] = res.u[k];
+
+ return;
+ } /* default */
+ }
+ }
+
+ case SpvOpConstantNull:
+ val->constant = vtn_null_constant(b, val->const_type);
+ break;
+
+ case SpvOpConstantSampler:
+ assert(!"OpConstantSampler requires Kernel Capability");
+ break;
+
+ default:
+ unreachable("Unhandled opcode");
+ }
+}
+
+static void
+set_mode_system_value(nir_variable_mode *mode)
+{
+ assert(*mode == nir_var_system_value || *mode == nir_var_shader_in);
+ *mode = nir_var_system_value;
+}
+
+static void
+vtn_get_builtin_location(struct vtn_builder *b,
+ SpvBuiltIn builtin, int *location,
+ nir_variable_mode *mode)
+{
+ switch (builtin) {
+ case SpvBuiltInPosition:
+ *location = VARYING_SLOT_POS;
+ break;
+ case SpvBuiltInPointSize:
+ *location = VARYING_SLOT_PSIZ;
+ break;
+ case SpvBuiltInClipDistance:
+ *location = VARYING_SLOT_CLIP_DIST0; /* XXX CLIP_DIST1? */
+ break;
+ case SpvBuiltInCullDistance:
+ /* XXX figure this out */
+ unreachable("unhandled builtin");
+ case SpvBuiltInVertexIndex:
+ *location = SYSTEM_VALUE_VERTEX_ID;
+ set_mode_system_value(mode);
+ break;
+ case SpvBuiltInVertexId:
+ /* Vulkan defines VertexID to be zero-based and reserves the new
+ * builtin keyword VertexIndex to indicate the non-zero-based value.
+ */
+ *location = SYSTEM_VALUE_VERTEX_ID_ZERO_BASE;
+ set_mode_system_value(mode);
+ break;
+ case SpvBuiltInInstanceIndex:
+ /* XXX */
+ case SpvBuiltInInstanceId:
+ *location = SYSTEM_VALUE_INSTANCE_ID;
+ set_mode_system_value(mode);
+ break;
+ case SpvBuiltInPrimitiveId:
+ *location = VARYING_SLOT_PRIMITIVE_ID;
+ *mode = nir_var_shader_out;
+ break;
+ case SpvBuiltInInvocationId:
+ *location = SYSTEM_VALUE_INVOCATION_ID;
+ set_mode_system_value(mode);
+ break;
+ case SpvBuiltInLayer:
+ *location = VARYING_SLOT_LAYER;
+ *mode = nir_var_shader_out;
+ break;
+ case SpvBuiltInViewportIndex:
+ *location = VARYING_SLOT_VIEWPORT;
+ if (b->shader->stage == MESA_SHADER_GEOMETRY)
+ *mode = nir_var_shader_out;
+ else if (b->shader->stage == MESA_SHADER_FRAGMENT)
+ *mode = nir_var_shader_in;
+ else
+ unreachable("invalid stage for SpvBuiltInViewportIndex");
+ break;
+ case SpvBuiltInTessLevelOuter:
+ case SpvBuiltInTessLevelInner:
+ case SpvBuiltInTessCoord:
+ case SpvBuiltInPatchVertices:
+ unreachable("no tessellation support");
+ case SpvBuiltInFragCoord:
+ *location = VARYING_SLOT_POS;
+ assert(*mode == nir_var_shader_in);
+ break;
+ case SpvBuiltInPointCoord:
+ *location = VARYING_SLOT_PNTC;
+ assert(*mode == nir_var_shader_in);
+ break;
+ case SpvBuiltInFrontFacing:
+ *location = VARYING_SLOT_FACE;
+ assert(*mode == nir_var_shader_in);
+ break;
+ case SpvBuiltInSampleId:
+ *location = SYSTEM_VALUE_SAMPLE_ID;
+ set_mode_system_value(mode);
+ break;
+ case SpvBuiltInSamplePosition:
+ *location = SYSTEM_VALUE_SAMPLE_POS;
+ set_mode_system_value(mode);
+ break;
+ case SpvBuiltInSampleMask:
+ *location = SYSTEM_VALUE_SAMPLE_MASK_IN; /* XXX out? */
+ set_mode_system_value(mode);
+ break;
+ case SpvBuiltInFragDepth:
+ *location = FRAG_RESULT_DEPTH;
+ assert(*mode == nir_var_shader_out);
+ break;
+ case SpvBuiltInNumWorkgroups:
+ *location = SYSTEM_VALUE_NUM_WORK_GROUPS;
+ set_mode_system_value(mode);
+ break;
+ case SpvBuiltInWorkgroupSize:
+ /* This should already be handled */
+ unreachable("unsupported builtin");
+ break;
+ case SpvBuiltInWorkgroupId:
+ *location = SYSTEM_VALUE_WORK_GROUP_ID;
+ set_mode_system_value(mode);
+ break;
+ case SpvBuiltInLocalInvocationId:
+ *location = SYSTEM_VALUE_LOCAL_INVOCATION_ID;
+ set_mode_system_value(mode);
+ break;
+ case SpvBuiltInLocalInvocationIndex:
+ *location = SYSTEM_VALUE_LOCAL_INVOCATION_INDEX;
+ set_mode_system_value(mode);
+ break;
+ case SpvBuiltInGlobalInvocationId:
+ *location = SYSTEM_VALUE_GLOBAL_INVOCATION_ID;
+ set_mode_system_value(mode);
+ break;
+ case SpvBuiltInHelperInvocation:
+ default:
+ unreachable("unsupported builtin");
+ }
+}
+
+static void
+var_decoration_cb(struct vtn_builder *b, struct vtn_value *val, int member,
+ const struct vtn_decoration *dec, void *void_var)
+{
+ assert(val->value_type == vtn_value_type_deref);
+ assert(val->deref->deref.child == NULL);
+ assert(val->deref->var == void_var);
+
+ nir_variable *var = void_var;
+ switch (dec->decoration) {
+ case SpvDecorationRelaxedPrecision:
+ break; /* FIXME: Do nothing with this for now. */
+ case SpvDecorationNoPerspective:
+ var->data.interpolation = INTERP_QUALIFIER_NOPERSPECTIVE;
+ break;
+ case SpvDecorationFlat:
+ var->data.interpolation = INTERP_QUALIFIER_FLAT;
+ break;
+ case SpvDecorationCentroid:
+ var->data.centroid = true;
+ break;
+ case SpvDecorationSample:
+ var->data.sample = true;
+ break;
+ case SpvDecorationInvariant:
+ var->data.invariant = true;
+ break;
+ case SpvDecorationConstant:
+ assert(var->constant_initializer != NULL);
+ var->data.read_only = true;
+ break;
+ case SpvDecorationNonWritable:
+ var->data.read_only = true;
+ break;
+ case SpvDecorationLocation:
+ var->data.location = dec->literals[0];
+ break;
+ case SpvDecorationComponent:
+ var->data.location_frac = dec->literals[0];
+ break;
+ case SpvDecorationIndex:
+ var->data.explicit_index = true;
+ var->data.index = dec->literals[0];
+ break;
+ case SpvDecorationBinding:
+ var->data.explicit_binding = true;
+ var->data.binding = dec->literals[0];
+ break;
+ case SpvDecorationDescriptorSet:
+ var->data.descriptor_set = dec->literals[0];
+ break;
+ case SpvDecorationBuiltIn: {
+ SpvBuiltIn builtin = dec->literals[0];
+
+ if (builtin == SpvBuiltInWorkgroupSize) {
+ /* This shouldn't be a builtin. It's actually a constant. */
+ var->data.mode = nir_var_global;
+ var->data.read_only = true;
+
+ nir_constant *val = rzalloc(var, nir_constant);
+ val->value.u[0] = b->shader->info.cs.local_size[0];
+ val->value.u[1] = b->shader->info.cs.local_size[1];
+ val->value.u[2] = b->shader->info.cs.local_size[2];
+ var->constant_initializer = val;
+ break;
+ }
+
+ nir_variable_mode mode = var->data.mode;
+ vtn_get_builtin_location(b, builtin, &var->data.location, &mode);
+ var->data.explicit_location = true;
+ var->data.mode = mode;
+ if (mode == nir_var_shader_in || mode == nir_var_system_value)
+ var->data.read_only = true;
+
+ if (builtin == SpvBuiltInFragCoord || builtin == SpvBuiltInSamplePosition)
+ var->data.origin_upper_left = b->origin_upper_left;
+
+ if (mode == nir_var_shader_out)
+ b->builtins[dec->literals[0]].out = var;
+ else
+ b->builtins[dec->literals[0]].in = var;
+ break;
+ }
+ case SpvDecorationRowMajor:
+ case SpvDecorationColMajor:
+ case SpvDecorationGLSLShared:
+ case SpvDecorationPatch:
+ case SpvDecorationRestrict:
+ case SpvDecorationAliased:
+ case SpvDecorationVolatile:
+ case SpvDecorationCoherent:
+ case SpvDecorationNonReadable:
+ case SpvDecorationUniform:
+ /* This is really nice but we have no use for it right now. */
+ case SpvDecorationCPacked:
+ case SpvDecorationSaturatedConversion:
+ case SpvDecorationStream:
+ case SpvDecorationOffset:
+ case SpvDecorationXfbBuffer:
+ case SpvDecorationFuncParamAttr:
+ case SpvDecorationFPRoundingMode:
+ case SpvDecorationFPFastMathMode:
+ case SpvDecorationLinkageAttributes:
+ case SpvDecorationSpecId:
+ break;
+ default:
+ unreachable("Unhandled variable decoration");
+ }
+}
+
+static nir_variable *
+get_builtin_variable(struct vtn_builder *b,
+ nir_variable_mode mode,
+ const struct glsl_type *type,
+ SpvBuiltIn builtin)
+{
+ nir_variable *var;
+ if (mode == nir_var_shader_out)
+ var = b->builtins[builtin].out;
+ else
+ var = b->builtins[builtin].in;
+
+ if (!var) {
+ int location;
+ vtn_get_builtin_location(b, builtin, &location, &mode);
+
+ var = nir_variable_create(b->shader, mode, type, "builtin");
+
+ var->data.location = location;
+ var->data.explicit_location = true;
+
+ if (builtin == SpvBuiltInFragCoord || builtin == SpvBuiltInSamplePosition)
+ var->data.origin_upper_left = b->origin_upper_left;
+
+ if (mode == nir_var_shader_out)
+ b->builtins[builtin].out = var;
+ else
+ b->builtins[builtin].in = var;
+ }
+
+ return var;
+}
+
+static struct vtn_ssa_value *
+_vtn_variable_load(struct vtn_builder *b,
+ nir_deref_var *src_deref, nir_deref *src_deref_tail)
+{
+ struct vtn_ssa_value *val = rzalloc(b, struct vtn_ssa_value);
+ val->type = src_deref_tail->type;
+
+ /* The deref tail may contain a deref to select a component of a vector (in
+ * other words, it might not be an actual tail) so we have to save it away
+ * here since we overwrite it later.
+ */
+ nir_deref *old_child = src_deref_tail->child;
+
+ if (glsl_type_is_vector_or_scalar(val->type)) {
+ /* Terminate the deref chain in case there is one more link to pick
+ * off a component of the vector.
+ */
+ src_deref_tail->child = NULL;
+
+ nir_intrinsic_instr *load =
+ nir_intrinsic_instr_create(b->shader, nir_intrinsic_load_var);
+ load->variables[0] =
+ nir_deref_as_var(nir_copy_deref(load, &src_deref->deref));
+ load->num_components = glsl_get_vector_elements(val->type);
+ nir_ssa_dest_init(&load->instr, &load->dest, load->num_components, NULL);
+
+ nir_builder_instr_insert(&b->nb, &load->instr);
+
+ if (src_deref->var->data.mode == nir_var_uniform &&
+ glsl_get_base_type(val->type) == GLSL_TYPE_BOOL) {
+ /* Uniform boolean loads need to be fixed up since they're defined
+ * to be zero/nonzero rather than NIR_FALSE/NIR_TRUE.
+ */
+ val->def = nir_ine(&b->nb, &load->dest.ssa, nir_imm_int(&b->nb, 0));
+ } else {
+ val->def = &load->dest.ssa;
+ }
+ } else if (glsl_get_base_type(val->type) == GLSL_TYPE_ARRAY ||
+ glsl_type_is_matrix(val->type)) {
+ unsigned elems = glsl_get_length(val->type);
+ val->elems = ralloc_array(b, struct vtn_ssa_value *, elems);
+
+ nir_deref_array *deref = nir_deref_array_create(b);
+ deref->deref_array_type = nir_deref_array_type_direct;
+ deref->deref.type = glsl_get_array_element(val->type);
+ src_deref_tail->child = &deref->deref;
+ for (unsigned i = 0; i < elems; i++) {
+ deref->base_offset = i;
+ val->elems[i] = _vtn_variable_load(b, src_deref, &deref->deref);
+ }
+ } else {
+ assert(glsl_get_base_type(val->type) == GLSL_TYPE_STRUCT);
+ unsigned elems = glsl_get_length(val->type);
+ val->elems = ralloc_array(b, struct vtn_ssa_value *, elems);
+
+ nir_deref_struct *deref = nir_deref_struct_create(b, 0);
+ src_deref_tail->child = &deref->deref;
+ for (unsigned i = 0; i < elems; i++) {
+ deref->index = i;
+ deref->deref.type = glsl_get_struct_field(val->type, i);
+ val->elems[i] = _vtn_variable_load(b, src_deref, &deref->deref);
+ }
+ }
+
+ src_deref_tail->child = old_child;
+
+ return val;
+}
+
+static void
+_vtn_variable_store(struct vtn_builder *b,
+ nir_deref_var *dest_deref, nir_deref *dest_deref_tail,
+ struct vtn_ssa_value *src)
+{
+ nir_deref *old_child = dest_deref_tail->child;
+
+ if (glsl_type_is_vector_or_scalar(src->type)) {
+ /* Terminate the deref chain in case there is one more link to pick
+ * off a component of the vector.
+ */
+ dest_deref_tail->child = NULL;
+
+ nir_intrinsic_instr *store =
+ nir_intrinsic_instr_create(b->shader, nir_intrinsic_store_var);
+ store->variables[0] =
+ nir_deref_as_var(nir_copy_deref(store, &dest_deref->deref));
+ store->num_components = glsl_get_vector_elements(src->type);
+ store->const_index[0] = (1 << store->num_components) - 1;
+ store->src[0] = nir_src_for_ssa(src->def);
+
+ nir_builder_instr_insert(&b->nb, &store->instr);
+ } else if (glsl_get_base_type(src->type) == GLSL_TYPE_ARRAY ||
+ glsl_type_is_matrix(src->type)) {
+ unsigned elems = glsl_get_length(src->type);
+
+ nir_deref_array *deref = nir_deref_array_create(b);
+ deref->deref_array_type = nir_deref_array_type_direct;
+ deref->deref.type = glsl_get_array_element(src->type);
+ dest_deref_tail->child = &deref->deref;
+ for (unsigned i = 0; i < elems; i++) {
+ deref->base_offset = i;
+ _vtn_variable_store(b, dest_deref, &deref->deref, src->elems[i]);
+ }
+ } else {
+ assert(glsl_get_base_type(src->type) == GLSL_TYPE_STRUCT);
+ unsigned elems = glsl_get_length(src->type);
+
+ nir_deref_struct *deref = nir_deref_struct_create(b, 0);
+ dest_deref_tail->child = &deref->deref;
+ for (unsigned i = 0; i < elems; i++) {
+ deref->index = i;
+ deref->deref.type = glsl_get_struct_field(src->type, i);
+ _vtn_variable_store(b, dest_deref, &deref->deref, src->elems[i]);
+ }
+ }
+
+ dest_deref_tail->child = old_child;
+}
+
+static nir_ssa_def *
+deref_array_offset(struct vtn_builder *b, nir_deref *deref)
+{
+ assert(deref->deref_type == nir_deref_type_array);
+ nir_deref_array *deref_array = nir_deref_as_array(deref);
+ nir_ssa_def *offset = nir_imm_int(&b->nb, deref_array->base_offset);
+
+ if (deref_array->deref_array_type == nir_deref_array_type_indirect)
+ offset = nir_iadd(&b->nb, offset, deref_array->indirect.ssa);
+
+ return offset;
+}
+
+static nir_ssa_def *
+get_vulkan_resource_index(struct vtn_builder *b,
+ nir_deref **deref, struct vtn_type **type)
+{
+ assert((*deref)->deref_type == nir_deref_type_var);
+ nir_variable *var = nir_deref_as_var(*deref)->var;
+
+ assert(var->interface_type && "variable is a block");
+
+ nir_ssa_def *array_index;
+ if ((*deref)->child && (*deref)->child->deref_type == nir_deref_type_array) {
+ *deref = (*deref)->child;
+ *type = (*type)->array_element;
+ array_index = deref_array_offset(b, *deref);
+ } else {
+ array_index = nir_imm_int(&b->nb, 0);
+ }
+
+ nir_intrinsic_instr *instr =
+ nir_intrinsic_instr_create(b->nb.shader,
+ nir_intrinsic_vulkan_resource_index);
+ instr->src[0] = nir_src_for_ssa(array_index);
+ instr->const_index[0] = var->data.descriptor_set;
+ instr->const_index[1] = var->data.binding;
+ instr->const_index[2] = var->data.mode;
+
+ nir_ssa_dest_init(&instr->instr, &instr->dest, 1, NULL);
+ nir_builder_instr_insert(&b->nb, &instr->instr);
+
+ return &instr->dest.ssa;
+}
+
+static void
+_vtn_load_store_tail(struct vtn_builder *b, nir_intrinsic_op op, bool load,
+ nir_ssa_def *index, nir_ssa_def *offset,
+ struct vtn_ssa_value **inout, const struct glsl_type *type)
+{
+ nir_intrinsic_instr *instr = nir_intrinsic_instr_create(b->nb.shader, op);
+ instr->num_components = glsl_get_vector_elements(type);
+
+ int src = 0;
+ if (!load) {
+ instr->const_index[0] = (1 << instr->num_components) - 1; /* write mask */
+ instr->src[src++] = nir_src_for_ssa((*inout)->def);
+ }
+
+ if (index)
+ instr->src[src++] = nir_src_for_ssa(index);
+
+ instr->src[src++] = nir_src_for_ssa(offset);
+
+ if (load) {
+ nir_ssa_dest_init(&instr->instr, &instr->dest,
+ instr->num_components, NULL);
+ (*inout)->def = &instr->dest.ssa;
+ }
+
+ nir_builder_instr_insert(&b->nb, &instr->instr);
+
+ if (load && glsl_get_base_type(type) == GLSL_TYPE_BOOL)
+ (*inout)->def = nir_ine(&b->nb, (*inout)->def, nir_imm_int(&b->nb, 0));
+}
+
+static void
+_vtn_block_load_store(struct vtn_builder *b, nir_intrinsic_op op, bool load,
+ nir_ssa_def *index, nir_ssa_def *offset, nir_deref *deref,
+ struct vtn_type *type, struct vtn_ssa_value **inout)
+{
+ if (load && deref == NULL && *inout == NULL)
+ *inout = vtn_create_ssa_value(b, type->type);
+
+ enum glsl_base_type base_type = glsl_get_base_type(type->type);
+ switch (base_type) {
+ case GLSL_TYPE_UINT:
+ case GLSL_TYPE_INT:
+ case GLSL_TYPE_FLOAT:
+ case GLSL_TYPE_BOOL:
+ /* This is where things get interesting. At this point, we've hit
+ * a vector, a scalar, or a matrix.
+ */
+ if (glsl_type_is_matrix(type->type)) {
+ if (deref == NULL) {
+ /* Loading the whole matrix */
+ struct vtn_ssa_value *transpose;
+ unsigned num_ops, vec_width;
+ if (type->row_major) {
+ num_ops = glsl_get_vector_elements(type->type);
+ vec_width = glsl_get_matrix_columns(type->type);
+ if (load) {
+ const struct glsl_type *transpose_type =
+ glsl_matrix_type(base_type, vec_width, num_ops);
+ *inout = vtn_create_ssa_value(b, transpose_type);
+ } else {
+ transpose = vtn_ssa_transpose(b, *inout);
+ inout = &transpose;
+ }
+ } else {
+ num_ops = glsl_get_matrix_columns(type->type);
+ vec_width = glsl_get_vector_elements(type->type);
+ }
+
+ for (unsigned i = 0; i < num_ops; i++) {
+ nir_ssa_def *elem_offset =
+ nir_iadd(&b->nb, offset,
+ nir_imm_int(&b->nb, i * type->stride));
+ _vtn_load_store_tail(b, op, load, index, elem_offset,
+ &(*inout)->elems[i],
+ glsl_vector_type(base_type, vec_width));
+ }
+
+ if (load && type->row_major)
+ *inout = vtn_ssa_transpose(b, *inout);
+
+ return;
+ } else if (type->row_major) {
+ /* Row-major but with a deref. */
+ nir_ssa_def *col_offset =
+ nir_imul(&b->nb, deref_array_offset(b, deref),
+ nir_imm_int(&b->nb, type->array_element->stride));
+ offset = nir_iadd(&b->nb, offset, col_offset);
+
+ if (deref->child) {
+ /* Picking off a single element */
+ nir_ssa_def *row_offset =
+ nir_imul(&b->nb, deref_array_offset(b, deref->child),
+ nir_imm_int(&b->nb, type->stride));
+ offset = nir_iadd(&b->nb, offset, row_offset);
+ _vtn_load_store_tail(b, op, load, index, offset, inout,
+ glsl_scalar_type(base_type));
+ return;
+ } else {
+ unsigned num_comps = glsl_get_vector_elements(type->type);
+ nir_ssa_def *comps[4];
+ for (unsigned i = 0; i < num_comps; i++) {
+ nir_ssa_def *elem_offset =
+ nir_iadd(&b->nb, offset,
+ nir_imm_int(&b->nb, i * type->stride));
+
+ struct vtn_ssa_value *comp = NULL, temp_val;
+ if (!load) {
+ temp_val.def = nir_channel(&b->nb, (*inout)->def, i);
+ temp_val.type = glsl_scalar_type(base_type);
+ comp = &temp_val;
+ }
+ _vtn_load_store_tail(b, op, load, index, elem_offset,
+ &comp, glsl_scalar_type(base_type));
+ comps[i] = comp->def;
+ }
+
+ if (load)
+ (*inout)->def = nir_vec(&b->nb, comps, num_comps);
+ return;
+ }
+ } else {
+ /* Column-major with a deref. Fall through to array case. */
+ }
+ } else if (deref == NULL) {
+ assert(glsl_type_is_vector_or_scalar(type->type));
+ _vtn_load_store_tail(b, op, load, index, offset, inout, type->type);
+ return;
+ } else {
+ /* Single component of a vector. Fall through to array case. */
+ }
+ /* Fall through */
+
+ case GLSL_TYPE_ARRAY:
+ if (deref) {
+ offset = nir_iadd(&b->nb, offset,
+ nir_imul(&b->nb, deref_array_offset(b, deref),
+ nir_imm_int(&b->nb, type->stride)));
+
+ _vtn_block_load_store(b, op, load, index, offset, deref->child,
+ type->array_element, inout);
+ return;
+ } else {
+ unsigned elems = glsl_get_length(type->type);
+ for (unsigned i = 0; i < elems; i++) {
+ nir_ssa_def *elem_off =
+ nir_iadd(&b->nb, offset, nir_imm_int(&b->nb, i * type->stride));
+ _vtn_block_load_store(b, op, load, index, elem_off, NULL,
+ type->array_element, &(*inout)->elems[i]);
+ }
+ return;
+ }
+ unreachable("Both branches above return");
+
+ case GLSL_TYPE_STRUCT:
+ if (deref) {
+ unsigned member = nir_deref_as_struct(deref)->index;
+ offset = nir_iadd(&b->nb, offset,
+ nir_imm_int(&b->nb, type->offsets[member]));
+
+ _vtn_block_load_store(b, op, load, index, offset, deref->child,
+ type->members[member], inout);
+ return;
+ } else {
+ unsigned elems = glsl_get_length(type->type);
+ for (unsigned i = 0; i < elems; i++) {
+ nir_ssa_def *elem_off =
+ nir_iadd(&b->nb, offset, nir_imm_int(&b->nb, type->offsets[i]));
+ _vtn_block_load_store(b, op, load, index, elem_off, NULL,
+ type->members[i], &(*inout)->elems[i]);
+ }
+ return;
+ }
+ unreachable("Both branches above return");
+
+ default:
+ unreachable("Invalid block member type");
+ }
+}
+
+static struct vtn_ssa_value *
+vtn_block_load(struct vtn_builder *b, nir_deref_var *src,
+ struct vtn_type *type)
+{
+ nir_intrinsic_op op;
+ if (src->var->data.mode == nir_var_uniform) {
+ if (src->var->data.descriptor_set >= 0) {
+ /* UBO load */
+ assert(src->var->data.binding >= 0);
+
+ op = nir_intrinsic_load_ubo;
+ } else {
+ /* Push constant load */
+ assert(src->var->data.descriptor_set == -1 &&
+ src->var->data.binding == -1);
+
+ op = nir_intrinsic_load_push_constant;
+ }
+ } else {
+ assert(src->var->data.mode == nir_var_shader_storage);
+ op = nir_intrinsic_load_ssbo;
+ }
+
+ nir_deref *block_deref = &src->deref;
+ nir_ssa_def *index = NULL;
+ if (op == nir_intrinsic_load_ubo || op == nir_intrinsic_load_ssbo)
+ index = get_vulkan_resource_index(b, &block_deref, &type);
+
+ struct vtn_ssa_value *value = NULL;
+ _vtn_block_load_store(b, op, true, index, nir_imm_int(&b->nb, 0),
+ block_deref->child, type, &value);
+ return value;
+}
+
+/*
+ * Gets the NIR-level deref tail, which may have as a child an array deref
+ * selecting which component due to OpAccessChain supporting per-component
+ * indexing in SPIR-V.
+ */
+
+static nir_deref *
+get_deref_tail(nir_deref_var *deref)
+{
+ nir_deref *cur = &deref->deref;
+ while (!glsl_type_is_vector_or_scalar(cur->type) && cur->child)
+ cur = cur->child;
+
+ return cur;
+}
+
+static nir_ssa_def *vtn_vector_extract(struct vtn_builder *b,
+ nir_ssa_def *src, unsigned index);
+
+static nir_ssa_def *vtn_vector_extract_dynamic(struct vtn_builder *b,
+ nir_ssa_def *src,
+ nir_ssa_def *index);
+
+static bool
+variable_is_external_block(nir_variable *var)
+{
+ return var->interface_type &&
+ glsl_type_is_struct(var->interface_type) &&
+ (var->data.mode == nir_var_uniform ||
+ var->data.mode == nir_var_shader_storage);
+}
+
+struct vtn_ssa_value *
+vtn_variable_load(struct vtn_builder *b, nir_deref_var *src,
+ struct vtn_type *src_type)
+{
+ if (variable_is_external_block(src->var))
+ return vtn_block_load(b, src, src_type);
+
+ nir_deref *src_tail = get_deref_tail(src);
+ struct vtn_ssa_value *val = _vtn_variable_load(b, src, src_tail);
+
+ if (src_tail->child) {
+ nir_deref_array *vec_deref = nir_deref_as_array(src_tail->child);
+ assert(vec_deref->deref.child == NULL);
+ val->type = vec_deref->deref.type;
+ if (vec_deref->deref_array_type == nir_deref_array_type_direct)
+ val->def = vtn_vector_extract(b, val->def, vec_deref->base_offset);
+ else
+ val->def = vtn_vector_extract_dynamic(b, val->def,
+ vec_deref->indirect.ssa);
+ }
+
+ return val;
+}
+
+static void
+vtn_block_store(struct vtn_builder *b, struct vtn_ssa_value *src,
+ nir_deref_var *dest, struct vtn_type *type)
+{
+ nir_deref *block_deref = &dest->deref;
+ nir_ssa_def *index = get_vulkan_resource_index(b, &block_deref, &type);
+
+ _vtn_block_load_store(b, nir_intrinsic_store_ssbo, false, index,
+ nir_imm_int(&b->nb, 0), block_deref->child,
+ type, &src);
+}
+
+static nir_ssa_def * vtn_vector_insert(struct vtn_builder *b,
+ nir_ssa_def *src, nir_ssa_def *insert,
+ unsigned index);
+
+static nir_ssa_def * vtn_vector_insert_dynamic(struct vtn_builder *b,
+ nir_ssa_def *src,
+ nir_ssa_def *insert,
+ nir_ssa_def *index);
+void
+vtn_variable_store(struct vtn_builder *b, struct vtn_ssa_value *src,
+ nir_deref_var *dest, struct vtn_type *dest_type)
+{
+ if (variable_is_external_block(dest->var)) {
+ assert(dest->var->data.mode == nir_var_shader_storage);
+ vtn_block_store(b, src, dest, dest_type);
+ } else {
+ nir_deref *dest_tail = get_deref_tail(dest);
+ if (dest_tail->child) {
+ struct vtn_ssa_value *val = _vtn_variable_load(b, dest, dest_tail);
+ nir_deref_array *deref = nir_deref_as_array(dest_tail->child);
+ assert(deref->deref.child == NULL);
+ if (deref->deref_array_type == nir_deref_array_type_direct)
+ val->def = vtn_vector_insert(b, val->def, src->def,
+ deref->base_offset);
+ else
+ val->def = vtn_vector_insert_dynamic(b, val->def, src->def,
+ deref->indirect.ssa);
+ _vtn_variable_store(b, dest, dest_tail, val);
+ } else {
+ _vtn_variable_store(b, dest, dest_tail, src);
+ }
+ }
+}
+
+static void
+vtn_variable_copy(struct vtn_builder *b,
+ nir_deref_var *dest, struct vtn_type *dest_type,
+ nir_deref_var *src, struct vtn_type *src_type)
+{
+ if (src->var->interface_type || dest->var->interface_type) {
+ struct vtn_ssa_value *val = vtn_variable_load(b, src, src_type);
+ vtn_variable_store(b, val, dest, dest_type);
+ } else {
+ nir_intrinsic_instr *copy =
+ nir_intrinsic_instr_create(b->shader, nir_intrinsic_copy_var);
+ copy->variables[0] = nir_deref_as_var(nir_copy_deref(copy, &dest->deref));
+ copy->variables[1] = nir_deref_as_var(nir_copy_deref(copy, &src->deref));
+
+ nir_builder_instr_insert(&b->nb, ©->instr);
+ }
+}
+
+/* Tries to compute the size of an interface block based on the strides and
+ * offsets that are provided to us in the SPIR-V source.
+ */
+static unsigned
+vtn_type_block_size(struct vtn_type *type)
+{
+ enum glsl_base_type base_type = glsl_get_base_type(type->type);
+ switch (base_type) {
+ case GLSL_TYPE_UINT:
+ case GLSL_TYPE_INT:
+ case GLSL_TYPE_FLOAT:
+ case GLSL_TYPE_BOOL:
+ case GLSL_TYPE_DOUBLE: {
+ unsigned cols = type->row_major ? glsl_get_vector_elements(type->type) :
+ glsl_get_matrix_columns(type->type);
+ if (cols > 1) {
+ assert(type->stride > 0);
+ return type->stride * cols;
+ } else if (base_type == GLSL_TYPE_DOUBLE) {
+ return glsl_get_vector_elements(type->type) * 8;
+ } else {
+ return glsl_get_vector_elements(type->type) * 4;
+ }
+ }
+
+ case GLSL_TYPE_STRUCT:
+ case GLSL_TYPE_INTERFACE: {
+ unsigned size = 0;
+ unsigned num_fields = glsl_get_length(type->type);
+ for (unsigned f = 0; f < num_fields; f++) {
+ unsigned field_end = type->offsets[f] +
+ vtn_type_block_size(type->members[f]);
+ size = MAX2(size, field_end);
+ }
+ return size;
+ }
+
+ case GLSL_TYPE_ARRAY:
+ assert(type->stride > 0);
+ assert(glsl_get_length(type->type) > 0);
+ return type->stride * glsl_get_length(type->type);
+
+ default:
+ assert(!"Invalid block type");
+ return 0;
+ }
+}
+
+static bool
+is_interface_type(struct vtn_type *type)
+{
+ return type->block || type->buffer_block ||
+ glsl_type_is_sampler(type->type) ||
+ glsl_type_is_image(type->type);
+}
+
+static void
+vtn_handle_variables(struct vtn_builder *b, SpvOp opcode,
+ const uint32_t *w, unsigned count)
+{
+ switch (opcode) {
+ case SpvOpVariable: {
+ struct vtn_type *type =
+ vtn_value(b, w[1], vtn_value_type_type)->type;
+ struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_deref);
+ SpvStorageClass storage_class = w[3];
+
+ nir_variable *var = rzalloc(b->shader, nir_variable);
+
+ var->type = type->type;
+ var->name = ralloc_strdup(var, val->name);
+
+ struct vtn_type *interface_type;
+ if (is_interface_type(type)) {
+ interface_type = type;
+ } else if (glsl_type_is_array(type->type) &&
+ is_interface_type(type->array_element)) {
+ interface_type = type->array_element;
+ } else {
+ interface_type = NULL;
+ }
+
+ if (interface_type)
+ var->interface_type = interface_type->type;
+
+ switch (storage_class) {
+ case SpvStorageClassUniform:
+ case SpvStorageClassUniformConstant:
+ if (interface_type && interface_type->buffer_block) {
+ var->data.mode = nir_var_shader_storage;
+ b->shader->info.num_ssbos++;
+ } else {
+ /* UBO's and samplers */
+ var->data.mode = nir_var_uniform;
+ var->data.read_only = true;
+ if (interface_type) {
+ if (glsl_type_is_image(interface_type->type)) {
+ b->shader->info.num_images++;
+ var->data.image.format = interface_type->image_format;
+
+ switch (interface_type->access_qualifier) {
+ case SpvAccessQualifierReadOnly:
+ var->data.image.read_only = true;
+ break;
+ case SpvAccessQualifierWriteOnly:
+ var->data.image.write_only = true;
+ break;
+ default:
+ break;
+ }
+ } else if (glsl_type_is_sampler(interface_type->type)) {
+ b->shader->info.num_textures++;
+ } else {
+ assert(glsl_type_is_struct(interface_type->type));
+ b->shader->info.num_ubos++;
+ }
+ }
+ }
+ break;
+ case SpvStorageClassPushConstant:
+ assert(interface_type && interface_type->block);
+ var->data.mode = nir_var_uniform;
+ var->data.read_only = true;
+ var->data.descriptor_set = -1;
+ var->data.binding = -1;
+
+ /* We have exactly one push constant block */
+ assert(b->shader->num_uniforms == 0);
+ b->shader->num_uniforms = vtn_type_block_size(type) * 4;
+ break;
+ case SpvStorageClassInput:
+ var->data.mode = nir_var_shader_in;
+ var->data.read_only = true;
+ break;
+ case SpvStorageClassOutput:
+ var->data.mode = nir_var_shader_out;
+ break;
+ case SpvStorageClassPrivate:
+ var->data.mode = nir_var_global;
+ var->interface_type = NULL;
+ break;
+ case SpvStorageClassFunction:
+ var->data.mode = nir_var_local;
+ var->interface_type = NULL;
+ break;
+ case SpvStorageClassWorkgroup:
+ var->data.mode = nir_var_shared;
+ break;
+ case SpvStorageClassCrossWorkgroup:
+ case SpvStorageClassGeneric:
+ case SpvStorageClassAtomicCounter:
+ default:
+ unreachable("Unhandled variable storage class");
+ }
+
+ if (count > 4) {
+ assert(count == 5);
+ nir_constant *constant =
+ vtn_value(b, w[4], vtn_value_type_constant)->constant;
+ var->constant_initializer = nir_constant_clone(constant, var);
+ }
+
+ val->deref = nir_deref_var_create(b, var);
+ val->deref_type = type;
+
+ /* We handle decorations first because decorations might give us
+ * location information. We use the data.explicit_location field to
+ * note that the location provided is the "final" location. If
+ * data.explicit_location == false, this means that it's relative to
+ * whatever the base location is.
+ */
+ vtn_foreach_decoration(b, val, var_decoration_cb, var);
+
+ if (!var->data.explicit_location) {
+ if (b->shader->stage == MESA_SHADER_FRAGMENT &&
+ var->data.mode == nir_var_shader_out) {
+ var->data.location += FRAG_RESULT_DATA0;
+ } else if (b->shader->stage == MESA_SHADER_VERTEX &&
+ var->data.mode == nir_var_shader_in) {
+ var->data.location += VERT_ATTRIB_GENERIC0;
+ } else if (var->data.mode == nir_var_shader_in ||
+ var->data.mode == nir_var_shader_out) {
+ var->data.location += VARYING_SLOT_VAR0;
+ }
+ }
+
+ /* XXX: Work around what appears to be a glslang bug. While the
+ * SPIR-V spec doesn't say that setting a descriptor set on a push
+ * constant is invalid, it certainly makes no sense. However, at
+ * some point, glslang started setting descriptor set 0 on push
+ * constants for some unknown reason. Hopefully this can be removed
+ * at some point in the future.
+ */
+ if (storage_class == SpvStorageClassPushConstant) {
+ var->data.descriptor_set = -1;
+ var->data.binding = -1;
+ }
+
+ /* Interface block variables aren't actually going to be referenced
+ * by the generated NIR, so we don't put them in the list
+ */
+ if (var->interface_type && glsl_type_is_struct(var->interface_type))
+ break;
+
+ if (var->data.mode == nir_var_local) {
+ nir_function_impl_add_variable(b->impl, var);
+ } else {
+ nir_shader_add_variable(b->shader, var);
+ }
+
+ break;
+ }
+
+ case SpvOpAccessChain:
+ case SpvOpInBoundsAccessChain: {
+ nir_deref_var *base;
+ struct vtn_value *base_val = vtn_untyped_value(b, w[3]);
+ if (base_val->value_type == vtn_value_type_sampled_image) {
+ /* This is rather insane. SPIR-V allows you to use OpSampledImage
+ * to combine an array of images with a single sampler to get an
+ * array of sampled images that all share the same sampler.
+ * Fortunately, this means that we can more-or-less ignore the
+ * sampler when crawling the access chain, but it does leave us
+ * with this rather awkward little special-case.
+ */
+ base = base_val->sampled_image->image;
+ } else {
+ assert(base_val->value_type == vtn_value_type_deref);
+ base = base_val->deref;
+ }
+
+ nir_deref_var *deref = nir_deref_as_var(nir_copy_deref(b, &base->deref));
+ struct vtn_type *deref_type = vtn_value(b, w[3], vtn_value_type_deref)->deref_type;
+
+ nir_deref *tail = &deref->deref;
+ while (tail->child)
+ tail = tail->child;
+
+ for (unsigned i = 0; i < count - 4; i++) {
+ assert(w[i + 4] < b->value_id_bound);
+ struct vtn_value *idx_val = &b->values[w[i + 4]];
+
+ enum glsl_base_type base_type = glsl_get_base_type(tail->type);
+ switch (base_type) {
+ case GLSL_TYPE_UINT:
+ case GLSL_TYPE_INT:
+ case GLSL_TYPE_FLOAT:
+ case GLSL_TYPE_DOUBLE:
+ case GLSL_TYPE_BOOL:
+ case GLSL_TYPE_ARRAY: {
+ nir_deref_array *deref_arr = nir_deref_array_create(b);
+ if (base_type == GLSL_TYPE_ARRAY ||
+ glsl_type_is_matrix(tail->type)) {
+ deref_type = deref_type->array_element;
+ } else {
+ assert(glsl_type_is_vector(tail->type));
+ deref_type = ralloc(b, struct vtn_type);
+ deref_type->type = glsl_scalar_type(base_type);
+ }
+
+ deref_arr->deref.type = deref_type->type;
+
+ if (idx_val->value_type == vtn_value_type_constant) {
+ unsigned idx = idx_val->constant->value.u[0];
+ deref_arr->deref_array_type = nir_deref_array_type_direct;
+ deref_arr->base_offset = idx;
+ } else {
+ assert(idx_val->value_type == vtn_value_type_ssa);
+ assert(glsl_type_is_scalar(idx_val->ssa->type));
+ deref_arr->deref_array_type = nir_deref_array_type_indirect;
+ deref_arr->base_offset = 0;
+ deref_arr->indirect = nir_src_for_ssa(idx_val->ssa->def);
+ }
+ tail->child = &deref_arr->deref;
+ break;
+ }
+
+ case GLSL_TYPE_STRUCT: {
+ assert(idx_val->value_type == vtn_value_type_constant);
+ unsigned idx = idx_val->constant->value.u[0];
+ deref_type = deref_type->members[idx];
+ nir_deref_struct *deref_struct = nir_deref_struct_create(b, idx);
+ deref_struct->deref.type = deref_type->type;
+ tail->child = &deref_struct->deref;
+ break;
+ }
+ default:
+ unreachable("Invalid type for deref");
+ }
+
+ if (deref_type->is_builtin) {
+ /* If we encounter a builtin, we throw away the ress of the
+ * access chain, jump to the builtin, and keep building.
+ */
+ const struct glsl_type *builtin_type = deref_type->type;
+
+ nir_deref_array *per_vertex_deref = NULL;
+ if (glsl_type_is_array(base->var->type)) {
+ /* This builtin is a per-vertex builtin */
+ assert(b->shader->stage == MESA_SHADER_GEOMETRY);
+ assert(base->var->data.mode == nir_var_shader_in);
+ builtin_type = glsl_array_type(builtin_type,
+ b->shader->info.gs.vertices_in);
+
+ /* The first non-var deref should be an array deref. */
+ assert(deref->deref.child->deref_type ==
+ nir_deref_type_array);
+ per_vertex_deref = nir_deref_as_array(deref->deref.child);
+ }
+
+ nir_variable *builtin = get_builtin_variable(b,
+ base->var->data.mode,
+ builtin_type,
+ deref_type->builtin);
+ deref = nir_deref_var_create(b, builtin);
+
+ if (per_vertex_deref) {
+ /* Since deref chains start at the variable, we can just
+ * steal that link and use it.
+ */
+ deref->deref.child = &per_vertex_deref->deref;
+ per_vertex_deref->deref.child = NULL;
+ per_vertex_deref->deref.type =
+ glsl_get_array_element(builtin_type);
+
+ tail = &per_vertex_deref->deref;
+ } else {
+ tail = &deref->deref;
+ }
+ } else {
+ tail = tail->child;
+ }
+ }
+
+ /* For uniform blocks, we don't resolve the access chain until we
+ * actually access the variable, so we need to keep around the original
+ * type of the variable.
+ */
+ if (variable_is_external_block(base->var))
+ deref_type = vtn_value(b, w[3], vtn_value_type_deref)->deref_type;
+
+ if (base_val->value_type == vtn_value_type_sampled_image) {
+ struct vtn_value *val =
+ vtn_push_value(b, w[2], vtn_value_type_sampled_image);
+ val->sampled_image = ralloc(b, struct vtn_sampled_image);
+ val->sampled_image->image = deref;
+ val->sampled_image->sampler = base_val->sampled_image->sampler;
+ } else {
+ struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_deref);
+ val->deref = deref;
+ val->deref_type = deref_type;
+ }
+
+ break;
+ }
+
+ case SpvOpCopyMemory: {
+ struct vtn_value *dest = vtn_value(b, w[1], vtn_value_type_deref);
+ struct vtn_value *src = vtn_value(b, w[2], vtn_value_type_deref);
+
+ vtn_variable_copy(b, dest->deref, dest->deref_type,
+ src->deref, src->deref_type);
+ break;
+ }
+
+ case SpvOpLoad: {
+ nir_deref_var *src = vtn_value(b, w[3], vtn_value_type_deref)->deref;
+ struct vtn_type *src_type =
+ vtn_value(b, w[3], vtn_value_type_deref)->deref_type;
+
+ if (src->var->interface_type &&
+ (glsl_type_is_sampler(src->var->interface_type) ||
+ glsl_type_is_image(src->var->interface_type))) {
+ vtn_push_value(b, w[2], vtn_value_type_deref)->deref = src;
+ return;
+ }
+
+ struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_ssa);
+ val->ssa = vtn_variable_load(b, src, src_type);
+ break;
+ }
+
+ case SpvOpStore: {
+ nir_deref_var *dest = vtn_value(b, w[1], vtn_value_type_deref)->deref;
+ struct vtn_type *dest_type =
+ vtn_value(b, w[1], vtn_value_type_deref)->deref_type;
+ struct vtn_ssa_value *src = vtn_ssa_value(b, w[2]);
+ vtn_variable_store(b, src, dest, dest_type);
+ break;
+ }
+
+ case SpvOpArrayLength: {
+ struct vtn_value *v_deref = vtn_value(b, w[3], vtn_value_type_deref);
+ struct vtn_type *type = v_deref->deref_type;
+ const uint32_t offset = type->offsets[w[4]];
+ const uint32_t stride = type->members[w[4]]->stride;
+ nir_deref *n_deref = &v_deref->deref->deref;
+ nir_ssa_def *index =
+ get_vulkan_resource_index(b, &n_deref, &type);
+ nir_intrinsic_instr *instr =
+ nir_intrinsic_instr_create(b->nb.shader,
+ nir_intrinsic_get_buffer_size);
+ instr->src[0] = nir_src_for_ssa(index);
+ nir_ssa_dest_init(&instr->instr, &instr->dest, 1, NULL);
+ nir_builder_instr_insert(&b->nb, &instr->instr);
+ nir_ssa_def *buf_size = &instr->dest.ssa;
+
+ /* array_length = max(buffer_size - offset, 0) / stride */
+ nir_ssa_def *array_length =
+ nir_idiv(&b->nb,
+ nir_imax(&b->nb,
+ nir_isub(&b->nb,
+ buf_size,
+ nir_imm_int(&b->nb, offset)),
+ nir_imm_int(&b->nb, 0u)),
+ nir_imm_int(&b->nb, stride));
+
+ struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_ssa);
+ val->ssa = vtn_create_ssa_value(b, glsl_uint_type());
+ val->ssa->def = array_length;
+ break;
+ }
+
+ case SpvOpCopyMemorySized:
+ default:
+ unreachable("Unhandled opcode");
+ }
+}
+
+static void
+vtn_handle_function_call(struct vtn_builder *b, SpvOp opcode,
+ const uint32_t *w, unsigned count)
+{
+ struct nir_function *callee =
+ vtn_value(b, w[3], vtn_value_type_function)->func->impl->function;
+
+ nir_call_instr *call = nir_call_instr_create(b->nb.shader, callee);
+ for (unsigned i = 0; i < call->num_params; i++) {
+ unsigned arg_id = w[4 + i];
+ struct vtn_value *arg = vtn_untyped_value(b, arg_id);
+ if (arg->value_type == vtn_value_type_deref) {
+ call->params[i] =
+ nir_deref_as_var(nir_copy_deref(call, &arg->deref->deref));
+ } else {
+ struct vtn_ssa_value *arg_ssa = vtn_ssa_value(b, arg_id);
+
+ /* Make a temporary to store the argument in */
+ nir_variable *tmp =
+ nir_local_variable_create(b->impl, arg_ssa->type, "arg_tmp");
+ call->params[i] = nir_deref_var_create(call, tmp);
+
+ vtn_variable_store(b, arg_ssa, call->params[i], arg->type);
+ }
+ }
+
+ nir_variable *out_tmp = NULL;
+ if (!glsl_type_is_void(callee->return_type)) {
+ out_tmp = nir_local_variable_create(b->impl, callee->return_type,
+ "out_tmp");
+ call->return_deref = nir_deref_var_create(call, out_tmp);
+ }
+
+ nir_builder_instr_insert(&b->nb, &call->instr);
+
+ if (glsl_type_is_void(callee->return_type)) {
+ vtn_push_value(b, w[2], vtn_value_type_undef);
+ } else {
+ struct vtn_type *rettype = vtn_value(b, w[1], vtn_value_type_type)->type;
+ struct vtn_value *retval = vtn_push_value(b, w[2], vtn_value_type_ssa);
+ retval->ssa = vtn_variable_load(b, call->return_deref, rettype);
+ }
+}
+
+struct vtn_ssa_value *
+vtn_create_ssa_value(struct vtn_builder *b, const struct glsl_type *type)
+{
+ struct vtn_ssa_value *val = rzalloc(b, struct vtn_ssa_value);
+ val->type = type;
+
+ if (!glsl_type_is_vector_or_scalar(type)) {
+ unsigned elems = glsl_get_length(type);
+ val->elems = ralloc_array(b, struct vtn_ssa_value *, elems);
+ for (unsigned i = 0; i < elems; i++) {
+ const struct glsl_type *child_type;
+
+ switch (glsl_get_base_type(type)) {
+ case GLSL_TYPE_INT:
+ case GLSL_TYPE_UINT:
+ case GLSL_TYPE_BOOL:
+ case GLSL_TYPE_FLOAT:
+ case GLSL_TYPE_DOUBLE:
+ child_type = glsl_get_column_type(type);
+ break;
+ case GLSL_TYPE_ARRAY:
+ child_type = glsl_get_array_element(type);
+ break;
+ case GLSL_TYPE_STRUCT:
+ child_type = glsl_get_struct_field(type, i);
+ break;
+ default:
+ unreachable("unkown base type");
+ }
+
+ val->elems[i] = vtn_create_ssa_value(b, child_type);
+ }
+ }
+
+ return val;
+}
+
+static nir_tex_src
+vtn_tex_src(struct vtn_builder *b, unsigned index, nir_tex_src_type type)
+{
+ nir_tex_src src;
+ src.src = nir_src_for_ssa(vtn_ssa_value(b, index)->def);
+ src.src_type = type;
+ return src;
+}
+
+static void
+vtn_handle_texture(struct vtn_builder *b, SpvOp opcode,
+ const uint32_t *w, unsigned count)
+{
+ if (opcode == SpvOpSampledImage) {
+ struct vtn_value *val =
+ vtn_push_value(b, w[2], vtn_value_type_sampled_image);
+ val->sampled_image = ralloc(b, struct vtn_sampled_image);
+ val->sampled_image->image =
+ vtn_value(b, w[3], vtn_value_type_deref)->deref;
+ val->sampled_image->sampler =
+ vtn_value(b, w[4], vtn_value_type_deref)->deref;
+ return;
+ }
+
+ struct vtn_type *ret_type = vtn_value(b, w[1], vtn_value_type_type)->type;
+ struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_ssa);
+
+ struct vtn_sampled_image sampled;
+ struct vtn_value *sampled_val = vtn_untyped_value(b, w[3]);
+ if (sampled_val->value_type == vtn_value_type_sampled_image) {
+ sampled = *sampled_val->sampled_image;
+ } else {
+ assert(sampled_val->value_type == vtn_value_type_deref);
+ sampled.image = NULL;
+ sampled.sampler = sampled_val->deref;
+ }
+
+ nir_tex_src srcs[8]; /* 8 should be enough */
+ nir_tex_src *p = srcs;
+
+ unsigned idx = 4;
+
+ bool has_coord = false;
+ switch (opcode) {
+ case SpvOpImageSampleImplicitLod:
+ case SpvOpImageSampleExplicitLod:
+ case SpvOpImageSampleDrefImplicitLod:
+ case SpvOpImageSampleDrefExplicitLod:
+ case SpvOpImageSampleProjImplicitLod:
+ case SpvOpImageSampleProjExplicitLod:
+ case SpvOpImageSampleProjDrefImplicitLod:
+ case SpvOpImageSampleProjDrefExplicitLod:
+ case SpvOpImageFetch:
+ case SpvOpImageGather:
+ case SpvOpImageDrefGather:
+ case SpvOpImageQueryLod: {
+ /* All these types have the coordinate as their first real argument */
+ struct vtn_ssa_value *coord = vtn_ssa_value(b, w[idx++]);
+ has_coord = true;
+ p->src = nir_src_for_ssa(coord->def);
+ p->src_type = nir_tex_src_coord;
+ p++;
+ break;
+ }
+
+ default:
+ break;
+ }
+
+ /* These all have an explicit depth value as their next source */
+ switch (opcode) {
+ case SpvOpImageSampleDrefImplicitLod:
+ case SpvOpImageSampleDrefExplicitLod:
+ case SpvOpImageSampleProjDrefImplicitLod:
+ case SpvOpImageSampleProjDrefExplicitLod:
+ (*p++) = vtn_tex_src(b, w[idx++], nir_tex_src_comparitor);
+ break;
+ default:
+ break;
+ }
+
+ /* Figure out the base texture operation */
+ nir_texop texop;
+ switch (opcode) {
+ case SpvOpImageSampleImplicitLod:
+ case SpvOpImageSampleDrefImplicitLod:
+ case SpvOpImageSampleProjImplicitLod:
+ case SpvOpImageSampleProjDrefImplicitLod:
+ texop = nir_texop_tex;
+ break;
+
+ case SpvOpImageSampleExplicitLod:
+ case SpvOpImageSampleDrefExplicitLod:
+ case SpvOpImageSampleProjExplicitLod:
+ case SpvOpImageSampleProjDrefExplicitLod:
+ texop = nir_texop_txl;
+ break;
+
+ case SpvOpImageFetch:
+ texop = nir_texop_txf;
+ break;
+
+ case SpvOpImageGather:
+ case SpvOpImageDrefGather:
+ texop = nir_texop_tg4;
+ break;
+
+ case SpvOpImageQuerySizeLod:
+ case SpvOpImageQuerySize:
+ texop = nir_texop_txs;
+ break;
+
+ case SpvOpImageQueryLod:
+ texop = nir_texop_lod;
+ break;
+
+ case SpvOpImageQueryLevels:
+ texop = nir_texop_query_levels;
+ break;
+
+ case SpvOpImageQuerySamples:
+ default:
+ unreachable("Unhandled opcode");
+ }
+
+ /* Now we need to handle some number of optional arguments */
+ if (idx < count) {
+ uint32_t operands = w[idx++];
+
+ if (operands & SpvImageOperandsBiasMask) {
+ assert(texop == nir_texop_tex);
+ texop = nir_texop_txb;
+ (*p++) = vtn_tex_src(b, w[idx++], nir_tex_src_bias);
+ }
+
+ if (operands & SpvImageOperandsLodMask) {
+ assert(texop == nir_texop_txl || texop == nir_texop_txf ||
+ texop == nir_texop_txs);
+ (*p++) = vtn_tex_src(b, w[idx++], nir_tex_src_lod);
+ }
+
+ if (operands & SpvImageOperandsGradMask) {
+ assert(texop == nir_texop_tex);
+ texop = nir_texop_txd;
+ (*p++) = vtn_tex_src(b, w[idx++], nir_tex_src_ddx);
+ (*p++) = vtn_tex_src(b, w[idx++], nir_tex_src_ddy);
+ }
+
+ if (operands & SpvImageOperandsOffsetMask ||
+ operands & SpvImageOperandsConstOffsetMask)
+ (*p++) = vtn_tex_src(b, w[idx++], nir_tex_src_offset);
+
+ if (operands & SpvImageOperandsConstOffsetsMask)
+ assert(!"Constant offsets to texture gather not yet implemented");
+
+ if (operands & SpvImageOperandsSampleMask) {
+ assert(texop == nir_texop_txf);
+ texop = nir_texop_txf_ms;
+ (*p++) = vtn_tex_src(b, w[idx++], nir_tex_src_ms_index);
+ }
+ }
+ /* We should have now consumed exactly all of the arguments */
+ assert(idx == count);
+
+ nir_tex_instr *instr = nir_tex_instr_create(b->shader, p - srcs);
+ instr->op = texop;
+
+ memcpy(instr->src, srcs, instr->num_srcs * sizeof(*instr->src));
+
+ const struct glsl_type *image_type;
+ if (sampled.image) {
+ image_type = nir_deref_tail(&sampled.image->deref)->type;
+ } else {
+ image_type = nir_deref_tail(&sampled.sampler->deref)->type;
+ }
+
+ instr->sampler_dim = glsl_get_sampler_dim(image_type);
+ instr->is_array = glsl_sampler_type_is_array(image_type);
+ instr->is_shadow = glsl_sampler_type_is_shadow(image_type);
+ instr->is_new_style_shadow = instr->is_shadow;
+
+ if (has_coord) {
+ switch (instr->sampler_dim) {
+ case GLSL_SAMPLER_DIM_1D:
+ case GLSL_SAMPLER_DIM_BUF:
+ instr->coord_components = 1;
+ break;
+ case GLSL_SAMPLER_DIM_2D:
+ case GLSL_SAMPLER_DIM_RECT:
+ instr->coord_components = 2;
+ break;
+ case GLSL_SAMPLER_DIM_3D:
+ case GLSL_SAMPLER_DIM_CUBE:
+ case GLSL_SAMPLER_DIM_MS:
+ instr->coord_components = 3;
+ break;
+ default:
+ assert("Invalid sampler type");
+ }
+
+ if (instr->is_array)
+ instr->coord_components++;
+ } else {
+ instr->coord_components = 0;
+ }
+
+ switch (glsl_get_sampler_result_type(image_type)) {
+ case GLSL_TYPE_FLOAT: instr->dest_type = nir_type_float; break;
+ case GLSL_TYPE_INT: instr->dest_type = nir_type_int; break;
+ case GLSL_TYPE_UINT: instr->dest_type = nir_type_uint; break;
+ case GLSL_TYPE_BOOL: instr->dest_type = nir_type_bool; break;
+ default:
+ unreachable("Invalid base type for sampler result");
+ }
+
+ instr->sampler =
+ nir_deref_as_var(nir_copy_deref(instr, &sampled.sampler->deref));
+ if (sampled.image) {
+ instr->texture =
+ nir_deref_as_var(nir_copy_deref(instr, &sampled.image->deref));
+ } else {
+ instr->texture = NULL;
+ }
+
+ nir_ssa_dest_init(&instr->instr, &instr->dest,
+ nir_tex_instr_dest_size(instr), NULL);
+
+ assert(glsl_get_vector_elements(ret_type->type) ==
+ nir_tex_instr_dest_size(instr));
+
+ val->ssa = vtn_create_ssa_value(b, ret_type->type);
+ val->ssa->def = &instr->dest.ssa;
+
+ nir_builder_instr_insert(&b->nb, &instr->instr);
+}
+
+static nir_ssa_def *
+get_image_coord(struct vtn_builder *b, uint32_t value)
+{
+ struct vtn_ssa_value *coord = vtn_ssa_value(b, value);
+
+ /* The image_load_store intrinsics assume a 4-dim coordinate */
+ unsigned dim = glsl_get_vector_elements(coord->type);
+ unsigned swizzle[4];
+ for (unsigned i = 0; i < 4; i++)
+ swizzle[i] = MIN2(i, dim - 1);
+
+ return nir_swizzle(&b->nb, coord->def, swizzle, 4, false);
+}
+
+static void
+vtn_handle_image(struct vtn_builder *b, SpvOp opcode,
+ const uint32_t *w, unsigned count)
+{
+ /* Just get this one out of the way */
+ if (opcode == SpvOpImageTexelPointer) {
+ struct vtn_value *val =
+ vtn_push_value(b, w[2], vtn_value_type_image_pointer);
+ val->image = ralloc(b, struct vtn_image_pointer);
+
+ val->image->deref = vtn_value(b, w[3], vtn_value_type_deref)->deref;
+ val->image->coord = get_image_coord(b, w[4]);
+ val->image->sample = vtn_ssa_value(b, w[5])->def;
+ return;
+ }
+
+ struct vtn_image_pointer image;
+
+ switch (opcode) {
+ case SpvOpAtomicExchange:
+ case SpvOpAtomicCompareExchange:
+ case SpvOpAtomicCompareExchangeWeak:
+ case SpvOpAtomicIIncrement:
+ case SpvOpAtomicIDecrement:
+ case SpvOpAtomicIAdd:
+ case SpvOpAtomicISub:
+ case SpvOpAtomicSMin:
+ case SpvOpAtomicUMin:
+ case SpvOpAtomicSMax:
+ case SpvOpAtomicUMax:
+ case SpvOpAtomicAnd:
+ case SpvOpAtomicOr:
+ case SpvOpAtomicXor:
+ image = *vtn_value(b, w[3], vtn_value_type_image_pointer)->image;
+ break;
+
+ case SpvOpImageQuerySize:
+ image.deref = vtn_value(b, w[3], vtn_value_type_deref)->deref;
+ image.coord = NULL;
+ image.sample = NULL;
+ break;
+
+ case SpvOpImageRead:
+ image.deref = vtn_value(b, w[3], vtn_value_type_deref)->deref;
+ image.coord = get_image_coord(b, w[4]);
+
+ if (count > 5 && (w[5] & SpvImageOperandsSampleMask)) {
+ assert(w[5] == SpvImageOperandsSampleMask);
+ image.sample = vtn_ssa_value(b, w[6])->def;
+ } else {
+ image.sample = nir_ssa_undef(&b->nb, 1);
+ }
+ break;
+
+ case SpvOpImageWrite:
+ image.deref = vtn_value(b, w[1], vtn_value_type_deref)->deref;
+ image.coord = get_image_coord(b, w[2]);
+
+ /* texel = w[3] */
+
+ if (count > 4 && (w[4] & SpvImageOperandsSampleMask)) {
+ assert(w[4] == SpvImageOperandsSampleMask);
+ image.sample = vtn_ssa_value(b, w[5])->def;
+ } else {
+ image.sample = nir_ssa_undef(&b->nb, 1);
+ }
+ break;
+
+ default:
+ unreachable("Invalid image opcode");
+ }
+
+ nir_intrinsic_op op;
+ switch (opcode) {
+#define OP(S, N) case SpvOp##S: op = nir_intrinsic_image_##N; break;
+ OP(ImageQuerySize, size)
+ OP(ImageRead, load)
+ OP(ImageWrite, store)
+ OP(AtomicExchange, atomic_exchange)
+ OP(AtomicCompareExchange, atomic_comp_swap)
+ OP(AtomicIIncrement, atomic_add)
+ OP(AtomicIDecrement, atomic_add)
+ OP(AtomicIAdd, atomic_add)
+ OP(AtomicISub, atomic_add)
+ OP(AtomicSMin, atomic_min)
+ OP(AtomicUMin, atomic_min)
+ OP(AtomicSMax, atomic_max)
+ OP(AtomicUMax, atomic_max)
+ OP(AtomicAnd, atomic_and)
+ OP(AtomicOr, atomic_or)
+ OP(AtomicXor, atomic_xor)
+#undef OP
+ default:
+ unreachable("Invalid image opcode");
+ }
+
+ nir_intrinsic_instr *intrin = nir_intrinsic_instr_create(b->shader, op);
+ intrin->variables[0] =
+ nir_deref_as_var(nir_copy_deref(&intrin->instr, &image.deref->deref));
+
+ /* ImageQuerySize doesn't take any extra parameters */
+ if (opcode != SpvOpImageQuerySize) {
+ /* The image coordinate is always 4 components but we may not have that
+ * many. Swizzle to compensate.
+ */
+ unsigned swiz[4];
+ for (unsigned i = 0; i < 4; i++)
+ swiz[i] = i < image.coord->num_components ? i : 0;
+ intrin->src[0] = nir_src_for_ssa(nir_swizzle(&b->nb, image.coord,
+ swiz, 4, false));
+ intrin->src[1] = nir_src_for_ssa(image.sample);
+ }
+
+ switch (opcode) {
+ case SpvOpImageQuerySize:
+ case SpvOpImageRead:
+ break;
+ case SpvOpImageWrite:
+ intrin->src[2] = nir_src_for_ssa(vtn_ssa_value(b, w[3])->def);
+ break;
+ case SpvOpAtomicIIncrement:
+ intrin->src[2] = nir_src_for_ssa(nir_imm_int(&b->nb, 1));
+ break;
+ case SpvOpAtomicIDecrement:
+ intrin->src[2] = nir_src_for_ssa(nir_imm_int(&b->nb, -1));
+ break;
+
+ case SpvOpAtomicExchange:
+ case SpvOpAtomicIAdd:
+ case SpvOpAtomicSMin:
+ case SpvOpAtomicUMin:
+ case SpvOpAtomicSMax:
+ case SpvOpAtomicUMax:
+ case SpvOpAtomicAnd:
+ case SpvOpAtomicOr:
+ case SpvOpAtomicXor:
+ intrin->src[2] = nir_src_for_ssa(vtn_ssa_value(b, w[6])->def);
+ break;
+
+ case SpvOpAtomicCompareExchange:
+ intrin->src[2] = nir_src_for_ssa(vtn_ssa_value(b, w[7])->def);
+ intrin->src[3] = nir_src_for_ssa(vtn_ssa_value(b, w[6])->def);
+ break;
+
+ case SpvOpAtomicISub:
+ intrin->src[2] = nir_src_for_ssa(nir_ineg(&b->nb, vtn_ssa_value(b, w[6])->def));
+ break;
+
+ default:
+ unreachable("Invalid image opcode");
+ }
+
+ if (opcode != SpvOpImageWrite) {
+ struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_ssa);
+ struct vtn_type *type = vtn_value(b, w[1], vtn_value_type_type)->type;
+ nir_ssa_dest_init(&intrin->instr, &intrin->dest, 4, NULL);
+
+ nir_builder_instr_insert(&b->nb, &intrin->instr);
+
+ /* The image intrinsics always return 4 channels but we may not want
+ * that many. Emit a mov to trim it down.
+ */
+ unsigned swiz[4] = {0, 1, 2, 3};
+ val->ssa = vtn_create_ssa_value(b, type->type);
+ val->ssa->def = nir_swizzle(&b->nb, &intrin->dest.ssa, swiz,
+ glsl_get_vector_elements(type->type), false);
+ } else {
+ nir_builder_instr_insert(&b->nb, &intrin->instr);
+ }
+}
+
+static nir_intrinsic_op
+get_ssbo_nir_atomic_op(SpvOp opcode)
+{
+ switch (opcode) {
+#define OP(S, N) case SpvOp##S: return nir_intrinsic_ssbo_##N;
+ OP(AtomicExchange, atomic_exchange)
+ OP(AtomicCompareExchange, atomic_comp_swap)
+ OP(AtomicIIncrement, atomic_add)
+ OP(AtomicIDecrement, atomic_add)
+ OP(AtomicIAdd, atomic_add)
+ OP(AtomicISub, atomic_add)
+ OP(AtomicSMin, atomic_imin)
+ OP(AtomicUMin, atomic_umin)
+ OP(AtomicSMax, atomic_imax)
+ OP(AtomicUMax, atomic_umax)
+ OP(AtomicAnd, atomic_and)
+ OP(AtomicOr, atomic_or)
+ OP(AtomicXor, atomic_xor)
+#undef OP
+ default:
+ unreachable("Invalid SSBO atomic");
+ }
+}
+
+static nir_intrinsic_op
+get_shared_nir_atomic_op(SpvOp opcode)
+{
+ switch (opcode) {
+#define OP(S, N) case SpvOp##S: return nir_intrinsic_var_##N;
+ OP(AtomicExchange, atomic_exchange)
+ OP(AtomicCompareExchange, atomic_comp_swap)
+ OP(AtomicIIncrement, atomic_add)
+ OP(AtomicIDecrement, atomic_add)
+ OP(AtomicIAdd, atomic_add)
+ OP(AtomicISub, atomic_add)
+ OP(AtomicSMin, atomic_imin)
+ OP(AtomicUMin, atomic_umin)
+ OP(AtomicSMax, atomic_imax)
+ OP(AtomicUMax, atomic_umax)
+ OP(AtomicAnd, atomic_and)
+ OP(AtomicOr, atomic_or)
+ OP(AtomicXor, atomic_xor)
+#undef OP
+ default:
+ unreachable("Invalid shared atomic");
+ }
+}
+
+static nir_ssa_def *
+get_ssbo_atomic_offset(struct vtn_builder *b, nir_deref *deref, struct vtn_type *type)
+{
+ nir_ssa_def *offset = nir_imm_int(&b->nb, 0);
+
+ while (deref->child) {
+ deref = deref->child;
+ switch (deref->deref_type) {
+ case nir_deref_type_array:
+ offset = nir_iadd(&b->nb, offset,
+ nir_imul(&b->nb, deref_array_offset(b, deref),
+ nir_imm_int(&b->nb, type->stride)));
+ type = type->array_element;
+ continue;
+
+ case nir_deref_type_struct: {
+ unsigned member = nir_deref_as_struct(deref)->index;
+ offset = nir_iadd(&b->nb, offset,
+ nir_imm_int(&b->nb, type->offsets[member]));
+ type = type->members[member];
+ continue;
+ }
+
+ default:
+ unreachable("Invalid deref type");
+ }
+ }
+
+ return offset;
+}
+
+static void
+fill_common_atomic_sources(struct vtn_builder *b, SpvOp opcode,
+ const uint32_t *w, nir_src *src)
+{
+ switch (opcode) {
+ case SpvOpAtomicIIncrement:
+ src[0] = nir_src_for_ssa(nir_imm_int(&b->nb, 1));
+ break;
+
+ case SpvOpAtomicIDecrement:
+ src[0] = nir_src_for_ssa(nir_imm_int(&b->nb, -1));
+ break;
+
+ case SpvOpAtomicISub:
+ src[0] =
+ nir_src_for_ssa(nir_ineg(&b->nb, vtn_ssa_value(b, w[6])->def));
+ break;
+
+ case SpvOpAtomicCompareExchange:
+ src[0] = nir_src_for_ssa(vtn_ssa_value(b, w[7])->def);
+ src[1] = nir_src_for_ssa(vtn_ssa_value(b, w[8])->def);
+ break;
+ /* Fall through */
+
+ case SpvOpAtomicExchange:
+ case SpvOpAtomicIAdd:
+ case SpvOpAtomicSMin:
+ case SpvOpAtomicUMin:
+ case SpvOpAtomicSMax:
+ case SpvOpAtomicUMax:
+ case SpvOpAtomicAnd:
+ case SpvOpAtomicOr:
+ case SpvOpAtomicXor:
+ src[0] = nir_src_for_ssa(vtn_ssa_value(b, w[6])->def);
+ break;
+
+ default:
+ unreachable("Invalid SPIR-V atomic");
+ }
+}
+
+static void
+vtn_handle_ssbo_or_shared_atomic(struct vtn_builder *b, SpvOp opcode,
+ const uint32_t *w, unsigned count)
+{
+ struct vtn_value *pointer = vtn_value(b, w[3], vtn_value_type_deref);
+ struct vtn_type *type = pointer->deref_type;
+ nir_deref *deref = &pointer->deref->deref;
+ nir_intrinsic_instr *atomic;
+
+ /*
+ SpvScope scope = w[4];
+ SpvMemorySemanticsMask semantics = w[5];
+ */
+
+ if (pointer->deref->var->data.mode == nir_var_shared) {
+ nir_intrinsic_op op = get_shared_nir_atomic_op(opcode);
+ atomic = nir_intrinsic_instr_create(b->nb.shader, op);
+ atomic->variables[0] =
+ nir_deref_as_var(nir_copy_deref(atomic, &pointer->deref->deref));
+ fill_common_atomic_sources(b, opcode, w, &atomic->src[0]);
+ } else {
+ nir_ssa_def *index = get_vulkan_resource_index(b, &deref, &type);
+ nir_ssa_def *offset = get_ssbo_atomic_offset(b, deref, type);
+ nir_intrinsic_op op = get_ssbo_nir_atomic_op(opcode);
+
+ atomic = nir_intrinsic_instr_create(b->nb.shader, op);
+ atomic->src[0] = nir_src_for_ssa(index);
+ atomic->src[1] = nir_src_for_ssa(offset);
+ fill_common_atomic_sources(b, opcode, w, &atomic->src[2]);
+ }
+
+ nir_ssa_dest_init(&atomic->instr, &atomic->dest, 1, NULL);
+
+ struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_ssa);
+ val->ssa = rzalloc(b, struct vtn_ssa_value);
+ val->ssa->def = &atomic->dest.ssa;
+ val->ssa->type = type->type;
+
+ nir_builder_instr_insert(&b->nb, &atomic->instr);
+}
+
+static nir_alu_instr *
+create_vec(nir_shader *shader, unsigned num_components)
+{
+ nir_op op;
+ switch (num_components) {
+ case 1: op = nir_op_fmov; break;
+ case 2: op = nir_op_vec2; break;
+ case 3: op = nir_op_vec3; break;
+ case 4: op = nir_op_vec4; break;
+ default: unreachable("bad vector size");
+ }
+
+ nir_alu_instr *vec = nir_alu_instr_create(shader, op);
+ nir_ssa_dest_init(&vec->instr, &vec->dest.dest, num_components, NULL);
+ vec->dest.write_mask = (1 << num_components) - 1;
+
+ return vec;
+}
+
+struct vtn_ssa_value *
+vtn_ssa_transpose(struct vtn_builder *b, struct vtn_ssa_value *src)
+{
+ if (src->transposed)
+ return src->transposed;
+
+ struct vtn_ssa_value *dest =
+ vtn_create_ssa_value(b, glsl_transposed_type(src->type));
+
+ for (unsigned i = 0; i < glsl_get_matrix_columns(dest->type); i++) {
+ nir_alu_instr *vec = create_vec(b->shader,
+ glsl_get_matrix_columns(src->type));
+ if (glsl_type_is_vector_or_scalar(src->type)) {
+ vec->src[0].src = nir_src_for_ssa(src->def);
+ vec->src[0].swizzle[0] = i;
+ } else {
+ for (unsigned j = 0; j < glsl_get_matrix_columns(src->type); j++) {
+ vec->src[j].src = nir_src_for_ssa(src->elems[j]->def);
+ vec->src[j].swizzle[0] = i;
+ }
+ }
+ nir_builder_instr_insert(&b->nb, &vec->instr);
+ dest->elems[i]->def = &vec->dest.dest.ssa;
+ }
+
+ dest->transposed = src;
+
+ return dest;
+}
+
+static nir_ssa_def *
+vtn_vector_extract(struct vtn_builder *b, nir_ssa_def *src, unsigned index)
+{
+ unsigned swiz[4] = { index };
+ return nir_swizzle(&b->nb, src, swiz, 1, true);
+}
+
+
+static nir_ssa_def *
+vtn_vector_insert(struct vtn_builder *b, nir_ssa_def *src, nir_ssa_def *insert,
+ unsigned index)
+{
+ nir_alu_instr *vec = create_vec(b->shader, src->num_components);
+
+ for (unsigned i = 0; i < src->num_components; i++) {
+ if (i == index) {
+ vec->src[i].src = nir_src_for_ssa(insert);
+ } else {
+ vec->src[i].src = nir_src_for_ssa(src);
+ vec->src[i].swizzle[0] = i;
+ }
+ }
+
+ nir_builder_instr_insert(&b->nb, &vec->instr);
+
+ return &vec->dest.dest.ssa;
+}
+
+static nir_ssa_def *
+vtn_vector_extract_dynamic(struct vtn_builder *b, nir_ssa_def *src,
+ nir_ssa_def *index)
+{
+ nir_ssa_def *dest = vtn_vector_extract(b, src, 0);
+ for (unsigned i = 1; i < src->num_components; i++)
+ dest = nir_bcsel(&b->nb, nir_ieq(&b->nb, index, nir_imm_int(&b->nb, i)),
+ vtn_vector_extract(b, src, i), dest);
+
+ return dest;
+}
+
+static nir_ssa_def *
+vtn_vector_insert_dynamic(struct vtn_builder *b, nir_ssa_def *src,
+ nir_ssa_def *insert, nir_ssa_def *index)
+{
+ nir_ssa_def *dest = vtn_vector_insert(b, src, insert, 0);
+ for (unsigned i = 1; i < src->num_components; i++)
+ dest = nir_bcsel(&b->nb, nir_ieq(&b->nb, index, nir_imm_int(&b->nb, i)),
+ vtn_vector_insert(b, src, insert, i), dest);
+
+ return dest;
+}
+
+static nir_ssa_def *
+vtn_vector_shuffle(struct vtn_builder *b, unsigned num_components,
+ nir_ssa_def *src0, nir_ssa_def *src1,
+ const uint32_t *indices)
+{
+ nir_alu_instr *vec = create_vec(b->shader, num_components);
+
+ nir_ssa_undef_instr *undef = nir_ssa_undef_instr_create(b->shader, 1);
+ nir_builder_instr_insert(&b->nb, &undef->instr);
+
+ for (unsigned i = 0; i < num_components; i++) {
+ uint32_t index = indices[i];
+ if (index == 0xffffffff) {
+ vec->src[i].src = nir_src_for_ssa(&undef->def);
+ } else if (index < src0->num_components) {
+ vec->src[i].src = nir_src_for_ssa(src0);
+ vec->src[i].swizzle[0] = index;
+ } else {
+ vec->src[i].src = nir_src_for_ssa(src1);
+ vec->src[i].swizzle[0] = index - src0->num_components;
+ }
+ }
+
+ nir_builder_instr_insert(&b->nb, &vec->instr);
+
+ return &vec->dest.dest.ssa;
+}
+
+/*
+ * Concatentates a number of vectors/scalars together to produce a vector
+ */
+static nir_ssa_def *
+vtn_vector_construct(struct vtn_builder *b, unsigned num_components,
+ unsigned num_srcs, nir_ssa_def **srcs)
+{
+ nir_alu_instr *vec = create_vec(b->shader, num_components);
+
+ unsigned dest_idx = 0;
+ for (unsigned i = 0; i < num_srcs; i++) {
+ nir_ssa_def *src = srcs[i];
+ for (unsigned j = 0; j < src->num_components; j++) {
+ vec->src[dest_idx].src = nir_src_for_ssa(src);
+ vec->src[dest_idx].swizzle[0] = j;
+ dest_idx++;
+ }
+ }
+
+ nir_builder_instr_insert(&b->nb, &vec->instr);
+
+ return &vec->dest.dest.ssa;
+}
+
+static struct vtn_ssa_value *
+vtn_composite_copy(void *mem_ctx, struct vtn_ssa_value *src)
+{
+ struct vtn_ssa_value *dest = rzalloc(mem_ctx, struct vtn_ssa_value);
+ dest->type = src->type;
+
+ if (glsl_type_is_vector_or_scalar(src->type)) {
+ dest->def = src->def;
+ } else {
+ unsigned elems = glsl_get_length(src->type);
+
+ dest->elems = ralloc_array(mem_ctx, struct vtn_ssa_value *, elems);
+ for (unsigned i = 0; i < elems; i++)
+ dest->elems[i] = vtn_composite_copy(mem_ctx, src->elems[i]);
+ }
+
+ return dest;
+}
+
+static struct vtn_ssa_value *
+vtn_composite_insert(struct vtn_builder *b, struct vtn_ssa_value *src,
+ struct vtn_ssa_value *insert, const uint32_t *indices,
+ unsigned num_indices)
+{
+ struct vtn_ssa_value *dest = vtn_composite_copy(b, src);
+
+ struct vtn_ssa_value *cur = dest;
+ unsigned i;
+ for (i = 0; i < num_indices - 1; i++) {
+ cur = cur->elems[indices[i]];
+ }
+
+ if (glsl_type_is_vector_or_scalar(cur->type)) {
+ /* According to the SPIR-V spec, OpCompositeInsert may work down to
+ * the component granularity. In that case, the last index will be
+ * the index to insert the scalar into the vector.
+ */
+
+ cur->def = vtn_vector_insert(b, cur->def, insert->def, indices[i]);
+ } else {
+ cur->elems[indices[i]] = insert;
+ }
+
+ return dest;
+}
+
+static struct vtn_ssa_value *
+vtn_composite_extract(struct vtn_builder *b, struct vtn_ssa_value *src,
+ const uint32_t *indices, unsigned num_indices)
+{
+ struct vtn_ssa_value *cur = src;
+ for (unsigned i = 0; i < num_indices; i++) {
+ if (glsl_type_is_vector_or_scalar(cur->type)) {
+ assert(i == num_indices - 1);
+ /* According to the SPIR-V spec, OpCompositeExtract may work down to
+ * the component granularity. The last index will be the index of the
+ * vector to extract.
+ */
+
+ struct vtn_ssa_value *ret = rzalloc(b, struct vtn_ssa_value);
+ ret->type = glsl_scalar_type(glsl_get_base_type(cur->type));
+ ret->def = vtn_vector_extract(b, cur->def, indices[i]);
+ return ret;
+ } else {
+ cur = cur->elems[indices[i]];
+ }
+ }
+
+ return cur;
+}
+
+static void
+vtn_handle_composite(struct vtn_builder *b, SpvOp opcode,
+ const uint32_t *w, unsigned count)
+{
+ struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_ssa);
+ const struct glsl_type *type =
+ vtn_value(b, w[1], vtn_value_type_type)->type->type;
+ val->ssa = vtn_create_ssa_value(b, type);
+
+ switch (opcode) {
+ case SpvOpVectorExtractDynamic:
+ val->ssa->def = vtn_vector_extract_dynamic(b, vtn_ssa_value(b, w[3])->def,
+ vtn_ssa_value(b, w[4])->def);
+ break;
+
+ case SpvOpVectorInsertDynamic:
+ val->ssa->def = vtn_vector_insert_dynamic(b, vtn_ssa_value(b, w[3])->def,
+ vtn_ssa_value(b, w[4])->def,
+ vtn_ssa_value(b, w[5])->def);
+ break;
+
+ case SpvOpVectorShuffle:
+ val->ssa->def = vtn_vector_shuffle(b, glsl_get_vector_elements(type),
+ vtn_ssa_value(b, w[3])->def,
+ vtn_ssa_value(b, w[4])->def,
+ w + 5);
+ break;
+
+ case SpvOpCompositeConstruct: {
+ unsigned elems = count - 3;
+ if (glsl_type_is_vector_or_scalar(type)) {
+ nir_ssa_def *srcs[4];
+ for (unsigned i = 0; i < elems; i++)
+ srcs[i] = vtn_ssa_value(b, w[3 + i])->def;
+ val->ssa->def =
+ vtn_vector_construct(b, glsl_get_vector_elements(type),
+ elems, srcs);
+ } else {
+ val->ssa->elems = ralloc_array(b, struct vtn_ssa_value *, elems);
+ for (unsigned i = 0; i < elems; i++)
+ val->ssa->elems[i] = vtn_ssa_value(b, w[3 + i]);
+ }
+ break;
+ }
+ case SpvOpCompositeExtract:
+ val->ssa = vtn_composite_extract(b, vtn_ssa_value(b, w[3]),
+ w + 4, count - 4);
+ break;
+
+ case SpvOpCompositeInsert:
+ val->ssa = vtn_composite_insert(b, vtn_ssa_value(b, w[4]),
+ vtn_ssa_value(b, w[3]),
+ w + 5, count - 5);
+ break;
+
+ case SpvOpCopyObject:
+ val->ssa = vtn_composite_copy(b, vtn_ssa_value(b, w[3]));
+ break;
+
+ default:
+ unreachable("unknown composite operation");
+ }
+}
+
+static void
+vtn_handle_barrier(struct vtn_builder *b, SpvOp opcode,
+ const uint32_t *w, unsigned count)
+{
+ nir_intrinsic_op intrinsic_op;
+ switch (opcode) {
+ case SpvOpEmitVertex:
+ case SpvOpEmitStreamVertex:
+ intrinsic_op = nir_intrinsic_emit_vertex;
+ break;
+ case SpvOpEndPrimitive:
+ case SpvOpEndStreamPrimitive:
+ intrinsic_op = nir_intrinsic_end_primitive;
+ break;
+ case SpvOpMemoryBarrier:
+ intrinsic_op = nir_intrinsic_memory_barrier;
+ break;
+ case SpvOpControlBarrier:
+ intrinsic_op = nir_intrinsic_barrier;
+ break;
+ default:
+ unreachable("unknown barrier instruction");
+ }
+
+ nir_intrinsic_instr *intrin =
+ nir_intrinsic_instr_create(b->shader, intrinsic_op);
+
+ if (opcode == SpvOpEmitStreamVertex || opcode == SpvOpEndStreamPrimitive)
+ intrin->const_index[0] = w[1];
+
+ nir_builder_instr_insert(&b->nb, &intrin->instr);
+}
+
+static unsigned
+gl_primitive_from_spv_execution_mode(SpvExecutionMode mode)
+{
+ switch (mode) {
+ case SpvExecutionModeInputPoints:
+ case SpvExecutionModeOutputPoints:
+ return 0; /* GL_POINTS */
+ case SpvExecutionModeInputLines:
+ return 1; /* GL_LINES */
+ case SpvExecutionModeInputLinesAdjacency:
+ return 0x000A; /* GL_LINE_STRIP_ADJACENCY_ARB */
+ case SpvExecutionModeTriangles:
+ return 4; /* GL_TRIANGLES */
+ case SpvExecutionModeInputTrianglesAdjacency:
+ return 0x000C; /* GL_TRIANGLES_ADJACENCY_ARB */
+ case SpvExecutionModeQuads:
+ return 7; /* GL_QUADS */
+ case SpvExecutionModeIsolines:
+ return 0x8E7A; /* GL_ISOLINES */
+ case SpvExecutionModeOutputLineStrip:
+ return 3; /* GL_LINE_STRIP */
+ case SpvExecutionModeOutputTriangleStrip:
+ return 5; /* GL_TRIANGLE_STRIP */
+ default:
+ assert(!"Invalid primitive type");
+ return 4;
+ }
+}
+
+static unsigned
+vertices_in_from_spv_execution_mode(SpvExecutionMode mode)
+{
+ switch (mode) {
+ case SpvExecutionModeInputPoints:
+ return 1;
+ case SpvExecutionModeInputLines:
+ return 2;
+ case SpvExecutionModeInputLinesAdjacency:
+ return 4;
+ case SpvExecutionModeTriangles:
+ return 3;
+ case SpvExecutionModeInputTrianglesAdjacency:
+ return 6;
+ default:
+ assert(!"Invalid GS input mode");
+ return 0;
+ }
+}
+
+static gl_shader_stage
+stage_for_execution_model(SpvExecutionModel model)
+{
+ switch (model) {
+ case SpvExecutionModelVertex:
+ return MESA_SHADER_VERTEX;
+ case SpvExecutionModelTessellationControl:
+ return MESA_SHADER_TESS_CTRL;
+ case SpvExecutionModelTessellationEvaluation:
+ return MESA_SHADER_TESS_EVAL;
+ case SpvExecutionModelGeometry:
+ return MESA_SHADER_GEOMETRY;
+ case SpvExecutionModelFragment:
+ return MESA_SHADER_FRAGMENT;
+ case SpvExecutionModelGLCompute:
+ return MESA_SHADER_COMPUTE;
+ default:
+ unreachable("Unsupported execution model");
+ }
+}
+
+static bool
+vtn_handle_preamble_instruction(struct vtn_builder *b, SpvOp opcode,
+ const uint32_t *w, unsigned count)
+{
+ switch (opcode) {
+ case SpvOpSource:
+ case SpvOpSourceExtension:
+ case SpvOpSourceContinued:
+ case SpvOpExtension:
+ /* Unhandled, but these are for debug so that's ok. */
+ break;
+
+ case SpvOpCapability:
+ switch ((SpvCapability)w[1]) {
+ case SpvCapabilityMatrix:
+ case SpvCapabilityShader:
+ case SpvCapabilityGeometry:
+ break;
+ default:
+ assert(!"Unsupported capability");
+ }
+ break;
+
+ case SpvOpExtInstImport:
+ vtn_handle_extension(b, opcode, w, count);
+ break;
+
+ case SpvOpMemoryModel:
+ assert(w[1] == SpvAddressingModelLogical);
+ assert(w[2] == SpvMemoryModelGLSL450);
+ break;
+
+ case SpvOpEntryPoint: {
+ struct vtn_value *entry_point = &b->values[w[2]];
+ /* Let this be a name label regardless */
+ unsigned name_words;
+ entry_point->name = vtn_string_literal(b, &w[3], count - 3, &name_words);
+
+ if (strcmp(entry_point->name, b->entry_point_name) != 0 ||
+ stage_for_execution_model(w[1]) != b->entry_point_stage)
+ break;
+
+ assert(b->entry_point == NULL);
+ b->entry_point = entry_point;
+ break;
+ }
+
+ case SpvOpString:
+ vtn_push_value(b, w[1], vtn_value_type_string)->str =
+ vtn_string_literal(b, &w[2], count - 2, NULL);
+ break;
+
+ case SpvOpName:
+ b->values[w[1]].name = vtn_string_literal(b, &w[2], count - 2, NULL);
+ break;
+
+ case SpvOpMemberName:
+ /* TODO */
+ break;
+
+ case SpvOpExecutionMode:
+ case SpvOpDecorationGroup:
+ case SpvOpDecorate:
+ case SpvOpMemberDecorate:
+ case SpvOpGroupDecorate:
+ case SpvOpGroupMemberDecorate:
+ vtn_handle_decoration(b, opcode, w, count);
+ break;
+
+ default:
+ return false; /* End of preamble */
+ }
+
+ return true;
+}
+
+static void
+vtn_handle_execution_mode(struct vtn_builder *b, struct vtn_value *entry_point,
+ const struct vtn_decoration *mode, void *data)
+{
+ assert(b->entry_point == entry_point);
+
+ switch(mode->exec_mode) {
+ case SpvExecutionModeOriginUpperLeft:
+ case SpvExecutionModeOriginLowerLeft:
+ b->origin_upper_left =
+ (mode->exec_mode == SpvExecutionModeOriginUpperLeft);
+ break;
+
+ case SpvExecutionModeEarlyFragmentTests:
+ assert(b->shader->stage == MESA_SHADER_FRAGMENT);
+ b->shader->info.fs.early_fragment_tests = true;
+ break;
+
+ case SpvExecutionModeInvocations:
+ assert(b->shader->stage == MESA_SHADER_GEOMETRY);
+ b->shader->info.gs.invocations = MAX2(1, mode->literals[0]);
+ break;
+
+ case SpvExecutionModeDepthReplacing:
+ assert(b->shader->stage == MESA_SHADER_FRAGMENT);
+ b->shader->info.fs.depth_layout = FRAG_DEPTH_LAYOUT_ANY;
+ break;
+ case SpvExecutionModeDepthGreater:
+ assert(b->shader->stage == MESA_SHADER_FRAGMENT);
+ b->shader->info.fs.depth_layout = FRAG_DEPTH_LAYOUT_GREATER;
+ break;
+ case SpvExecutionModeDepthLess:
+ assert(b->shader->stage == MESA_SHADER_FRAGMENT);
+ b->shader->info.fs.depth_layout = FRAG_DEPTH_LAYOUT_LESS;
+ break;
+ case SpvExecutionModeDepthUnchanged:
+ assert(b->shader->stage == MESA_SHADER_FRAGMENT);
+ b->shader->info.fs.depth_layout = FRAG_DEPTH_LAYOUT_UNCHANGED;
+ break;
+
+ case SpvExecutionModeLocalSize:
+ assert(b->shader->stage == MESA_SHADER_COMPUTE);
+ b->shader->info.cs.local_size[0] = mode->literals[0];
+ b->shader->info.cs.local_size[1] = mode->literals[1];
+ b->shader->info.cs.local_size[2] = mode->literals[2];
+ break;
+ case SpvExecutionModeLocalSizeHint:
+ break; /* Nothing do do with this */
+
+ case SpvExecutionModeOutputVertices:
+ assert(b->shader->stage == MESA_SHADER_GEOMETRY);
+ b->shader->info.gs.vertices_out = mode->literals[0];
+ break;
+
+ case SpvExecutionModeInputPoints:
+ case SpvExecutionModeInputLines:
+ case SpvExecutionModeInputLinesAdjacency:
+ case SpvExecutionModeTriangles:
+ case SpvExecutionModeInputTrianglesAdjacency:
+ case SpvExecutionModeQuads:
+ case SpvExecutionModeIsolines:
+ if (b->shader->stage == MESA_SHADER_GEOMETRY) {
+ b->shader->info.gs.vertices_in =
+ vertices_in_from_spv_execution_mode(mode->exec_mode);
+ } else {
+ assert(!"Tesselation shaders not yet supported");
+ }
+ break;
+
+ case SpvExecutionModeOutputPoints:
+ case SpvExecutionModeOutputLineStrip:
+ case SpvExecutionModeOutputTriangleStrip:
+ assert(b->shader->stage == MESA_SHADER_GEOMETRY);
+ b->shader->info.gs.output_primitive =
+ gl_primitive_from_spv_execution_mode(mode->exec_mode);
+ break;
+
+ case SpvExecutionModeSpacingEqual:
+ case SpvExecutionModeSpacingFractionalEven:
+ case SpvExecutionModeSpacingFractionalOdd:
+ case SpvExecutionModeVertexOrderCw:
+ case SpvExecutionModeVertexOrderCcw:
+ case SpvExecutionModePointMode:
+ assert(!"TODO: Add tessellation metadata");
+ break;
+
+ case SpvExecutionModePixelCenterInteger:
+ case SpvExecutionModeXfb:
+ assert(!"Unhandled execution mode");
+ break;
+
+ case SpvExecutionModeVecTypeHint:
+ case SpvExecutionModeContractionOff:
+ break; /* OpenCL */
+ }
+}
+
+static bool
+vtn_handle_variable_or_type_instruction(struct vtn_builder *b, SpvOp opcode,
+ const uint32_t *w, unsigned count)
+{
+ switch (opcode) {
+ case SpvOpSource:
+ case SpvOpSourceContinued:
+ case SpvOpSourceExtension:
+ case SpvOpExtension:
+ case SpvOpCapability:
+ case SpvOpExtInstImport:
+ case SpvOpMemoryModel:
+ case SpvOpEntryPoint:
+ case SpvOpExecutionMode:
+ case SpvOpString:
+ case SpvOpName:
+ case SpvOpMemberName:
+ case SpvOpDecorationGroup:
+ case SpvOpDecorate:
+ case SpvOpMemberDecorate:
+ case SpvOpGroupDecorate:
+ case SpvOpGroupMemberDecorate:
+ assert(!"Invalid opcode types and variables section");
+ break;
+
+ case SpvOpTypeVoid:
+ case SpvOpTypeBool:
+ case SpvOpTypeInt:
+ case SpvOpTypeFloat:
+ case SpvOpTypeVector:
+ case SpvOpTypeMatrix:
+ case SpvOpTypeImage:
+ case SpvOpTypeSampler:
+ case SpvOpTypeSampledImage:
+ case SpvOpTypeArray:
+ case SpvOpTypeRuntimeArray:
+ case SpvOpTypeStruct:
+ case SpvOpTypeOpaque:
+ case SpvOpTypePointer:
+ case SpvOpTypeFunction:
+ case SpvOpTypeEvent:
+ case SpvOpTypeDeviceEvent:
+ case SpvOpTypeReserveId:
+ case SpvOpTypeQueue:
+ case SpvOpTypePipe:
+ vtn_handle_type(b, opcode, w, count);
+ break;
+
+ case SpvOpConstantTrue:
+ case SpvOpConstantFalse:
+ case SpvOpConstant:
+ case SpvOpConstantComposite:
+ case SpvOpConstantSampler:
+ case SpvOpConstantNull:
+ case SpvOpSpecConstantTrue:
+ case SpvOpSpecConstantFalse:
+ case SpvOpSpecConstant:
+ case SpvOpSpecConstantComposite:
+ case SpvOpSpecConstantOp:
+ vtn_handle_constant(b, opcode, w, count);
+ break;
+
+ case SpvOpVariable:
+ vtn_handle_variables(b, opcode, w, count);
+ break;
+
+ default:
+ return false; /* End of preamble */
+ }
+
+ return true;
+}
+
+static bool
+vtn_handle_body_instruction(struct vtn_builder *b, SpvOp opcode,
+ const uint32_t *w, unsigned count)
+{
+ switch (opcode) {
+ case SpvOpLabel:
+ break;
+
+ case SpvOpLoopMerge:
+ case SpvOpSelectionMerge:
+ /* This is handled by cfg pre-pass and walk_blocks */
+ break;
+
+ case SpvOpUndef: {
+ struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_undef);
+ val->type = vtn_value(b, w[1], vtn_value_type_type)->type;
+ break;
+ }
+
+ case SpvOpExtInst:
+ vtn_handle_extension(b, opcode, w, count);
+ break;
+
+ case SpvOpVariable:
+ case SpvOpLoad:
+ case SpvOpStore:
+ case SpvOpCopyMemory:
+ case SpvOpCopyMemorySized:
+ case SpvOpAccessChain:
+ case SpvOpInBoundsAccessChain:
+ case SpvOpArrayLength:
+ vtn_handle_variables(b, opcode, w, count);
+ break;
+
+ case SpvOpFunctionCall:
+ vtn_handle_function_call(b, opcode, w, count);
+ break;
+
+ case SpvOpSampledImage:
+ case SpvOpImageSampleImplicitLod:
+ case SpvOpImageSampleExplicitLod:
+ case SpvOpImageSampleDrefImplicitLod:
+ case SpvOpImageSampleDrefExplicitLod:
+ case SpvOpImageSampleProjImplicitLod:
+ case SpvOpImageSampleProjExplicitLod:
+ case SpvOpImageSampleProjDrefImplicitLod:
+ case SpvOpImageSampleProjDrefExplicitLod:
+ case SpvOpImageFetch:
+ case SpvOpImageGather:
+ case SpvOpImageDrefGather:
+ case SpvOpImageQuerySizeLod:
+ case SpvOpImageQueryLod:
+ case SpvOpImageQueryLevels:
+ case SpvOpImageQuerySamples:
+ vtn_handle_texture(b, opcode, w, count);
+ break;
+
+ case SpvOpImageRead:
+ case SpvOpImageWrite:
+ case SpvOpImageTexelPointer:
+ vtn_handle_image(b, opcode, w, count);
+ break;
+
+ case SpvOpImageQuerySize: {
+ nir_deref_var *image = vtn_value(b, w[3], vtn_value_type_deref)->deref;
+ const struct glsl_type *image_type = nir_deref_tail(&image->deref)->type;
+ if (glsl_type_is_image(image_type)) {
+ vtn_handle_image(b, opcode, w, count);
+ } else {
+ vtn_handle_texture(b, opcode, w, count);
+ }
+ break;
+ }
+
+ case SpvOpAtomicExchange:
+ case SpvOpAtomicCompareExchange:
+ case SpvOpAtomicCompareExchangeWeak:
+ case SpvOpAtomicIIncrement:
+ case SpvOpAtomicIDecrement:
+ case SpvOpAtomicIAdd:
+ case SpvOpAtomicISub:
+ case SpvOpAtomicSMin:
+ case SpvOpAtomicUMin:
+ case SpvOpAtomicSMax:
+ case SpvOpAtomicUMax:
+ case SpvOpAtomicAnd:
+ case SpvOpAtomicOr:
+ case SpvOpAtomicXor: {
+ struct vtn_value *pointer = vtn_untyped_value(b, w[3]);
+ if (pointer->value_type == vtn_value_type_image_pointer) {
+ vtn_handle_image(b, opcode, w, count);
+ } else {
+ assert(pointer->value_type == vtn_value_type_deref);
+ vtn_handle_ssbo_or_shared_atomic(b, opcode, w, count);
+ }
+ break;
+ }
+
+ case SpvOpSNegate:
+ case SpvOpFNegate:
+ case SpvOpNot:
+ case SpvOpAny:
+ case SpvOpAll:
+ case SpvOpConvertFToU:
+ case SpvOpConvertFToS:
+ case SpvOpConvertSToF:
+ case SpvOpConvertUToF:
+ case SpvOpUConvert:
+ case SpvOpSConvert:
+ case SpvOpFConvert:
+ case SpvOpQuantizeToF16:
+ case SpvOpConvertPtrToU:
+ case SpvOpConvertUToPtr:
+ case SpvOpPtrCastToGeneric:
+ case SpvOpGenericCastToPtr:
+ case SpvOpBitcast:
+ case SpvOpIsNan:
+ case SpvOpIsInf:
+ case SpvOpIsFinite:
+ case SpvOpIsNormal:
+ case SpvOpSignBitSet:
+ case SpvOpLessOrGreater:
+ case SpvOpOrdered:
+ case SpvOpUnordered:
+ case SpvOpIAdd:
+ case SpvOpFAdd:
+ case SpvOpISub:
+ case SpvOpFSub:
+ case SpvOpIMul:
+ case SpvOpFMul:
+ case SpvOpUDiv:
+ case SpvOpSDiv:
+ case SpvOpFDiv:
+ case SpvOpUMod:
+ case SpvOpSRem:
+ case SpvOpSMod:
+ case SpvOpFRem:
+ case SpvOpFMod:
+ case SpvOpVectorTimesScalar:
+ case SpvOpDot:
+ case SpvOpIAddCarry:
+ case SpvOpISubBorrow:
+ case SpvOpUMulExtended:
+ case SpvOpSMulExtended:
+ case SpvOpShiftRightLogical:
+ case SpvOpShiftRightArithmetic:
+ case SpvOpShiftLeftLogical:
+ case SpvOpLogicalEqual:
+ case SpvOpLogicalNotEqual:
+ case SpvOpLogicalOr:
+ case SpvOpLogicalAnd:
+ case SpvOpLogicalNot:
+ case SpvOpBitwiseOr:
+ case SpvOpBitwiseXor:
+ case SpvOpBitwiseAnd:
+ case SpvOpSelect:
+ case SpvOpIEqual:
+ case SpvOpFOrdEqual:
+ case SpvOpFUnordEqual:
+ case SpvOpINotEqual:
+ case SpvOpFOrdNotEqual:
+ case SpvOpFUnordNotEqual:
+ case SpvOpULessThan:
+ case SpvOpSLessThan:
+ case SpvOpFOrdLessThan:
+ case SpvOpFUnordLessThan:
+ case SpvOpUGreaterThan:
+ case SpvOpSGreaterThan:
+ case SpvOpFOrdGreaterThan:
+ case SpvOpFUnordGreaterThan:
+ case SpvOpULessThanEqual:
+ case SpvOpSLessThanEqual:
+ case SpvOpFOrdLessThanEqual:
+ case SpvOpFUnordLessThanEqual:
+ case SpvOpUGreaterThanEqual:
+ case SpvOpSGreaterThanEqual:
+ case SpvOpFOrdGreaterThanEqual:
+ case SpvOpFUnordGreaterThanEqual:
+ case SpvOpDPdx:
+ case SpvOpDPdy:
+ case SpvOpFwidth:
+ case SpvOpDPdxFine:
+ case SpvOpDPdyFine:
+ case SpvOpFwidthFine:
+ case SpvOpDPdxCoarse:
+ case SpvOpDPdyCoarse:
+ case SpvOpFwidthCoarse:
+ case SpvOpBitFieldInsert:
+ case SpvOpBitFieldSExtract:
+ case SpvOpBitFieldUExtract:
+ case SpvOpBitReverse:
+ case SpvOpBitCount:
+ case SpvOpTranspose:
+ case SpvOpOuterProduct:
+ case SpvOpMatrixTimesScalar:
+ case SpvOpVectorTimesMatrix:
+ case SpvOpMatrixTimesVector:
+ case SpvOpMatrixTimesMatrix:
+ vtn_handle_alu(b, opcode, w, count);
+ break;
+
+ case SpvOpVectorExtractDynamic:
+ case SpvOpVectorInsertDynamic:
+ case SpvOpVectorShuffle:
+ case SpvOpCompositeConstruct:
+ case SpvOpCompositeExtract:
+ case SpvOpCompositeInsert:
+ case SpvOpCopyObject:
+ vtn_handle_composite(b, opcode, w, count);
+ break;
+
+ case SpvOpEmitVertex:
+ case SpvOpEndPrimitive:
+ case SpvOpEmitStreamVertex:
+ case SpvOpEndStreamPrimitive:
+ case SpvOpControlBarrier:
+ case SpvOpMemoryBarrier:
+ vtn_handle_barrier(b, opcode, w, count);
+ break;
+
+ default:
+ unreachable("Unhandled opcode");
+ }
+
+ return true;
+}
+
+nir_function *
+spirv_to_nir(const uint32_t *words, size_t word_count,
+ struct nir_spirv_specialization *spec, unsigned num_spec,
+ gl_shader_stage stage, const char *entry_point_name,
+ const nir_shader_compiler_options *options)
+{
+ const uint32_t *word_end = words + word_count;
+
+ /* Handle the SPIR-V header (first 4 dwords) */
+ assert(word_count > 5);
+
+ assert(words[0] == SpvMagicNumber);
+ assert(words[1] >= 0x10000);
+ /* words[2] == generator magic */
+ unsigned value_id_bound = words[3];
+ assert(words[4] == 0);
+
+ words+= 5;
+
+ /* Initialize the stn_builder object */
+ struct vtn_builder *b = rzalloc(NULL, struct vtn_builder);
+ b->value_id_bound = value_id_bound;
+ b->values = rzalloc_array(b, struct vtn_value, value_id_bound);
+ exec_list_make_empty(&b->functions);
+ b->entry_point_stage = stage;
+ b->entry_point_name = entry_point_name;
+
+ /* Handle all the preamble instructions */
+ words = vtn_foreach_instruction(b, words, word_end,
+ vtn_handle_preamble_instruction);
+
+ if (b->entry_point == NULL) {
+ assert(!"Entry point not found");
+ ralloc_free(b);
+ return NULL;
+ }
+
+ b->shader = nir_shader_create(NULL, stage, options);
+
+ /* Parse execution modes */
+ vtn_foreach_execution_mode(b, b->entry_point,
+ vtn_handle_execution_mode, NULL);
+
+ b->specializations = spec;
+ b->num_specializations = num_spec;
+
+ /* Handle all variable, type, and constant instructions */
+ words = vtn_foreach_instruction(b, words, word_end,
+ vtn_handle_variable_or_type_instruction);
+
+ vtn_build_cfg(b, words, word_end);
+
+ foreach_list_typed(struct vtn_function, func, node, &b->functions) {
+ b->impl = func->impl;
+ b->const_table = _mesa_hash_table_create(b, _mesa_hash_pointer,
+ _mesa_key_pointer_equal);
+
+ vtn_function_emit(b, func, vtn_handle_body_instruction);
+ }
+
+ assert(b->entry_point->value_type == vtn_value_type_function);
+ nir_function *entry_point = b->entry_point->func->impl->function;
+ assert(entry_point);
+
+ ralloc_free(b);
+
+ return entry_point;
+}
--- /dev/null
+/*
+ * Copyright © 2016 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include "vtn_private.h"
+
+/*
+ * Normally, column vectors in SPIR-V correspond to a single NIR SSA
+ * definition. But for matrix multiplies, we want to do one routine for
+ * multiplying a matrix by a matrix and then pretend that vectors are matrices
+ * with one column. So we "wrap" these things, and unwrap the result before we
+ * send it off.
+ */
+
+static struct vtn_ssa_value *
+wrap_matrix(struct vtn_builder *b, struct vtn_ssa_value *val)
+{
+ if (val == NULL)
+ return NULL;
+
+ if (glsl_type_is_matrix(val->type))
+ return val;
+
+ struct vtn_ssa_value *dest = rzalloc(b, struct vtn_ssa_value);
+ dest->type = val->type;
+ dest->elems = ralloc_array(b, struct vtn_ssa_value *, 1);
+ dest->elems[0] = val;
+
+ return dest;
+}
+
+static struct vtn_ssa_value *
+unwrap_matrix(struct vtn_ssa_value *val)
+{
+ if (glsl_type_is_matrix(val->type))
+ return val;
+
+ return val->elems[0];
+}
+
+static struct vtn_ssa_value *
+matrix_multiply(struct vtn_builder *b,
+ struct vtn_ssa_value *_src0, struct vtn_ssa_value *_src1)
+{
+
+ struct vtn_ssa_value *src0 = wrap_matrix(b, _src0);
+ struct vtn_ssa_value *src1 = wrap_matrix(b, _src1);
+ struct vtn_ssa_value *src0_transpose = wrap_matrix(b, _src0->transposed);
+ struct vtn_ssa_value *src1_transpose = wrap_matrix(b, _src1->transposed);
+
+ unsigned src0_rows = glsl_get_vector_elements(src0->type);
+ unsigned src0_columns = glsl_get_matrix_columns(src0->type);
+ unsigned src1_columns = glsl_get_matrix_columns(src1->type);
+
+ const struct glsl_type *dest_type;
+ if (src1_columns > 1) {
+ dest_type = glsl_matrix_type(glsl_get_base_type(src0->type),
+ src0_rows, src1_columns);
+ } else {
+ dest_type = glsl_vector_type(glsl_get_base_type(src0->type), src0_rows);
+ }
+ struct vtn_ssa_value *dest = vtn_create_ssa_value(b, dest_type);
+
+ dest = wrap_matrix(b, dest);
+
+ bool transpose_result = false;
+ if (src0_transpose && src1_transpose) {
+ /* transpose(A) * transpose(B) = transpose(B * A) */
+ src1 = src0_transpose;
+ src0 = src1_transpose;
+ src0_transpose = NULL;
+ src1_transpose = NULL;
+ transpose_result = true;
+ }
+
+ if (src0_transpose && !src1_transpose &&
+ glsl_get_base_type(src0->type) == GLSL_TYPE_FLOAT) {
+ /* We already have the rows of src0 and the columns of src1 available,
+ * so we can just take the dot product of each row with each column to
+ * get the result.
+ */
+
+ for (unsigned i = 0; i < src1_columns; i++) {
+ nir_ssa_def *vec_src[4];
+ for (unsigned j = 0; j < src0_rows; j++) {
+ vec_src[j] = nir_fdot(&b->nb, src0_transpose->elems[j]->def,
+ src1->elems[i]->def);
+ }
+ dest->elems[i]->def = nir_vec(&b->nb, vec_src, src0_rows);
+ }
+ } else {
+ /* We don't handle the case where src1 is transposed but not src0, since
+ * the general case only uses individual components of src1 so the
+ * optimizer should chew through the transpose we emitted for src1.
+ */
+
+ for (unsigned i = 0; i < src1_columns; i++) {
+ /* dest[i] = sum(src0[j] * src1[i][j] for all j) */
+ dest->elems[i]->def =
+ nir_fmul(&b->nb, src0->elems[0]->def,
+ nir_channel(&b->nb, src1->elems[i]->def, 0));
+ for (unsigned j = 1; j < src0_columns; j++) {
+ dest->elems[i]->def =
+ nir_fadd(&b->nb, dest->elems[i]->def,
+ nir_fmul(&b->nb, src0->elems[j]->def,
+ nir_channel(&b->nb, src1->elems[i]->def, j)));
+ }
+ }
+ }
+
+ dest = unwrap_matrix(dest);
+
+ if (transpose_result)
+ dest = vtn_ssa_transpose(b, dest);
+
+ return dest;
+}
+
+static struct vtn_ssa_value *
+mat_times_scalar(struct vtn_builder *b,
+ struct vtn_ssa_value *mat,
+ nir_ssa_def *scalar)
+{
+ struct vtn_ssa_value *dest = vtn_create_ssa_value(b, mat->type);
+ for (unsigned i = 0; i < glsl_get_matrix_columns(mat->type); i++) {
+ if (glsl_get_base_type(mat->type) == GLSL_TYPE_FLOAT)
+ dest->elems[i]->def = nir_fmul(&b->nb, mat->elems[i]->def, scalar);
+ else
+ dest->elems[i]->def = nir_imul(&b->nb, mat->elems[i]->def, scalar);
+ }
+
+ return dest;
+}
+
+static void
+vtn_handle_matrix_alu(struct vtn_builder *b, SpvOp opcode,
+ struct vtn_value *dest,
+ struct vtn_ssa_value *src0, struct vtn_ssa_value *src1)
+{
+ switch (opcode) {
+ case SpvOpFNegate: {
+ dest->ssa = vtn_create_ssa_value(b, src0->type);
+ unsigned cols = glsl_get_matrix_columns(src0->type);
+ for (unsigned i = 0; i < cols; i++)
+ dest->ssa->elems[i]->def = nir_fneg(&b->nb, src0->elems[i]->def);
+ break;
+ }
+
+ case SpvOpFAdd: {
+ dest->ssa = vtn_create_ssa_value(b, src0->type);
+ unsigned cols = glsl_get_matrix_columns(src0->type);
+ for (unsigned i = 0; i < cols; i++)
+ dest->ssa->elems[i]->def =
+ nir_fadd(&b->nb, src0->elems[i]->def, src1->elems[i]->def);
+ break;
+ }
+
+ case SpvOpFSub: {
+ dest->ssa = vtn_create_ssa_value(b, src0->type);
+ unsigned cols = glsl_get_matrix_columns(src0->type);
+ for (unsigned i = 0; i < cols; i++)
+ dest->ssa->elems[i]->def =
+ nir_fsub(&b->nb, src0->elems[i]->def, src1->elems[i]->def);
+ break;
+ }
+
+ case SpvOpTranspose:
+ dest->ssa = vtn_ssa_transpose(b, src0);
+ break;
+
+ case SpvOpMatrixTimesScalar:
+ if (src0->transposed) {
+ dest->ssa = vtn_ssa_transpose(b, mat_times_scalar(b, src0->transposed,
+ src1->def));
+ } else {
+ dest->ssa = mat_times_scalar(b, src0, src1->def);
+ }
+ break;
+
+ case SpvOpVectorTimesMatrix:
+ case SpvOpMatrixTimesVector:
+ case SpvOpMatrixTimesMatrix:
+ if (opcode == SpvOpVectorTimesMatrix) {
+ dest->ssa = matrix_multiply(b, vtn_ssa_transpose(b, src1), src0);
+ } else {
+ dest->ssa = matrix_multiply(b, src0, src1);
+ }
+ break;
+
+ default: unreachable("unknown matrix opcode");
+ }
+}
+
+nir_op
+vtn_nir_alu_op_for_spirv_opcode(SpvOp opcode, bool *swap)
+{
+ /* Indicates that the first two arguments should be swapped. This is
+ * used for implementing greater-than and less-than-or-equal.
+ */
+ *swap = false;
+
+ switch (opcode) {
+ case SpvOpSNegate: return nir_op_ineg;
+ case SpvOpFNegate: return nir_op_fneg;
+ case SpvOpNot: return nir_op_inot;
+ case SpvOpIAdd: return nir_op_iadd;
+ case SpvOpFAdd: return nir_op_fadd;
+ case SpvOpISub: return nir_op_isub;
+ case SpvOpFSub: return nir_op_fsub;
+ case SpvOpIMul: return nir_op_imul;
+ case SpvOpFMul: return nir_op_fmul;
+ case SpvOpUDiv: return nir_op_udiv;
+ case SpvOpSDiv: return nir_op_idiv;
+ case SpvOpFDiv: return nir_op_fdiv;
+ case SpvOpUMod: return nir_op_umod;
+ case SpvOpSMod: return nir_op_imod;
+ case SpvOpFMod: return nir_op_fmod;
+ case SpvOpSRem: return nir_op_irem;
+ case SpvOpFRem: return nir_op_frem;
+
+ case SpvOpShiftRightLogical: return nir_op_ushr;
+ case SpvOpShiftRightArithmetic: return nir_op_ishr;
+ case SpvOpShiftLeftLogical: return nir_op_ishl;
+ case SpvOpLogicalOr: return nir_op_ior;
+ case SpvOpLogicalEqual: return nir_op_ieq;
+ case SpvOpLogicalNotEqual: return nir_op_ine;
+ case SpvOpLogicalAnd: return nir_op_iand;
+ case SpvOpLogicalNot: return nir_op_inot;
+ case SpvOpBitwiseOr: return nir_op_ior;
+ case SpvOpBitwiseXor: return nir_op_ixor;
+ case SpvOpBitwiseAnd: return nir_op_iand;
+ case SpvOpSelect: return nir_op_bcsel;
+ case SpvOpIEqual: return nir_op_ieq;
+
+ case SpvOpBitFieldInsert: return nir_op_bitfield_insert;
+ case SpvOpBitFieldSExtract: return nir_op_ibitfield_extract;
+ case SpvOpBitFieldUExtract: return nir_op_ubitfield_extract;
+ case SpvOpBitReverse: return nir_op_bitfield_reverse;
+ case SpvOpBitCount: return nir_op_bit_count;
+
+ /* Comparisons: (TODO: How do we want to handled ordered/unordered?) */
+ case SpvOpFOrdEqual: return nir_op_feq;
+ case SpvOpFUnordEqual: return nir_op_feq;
+ case SpvOpINotEqual: return nir_op_ine;
+ case SpvOpFOrdNotEqual: return nir_op_fne;
+ case SpvOpFUnordNotEqual: return nir_op_fne;
+ case SpvOpULessThan: return nir_op_ult;
+ case SpvOpSLessThan: return nir_op_ilt;
+ case SpvOpFOrdLessThan: return nir_op_flt;
+ case SpvOpFUnordLessThan: return nir_op_flt;
+ case SpvOpUGreaterThan: *swap = true; return nir_op_ult;
+ case SpvOpSGreaterThan: *swap = true; return nir_op_ilt;
+ case SpvOpFOrdGreaterThan: *swap = true; return nir_op_flt;
+ case SpvOpFUnordGreaterThan: *swap = true; return nir_op_flt;
+ case SpvOpULessThanEqual: *swap = true; return nir_op_uge;
+ case SpvOpSLessThanEqual: *swap = true; return nir_op_ige;
+ case SpvOpFOrdLessThanEqual: *swap = true; return nir_op_fge;
+ case SpvOpFUnordLessThanEqual: *swap = true; return nir_op_fge;
+ case SpvOpUGreaterThanEqual: return nir_op_uge;
+ case SpvOpSGreaterThanEqual: return nir_op_ige;
+ case SpvOpFOrdGreaterThanEqual: return nir_op_fge;
+ case SpvOpFUnordGreaterThanEqual: return nir_op_fge;
+
+ /* Conversions: */
+ case SpvOpConvertFToU: return nir_op_f2u;
+ case SpvOpConvertFToS: return nir_op_f2i;
+ case SpvOpConvertSToF: return nir_op_i2f;
+ case SpvOpConvertUToF: return nir_op_u2f;
+ case SpvOpBitcast: return nir_op_imov;
+ case SpvOpUConvert:
+ case SpvOpQuantizeToF16: return nir_op_fquantize2f16;
+ /* TODO: NIR is 32-bit only; these are no-ops. */
+ case SpvOpSConvert: return nir_op_imov;
+ case SpvOpFConvert: return nir_op_fmov;
+
+ /* Derivatives: */
+ case SpvOpDPdx: return nir_op_fddx;
+ case SpvOpDPdy: return nir_op_fddy;
+ case SpvOpDPdxFine: return nir_op_fddx_fine;
+ case SpvOpDPdyFine: return nir_op_fddy_fine;
+ case SpvOpDPdxCoarse: return nir_op_fddx_coarse;
+ case SpvOpDPdyCoarse: return nir_op_fddy_coarse;
+
+ default:
+ unreachable("No NIR equivalent");
+ }
+}
+
+void
+vtn_handle_alu(struct vtn_builder *b, SpvOp opcode,
+ const uint32_t *w, unsigned count)
+{
+ struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_ssa);
+ const struct glsl_type *type =
+ vtn_value(b, w[1], vtn_value_type_type)->type->type;
+
+ /* Collect the various SSA sources */
+ const unsigned num_inputs = count - 3;
+ struct vtn_ssa_value *vtn_src[4] = { NULL, };
+ for (unsigned i = 0; i < num_inputs; i++)
+ vtn_src[i] = vtn_ssa_value(b, w[i + 3]);
+
+ if (glsl_type_is_matrix(vtn_src[0]->type) ||
+ (num_inputs >= 2 && glsl_type_is_matrix(vtn_src[1]->type))) {
+ vtn_handle_matrix_alu(b, opcode, val, vtn_src[0], vtn_src[1]);
+ return;
+ }
+
+ val->ssa = vtn_create_ssa_value(b, type);
+ nir_ssa_def *src[4] = { NULL, };
+ for (unsigned i = 0; i < num_inputs; i++) {
+ assert(glsl_type_is_vector_or_scalar(vtn_src[i]->type));
+ src[i] = vtn_src[i]->def;
+ }
+
+ switch (opcode) {
+ case SpvOpAny:
+ if (src[0]->num_components == 1) {
+ val->ssa->def = nir_imov(&b->nb, src[0]);
+ } else {
+ nir_op op;
+ switch (src[0]->num_components) {
+ case 2: op = nir_op_bany_inequal2; break;
+ case 3: op = nir_op_bany_inequal3; break;
+ case 4: op = nir_op_bany_inequal4; break;
+ }
+ val->ssa->def = nir_build_alu(&b->nb, op, src[0],
+ nir_imm_int(&b->nb, NIR_FALSE),
+ NULL, NULL);
+ }
+ return;
+
+ case SpvOpAll:
+ if (src[0]->num_components == 1) {
+ val->ssa->def = nir_imov(&b->nb, src[0]);
+ } else {
+ nir_op op;
+ switch (src[0]->num_components) {
+ case 2: op = nir_op_ball_iequal2; break;
+ case 3: op = nir_op_ball_iequal3; break;
+ case 4: op = nir_op_ball_iequal4; break;
+ }
+ val->ssa->def = nir_build_alu(&b->nb, op, src[0],
+ nir_imm_int(&b->nb, NIR_TRUE),
+ NULL, NULL);
+ }
+ return;
+
+ case SpvOpOuterProduct: {
+ for (unsigned i = 0; i < src[1]->num_components; i++) {
+ val->ssa->elems[i]->def =
+ nir_fmul(&b->nb, src[0], nir_channel(&b->nb, src[1], i));
+ }
+ return;
+ }
+
+ case SpvOpDot:
+ val->ssa->def = nir_fdot(&b->nb, src[0], src[1]);
+ return;
+
+ case SpvOpIAddCarry:
+ assert(glsl_type_is_struct(val->ssa->type));
+ val->ssa->elems[0]->def = nir_iadd(&b->nb, src[0], src[1]);
+ val->ssa->elems[1]->def = nir_uadd_carry(&b->nb, src[0], src[1]);
+ return;
+
+ case SpvOpISubBorrow:
+ assert(glsl_type_is_struct(val->ssa->type));
+ val->ssa->elems[0]->def = nir_isub(&b->nb, src[0], src[1]);
+ val->ssa->elems[1]->def = nir_usub_borrow(&b->nb, src[0], src[1]);
+ return;
+
+ case SpvOpUMulExtended:
+ assert(glsl_type_is_struct(val->ssa->type));
+ val->ssa->elems[0]->def = nir_imul(&b->nb, src[0], src[1]);
+ val->ssa->elems[1]->def = nir_umul_high(&b->nb, src[0], src[1]);
+ return;
+
+ case SpvOpSMulExtended:
+ assert(glsl_type_is_struct(val->ssa->type));
+ val->ssa->elems[0]->def = nir_imul(&b->nb, src[0], src[1]);
+ val->ssa->elems[1]->def = nir_imul_high(&b->nb, src[0], src[1]);
+ return;
+
+ case SpvOpFwidth:
+ val->ssa->def = nir_fadd(&b->nb,
+ nir_fabs(&b->nb, nir_fddx(&b->nb, src[0])),
+ nir_fabs(&b->nb, nir_fddx(&b->nb, src[1])));
+ return;
+ case SpvOpFwidthFine:
+ val->ssa->def = nir_fadd(&b->nb,
+ nir_fabs(&b->nb, nir_fddx_fine(&b->nb, src[0])),
+ nir_fabs(&b->nb, nir_fddx_fine(&b->nb, src[1])));
+ return;
+ case SpvOpFwidthCoarse:
+ val->ssa->def = nir_fadd(&b->nb,
+ nir_fabs(&b->nb, nir_fddx_coarse(&b->nb, src[0])),
+ nir_fabs(&b->nb, nir_fddx_coarse(&b->nb, src[1])));
+ return;
+
+ case SpvOpVectorTimesScalar:
+ /* The builder will take care of splatting for us. */
+ val->ssa->def = nir_fmul(&b->nb, src[0], src[1]);
+ return;
+
+ case SpvOpIsNan:
+ val->ssa->def = nir_fne(&b->nb, src[0], src[0]);
+ return;
+
+ case SpvOpIsInf:
+ val->ssa->def = nir_feq(&b->nb, nir_fabs(&b->nb, src[0]),
+ nir_imm_float(&b->nb, INFINITY));
+ return;
+
+ default: {
+ bool swap;
+ nir_op op = vtn_nir_alu_op_for_spirv_opcode(opcode, &swap);
+
+ if (swap) {
+ nir_ssa_def *tmp = src[0];
+ src[0] = src[1];
+ src[1] = tmp;
+ }
+
+ val->ssa->def = nir_build_alu(&b->nb, op, src[0], src[1], src[2], src[3]);
+ return;
+ } /* default */
+ }
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include "vtn_private.h"
+#include "nir/nir_vla.h"
+
+static bool
+vtn_cfg_handle_prepass_instruction(struct vtn_builder *b, SpvOp opcode,
+ const uint32_t *w, unsigned count)
+{
+ switch (opcode) {
+ case SpvOpFunction: {
+ assert(b->func == NULL);
+ b->func = rzalloc(b, struct vtn_function);
+
+ list_inithead(&b->func->body);
+ b->func->control = w[3];
+
+ const struct glsl_type *result_type =
+ vtn_value(b, w[1], vtn_value_type_type)->type->type;
+ struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_function);
+ val->func = b->func;
+
+ const struct glsl_type *func_type =
+ vtn_value(b, w[4], vtn_value_type_type)->type->type;
+
+ assert(glsl_get_function_return_type(func_type) == result_type);
+
+ nir_function *func =
+ nir_function_create(b->shader, ralloc_strdup(b->shader, val->name));
+
+ func->num_params = glsl_get_length(func_type);
+ func->params = ralloc_array(b->shader, nir_parameter, func->num_params);
+ for (unsigned i = 0; i < func->num_params; i++) {
+ const struct glsl_function_param *param =
+ glsl_get_function_param(func_type, i);
+ func->params[i].type = param->type;
+ if (param->in) {
+ if (param->out) {
+ func->params[i].param_type = nir_parameter_inout;
+ } else {
+ func->params[i].param_type = nir_parameter_in;
+ }
+ } else {
+ if (param->out) {
+ func->params[i].param_type = nir_parameter_out;
+ } else {
+ assert(!"Parameter is neither in nor out");
+ }
+ }
+ }
+
+ func->return_type = glsl_get_function_return_type(func_type);
+
+ b->func->impl = nir_function_impl_create(func);
+ if (!glsl_type_is_void(func->return_type)) {
+ b->func->impl->return_var =
+ nir_local_variable_create(b->func->impl, func->return_type, "ret");
+ }
+
+ b->func_param_idx = 0;
+ break;
+ }
+
+ case SpvOpFunctionEnd:
+ b->func->end = w;
+ b->func = NULL;
+ break;
+
+ case SpvOpFunctionParameter: {
+ struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_deref);
+
+ assert(b->func_param_idx < b->func->impl->num_params);
+ unsigned idx = b->func_param_idx++;
+
+ nir_variable *param =
+ nir_local_variable_create(b->func->impl,
+ b->func->impl->function->params[idx].type,
+ val->name);
+
+ b->func->impl->params[idx] = param;
+ val->deref = nir_deref_var_create(b, param);
+ val->deref_type = vtn_value(b, w[1], vtn_value_type_type)->type;
+ break;
+ }
+
+ case SpvOpLabel: {
+ assert(b->block == NULL);
+ b->block = rzalloc(b, struct vtn_block);
+ b->block->node.type = vtn_cf_node_type_block;
+ b->block->label = w;
+ vtn_push_value(b, w[1], vtn_value_type_block)->block = b->block;
+
+ if (b->func->start_block == NULL) {
+ /* This is the first block encountered for this function. In this
+ * case, we set the start block and add it to the list of
+ * implemented functions that we'll walk later.
+ */
+ b->func->start_block = b->block;
+ exec_list_push_tail(&b->functions, &b->func->node);
+ }
+ break;
+ }
+
+ case SpvOpSelectionMerge:
+ case SpvOpLoopMerge:
+ assert(b->block && b->block->merge == NULL);
+ b->block->merge = w;
+ break;
+
+ case SpvOpBranch:
+ case SpvOpBranchConditional:
+ case SpvOpSwitch:
+ case SpvOpKill:
+ case SpvOpReturn:
+ case SpvOpReturnValue:
+ case SpvOpUnreachable:
+ assert(b->block && b->block->branch == NULL);
+ b->block->branch = w;
+ b->block = NULL;
+ break;
+
+ default:
+ /* Continue on as per normal */
+ return true;
+ }
+
+ return true;
+}
+
+static void
+vtn_add_case(struct vtn_builder *b, struct vtn_switch *swtch,
+ struct vtn_block *break_block,
+ uint32_t block_id, uint32_t val, bool is_default)
+{
+ struct vtn_block *case_block =
+ vtn_value(b, block_id, vtn_value_type_block)->block;
+
+ /* Don't create dummy cases that just break */
+ if (case_block == break_block)
+ return;
+
+ if (case_block->switch_case == NULL) {
+ struct vtn_case *c = ralloc(b, struct vtn_case);
+
+ list_inithead(&c->body);
+ c->start_block = case_block;
+ c->fallthrough = NULL;
+ nir_array_init(&c->values, b);
+ c->is_default = false;
+ c->visited = false;
+
+ list_addtail(&c->link, &swtch->cases);
+
+ case_block->switch_case = c;
+ }
+
+ if (is_default) {
+ case_block->switch_case->is_default = true;
+ } else {
+ nir_array_add(&case_block->switch_case->values, uint32_t, val);
+ }
+}
+
+/* This function performs a depth-first search of the cases and puts them
+ * in fall-through order.
+ */
+static void
+vtn_order_case(struct vtn_switch *swtch, struct vtn_case *cse)
+{
+ if (cse->visited)
+ return;
+
+ cse->visited = true;
+
+ list_del(&cse->link);
+
+ if (cse->fallthrough) {
+ vtn_order_case(swtch, cse->fallthrough);
+
+ /* If we have a fall-through, place this case right before the case it
+ * falls through to. This ensures that fallthroughs come one after
+ * the other. These two can never get separated because that would
+ * imply something else falling through to the same case. Also, this
+ * can't break ordering because the DFS ensures that this case is
+ * visited before anything that falls through to it.
+ */
+ list_addtail(&cse->link, &cse->fallthrough->link);
+ } else {
+ list_add(&cse->link, &swtch->cases);
+ }
+}
+
+static enum vtn_branch_type
+vtn_get_branch_type(struct vtn_block *block,
+ struct vtn_case *swcase, struct vtn_block *switch_break,
+ struct vtn_block *loop_break, struct vtn_block *loop_cont)
+{
+ if (block->switch_case) {
+ /* This branch is actually a fallthrough */
+ assert(swcase->fallthrough == NULL ||
+ swcase->fallthrough == block->switch_case);
+ swcase->fallthrough = block->switch_case;
+ return vtn_branch_type_switch_fallthrough;
+ } else if (block == switch_break) {
+ return vtn_branch_type_switch_break;
+ } else if (block == loop_break) {
+ return vtn_branch_type_loop_break;
+ } else if (block == loop_cont) {
+ return vtn_branch_type_loop_continue;
+ } else {
+ return vtn_branch_type_none;
+ }
+}
+
+static void
+vtn_cfg_walk_blocks(struct vtn_builder *b, struct list_head *cf_list,
+ struct vtn_block *start, struct vtn_case *switch_case,
+ struct vtn_block *switch_break,
+ struct vtn_block *loop_break, struct vtn_block *loop_cont,
+ struct vtn_block *end)
+{
+ struct vtn_block *block = start;
+ while (block != end) {
+ if (block->merge && (*block->merge & SpvOpCodeMask) == SpvOpLoopMerge &&
+ !block->loop) {
+ struct vtn_loop *loop = ralloc(b, struct vtn_loop);
+
+ loop->node.type = vtn_cf_node_type_loop;
+ list_inithead(&loop->body);
+ list_inithead(&loop->cont_body);
+ loop->control = block->merge[3];
+
+ list_addtail(&loop->node.link, cf_list);
+ block->loop = loop;
+
+ struct vtn_block *new_loop_break =
+ vtn_value(b, block->merge[1], vtn_value_type_block)->block;
+ struct vtn_block *new_loop_cont =
+ vtn_value(b, block->merge[2], vtn_value_type_block)->block;
+
+ /* Note: This recursive call will start with the current block as
+ * its start block. If we weren't careful, we would get here
+ * again and end up in infinite recursion. This is why we set
+ * block->loop above and check for it before creating one. This
+ * way, we only create the loop once and the second call that
+ * tries to handle this loop goes to the cases below and gets
+ * handled as a regular block.
+ *
+ * Note: When we make the recursive walk calls, we pass NULL for
+ * the switch break since you have to break out of the loop first.
+ * We do, however, still pass the current switch case because it's
+ * possible that the merge block for the loop is the start of
+ * another case.
+ */
+ vtn_cfg_walk_blocks(b, &loop->body, block, switch_case, NULL,
+ new_loop_break, new_loop_cont, NULL );
+ vtn_cfg_walk_blocks(b, &loop->cont_body, new_loop_cont, NULL, NULL,
+ new_loop_break, NULL, block);
+
+ block = new_loop_break;
+ continue;
+ }
+
+ assert(block->node.link.next == NULL);
+ list_addtail(&block->node.link, cf_list);
+
+ switch (*block->branch & SpvOpCodeMask) {
+ case SpvOpBranch: {
+ struct vtn_block *branch_block =
+ vtn_value(b, block->branch[1], vtn_value_type_block)->block;
+
+ block->branch_type = vtn_get_branch_type(branch_block,
+ switch_case, switch_break,
+ loop_break, loop_cont);
+
+ if (block->branch_type != vtn_branch_type_none)
+ return;
+
+ block = branch_block;
+ continue;
+ }
+
+ case SpvOpReturn:
+ case SpvOpReturnValue:
+ block->branch_type = vtn_branch_type_return;
+ return;
+
+ case SpvOpKill:
+ block->branch_type = vtn_branch_type_discard;
+ return;
+
+ case SpvOpBranchConditional: {
+ struct vtn_block *then_block =
+ vtn_value(b, block->branch[2], vtn_value_type_block)->block;
+ struct vtn_block *else_block =
+ vtn_value(b, block->branch[3], vtn_value_type_block)->block;
+
+ struct vtn_if *if_stmt = ralloc(b, struct vtn_if);
+
+ if_stmt->node.type = vtn_cf_node_type_if;
+ if_stmt->condition = block->branch[1];
+ list_inithead(&if_stmt->then_body);
+ list_inithead(&if_stmt->else_body);
+
+ list_addtail(&if_stmt->node.link, cf_list);
+
+ if (block->merge &&
+ (*block->merge & SpvOpCodeMask) == SpvOpSelectionMerge) {
+ if_stmt->control = block->merge[2];
+ }
+
+ if_stmt->then_type = vtn_get_branch_type(then_block,
+ switch_case, switch_break,
+ loop_break, loop_cont);
+ if_stmt->else_type = vtn_get_branch_type(else_block,
+ switch_case, switch_break,
+ loop_break, loop_cont);
+
+ if (if_stmt->then_type == vtn_branch_type_none &&
+ if_stmt->else_type == vtn_branch_type_none) {
+ /* Neither side of the if is something we can short-circuit. */
+ assert((*block->merge & SpvOpCodeMask) == SpvOpSelectionMerge);
+ struct vtn_block *merge_block =
+ vtn_value(b, block->merge[1], vtn_value_type_block)->block;
+
+ vtn_cfg_walk_blocks(b, &if_stmt->then_body, then_block,
+ switch_case, switch_break,
+ loop_break, loop_cont, merge_block);
+ vtn_cfg_walk_blocks(b, &if_stmt->else_body, else_block,
+ switch_case, switch_break,
+ loop_break, loop_cont, merge_block);
+
+ enum vtn_branch_type merge_type =
+ vtn_get_branch_type(merge_block, switch_case, switch_break,
+ loop_break, loop_cont);
+ if (merge_type == vtn_branch_type_none) {
+ block = merge_block;
+ continue;
+ } else {
+ return;
+ }
+ } else if (if_stmt->then_type != vtn_branch_type_none &&
+ if_stmt->else_type != vtn_branch_type_none) {
+ /* Both sides were short-circuited. We're done here. */
+ return;
+ } else {
+ /* Exeactly one side of the branch could be short-circuited.
+ * We set the branch up as a predicated break/continue and we
+ * continue on with the other side as if it were what comes
+ * after the if.
+ */
+ if (if_stmt->then_type == vtn_branch_type_none) {
+ block = then_block;
+ } else {
+ block = else_block;
+ }
+ continue;
+ }
+ unreachable("Should have returned or continued");
+ }
+
+ case SpvOpSwitch: {
+ assert((*block->merge & SpvOpCodeMask) == SpvOpSelectionMerge);
+ struct vtn_block *break_block =
+ vtn_value(b, block->merge[1], vtn_value_type_block)->block;
+
+ struct vtn_switch *swtch = ralloc(b, struct vtn_switch);
+
+ swtch->node.type = vtn_cf_node_type_switch;
+ swtch->selector = block->branch[1];
+ list_inithead(&swtch->cases);
+
+ list_addtail(&swtch->node.link, cf_list);
+
+ /* First, we go through and record all of the cases. */
+ const uint32_t *branch_end =
+ block->branch + (block->branch[0] >> SpvWordCountShift);
+
+ vtn_add_case(b, swtch, break_block, block->branch[2], 0, true);
+ for (const uint32_t *w = block->branch + 3; w < branch_end; w += 2)
+ vtn_add_case(b, swtch, break_block, w[1], w[0], false);
+
+ /* Now, we go through and walk the blocks. While we walk through
+ * the blocks, we also gather the much-needed fall-through
+ * information.
+ */
+ list_for_each_entry(struct vtn_case, cse, &swtch->cases, link) {
+ assert(cse->start_block != break_block);
+ vtn_cfg_walk_blocks(b, &cse->body, cse->start_block, cse,
+ break_block, NULL, loop_cont, NULL);
+ }
+
+ /* Finally, we walk over all of the cases one more time and put
+ * them in fall-through order.
+ */
+ for (const uint32_t *w = block->branch + 2; w < branch_end; w += 2) {
+ struct vtn_block *case_block =
+ vtn_value(b, *w, vtn_value_type_block)->block;
+
+ if (case_block == break_block)
+ continue;
+
+ assert(case_block->switch_case);
+
+ vtn_order_case(swtch, case_block->switch_case);
+ }
+
+ block = break_block;
+ continue;
+ }
+
+ case SpvOpUnreachable:
+ return;
+
+ default:
+ unreachable("Unhandled opcode");
+ }
+ }
+}
+
+void
+vtn_build_cfg(struct vtn_builder *b, const uint32_t *words, const uint32_t *end)
+{
+ vtn_foreach_instruction(b, words, end,
+ vtn_cfg_handle_prepass_instruction);
+
+ foreach_list_typed(struct vtn_function, func, node, &b->functions) {
+ vtn_cfg_walk_blocks(b, &func->body, func->start_block,
+ NULL, NULL, NULL, NULL, NULL);
+ }
+}
+
+static bool
+vtn_handle_phis_first_pass(struct vtn_builder *b, SpvOp opcode,
+ const uint32_t *w, unsigned count)
+{
+ if (opcode == SpvOpLabel)
+ return true; /* Nothing to do */
+
+ /* If this isn't a phi node, stop. */
+ if (opcode != SpvOpPhi)
+ return false;
+
+ /* For handling phi nodes, we do a poor-man's out-of-ssa on the spot.
+ * For each phi, we create a variable with the appropreate type and
+ * do a load from that variable. Then, in a second pass, we add
+ * stores to that variable to each of the predecessor blocks.
+ *
+ * We could do something more intelligent here. However, in order to
+ * handle loops and things properly, we really need dominance
+ * information. It would end up basically being the into-SSA
+ * algorithm all over again. It's easier if we just let
+ * lower_vars_to_ssa do that for us instead of repeating it here.
+ */
+ struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_ssa);
+
+ struct vtn_type *type = vtn_value(b, w[1], vtn_value_type_type)->type;
+ nir_variable *phi_var =
+ nir_local_variable_create(b->nb.impl, type->type, "phi");
+ _mesa_hash_table_insert(b->phi_table, w, phi_var);
+
+ val->ssa = vtn_variable_load(b, nir_deref_var_create(b, phi_var), type);
+
+ return true;
+}
+
+static bool
+vtn_handle_phi_second_pass(struct vtn_builder *b, SpvOp opcode,
+ const uint32_t *w, unsigned count)
+{
+ if (opcode != SpvOpPhi)
+ return true;
+
+ struct hash_entry *phi_entry = _mesa_hash_table_search(b->phi_table, w);
+ assert(phi_entry);
+ nir_variable *phi_var = phi_entry->data;
+
+ struct vtn_type *type = vtn_value(b, w[1], vtn_value_type_type)->type;
+
+ for (unsigned i = 3; i < count; i += 2) {
+ struct vtn_ssa_value *src = vtn_ssa_value(b, w[i]);
+ struct vtn_block *pred =
+ vtn_value(b, w[i + 1], vtn_value_type_block)->block;
+
+ b->nb.cursor = nir_after_block_before_jump(pred->end_block);
+
+ vtn_variable_store(b, src, nir_deref_var_create(b, phi_var), type);
+ }
+
+ return true;
+}
+
+static void
+vtn_emit_branch(struct vtn_builder *b, enum vtn_branch_type branch_type,
+ nir_variable *switch_fall_var, bool *has_switch_break)
+{
+ switch (branch_type) {
+ case vtn_branch_type_switch_break:
+ nir_store_var(&b->nb, switch_fall_var, nir_imm_int(&b->nb, NIR_FALSE), 1);
+ *has_switch_break = true;
+ break;
+ case vtn_branch_type_switch_fallthrough:
+ break; /* Nothing to do */
+ case vtn_branch_type_loop_break:
+ nir_jump(&b->nb, nir_jump_break);
+ break;
+ case vtn_branch_type_loop_continue:
+ nir_jump(&b->nb, nir_jump_continue);
+ break;
+ case vtn_branch_type_return:
+ nir_jump(&b->nb, nir_jump_return);
+ break;
+ case vtn_branch_type_discard: {
+ nir_intrinsic_instr *discard =
+ nir_intrinsic_instr_create(b->nb.shader, nir_intrinsic_discard);
+ nir_builder_instr_insert(&b->nb, &discard->instr);
+ break;
+ }
+ default:
+ unreachable("Invalid branch type");
+ }
+}
+
+static void
+vtn_emit_cf_list(struct vtn_builder *b, struct list_head *cf_list,
+ nir_variable *switch_fall_var, bool *has_switch_break,
+ vtn_instruction_handler handler)
+{
+ list_for_each_entry(struct vtn_cf_node, node, cf_list, link) {
+ switch (node->type) {
+ case vtn_cf_node_type_block: {
+ struct vtn_block *block = (struct vtn_block *)node;
+
+ const uint32_t *block_start = block->label;
+ const uint32_t *block_end = block->merge ? block->merge :
+ block->branch;
+
+ block_start = vtn_foreach_instruction(b, block_start, block_end,
+ vtn_handle_phis_first_pass);
+
+ vtn_foreach_instruction(b, block_start, block_end, handler);
+
+ block->end_block = nir_cursor_current_block(b->nb.cursor);
+
+ if ((*block->branch & SpvOpCodeMask) == SpvOpReturnValue) {
+ struct vtn_ssa_value *src = vtn_ssa_value(b, block->branch[1]);
+ vtn_variable_store(b, src,
+ nir_deref_var_create(b, b->impl->return_var),
+ NULL);
+ }
+
+ if (block->branch_type != vtn_branch_type_none) {
+ vtn_emit_branch(b, block->branch_type,
+ switch_fall_var, has_switch_break);
+ }
+
+ break;
+ }
+
+ case vtn_cf_node_type_if: {
+ struct vtn_if *vtn_if = (struct vtn_if *)node;
+
+ nir_if *if_stmt = nir_if_create(b->shader);
+ if_stmt->condition =
+ nir_src_for_ssa(vtn_ssa_value(b, vtn_if->condition)->def);
+ nir_cf_node_insert(b->nb.cursor, &if_stmt->cf_node);
+
+ bool sw_break = false;
+
+ b->nb.cursor = nir_after_cf_list(&if_stmt->then_list);
+ if (vtn_if->then_type == vtn_branch_type_none) {
+ vtn_emit_cf_list(b, &vtn_if->then_body,
+ switch_fall_var, &sw_break, handler);
+ } else {
+ vtn_emit_branch(b, vtn_if->then_type, switch_fall_var, &sw_break);
+ }
+
+ b->nb.cursor = nir_after_cf_list(&if_stmt->else_list);
+ if (vtn_if->else_type == vtn_branch_type_none) {
+ vtn_emit_cf_list(b, &vtn_if->else_body,
+ switch_fall_var, &sw_break, handler);
+ } else {
+ vtn_emit_branch(b, vtn_if->else_type, switch_fall_var, &sw_break);
+ }
+
+ b->nb.cursor = nir_after_cf_node(&if_stmt->cf_node);
+
+ /* If we encountered a switch break somewhere inside of the if,
+ * then it would have been handled correctly by calling
+ * emit_cf_list or emit_branch for the interrior. However, we
+ * need to predicate everything following on wether or not we're
+ * still going.
+ */
+ if (sw_break) {
+ *has_switch_break = true;
+
+ nir_if *switch_if = nir_if_create(b->shader);
+ switch_if->condition =
+ nir_src_for_ssa(nir_load_var(&b->nb, switch_fall_var));
+ nir_cf_node_insert(b->nb.cursor, &switch_if->cf_node);
+
+ b->nb.cursor = nir_after_cf_list(&if_stmt->then_list);
+ }
+ break;
+ }
+
+ case vtn_cf_node_type_loop: {
+ struct vtn_loop *vtn_loop = (struct vtn_loop *)node;
+
+ nir_loop *loop = nir_loop_create(b->shader);
+ nir_cf_node_insert(b->nb.cursor, &loop->cf_node);
+
+ b->nb.cursor = nir_after_cf_list(&loop->body);
+ vtn_emit_cf_list(b, &vtn_loop->body, NULL, NULL, handler);
+
+ if (!list_empty(&vtn_loop->cont_body)) {
+ /* If we have a non-trivial continue body then we need to put
+ * it at the beginning of the loop with a flag to ensure that
+ * it doesn't get executed in the first iteration.
+ */
+ nir_variable *do_cont =
+ nir_local_variable_create(b->nb.impl, glsl_bool_type(), "cont");
+
+ b->nb.cursor = nir_before_cf_node(&loop->cf_node);
+ nir_store_var(&b->nb, do_cont, nir_imm_int(&b->nb, NIR_FALSE), 1);
+
+ b->nb.cursor = nir_before_cf_list(&loop->body);
+ nir_if *cont_if = nir_if_create(b->shader);
+ cont_if->condition = nir_src_for_ssa(nir_load_var(&b->nb, do_cont));
+ nir_cf_node_insert(b->nb.cursor, &cont_if->cf_node);
+
+ b->nb.cursor = nir_after_cf_list(&cont_if->then_list);
+ vtn_emit_cf_list(b, &vtn_loop->cont_body, NULL, NULL, handler);
+
+ b->nb.cursor = nir_after_cf_node(&cont_if->cf_node);
+ nir_store_var(&b->nb, do_cont, nir_imm_int(&b->nb, NIR_TRUE), 1);
+
+ b->has_loop_continue = true;
+ }
+
+ b->nb.cursor = nir_after_cf_node(&loop->cf_node);
+ break;
+ }
+
+ case vtn_cf_node_type_switch: {
+ struct vtn_switch *vtn_switch = (struct vtn_switch *)node;
+
+ /* First, we create a variable to keep track of whether or not the
+ * switch is still going at any given point. Any switch breaks
+ * will set this variable to false.
+ */
+ nir_variable *fall_var =
+ nir_local_variable_create(b->nb.impl, glsl_bool_type(), "fall");
+ nir_store_var(&b->nb, fall_var, nir_imm_int(&b->nb, NIR_FALSE), 1);
+
+ /* Next, we gather up all of the conditions. We have to do this
+ * up-front because we also need to build an "any" condition so
+ * that we can use !any for default.
+ */
+ const int num_cases = list_length(&vtn_switch->cases);
+ NIR_VLA(nir_ssa_def *, conditions, num_cases);
+
+ nir_ssa_def *sel = vtn_ssa_value(b, vtn_switch->selector)->def;
+ /* An accumulation of all conditions. Used for the default */
+ nir_ssa_def *any = NULL;
+
+ int i = 0;
+ list_for_each_entry(struct vtn_case, cse, &vtn_switch->cases, link) {
+ if (cse->is_default) {
+ conditions[i++] = NULL;
+ continue;
+ }
+
+ nir_ssa_def *cond = NULL;
+ nir_array_foreach(&cse->values, uint32_t, val) {
+ nir_ssa_def *is_val =
+ nir_ieq(&b->nb, sel, nir_imm_int(&b->nb, *val));
+
+ cond = cond ? nir_ior(&b->nb, cond, is_val) : is_val;
+ }
+
+ any = any ? nir_ior(&b->nb, any, cond) : cond;
+ conditions[i++] = cond;
+ }
+ assert(i == num_cases);
+
+ /* Now we can walk the list of cases and actually emit code */
+ i = 0;
+ list_for_each_entry(struct vtn_case, cse, &vtn_switch->cases, link) {
+ /* Figure out the condition */
+ nir_ssa_def *cond = conditions[i++];
+ if (cse->is_default) {
+ assert(cond == NULL);
+ cond = nir_inot(&b->nb, any);
+ }
+ /* Take fallthrough into account */
+ cond = nir_ior(&b->nb, cond, nir_load_var(&b->nb, fall_var));
+
+ nir_if *case_if = nir_if_create(b->nb.shader);
+ case_if->condition = nir_src_for_ssa(cond);
+ nir_cf_node_insert(b->nb.cursor, &case_if->cf_node);
+
+ bool has_break = false;
+ b->nb.cursor = nir_after_cf_list(&case_if->then_list);
+ nir_store_var(&b->nb, fall_var, nir_imm_int(&b->nb, NIR_TRUE), 1);
+ vtn_emit_cf_list(b, &cse->body, fall_var, &has_break, handler);
+ (void)has_break; /* We don't care */
+
+ b->nb.cursor = nir_after_cf_node(&case_if->cf_node);
+ }
+ assert(i == num_cases);
+
+ break;
+ }
+
+ default:
+ unreachable("Invalid CF node type");
+ }
+ }
+}
+
+void
+vtn_function_emit(struct vtn_builder *b, struct vtn_function *func,
+ vtn_instruction_handler instruction_handler)
+{
+ nir_builder_init(&b->nb, func->impl);
+ b->nb.cursor = nir_after_cf_list(&func->impl->body);
+ b->has_loop_continue = false;
+ b->phi_table = _mesa_hash_table_create(b, _mesa_hash_pointer,
+ _mesa_key_pointer_equal);
+
+ vtn_emit_cf_list(b, &func->body, NULL, NULL, instruction_handler);
+
+ vtn_foreach_instruction(b, func->start_block->label, func->end,
+ vtn_handle_phi_second_pass);
+
+ /* Continue blocks for loops get inserted before the body of the loop
+ * but instructions in the continue may use SSA defs in the loop body.
+ * Therefore, we need to repair SSA to insert the needed phi nodes.
+ */
+ if (b->has_loop_continue)
+ nir_repair_ssa_impl(func->impl);
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ * Authors:
+ * Jason Ekstrand (jason@jlekstrand.net)
+ *
+ */
+
+#include "vtn_private.h"
+#include "GLSL.std.450.h"
+
+#define M_PIf ((float) M_PI)
+#define M_PI_2f ((float) M_PI_2)
+#define M_PI_4f ((float) M_PI_4)
+
+static nir_ssa_def *
+build_mat2_det(nir_builder *b, nir_ssa_def *col[2])
+{
+ unsigned swiz[4] = {1, 0, 0, 0};
+ nir_ssa_def *p = nir_fmul(b, col[0], nir_swizzle(b, col[1], swiz, 2, true));
+ return nir_fsub(b, nir_channel(b, p, 0), nir_channel(b, p, 1));
+}
+
+static nir_ssa_def *
+build_mat3_det(nir_builder *b, nir_ssa_def *col[3])
+{
+ unsigned yzx[4] = {1, 2, 0, 0};
+ unsigned zxy[4] = {2, 0, 1, 0};
+
+ nir_ssa_def *prod0 =
+ nir_fmul(b, col[0],
+ nir_fmul(b, nir_swizzle(b, col[1], yzx, 3, true),
+ nir_swizzle(b, col[2], zxy, 3, true)));
+ nir_ssa_def *prod1 =
+ nir_fmul(b, col[0],
+ nir_fmul(b, nir_swizzle(b, col[1], zxy, 3, true),
+ nir_swizzle(b, col[2], yzx, 3, true)));
+
+ nir_ssa_def *diff = nir_fsub(b, prod0, prod1);
+
+ return nir_fadd(b, nir_channel(b, diff, 0),
+ nir_fadd(b, nir_channel(b, diff, 1),
+ nir_channel(b, diff, 2)));
+}
+
+static nir_ssa_def *
+build_mat4_det(nir_builder *b, nir_ssa_def **col)
+{
+ nir_ssa_def *subdet[4];
+ for (unsigned i = 0; i < 4; i++) {
+ unsigned swiz[3];
+ for (unsigned j = 0; j < 4; j++)
+ swiz[j - (j > i)] = j;
+
+ nir_ssa_def *subcol[3];
+ subcol[0] = nir_swizzle(b, col[1], swiz, 3, true);
+ subcol[1] = nir_swizzle(b, col[2], swiz, 3, true);
+ subcol[2] = nir_swizzle(b, col[3], swiz, 3, true);
+
+ subdet[i] = build_mat3_det(b, subcol);
+ }
+
+ nir_ssa_def *prod = nir_fmul(b, col[0], nir_vec(b, subdet, 4));
+
+ return nir_fadd(b, nir_fsub(b, nir_channel(b, prod, 0),
+ nir_channel(b, prod, 1)),
+ nir_fsub(b, nir_channel(b, prod, 2),
+ nir_channel(b, prod, 3)));
+}
+
+static nir_ssa_def *
+build_mat_det(struct vtn_builder *b, struct vtn_ssa_value *src)
+{
+ unsigned size = glsl_get_vector_elements(src->type);
+
+ nir_ssa_def *cols[4];
+ for (unsigned i = 0; i < size; i++)
+ cols[i] = src->elems[i]->def;
+
+ switch(size) {
+ case 2: return build_mat2_det(&b->nb, cols);
+ case 3: return build_mat3_det(&b->nb, cols);
+ case 4: return build_mat4_det(&b->nb, cols);
+ default:
+ unreachable("Invalid matrix size");
+ }
+}
+
+/* Computes the determinate of the submatrix given by taking src and
+ * removing the specified row and column.
+ */
+static nir_ssa_def *
+build_mat_subdet(struct nir_builder *b, struct vtn_ssa_value *src,
+ unsigned size, unsigned row, unsigned col)
+{
+ assert(row < size && col < size);
+ if (size == 2) {
+ return nir_channel(b, src->elems[1 - col]->def, 1 - row);
+ } else {
+ /* Swizzle to get all but the specified row */
+ unsigned swiz[3];
+ for (unsigned j = 0; j < 4; j++)
+ swiz[j - (j > row)] = j;
+
+ /* Grab all but the specified column */
+ nir_ssa_def *subcol[3];
+ for (unsigned j = 0; j < size; j++) {
+ if (j != col) {
+ subcol[j - (j > col)] = nir_swizzle(b, src->elems[j]->def,
+ swiz, size - 1, true);
+ }
+ }
+
+ if (size == 3) {
+ return build_mat2_det(b, subcol);
+ } else {
+ assert(size == 4);
+ return build_mat3_det(b, subcol);
+ }
+ }
+}
+
+static struct vtn_ssa_value *
+matrix_inverse(struct vtn_builder *b, struct vtn_ssa_value *src)
+{
+ nir_ssa_def *adj_col[4];
+ unsigned size = glsl_get_vector_elements(src->type);
+
+ /* Build up an adjugate matrix */
+ for (unsigned c = 0; c < size; c++) {
+ nir_ssa_def *elem[4];
+ for (unsigned r = 0; r < size; r++) {
+ elem[r] = build_mat_subdet(&b->nb, src, size, c, r);
+
+ if ((r + c) % 2)
+ elem[r] = nir_fneg(&b->nb, elem[r]);
+ }
+
+ adj_col[c] = nir_vec(&b->nb, elem, size);
+ }
+
+ nir_ssa_def *det_inv = nir_frcp(&b->nb, build_mat_det(b, src));
+
+ struct vtn_ssa_value *val = vtn_create_ssa_value(b, src->type);
+ for (unsigned i = 0; i < size; i++)
+ val->elems[i]->def = nir_fmul(&b->nb, adj_col[i], det_inv);
+
+ return val;
+}
+
+static nir_ssa_def*
+build_length(nir_builder *b, nir_ssa_def *vec)
+{
+ switch (vec->num_components) {
+ case 1: return nir_fsqrt(b, nir_fmul(b, vec, vec));
+ case 2: return nir_fsqrt(b, nir_fdot2(b, vec, vec));
+ case 3: return nir_fsqrt(b, nir_fdot3(b, vec, vec));
+ case 4: return nir_fsqrt(b, nir_fdot4(b, vec, vec));
+ default:
+ unreachable("Invalid number of components");
+ }
+}
+
+static inline nir_ssa_def *
+build_fclamp(nir_builder *b,
+ nir_ssa_def *x, nir_ssa_def *min_val, nir_ssa_def *max_val)
+{
+ return nir_fmin(b, nir_fmax(b, x, min_val), max_val);
+}
+
+/**
+ * Return e^x.
+ */
+static nir_ssa_def *
+build_exp(nir_builder *b, nir_ssa_def *x)
+{
+ return nir_fexp2(b, nir_fmul(b, x, nir_imm_float(b, M_LOG2E)));
+}
+
+/**
+ * Return ln(x) - the natural logarithm of x.
+ */
+static nir_ssa_def *
+build_log(nir_builder *b, nir_ssa_def *x)
+{
+ return nir_fmul(b, nir_flog2(b, x), nir_imm_float(b, 1.0 / M_LOG2E));
+}
+
+static nir_ssa_def *
+build_asin(nir_builder *b, nir_ssa_def *x)
+{
+ nir_ssa_def *abs_x = nir_fabs(b, x);
+ return nir_fmul(b, nir_fsign(b, x),
+ nir_fsub(b, nir_imm_float(b, M_PI_2f),
+ nir_fmul(b, nir_fsqrt(b, nir_fsub(b, nir_imm_float(b, 1.0f), abs_x)),
+ nir_fadd(b, nir_imm_float(b, M_PI_2f),
+ nir_fmul(b, abs_x,
+ nir_fadd(b, nir_imm_float(b, M_PI_4f - 1.0f),
+ nir_fmul(b, abs_x,
+ nir_fadd(b, nir_imm_float(b, 0.086566724f),
+ nir_fmul(b, abs_x,
+ nir_imm_float(b, -0.03102955f))))))))));
+}
+
+/**
+ * Compute xs[0] + xs[1] + xs[2] + ... using fadd.
+ */
+static nir_ssa_def *
+build_fsum(nir_builder *b, nir_ssa_def **xs, int terms)
+{
+ nir_ssa_def *accum = xs[0];
+
+ for (int i = 1; i < terms; i++)
+ accum = nir_fadd(b, accum, xs[i]);
+
+ return accum;
+}
+
+static nir_ssa_def *
+build_atan(nir_builder *b, nir_ssa_def *y_over_x)
+{
+ nir_ssa_def *abs_y_over_x = nir_fabs(b, y_over_x);
+ nir_ssa_def *one = nir_imm_float(b, 1.0f);
+
+ /*
+ * range-reduction, first step:
+ *
+ * / y_over_x if |y_over_x| <= 1.0;
+ * x = <
+ * \ 1.0 / y_over_x otherwise
+ */
+ nir_ssa_def *x = nir_fdiv(b, nir_fmin(b, abs_y_over_x, one),
+ nir_fmax(b, abs_y_over_x, one));
+
+ /*
+ * approximate atan by evaluating polynomial:
+ *
+ * x * 0.9999793128310355 - x^3 * 0.3326756418091246 +
+ * x^5 * 0.1938924977115610 - x^7 * 0.1173503194786851 +
+ * x^9 * 0.0536813784310406 - x^11 * 0.0121323213173444
+ */
+ nir_ssa_def *x_2 = nir_fmul(b, x, x);
+ nir_ssa_def *x_3 = nir_fmul(b, x_2, x);
+ nir_ssa_def *x_5 = nir_fmul(b, x_3, x_2);
+ nir_ssa_def *x_7 = nir_fmul(b, x_5, x_2);
+ nir_ssa_def *x_9 = nir_fmul(b, x_7, x_2);
+ nir_ssa_def *x_11 = nir_fmul(b, x_9, x_2);
+
+ nir_ssa_def *polynomial_terms[] = {
+ nir_fmul(b, x, nir_imm_float(b, 0.9999793128310355f)),
+ nir_fmul(b, x_3, nir_imm_float(b, -0.3326756418091246f)),
+ nir_fmul(b, x_5, nir_imm_float(b, 0.1938924977115610f)),
+ nir_fmul(b, x_7, nir_imm_float(b, -0.1173503194786851f)),
+ nir_fmul(b, x_9, nir_imm_float(b, 0.0536813784310406f)),
+ nir_fmul(b, x_11, nir_imm_float(b, -0.0121323213173444f)),
+ };
+
+ nir_ssa_def *tmp =
+ build_fsum(b, polynomial_terms, ARRAY_SIZE(polynomial_terms));
+
+ /* range-reduction fixup */
+ tmp = nir_fadd(b, tmp,
+ nir_fmul(b,
+ nir_b2f(b, nir_flt(b, one, abs_y_over_x)),
+ nir_fadd(b, nir_fmul(b, tmp,
+ nir_imm_float(b, -2.0f)),
+ nir_imm_float(b, M_PI_2f))));
+
+ /* sign fixup */
+ return nir_fmul(b, tmp, nir_fsign(b, y_over_x));
+}
+
+static nir_ssa_def *
+build_atan2(nir_builder *b, nir_ssa_def *y, nir_ssa_def *x)
+{
+ nir_ssa_def *zero = nir_imm_float(b, 0.0f);
+
+ /* If |x| >= 1.0e-8 * |y|: */
+ nir_if *if_stmt = nir_if_create(b->shader);
+ if_stmt->condition = nir_src_for_ssa(
+ nir_fge(b, nir_fabs(b, x),
+ nir_fmul(b, nir_imm_float(b, 1.0e-8f), nir_fabs(b, y))));
+ nir_builder_cf_insert(b, &if_stmt->cf_node);
+
+ /* Then...call atan(y/x) and fix it up: */
+ b->cursor = nir_after_cf_list(&if_stmt->then_list);
+ nir_ssa_def *atan1 = build_atan(b, nir_fdiv(b, y, x));
+ nir_ssa_def *r_then =
+ nir_bcsel(b, nir_flt(b, x, zero),
+ nir_fadd(b, atan1,
+ nir_bcsel(b, nir_fge(b, y, zero),
+ nir_imm_float(b, M_PIf),
+ nir_imm_float(b, -M_PIf))),
+ atan1);
+
+ /* Else... */
+ b->cursor = nir_after_cf_list(&if_stmt->else_list);
+ nir_ssa_def *r_else =
+ nir_fmul(b, nir_fsign(b, y), nir_imm_float(b, M_PI_2f));
+
+ b->cursor = nir_after_cf_node(&if_stmt->cf_node);
+
+ nir_phi_instr *phi = nir_phi_instr_create(b->shader);
+ nir_ssa_dest_init(&phi->instr, &phi->dest, r_then->num_components, NULL);
+
+ nir_phi_src *phi_src0 = ralloc(phi, nir_phi_src);
+ nir_phi_src *phi_src1 = ralloc(phi, nir_phi_src);
+
+ phi_src0->pred = nir_cf_node_as_block((nir_cf_node *) exec_list_get_head(&if_stmt->then_list));
+ phi_src0->src = nir_src_for_ssa(r_then);
+ exec_list_push_tail(&phi->srcs, &phi_src0->node);
+ phi_src1->pred = nir_cf_node_as_block((nir_cf_node *) exec_list_get_head(&if_stmt->else_list));
+ phi_src1->src = nir_src_for_ssa(r_else);
+ exec_list_push_tail(&phi->srcs, &phi_src1->node);
+
+ nir_builder_instr_insert(b, &phi->instr);
+
+ return &phi->dest.ssa;
+}
+
+static nir_ssa_def *
+build_frexp(nir_builder *b, nir_ssa_def *x, nir_ssa_def **exponent)
+{
+ nir_ssa_def *abs_x = nir_fabs(b, x);
+ nir_ssa_def *zero = nir_imm_float(b, 0.0f);
+
+ /* Single-precision floating-point values are stored as
+ * 1 sign bit;
+ * 8 exponent bits;
+ * 23 mantissa bits.
+ *
+ * An exponent shift of 23 will shift the mantissa out, leaving only the
+ * exponent and sign bit (which itself may be zero, if the absolute value
+ * was taken before the bitcast and shift.
+ */
+ nir_ssa_def *exponent_shift = nir_imm_int(b, 23);
+ nir_ssa_def *exponent_bias = nir_imm_int(b, -126);
+
+ nir_ssa_def *sign_mantissa_mask = nir_imm_int(b, 0x807fffffu);
+
+ /* Exponent of floating-point values in the range [0.5, 1.0). */
+ nir_ssa_def *exponent_value = nir_imm_int(b, 0x3f000000u);
+
+ nir_ssa_def *is_not_zero = nir_fne(b, abs_x, zero);
+
+ *exponent =
+ nir_iadd(b, nir_ushr(b, abs_x, exponent_shift),
+ nir_bcsel(b, is_not_zero, exponent_bias, zero));
+
+ return nir_ior(b, nir_iand(b, x, sign_mantissa_mask),
+ nir_bcsel(b, is_not_zero, exponent_value, zero));
+}
+
+static void
+handle_glsl450_alu(struct vtn_builder *b, enum GLSLstd450 entrypoint,
+ const uint32_t *w, unsigned count)
+{
+ struct nir_builder *nb = &b->nb;
+ const struct glsl_type *dest_type =
+ vtn_value(b, w[1], vtn_value_type_type)->type->type;
+
+ struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_ssa);
+ val->ssa = vtn_create_ssa_value(b, dest_type);
+
+ /* Collect the various SSA sources */
+ unsigned num_inputs = count - 5;
+ nir_ssa_def *src[3];
+ for (unsigned i = 0; i < num_inputs; i++)
+ src[i] = vtn_ssa_value(b, w[i + 5])->def;
+
+ nir_op op;
+ switch (entrypoint) {
+ case GLSLstd450Round: op = nir_op_fround_even; break; /* TODO */
+ case GLSLstd450RoundEven: op = nir_op_fround_even; break;
+ case GLSLstd450Trunc: op = nir_op_ftrunc; break;
+ case GLSLstd450FAbs: op = nir_op_fabs; break;
+ case GLSLstd450SAbs: op = nir_op_iabs; break;
+ case GLSLstd450FSign: op = nir_op_fsign; break;
+ case GLSLstd450SSign: op = nir_op_isign; break;
+ case GLSLstd450Floor: op = nir_op_ffloor; break;
+ case GLSLstd450Ceil: op = nir_op_fceil; break;
+ case GLSLstd450Fract: op = nir_op_ffract; break;
+ case GLSLstd450Radians:
+ val->ssa->def = nir_fmul(nb, src[0], nir_imm_float(nb, 0.01745329251));
+ return;
+ case GLSLstd450Degrees:
+ val->ssa->def = nir_fmul(nb, src[0], nir_imm_float(nb, 57.2957795131));
+ return;
+ case GLSLstd450Sin: op = nir_op_fsin; break;
+ case GLSLstd450Cos: op = nir_op_fcos; break;
+ case GLSLstd450Tan:
+ val->ssa->def = nir_fdiv(nb, nir_fsin(nb, src[0]),
+ nir_fcos(nb, src[0]));
+ return;
+ case GLSLstd450Pow: op = nir_op_fpow; break;
+ case GLSLstd450Exp2: op = nir_op_fexp2; break;
+ case GLSLstd450Log2: op = nir_op_flog2; break;
+ case GLSLstd450Sqrt: op = nir_op_fsqrt; break;
+ case GLSLstd450InverseSqrt: op = nir_op_frsq; break;
+
+ case GLSLstd450Modf: {
+ val->ssa->def = nir_ffract(nb, src[0]);
+ nir_deref_var *out = vtn_value(b, w[6], vtn_value_type_deref)->deref;
+ nir_store_deref_var(nb, out, nir_ffloor(nb, src[0]), 0xf);
+ return;
+ }
+
+ op = nir_op_fmod; break;
+ case GLSLstd450FMin: op = nir_op_fmin; break;
+ case GLSLstd450UMin: op = nir_op_umin; break;
+ case GLSLstd450SMin: op = nir_op_imin; break;
+ case GLSLstd450FMax: op = nir_op_fmax; break;
+ case GLSLstd450UMax: op = nir_op_umax; break;
+ case GLSLstd450SMax: op = nir_op_imax; break;
+ case GLSLstd450FMix: op = nir_op_flrp; break;
+ case GLSLstd450Step:
+ val->ssa->def = nir_sge(nb, src[1], src[0]);
+ return;
+
+ case GLSLstd450Fma: op = nir_op_ffma; break;
+ case GLSLstd450Ldexp: op = nir_op_ldexp; break;
+
+ /* Packing/Unpacking functions */
+ case GLSLstd450PackSnorm4x8: op = nir_op_pack_snorm_4x8; break;
+ case GLSLstd450PackUnorm4x8: op = nir_op_pack_unorm_4x8; break;
+ case GLSLstd450PackSnorm2x16: op = nir_op_pack_snorm_2x16; break;
+ case GLSLstd450PackUnorm2x16: op = nir_op_pack_unorm_2x16; break;
+ case GLSLstd450PackHalf2x16: op = nir_op_pack_half_2x16; break;
+ case GLSLstd450UnpackSnorm4x8: op = nir_op_unpack_snorm_4x8; break;
+ case GLSLstd450UnpackUnorm4x8: op = nir_op_unpack_unorm_4x8; break;
+ case GLSLstd450UnpackSnorm2x16: op = nir_op_unpack_snorm_2x16; break;
+ case GLSLstd450UnpackUnorm2x16: op = nir_op_unpack_unorm_2x16; break;
+ case GLSLstd450UnpackHalf2x16: op = nir_op_unpack_half_2x16; break;
+
+ case GLSLstd450Length:
+ val->ssa->def = build_length(nb, src[0]);
+ return;
+ case GLSLstd450Distance:
+ val->ssa->def = build_length(nb, nir_fsub(nb, src[0], src[1]));
+ return;
+ case GLSLstd450Normalize:
+ val->ssa->def = nir_fdiv(nb, src[0], build_length(nb, src[0]));
+ return;
+
+ case GLSLstd450Exp:
+ val->ssa->def = build_exp(nb, src[0]);
+ return;
+
+ case GLSLstd450Log:
+ val->ssa->def = build_log(nb, src[0]);
+ return;
+
+ case GLSLstd450FClamp:
+ val->ssa->def = build_fclamp(nb, src[0], src[1], src[2]);
+ return;
+ case GLSLstd450UClamp:
+ val->ssa->def = nir_umin(nb, nir_umax(nb, src[0], src[1]), src[2]);
+ return;
+ case GLSLstd450SClamp:
+ val->ssa->def = nir_imin(nb, nir_imax(nb, src[0], src[1]), src[2]);
+ return;
+
+ case GLSLstd450Cross: {
+ unsigned yzx[4] = { 1, 2, 0, 0 };
+ unsigned zxy[4] = { 2, 0, 1, 0 };
+ val->ssa->def =
+ nir_fsub(nb, nir_fmul(nb, nir_swizzle(nb, src[0], yzx, 3, true),
+ nir_swizzle(nb, src[1], zxy, 3, true)),
+ nir_fmul(nb, nir_swizzle(nb, src[0], zxy, 3, true),
+ nir_swizzle(nb, src[1], yzx, 3, true)));
+ return;
+ }
+
+ case GLSLstd450SmoothStep: {
+ /* t = clamp((x - edge0) / (edge1 - edge0), 0, 1) */
+ nir_ssa_def *t =
+ build_fclamp(nb, nir_fdiv(nb, nir_fsub(nb, src[2], src[0]),
+ nir_fsub(nb, src[1], src[0])),
+ nir_imm_float(nb, 0.0), nir_imm_float(nb, 1.0));
+ /* result = t * t * (3 - 2 * t) */
+ val->ssa->def =
+ nir_fmul(nb, t, nir_fmul(nb, t,
+ nir_fsub(nb, nir_imm_float(nb, 3.0),
+ nir_fmul(nb, nir_imm_float(nb, 2.0), t))));
+ return;
+ }
+
+ case GLSLstd450FaceForward:
+ val->ssa->def =
+ nir_bcsel(nb, nir_flt(nb, nir_fdot(nb, src[2], src[1]),
+ nir_imm_float(nb, 0.0)),
+ src[0], nir_fneg(nb, src[0]));
+ return;
+
+ case GLSLstd450Reflect:
+ /* I - 2 * dot(N, I) * N */
+ val->ssa->def =
+ nir_fsub(nb, src[0], nir_fmul(nb, nir_imm_float(nb, 2.0),
+ nir_fmul(nb, nir_fdot(nb, src[0], src[1]),
+ src[1])));
+ return;
+
+ case GLSLstd450Refract: {
+ nir_ssa_def *I = src[0];
+ nir_ssa_def *N = src[1];
+ nir_ssa_def *eta = src[2];
+ nir_ssa_def *n_dot_i = nir_fdot(nb, N, I);
+ nir_ssa_def *one = nir_imm_float(nb, 1.0);
+ nir_ssa_def *zero = nir_imm_float(nb, 0.0);
+ /* k = 1.0 - eta * eta * (1.0 - dot(N, I) * dot(N, I)) */
+ nir_ssa_def *k =
+ nir_fsub(nb, one, nir_fmul(nb, eta, nir_fmul(nb, eta,
+ nir_fsub(nb, one, nir_fmul(nb, n_dot_i, n_dot_i)))));
+ nir_ssa_def *result =
+ nir_fsub(nb, nir_fmul(nb, eta, I),
+ nir_fmul(nb, nir_fadd(nb, nir_fmul(nb, eta, n_dot_i),
+ nir_fsqrt(nb, k)), N));
+ /* XXX: bcsel, or if statement? */
+ val->ssa->def = nir_bcsel(nb, nir_flt(nb, k, zero), zero, result);
+ return;
+ }
+
+ case GLSLstd450Sinh:
+ /* 0.5 * (e^x - e^(-x)) */
+ val->ssa->def =
+ nir_fmul(nb, nir_imm_float(nb, 0.5f),
+ nir_fsub(nb, build_exp(nb, src[0]),
+ build_exp(nb, nir_fneg(nb, src[0]))));
+ return;
+
+ case GLSLstd450Cosh:
+ /* 0.5 * (e^x + e^(-x)) */
+ val->ssa->def =
+ nir_fmul(nb, nir_imm_float(nb, 0.5f),
+ nir_fadd(nb, build_exp(nb, src[0]),
+ build_exp(nb, nir_fneg(nb, src[0]))));
+ return;
+
+ case GLSLstd450Tanh:
+ /* (e^x - e^(-x)) / (e^x + e^(-x)) */
+ val->ssa->def =
+ nir_fdiv(nb, nir_fsub(nb, build_exp(nb, src[0]),
+ build_exp(nb, nir_fneg(nb, src[0]))),
+ nir_fadd(nb, build_exp(nb, src[0]),
+ build_exp(nb, nir_fneg(nb, src[0]))));
+ return;
+
+ case GLSLstd450Asinh:
+ val->ssa->def = nir_fmul(nb, nir_fsign(nb, src[0]),
+ build_log(nb, nir_fadd(nb, nir_fabs(nb, src[0]),
+ nir_fsqrt(nb, nir_fadd(nb, nir_fmul(nb, src[0], src[0]),
+ nir_imm_float(nb, 1.0f))))));
+ return;
+ case GLSLstd450Acosh:
+ val->ssa->def = build_log(nb, nir_fadd(nb, src[0],
+ nir_fsqrt(nb, nir_fsub(nb, nir_fmul(nb, src[0], src[0]),
+ nir_imm_float(nb, 1.0f)))));
+ return;
+ case GLSLstd450Atanh: {
+ nir_ssa_def *one = nir_imm_float(nb, 1.0);
+ val->ssa->def = nir_fmul(nb, nir_imm_float(nb, 0.5f),
+ build_log(nb, nir_fdiv(nb, nir_fadd(nb, one, src[0]),
+ nir_fsub(nb, one, src[0]))));
+ return;
+ }
+
+ case GLSLstd450FindILsb: op = nir_op_find_lsb; break;
+ case GLSLstd450FindSMsb: op = nir_op_ifind_msb; break;
+ case GLSLstd450FindUMsb: op = nir_op_ufind_msb; break;
+
+ case GLSLstd450Asin:
+ val->ssa->def = build_asin(nb, src[0]);
+ return;
+
+ case GLSLstd450Acos:
+ val->ssa->def = nir_fsub(nb, nir_imm_float(nb, M_PI_2f),
+ build_asin(nb, src[0]));
+ return;
+
+ case GLSLstd450Atan:
+ val->ssa->def = build_atan(nb, src[0]);
+ return;
+
+ case GLSLstd450Atan2:
+ val->ssa->def = build_atan2(nb, src[0], src[1]);
+ return;
+
+ case GLSLstd450Frexp: {
+ nir_ssa_def *exponent;
+ val->ssa->def = build_frexp(nb, src[0], &exponent);
+ nir_deref_var *out = vtn_value(b, w[6], vtn_value_type_deref)->deref;
+ nir_store_deref_var(nb, out, exponent, 0xf);
+ return;
+ }
+
+ case GLSLstd450FrexpStruct: {
+ assert(glsl_type_is_struct(val->ssa->type));
+ val->ssa->elems[0]->def = build_frexp(nb, src[0],
+ &val->ssa->elems[1]->def);
+ return;
+ }
+
+ case GLSLstd450ModfStruct:
+ case GLSLstd450PackDouble2x32:
+ case GLSLstd450UnpackDouble2x32:
+ case GLSLstd450IMix:
+ default:
+ unreachable("Unhandled opcode");
+ }
+
+ nir_alu_instr *instr = nir_alu_instr_create(b->shader, op);
+ nir_ssa_dest_init(&instr->instr, &instr->dest.dest,
+ glsl_get_vector_elements(val->ssa->type), val->name);
+ instr->dest.write_mask = (1 << instr->dest.dest.ssa.num_components) - 1;
+ val->ssa->def = &instr->dest.dest.ssa;
+
+ for (unsigned i = 0; i < nir_op_infos[op].num_inputs; i++)
+ instr->src[i].src = nir_src_for_ssa(src[i]);
+
+ nir_builder_instr_insert(nb, &instr->instr);
+}
+
+bool
+vtn_handle_glsl450_instruction(struct vtn_builder *b, uint32_t ext_opcode,
+ const uint32_t *w, unsigned count)
+{
+ switch ((enum GLSLstd450)ext_opcode) {
+ case GLSLstd450Determinant: {
+ struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_ssa);
+ val->ssa = rzalloc(b, struct vtn_ssa_value);
+ val->ssa->type = vtn_value(b, w[1], vtn_value_type_type)->type->type;
+ val->ssa->def = build_mat_det(b, vtn_ssa_value(b, w[5]));
+ break;
+ }
+
+ case GLSLstd450MatrixInverse: {
+ struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_ssa);
+ val->ssa = matrix_inverse(b, vtn_ssa_value(b, w[5]));
+ break;
+ }
+
+ case GLSLstd450InterpolateAtCentroid:
+ case GLSLstd450InterpolateAtSample:
+ case GLSLstd450InterpolateAtOffset:
+ unreachable("Unhandled opcode");
+
+ default:
+ handle_glsl450_alu(b, (enum GLSLstd450)ext_opcode, w, count);
+ }
+
+ return true;
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ * Authors:
+ * Jason Ekstrand (jason@jlekstrand.net)
+ *
+ */
+
+#include "nir/nir.h"
+#include "nir/nir_builder.h"
+#include "nir/nir_array.h"
+#include "nir_spirv.h"
+#include "spirv.h"
+
+struct vtn_builder;
+struct vtn_decoration;
+
+enum vtn_value_type {
+ vtn_value_type_invalid = 0,
+ vtn_value_type_undef,
+ vtn_value_type_string,
+ vtn_value_type_decoration_group,
+ vtn_value_type_type,
+ vtn_value_type_constant,
+ vtn_value_type_deref,
+ vtn_value_type_function,
+ vtn_value_type_block,
+ vtn_value_type_ssa,
+ vtn_value_type_extension,
+ vtn_value_type_image_pointer,
+ vtn_value_type_sampled_image,
+};
+
+enum vtn_branch_type {
+ vtn_branch_type_none,
+ vtn_branch_type_switch_break,
+ vtn_branch_type_switch_fallthrough,
+ vtn_branch_type_loop_break,
+ vtn_branch_type_loop_continue,
+ vtn_branch_type_discard,
+ vtn_branch_type_return,
+};
+
+enum vtn_cf_node_type {
+ vtn_cf_node_type_block,
+ vtn_cf_node_type_if,
+ vtn_cf_node_type_loop,
+ vtn_cf_node_type_switch,
+};
+
+struct vtn_cf_node {
+ struct list_head link;
+ enum vtn_cf_node_type type;
+};
+
+struct vtn_loop {
+ struct vtn_cf_node node;
+
+ /* The main body of the loop */
+ struct list_head body;
+
+ /* The "continue" part of the loop. This gets executed after the body
+ * and is where you go when you hit a continue.
+ */
+ struct list_head cont_body;
+
+ SpvLoopControlMask control;
+};
+
+struct vtn_if {
+ struct vtn_cf_node node;
+
+ uint32_t condition;
+
+ enum vtn_branch_type then_type;
+ struct list_head then_body;
+
+ enum vtn_branch_type else_type;
+ struct list_head else_body;
+
+ SpvSelectionControlMask control;
+};
+
+struct vtn_case {
+ struct list_head link;
+
+ struct list_head body;
+
+ /* The block that starts this case */
+ struct vtn_block *start_block;
+
+ /* The fallthrough case, if any */
+ struct vtn_case *fallthrough;
+
+ /* The uint32_t values that map to this case */
+ nir_array values;
+
+ /* True if this is the default case */
+ bool is_default;
+
+ /* Initialized to false; used when sorting the list of cases */
+ bool visited;
+};
+
+struct vtn_switch {
+ struct vtn_cf_node node;
+
+ uint32_t selector;
+
+ struct list_head cases;
+};
+
+struct vtn_block {
+ struct vtn_cf_node node;
+
+ /** A pointer to the label instruction */
+ const uint32_t *label;
+
+ /** A pointer to the merge instruction (or NULL if non exists) */
+ const uint32_t *merge;
+
+ /** A pointer to the branch instruction that ends this block */
+ const uint32_t *branch;
+
+ enum vtn_branch_type branch_type;
+
+ /** Points to the loop that this block starts (if it starts a loop) */
+ struct vtn_loop *loop;
+
+ /** Points to the switch case started by this block (if any) */
+ struct vtn_case *switch_case;
+
+ /** The last block in this SPIR-V block. */
+ nir_block *end_block;
+};
+
+struct vtn_function {
+ struct exec_node node;
+
+ nir_function_impl *impl;
+ struct vtn_block *start_block;
+
+ struct list_head body;
+
+ const uint32_t *end;
+
+ SpvFunctionControlMask control;
+};
+
+typedef bool (*vtn_instruction_handler)(struct vtn_builder *, uint32_t,
+ const uint32_t *, unsigned);
+
+void vtn_build_cfg(struct vtn_builder *b, const uint32_t *words,
+ const uint32_t *end);
+void vtn_function_emit(struct vtn_builder *b, struct vtn_function *func,
+ vtn_instruction_handler instruction_handler);
+
+const uint32_t *
+vtn_foreach_instruction(struct vtn_builder *b, const uint32_t *start,
+ const uint32_t *end, vtn_instruction_handler handler);
+
+struct vtn_ssa_value {
+ union {
+ nir_ssa_def *def;
+ struct vtn_ssa_value **elems;
+ };
+
+ /* For matrices, if this is non-NULL, then this value is actually the
+ * transpose of some other value. The value that `transposed` points to
+ * always dominates this value.
+ */
+ struct vtn_ssa_value *transposed;
+
+ const struct glsl_type *type;
+};
+
+struct vtn_type {
+ const struct glsl_type *type;
+
+ /* for matrices, whether the matrix is stored row-major */
+ bool row_major;
+
+ /* for structs, the offset of each member */
+ unsigned *offsets;
+
+ /* for structs, whether it was decorated as a "non-SSBO-like" block */
+ bool block;
+
+ /* for structs, whether it was decorated as an "SSBO-like" block */
+ bool buffer_block;
+
+ /* for structs with block == true, whether this is a builtin block (i.e. a
+ * block that contains only builtins).
+ */
+ bool builtin_block;
+
+ /* Image format for image_load_store type images */
+ unsigned image_format;
+
+ /* Access qualifier for storage images */
+ SpvAccessQualifier access_qualifier;
+
+ /* for arrays and matrices, the array stride */
+ unsigned stride;
+
+ /* for arrays, the vtn_type for the elements of the array */
+ struct vtn_type *array_element;
+
+ /* for structures, the vtn_type for each member */
+ struct vtn_type **members;
+
+ /* Whether this type, or a parent type, has been decorated as a builtin */
+ bool is_builtin;
+
+ SpvBuiltIn builtin;
+};
+
+struct vtn_image_pointer {
+ nir_deref_var *deref;
+ nir_ssa_def *coord;
+ nir_ssa_def *sample;
+};
+
+struct vtn_sampled_image {
+ nir_deref_var *image; /* Image or array of images */
+ nir_deref_var *sampler; /* Sampler */
+};
+
+struct vtn_value {
+ enum vtn_value_type value_type;
+ const char *name;
+ struct vtn_decoration *decoration;
+ union {
+ void *ptr;
+ char *str;
+ struct vtn_type *type;
+ struct {
+ nir_constant *constant;
+ const struct glsl_type *const_type;
+ };
+ struct {
+ nir_deref_var *deref;
+ struct vtn_type *deref_type;
+ };
+ struct vtn_image_pointer *image;
+ struct vtn_sampled_image *sampled_image;
+ struct vtn_function *func;
+ struct vtn_block *block;
+ struct vtn_ssa_value *ssa;
+ vtn_instruction_handler ext_handler;
+ };
+};
+
+#define VTN_DEC_DECORATION -1
+#define VTN_DEC_EXECUTION_MODE -2
+#define VTN_DEC_STRUCT_MEMBER0 0
+
+struct vtn_decoration {
+ struct vtn_decoration *next;
+
+ /* Specifies how to apply this decoration. Negative values represent a
+ * decoration or execution mode. (See the VTN_DEC_ #defines above.)
+ * Non-negative values specify that it applies to a structure member.
+ */
+ int scope;
+
+ const uint32_t *literals;
+ struct vtn_value *group;
+
+ union {
+ SpvDecoration decoration;
+ SpvExecutionMode exec_mode;
+ };
+};
+
+struct vtn_builder {
+ nir_builder nb;
+
+ nir_shader *shader;
+ nir_function_impl *impl;
+ struct vtn_block *block;
+
+ /* Current file, line, and column. Useful for debugging. Set
+ * automatically by vtn_foreach_instruction.
+ */
+ char *file;
+ int line, col;
+
+ /*
+ * In SPIR-V, constants are global, whereas in NIR, the load_const
+ * instruction we use is per-function. So while we parse each function, we
+ * keep a hash table of constants we've resolved to nir_ssa_value's so
+ * far, and we lazily resolve them when we see them used in a function.
+ */
+ struct hash_table *const_table;
+
+ /*
+ * Map from phi instructions (pointer to the start of the instruction)
+ * to the variable corresponding to it.
+ */
+ struct hash_table *phi_table;
+
+ unsigned num_specializations;
+ struct nir_spirv_specialization *specializations;
+
+ /*
+ * NIR variable for each SPIR-V builtin.
+ */
+ struct {
+ nir_variable *in;
+ nir_variable *out;
+ } builtins[42]; /* XXX need symbolic constant from SPIR-V header */
+
+ unsigned value_id_bound;
+ struct vtn_value *values;
+
+ gl_shader_stage entry_point_stage;
+ const char *entry_point_name;
+ struct vtn_value *entry_point;
+ bool origin_upper_left;
+
+ struct vtn_function *func;
+ struct exec_list functions;
+
+ /* Current function parameter index */
+ unsigned func_param_idx;
+
+ bool has_loop_continue;
+};
+
+static inline struct vtn_value *
+vtn_push_value(struct vtn_builder *b, uint32_t value_id,
+ enum vtn_value_type value_type)
+{
+ assert(value_id < b->value_id_bound);
+ assert(b->values[value_id].value_type == vtn_value_type_invalid);
+
+ b->values[value_id].value_type = value_type;
+
+ return &b->values[value_id];
+}
+
+static inline struct vtn_value *
+vtn_untyped_value(struct vtn_builder *b, uint32_t value_id)
+{
+ assert(value_id < b->value_id_bound);
+ return &b->values[value_id];
+}
+
+static inline struct vtn_value *
+vtn_value(struct vtn_builder *b, uint32_t value_id,
+ enum vtn_value_type value_type)
+{
+ struct vtn_value *val = vtn_untyped_value(b, value_id);
+ assert(val->value_type == value_type);
+ return val;
+}
+
+struct vtn_ssa_value *vtn_ssa_value(struct vtn_builder *b, uint32_t value_id);
+
+struct vtn_ssa_value *vtn_create_ssa_value(struct vtn_builder *b,
+ const struct glsl_type *type);
+
+struct vtn_ssa_value *vtn_ssa_transpose(struct vtn_builder *b,
+ struct vtn_ssa_value *src);
+
+struct vtn_ssa_value *
+vtn_variable_load(struct vtn_builder *b, nir_deref_var *src,
+ struct vtn_type *src_type);
+
+void vtn_variable_store(struct vtn_builder *b, struct vtn_ssa_value *src,
+ nir_deref_var *dest, struct vtn_type *dest_type);
+
+
+typedef void (*vtn_decoration_foreach_cb)(struct vtn_builder *,
+ struct vtn_value *,
+ int member,
+ const struct vtn_decoration *,
+ void *);
+
+void vtn_foreach_decoration(struct vtn_builder *b, struct vtn_value *value,
+ vtn_decoration_foreach_cb cb, void *data);
+
+typedef void (*vtn_execution_mode_foreach_cb)(struct vtn_builder *,
+ struct vtn_value *,
+ const struct vtn_decoration *,
+ void *);
+
+void vtn_foreach_execution_mode(struct vtn_builder *b, struct vtn_value *value,
+ vtn_execution_mode_foreach_cb cb, void *data);
+
+nir_op vtn_nir_alu_op_for_spirv_opcode(SpvOp opcode, bool *swap);
+
+void vtn_handle_alu(struct vtn_builder *b, SpvOp opcode,
+ const uint32_t *w, unsigned count);
+
+bool vtn_handle_glsl450_instruction(struct vtn_builder *b, uint32_t ext_opcode,
+ const uint32_t *words, unsigned count);
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ * Authors:
+ * Jason Ekstrand (jason@jlekstrand.net)
+ *
+ */
+
+/*
+ * A simple executable that opens a SPIR-V shader, converts it to NIR, and
+ * dumps out the result. This should be useful for testing the
+ * spirv_to_nir code.
+ */
+
+#include "spirv/nir_spirv.h"
+
+#include <sys/mman.h>
+#include <sys/types.h>
+#include <fcntl.h>
+#include <unistd.h>
+
+int main(int argc, char **argv)
+{
+ int fd = open(argv[1], O_RDONLY);
+ off_t len = lseek(fd, 0, SEEK_END);
+
+ assert(len % 4 == 0);
+ size_t word_count = len / 4;
+
+ const void *map = mmap(NULL, len, PROT_READ, MAP_PRIVATE, fd, 0);
+ assert(map != NULL);
+
+ nir_function *func = spirv_to_nir(map, word_count, NULL, 0,
+ MESA_SHADER_FRAGMENT, "main", NULL);
+ nir_print_shader(func->shader, stderr);
+}
~nir_cf_test();
nir_builder b;
- nir_shader *shader;
- nir_function_impl *impl;
};
nir_cf_test::nir_cf_test()
{
static const nir_shader_compiler_options options = { };
- shader = nir_shader_create(NULL, MESA_SHADER_VERTEX, &options);
- nir_function *func = nir_function_create(shader, "main");
- nir_function_overload *overload = nir_function_overload_create(func);
- impl = nir_function_impl_create(overload);
-
- nir_builder_init(&b, impl);
+ nir_builder_init_simple_shader(&b, NULL, MESA_SHADER_VERTEX, &options);
}
nir_cf_test::~nir_cf_test()
{
- ralloc_free(shader);
+ ralloc_free(b.shader);
}
TEST_F(nir_cf_test, delete_break_in_loop)
*
* while (...) { break; }
*/
- nir_loop *loop = nir_loop_create(shader);
- nir_cf_node_insert(nir_after_cf_list(&impl->body), &loop->cf_node);
+ nir_loop *loop = nir_loop_create(b.shader);
+ nir_cf_node_insert(nir_after_cf_list(&b.impl->body), &loop->cf_node);
b.cursor = nir_after_cf_list(&loop->body);
- nir_jump_instr *jump = nir_jump_instr_create(shader, nir_jump_break);
+ nir_jump_instr *jump = nir_jump_instr_create(b.shader, nir_jump_break);
nir_builder_instr_insert(&b, &jump->instr);
/* At this point, we should have:
* block block_3:
* }
*/
- nir_block *block_0 = nir_start_block(impl);
+ nir_block *block_0 = nir_start_block(b.impl);
nir_block *block_1 = nir_cf_node_as_block(nir_loop_first_cf_node(loop));
nir_block *block_2 = nir_cf_node_as_block(nir_cf_node_next(&loop->cf_node));
- nir_block *block_3 = impl->end_block;
+ nir_block *block_3 = b.impl->end_block;
ASSERT_EQ(nir_cf_node_block, block_0->cf_node.type);
ASSERT_EQ(nir_cf_node_block, block_1->cf_node.type);
ASSERT_EQ(nir_cf_node_block, block_2->cf_node.type);
EXPECT_TRUE(_mesa_set_search(block_2->predecessors, block_1));
EXPECT_TRUE(_mesa_set_search(block_3->predecessors, block_2));
- nir_print_shader(shader, stderr);
+ nir_print_shader(b.shader, stderr);
/* Now remove the break. */
nir_instr_remove(&jump->instr);
- nir_print_shader(shader, stderr);
+ nir_print_shader(b.shader, stderr);
/* At this point, we should have:
*
EXPECT_TRUE(_mesa_set_search(block_2->predecessors, block_1));
EXPECT_TRUE(_mesa_set_search(block_3->predecessors, block_2));
- nir_metadata_require(impl, nir_metadata_dominance);
+ nir_metadata_require(b.impl, nir_metadata_dominance);
}
#include "util/ralloc.h"
#include "util/strtod.h"
+extern "C" void
+_mesa_error_no_memory(const char *caller)
+{
+ fprintf(stderr, "Mesa error: out of memory in %s", caller);
+}
+
void
_mesa_warning(struct gl_context *ctx, const char *fmt, ...)
{
shProg->InterfaceBlockStageIndex[i] = NULL;
}
+ ralloc_free(shProg->UboInterfaceBlockIndex);
+ shProg->UboInterfaceBlockIndex = NULL;
+ ralloc_free(shProg->SsboInterfaceBlockIndex);
+ shProg->SsboInterfaceBlockIndex = NULL;
+
ralloc_free(shProg->AtomicBuffers);
shProg->AtomicBuffers = NULL;
shProg->NumAtomicBuffers = 0;
ctx->Extensions.ARB_gpu_shader_fp64 = true;
ctx->Extensions.ARB_sample_shading = true;
ctx->Extensions.ARB_shader_bit_encoding = true;
+ ctx->Extensions.ARB_shader_draw_parameters = true;
ctx->Extensions.ARB_shader_stencil_export = true;
ctx->Extensions.ARB_shader_subroutine = true;
ctx->Extensions.ARB_shader_texture_lod = true;
struct dri3_drawable *priv = (struct dri3_drawable *)
GetGLXDRIDrawable(gc->currentDpy, gc->currentDrawable);
- loader_dri3_wait_x(&priv->loader_drawable);
+ if (priv)
+ loader_dri3_wait_x(&priv->loader_drawable);
}
static void
struct dri3_drawable *priv = (struct dri3_drawable *)
GetGLXDRIDrawable(gc->currentDpy, gc->currentDrawable);
- loader_dri3_wait_gl(&priv->loader_drawable);
+ if (priv)
+ loader_dri3_wait_gl(&priv->loader_drawable);
}
/**
--- /dev/null
+/isl_format_layout.c
--- /dev/null
+# Copyright 2015 Intel Corporation
+#
+# Permission is hereby granted, free of charge, to any person obtaining a
+# copy of this software and associated documentation files (the "Software"),
+# to deal in the Software without restriction, including without limitation
+# the rights to use, copy, modify, merge, publish, distribute, sublicense,
+# and/or sell copies of the Software, and to permit persons to whom the
+# Software is furnished to do so, subject to the following conditions:
+#
+# The above copyright notice and this permission notice (including the next
+# paragraph) shall be included in all copies or substantial portions of the
+# Software.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+# THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+# IN THE SOFTWARE.
+
+SUBDIRS = .
+
+noinst_LTLIBRARIES = libisl.la
+
+EXTRA_DIST = tests
+
+# The gallium includes are for the util/u_math.h include from main/macros.h
+AM_CPPFLAGS = \
+ $(INTEL_CFLAGS) \
+ $(VALGRIND_CFLAGS) \
+ $(DEFINES) \
+ -I$(top_srcdir)/include \
+ -I$(top_srcdir)/src \
+ -I$(top_srcdir)/src/mapi \
+ -I$(top_srcdir)/src/mesa \
+ -I$(top_srcdir)/src/mesa/drivers/dri/common \
+ -I$(top_srcdir)/src/mesa/drivers/dri/i965 \
+ -I$(top_srcdir)/src/gallium/auxiliary \
+ -I$(top_srcdir)/src/gallium/include \
+ -I$(top_builddir)/src
+
+libisl_la_CFLAGS = $(CFLAGS) -Wno-override-init
+
+libisl_la_SOURCES = \
+ isl.c \
+ isl.h \
+ isl_format.c \
+ isl_format_layout.c \
+ isl_gen4.c \
+ isl_gen4.h \
+ isl_gen6.c \
+ isl_gen6.h \
+ isl_gen7.c \
+ isl_gen7.h \
+ isl_gen8.c \
+ isl_gen8.h \
+ isl_gen9.c \
+ isl_gen9.h \
+ isl_image.c \
+ $(NULL)
+
+BUILT_SOURCES = \
+ isl_format_layout.c
+
+isl_format_layout.c: isl_format_layout_gen.bash \
+ isl_format_layout.csv
+ $(AM_V_GEN)$(srcdir)/isl_format_layout_gen.bash \
+ <$(srcdir)/isl_format_layout.csv >$@
+
+# ----------------------------------------------------------------------------
+# Tests
+# ----------------------------------------------------------------------------
+
+TESTS = tests/isl_surf_get_image_offset_test
+
+check_PROGRAMS = $(TESTS)
+
+# Link tests to lib965_compiler.la for brw_get_device_info().
+tests_ldadd = \
+ libisl.la \
+ $(top_builddir)/src/mesa/drivers/dri/i965/libi965_compiler.la
+
+tests_isl_surf_get_image_offset_test_SOURCES = \
+ tests/isl_surf_get_image_offset_test.c
+tests_isl_surf_get_image_offset_test_LDADD = $(tests_ldadd)
+
+# ----------------------------------------------------------------------------
+
+include $(top_srcdir)/install-lib-links.mk
--- /dev/null
+Intel Surface Layout
+
+Introduction
+============
+isl is a small library that calculates the layout of Intel GPU surfaces, queries
+those layouts, and queries the properties of surface formats.
+
+
+Independence from User APIs
+===========================
+isl's API is independent of any user-facing graphics API, such as OpenGL and
+Vulkan. This independence allows isl to be used a shared component by multiple
+Intel drivers.
+
+Rather than mimic the user-facing APIs, the isl API attempts to reflect Intel
+hardware: the actual memory layout of Intel GPU surfaces and how one programs
+the GPU to use those surfaces. For example:
+
+ - The tokens of `enum isl_format` (such as `ISL_FORMAT_R8G8B8A8_UNORM`)
+ match those of the hardware enum `SURFACE_FORMAT` rather than the OpenGL
+ or Vulkan format tokens. And the values of `isl_format` and
+ `SURFACE_FORMAT` are identical.
+
+ - The OpenGL and Vulkan APIs contain depth and stencil formats. However the
+ hardware enum `SURFACE_FORMAT` does not, and therefore neither does `enum
+ isl_format`. Rather than define new pixel formats that have no hardware
+ counterpart, isl records the intent to use a surface as a depth or stencil
+ buffer with the usage flags `ISL_SURF_USAGE_DEPTH_BIT` and
+ `ISL_SURF_USAGE_STENCIL_BIT`.
+
+ - `struct isl_surf` distinguishes between the surface's logical dimension
+ from the user API's perspective (`enum isl_surf_dim`, which may be 1D, 2D,
+ or 3D) and the layout of those dimensions in memory (`enum isl_dim_layout`).
+
+
+Surface Units
+=============
+
+Intro
+-----
+ISL takes care in its equations to correctly handle conversion among surface
+units (such as pixels and compression blocks) and to carefully distinguish
+between a surface's logical layout in the client API and its physical layout
+in memory.
+
+Symbol names often explicitly declare their unit with a suffix:
+
+ - px: logical pixels
+ - sa: physical surface samples
+ - el: physical surface elements
+ - sa_rows: rows of physical surface samples
+ - el_rows: rows of physical surface elements
+
+Logical units are independent of hardware generation and are closely related
+to the user-facing API (OpenGL and Vulkan). Physical units are dependent on
+hardware generation and reflect the surface's layout in memory.
+
+Definitions
+-----------
+- Logical Pixels (px):
+
+ The surface's layout from the perspective of the client API (OpenGL and
+ Vulkan) is in units of logical pixels. Logical pixels are independent of the
+ surface's layout in memory.
+
+ A surface's width and height, in units of logical pixels, is not affected by
+ the surface's sample count. For example, consider a VkImage created with
+ VkImageCreateInfo{width=w0, height=h0, samples=s0}. The surface's width and
+ height at level 0 is, in units of logical pixels, w0 and h0 regardless of
+ the value of s0.
+
+ For example, the logical array length of a 3D surface is always 1, even on
+ Gen9 where the surface's memory layout is that of an array surface
+ (ISL_DIM_LAYOUT_GEN4_2D).
+
+- Physical Surface Samples (sa):
+
+ For a multisampled surface, this unit has the obvious meaning.
+ A singlesampled surface, from ISL's perspective, is simply a multisampled
+ surface whose sample count is 1.
+
+ For example, consider a 2D single-level non-array surface with samples=4,
+ width_px=64, and height_px=64 (note that the suffix 'px' indicates logical
+ pixels). If the surface's multisample layout is ISL_MSAA_LAYOUT_INTERLEAVED,
+ then the extent of level 0 is, in units of physical surface samples,
+ width_sa=128, height_sa=128, depth_sa=1, array_length_sa=1. If
+ ISL_MSAA_LAYOUT_ARRAY, then width_sa=64, height_sa=64, depth_sa=1,
+ array_length_sa=4.
+
+- Physical Surface Elements (el):
+
+ This unit allows ISL to treat compressed and uncompressed formats
+ identically in many calculations.
+
+ If the surface's pixel format is compressed, such as ETC2, then a surface
+ element is equivalent to a compression block. If uncompressed, then
+ a surface element is equivalent to a surface sample. As a corollary, for
+ a given surface a surface element is at least as large as a surface sample.
+
+Errata
+------
+ISL acquired the term 'surface element' from the Broadwell PRM [1], which
+defines it as follows:
+
+ An element is defined as a pixel in uncompresed surface formats, and as
+ a compression block in compressed surface formats. For MSFMT_DEPTH_STENCIL
+ type multisampled surfaces, an element is a sample.
+
+
+References
+==========
+[1]: Broadwell PRM >> Volume 2d: Command Reference: Structures >>
+ RENDER_SURFACE_STATE Surface Vertical Alignment (p325)
--- /dev/null
+/*
+ * Copyright 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <assert.h>
+
+#include "isl.h"
+#include "isl_gen4.h"
+#include "isl_gen6.h"
+#include "isl_gen7.h"
+#include "isl_gen8.h"
+#include "isl_gen9.h"
+#include "isl_priv.h"
+
+void PRINTFLIKE(3, 4) UNUSED
+__isl_finishme(const char *file, int line, const char *fmt, ...)
+{
+ va_list ap;
+ char buf[512];
+
+ va_start(ap, fmt);
+ vsnprintf(buf, sizeof(buf), fmt, ap);
+ va_end(ap);
+
+ fprintf(stderr, "%s:%d: FINISHME: %s\n", file, line, buf);
+}
+
+void
+isl_device_init(struct isl_device *dev,
+ const struct brw_device_info *info,
+ bool has_bit6_swizzling)
+{
+ dev->info = info;
+ dev->use_separate_stencil = ISL_DEV_GEN(dev) >= 6;
+ dev->has_bit6_swizzling = has_bit6_swizzling;
+
+ /* The ISL_DEV macros may be defined in the CFLAGS, thus hardcoding some
+ * device properties at buildtime. Verify that the macros with the device
+ * properties chosen during runtime.
+ */
+ assert(ISL_DEV_GEN(dev) == dev->info->gen);
+ assert(ISL_DEV_USE_SEPARATE_STENCIL(dev) == dev->use_separate_stencil);
+
+ /* Did we break hiz or stencil? */
+ if (ISL_DEV_USE_SEPARATE_STENCIL(dev))
+ assert(info->has_hiz_and_separate_stencil);
+ if (info->must_use_separate_stencil)
+ assert(ISL_DEV_USE_SEPARATE_STENCIL(dev));
+}
+
+/**
+ * @param[out] info is written only on success
+ */
+bool
+isl_tiling_get_info(const struct isl_device *dev,
+ enum isl_tiling tiling,
+ uint32_t format_block_size,
+ struct isl_tile_info *tile_info)
+{
+ const uint32_t bs = format_block_size;
+ uint32_t width, height;
+
+ assert(bs > 0);
+
+ switch (tiling) {
+ case ISL_TILING_LINEAR:
+ width = 1;
+ height = 1;
+ break;
+
+ case ISL_TILING_X:
+ width = 1 << 9;
+ height = 1 << 3;
+ break;
+
+ case ISL_TILING_Y0:
+ width = 1 << 7;
+ height = 1 << 5;
+ break;
+
+ case ISL_TILING_W:
+ /* XXX: Should W tile be same as Y? */
+ width = 1 << 6;
+ height = 1 << 6;
+ break;
+
+ case ISL_TILING_Yf:
+ case ISL_TILING_Ys: {
+ if (ISL_DEV_GEN(dev) < 9)
+ return false;
+
+ if (!isl_is_pow2(bs))
+ return false;
+
+ bool is_Ys = tiling == ISL_TILING_Ys;
+
+ width = 1 << (6 + (ffs(bs) / 2) + (2 * is_Ys));
+ height = 1 << (6 - (ffs(bs) / 2) + (2 * is_Ys));
+ break;
+ }
+ } /* end switch */
+
+ *tile_info = (struct isl_tile_info) {
+ .tiling = tiling,
+ .width = width,
+ .height = height,
+ .size = width * height,
+ };
+
+ return true;
+}
+
+void
+isl_tiling_get_extent(const struct isl_device *dev,
+ enum isl_tiling tiling,
+ uint32_t format_block_size,
+ struct isl_extent2d *e)
+{
+ struct isl_tile_info tile_info;
+ isl_tiling_get_info(dev, tiling, format_block_size, &tile_info);
+ *e = isl_extent2d(tile_info.width, tile_info.height);
+}
+
+/**
+ * @param[out] tiling is set only on success
+ */
+bool
+isl_surf_choose_tiling(const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info,
+ enum isl_tiling *tiling)
+{
+ isl_tiling_flags_t tiling_flags = info->tiling_flags;
+
+ if (ISL_DEV_GEN(dev) >= 7) {
+ gen7_filter_tiling(dev, info, &tiling_flags);
+ } else {
+ isl_finishme("%s: gen%u", __func__, ISL_DEV_GEN(dev));
+ gen7_filter_tiling(dev, info, &tiling_flags);
+ }
+
+ #define CHOOSE(__tiling) \
+ do { \
+ if (tiling_flags & (1u << (__tiling))) { \
+ *tiling = (__tiling); \
+ return true; \
+ } \
+ } while (0)
+
+ /* Of the tiling modes remaining, choose the one that offers the best
+ * performance.
+ */
+
+ if (info->dim == ISL_SURF_DIM_1D) {
+ /* Prefer linear for 1D surfaces because they do not benefit from
+ * tiling. To the contrary, tiling leads to wasted memory and poor
+ * memory locality due to the swizzling and alignment restrictions
+ * required in tiled surfaces.
+ */
+ CHOOSE(ISL_TILING_LINEAR);
+ }
+
+ CHOOSE(ISL_TILING_Ys);
+ CHOOSE(ISL_TILING_Yf);
+ CHOOSE(ISL_TILING_Y0);
+ CHOOSE(ISL_TILING_X);
+ CHOOSE(ISL_TILING_W);
+ CHOOSE(ISL_TILING_LINEAR);
+
+ #undef CHOOSE
+
+ /* No tiling mode accomodates the inputs. */
+ return false;
+}
+
+static bool
+isl_choose_msaa_layout(const struct isl_device *dev,
+ const struct isl_surf_init_info *info,
+ enum isl_tiling tiling,
+ enum isl_msaa_layout *msaa_layout)
+{
+ if (ISL_DEV_GEN(dev) >= 8) {
+ return gen8_choose_msaa_layout(dev, info, tiling, msaa_layout);
+ } else if (ISL_DEV_GEN(dev) >= 7) {
+ return gen7_choose_msaa_layout(dev, info, tiling, msaa_layout);
+ } else if (ISL_DEV_GEN(dev) >= 6) {
+ return gen6_choose_msaa_layout(dev, info, tiling, msaa_layout);
+ } else {
+ return gen4_choose_msaa_layout(dev, info, tiling, msaa_layout);
+ }
+}
+
+static void
+isl_msaa_interleaved_scale_px_to_sa(uint32_t samples,
+ uint32_t *width, uint32_t *height)
+{
+ assert(isl_is_pow2(samples));
+
+ /* From the Broadwell PRM >> Volume 5: Memory Views >> Computing Mip Level
+ * Sizes (p133):
+ *
+ * If the surface is multisampled and it is a depth or stencil surface
+ * or Multisampled Surface StorageFormat in SURFACE_STATE is
+ * MSFMT_DEPTH_STENCIL, W_L and H_L must be adjusted as follows before
+ * proceeding: [...]
+ */
+ if (width)
+ *width = isl_align(*width, 2) << ((ffs(samples) - 0) / 2);
+ if (height)
+ *height = isl_align(*height, 2) << ((ffs(samples) - 1) / 2);
+}
+
+static enum isl_array_pitch_span
+isl_choose_array_pitch_span(const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info,
+ enum isl_dim_layout dim_layout,
+ const struct isl_extent4d *phys_level0_sa)
+{
+ switch (dim_layout) {
+ case ISL_DIM_LAYOUT_GEN9_1D:
+ case ISL_DIM_LAYOUT_GEN4_2D:
+ if (ISL_DEV_GEN(dev) >= 8) {
+ /* QPitch becomes programmable in Broadwell. So choose the
+ * most compact QPitch possible in order to conserve memory.
+ *
+ * From the Broadwell PRM >> Volume 2d: Command Reference: Structures
+ * >> RENDER_SURFACE_STATE Surface QPitch (p325):
+ *
+ * - Software must ensure that this field is set to a value
+ * sufficiently large such that the array slices in the surface
+ * do not overlap. Refer to the Memory Data Formats section for
+ * information on how surfaces are stored in memory.
+ *
+ * - This field specifies the distance in rows between array
+ * slices. It is used only in the following cases:
+ *
+ * - Surface Array is enabled OR
+ * - Number of Mulitsamples is not NUMSAMPLES_1 and
+ * Multisampled Surface Storage Format set to MSFMT_MSS OR
+ * - Surface Type is SURFTYPE_CUBE
+ */
+ return ISL_ARRAY_PITCH_SPAN_COMPACT;
+ } else if (ISL_DEV_GEN(dev) >= 7) {
+ /* Note that Ivybridge introduces
+ * RENDER_SURFACE_STATE.SurfaceArraySpacing, which provides the
+ * driver more control over the QPitch.
+ */
+
+ if (phys_level0_sa->array_len == 1) {
+ /* The hardware will never use the QPitch. So choose the most
+ * compact QPitch possible in order to conserve memory.
+ */
+ return ISL_ARRAY_PITCH_SPAN_COMPACT;
+ }
+
+ if (isl_surf_usage_is_depth_or_stencil(info->usage)) {
+ /* From the Ivybridge PRM >> Volume 1 Part 1: Graphics Core >>
+ * Section 6.18.4.7: Surface Arrays (p112):
+ *
+ * If Surface Array Spacing is set to ARYSPC_FULL (note that
+ * the depth buffer and stencil buffer have an implied value of
+ * ARYSPC_FULL):
+ */
+ return ISL_ARRAY_PITCH_SPAN_COMPACT;
+ }
+
+ if (info->levels == 1) {
+ /* We are able to set RENDER_SURFACE_STATE.SurfaceArraySpacing
+ * to ARYSPC_LOD0.
+ */
+ return ISL_ARRAY_PITCH_SPAN_COMPACT;
+ }
+
+ return ISL_ARRAY_PITCH_SPAN_FULL;
+ } else if ((ISL_DEV_GEN(dev) == 5 || ISL_DEV_GEN(dev) == 6) &&
+ ISL_DEV_USE_SEPARATE_STENCIL(dev) &&
+ isl_surf_usage_is_stencil(info->usage)) {
+ /* [ILK-SNB] Errata from the Sandy Bridge PRM >> Volume 4 Part 1:
+ * Graphics Core >> Section 7.18.3.7: Surface Arrays:
+ *
+ * The separate stencil buffer does not support mip mapping, thus
+ * the storage for LODs other than LOD 0 is not needed.
+ */
+ assert(info->levels == 1);
+ assert(phys_level0_sa->array_len == 1);
+ return ISL_ARRAY_PITCH_SPAN_COMPACT;
+ } else {
+ if ((ISL_DEV_GEN(dev) == 5 || ISL_DEV_GEN(dev) == 6) &&
+ ISL_DEV_USE_SEPARATE_STENCIL(dev) &&
+ isl_surf_usage_is_stencil(info->usage)) {
+ /* [ILK-SNB] Errata from the Sandy Bridge PRM >> Volume 4 Part 1:
+ * Graphics Core >> Section 7.18.3.7: Surface Arrays:
+ *
+ * The separate stencil buffer does not support mip mapping,
+ * thus the storage for LODs other than LOD 0 is not needed.
+ */
+ assert(info->levels == 1);
+ assert(phys_level0_sa->array_len == 1);
+ return ISL_ARRAY_PITCH_SPAN_COMPACT;
+ }
+
+ if (phys_level0_sa->array_len == 1) {
+ /* The hardware will never use the QPitch. So choose the most
+ * compact QPitch possible in order to conserve memory.
+ */
+ return ISL_ARRAY_PITCH_SPAN_COMPACT;
+ }
+
+ return ISL_ARRAY_PITCH_SPAN_FULL;
+ }
+
+ case ISL_DIM_LAYOUT_GEN4_3D:
+ /* The hardware will never use the QPitch. So choose the most
+ * compact QPitch possible in order to conserve memory.
+ */
+ return ISL_ARRAY_PITCH_SPAN_COMPACT;
+ }
+
+ unreachable("bad isl_dim_layout");
+ return ISL_ARRAY_PITCH_SPAN_FULL;
+}
+
+static void
+isl_choose_image_alignment_el(const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info,
+ enum isl_tiling tiling,
+ enum isl_msaa_layout msaa_layout,
+ struct isl_extent3d *image_align_el)
+{
+ if (ISL_DEV_GEN(dev) >= 9) {
+ gen9_choose_image_alignment_el(dev, info, tiling, msaa_layout,
+ image_align_el);
+ } else if (ISL_DEV_GEN(dev) >= 8) {
+ gen8_choose_image_alignment_el(dev, info, tiling, msaa_layout,
+ image_align_el);
+ } else if (ISL_DEV_GEN(dev) >= 7) {
+ gen7_choose_image_alignment_el(dev, info, tiling, msaa_layout,
+ image_align_el);
+ } else if (ISL_DEV_GEN(dev) >= 6) {
+ gen6_choose_image_alignment_el(dev, info, tiling, msaa_layout,
+ image_align_el);
+ } else {
+ gen4_choose_image_alignment_el(dev, info, tiling, msaa_layout,
+ image_align_el);
+ }
+}
+
+static enum isl_dim_layout
+isl_surf_choose_dim_layout(const struct isl_device *dev,
+ enum isl_surf_dim logical_dim)
+{
+ if (ISL_DEV_GEN(dev) >= 9) {
+ switch (logical_dim) {
+ case ISL_SURF_DIM_1D:
+ return ISL_DIM_LAYOUT_GEN9_1D;
+ case ISL_SURF_DIM_2D:
+ case ISL_SURF_DIM_3D:
+ return ISL_DIM_LAYOUT_GEN4_2D;
+ }
+ } else {
+ switch (logical_dim) {
+ case ISL_SURF_DIM_1D:
+ case ISL_SURF_DIM_2D:
+ return ISL_DIM_LAYOUT_GEN4_2D;
+ case ISL_SURF_DIM_3D:
+ return ISL_DIM_LAYOUT_GEN4_3D;
+ }
+ }
+
+ unreachable("bad isl_surf_dim");
+ return ISL_DIM_LAYOUT_GEN4_2D;
+}
+
+/**
+ * Calculate the physical extent of the surface's first level, in units of
+ * surface samples. The result is aligned to the format's compression block.
+ */
+static void
+isl_calc_phys_level0_extent_sa(const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info,
+ enum isl_dim_layout dim_layout,
+ enum isl_tiling tiling,
+ enum isl_msaa_layout msaa_layout,
+ struct isl_extent4d *phys_level0_sa)
+{
+ const struct isl_format_layout *fmtl = isl_format_get_layout(info->format);
+
+ if (isl_format_is_yuv(info->format))
+ isl_finishme("%s:%s: YUV format", __FILE__, __func__);
+
+ switch (info->dim) {
+ case ISL_SURF_DIM_1D:
+ assert(info->height == 1);
+ assert(info->depth == 1);
+ assert(info->samples == 1);
+ assert(!isl_format_is_compressed(info->format));
+
+ switch (dim_layout) {
+ case ISL_DIM_LAYOUT_GEN4_3D:
+ unreachable("bad isl_dim_layout");
+
+ case ISL_DIM_LAYOUT_GEN9_1D:
+ case ISL_DIM_LAYOUT_GEN4_2D:
+ *phys_level0_sa = (struct isl_extent4d) {
+ .w = info->width,
+ .h = 1,
+ .d = 1,
+ .a = info->array_len,
+ };
+ break;
+ }
+ break;
+
+ case ISL_SURF_DIM_2D:
+ assert(dim_layout == ISL_DIM_LAYOUT_GEN4_2D);
+
+ if (tiling == ISL_TILING_Ys && info->samples > 1)
+ isl_finishme("%s:%s: multisample TileYs layout", __FILE__, __func__);
+
+ switch (msaa_layout) {
+ case ISL_MSAA_LAYOUT_NONE:
+ assert(info->depth == 1);
+ assert(info->samples == 1);
+
+ *phys_level0_sa = (struct isl_extent4d) {
+ .w = isl_align(info->width, fmtl->bw),
+ .h = isl_align(info->height, fmtl->bh),
+ .d = 1,
+ .a = info->array_len,
+ };
+ break;
+
+ case ISL_MSAA_LAYOUT_ARRAY:
+ assert(info->depth == 1);
+ assert(info->array_len == 1);
+ assert(!isl_format_is_compressed(info->format));
+
+ *phys_level0_sa = (struct isl_extent4d) {
+ .w = info->width,
+ .h = info->height,
+ .d = 1,
+ .a = info->samples,
+ };
+ break;
+
+ case ISL_MSAA_LAYOUT_INTERLEAVED:
+ assert(info->depth == 1);
+ assert(info->array_len == 1);
+ assert(!isl_format_is_compressed(info->format));
+
+ *phys_level0_sa = (struct isl_extent4d) {
+ .w = info->width,
+ .h = info->height,
+ .d = 1,
+ .a = 1,
+ };
+
+ isl_msaa_interleaved_scale_px_to_sa(info->samples,
+ &phys_level0_sa->w,
+ &phys_level0_sa->h);
+ break;
+ }
+ break;
+
+ case ISL_SURF_DIM_3D:
+ assert(info->array_len == 1);
+ assert(info->samples == 1);
+
+ if (fmtl->bd > 1) {
+ isl_finishme("%s:%s: compression block with depth > 1",
+ __FILE__, __func__);
+ }
+
+ switch (dim_layout) {
+ case ISL_DIM_LAYOUT_GEN9_1D:
+ unreachable("bad isl_dim_layout");
+
+ case ISL_DIM_LAYOUT_GEN4_2D:
+ assert(ISL_DEV_GEN(dev) >= 9);
+
+ *phys_level0_sa = (struct isl_extent4d) {
+ .w = isl_align(info->width, fmtl->bw),
+ .h = isl_align(info->height, fmtl->bh),
+ .d = 1,
+ .a = info->depth,
+ };
+ break;
+
+ case ISL_DIM_LAYOUT_GEN4_3D:
+ assert(ISL_DEV_GEN(dev) < 9);
+ *phys_level0_sa = (struct isl_extent4d) {
+ .w = isl_align(info->width, fmtl->bw),
+ .h = isl_align(info->height, fmtl->bh),
+ .d = info->depth,
+ .a = 1,
+ };
+ break;
+ }
+ break;
+ }
+}
+
+/**
+ * A variant of isl_calc_phys_slice0_extent_sa() specific to
+ * ISL_DIM_LAYOUT_GEN4_2D.
+ */
+static void
+isl_calc_phys_slice0_extent_sa_gen4_2d(
+ const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info,
+ enum isl_msaa_layout msaa_layout,
+ const struct isl_extent3d *image_align_sa,
+ const struct isl_extent4d *phys_level0_sa,
+ struct isl_extent2d *phys_slice0_sa)
+{
+ const struct isl_format_layout *fmtl = isl_format_get_layout(info->format);
+
+ assert(phys_level0_sa->depth == 1);
+
+ if (info->levels == 1 && msaa_layout != ISL_MSAA_LAYOUT_INTERLEAVED) {
+ /* Do not pad the surface to the image alignment. Instead, pad it only
+ * to the pixel format's block alignment.
+ *
+ * For tiled surfaces, using a reduced alignment here avoids wasting CPU
+ * cycles on the below mipmap layout caluclations. Reducing the
+ * alignment here is safe because we later align the row pitch and array
+ * pitch to the tile boundary. It is safe even for
+ * ISL_MSAA_LAYOUT_INTERLEAVED, because phys_level0_sa is already scaled
+ * to accomodate the interleaved samples.
+ *
+ * For linear surfaces, reducing the alignment here permits us to later
+ * choose an arbitrary, non-aligned row pitch. If the surface backs
+ * a VkBuffer, then an arbitrary pitch may be needed to accomodate
+ * VkBufferImageCopy::bufferRowLength.
+ */
+ *phys_slice0_sa = (struct isl_extent2d) {
+ .w = isl_align_npot(phys_level0_sa->w, fmtl->bw),
+ .h = isl_align_npot(phys_level0_sa->h, fmtl->bh),
+ };
+ return;
+ }
+
+ uint32_t slice_top_w = 0;
+ uint32_t slice_bottom_w = 0;
+ uint32_t slice_left_h = 0;
+ uint32_t slice_right_h = 0;
+
+ uint32_t W0 = phys_level0_sa->w;
+ uint32_t H0 = phys_level0_sa->h;
+
+ for (uint32_t l = 0; l < info->levels; ++l) {
+ uint32_t W = isl_minify(W0, l);
+ uint32_t H = isl_minify(H0, l);
+
+ if (msaa_layout == ISL_MSAA_LAYOUT_INTERLEAVED) {
+ /* From the Broadwell PRM >> Volume 5: Memory Views >> Computing Mip Level
+ * Sizes (p133):
+ *
+ * If the surface is multisampled and it is a depth or stencil
+ * surface or Multisampled Surface StorageFormat in
+ * SURFACE_STATE is MSFMT_DEPTH_STENCIL, W_L and H_L must be
+ * adjusted as follows before proceeding: [...]
+ */
+ isl_msaa_interleaved_scale_px_to_sa(info->samples, &W, &H);
+ }
+
+ uint32_t w = isl_align_npot(W, image_align_sa->w);
+ uint32_t h = isl_align_npot(H, image_align_sa->h);
+
+ if (l == 0) {
+ slice_top_w = w;
+ slice_left_h = h;
+ slice_right_h = h;
+ } else if (l == 1) {
+ slice_bottom_w = w;
+ slice_left_h += h;
+ } else if (l == 2) {
+ slice_bottom_w += w;
+ slice_right_h += h;
+ } else {
+ slice_right_h += h;
+ }
+ }
+
+ *phys_slice0_sa = (struct isl_extent2d) {
+ .w = MAX(slice_top_w, slice_bottom_w),
+ .h = MAX(slice_left_h, slice_right_h),
+ };
+}
+
+/**
+ * A variant of isl_calc_phys_slice0_extent_sa() specific to
+ * ISL_DIM_LAYOUT_GEN4_3D.
+ */
+static void
+isl_calc_phys_slice0_extent_sa_gen4_3d(
+ const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info,
+ const struct isl_extent3d *image_align_sa,
+ const struct isl_extent4d *phys_level0_sa,
+ struct isl_extent2d *phys_slice0_sa)
+{
+ assert(info->samples == 1);
+ assert(phys_level0_sa->array_len == 1);
+
+ uint32_t slice_w = 0;
+ uint32_t slice_h = 0;
+
+ uint32_t W0 = phys_level0_sa->w;
+ uint32_t H0 = phys_level0_sa->h;
+ uint32_t D0 = phys_level0_sa->d;
+
+ for (uint32_t l = 0; l < info->levels; ++l) {
+ uint32_t level_w = isl_align_npot(isl_minify(W0, l), image_align_sa->w);
+ uint32_t level_h = isl_align_npot(isl_minify(H0, l), image_align_sa->h);
+ uint32_t level_d = isl_align_npot(isl_minify(D0, l), image_align_sa->d);
+
+ uint32_t max_layers_horiz = MIN(level_d, 1u << l);
+ uint32_t max_layers_vert = isl_align(level_d, 1u << l) / (1u << l);
+
+ slice_w = MAX(slice_w, level_w * max_layers_horiz);
+ slice_h += level_h * max_layers_vert;
+ }
+
+ *phys_slice0_sa = (struct isl_extent2d) {
+ .w = slice_w,
+ .h = slice_h,
+ };
+}
+
+/**
+ * A variant of isl_calc_phys_slice0_extent_sa() specific to
+ * ISL_DIM_LAYOUT_GEN9_1D.
+ */
+static void
+isl_calc_phys_slice0_extent_sa_gen9_1d(
+ const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info,
+ const struct isl_extent3d *image_align_sa,
+ const struct isl_extent4d *phys_level0_sa,
+ struct isl_extent2d *phys_slice0_sa)
+{
+ const struct isl_format_layout *fmtl = isl_format_get_layout(info->format);
+
+ assert(phys_level0_sa->height == 1);
+ assert(phys_level0_sa->depth == 1);
+ assert(info->samples == 1);
+ assert(image_align_sa->w >= fmtl->bw);
+
+ uint32_t slice_w = 0;
+ const uint32_t W0 = phys_level0_sa->w;
+
+ for (uint32_t l = 0; l < info->levels; ++l) {
+ uint32_t W = isl_minify(W0, l);
+ uint32_t w = isl_align_npot(W, image_align_sa->w);
+
+ slice_w += w;
+ }
+
+ *phys_slice0_sa = isl_extent2d(slice_w, 1);
+}
+
+/**
+ * Calculate the physical extent of the surface's first array slice, in units
+ * of surface samples. If the surface is multi-leveled, then the result will
+ * be aligned to \a image_align_sa.
+ */
+static void
+isl_calc_phys_slice0_extent_sa(const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info,
+ enum isl_dim_layout dim_layout,
+ enum isl_msaa_layout msaa_layout,
+ const struct isl_extent3d *image_align_sa,
+ const struct isl_extent4d *phys_level0_sa,
+ struct isl_extent2d *phys_slice0_sa)
+{
+ switch (dim_layout) {
+ case ISL_DIM_LAYOUT_GEN9_1D:
+ isl_calc_phys_slice0_extent_sa_gen9_1d(dev, info,
+ image_align_sa, phys_level0_sa,
+ phys_slice0_sa);
+ return;
+ case ISL_DIM_LAYOUT_GEN4_2D:
+ isl_calc_phys_slice0_extent_sa_gen4_2d(dev, info, msaa_layout,
+ image_align_sa, phys_level0_sa,
+ phys_slice0_sa);
+ return;
+ case ISL_DIM_LAYOUT_GEN4_3D:
+ isl_calc_phys_slice0_extent_sa_gen4_3d(dev, info, image_align_sa,
+ phys_level0_sa, phys_slice0_sa);
+ return;
+ }
+}
+
+/**
+ * Calculate the pitch between physical array slices, in units of rows of
+ * surface elements.
+ */
+static uint32_t
+isl_calc_array_pitch_el_rows(const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info,
+ const struct isl_tile_info *tile_info,
+ enum isl_dim_layout dim_layout,
+ enum isl_array_pitch_span array_pitch_span,
+ const struct isl_extent3d *image_align_sa,
+ const struct isl_extent4d *phys_level0_sa,
+ const struct isl_extent2d *phys_slice0_sa)
+{
+ const struct isl_format_layout *fmtl = isl_format_get_layout(info->format);
+ uint32_t pitch_sa_rows = 0;
+
+ switch (dim_layout) {
+ case ISL_DIM_LAYOUT_GEN9_1D:
+ /* Each row is an array slice */
+ pitch_sa_rows = 1;
+ break;
+ case ISL_DIM_LAYOUT_GEN4_2D:
+ switch (array_pitch_span) {
+ case ISL_ARRAY_PITCH_SPAN_COMPACT:
+ pitch_sa_rows = isl_align_npot(phys_slice0_sa->h, image_align_sa->h);
+ break;
+ case ISL_ARRAY_PITCH_SPAN_FULL: {
+ /* The QPitch equation is found in the Broadwell PRM >> Volume 5:
+ * Memory Views >> Common Surface Formats >> Surface Layout >> 2D
+ * Surfaces >> Surface Arrays.
+ */
+ uint32_t H0_sa = phys_level0_sa->h;
+ uint32_t H1_sa = isl_minify(H0_sa, 1);
+
+ uint32_t h0_sa = isl_align_npot(H0_sa, image_align_sa->h);
+ uint32_t h1_sa = isl_align_npot(H1_sa, image_align_sa->h);
+
+ uint32_t m;
+ if (ISL_DEV_GEN(dev) >= 7) {
+ /* The QPitch equation changed slightly in Ivybridge. */
+ m = 12;
+ } else {
+ m = 11;
+ }
+
+ pitch_sa_rows = h0_sa + h1_sa + (m * image_align_sa->h);
+
+ if (ISL_DEV_GEN(dev) == 6 && info->samples > 1 &&
+ (info->height % 4 == 1)) {
+ /* [SNB] Errata from the Sandy Bridge PRM >> Volume 4 Part 1:
+ * Graphics Core >> Section 7.18.3.7: Surface Arrays:
+ *
+ * [SNB] Errata: Sampler MSAA Qpitch will be 4 greater than
+ * the value calculated in the equation above , for every
+ * other odd Surface Height starting from 1 i.e. 1,5,9,13.
+ *
+ * XXX(chadv): Is the errata natural corollary of the physical
+ * layout of interleaved samples?
+ */
+ pitch_sa_rows += 4;
+ }
+
+ pitch_sa_rows = isl_align_npot(pitch_sa_rows, fmtl->bh);
+ } /* end case */
+ break;
+ }
+ break;
+ case ISL_DIM_LAYOUT_GEN4_3D:
+ assert(array_pitch_span == ISL_ARRAY_PITCH_SPAN_COMPACT);
+ pitch_sa_rows = isl_align_npot(phys_slice0_sa->h, image_align_sa->h);
+ break;
+ default:
+ unreachable("bad isl_dim_layout");
+ break;
+ }
+
+ assert(pitch_sa_rows % fmtl->bh == 0);
+ uint32_t pitch_el_rows = pitch_sa_rows / fmtl->bh;
+
+ if (ISL_DEV_GEN(dev) >= 9 &&
+ info->dim == ISL_SURF_DIM_3D &&
+ tile_info->tiling != ISL_TILING_LINEAR) {
+ /* From the Skylake BSpec >> RENDER_SURFACE_STATE >> Surface QPitch:
+ *
+ * Tile Mode != Linear: This field must be set to an integer multiple
+ * of the tile height
+ */
+ pitch_el_rows = isl_align(pitch_el_rows, tile_info->height);
+ }
+
+ return pitch_el_rows;
+}
+
+/**
+ * Calculate the pitch of each surface row, in bytes.
+ */
+static uint32_t
+isl_calc_row_pitch(const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info,
+ const struct isl_tile_info *tile_info,
+ const struct isl_extent3d *image_align_sa,
+ const struct isl_extent2d *phys_slice0_sa)
+{
+ const struct isl_format_layout *fmtl = isl_format_get_layout(info->format);
+
+ uint32_t row_pitch = info->min_pitch;
+
+ /* First, align the surface to a cache line boundary, as the PRM explains
+ * below.
+ *
+ * From the Broadwell PRM >> Volume 5: Memory Views >> Common Surface
+ * Formats >> Surface Padding Requirements >> Render Target and Media
+ * Surfaces:
+ *
+ * The data port accesses data (pixels) outside of the surface if they
+ * are contained in the same cache request as pixels that are within the
+ * surface. These pixels will not be returned by the requesting message,
+ * however if these pixels lie outside of defined pages in the GTT,
+ * a GTT error will result when the cache request is processed. In order
+ * to avoid these GTT errors, “padding” at the bottom of the surface is
+ * sometimes necessary.
+ *
+ * From the Broadwell PRM >> Volume 5: Memory Views >> Common Surface
+ * Formats >> Surface Padding Requirements >> Sampling Engine Surfaces:
+ *
+ * The sampling engine accesses texels outside of the surface if they
+ * are contained in the same cache line as texels that are within the
+ * surface. These texels will not participate in any calculation
+ * performed by the sampling engine and will not affect the result of
+ * any sampling engine operation, however if these texels lie outside of
+ * defined pages in the GTT, a GTT error will result when the cache line
+ * is accessed. In order to avoid these GTT errors, “padding” at the
+ * bottom and right side of a sampling engine surface is sometimes
+ * necessary.
+ *
+ * It is possible that a cache line will straddle a page boundary if the
+ * base address or pitch is not aligned. All pages included in the cache
+ * lines that are part of the surface must map to valid GTT entries to
+ * avoid errors. To determine the necessary padding on the bottom and
+ * right side of the surface, refer to the table in Alignment Unit Size
+ * section for the i and j parameters for the surface format in use. The
+ * surface must then be extended to the next multiple of the alignment
+ * unit size in each dimension, and all texels contained in this
+ * extended surface must have valid GTT entries.
+ *
+ * For example, suppose the surface size is 15 texels by 10 texels and
+ * the alignment parameters are i=4 and j=2. In this case, the extended
+ * surface would be 16 by 10. Note that these calculations are done in
+ * texels, and must be converted to bytes based on the surface format
+ * being used to determine whether additional pages need to be defined.
+ */
+ assert(phys_slice0_sa->w % fmtl->bw == 0);
+ row_pitch = MAX(row_pitch, fmtl->bs * phys_slice0_sa->w);
+
+ switch (tile_info->tiling) {
+ case ISL_TILING_LINEAR:
+ /* From the Broadwel PRM >> Volume 2d: Command Reference: Structures >>
+ * RENDER_SURFACE_STATE Surface Pitch (p349):
+ *
+ * - For linear render target surfaces and surfaces accessed with the
+ * typed data port messages, the pitch must be a multiple of the
+ * element size for non-YUV surface formats. Pitch must be
+ * a multiple of 2 * element size for YUV surface formats.
+ *
+ * - [Requirements for SURFTYPE_BUFFER and SURFTYPE_STRBUF, which we
+ * ignore because isl doesn't do buffers.]
+ *
+ * - For other linear surfaces, the pitch can be any multiple of
+ * bytes.
+ */
+ if (info->usage & ISL_SURF_USAGE_RENDER_TARGET_BIT) {
+ if (isl_format_is_yuv(info->format)) {
+ row_pitch = isl_align_npot(row_pitch, 2 * fmtl->bs);
+ } else {
+ row_pitch = isl_align_npot(row_pitch, fmtl->bs);
+ }
+ }
+ break;
+ default:
+ /* From the Broadwel PRM >> Volume 2d: Command Reference: Structures >>
+ * RENDER_SURFACE_STATE Surface Pitch (p349):
+ *
+ * - For tiled surfaces, the pitch must be a multiple of the tile
+ * width.
+ */
+ row_pitch = isl_align(row_pitch, tile_info->width);
+ break;
+ }
+
+ return row_pitch;
+}
+
+/**
+ * Calculate the surface's total height, including padding, in units of
+ * surface elements.
+ */
+static uint32_t
+isl_calc_total_height_el(const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info,
+ const struct isl_tile_info *tile_info,
+ uint32_t phys_array_len,
+ uint32_t row_pitch,
+ uint32_t array_pitch_el_rows)
+{
+ const struct isl_format_layout *fmtl = isl_format_get_layout(info->format);
+
+ uint32_t total_h_el = phys_array_len * array_pitch_el_rows;
+ uint32_t pad_bytes = 0;
+
+ /* From the Broadwell PRM >> Volume 5: Memory Views >> Common Surface
+ * Formats >> Surface Padding Requirements >> Render Target and Media
+ * Surfaces:
+ *
+ * The data port accesses data (pixels) outside of the surface if they
+ * are contained in the same cache request as pixels that are within the
+ * surface. These pixels will not be returned by the requesting message,
+ * however if these pixels lie outside of defined pages in the GTT,
+ * a GTT error will result when the cache request is processed. In
+ * order to avoid these GTT errors, “padding” at the bottom of the
+ * surface is sometimes necessary.
+ *
+ * From the Broadwell PRM >> Volume 5: Memory Views >> Common Surface
+ * Formats >> Surface Padding Requirements >> Sampling Engine Surfaces:
+ *
+ * ... Lots of padding requirements, all listed separately below.
+ */
+
+ /* We can safely ignore the first padding requirement, quoted below,
+ * because isl doesn't do buffers.
+ *
+ * - [pre-BDW] For buffers, which have no inherent “height,” padding
+ * requirements are different. A buffer must be padded to the next
+ * multiple of 256 array elements, with an additional 16 bytes added
+ * beyond that to account for the L1 cache line.
+ */
+
+ /*
+ * - For compressed textures [...], padding at the bottom of the surface
+ * is to an even compressed row.
+ */
+ if (isl_format_is_compressed(info->format))
+ total_h_el = isl_align(total_h_el, 2);
+
+ /*
+ * - For cube surfaces, an additional two rows of padding are required
+ * at the bottom of the surface.
+ */
+ if (info->usage & ISL_SURF_USAGE_CUBE_BIT)
+ total_h_el += 2;
+
+ /*
+ * - For packed YUV, 96 bpt, 48 bpt, and 24 bpt surface formats,
+ * additional padding is required. These surfaces require an extra row
+ * plus 16 bytes of padding at the bottom in addition to the general
+ * padding requirements.
+ */
+ if (isl_format_is_yuv(info->format) &&
+ (fmtl->bs == 96 || fmtl->bs == 48|| fmtl->bs == 24)) {
+ total_h_el += 1;
+ pad_bytes += 16;
+ }
+
+ /*
+ * - For linear surfaces, additional padding of 64 bytes is required at
+ * the bottom of the surface. This is in addition to the padding
+ * required above.
+ */
+ if (tile_info->tiling == ISL_TILING_LINEAR)
+ pad_bytes += 64;
+
+ /* The below text weakens, not strengthens, the padding requirements for
+ * linear surfaces. Therefore we can safely ignore it.
+ *
+ * - [BDW+] For SURFTYPE_BUFFER, SURFTYPE_1D, and SURFTYPE_2D non-array,
+ * non-MSAA, non-mip-mapped surfaces in linear memory, the only
+ * padding requirement is to the next aligned 64-byte boundary beyond
+ * the end of the surface. The rest of the padding requirements
+ * documented above do not apply to these surfaces.
+ */
+
+ /*
+ * - [SKL+] For SURFTYPE_2D and SURFTYPE_3D with linear mode and
+ * height % 4 != 0, the surface must be padded with
+ * 4-(height % 4)*Surface Pitch # of bytes.
+ */
+ if (ISL_DEV_GEN(dev) >= 9 &&
+ tile_info->tiling == ISL_TILING_LINEAR &&
+ (info->dim == ISL_SURF_DIM_2D || info->dim == ISL_SURF_DIM_3D)) {
+ total_h_el = isl_align(total_h_el, 4);
+ }
+
+ /*
+ * - [SKL+] For SURFTYPE_1D with linear mode, the surface must be padded
+ * to 4 times the Surface Pitch # of bytes
+ */
+ if (ISL_DEV_GEN(dev) >= 9 &&
+ tile_info->tiling == ISL_TILING_LINEAR &&
+ info->dim == ISL_SURF_DIM_1D) {
+ total_h_el += 4;
+ }
+
+ /* Be sloppy. Align any leftover padding to a row boundary. */
+ total_h_el += isl_align_div_npot(pad_bytes, row_pitch);
+
+ return total_h_el;
+}
+
+bool
+isl_surf_init_s(const struct isl_device *dev,
+ struct isl_surf *surf,
+ const struct isl_surf_init_info *restrict info)
+{
+ const struct isl_format_layout *fmtl = isl_format_get_layout(info->format);
+
+ const struct isl_extent4d logical_level0_px = {
+ .w = info->width,
+ .h = info->height,
+ .d = info->depth,
+ .a = info->array_len,
+ };
+
+ enum isl_dim_layout dim_layout =
+ isl_surf_choose_dim_layout(dev, info->dim);
+
+ enum isl_tiling tiling;
+ if (!isl_surf_choose_tiling(dev, info, &tiling))
+ return false;
+
+ struct isl_tile_info tile_info;
+ if (!isl_tiling_get_info(dev, tiling, fmtl->bs, &tile_info))
+ return false;
+
+ enum isl_msaa_layout msaa_layout;
+ if (!isl_choose_msaa_layout(dev, info, tiling, &msaa_layout))
+ return false;
+
+ struct isl_extent3d image_align_el;
+ isl_choose_image_alignment_el(dev, info, tiling, msaa_layout,
+ &image_align_el);
+
+ struct isl_extent3d image_align_sa =
+ isl_extent3d_el_to_sa(info->format, image_align_el);
+
+ struct isl_extent4d phys_level0_sa;
+ isl_calc_phys_level0_extent_sa(dev, info, dim_layout, tiling, msaa_layout,
+ &phys_level0_sa);
+ assert(phys_level0_sa.w % fmtl->bw == 0);
+ assert(phys_level0_sa.h % fmtl->bh == 0);
+
+ enum isl_array_pitch_span array_pitch_span =
+ isl_choose_array_pitch_span(dev, info, dim_layout, &phys_level0_sa);
+
+ struct isl_extent2d phys_slice0_sa;
+ isl_calc_phys_slice0_extent_sa(dev, info, dim_layout, msaa_layout,
+ &image_align_sa, &phys_level0_sa,
+ &phys_slice0_sa);
+ assert(phys_slice0_sa.w % fmtl->bw == 0);
+ assert(phys_slice0_sa.h % fmtl->bh == 0);
+
+ const uint32_t row_pitch = isl_calc_row_pitch(dev, info, &tile_info,
+ &image_align_sa,
+ &phys_slice0_sa);
+
+ const uint32_t array_pitch_el_rows =
+ isl_calc_array_pitch_el_rows(dev, info, &tile_info, dim_layout,
+ array_pitch_span, &image_align_sa,
+ &phys_level0_sa, &phys_slice0_sa);
+
+ const uint32_t total_h_el =
+ isl_calc_total_height_el(dev, info, &tile_info,
+ phys_level0_sa.array_len, row_pitch,
+ array_pitch_el_rows);
+
+ const uint32_t total_h_sa = total_h_el * fmtl->bh;
+ const uint32_t size = row_pitch * isl_align(total_h_sa, tile_info.height);
+
+ /* Alignment of surface base address, in bytes */
+ uint32_t base_alignment = MAX(1, info->min_alignment);
+ assert(isl_is_pow2(base_alignment) && isl_is_pow2(tile_info.size));
+ base_alignment = MAX(base_alignment, tile_info.size);
+
+ *surf = (struct isl_surf) {
+ .dim = info->dim,
+ .dim_layout = dim_layout,
+ .msaa_layout = msaa_layout,
+ .tiling = tiling,
+ .format = info->format,
+
+ .levels = info->levels,
+ .samples = info->samples,
+
+ .image_alignment_el = image_align_el,
+ .logical_level0_px = logical_level0_px,
+ .phys_level0_sa = phys_level0_sa,
+
+ .size = size,
+ .alignment = base_alignment,
+ .row_pitch = row_pitch,
+ .array_pitch_el_rows = array_pitch_el_rows,
+ .array_pitch_span = array_pitch_span,
+
+ .usage = info->usage,
+ };
+
+ return true;
+}
+
+/**
+ * A variant of isl_surf_get_image_offset_sa() specific to
+ * ISL_DIM_LAYOUT_GEN4_2D.
+ */
+static void
+get_image_offset_sa_gen4_2d(const struct isl_surf *surf,
+ uint32_t level, uint32_t layer,
+ uint32_t *x_offset_sa,
+ uint32_t *y_offset_sa)
+{
+ assert(level < surf->levels);
+ assert(layer < surf->phys_level0_sa.array_len);
+ assert(surf->phys_level0_sa.depth == 1);
+
+ const struct isl_extent3d image_align_sa =
+ isl_surf_get_image_alignment_sa(surf);
+
+ const uint32_t W0 = surf->phys_level0_sa.width;
+ const uint32_t H0 = surf->phys_level0_sa.height;
+
+ uint32_t x = 0;
+ uint32_t y = layer * isl_surf_get_array_pitch_sa_rows(surf);
+
+ for (uint32_t l = 0; l < level; ++l) {
+ if (l == 1) {
+ uint32_t W = isl_minify(W0, l);
+
+ if (surf->msaa_layout == ISL_MSAA_LAYOUT_INTERLEAVED)
+ isl_msaa_interleaved_scale_px_to_sa(surf->samples, &W, NULL);
+
+ x += isl_align_npot(W, image_align_sa.w);
+ } else {
+ uint32_t H = isl_minify(H0, l);
+
+ if (surf->msaa_layout == ISL_MSAA_LAYOUT_INTERLEAVED)
+ isl_msaa_interleaved_scale_px_to_sa(surf->samples, NULL, &H);
+
+ y += isl_align_npot(H, image_align_sa.h);
+ }
+ }
+
+ *x_offset_sa = x;
+ *y_offset_sa = y;
+}
+
+/**
+ * A variant of isl_surf_get_image_offset_sa() specific to
+ * ISL_DIM_LAYOUT_GEN4_3D.
+ */
+static void
+get_image_offset_sa_gen4_3d(const struct isl_surf *surf,
+ uint32_t level, uint32_t logical_z_offset_px,
+ uint32_t *x_offset_sa,
+ uint32_t *y_offset_sa)
+{
+ assert(level < surf->levels);
+ assert(logical_z_offset_px < isl_minify(surf->phys_level0_sa.depth, level));
+ assert(surf->phys_level0_sa.array_len == 1);
+
+ const struct isl_extent3d image_align_sa =
+ isl_surf_get_image_alignment_sa(surf);
+
+ const uint32_t W0 = surf->phys_level0_sa.width;
+ const uint32_t H0 = surf->phys_level0_sa.height;
+ const uint32_t D0 = surf->phys_level0_sa.depth;
+
+ uint32_t x = 0;
+ uint32_t y = 0;
+
+ for (uint32_t l = 0; l < level; ++l) {
+ const uint32_t level_h = isl_align_npot(isl_minify(H0, l), image_align_sa.h);
+ const uint32_t level_d = isl_align_npot(isl_minify(D0, l), image_align_sa.d);
+ const uint32_t max_layers_vert = isl_align(level_d, 1u << l) / (1u << l);
+
+ y += level_h * max_layers_vert;
+ }
+
+ const uint32_t level_w = isl_align_npot(isl_minify(W0, level), image_align_sa.w);
+ const uint32_t level_h = isl_align_npot(isl_minify(H0, level), image_align_sa.h);
+ const uint32_t level_d = isl_align_npot(isl_minify(D0, level), image_align_sa.d);
+
+ const uint32_t max_layers_horiz = MIN(level_d, 1u << level);
+ const uint32_t max_layers_vert = isl_align_div(level_d, 1u << level);
+
+ x += level_w * (logical_z_offset_px % max_layers_horiz);
+ y += level_h * (logical_z_offset_px / max_layers_vert);
+
+ *x_offset_sa = x;
+ *y_offset_sa = y;
+}
+
+/**
+ * A variant of isl_surf_get_image_offset_sa() specific to
+ * ISL_DIM_LAYOUT_GEN9_1D.
+ */
+static void
+get_image_offset_sa_gen9_1d(const struct isl_surf *surf,
+ uint32_t level, uint32_t layer,
+ uint32_t *x_offset_sa,
+ uint32_t *y_offset_sa)
+{
+ assert(level < surf->levels);
+ assert(layer < surf->phys_level0_sa.array_len);
+ assert(surf->phys_level0_sa.height == 1);
+ assert(surf->phys_level0_sa.depth == 1);
+ assert(surf->samples == 1);
+
+ const uint32_t W0 = surf->phys_level0_sa.width;
+ const struct isl_extent3d image_align_sa =
+ isl_surf_get_image_alignment_sa(surf);
+
+ uint32_t x = layer * isl_surf_get_array_pitch_sa_rows(surf);
+
+ for (uint32_t l = 0; l < level; ++l) {
+ uint32_t W = isl_minify(W0, l);
+ uint32_t w = isl_align_npot(W, image_align_sa.w);
+
+ x += w;
+ }
+
+ *x_offset_sa = x;
+ *y_offset_sa = 0;
+}
+
+void
+isl_surf_get_image_offset_sa(const struct isl_surf *surf,
+ uint32_t level,
+ uint32_t logical_array_layer,
+ uint32_t logical_z_offset_px,
+ uint32_t *x_offset_sa,
+ uint32_t *y_offset_sa)
+{
+ assert(level < surf->levels);
+ assert(logical_array_layer < surf->logical_level0_px.array_len);
+ assert(logical_z_offset_px
+ < isl_minify(surf->logical_level0_px.depth, level));
+
+ switch (surf->dim_layout) {
+ case ISL_DIM_LAYOUT_GEN9_1D:
+ get_image_offset_sa_gen9_1d(surf, level, logical_array_layer,
+ x_offset_sa, y_offset_sa);
+ break;
+ case ISL_DIM_LAYOUT_GEN4_2D:
+ get_image_offset_sa_gen4_2d(surf, level, logical_array_layer,
+ x_offset_sa, y_offset_sa);
+ break;
+ case ISL_DIM_LAYOUT_GEN4_3D:
+ get_image_offset_sa_gen4_3d(surf, level, logical_z_offset_px,
+ x_offset_sa, y_offset_sa);
+ break;
+ }
+}
--- /dev/null
+/*
+ * Copyright 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+/**
+ * @file
+ * @brief Intel Surface Layout
+ *
+ * Header Layout
+ * -------------
+ * The header is ordered as:
+ * - forward declarations
+ * - macros that may be overridden at compile-time for specific gens
+ * - enums and constants
+ * - structs and unions
+ * - functions
+ */
+
+#pragma once
+
+#include <assert.h>
+#include <stdbool.h>
+#include <stdint.h>
+
+#include "util/macros.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+struct brw_device_info;
+struct brw_image_param;
+
+#ifndef ISL_DEV_GEN
+/**
+ * @brief Get the hardware generation of isl_device.
+ *
+ * You can define this as a compile-time constant in the CFLAGS. For example,
+ * `gcc -DISL_DEV_GEN(dev)=9 ...`.
+ */
+#define ISL_DEV_GEN(__dev) ((__dev)->info->gen)
+#endif
+
+#ifndef ISL_DEV_USE_SEPARATE_STENCIL
+/**
+ * You can define this as a compile-time constant in the CFLAGS. For example,
+ * `gcc -DISL_DEV_USE_SEPARATE_STENCIL(dev)=1 ...`.
+ */
+#define ISL_DEV_USE_SEPARATE_STENCIL(__dev) ((__dev)->use_separate_stencil)
+#endif
+
+/**
+ * Hardware enumeration SURFACE_FORMAT.
+ *
+ * For the official list, see Broadwell PRM: Volume 2b: Command Reference:
+ * Enumerations: SURFACE_FORMAT.
+ */
+enum isl_format {
+ ISL_FORMAT_R32G32B32A32_FLOAT = 0,
+ ISL_FORMAT_R32G32B32A32_SINT = 1,
+ ISL_FORMAT_R32G32B32A32_UINT = 2,
+ ISL_FORMAT_R32G32B32A32_UNORM = 3,
+ ISL_FORMAT_R32G32B32A32_SNORM = 4,
+ ISL_FORMAT_R64G64_FLOAT = 5,
+ ISL_FORMAT_R32G32B32X32_FLOAT = 6,
+ ISL_FORMAT_R32G32B32A32_SSCALED = 7,
+ ISL_FORMAT_R32G32B32A32_USCALED = 8,
+ ISL_FORMAT_R32G32B32A32_SFIXED = 32,
+ ISL_FORMAT_R64G64_PASSTHRU = 33,
+ ISL_FORMAT_R32G32B32_FLOAT = 64,
+ ISL_FORMAT_R32G32B32_SINT = 65,
+ ISL_FORMAT_R32G32B32_UINT = 66,
+ ISL_FORMAT_R32G32B32_UNORM = 67,
+ ISL_FORMAT_R32G32B32_SNORM = 68,
+ ISL_FORMAT_R32G32B32_SSCALED = 69,
+ ISL_FORMAT_R32G32B32_USCALED = 70,
+ ISL_FORMAT_R32G32B32_SFIXED = 80,
+ ISL_FORMAT_R16G16B16A16_UNORM = 128,
+ ISL_FORMAT_R16G16B16A16_SNORM = 129,
+ ISL_FORMAT_R16G16B16A16_SINT = 130,
+ ISL_FORMAT_R16G16B16A16_UINT = 131,
+ ISL_FORMAT_R16G16B16A16_FLOAT = 132,
+ ISL_FORMAT_R32G32_FLOAT = 133,
+ ISL_FORMAT_R32G32_SINT = 134,
+ ISL_FORMAT_R32G32_UINT = 135,
+ ISL_FORMAT_R32_FLOAT_X8X24_TYPELESS = 136,
+ ISL_FORMAT_X32_TYPELESS_G8X24_UINT = 137,
+ ISL_FORMAT_L32A32_FLOAT = 138,
+ ISL_FORMAT_R32G32_UNORM = 139,
+ ISL_FORMAT_R32G32_SNORM = 140,
+ ISL_FORMAT_R64_FLOAT = 141,
+ ISL_FORMAT_R16G16B16X16_UNORM = 142,
+ ISL_FORMAT_R16G16B16X16_FLOAT = 143,
+ ISL_FORMAT_A32X32_FLOAT = 144,
+ ISL_FORMAT_L32X32_FLOAT = 145,
+ ISL_FORMAT_I32X32_FLOAT = 146,
+ ISL_FORMAT_R16G16B16A16_SSCALED = 147,
+ ISL_FORMAT_R16G16B16A16_USCALED = 148,
+ ISL_FORMAT_R32G32_SSCALED = 149,
+ ISL_FORMAT_R32G32_USCALED = 150,
+ ISL_FORMAT_R32G32_SFIXED = 160,
+ ISL_FORMAT_R64_PASSTHRU = 161,
+ ISL_FORMAT_B8G8R8A8_UNORM = 192,
+ ISL_FORMAT_B8G8R8A8_UNORM_SRGB = 193,
+ ISL_FORMAT_R10G10B10A2_UNORM = 194,
+ ISL_FORMAT_R10G10B10A2_UNORM_SRGB = 195,
+ ISL_FORMAT_R10G10B10A2_UINT = 196,
+ ISL_FORMAT_R10G10B10_SNORM_A2_UNORM = 197,
+ ISL_FORMAT_R8G8B8A8_UNORM = 199,
+ ISL_FORMAT_R8G8B8A8_UNORM_SRGB = 200,
+ ISL_FORMAT_R8G8B8A8_SNORM = 201,
+ ISL_FORMAT_R8G8B8A8_SINT = 202,
+ ISL_FORMAT_R8G8B8A8_UINT = 203,
+ ISL_FORMAT_R16G16_UNORM = 204,
+ ISL_FORMAT_R16G16_SNORM = 205,
+ ISL_FORMAT_R16G16_SINT = 206,
+ ISL_FORMAT_R16G16_UINT = 207,
+ ISL_FORMAT_R16G16_FLOAT = 208,
+ ISL_FORMAT_B10G10R10A2_UNORM = 209,
+ ISL_FORMAT_B10G10R10A2_UNORM_SRGB = 210,
+ ISL_FORMAT_R11G11B10_FLOAT = 211,
+ ISL_FORMAT_R32_SINT = 214,
+ ISL_FORMAT_R32_UINT = 215,
+ ISL_FORMAT_R32_FLOAT = 216,
+ ISL_FORMAT_R24_UNORM_X8_TYPELESS = 217,
+ ISL_FORMAT_X24_TYPELESS_G8_UINT = 218,
+ ISL_FORMAT_L32_UNORM = 221,
+ ISL_FORMAT_A32_UNORM = 222,
+ ISL_FORMAT_L16A16_UNORM = 223,
+ ISL_FORMAT_I24X8_UNORM = 224,
+ ISL_FORMAT_L24X8_UNORM = 225,
+ ISL_FORMAT_A24X8_UNORM = 226,
+ ISL_FORMAT_I32_FLOAT = 227,
+ ISL_FORMAT_L32_FLOAT = 228,
+ ISL_FORMAT_A32_FLOAT = 229,
+ ISL_FORMAT_X8B8_UNORM_G8R8_SNORM = 230,
+ ISL_FORMAT_A8X8_UNORM_G8R8_SNORM = 231,
+ ISL_FORMAT_B8X8_UNORM_G8R8_SNORM = 232,
+ ISL_FORMAT_B8G8R8X8_UNORM = 233,
+ ISL_FORMAT_B8G8R8X8_UNORM_SRGB = 234,
+ ISL_FORMAT_R8G8B8X8_UNORM = 235,
+ ISL_FORMAT_R8G8B8X8_UNORM_SRGB = 236,
+ ISL_FORMAT_R9G9B9E5_SHAREDEXP = 237,
+ ISL_FORMAT_B10G10R10X2_UNORM = 238,
+ ISL_FORMAT_L16A16_FLOAT = 240,
+ ISL_FORMAT_R32_UNORM = 241,
+ ISL_FORMAT_R32_SNORM = 242,
+ ISL_FORMAT_R10G10B10X2_USCALED = 243,
+ ISL_FORMAT_R8G8B8A8_SSCALED = 244,
+ ISL_FORMAT_R8G8B8A8_USCALED = 245,
+ ISL_FORMAT_R16G16_SSCALED = 246,
+ ISL_FORMAT_R16G16_USCALED = 247,
+ ISL_FORMAT_R32_SSCALED = 248,
+ ISL_FORMAT_R32_USCALED = 249,
+ ISL_FORMAT_B5G6R5_UNORM = 256,
+ ISL_FORMAT_B5G6R5_UNORM_SRGB = 257,
+ ISL_FORMAT_B5G5R5A1_UNORM = 258,
+ ISL_FORMAT_B5G5R5A1_UNORM_SRGB = 259,
+ ISL_FORMAT_B4G4R4A4_UNORM = 260,
+ ISL_FORMAT_B4G4R4A4_UNORM_SRGB = 261,
+ ISL_FORMAT_R8G8_UNORM = 262,
+ ISL_FORMAT_R8G8_SNORM = 263,
+ ISL_FORMAT_R8G8_SINT = 264,
+ ISL_FORMAT_R8G8_UINT = 265,
+ ISL_FORMAT_R16_UNORM = 266,
+ ISL_FORMAT_R16_SNORM = 267,
+ ISL_FORMAT_R16_SINT = 268,
+ ISL_FORMAT_R16_UINT = 269,
+ ISL_FORMAT_R16_FLOAT = 270,
+ ISL_FORMAT_A8P8_UNORM_PALETTE0 = 271,
+ ISL_FORMAT_A8P8_UNORM_PALETTE1 = 272,
+ ISL_FORMAT_I16_UNORM = 273,
+ ISL_FORMAT_L16_UNORM = 274,
+ ISL_FORMAT_A16_UNORM = 275,
+ ISL_FORMAT_L8A8_UNORM = 276,
+ ISL_FORMAT_I16_FLOAT = 277,
+ ISL_FORMAT_L16_FLOAT = 278,
+ ISL_FORMAT_A16_FLOAT = 279,
+ ISL_FORMAT_L8A8_UNORM_SRGB = 280,
+ ISL_FORMAT_R5G5_SNORM_B6_UNORM = 281,
+ ISL_FORMAT_B5G5R5X1_UNORM = 282,
+ ISL_FORMAT_B5G5R5X1_UNORM_SRGB = 283,
+ ISL_FORMAT_R8G8_SSCALED = 284,
+ ISL_FORMAT_R8G8_USCALED = 285,
+ ISL_FORMAT_R16_SSCALED = 286,
+ ISL_FORMAT_R16_USCALED = 287,
+ ISL_FORMAT_P8A8_UNORM_PALETTE0 = 290,
+ ISL_FORMAT_P8A8_UNORM_PALETTE1 = 291,
+ ISL_FORMAT_A1B5G5R5_UNORM = 292,
+ ISL_FORMAT_A4B4G4R4_UNORM = 293,
+ ISL_FORMAT_L8A8_UINT = 294,
+ ISL_FORMAT_L8A8_SINT = 295,
+ ISL_FORMAT_R8_UNORM = 320,
+ ISL_FORMAT_R8_SNORM = 321,
+ ISL_FORMAT_R8_SINT = 322,
+ ISL_FORMAT_R8_UINT = 323,
+ ISL_FORMAT_A8_UNORM = 324,
+ ISL_FORMAT_I8_UNORM = 325,
+ ISL_FORMAT_L8_UNORM = 326,
+ ISL_FORMAT_P4A4_UNORM_PALETTE0 = 327,
+ ISL_FORMAT_A4P4_UNORM_PALETTE0 = 328,
+ ISL_FORMAT_R8_SSCALED = 329,
+ ISL_FORMAT_R8_USCALED = 330,
+ ISL_FORMAT_P8_UNORM_PALETTE0 = 331,
+ ISL_FORMAT_L8_UNORM_SRGB = 332,
+ ISL_FORMAT_P8_UNORM_PALETTE1 = 333,
+ ISL_FORMAT_P4A4_UNORM_PALETTE1 = 334,
+ ISL_FORMAT_A4P4_UNORM_PALETTE1 = 335,
+ ISL_FORMAT_Y8_UNORM = 336,
+ ISL_FORMAT_L8_UINT = 338,
+ ISL_FORMAT_L8_SINT = 339,
+ ISL_FORMAT_I8_UINT = 340,
+ ISL_FORMAT_I8_SINT = 341,
+ ISL_FORMAT_DXT1_RGB_SRGB = 384,
+ ISL_FORMAT_R1_UNORM = 385,
+ ISL_FORMAT_YCRCB_NORMAL = 386,
+ ISL_FORMAT_YCRCB_SWAPUVY = 387,
+ ISL_FORMAT_P2_UNORM_PALETTE0 = 388,
+ ISL_FORMAT_P2_UNORM_PALETTE1 = 389,
+ ISL_FORMAT_BC1_UNORM = 390,
+ ISL_FORMAT_BC2_UNORM = 391,
+ ISL_FORMAT_BC3_UNORM = 392,
+ ISL_FORMAT_BC4_UNORM = 393,
+ ISL_FORMAT_BC5_UNORM = 394,
+ ISL_FORMAT_BC1_UNORM_SRGB = 395,
+ ISL_FORMAT_BC2_UNORM_SRGB = 396,
+ ISL_FORMAT_BC3_UNORM_SRGB = 397,
+ ISL_FORMAT_MONO8 = 398,
+ ISL_FORMAT_YCRCB_SWAPUV = 399,
+ ISL_FORMAT_YCRCB_SWAPY = 400,
+ ISL_FORMAT_DXT1_RGB = 401,
+ ISL_FORMAT_FXT1 = 402,
+ ISL_FORMAT_R8G8B8_UNORM = 403,
+ ISL_FORMAT_R8G8B8_SNORM = 404,
+ ISL_FORMAT_R8G8B8_SSCALED = 405,
+ ISL_FORMAT_R8G8B8_USCALED = 406,
+ ISL_FORMAT_R64G64B64A64_FLOAT = 407,
+ ISL_FORMAT_R64G64B64_FLOAT = 408,
+ ISL_FORMAT_BC4_SNORM = 409,
+ ISL_FORMAT_BC5_SNORM = 410,
+ ISL_FORMAT_R16G16B16_FLOAT = 411,
+ ISL_FORMAT_R16G16B16_UNORM = 412,
+ ISL_FORMAT_R16G16B16_SNORM = 413,
+ ISL_FORMAT_R16G16B16_SSCALED = 414,
+ ISL_FORMAT_R16G16B16_USCALED = 415,
+ ISL_FORMAT_BC6H_SF16 = 417,
+ ISL_FORMAT_BC7_UNORM = 418,
+ ISL_FORMAT_BC7_UNORM_SRGB = 419,
+ ISL_FORMAT_BC6H_UF16 = 420,
+ ISL_FORMAT_PLANAR_420_8 = 421,
+ ISL_FORMAT_R8G8B8_UNORM_SRGB = 424,
+ ISL_FORMAT_ETC1_RGB8 = 425,
+ ISL_FORMAT_ETC2_RGB8 = 426,
+ ISL_FORMAT_EAC_R11 = 427,
+ ISL_FORMAT_EAC_RG11 = 428,
+ ISL_FORMAT_EAC_SIGNED_R11 = 429,
+ ISL_FORMAT_EAC_SIGNED_RG11 = 430,
+ ISL_FORMAT_ETC2_SRGB8 = 431,
+ ISL_FORMAT_R16G16B16_UINT = 432,
+ ISL_FORMAT_R16G16B16_SINT = 433,
+ ISL_FORMAT_R32_SFIXED = 434,
+ ISL_FORMAT_R10G10B10A2_SNORM = 435,
+ ISL_FORMAT_R10G10B10A2_USCALED = 436,
+ ISL_FORMAT_R10G10B10A2_SSCALED = 437,
+ ISL_FORMAT_R10G10B10A2_SINT = 438,
+ ISL_FORMAT_B10G10R10A2_SNORM = 439,
+ ISL_FORMAT_B10G10R10A2_USCALED = 440,
+ ISL_FORMAT_B10G10R10A2_SSCALED = 441,
+ ISL_FORMAT_B10G10R10A2_UINT = 442,
+ ISL_FORMAT_B10G10R10A2_SINT = 443,
+ ISL_FORMAT_R64G64B64A64_PASSTHRU = 444,
+ ISL_FORMAT_R64G64B64_PASSTHRU = 445,
+ ISL_FORMAT_ETC2_RGB8_PTA = 448,
+ ISL_FORMAT_ETC2_SRGB8_PTA = 449,
+ ISL_FORMAT_ETC2_EAC_RGBA8 = 450,
+ ISL_FORMAT_ETC2_EAC_SRGB8_A8 = 451,
+ ISL_FORMAT_R8G8B8_UINT = 456,
+ ISL_FORMAT_R8G8B8_SINT = 457,
+ ISL_FORMAT_RAW = 511,
+
+ /* Hardware doesn't understand this out-of-band value */
+ ISL_FORMAT_UNSUPPORTED = UINT16_MAX,
+};
+
+/**
+ * Numerical base type for channels of isl_format.
+ */
+enum isl_base_type {
+ ISL_VOID,
+ ISL_RAW,
+ ISL_UNORM,
+ ISL_SNORM,
+ ISL_UFLOAT,
+ ISL_SFLOAT,
+ ISL_UFIXED,
+ ISL_SFIXED,
+ ISL_UINT,
+ ISL_SINT,
+ ISL_USCALED,
+ ISL_SSCALED,
+};
+
+/**
+ * Colorspace of isl_format.
+ */
+enum isl_colorspace {
+ ISL_COLORSPACE_NONE = 0,
+ ISL_COLORSPACE_LINEAR,
+ ISL_COLORSPACE_SRGB,
+ ISL_COLORSPACE_YUV,
+};
+
+/**
+ * Texture compression mode of isl_format.
+ */
+enum isl_txc {
+ ISL_TXC_NONE = 0,
+ ISL_TXC_DXT1,
+ ISL_TXC_DXT3,
+ ISL_TXC_DXT5,
+ ISL_TXC_FXT1,
+ ISL_TXC_RGTC1,
+ ISL_TXC_RGTC2,
+ ISL_TXC_BPTC,
+ ISL_TXC_ETC1,
+ ISL_TXC_ETC2,
+};
+
+/**
+ * @brief Hardware tile mode
+ *
+ * WARNING: These values differ from the hardware enum values, which are
+ * unstable across hardware generations.
+ *
+ * Note that legacy Y tiling is ISL_TILING_Y0 instead of ISL_TILING_Y, to
+ * clearly distinguish it from Yf and Ys.
+ */
+enum isl_tiling {
+ ISL_TILING_LINEAR = 0,
+ ISL_TILING_W,
+ ISL_TILING_X,
+ ISL_TILING_Y0, /**< Legacy Y tiling */
+ ISL_TILING_Yf, /**< Standard 4K tiling. The 'f' means "four". */
+ ISL_TILING_Ys, /**< Standard 64K tiling. The 's' means "sixty-four". */
+};
+
+/**
+ * @defgroup Tiling Flags
+ * @{
+ */
+typedef uint32_t isl_tiling_flags_t;
+#define ISL_TILING_LINEAR_BIT (1u << ISL_TILING_LINEAR)
+#define ISL_TILING_W_BIT (1u << ISL_TILING_W)
+#define ISL_TILING_X_BIT (1u << ISL_TILING_X)
+#define ISL_TILING_Y0_BIT (1u << ISL_TILING_Y0)
+#define ISL_TILING_Yf_BIT (1u << ISL_TILING_Yf)
+#define ISL_TILING_Ys_BIT (1u << ISL_TILING_Ys)
+#define ISL_TILING_ANY_MASK (~0u)
+#define ISL_TILING_NON_LINEAR_MASK (~ISL_TILING_LINEAR_BIT)
+
+/** Any Y tiling, including legacy Y tiling. */
+#define ISL_TILING_ANY_Y_MASK (ISL_TILING_Y0_BIT | \
+ ISL_TILING_Yf_BIT | \
+ ISL_TILING_Ys_BIT)
+
+/** The Skylake BSpec refers to Yf and Ys as "standard tiling formats". */
+#define ISL_TILING_STD_Y_MASK (ISL_TILING_Yf_BIT | \
+ ISL_TILING_Ys_BIT)
+/** @} */
+
+/**
+ * @brief Logical dimension of surface.
+ *
+ * Note: There is no dimension for cube map surfaces. ISL interprets cube maps
+ * as 2D array surfaces.
+ */
+enum isl_surf_dim {
+ ISL_SURF_DIM_1D,
+ ISL_SURF_DIM_2D,
+ ISL_SURF_DIM_3D,
+};
+
+/**
+ * @brief Physical layout of the surface's dimensions.
+ */
+enum isl_dim_layout {
+ /**
+ * For details, see the G35 PRM >> Volume 1: Graphics Core >> Section
+ * 6.17.3: 2D Surfaces.
+ *
+ * On many gens, 1D surfaces share the same layout as 2D surfaces. From
+ * the G35 PRM >> Volume 1: Graphics Core >> Section 6.17.2: 1D Surfaces:
+ *
+ * One-dimensional surfaces are identical to 2D surfaces with height of
+ * one.
+ *
+ * @invariant isl_surf::phys_level0_sa::depth == 1
+ */
+ ISL_DIM_LAYOUT_GEN4_2D,
+
+ /**
+ * For details, see the G35 PRM >> Volume 1: Graphics Core >> Section
+ * 6.17.5: 3D Surfaces.
+ *
+ * @invariant isl_surf::phys_level0_sa::array_len == 1
+ */
+ ISL_DIM_LAYOUT_GEN4_3D,
+
+ /**
+ * For details, see the Skylake BSpec >> Memory Views >> Common Surface
+ * Formats >> Surface Layout and Tiling >> » 1D Surfaces.
+ */
+ ISL_DIM_LAYOUT_GEN9_1D,
+};
+
+/* TODO(chadv): Explain */
+enum isl_array_pitch_span {
+ ISL_ARRAY_PITCH_SPAN_FULL,
+ ISL_ARRAY_PITCH_SPAN_COMPACT,
+};
+
+/**
+ * @defgroup Surface Usage
+ * @{
+ */
+typedef uint64_t isl_surf_usage_flags_t;
+#define ISL_SURF_USAGE_RENDER_TARGET_BIT (1u << 0)
+#define ISL_SURF_USAGE_DEPTH_BIT (1u << 1)
+#define ISL_SURF_USAGE_STENCIL_BIT (1u << 2)
+#define ISL_SURF_USAGE_TEXTURE_BIT (1u << 3)
+#define ISL_SURF_USAGE_CUBE_BIT (1u << 4)
+#define ISL_SURF_USAGE_DISABLE_AUX_BIT (1u << 5)
+#define ISL_SURF_USAGE_DISPLAY_BIT (1u << 6)
+#define ISL_SURF_USAGE_DISPLAY_ROTATE_90_BIT (1u << 7)
+#define ISL_SURF_USAGE_DISPLAY_ROTATE_180_BIT (1u << 8)
+#define ISL_SURF_USAGE_DISPLAY_ROTATE_270_BIT (1u << 9)
+#define ISL_SURF_USAGE_DISPLAY_FLIP_X_BIT (1u << 10)
+#define ISL_SURF_USAGE_DISPLAY_FLIP_Y_BIT (1u << 11)
+/** @} */
+
+/**
+ * @brief Multisample Format
+ */
+enum isl_msaa_layout {
+ /**
+ * @brief Suface is single-sampled.
+ */
+ ISL_MSAA_LAYOUT_NONE,
+
+ /**
+ * @brief [SNB+] Interleaved Multisample Format
+ *
+ * In this format, multiple samples are interleaved into each cacheline.
+ * In other words, the sample index is swizzled into the low 6 bits of the
+ * surface's virtual address space.
+ *
+ * For example, suppose the surface is legacy Y tiled, is 4x multisampled,
+ * and its pixel format is 32bpp. Then the first cacheline is arranged
+ * thus:
+ *
+ * (0,0,0) (0,1,0) (0,0,1) (1,0,1)
+ * (1,0,0) (1,1,0) (0,1,1) (1,1,1)
+ *
+ * (0,0,2) (1,0,2) (0,0,3) (1,0,3)
+ * (0,1,2) (1,1,2) (0,1,3) (1,1,3)
+ *
+ * The hardware docs refer to this format with multiple terms. In
+ * Sandybridge, this is the only multisample format; so no term is used.
+ * The Ivybridge docs refer to surfaces in this format as IMS (Interleaved
+ * Multisample Surface). Later hardware docs additionally refer to this
+ * format as MSFMT_DEPTH_STENCIL (because the format is deprecated for
+ * color surfaces).
+ *
+ * See the Sandybridge PRM, Volume 4, Part 1, Section 2.7 "Multisampled
+ * Surface Behavior".
+ *
+ * See the Ivybridge PRM, Volume 1, Part 1, Section 6.18.4.1 "Interleaved
+ * Multisampled Surfaces".
+ */
+ ISL_MSAA_LAYOUT_INTERLEAVED,
+
+ /**
+ * @brief [IVB+] Array Multisample Format
+ *
+ * In this format, the surface's physical layout resembles that of a
+ * 2D array surface.
+ *
+ * Suppose the multisample surface's logical extent is (w, h) and its
+ * sample count is N. Then surface's physical extent is the same as
+ * a singlesample 2D surface whose logical extent is (w, h) and array
+ * length is N. Array slice `i` contains the pixel values for sample
+ * index `i`.
+ *
+ * The Ivybridge docs refer to surfaces in this format as UMS
+ * (Uncompressed Multsample Layout) and CMS (Compressed Multisample
+ * Surface). The Broadwell docs additionally refer to this format as
+ * MSFMT_MSS (MSS=Multisample Surface Storage).
+ *
+ * See the Broadwell PRM, Volume 5 "Memory Views", Section "Uncompressed
+ * Multisample Surfaces".
+ *
+ * See the Broadwell PRM, Volume 5 "Memory Views", Section "Compressed
+ * Multisample Surfaces".
+ */
+ ISL_MSAA_LAYOUT_ARRAY,
+};
+
+
+struct isl_device {
+ const struct brw_device_info *info;
+ bool use_separate_stencil;
+ bool has_bit6_swizzling;
+};
+
+struct isl_extent2d {
+ union { uint32_t w, width; };
+ union { uint32_t h, height; };
+};
+
+struct isl_extent3d {
+ union { uint32_t w, width; };
+ union { uint32_t h, height; };
+ union { uint32_t d, depth; };
+};
+
+struct isl_extent4d {
+ union { uint32_t w, width; };
+ union { uint32_t h, height; };
+ union { uint32_t d, depth; };
+ union { uint32_t a, array_len; };
+};
+
+struct isl_channel_layout {
+ enum isl_base_type type;
+ uint8_t bits; /**< Size in bits */
+};
+
+/**
+ * Each format has 3D block extent (width, height, depth). The block extent
+ * of compressed formats is that of the format's compression block. For
+ * example, the block extent of ISL_FORMAT_ETC2_RGB8 is (w=4, h=4, d=1).
+ * The block extent of uncompressed pixel formats, such as
+ * ISL_FORMAT_R8G8B8A8_UNORM, is is (w=1, h=1, d=1).
+ */
+struct isl_format_layout {
+ enum isl_format format;
+
+ uint8_t bs; /**< Block size, in bytes, rounded towards 0 */
+ uint8_t bw; /**< Block width, in pixels */
+ uint8_t bh; /**< Block height, in pixels */
+ uint8_t bd; /**< Block depth, in pixels */
+
+ struct {
+ struct isl_channel_layout r; /**< Red channel */
+ struct isl_channel_layout g; /**< Green channel */
+ struct isl_channel_layout b; /**< Blue channel */
+ struct isl_channel_layout a; /**< Alpha channel */
+ struct isl_channel_layout l; /**< Luminance channel */
+ struct isl_channel_layout i; /**< Intensity channel */
+ struct isl_channel_layout p; /**< Palette channel */
+ } channels;
+
+ enum isl_colorspace colorspace;
+ enum isl_txc txc;
+};
+
+struct isl_tile_info {
+ enum isl_tiling tiling;
+ uint32_t width; /**< in bytes */
+ uint32_t height; /**< in rows of memory */
+ uint32_t size; /**< in bytes */
+};
+
+/**
+ * @brief Input to surface initialization
+ *
+ * @invariant width >= 1
+ * @invariant height >= 1
+ * @invariant depth >= 1
+ * @invariant levels >= 1
+ * @invariant samples >= 1
+ * @invariant array_len >= 1
+ *
+ * @invariant if 1D then height == 1 and depth == 1 and samples == 1
+ * @invariant if 2D then depth == 1
+ * @invariant if 3D then array_len == 1 and samples == 1
+ */
+struct isl_surf_init_info {
+ enum isl_surf_dim dim;
+ enum isl_format format;
+
+ uint32_t width;
+ uint32_t height;
+ uint32_t depth;
+ uint32_t levels;
+ uint32_t array_len;
+ uint32_t samples;
+
+ /** Lower bound for isl_surf::alignment, in bytes. */
+ uint32_t min_alignment;
+
+ /** Lower bound for isl_surf::pitch, in bytes. */
+ uint32_t min_pitch;
+
+ isl_surf_usage_flags_t usage;
+
+ /** Flags that alter how ISL selects isl_surf::tiling. */
+ isl_tiling_flags_t tiling_flags;
+};
+
+struct isl_surf {
+ enum isl_surf_dim dim;
+ enum isl_dim_layout dim_layout;
+ enum isl_msaa_layout msaa_layout;
+ enum isl_tiling tiling;
+ enum isl_format format;
+
+ /**
+ * Alignment of the upper-left sample of each subimage, in units of surface
+ * elements.
+ */
+ struct isl_extent3d image_alignment_el;
+
+ /**
+ * Logical extent of the surface's base level, in units of pixels. This is
+ * identical to the extent defined in isl_surf_init_info.
+ */
+ struct isl_extent4d logical_level0_px;
+
+ /**
+ * Physical extent of the surface's base level, in units of physical
+ * surface samples and aligned to the format's compression block.
+ *
+ * Consider isl_dim_layout as an operator that transforms a logical surface
+ * layout to a physical surface layout. Then
+ *
+ * logical_layout := (isl_surf::dim, isl_surf::logical_level0_px)
+ * isl_surf::phys_level0_sa := isl_surf::dim_layout * logical_layout
+ */
+ struct isl_extent4d phys_level0_sa;
+
+ uint32_t levels;
+ uint32_t samples;
+
+ /** Total size of the surface, in bytes. */
+ uint32_t size;
+
+ /** Required alignment for the surface's base address. */
+ uint32_t alignment;
+
+ /**
+ * Pitch between vertically adjacent surface elements, in bytes.
+ */
+ uint32_t row_pitch;
+
+ /**
+ * Pitch between physical array slices, in rows of surface elements.
+ */
+ uint32_t array_pitch_el_rows;
+
+ enum isl_array_pitch_span array_pitch_span;
+
+ /** Copy of isl_surf_init_info::usage. */
+ isl_surf_usage_flags_t usage;
+};
+
+extern const struct isl_format_layout isl_format_layouts[];
+
+void
+isl_device_init(struct isl_device *dev,
+ const struct brw_device_info *info,
+ bool has_bit6_swizzling);
+
+static inline const struct isl_format_layout * ATTRIBUTE_CONST
+isl_format_get_layout(enum isl_format fmt)
+{
+ return &isl_format_layouts[fmt];
+}
+
+bool
+isl_format_has_sint_channel(enum isl_format fmt) ATTRIBUTE_CONST;
+
+static inline bool
+isl_format_is_compressed(enum isl_format fmt)
+{
+ const struct isl_format_layout *fmtl = isl_format_get_layout(fmt);
+
+ return fmtl->txc != ISL_TXC_NONE;
+}
+
+static inline bool
+isl_format_has_bc_compression(enum isl_format fmt)
+{
+ switch (isl_format_get_layout(fmt)->txc) {
+ case ISL_TXC_DXT1:
+ case ISL_TXC_DXT3:
+ case ISL_TXC_DXT5:
+ return true;
+ case ISL_TXC_NONE:
+ case ISL_TXC_FXT1:
+ case ISL_TXC_RGTC1:
+ case ISL_TXC_RGTC2:
+ case ISL_TXC_BPTC:
+ case ISL_TXC_ETC1:
+ case ISL_TXC_ETC2:
+ return false;
+ }
+
+ unreachable("bad texture compression mode");
+ return false;
+}
+
+static inline bool
+isl_format_is_yuv(enum isl_format fmt)
+{
+ const struct isl_format_layout *fmtl = isl_format_get_layout(fmt);
+
+ return fmtl->colorspace == ISL_COLORSPACE_YUV;
+}
+
+static inline bool
+isl_format_block_is_1x1x1(enum isl_format fmt)
+{
+ const struct isl_format_layout *fmtl = isl_format_get_layout(fmt);
+
+ return fmtl->bw == 1 && fmtl->bh == 1 && fmtl->bd == 1;
+}
+
+static inline bool
+isl_format_is_rgb(enum isl_format fmt)
+{
+ return isl_format_layouts[fmt].channels.r.bits > 0 &&
+ isl_format_layouts[fmt].channels.g.bits > 0 &&
+ isl_format_layouts[fmt].channels.b.bits > 0 &&
+ isl_format_layouts[fmt].channels.a.bits == 0;
+}
+
+enum isl_format isl_format_rgb_to_rgba(enum isl_format rgb) ATTRIBUTE_CONST;
+enum isl_format isl_format_rgb_to_rgbx(enum isl_format rgb) ATTRIBUTE_CONST;
+
+bool isl_is_storage_image_format(enum isl_format fmt);
+
+enum isl_format
+isl_lower_storage_image_format(const struct isl_device *dev,
+ enum isl_format fmt);
+
+static inline bool
+isl_tiling_is_std_y(enum isl_tiling tiling)
+{
+ return (1u << tiling) & ISL_TILING_STD_Y_MASK;
+}
+
+bool
+isl_tiling_get_info(const struct isl_device *dev,
+ enum isl_tiling tiling,
+ uint32_t format_block_size,
+ struct isl_tile_info *info);
+
+void
+isl_tiling_get_extent(const struct isl_device *dev,
+ enum isl_tiling tiling,
+ uint32_t format_block_size,
+ struct isl_extent2d *e);
+bool
+isl_surf_choose_tiling(const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info,
+ enum isl_tiling *tiling);
+
+static inline bool
+isl_surf_usage_is_display(isl_surf_usage_flags_t usage)
+{
+ return usage & ISL_SURF_USAGE_DISPLAY_BIT;
+}
+
+static inline bool
+isl_surf_usage_is_depth(isl_surf_usage_flags_t usage)
+{
+ return usage & ISL_SURF_USAGE_DEPTH_BIT;
+}
+
+static inline bool
+isl_surf_usage_is_stencil(isl_surf_usage_flags_t usage)
+{
+ return usage & ISL_SURF_USAGE_STENCIL_BIT;
+}
+
+static inline bool
+isl_surf_usage_is_depth_and_stencil(isl_surf_usage_flags_t usage)
+{
+ return (usage & ISL_SURF_USAGE_DEPTH_BIT) &&
+ (usage & ISL_SURF_USAGE_STENCIL_BIT);
+}
+
+static inline bool
+isl_surf_usage_is_depth_or_stencil(isl_surf_usage_flags_t usage)
+{
+ return usage & (ISL_SURF_USAGE_DEPTH_BIT | ISL_SURF_USAGE_STENCIL_BIT);
+}
+
+static inline bool
+isl_surf_info_is_z16(const struct isl_surf_init_info *info)
+{
+ return (info->usage & ISL_SURF_USAGE_DEPTH_BIT) &&
+ (info->format == ISL_FORMAT_R16_UNORM);
+}
+
+static inline bool
+isl_surf_info_is_z32_float(const struct isl_surf_init_info *info)
+{
+ return (info->usage & ISL_SURF_USAGE_DEPTH_BIT) &&
+ (info->format == ISL_FORMAT_R32_FLOAT);
+}
+
+static inline struct isl_extent2d
+isl_extent2d(uint32_t width, uint32_t height)
+{
+ return (struct isl_extent2d) { .w = width, .h = height };
+}
+
+static inline struct isl_extent3d
+isl_extent3d(uint32_t width, uint32_t height, uint32_t depth)
+{
+ return (struct isl_extent3d) { .w = width, .h = height, .d = depth };
+}
+
+static inline struct isl_extent4d
+isl_extent4d(uint32_t width, uint32_t height, uint32_t depth,
+ uint32_t array_len)
+{
+ return (struct isl_extent4d) {
+ .w = width,
+ .h = height,
+ .d = depth,
+ .a = array_len,
+ };
+}
+
+#define isl_surf_init(dev, surf, ...) \
+ isl_surf_init_s((dev), (surf), \
+ &(struct isl_surf_init_info) { __VA_ARGS__ });
+
+bool
+isl_surf_init_s(const struct isl_device *dev,
+ struct isl_surf *surf,
+ const struct isl_surf_init_info *restrict info);
+
+/**
+ * Alignment of the upper-left sample of each subimage, in units of surface
+ * elements.
+ */
+static inline struct isl_extent3d
+isl_surf_get_image_alignment_el(const struct isl_surf *surf)
+{
+ return surf->image_alignment_el;
+}
+
+/**
+ * Alignment of the upper-left sample of each subimage, in units of surface
+ * samples.
+ */
+static inline struct isl_extent3d
+isl_surf_get_image_alignment_sa(const struct isl_surf *surf)
+{
+ const struct isl_format_layout *fmtl = isl_format_get_layout(surf->format);
+
+ return (struct isl_extent3d) {
+ .w = fmtl->bw * surf->image_alignment_el.w,
+ .h = fmtl->bh * surf->image_alignment_el.h,
+ .d = fmtl->bd * surf->image_alignment_el.d,
+ };
+}
+
+/**
+ * Pitch between vertically adjacent surface elements, in bytes.
+ */
+static inline uint32_t
+isl_surf_get_row_pitch(const struct isl_surf *surf)
+{
+ return surf->row_pitch;
+}
+
+/**
+ * Pitch between vertically adjacent surface elements, in units of surface elements.
+ */
+static inline uint32_t
+isl_surf_get_row_pitch_el(const struct isl_surf *surf)
+{
+ const struct isl_format_layout *fmtl = isl_format_get_layout(surf->format);
+
+ assert(surf->row_pitch % fmtl->bs == 0);
+ return surf->row_pitch / fmtl->bs;
+}
+
+/**
+ * Pitch between physical array slices, in rows of surface elements.
+ */
+static inline uint32_t
+isl_surf_get_array_pitch_el_rows(const struct isl_surf *surf)
+{
+ return surf->array_pitch_el_rows;
+}
+
+/**
+ * Pitch between physical array slices, in units of surface elements.
+ */
+static inline uint32_t
+isl_surf_get_array_pitch_el(const struct isl_surf *surf)
+{
+ return isl_surf_get_array_pitch_el_rows(surf) *
+ isl_surf_get_row_pitch_el(surf);
+}
+
+/**
+ * Pitch between physical array slices, in rows of surface samples.
+ */
+static inline uint32_t
+isl_surf_get_array_pitch_sa_rows(const struct isl_surf *surf)
+{
+ const struct isl_format_layout *fmtl = isl_format_get_layout(surf->format);
+ return fmtl->bh * isl_surf_get_array_pitch_el_rows(surf);
+}
+
+/**
+ * Pitch between physical array slices, in bytes.
+ */
+static inline uint32_t
+isl_surf_get_array_pitch(const struct isl_surf *surf)
+{
+ return isl_surf_get_array_pitch_sa_rows(surf) * surf->row_pitch;
+}
+
+/**
+ * Get the offset to an subimage within the surface, in units of surface
+ * samples.
+ *
+ * @invariant level < surface levels
+ * @invariant logical_array_layer < logical array length of surface
+ * @invariant logical_z_offset_px < logical depth of surface at level
+ */
+void
+isl_surf_get_image_offset_sa(const struct isl_surf *surf,
+ uint32_t level,
+ uint32_t logical_array_layer,
+ uint32_t logical_z_offset_px,
+ uint32_t *x_offset_sa,
+ uint32_t *y_offset_sa);
+
+#ifdef __cplusplus
+}
+#endif
--- /dev/null
+/*
+ * Copyright 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <assert.h>
+
+#include "isl.h"
+
+bool
+isl_format_has_sint_channel(enum isl_format fmt)
+{
+ const struct isl_format_layout *fmtl = isl_format_get_layout(fmt);
+
+ return fmtl->channels.r.type == ISL_SINT ||
+ fmtl->channels.g.type == ISL_SINT ||
+ fmtl->channels.b.type == ISL_SINT ||
+ fmtl->channels.a.type == ISL_SINT ||
+ fmtl->channels.l.type == ISL_SINT ||
+ fmtl->channels.i.type == ISL_SINT ||
+ fmtl->channels.p.type == ISL_SINT ||
+ fmtl->channels.g.type == ISL_SINT;
+}
+
+enum isl_format
+isl_format_rgb_to_rgba(enum isl_format rgb)
+{
+ assert(isl_format_is_rgb(rgb));
+
+ switch (rgb) {
+ case ISL_FORMAT_R32G32B32_FLOAT: return ISL_FORMAT_R32G32B32A32_FLOAT;
+ case ISL_FORMAT_R32G32B32_SINT: return ISL_FORMAT_R32G32B32A32_SINT;
+ case ISL_FORMAT_R32G32B32_UINT: return ISL_FORMAT_R32G32B32A32_UINT;
+ case ISL_FORMAT_R32G32B32_UNORM: return ISL_FORMAT_R32G32B32A32_UNORM;
+ case ISL_FORMAT_R32G32B32_SNORM: return ISL_FORMAT_R32G32B32A32_SNORM;
+ case ISL_FORMAT_R32G32B32_SSCALED: return ISL_FORMAT_R32G32B32A32_SSCALED;
+ case ISL_FORMAT_R32G32B32_USCALED: return ISL_FORMAT_R32G32B32A32_USCALED;
+ case ISL_FORMAT_R32G32B32_SFIXED: return ISL_FORMAT_R32G32B32A32_SFIXED;
+ case ISL_FORMAT_R8G8B8_UNORM: return ISL_FORMAT_R8G8B8A8_UNORM;
+ case ISL_FORMAT_R8G8B8_SNORM: return ISL_FORMAT_R8G8B8A8_SNORM;
+ case ISL_FORMAT_R8G8B8_SSCALED: return ISL_FORMAT_R8G8B8A8_SSCALED;
+ case ISL_FORMAT_R8G8B8_USCALED: return ISL_FORMAT_R8G8B8A8_USCALED;
+ case ISL_FORMAT_R16G16B16_FLOAT: return ISL_FORMAT_R16G16B16A16_FLOAT;
+ case ISL_FORMAT_R16G16B16_UNORM: return ISL_FORMAT_R16G16B16A16_UNORM;
+ case ISL_FORMAT_R16G16B16_SNORM: return ISL_FORMAT_R16G16B16A16_SNORM;
+ case ISL_FORMAT_R16G16B16_SSCALED: return ISL_FORMAT_R16G16B16A16_SSCALED;
+ case ISL_FORMAT_R16G16B16_USCALED: return ISL_FORMAT_R16G16B16A16_USCALED;
+ case ISL_FORMAT_R8G8B8_UNORM_SRGB: return ISL_FORMAT_R8G8B8A8_UNORM_SRGB;
+ case ISL_FORMAT_R16G16B16_UINT: return ISL_FORMAT_R16G16B16A16_UINT;
+ case ISL_FORMAT_R16G16B16_SINT: return ISL_FORMAT_R16G16B16A16_SINT;
+ case ISL_FORMAT_R8G8B8_UINT: return ISL_FORMAT_R8G8B8A8_UINT;
+ case ISL_FORMAT_R8G8B8_SINT: return ISL_FORMAT_R8G8B8A8_SINT;
+ default:
+ return ISL_FORMAT_UNSUPPORTED;
+ }
+}
+
+enum isl_format
+isl_format_rgb_to_rgbx(enum isl_format rgb)
+{
+ assert(isl_format_is_rgb(rgb));
+
+ switch (rgb) {
+ case ISL_FORMAT_R32G32B32_FLOAT:
+ return ISL_FORMAT_R32G32B32X32_FLOAT;
+ case ISL_FORMAT_R16G16B16_UNORM:
+ return ISL_FORMAT_R16G16B16X16_UNORM;
+ case ISL_FORMAT_R16G16B16_FLOAT:
+ return ISL_FORMAT_R16G16B16X16_FLOAT;
+ case ISL_FORMAT_R8G8B8_UNORM:
+ return ISL_FORMAT_R8G8B8X8_UNORM;
+ case ISL_FORMAT_R8G8B8_UNORM_SRGB:
+ return ISL_FORMAT_R8G8B8X8_UNORM_SRGB;
+ default:
+ return ISL_FORMAT_UNSUPPORTED;
+ }
+}
--- /dev/null
+# Copyright 2015 Intel Corporation
+#
+# Permission is hereby granted, free of charge, to any person obtaining a
+# copy of this software and associated documentation files (the "Software"),
+# to deal in the Software without restriction, including without limitation
+# the rights to use, copy, modify, merge, publish, distribute, sublicense,
+# and/or sell copies of the Software, and to permit persons to whom the
+# Software is furnished to do so, subject to the following conditions:
+#
+# The above copyright notice and this permission notice (including the next
+# paragraph) shall be included in all copies or substantial portions of the
+# Software.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+# THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+# IN THE SOFTWARE.
+
+#
+# @file
+# @brief Layout of all hardware surface formats
+#
+# For the official list, see Broadwell PRM: Volume 2b: Command Reference:
+# Enumerations: SURFACE_FORMAT.
+#
+
+
+# Columns:
+# name: format name in PRM
+# bpb: bits per block
+# bw: block width, in pixels
+# bh: block height, in pixels
+# bd: block depth, in pixels
+# r: red channel, data type and bitwidth
+# g: green channel
+# b: blue channel
+# a: alpha channel
+# l: luminance channel
+# i: intensity channel
+# p: palette channel
+# space: colorspace
+# txc: texture compression
+#
+# Data Types:
+# x: void
+# r: raw
+# un: unorm
+# sn: snorm
+# uf: ufloat
+# sf: sfloat
+# ux: ufixed
+# sx: sfixed
+# ui: uint
+# si: sint
+# us: uscaled
+# ss: sscaled
+
+
+# Table is aligned with the Vim commands below, using the Align plugin:
+# :AlignCtrl lr+ p8000000000000P1
+# /^# name/,$ Align,
+
+# name , bpb, bw, bh, bd, r, g, b, a, l, i, p, space, txc
+R32G32B32A32_FLOAT , 128, 1, 1, 1, sf32, sf32, sf32, sf32, , , , linear,
+R32G32B32A32_SINT , 128, 1, 1, 1, si32, si32, si32, si32, , , , linear,
+R32G32B32A32_UINT , 128, 1, 1, 1, ui32, ui32, ui32, ui32, , , , linear,
+R32G32B32A32_UNORM , 128, 1, 1, 1, un32, un32, un32, un32, , , , linear,
+R32G32B32A32_SNORM , 128, 1, 1, 1, sn32, sn32, sn32, sn32, , , , linear,
+R64G64_FLOAT , 128, 1, 1, 1, sf64, sf64, , , , , , linear,
+R32G32B32X32_FLOAT , 128, 1, 1, 1, sf32, sf32, sf32, x32, , , , linear,
+R32G32B32A32_SSCALED , 128, 1, 1, 1, ss32, ss32, ss32, ss32, , , , linear,
+R32G32B32A32_USCALED , 128, 1, 1, 1, us32, us32, us32, us32, , , , linear,
+R32G32B32A32_SFIXED , 128, 1, 1, 1, sx32, sx32, sx32, sx32, , , , linear,
+R64G64_PASSTHRU , 128, 1, 1, 1, r64, r64, , , , , , ,
+R32G32B32_FLOAT , 96, 1, 1, 1, sf32, sf32, sf32, , , , , linear,
+R32G32B32_SINT , 96, 1, 1, 1, si32, si32, si32, , , , , linear,
+R32G32B32_UINT , 96, 1, 1, 1, ui32, ui32, ui32, , , , , linear,
+R32G32B32_UNORM , 96, 1, 1, 1, un32, un32, un32, , , , , linear,
+R32G32B32_SNORM , 96, 1, 1, 1, sn32, sn32, sn32, , , , , linear,
+R32G32B32_SSCALED , 96, 1, 1, 1, ss32, ss32, ss32, , , , , linear,
+R32G32B32_USCALED , 96, 1, 1, 1, us32, us32, us32, , , , , linear,
+R32G32B32_SFIXED , 96, 1, 1, 1, sx32, sx32, sx32, , , , , linear,
+R16G16B16A16_UNORM , 64, 1, 1, 1, un16, un16, un16, un16, , , , linear,
+R16G16B16A16_SNORM , 64, 1, 1, 1, sn16, sn16, sn16, sn16, , , , linear,
+R16G16B16A16_SINT , 64, 1, 1, 1, si16, si16, si16, si16, , , , linear,
+R16G16B16A16_UINT , 64, 1, 1, 1, ui16, ui16, ui16, ui16, , , , linear,
+R16G16B16A16_FLOAT , 64, 1, 1, 1, sf16, sf16, sf16, sf16, , , , linear,
+R32G32_FLOAT , 64, 1, 1, 1, sf32, sf32, , , , , , linear,
+R32G32_SINT , 64, 1, 1, 1, si32, si32, , , , , , linear,
+R32G32_UINT , 64, 1, 1, 1, ui32, ui32, , , , , , linear,
+R32_FLOAT_X8X24_TYPELESS , 64, 1, 1, 1, sf32, x8, x24, , , , , linear,
+X32_TYPELESS_G8X24_UINT , 64, 1, 1, 1, x32, ui8, x24, , , , , linear,
+L32A32_FLOAT , 64, 1, 1, 1, , , , sf32, sf32, , , linear,
+R32G32_UNORM , 64, 1, 1, 1, un32, un32, , , , , , linear,
+R32G32_SNORM , 64, 1, 1, 1, sn32, sn32, , , , , , linear,
+R64_FLOAT , 64, 1, 1, 1, sf64, , , , , , , linear,
+R16G16B16X16_UNORM , 64, 1, 1, 1, un16, un16, un16, x16, , , , linear,
+R16G16B16X16_FLOAT , 64, 1, 1, 1, sf16, sf16, sf16, x16, , , , linear,
+A32X32_FLOAT , 64, 1, 1, 1, , , , sf32, x32, , , alpha,
+L32X32_FLOAT , 64, 1, 1, 1, , , , x32, sf32, , , linear,
+I32X32_FLOAT , 64, 1, 1, 1, , , , x32, , sf32, , linear,
+R16G16B16A16_SSCALED , 64, 1, 1, 1, ss16, ss16, ss16, ss16, , , , linear,
+R16G16B16A16_USCALED , 64, 1, 1, 1, us16, us16, us16, us16, , , , linear,
+R32G32_SSCALED , 64, 1, 1, 1, ss32, ss32, , , , , , linear,
+R32G32_USCALED , 64, 1, 1, 1, us32, us32, , , , , , linear,
+R32G32_SFIXED , 64, 1, 1, 1, sx32, sx32, , , , , , linear,
+R64_PASSTHRU , 64, 1, 1, 1, r64, , , , , , , ,
+B8G8R8A8_UNORM , 32, 1, 1, 1, un8, un8, un8, un8, , , , linear,
+B8G8R8A8_UNORM_SRGB , 32, 1, 1, 1, un8, un8, un8, un8, , , , srgb,
+R10G10B10A2_UNORM , 32, 1, 1, 1, un10, un10, un10, un2, , , , linear,
+R10G10B10A2_UNORM_SRGB , 32, 1, 1, 1, un10, un10, un10, un2, , , , srgb,
+R10G10B10A2_UINT , 32, 1, 1, 1, ui10, ui10, ui10, ui2, , , , linear,
+R10G10B10_SNORM_A2_UNORM , 32, 1, 1, 1, sn10, sn10, sn10, un2, , , , linear,
+R8G8B8A8_UNORM , 32, 1, 1, 1, un8, un8, un8, un8, , , , linear,
+R8G8B8A8_UNORM_SRGB , 32, 1, 1, 1, un8, un8, un8, un8, , , , srgb,
+R8G8B8A8_SNORM , 32, 1, 1, 1, sn8, sn8, sn8, sn8, , , , linear,
+R8G8B8A8_SINT , 32, 1, 1, 1, si8, si8, si8, si8, , , , linear,
+R8G8B8A8_UINT , 32, 1, 1, 1, ui8, ui8, ui8, ui8, , , , linear,
+R16G16_UNORM , 32, 1, 1, 1, un16, un16, , , , , , linear,
+R16G16_SNORM , 32, 1, 1, 1, sn16, sn16, , , , , , linear,
+R16G16_SINT , 32, 1, 1, 1, si16, si16, , , , , , linear,
+R16G16_UINT , 32, 1, 1, 1, ui16, ui16, , , , , , linear,
+R16G16_FLOAT , 32, 1, 1, 1, sf16, sf16, , , , , , linear,
+B10G10R10A2_UNORM , 32, 1, 1, 1, un10, un10, un10, un2, , , , linear,
+B10G10R10A2_UNORM_SRGB , 32, 1, 1, 1, un10, un10, un10, un2, , , , srgb,
+R11G11B10_FLOAT , 32, 1, 1, 1, sf11, sf11, sf10, , , , , linear,
+R32_SINT , 32, 1, 1, 1, si32, , , , , , , linear,
+R32_UINT , 32, 1, 1, 1, ui32, , , , , , , linear,
+R32_FLOAT , 32, 1, 1, 1, sf32, , , , , , , linear,
+R24_UNORM_X8_TYPELESS , 32, 1, 1, 1, un24, x8, , , , , , linear,
+X24_TYPELESS_G8_UINT , 32, 1, 1, 1, x24, ui8, , , , , , linear,
+L32_UNORM , 32, 1, 1, 1, , , , , un32, , , linear,
+A32_UNORM , 32, 1, 1, 1, , , , un32, , , , alpha,
+L16A16_UNORM , 32, 1, 1, 1, , , , un16, un16, , , linear,
+I24X8_UNORM , 32, 1, 1, 1, , , , x8, , un24, , linear,
+L24X8_UNORM , 32, 1, 1, 1, , , , x8, un24, , , linear,
+A24X8_UNORM , 32, 1, 1, 1, , , , un24, x8, , , alpha,
+I32_FLOAT , 32, 1, 1, 1, , , , , , sf32, , linear,
+L32_FLOAT , 32, 1, 1, 1, , , , , sf32, , , linear,
+A32_FLOAT , 32, 1, 1, 1, , , , sf32, , , , alpha,
+X8B8_UNORM_G8R8_SNORM , 32, 1, 1, 1, sn8, sn8, un8, x8, , , , linear,
+A8X8_UNORM_G8R8_SNORM , 32, 1, 1, 1, sn8, sn8, x8, un8, , , , linear,
+B8X8_UNORM_G8R8_SNORM , 32, 1, 1, 1, sn8, sn8, un8, x8, , , , linear,
+B8G8R8X8_UNORM , 32, 1, 1, 1, un8, un8, un8, x8, , , , linear,
+B8G8R8X8_UNORM_SRGB , 32, 1, 1, 1, un8, un8, un8, x8, , , , srgb,
+R8G8B8X8_UNORM , 32, 1, 1, 1, un8, un8, un8, x8, , , , linear,
+R8G8B8X8_UNORM_SRGB , 32, 1, 1, 1, un8, un8, un8, x8, , , , srgb,
+R9G9B9E5_SHAREDEXP , 32, 1, 1, 1, ui9, ui9, ui9, , , , , linear,
+B10G10R10X2_UNORM , 32, 1, 1, 1, un10, un10, un10, x2, , , , linear,
+L16A16_FLOAT , 32, 1, 1, 1, , , , sf16, sf16, , , linear,
+R32_UNORM , 32, 1, 1, 1, un32, , , , , , , linear,
+R32_SNORM , 32, 1, 1, 1, sn32, , , , , , , linear,
+R10G10B10X2_USCALED , 32, 1, 1, 1, us10, us10, us10, x2, , , , linear,
+R8G8B8A8_SSCALED , 32, 1, 1, 1, ss8, ss8, ss8, ss8, , , , linear,
+R8G8B8A8_USCALED , 32, 1, 1, 1, us8, us8, us8, us8, , , , linear,
+R16G16_SSCALED , 32, 1, 1, 1, ss16, ss6, , , , , , linear,
+R16G16_USCALED , 32, 1, 1, 1, us16, us16, , , , , , linear,
+R32_SSCALED , 32, 1, 1, 1, ss32, , , , , , , linear,
+R32_USCALED , 32, 1, 1, 1, us32, , , , , , , linear,
+B5G6R5_UNORM , 16, 1, 1, 1, un5, un6, un5, , , , , linear,
+B5G6R5_UNORM_SRGB , 16, 1, 1, 1, un5, un6, un5, , , , , srgb,
+B5G5R5A1_UNORM , 16, 1, 1, 1, un5, un5, un5, un1, , , , linear,
+B5G5R5A1_UNORM_SRGB , 16, 1, 1, 1, un5, un5, un5, un1, , , , srgb,
+B4G4R4A4_UNORM , 16, 1, 1, 1, un4, un4, un4, un4, , , , linear,
+B4G4R4A4_UNORM_SRGB , 16, 1, 1, 1, un4, un4, un4, un4, , , , srgb,
+R8G8_UNORM , 16, 1, 1, 1, un8, un8, , , , , , linear,
+R8G8_SNORM , 16, 1, 1, 1, sn8, sn8, , , , , , linear,
+R8G8_SINT , 16, 1, 1, 1, si8, si8, , , , , , linear,
+R8G8_UINT , 16, 1, 1, 1, ui8, ui8, , , , , , linear,
+R16_UNORM , 16, 1, 1, 1, un16, , , , , , , linear,
+R16_SNORM , 16, 1, 1, 1, sn16, , , , , , , linear,
+R16_SINT , 16, 1, 1, 1, si16, , , , , , , linear,
+R16_UINT , 16, 1, 1, 1, ui16, , , , , , , linear,
+R16_FLOAT , 16, 1, 1, 1, sf16, , , , , , , linear,
+A8P8_UNORM_PALETTE0 , 16, 1, 1, 1, , , , un8, , , un8, linear,
+A8P8_UNORM_PALETTE1 , 16, 1, 1, 1, , , , un8, , , un8, linear,
+I16_UNORM , 16, 1, 1, 1, , , , , , un16, , linear,
+L16_UNORM , 16, 1, 1, 1, , , , , un16, , , linear,
+A16_UNORM , 16, 1, 1, 1, , , , un16, , , , alpha,
+L8A8_UNORM , 16, 1, 1, 1, , , , un8, un8, , , linear,
+I16_FLOAT , 16, 1, 1, 1, , , , , , sf16, , linear,
+L16_FLOAT , 16, 1, 1, 1, , , , , sf16, , , linear,
+A16_FLOAT , 16, 1, 1, 1, , , , sf16, , , , alpha,
+L8A8_UNORM_SRGB , 16, 1, 1, 1, , , , un8, un8, , , srgb,
+R5G5_SNORM_B6_UNORM , 16, 1, 1, 1, sn5, sn5, un6, , , , , linear,
+B5G5R5X1_UNORM , 16, 1, 1, 1, un5, un5, un5, x1, , , , linear,
+B5G5R5X1_UNORM_SRGB , 16, 1, 1, 1, un5, un5, un5, x1, , , , srgb,
+R8G8_SSCALED , 16, 1, 1, 1, ss8, ss8, , , , , , linear,
+R8G8_USCALED , 16, 1, 1, 1, us8, us8, , , , , , linear,
+R16_SSCALED , 16, 1, 1, 1, ss16, , , , , , , linear,
+R16_USCALED , 16, 1, 1, 1, us16, , , , , , , linear,
+P8A8_UNORM_PALETTE0 , 16, 1, 1, 1, , , , un8, , , un8, linear,
+P8A8_UNORM_PALETTE1 , 16, 1, 1, 1, , , , un8, , , un8, linear,
+A1B5G5R5_UNORM , 16, 1, 1, 1, un5, un5, un5, un1, , , , linear,
+A4B4G4R4_UNORM , 16, 1, 1, 1, un4, un4, un4, un4, , , , linear,
+L8A8_UINT , 16, 1, 1, 1, , , , ui8, ui8, , , linear,
+L8A8_SINT , 16, 1, 1, 1, , , , si8, si8, , , linear,
+R8_UNORM , 8, 1, 1, 1, un8, , , , , , , linear,
+R8_SNORM , 8, 1, 1, 1, sn8, , , , , , , linear,
+R8_SINT , 8, 1, 1, 1, si8, , , , , , , linear,
+R8_UINT , 8, 1, 1, 1, ui8, , , , , , , linear,
+A8_UNORM , 8, 1, 1, 1, , , , un8, , , , alpha,
+I8_UNORM , 8, 1, 1, 1, , , , , , un8, , linear,
+L8_UNORM , 8, 1, 1, 1, , , , , un8, , , linear,
+P4A4_UNORM_PALETTE0 , 8, 1, 1, 1, , , , un4, , , un4, linear,
+A4P4_UNORM_PALETTE0 , 8, 1, 1, 1, , , , un4, , , un4, linear,
+R8_SSCALED , 8, 1, 1, 1, ss8, , , , , , , linear,
+R8_USCALED , 8, 1, 1, 1, us8, , , , , , , linear,
+P8_UNORM_PALETTE0 , 8, 1, 1, 1, , , , , , , un8, linear,
+L8_UNORM_SRGB , 8, 1, 1, 1, , , , , un8, , , linear,
+P8_UNORM_PALETTE1 , 8, 1, 1, 1, , , , , , , un8, linear,
+P4A4_UNORM_PALETTE1 , 8, 1, 1, 1, , , , un4, , , un4, linear,
+A4P4_UNORM_PALETTE1 , 8, 1, 1, 1, , , , un4, , , un4, linear,
+Y8_UNORM , 0, 0, 0, 0, , , , , , , , yuv,
+L8_UINT , 8, 1, 1, 1, , , , , ui8, , , linear,
+L8_SINT , 8, 1, 1, 1, , , , , si8, , , linear,
+I8_UINT , 8, 1, 1, 1, , , , , , ui8, , linear,
+I8_SINT , 8, 1, 1, 1, , , , , , si8, , linear,
+DXT1_RGB_SRGB , 64, 4, 4, 1, un4, un4, un4, , , , , srgb, dxt1
+R1_UNORM , 1, 1, 1, 1, un1, , , , , , , linear,
+YCRCB_NORMAL , 0, 0, 0, 0, , , , , , , , yuv,
+YCRCB_SWAPUVY , 0, 0, 0, 0, , , , , , , , yuv,
+P2_UNORM_PALETTE0 , 2, 1, 1, 1, , , , , , , un2, linear,
+P2_UNORM_PALETTE1 , 2, 1, 1, 1, , , , , , , un2, linear,
+BC1_UNORM , 64, 4, 4, 1, un4, un4, un4, un4, , , , linear, dxt1
+BC2_UNORM , 128, 4, 4, 1, un4, un4, un4, un4, , , , linear, dxt3
+BC3_UNORM , 128, 4, 4, 1, un4, un4, un4, un4, , , , linear, dxt5
+BC4_UNORM , 64, 4, 4, 1, un8, , , , , , , linear, rgtc1
+BC5_UNORM , 128, 4, 4, 1, un8, un8, , , , , , linear, rgtc2
+BC1_UNORM_SRGB , 64, 4, 4, 1, un4, un4, un4, un4, , , , srgb, dxt1
+BC2_UNORM_SRGB , 128, 4, 4, 1, un4, un4, un4, un4, , , , srgb, dxt3
+BC3_UNORM_SRGB , 128, 4, 4, 1, un4, un4, un4, un4, , , , srgb, dxt5
+MONO8 , 1, 1, 1, 1, , , , , , , , ,
+YCRCB_SWAPUV , 0, 0, 0, 0, , , , , , , , yuv,
+YCRCB_SWAPY , 0, 0, 0, 0, , , , , , , , yuv,
+DXT1_RGB , 64, 4, 4, 1, un4, un4, un4, , , , , linear, dxt1
+FXT1 , 128, 8, 4, 1, un4, un4, un4, , , , , linear, fxt1
+R8G8B8_UNORM , 24, 1, 1, 1, un8, un8, un8, , , , , linear,
+R8G8B8_SNORM , 24, 1, 1, 1, sn8, sn8, sn8, , , , , linear,
+R8G8B8_SSCALED , 24, 1, 1, 1, ss8, ss8, ss8, , , , , linear,
+R8G8B8_USCALED , 24, 1, 1, 1, us8, us8, us8, , , , , linear,
+R64G64B64A64_FLOAT , 256, 1, 1, 1, sf64, sf64, sf64, sf64, , , , linear,
+R64G64B64_FLOAT , 196, 1, 1, 1, sf64, sf64, sf64, , , , , linear,
+BC4_SNORM , 64, 4, 4, 1, sn8, , , , , , , linear, rgtc1
+BC5_SNORM , 128, 4, 4, 1, sn8, sn8, , , , , , linear, rgtc2
+R16G16B16_FLOAT , 48, 1, 1, 1, sf16, sf16, sf16, , , , , linear,
+R16G16B16_UNORM , 48, 1, 1, 1, un16, un16, un16, , , , , linear,
+R16G16B16_SNORM , 48, 1, 1, 1, sn16, sn16, sn16, , , , , linear,
+R16G16B16_SSCALED , 48, 1, 1, 1, ss16, ss16, ss16, , , , , linear,
+R16G16B16_USCALED , 48, 1, 1, 1, us16, us16, us16, , , , , linear,
+BC6H_SF16 , 128, 4, 4, 1, sf16, sf16, sf16, , , , , linear, bptc
+BC7_UNORM , 128, 4, 4, 1, un8, un8, un8, un8, , , , linear, bptc
+BC7_UNORM_SRGB , 128, 4, 4, 1, un8, un8, un8, un8, , , , srgb, bptc
+BC6H_UF16 , 128, 4, 4, 1, uf16, uf16, uf16, , , , , linear, bptc
+PLANAR_420_8 , 0, 0, 0, 0, , , , , , , , yuv,
+R8G8B8_UNORM_SRGB , 24, 1, 1, 1, un8, un8, un8, , , , , srgb,
+ETC1_RGB8 , 64, 4, 4, 1, un8, un8, un8, , , , , linear, etc1
+ETC2_RGB8 , 64, 4, 4, 1, un8, un8, un8, , , , , linear, etc2
+EAC_R11 , 64, 4, 4, 1, un11, , , , , , , linear, etc2
+EAC_RG11 , 128, 4, 4, 1, un11, un11, , , , , , linear, etc2
+EAC_SIGNED_R11 , 64, 4, 4, 1, sn11, , , , , , , linear, etc2
+EAC_SIGNED_RG11 , 128, 4, 4, 1, sn11, sn11, , , , , , linear, etc2
+ETC2_SRGB8 , 64, 4, 4, 1, un8, un8, un8, , , , , srgb, etc2
+R16G16B16_UINT , 48, 1, 1, 1, ui16, ui16, ui16, , , , , linear,
+R16G16B16_SINT , 48, 1, 1, 1, si16, si16, si16, , , , , linear,
+R32_SFIXED , 32, 1, 1, 1, sx16, , , , , , , linear,
+R10G10B10A2_SNORM , 32, 1, 1, 1, sn10, sn10, sn10, sn2, , , , linear,
+R10G10B10A2_USCALED , 32, 1, 1, 1, us10, us10, us10, us2, , , , linear,
+R10G10B10A2_SSCALED , 32, 1, 1, 1, ss10, ss10, ss10, ss2, , , , linear,
+R10G10B10A2_SINT , 32, 1, 1, 1, si10, si10, si10, si2, , , , linear,
+B10G10R10A2_SNORM , 32, 1, 1, 1, sn10, sn10, sn10, sn2, , , , linear,
+B10G10R10A2_USCALED , 32, 1, 1, 1, us10, us10, us10, us2, , , , linear,
+B10G10R10A2_SSCALED , 32, 1, 1, 1, ss10, ss10, ss10, ss2, , , , linear,
+B10G10R10A2_UINT , 32, 1, 1, 1, ui10, ui10, ui10, ui2, , , , linear,
+B10G10R10A2_SINT , 32, 1, 1, 1, si10, si10, si10, si2, , , , linear,
+R64G64B64A64_PASSTHRU , 256, 1, 1, 1, r64, r64, r64, r64, , , , ,
+R64G64B64_PASSTHRU , 192, 1, 1, 1, r64, r64, r64, , , , , ,
+ETC2_RGB8_PTA , 64, 4, 4, 1, un8, un8, un8, un1, , , , linear, etc2
+ETC2_SRGB8_PTA , 64, 4, 4, 1, un8, un8, un8, un1, , , , srgb, etc2
+ETC2_EAC_RGBA8 , 128, 4, 4, 1, un8, un8, un8, un8, , , , linear, etc2
+ETC2_EAC_SRGB8_A8 , 128, 4, 4, 1, un8, un8, un8, un8, , , , srgb, etc2
+R8G8B8_UINT , 24, 1, 1, 1, ui8, ui8, ui8, , , , , linear,
+R8G8B8_SINT , 24, 1, 1, 1, si8, si8, si8, , , , , linear,
+RAW , 0, 0, 0, 0, , , , , , , , ,
--- /dev/null
+#!/usr/bin/env bash
+#
+# Copyright 2015 Intel Corporation
+#
+# Permission is hereby granted, free of charge, to any person obtaining a
+# copy of this software and associated documentation files (the "Software"),
+# to deal in the Software without restriction, including without limitation
+# the rights to use, copy, modify, merge, publish, distribute, sublicense,
+# and/or sell copies of the Software, and to permit persons to whom the
+# Software is furnished to do so, subject to the following conditions:
+#
+# The above copyright notice and this permission notice (including the next
+# paragraph) shall be included in all copies or substantial portions of the
+# Software.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+# THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+# IN THE SOFTWARE.
+
+set -eu
+set -o pipefail
+
+cat <<'EOF'
+/*
+ * Copyright 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include "isl.h"
+
+const struct isl_format_layout
+isl_format_layouts[] = {
+EOF
+
+sed -r '
+# Delete comment lines and empty lines
+/^[[:space:]]*#/d
+/^[[:space:]]*$/d
+
+# Delete spaces
+s/[[:space:]]//g
+
+# Translate formats
+s/^([A-Za-z0-9_]+),*/ISL_FORMAT_\1,/
+
+# Translate data type of channels
+s/\<x([0-9]+),/ISL_VOID@\1,/g
+s/\<r([0-9]+),/ISL_RAW@\1,/g
+s/\<un([0-9]+),/ISL_UNORM@\1,/g
+s/\<sn([0-9]+),/ISL_SNORM@\1,/g
+s/\<uf([0-9]+),/ISL_UFLOAT@\1,/g
+s/\<sf([0-9]+),/ISL_SFLOAT@\1,/g
+s/\<ux([0-9]+),/ISL_UFIXED@\1,/g
+s/\<sx([0-9]+),/ISL_SFIXED@\1,/g
+s/\<ui([0-9]+),/ISL_UINT@\1,/g
+s/\<si([0-9]+),/ISL_SINT@\1,/g
+s/\<us([0-9]+),/ISL_USCALED@\1,/g
+s/\<ss([0-9]+),/ISL_SSCALED@\1,/g
+
+# Translate colorspaces
+# Interpret alpha-only formats as having no colorspace.
+s/\<(linear|srgb|yuv)\>/ISL_COLORSPACE_\1/
+s/\<alpha\>//
+
+# Translate texture compression
+s/\<(dxt|fxt|rgtc|bptc|etc)([0-9]*)\>/ISL_TXC_\1\2/
+' |
+tr 'a-z' 'A-Z' | # Convert to uppersace
+while IFS=, read -r format bpb bw bh bd \
+ red green blue alpha \
+ luminance intensity palette \
+ colorspace txc
+do
+ : ${colorspace:=ISL_COLORSPACE_NONE}
+ : ${txc:=ISL_TXC_NONE}
+
+ cat <<EOF
+ [$format] = {
+ $format,
+ .bs = $((bpb/8)),
+ .bw = $bw, .bh = $bh, .bd = $bd,
+ .channels = {
+ .r = { $red },
+ .g = { $green },
+ .b = { $blue },
+ .a = { $alpha },
+ .l = { $luminance },
+ .i = { $intensity },
+ .p = { $palette },
+ },
+ .colorspace = $colorspace,
+ .txc = $txc,
+ },
+
+EOF
+done |
+sed -r '
+# Collapse empty channels
+s/\{ \}/{}/
+
+# Split non-empty channels into two members: base type and bit size
+s/@/, /
+'
+
+# Terminate the table
+printf '};\n'
--- /dev/null
+/*
+ * Copyright 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include "isl_gen4.h"
+#include "isl_priv.h"
+
+bool
+gen4_choose_msaa_layout(const struct isl_device *dev,
+ const struct isl_surf_init_info *info,
+ enum isl_tiling tiling,
+ enum isl_msaa_layout *msaa_layout)
+{
+ /* Gen4 and Gen5 do not support MSAA */
+ assert(info->samples >= 1);
+
+ *msaa_layout = ISL_MSAA_LAYOUT_NONE;
+ return true;
+}
+
+void
+gen4_choose_image_alignment_el(const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info,
+ enum isl_tiling tiling,
+ enum isl_msaa_layout msaa_layout,
+ struct isl_extent3d *image_align_el)
+{
+ assert(info->samples == 1);
+ assert(msaa_layout == ISL_MSAA_LAYOUT_NONE);
+ assert(!isl_tiling_is_std_y(tiling));
+
+ /* Note that neither the surface's horizontal nor vertical image alignment
+ * is programmable on gen4 nor gen5.
+ *
+ * From the G35 PRM (2008-01), Volume 1 Graphics Core, Section 6.17.3.4
+ * Alignment Unit Size:
+ *
+ * Note that the compressed formats are padded to a full compression
+ * cell.
+ *
+ * +------------------------+--------+--------+
+ * | format | halign | valign |
+ * +------------------------+--------+--------+
+ * | YUV 4:2:2 formats | 4 | 2 |
+ * | uncompressed formats | 4 | 2 |
+ * +------------------------+--------+--------+
+ */
+
+ if (isl_format_is_compressed(info->format)) {
+ *image_align_el = isl_extent3d(1, 1, 1);
+ return;
+ }
+
+ *image_align_el = isl_extent3d(4, 2, 1);
+}
--- /dev/null
+/*
+ * Copyright 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#pragma once
+
+#include "isl_priv.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+bool
+gen4_choose_msaa_layout(const struct isl_device *dev,
+ const struct isl_surf_init_info *info,
+ enum isl_tiling tiling,
+ enum isl_msaa_layout *msaa_layout);
+
+void
+gen4_choose_image_alignment_el(const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info,
+ enum isl_tiling tiling,
+ enum isl_msaa_layout msaa_layout,
+ struct isl_extent3d *image_align_el);
+
+#ifdef __cplusplus
+}
+#endif
--- /dev/null
+/*
+ * Copyright 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include "isl_gen6.h"
+#include "isl_priv.h"
+
+bool
+gen6_choose_msaa_layout(const struct isl_device *dev,
+ const struct isl_surf_init_info *info,
+ enum isl_tiling tiling,
+ enum isl_msaa_layout *msaa_layout)
+{
+ const struct isl_format_layout *fmtl = isl_format_get_layout(info->format);
+
+ assert(ISL_DEV_GEN(dev) == 6);
+ assert(info->samples >= 1);
+
+ if (info->samples == 1) {
+ *msaa_layout = ISL_MSAA_LAYOUT_NONE;
+ return false;
+ }
+
+ /* From the Sandybridge PRM, Volume 4 Part 1 p72, SURFACE_STATE, Surface
+ * Format:
+ *
+ * If Number of Multisamples is set to a value other than
+ * MULTISAMPLECOUNT_1, this field cannot be set to the following
+ * formats:
+ *
+ * - any format with greater than 64 bits per element
+ * - any compressed texture format (BC*)
+ * - any YCRCB* format
+ */
+ if (fmtl->bs > 8)
+ return false;
+ if (isl_format_is_compressed(info->format))
+ return false;
+ if (isl_format_is_yuv(info->format))
+ return false;
+
+ /* From the Sandybridge PRM, Volume 4 Part 1 p85, SURFACE_STATE, Number of
+ * Multisamples:
+ *
+ * If this field is any value other than MULTISAMPLECOUNT_1 the
+ * following restrictions apply:
+ *
+ * - the Surface Type must be SURFTYPE_2D
+ * - [...]
+ */
+ if (info->dim != ISL_SURF_DIM_2D)
+ return false;
+
+ /* More obvious restrictions */
+ if (isl_surf_usage_is_display(info->usage))
+ return false;
+ if (tiling == ISL_TILING_LINEAR)
+ return false;
+ if (info->levels > 1)
+ return false;
+
+ *msaa_layout = ISL_MSAA_LAYOUT_INTERLEAVED;
+ return true;
+}
+
+void
+gen6_choose_image_alignment_el(const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info,
+ enum isl_tiling tiling,
+ enum isl_msaa_layout msaa_layout,
+ struct isl_extent3d *image_align_el)
+{
+ /* Note that the surface's horizontal image alignment is not programmable
+ * on Sandybridge.
+ *
+ * From the Sandybridge PRM (2011-05), Volume 1, Part 1, Section 7.18.3.4
+ * Alignment Unit Size:
+ *
+ * Note that the compressed formats are padded to a full compression cell.
+ *
+ * +------------------------+--------+--------+
+ * | format | halign | valign |
+ * +------------------------+--------+--------+
+ * | YUV 4:2:2 formats | 4 | * |
+ * | uncompressed formats | 4 | * |
+ * +------------------------+--------+--------+
+ *
+ * * For these formats, the vertical alignment factor “j” is determined
+ * as follows:
+ * - j = 4 for any depth buffer
+ * - j = 2 for separate stencil buffer
+ * - j = 4 for any render target surface is multisampled (4x)
+ * - j = 2 for all other render target surface
+ *
+ * From the Sandrybridge PRM (2011-05), Volume 4, Part 1, Section 2.11.2
+ * SURFACE_STATE, Surface Vertical Alignment:
+ *
+ * - This field must be set to VALIGN_2 if the Surface Format is 96 bits
+ * per element (BPE).
+ *
+ * - Value of 1 [VALIGN_4] is not supported for format YCRCB_NORMAL
+ * (0x182), YCRCB_SWAPUVY (0x183), YCRCB_SWAPUV (0x18f), YCRCB_SWAPY
+ * (0x190)
+ */
+
+ if (isl_format_is_compressed(info->format)) {
+ *image_align_el = isl_extent3d(1, 1, 1);
+ return;
+ }
+
+ if (isl_format_is_yuv(info->format)) {
+ *image_align_el = isl_extent3d(4, 2, 1);
+ return;
+ }
+
+ if (info->samples > 1) {
+ *image_align_el = isl_extent3d(4, 4, 1);
+ return;
+ }
+
+ if (isl_surf_usage_is_depth_or_stencil(info->usage) &&
+ !ISL_DEV_USE_SEPARATE_STENCIL(dev)) {
+ /* interleaved depthstencil buffer */
+ *image_align_el = isl_extent3d(4, 4, 1);
+ return;
+ }
+
+ if (isl_surf_usage_is_depth(info->usage)) {
+ /* separate depth buffer */
+ *image_align_el = isl_extent3d(4, 4, 1);
+ return;
+ }
+
+ if (isl_surf_usage_is_stencil(info->usage)) {
+ /* separate stencil buffer */
+ *image_align_el = isl_extent3d(4, 2, 1);
+ return;
+ }
+
+ *image_align_el = isl_extent3d(4, 2, 1);
+}
--- /dev/null
+/*
+ * Copyright 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#pragma once
+
+#include "isl_priv.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+bool
+gen6_choose_msaa_layout(const struct isl_device *dev,
+ const struct isl_surf_init_info *info,
+ enum isl_tiling tiling,
+ enum isl_msaa_layout *msaa_layout);
+
+void
+gen6_choose_image_alignment_el(const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info,
+ enum isl_tiling tiling,
+ enum isl_msaa_layout msaa_layout,
+ struct isl_extent3d *image_align_el);
+
+#ifdef __cplusplus
+}
+#endif
--- /dev/null
+/*
+ * Copyright 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include "isl_gen7.h"
+#include "isl_priv.h"
+
+bool
+gen7_choose_msaa_layout(const struct isl_device *dev,
+ const struct isl_surf_init_info *info,
+ enum isl_tiling tiling,
+ enum isl_msaa_layout *msaa_layout)
+{
+ const struct isl_format_layout *fmtl = isl_format_get_layout(info->format);
+
+ bool require_array = false;
+ bool require_interleaved = false;
+
+ assert(ISL_DEV_GEN(dev) == 7);
+ assert(info->samples >= 1);
+
+ if (info->samples == 1) {
+ *msaa_layout = ISL_MSAA_LAYOUT_NONE;
+ return true;
+ }
+
+ /* From the Ivybridge PRM, Volume 4 Part 1 p63, SURFACE_STATE, Surface
+ * Format:
+ *
+ * If Number of Multisamples is set to a value other than
+ * MULTISAMPLECOUNT_1, this field cannot be set to the following
+ * formats: any format with greater than 64 bits per element, any
+ * compressed texture format (BC*), and any YCRCB* format.
+ */
+ if (fmtl->bs > 8)
+ return false;
+ if (isl_format_is_compressed(info->format))
+ return false;
+ if (isl_format_is_yuv(info->format))
+ return false;
+
+ /* From the Ivybridge PRM, Volume 4 Part 1 p73, SURFACE_STATE, Number of
+ * Multisamples:
+ *
+ * - If this field is any value other than MULTISAMPLECOUNT_1, the
+ * Surface Type must be SURFTYPE_2D.
+ *
+ * - If this field is any value other than MULTISAMPLECOUNT_1, Surface
+ * Min LOD, Mip Count / LOD, and Resource Min LOD must be set to zero
+ */
+ if (info->dim != ISL_SURF_DIM_2D)
+ return false;
+ if (info->levels > 1)
+ return false;
+
+ /* The Ivyrbridge PRM insists twice that signed integer formats cannot be
+ * multisampled.
+ *
+ * From the Ivybridge PRM, Volume 4 Part 1 p73, SURFACE_STATE, Number of
+ * Multisamples:
+ *
+ * - This field must be set to MULTISAMPLECOUNT_1 for SINT MSRTs when
+ * all RT channels are not written.
+ *
+ * And errata from the Ivybridge PRM, Volume 4 Part 1 p77,
+ * RENDER_SURFACE_STATE, MCS Enable:
+ *
+ * This field must be set to 0 [MULTISAMPLECOUNT_1] for all SINT MSRTs
+ * when all RT channels are not written.
+ *
+ * Note that the above SINT restrictions apply only to *MSRTs* (that is,
+ * *multisampled* render targets). The restrictions seem to permit an MCS
+ * if the render target is singlesampled.
+ */
+ if (isl_format_has_sint_channel(info->format))
+ return false;
+
+ /* More obvious restrictions */
+ if (isl_surf_usage_is_display(info->usage))
+ return false;
+ if (tiling == ISL_TILING_LINEAR)
+ return false;
+
+ /* From the Ivybridge PRM, Volume 4 Part 1 p72, SURFACE_STATE, Multisampled
+ * Suface Storage Format:
+ *
+ * +---------------------+----------------------------------------------------------------+
+ * | MSFMT_MSS | Multsampled surface was/is rendered as a render target |
+ * | MSFMT_DEPTH_STENCIL | Multisampled surface was rendered as a depth or stencil buffer |
+ * +---------------------+----------------------------------------------------------------+
+ *
+ * In the table above, MSFMT_MSS refers to ISL_MSAA_LAYOUT_ARRAY, and
+ * MSFMT_DEPTH_STENCIL refers to ISL_MSAA_LAYOUT_INTERLEAVED.
+ */
+ if (isl_surf_usage_is_depth_or_stencil(info->usage))
+ require_interleaved = true;
+
+ /* From the Ivybridge PRM, Volume 4 Part 1 p72, SURFACE_STATE, Multisampled
+ * Suface Storage Format:
+ *
+ * If the surface’s Number of Multisamples is MULTISAMPLECOUNT_8, Width
+ * is >= 8192 (meaning the actual surface width is >= 8193 pixels), this
+ * field must be set to MSFMT_MSS.
+ */
+ if (info->samples == 8 && info->width == 8192)
+ require_array = true;
+
+ /* From the Ivybridge PRM, Volume 4 Part 1 p72, SURFACE_STATE, Multisampled
+ * Suface Storage Format:
+ *
+ * If the surface’s Number of Multisamples is MULTISAMPLECOUNT_8,
+ * ((Depth+1) * (Height+1)) is > 4,194,304, OR if the surface’s Number
+ * of Multisamples is MULTISAMPLECOUNT_4, ((Depth+1) * (Height+1)) is
+ * > 8,388,608, this field must be set to MSFMT_DEPTH_STENCIL.
+ */
+ if ((info->samples == 8 && info->height > 4194304u) ||
+ (info->samples == 4 && info->height > 8388608u))
+ require_interleaved = true;
+
+ /* From the Ivybridge PRM, Volume 4 Part 1 p72, SURFACE_STATE, Multisampled
+ * Suface Storage Format:
+ *
+ * This field must be set to MSFMT_DEPTH_STENCIL if Surface Format is
+ * one of the following: I24X8_UNORM, L24X8_UNORM, A24X8_UNORM, or
+ * R24_UNORM_X8_TYPELESS.
+ */
+ if (info->format == ISL_FORMAT_I24X8_UNORM ||
+ info->format == ISL_FORMAT_L24X8_UNORM ||
+ info->format == ISL_FORMAT_A24X8_UNORM ||
+ info->format == ISL_FORMAT_R24_UNORM_X8_TYPELESS)
+ require_interleaved = true;
+
+ if (require_array && require_interleaved)
+ return false;
+
+ if (require_interleaved) {
+ *msaa_layout = ISL_MSAA_LAYOUT_INTERLEAVED;
+ return true;
+ }
+
+ /* Default to the array layout because it permits multisample
+ * compression.
+ */
+ *msaa_layout = ISL_MSAA_LAYOUT_ARRAY;
+ return true;
+}
+
+static bool
+gen7_format_needs_valign2(const struct isl_device *dev,
+ enum isl_format format)
+{
+ /* This workaround applies only to gen7 */
+ if (ISL_DEV_GEN(dev) > 7)
+ return false;
+
+ /* From the Ivybridge PRM (2012-05-31), Volume 4, Part 1, Section 2.12.1,
+ * RENDER_SURFACE_STATE Surface Vertical Alignment:
+ *
+ * - Value of 1 [VALIGN_4] is not supported for format YCRCB_NORMAL
+ * (0x182), YCRCB_SWAPUVY (0x183), YCRCB_SWAPUV (0x18f), YCRCB_SWAPY
+ * (0x190)
+ *
+ * - VALIGN_4 is not supported for surface format R32G32B32_FLOAT.
+ */
+ return isl_format_is_yuv(format) ||
+ format == ISL_FORMAT_R32G32B32_FLOAT;
+}
+
+/**
+ * @brief Filter out tiling flags that are incompatible with the surface.
+ *
+ * The resultant outgoing @a flags is a subset of the incoming @a flags. The
+ * outgoing flags may be empty (0x0) if the incoming flags were too
+ * restrictive.
+ *
+ * For example, if the surface will be used for a display
+ * (ISL_SURF_USAGE_DISPLAY_BIT), then this function filters out all tiling
+ * flags except ISL_TILING_X_BIT and ISL_TILING_LINEAR_BIT.
+ */
+void
+gen7_filter_tiling(const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info,
+ isl_tiling_flags_t *flags)
+{
+ /* IVB+ requires separate stencil */
+ assert(ISL_DEV_USE_SEPARATE_STENCIL(dev));
+
+ /* Clear flags unsupported on this hardware */
+ if (ISL_DEV_GEN(dev) < 9) {
+ *flags &= ~ISL_TILING_Yf_BIT;
+ *flags &= ~ISL_TILING_Ys_BIT;
+ }
+
+ /* And... clear the Yf and Ys bits anyway because Anvil doesn't support
+ * them yet.
+ */
+ *flags &= ~ISL_TILING_Yf_BIT; /* FINISHME[SKL]: Support Yf */
+ *flags &= ~ISL_TILING_Ys_BIT; /* FINISHME[SKL]: Support Ys */
+
+ if (isl_surf_usage_is_depth(info->usage)) {
+ /* Depth requires Y. */
+ *flags &= ISL_TILING_ANY_Y_MASK;
+ }
+
+ /* Separate stencil requires W tiling, and W tiling requires separate
+ * stencil.
+ */
+ if (isl_surf_usage_is_stencil(info->usage)) {
+ *flags &= ISL_TILING_W_BIT;
+ } else {
+ *flags &= ~ISL_TILING_W_BIT;
+ }
+
+ if (info->usage & (ISL_SURF_USAGE_DISPLAY_ROTATE_90_BIT |
+ ISL_SURF_USAGE_DISPLAY_ROTATE_180_BIT |
+ ISL_SURF_USAGE_DISPLAY_ROTATE_270_BIT)) {
+ assert(*flags & ISL_SURF_USAGE_DISPLAY_BIT);
+ isl_finishme("%s:%s: handle rotated display surfaces",
+ __FILE__, __func__);
+ }
+
+ if (info->usage & (ISL_SURF_USAGE_DISPLAY_FLIP_X_BIT |
+ ISL_SURF_USAGE_DISPLAY_FLIP_Y_BIT)) {
+ assert(*flags & ISL_SURF_USAGE_DISPLAY_BIT);
+ isl_finishme("%s:%s: handle flipped display surfaces",
+ __FILE__, __func__);
+ }
+
+ if (info->usage & ISL_SURF_USAGE_DISPLAY_BIT) {
+ /* Before Skylake, the display engine does not accept Y */
+ /* FINISHME[SKL]: Y tiling for display surfaces */
+ *flags &= (ISL_TILING_LINEAR_BIT | ISL_TILING_X_BIT);
+ }
+
+ if (info->samples > 1) {
+ /* From the Sandybridge PRM, Volume 4 Part 1, SURFACE_STATE Tiled
+ * Surface:
+ *
+ * For multisample render targets, this field must be 1 (true). MSRTs
+ * can only be tiled.
+ *
+ * Multisample surfaces never require X tiling, and Y tiling generally
+ * performs better than X. So choose Y. (Unless it's stencil, then it
+ * must be W).
+ */
+ *flags &= (ISL_TILING_ANY_Y_MASK | ISL_TILING_W_BIT);
+ }
+
+ /* workaround */
+ if (ISL_DEV_GEN(dev) == 7 &&
+ gen7_format_needs_valign2(dev, info->format) &&
+ (info->usage & ISL_SURF_USAGE_RENDER_TARGET_BIT) &&
+ info->samples == 1) {
+ /* Y tiling is illegal. From the Ivybridge PRM, Vol4 Part1 2.12.2.1,
+ * SURFACE_STATE Surface Vertical Alignment:
+ *
+ * This field must be set to VALIGN_4 for all tiled Y Render Target
+ * surfaces.
+ */
+ *flags &= ~ISL_TILING_Y0_BIT;
+ }
+}
+
+/**
+ * Choose horizontal subimage alignment, in units of surface elements.
+ */
+static uint32_t
+gen7_choose_halign_el(const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info)
+{
+ if (isl_format_is_compressed(info->format))
+ return 1;
+
+ /* From the Ivybridge PRM (2012-05-31), Volume 4, Part 1, Section 2.12.1,
+ * RENDER_SURFACE_STATE Surface Hoizontal Alignment:
+ *
+ * - This field is intended to be set to HALIGN_8 only if the surface
+ * was rendered as a depth buffer with Z16 format or a stencil buffer,
+ * since these surfaces support only alignment of 8. Use of HALIGN_8
+ * for other surfaces is supported, but uses more memory.
+ */
+ if (isl_surf_info_is_z16(info) ||
+ isl_surf_usage_is_stencil(info->usage))
+ return 8;
+
+ return 4;
+}
+
+/**
+ * Choose vertical subimage alignment, in units of surface elements.
+ */
+static uint32_t
+gen7_choose_valign_el(const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info,
+ enum isl_tiling tiling)
+{
+ bool require_valign2 = false;
+ bool require_valign4 = false;
+
+ if (isl_format_is_compressed(info->format))
+ return 1;
+
+ if (gen7_format_needs_valign2(dev, info->format))
+ require_valign2 = true;
+
+ /* From the Ivybridge PRM, Volume 4, Part 1, Section 2.12.1:
+ * RENDER_SURFACE_STATE Surface Vertical Alignment:
+ *
+ * - This field is intended to be set to VALIGN_4 if the surface was
+ * rendered as a depth buffer, for a multisampled (4x) render target,
+ * or for a multisampled (8x) render target, since these surfaces
+ * support only alignment of 4. Use of VALIGN_4 for other surfaces is
+ * supported, but uses more memory. This field must be set to
+ * VALIGN_4 for all tiled Y Render Target surfaces.
+ *
+ */
+ if (isl_surf_usage_is_depth(info->usage) ||
+ info->samples > 1 ||
+ tiling == ISL_TILING_Y0) {
+ require_valign4 = true;
+ }
+
+ if (isl_surf_usage_is_stencil(info->usage)) {
+ /* The Ivybridge PRM states that the stencil buffer's vertical alignment
+ * is 8 [Ivybridge PRM, Volume 1, Part 1, Section 6.18.4.4 Alignment
+ * Unit Size]. However, valign=8 is outside the set of valid values of
+ * RENDER_SURFACE_STATE.SurfaceVerticalAlignment, which is VALIGN_2
+ * (0x0) and VALIGN_4 (0x1).
+ *
+ * The PRM is generally confused about the width, height, and alignment
+ * of the stencil buffer; and this confusion appears elsewhere. For
+ * example, the following PRM text effectively converts the stencil
+ * buffer's 8-pixel alignment to a 4-pixel alignment [Ivybridge PRM,
+ * Volume 1, Part 1, Section
+ * 6.18.4.2 Base Address and LOD Calculation]:
+ *
+ * For separate stencil buffer, the width must be mutiplied by 2 and
+ * height divided by 2 as follows:
+ *
+ * w_L = 2*i*ceil(W_L/i)
+ * h_L = 1/2*j*ceil(H_L/j)
+ *
+ * The root of the confusion is that, in W tiling, each pair of rows is
+ * interleaved into one.
+ *
+ * FINISHME(chadv): Decide to set valign=4 or valign=8 after isl's API
+ * is more polished.
+ */
+ require_valign4 = true;
+ }
+
+ assert(!require_valign2 || !require_valign4);
+
+ if (require_valign4)
+ return 4;
+
+ /* Prefer VALIGN_2 because it conserves memory. */
+ return 2;
+}
+
+void
+gen7_choose_image_alignment_el(const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info,
+ enum isl_tiling tiling,
+ enum isl_msaa_layout msaa_layout,
+ struct isl_extent3d *image_align_el)
+{
+ /* IVB+ does not support combined depthstencil. */
+ assert(!isl_surf_usage_is_depth_and_stencil(info->usage));
+
+ *image_align_el = (struct isl_extent3d) {
+ .w = gen7_choose_halign_el(dev, info),
+ .h = gen7_choose_valign_el(dev, info, tiling),
+ .d = 1,
+ };
+}
--- /dev/null
+/*
+ * Copyright 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#pragma once
+
+#include "isl_priv.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+void
+gen7_filter_tiling(const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info,
+ isl_tiling_flags_t *flags);
+
+bool
+gen7_choose_msaa_layout(const struct isl_device *dev,
+ const struct isl_surf_init_info *info,
+ enum isl_tiling tiling,
+ enum isl_msaa_layout *msaa_layout);
+
+void
+gen7_choose_image_alignment_el(const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info,
+ enum isl_tiling tiling,
+ enum isl_msaa_layout msaa_layout,
+ struct isl_extent3d *image_align_el);
+
+#ifdef __cplusplus
+}
+#endif
--- /dev/null
+/*
+ * Copyright 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include "isl_gen8.h"
+#include "isl_priv.h"
+
+bool
+gen8_choose_msaa_layout(const struct isl_device *dev,
+ const struct isl_surf_init_info *info,
+ enum isl_tiling tiling,
+ enum isl_msaa_layout *msaa_layout)
+{
+ bool require_array = false;
+ bool require_interleaved = false;
+
+ assert(info->samples >= 1);
+
+ if (info->samples == 1) {
+ *msaa_layout = ISL_MSAA_LAYOUT_NONE;
+ return true;
+ }
+
+ /* From the Broadwell PRM >> Volume2d: Command Structures >>
+ * RENDER_SURFACE_STATE Tile Mode:
+ *
+ * - If Number of Multisamples is not MULTISAMPLECOUNT_1, this field
+ * must be YMAJOR.
+ *
+ * As usual, though, stencil is special.
+ */
+ if (!isl_tiling_is_std_y(tiling) && !isl_surf_usage_is_stencil(info->usage))
+ return false;
+
+ /* From the Broadwell PRM >> Volume2d: Command Structures >>
+ * RENDER_SURFACE_STATE Multisampled Surface Storage Format:
+ *
+ * All multisampled render target surfaces must have this field set to
+ * MSFMT_MSS
+ */
+ if (info->usage & ISL_SURF_USAGE_RENDER_TARGET_BIT)
+ require_array = true;
+
+ /* From the Broadwell PRM >> Volume2d: Command Structures >>
+ * RENDER_SURFACE_STATE Number of Multisamples:
+ *
+ * - If this field is any value other than MULTISAMPLECOUNT_1, the
+ * Surface Type must be SURFTYPE_2D This field must be set to
+ * MULTISAMPLECOUNT_1 unless the surface is a Sampling Engine surface
+ * or Render Target surface.
+ *
+ * - If this field is any value other than MULTISAMPLECOUNT_1, Surface
+ * Min LOD, Mip Count / LOD, and Resource Min LOD must be set to zero.
+ */
+ if (info->dim != ISL_SURF_DIM_2D)
+ return false;
+ if (info->levels > 1)
+ return false;
+
+ /* More obvious restrictions */
+ if (isl_surf_usage_is_display(info->usage))
+ return false;
+ if (isl_format_is_compressed(info->format))
+ return false;
+ if (isl_format_is_yuv(info->format))
+ return false;
+
+ if (isl_surf_usage_is_depth_or_stencil(info->usage))
+ require_interleaved = true;
+
+ if (require_array && require_interleaved)
+ return false;
+
+ if (require_interleaved) {
+ *msaa_layout = ISL_MSAA_LAYOUT_INTERLEAVED;
+ return true;
+ }
+
+ *msaa_layout = ISL_MSAA_LAYOUT_ARRAY;
+ return true;
+}
+
+/**
+ * Choose horizontal subimage alignment, in units of surface elements.
+ */
+static uint32_t
+gen8_choose_halign_el(const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info)
+{
+ if (isl_format_is_compressed(info->format))
+ return 1;
+
+ /* From the Broadwell PRM, Volume 2d "Command Reference: Structures",
+ * RENDER_SURFACE_STATE Surface Horizontal Alignment, p326:
+ *
+ * - This field is intended to be set to HALIGN_8 only if the surface
+ * was rendered as a depth buffer with Z16 format or a stencil buffer.
+ * In this case it must be set to HALIGN_8 since these surfaces
+ * support only alignment of 8. [...]
+ */
+ if (isl_surf_info_is_z16(info))
+ return 8;
+ if (isl_surf_usage_is_stencil(info->usage))
+ return 8;
+
+ /* From the Broadwell PRM, Volume 2d "Command Reference: Structures",
+ * RENDER_SURFACE_STATE Surface Horizontal Alignment, p326:
+ *
+ * [...] For Z32 formats it must be set to HALIGN_4.
+ */
+ if (isl_surf_usage_is_depth(info->usage))
+ return 4;
+
+ if (!(info->usage & ISL_SURF_USAGE_DISABLE_AUX_BIT)) {
+ /* From the Broadwell PRM, Volume 2d "Command Reference: Structures",
+ * RENDER_SURFACE_STATE Surface Horizontal Alignment, p326:
+ *
+ * - When Auxiliary Surface Mode is set to AUX_CCS_D or AUX_CCS_E,
+ * HALIGN 16 must be used.
+ *
+ * This case handles color surfaces that may own an auxiliary MCS, CCS_D,
+ * or CCS_E. Depth buffers, including those that own an auxiliary HiZ
+ * surface, are handled above and do not require HALIGN_16.
+ */
+ assert(!isl_surf_usage_is_depth(info->usage));
+ return 16;
+ }
+
+ /* XXX(chadv): I believe the hardware requires each image to be
+ * cache-aligned. If that's true, then defaulting to halign=4 is wrong for
+ * many formats. Depending on the format's block size, we may need to
+ * increase halign to 8.
+ */
+ return 4;
+}
+
+/**
+ * Choose vertical subimage alignment, in units of surface elements.
+ */
+static uint32_t
+gen8_choose_valign_el(const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info)
+{
+ /* From the Broadwell PRM > Volume 2d: Command Reference: Structures
+ * > RENDER_SURFACE_STATE Surface Vertical Alignment (p325):
+ *
+ * - For Sampling Engine and Render Target Surfaces: This field
+ * specifies the vertical alignment requirement in elements for the
+ * surface. [...] An element is defined as a pixel in uncompresed
+ * surface formats, and as a compression block in compressed surface
+ * formats. For MSFMT_DEPTH_STENCIL type multisampled surfaces, an
+ * element is a sample.
+ *
+ * - This field is intended to be set to VALIGN_4 if the surface was
+ * rendered as a depth buffer, for a multisampled (4x) render target,
+ * or for a multisampled (8x) render target, since these surfaces
+ * support only alignment of 4. Use of VALIGN_4 for other surfaces is
+ * supported, but increases memory usage.
+ *
+ * - This field is intended to be set to VALIGN_8 only if the surface
+ * was rendered as a stencil buffer, since stencil buffer surfaces
+ * support only alignment of 8. If set to VALIGN_8, Surface Format
+ * must be R8_UINT.
+ */
+
+ if (isl_format_is_compressed(info->format))
+ return 1;
+
+ if (isl_surf_usage_is_stencil(info->usage))
+ return 8;
+
+ return 4;
+}
+
+void
+gen8_choose_image_alignment_el(const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info,
+ enum isl_tiling tiling,
+ enum isl_msaa_layout msaa_layout,
+ struct isl_extent3d *image_align_el)
+{
+ assert(!isl_tiling_is_std_y(tiling));
+
+ /* The below text from the Broadwell PRM provides some insight into the
+ * hardware's requirements for LOD alignment. From the Broadwell PRM >>
+ * Volume 5: Memory Views >> Surface Layout >> 2D Surfaces:
+ *
+ * These [2D surfaces] must adhere to the following memory organization
+ * rules:
+ *
+ * - For non-compressed texture formats, each mipmap must start on an
+ * even row within the monolithic rectangular area. For
+ * 1-texel-high mipmaps, this may require a row of padding below
+ * the previous mipmap. This restriction does not apply to any
+ * compressed texture formats; each subsequent (lower-res)
+ * compressed mipmap is positioned directly below the previous
+ * mipmap.
+ *
+ * - Vertical alignment restrictions vary with memory tiling type:
+ * 1 DWord for linear, 16-byte (DQWord) for tiled. (Note that tiled
+ * mipmaps are not required to start at the left edge of a tile
+ * row.)
+ */
+
+ *image_align_el = (struct isl_extent3d) {
+ .w = gen8_choose_halign_el(dev, info),
+ .h = gen8_choose_valign_el(dev, info),
+ .d = 1,
+ };
+}
--- /dev/null
+/*
+ * Copyright 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#pragma once
+
+#include "isl_priv.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+bool
+gen8_choose_msaa_layout(const struct isl_device *dev,
+ const struct isl_surf_init_info *info,
+ enum isl_tiling tiling,
+ enum isl_msaa_layout *msaa_layout);
+
+void
+gen8_choose_image_alignment_el(const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info,
+ enum isl_tiling tiling,
+ enum isl_msaa_layout msaa_layout,
+ struct isl_extent3d *image_align_el);
+
+#ifdef __cplusplus
+}
+#endif
--- /dev/null
+/*
+ * Copyright 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include "isl_gen8.h"
+#include "isl_gen9.h"
+#include "isl_priv.h"
+
+/**
+ * Calculate the surface's subimage alignment, in units of surface samples,
+ * for the standard tiling formats Yf and Ys.
+ */
+static void
+gen9_calc_std_image_alignment_sa(const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info,
+ enum isl_tiling tiling,
+ enum isl_msaa_layout msaa_layout,
+ struct isl_extent3d *align_sa)
+{
+ const struct isl_format_layout *fmtl = isl_format_get_layout(info->format);
+
+ assert(isl_tiling_is_std_y(tiling));
+
+ const uint32_t bs = fmtl->bs;
+ const uint32_t is_Ys = tiling == ISL_TILING_Ys;
+
+ switch (info->dim) {
+ case ISL_SURF_DIM_1D:
+ /* See the Skylake BSpec > Memory Views > Common Surface Formats > Surface
+ * Layout and Tiling > 1D Surfaces > 1D Alignment Requirements.
+ */
+ *align_sa = (struct isl_extent3d) {
+ .w = 1 << (12 - (ffs(bs) - 1) + (4 * is_Ys)),
+ .h = 1,
+ .d = 1,
+ };
+ return;
+ case ISL_SURF_DIM_2D:
+ /* See the Skylake BSpec > Memory Views > Common Surface Formats >
+ * Surface Layout and Tiling > 2D Surfaces > 2D/CUBE Alignment
+ * Requirements.
+ */
+ *align_sa = (struct isl_extent3d) {
+ .w = 1 << (6 - ((ffs(bs) - 1) / 2) + (4 * is_Ys)),
+ .h = 1 << (6 - ((ffs(bs) - 0) / 2) + (4 * is_Ys)),
+ .d = 1,
+ };
+
+ if (is_Ys) {
+ /* FINISHME(chadv): I don't trust this code. Untested. */
+ isl_finishme("%s:%s: [SKL+] multisample TileYs", __FILE__, __func__);
+
+ switch (msaa_layout) {
+ case ISL_MSAA_LAYOUT_NONE:
+ case ISL_MSAA_LAYOUT_INTERLEAVED:
+ break;
+ case ISL_MSAA_LAYOUT_ARRAY:
+ align_sa->w >>= (ffs(info->samples) - 0) / 2;
+ align_sa->h >>= (ffs(info->samples) - 1) / 2;
+ break;
+ }
+ }
+ return;
+
+ case ISL_SURF_DIM_3D:
+ /* See the Skylake BSpec > Memory Views > Common Surface Formats > Surface
+ * Layout and Tiling > 1D Surfaces > 1D Alignment Requirements.
+ */
+ *align_sa = (struct isl_extent3d) {
+ .w = 1 << (4 - ((ffs(bs) + 1) / 3) + (4 * is_Ys)),
+ .h = 1 << (4 - ((ffs(bs) - 1) / 3) + (2 * is_Ys)),
+ .d = 1 << (4 - ((ffs(bs) - 0) / 3) + (2 * is_Ys)),
+ };
+ return;
+ }
+
+ unreachable("bad isl_surface_type");
+}
+
+void
+gen9_choose_image_alignment_el(const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info,
+ enum isl_tiling tiling,
+ enum isl_msaa_layout msaa_layout,
+ struct isl_extent3d *image_align_el)
+{
+ /* This BSpec text provides some insight into the hardware's alignment
+ * requirements [Skylake BSpec > Memory Views > Common Surface Formats >
+ * Surface Layout and Tiling > 2D Surfaces]:
+ *
+ * An LOD must be aligned to a cache-line except for some special cases
+ * related to Planar YUV surfaces. In general, the cache-alignment
+ * restriction implies there is a minimum height for an LOD of 4 texels.
+ * So, LODs which are smaller than 4 high are padded.
+ *
+ * From the Skylake BSpec, RENDER_SURFACE_STATE Surface Vertical Alignment:
+ *
+ * - For Sampling Engine and Render Target Surfaces: This field
+ * specifies the vertical alignment requirement in elements for the
+ * surface. [...] An element is defined as a pixel in uncompresed
+ * surface formats, and as a compression block in compressed surface
+ * formats. For MSFMT_DEPTH_STENCIL type multisampled surfaces, an
+ * element is a sample.
+ *
+ * - This field is used for 2D, CUBE, and 3D surface alignment when Tiled
+ * Resource Mode is TRMODE_NONE (Tiled Resource Mode is disabled).
+ * This field is ignored for 1D surfaces and also when Tiled Resource
+ * Mode is not TRMODE_NONE (e.g. Tiled Resource Mode is enabled).
+ *
+ * See the appropriate Alignment table in the "Surface Layout and
+ * Tiling" section under Common Surface Formats for the table of
+ * alignment values for Tiled Resrouces.
+ *
+ * - For uncompressed surfaces, the units of "j" are rows of pixels on
+ * the physical surface. For compressed texture formats, the units of
+ * "j" are in compression blocks, thus each increment in "j" is equal
+ * to h pixels, where h is the height of the compression block in
+ * pixels.
+ *
+ * - Valid Values: VALIGN_4, VALIGN_8, VALIGN_16
+ *
+ * From the Skylake BSpec, RENDER_SURFACE_STATE Surface Horizontal
+ * Alignment:
+ *
+ * - For uncompressed surfaces, the units of "i" are pixels on the
+ * physical surface. For compressed texture formats, the units of "i"
+ * are in compression blocks, thus each increment in "i" is equal to
+ * w pixels, where w is the width of the compression block in pixels.
+ *
+ * - Valid Values: HALIGN_4, HALIGN_8, HALIGN_16
+ */
+
+ if (isl_tiling_is_std_y(tiling)) {
+ struct isl_extent3d image_align_sa;
+ gen9_calc_std_image_alignment_sa(dev, info, tiling, msaa_layout,
+ &image_align_sa);
+
+ *image_align_el = isl_extent3d_sa_to_el(info->format, image_align_sa);
+ return;
+ }
+
+ if (info->dim == ISL_SURF_DIM_1D) {
+ /* See the Skylake BSpec > Memory Views > Common Surface Formats > Surface
+ * Layout and Tiling > 1D Surfaces > 1D Alignment Requirements.
+ */
+ *image_align_el = isl_extent3d(64, 1, 1);
+ return;
+ }
+
+ if (isl_format_is_compressed(info->format)) {
+ /* On Gen9, the meaning of RENDER_SURFACE_STATE's
+ * SurfaceHorizontalAlignment and SurfaceVerticalAlignment changed for
+ * compressed formats. They now indicate a multiple of the compression
+ * block. For example, if the compression mode is ETC2 then HALIGN_4
+ * indicates a horizontal alignment of 16 pixels.
+ *
+ * To avoid wasting memory, choose the smallest alignment possible:
+ * HALIGN_4 and VALIGN_4.
+ */
+ *image_align_el = isl_extent3d(4, 4, 1);
+ return;
+ }
+
+ gen8_choose_image_alignment_el(dev, info, tiling, msaa_layout,
+ image_align_el);
+}
--- /dev/null
+/*
+ * Copyright 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#pragma once
+
+#include "isl_priv.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+void
+gen9_choose_image_alignment_el(const struct isl_device *dev,
+ const struct isl_surf_init_info *restrict info,
+ enum isl_tiling tiling,
+ enum isl_msaa_layout msaa_layout,
+ struct isl_extent3d *image_align_el);
+
+#ifdef __cplusplus
+}
+#endif
--- /dev/null
+/*
+ * Copyright 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include "isl.h"
+#include "brw_compiler.h"
+
+bool
+isl_is_storage_image_format(enum isl_format format)
+{
+ /* XXX: Maybe we should put this in the CSV? */
+
+ switch (format) {
+ case ISL_FORMAT_R32G32B32A32_UINT:
+ case ISL_FORMAT_R32G32B32A32_SINT:
+ case ISL_FORMAT_R32G32B32A32_FLOAT:
+ case ISL_FORMAT_R32_UINT:
+ case ISL_FORMAT_R32_SINT:
+ case ISL_FORMAT_R32_FLOAT:
+ case ISL_FORMAT_R16G16B16A16_UINT:
+ case ISL_FORMAT_R16G16B16A16_SINT:
+ case ISL_FORMAT_R16G16B16A16_FLOAT:
+ case ISL_FORMAT_R32G32_UINT:
+ case ISL_FORMAT_R32G32_SINT:
+ case ISL_FORMAT_R32G32_FLOAT:
+ case ISL_FORMAT_R8G8B8A8_UINT:
+ case ISL_FORMAT_R8G8B8A8_SINT:
+ case ISL_FORMAT_R16G16_UINT:
+ case ISL_FORMAT_R16G16_SINT:
+ case ISL_FORMAT_R16G16_FLOAT:
+ case ISL_FORMAT_R8G8_UINT:
+ case ISL_FORMAT_R8G8_SINT:
+ case ISL_FORMAT_R16_UINT:
+ case ISL_FORMAT_R16_FLOAT:
+ case ISL_FORMAT_R16_SINT:
+ case ISL_FORMAT_R8_UINT:
+ case ISL_FORMAT_R8_SINT:
+ case ISL_FORMAT_R10G10B10A2_UINT:
+ case ISL_FORMAT_R10G10B10A2_UNORM:
+ case ISL_FORMAT_R11G11B10_FLOAT:
+ case ISL_FORMAT_R16G16B16A16_UNORM:
+ case ISL_FORMAT_R16G16B16A16_SNORM:
+ case ISL_FORMAT_R8G8B8A8_UNORM:
+ case ISL_FORMAT_R8G8B8A8_SNORM:
+ case ISL_FORMAT_R16G16_UNORM:
+ case ISL_FORMAT_R16G16_SNORM:
+ case ISL_FORMAT_R8G8_UNORM:
+ case ISL_FORMAT_R8G8_SNORM:
+ case ISL_FORMAT_R16_UNORM:
+ case ISL_FORMAT_R16_SNORM:
+ case ISL_FORMAT_R8_UNORM:
+ case ISL_FORMAT_R8_SNORM:
+ return true;
+ default:
+ return false;
+ }
+}
+
+enum isl_format
+isl_lower_storage_image_format(const struct isl_device *dev,
+ enum isl_format format)
+{
+ switch (format) {
+ /* These are never lowered. Up to BDW we'll have to fall back to untyped
+ * surface access for 128bpp formats.
+ */
+ case ISL_FORMAT_R32G32B32A32_UINT:
+ case ISL_FORMAT_R32G32B32A32_SINT:
+ case ISL_FORMAT_R32G32B32A32_FLOAT:
+ case ISL_FORMAT_R32_UINT:
+ case ISL_FORMAT_R32_SINT:
+ case ISL_FORMAT_R32_FLOAT:
+ return format;
+
+ /* From HSW to BDW the only 64bpp format supported for typed access is
+ * RGBA_UINT16. IVB falls back to untyped.
+ */
+ case ISL_FORMAT_R16G16B16A16_UINT:
+ case ISL_FORMAT_R16G16B16A16_SINT:
+ case ISL_FORMAT_R16G16B16A16_FLOAT:
+ case ISL_FORMAT_R32G32_UINT:
+ case ISL_FORMAT_R32G32_SINT:
+ case ISL_FORMAT_R32G32_FLOAT:
+ return (ISL_DEV_GEN(dev) >= 9 ? format :
+ ISL_DEV_GEN(dev) >= 8 || dev->info->is_haswell ?
+ ISL_FORMAT_R16G16B16A16_UINT :
+ ISL_FORMAT_R32G32_UINT);
+
+ /* Up to BDW no SINT or FLOAT formats of less than 32 bits per component
+ * are supported. IVB doesn't support formats with more than one component
+ * for typed access. For 8 and 16 bpp formats IVB relies on the
+ * undocumented behavior that typed reads from R_UINT8 and R_UINT16
+ * surfaces actually do a 32-bit misaligned read. The alternative would be
+ * to use two surface state entries with different formats for each image,
+ * one for reading (using R_UINT32) and another one for writing (using
+ * R_UINT8 or R_UINT16), but that would complicate the shaders we generate
+ * even more.
+ */
+ case ISL_FORMAT_R8G8B8A8_UINT:
+ case ISL_FORMAT_R8G8B8A8_SINT:
+ return (ISL_DEV_GEN(dev) >= 9 ? format :
+ ISL_DEV_GEN(dev) >= 8 || dev->info->is_haswell ?
+ ISL_FORMAT_R8G8B8A8_UINT : ISL_FORMAT_R32_UINT);
+
+ case ISL_FORMAT_R16G16_UINT:
+ case ISL_FORMAT_R16G16_SINT:
+ case ISL_FORMAT_R16G16_FLOAT:
+ return (ISL_DEV_GEN(dev) >= 9 ? format :
+ ISL_DEV_GEN(dev) >= 8 || dev->info->is_haswell ?
+ ISL_FORMAT_R16G16_UINT : ISL_FORMAT_R32_UINT);
+
+ case ISL_FORMAT_R8G8_UINT:
+ case ISL_FORMAT_R8G8_SINT:
+ return (ISL_DEV_GEN(dev) >= 9 ? format :
+ ISL_DEV_GEN(dev) >= 8 || dev->info->is_haswell ?
+ ISL_FORMAT_R8G8_UINT : ISL_FORMAT_R16_UINT);
+
+ case ISL_FORMAT_R16_UINT:
+ case ISL_FORMAT_R16_FLOAT:
+ case ISL_FORMAT_R16_SINT:
+ return (ISL_DEV_GEN(dev) >= 9 ? format : ISL_FORMAT_R16_UINT);
+
+ case ISL_FORMAT_R8_UINT:
+ case ISL_FORMAT_R8_SINT:
+ return (ISL_DEV_GEN(dev) >= 9 ? format : ISL_FORMAT_R8_UINT);
+
+ /* Neither the 2/10/10/10 nor the 11/11/10 packed formats are supported
+ * by the hardware.
+ */
+ case ISL_FORMAT_R10G10B10A2_UINT:
+ case ISL_FORMAT_R10G10B10A2_UNORM:
+ case ISL_FORMAT_R11G11B10_FLOAT:
+ return ISL_FORMAT_R32_UINT;
+
+ /* No normalized fixed-point formats are supported by the hardware. */
+ case ISL_FORMAT_R16G16B16A16_UNORM:
+ case ISL_FORMAT_R16G16B16A16_SNORM:
+ return (ISL_DEV_GEN(dev) >= 8 || dev->info->is_haswell ?
+ ISL_FORMAT_R16G16B16A16_UINT :
+ ISL_FORMAT_R32G32_UINT);
+
+ case ISL_FORMAT_R8G8B8A8_UNORM:
+ case ISL_FORMAT_R8G8B8A8_SNORM:
+ return (ISL_DEV_GEN(dev) >= 8 || dev->info->is_haswell ?
+ ISL_FORMAT_R8G8B8A8_UINT : ISL_FORMAT_R32_UINT);
+
+ case ISL_FORMAT_R16G16_UNORM:
+ case ISL_FORMAT_R16G16_SNORM:
+ return (ISL_DEV_GEN(dev) >= 8 || dev->info->is_haswell ?
+ ISL_FORMAT_R16G16_UINT : ISL_FORMAT_R32_UINT);
+
+ case ISL_FORMAT_R8G8_UNORM:
+ case ISL_FORMAT_R8G8_SNORM:
+ return (ISL_DEV_GEN(dev) >= 8 || dev->info->is_haswell ?
+ ISL_FORMAT_R8G8_UINT : ISL_FORMAT_R16_UINT);
+
+ case ISL_FORMAT_R16_UNORM:
+ case ISL_FORMAT_R16_SNORM:
+ return ISL_FORMAT_R16_UINT;
+
+ case ISL_FORMAT_R8_UNORM:
+ case ISL_FORMAT_R8_SNORM:
+ return ISL_FORMAT_R8_UINT;
+
+ default:
+ assert(!"Unknown image format");
+ return ISL_FORMAT_UNSUPPORTED;
+ }
+}
--- /dev/null
+/*
+ * Copyright 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#pragma once
+
+#include <assert.h>
+
+#include "brw_device_info.h"
+#include "mesa/main/imports.h"
+#include "util/macros.h"
+
+#include "isl.h"
+
+#define isl_finishme(format, ...) \
+ __isl_finishme(__FILE__, __LINE__, format, ##__VA_ARGS__)
+
+void PRINTFLIKE(3, 4) UNUSED
+__isl_finishme(const char *file, int line, const char *fmt, ...);
+
+#define MIN(a, b) ((a) < (b) ? (a) : (b))
+#define MAX(a, b) ((a) > (b) ? (a) : (b))
+
+static inline uint32_t
+ffs(uint32_t n) {
+ return __builtin_ffs(n);
+}
+
+static inline bool
+isl_is_pow2(uintmax_t n)
+{
+ return !(n & (n - 1));
+}
+
+/**
+ * Alignment must be a power of 2.
+ */
+static inline bool
+isl_is_aligned(uintmax_t n, uintmax_t a)
+{
+ assert(isl_is_pow2(a));
+ return (n & (a - 1)) == 0;
+}
+
+/**
+ * Alignment must be a power of 2.
+ */
+static inline uintmax_t
+isl_align(uintmax_t n, uintmax_t a)
+{
+ assert(a != 0 && isl_is_pow2(a));
+ return (n + a - 1) & ~(a - 1);
+}
+
+static inline uintmax_t
+isl_align_npot(uintmax_t n, uintmax_t a)
+{
+ assert(a > 0);
+ return ((n + a - 1) / a) * a;
+}
+
+/**
+ * Alignment must be a power of 2.
+ */
+static inline uintmax_t
+isl_align_div(uintmax_t n, uintmax_t a)
+{
+ return isl_align(n, a) / a;
+}
+
+static inline uintmax_t
+isl_align_div_npot(uintmax_t n, uintmax_t a)
+{
+ return isl_align_npot(n, a) / a;
+}
+
+/**
+ * Log base 2, rounding towards zero.
+ */
+static inline uint32_t
+isl_log2u(uint32_t n)
+{
+ assert(n != 0);
+ return 31 - __builtin_clz(n);
+}
+
+static inline uint32_t
+isl_minify(uint32_t n, uint32_t levels)
+{
+ if (unlikely(n == 0))
+ return 0;
+ else
+ return MAX(n >> levels, 1);
+}
+
+static inline struct isl_extent3d
+isl_extent3d_sa_to_el(enum isl_format fmt, struct isl_extent3d extent_sa)
+{
+ const struct isl_format_layout *fmtl = isl_format_get_layout(fmt);
+
+ assert(extent_sa.w % fmtl->bw == 0);
+ assert(extent_sa.h % fmtl->bh == 0);
+ assert(extent_sa.d % fmtl->bd == 0);
+
+ return (struct isl_extent3d) {
+ .w = extent_sa.w / fmtl->bw,
+ .h = extent_sa.h / fmtl->bh,
+ .d = extent_sa.d / fmtl->bd,
+ };
+}
+
+static inline struct isl_extent3d
+isl_extent3d_el_to_sa(enum isl_format fmt, struct isl_extent3d extent_el)
+{
+ const struct isl_format_layout *fmtl = isl_format_get_layout(fmt);
+
+ return (struct isl_extent3d) {
+ .w = extent_el.w * fmtl->bw,
+ .h = extent_el.h * fmtl->bh,
+ .d = extent_el.d * fmtl->bd,
+ };
+}
--- /dev/null
+/isl_surf_get_image_offset_test
--- /dev/null
+/*
+ * Copyright 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include <assert.h>
+#include <stdbool.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+#include "brw_device_info.h"
+#include "isl.h"
+
+#define BDW_GT2_DEVID 0x161a
+
+// An asssert that works regardless of NDEBUG.
+#define t_assert(cond) \
+ do { \
+ if (!(cond)) { \
+ fprintf(stderr, "%s:%d: assertion failed\n", __FILE__, __LINE__); \
+ abort(); \
+ } \
+ } while (0)
+
+static void
+t_assert_extent4d(const struct isl_extent4d *e, uint32_t width,
+ uint32_t height, uint32_t depth, uint32_t array_len)
+{
+ t_assert(e->width == width);
+ t_assert(e->height == height);
+ t_assert(e->depth == depth);
+ t_assert(e->array_len == array_len);
+}
+
+static void
+t_assert_image_alignment_el(const struct isl_surf *surf,
+ uint32_t w, uint32_t h, uint32_t d)
+{
+ struct isl_extent3d align_el;
+
+ align_el = isl_surf_get_image_alignment_el(surf);
+ t_assert(align_el.w == w);
+ t_assert(align_el.h == h);
+ t_assert(align_el.d == d);
+
+}
+
+static void
+t_assert_image_alignment_sa(const struct isl_surf *surf,
+ uint32_t w, uint32_t h, uint32_t d)
+{
+ struct isl_extent3d align_sa;
+
+ align_sa = isl_surf_get_image_alignment_sa(surf);
+ t_assert(align_sa.w == w);
+ t_assert(align_sa.h == h);
+ t_assert(align_sa.d == d);
+
+}
+
+static void
+t_assert_offset(const struct isl_surf *surf,
+ uint32_t level,
+ uint32_t logical_array_layer,
+ uint32_t logical_z_offset_px,
+ uint32_t expected_x_offset_sa,
+ uint32_t expected_y_offset_sa)
+{
+ uint32_t x, y;
+ isl_surf_get_image_offset_sa(surf, level, logical_array_layer,
+ logical_z_offset_px, &x, &y);
+
+ t_assert(x == expected_x_offset_sa);
+ t_assert(y == expected_y_offset_sa);
+}
+
+static void
+t_assert_phys_level0_sa(const struct isl_surf *surf, uint32_t width,
+ uint32_t height, uint32_t depth, uint32_t array_len)
+{
+ t_assert_extent4d(&surf->phys_level0_sa, width, height, depth, array_len);
+}
+
+static void
+test_bdw_2d_r8g8b8a8_unorm_512x512_a1_s1_noaux_y0(void)
+{
+ bool ok;
+
+ struct isl_device dev;
+ isl_device_init(&dev, brw_get_device_info(BDW_GT2_DEVID));
+
+ struct isl_surf surf;
+ ok = isl_surf_init(&dev, &surf,
+ .dim = ISL_SURF_DIM_2D,
+ .format = ISL_FORMAT_R8G8B8A8_UNORM,
+ .width = 512,
+ .height = 512,
+ .depth = 1,
+ .levels = 10,
+ .array_len = 1,
+ .samples = 1,
+ .usage = ISL_SURF_USAGE_TEXTURE_BIT |
+ ISL_SURF_USAGE_DISABLE_AUX_BIT,
+ .tiling_flags = ISL_TILING_Y0_BIT);
+ t_assert(ok);
+
+ t_assert_image_alignment_el(&surf, 4, 4, 1);
+ t_assert_image_alignment_sa(&surf, 4, 4, 1);
+ t_assert_phys_level0_sa(&surf, 512, 512, 1, 1);
+ t_assert(isl_surf_get_array_pitch_el_rows(&surf) >= 772);
+ t_assert(isl_surf_get_array_pitch_sa_rows(&surf) >= 772);
+
+ t_assert_offset(&surf, 0, 0, 0, 0, 0); // +0, +0
+ t_assert_offset(&surf, 1, 0, 0, 0, 512); // +0, +512
+ t_assert_offset(&surf, 2, 0, 0, 256, 512); // +256, +0
+ t_assert_offset(&surf, 3, 0, 0, 256, 640); // +0, +128
+ t_assert_offset(&surf, 4, 0, 0, 256, 704); // +0, +64
+ t_assert_offset(&surf, 5, 0, 0, 256, 736); // +0, +32
+ t_assert_offset(&surf, 6, 0, 0, 256, 752); // +0, +16
+ t_assert_offset(&surf, 7, 0, 0, 256, 760); // +0, +8
+ t_assert_offset(&surf, 8, 0, 0, 256, 764); // +0, +4
+ t_assert_offset(&surf, 9, 0, 0, 256, 768); // +0, +4
+}
+
+static void
+test_bdw_2d_r8g8b8a8_unorm_1024x1024_a6_s1_noaux_y0(void)
+{
+ bool ok;
+
+ struct isl_device dev;
+ isl_device_init(&dev, brw_get_device_info(BDW_GT2_DEVID));
+
+ struct isl_surf surf;
+ ok = isl_surf_init(&dev, &surf,
+ .dim = ISL_SURF_DIM_2D,
+ .format = ISL_FORMAT_R8G8B8A8_UNORM,
+ .width = 1024,
+ .height = 1024,
+ .depth = 1,
+ .levels = 11,
+ .array_len = 6,
+ .samples = 1,
+ .usage = ISL_SURF_USAGE_TEXTURE_BIT |
+ ISL_SURF_USAGE_DISABLE_AUX_BIT,
+ .tiling_flags = ISL_TILING_Y0_BIT);
+ t_assert(ok);
+
+ t_assert_image_alignment_el(&surf, 4, 4, 1);
+ t_assert_image_alignment_sa(&surf, 4, 4, 1);
+ t_assert_image_alignment_sa(&surf, 4, 4, 1);
+ t_assert(isl_surf_get_array_pitch_el_rows(&surf) >= 1540);
+ t_assert(isl_surf_get_array_pitch_sa_rows(&surf) >= 1540);
+
+ for (uint32_t a = 0; a < 6; ++a) {
+ uint32_t b = a * isl_surf_get_array_pitch_sa_rows(&surf);
+
+ t_assert_offset(&surf, 0, a, 0, 0, b + 0); // +0, +0
+ t_assert_offset(&surf, 1, a, 0, 0, b + 1024); // +0, +1024
+ t_assert_offset(&surf, 2, a, 0, 512, b + 1024); // +512, +0
+ t_assert_offset(&surf, 3, a, 0, 512, b + 1280); // +0, +256
+ t_assert_offset(&surf, 4, a, 0, 512, b + 1408); // +0, +128
+ t_assert_offset(&surf, 5, a, 0, 512, b + 1472); // +0, +64
+ t_assert_offset(&surf, 6, a, 0, 512, b + 1504); // +0, +32
+ t_assert_offset(&surf, 7, a, 0, 512, b + 1520); // +0, +16
+ t_assert_offset(&surf, 8, a, 0, 512, b + 1528); // +0, +8
+ t_assert_offset(&surf, 9, a, 0, 512, b + 1532); // +0, +4
+ t_assert_offset(&surf, 10, a, 0, 512, b + 1536); // +0, +4
+ }
+}
+
+int main(void)
+{
+ /* FINISHME: Add tests for npot sizes */
+ /* FINISHME: Add tests for 1D surfaces */
+ /* FINISHME: Add tests for 3D surfaces */
+
+ test_bdw_2d_r8g8b8a8_unorm_512x512_a1_s1_noaux_y0();
+ test_bdw_2d_r8g8b8a8_unorm_1024x1024_a6_s1_noaux_y0();
+}
--- /dev/null
+<?xml version="1.0"?>
+<!DOCTYPE OpenGLAPI SYSTEM "gl_API.dtd">
+
+<OpenGLAPI>
+
+<category name="GL_ARB_draw_indirect" number="154">
+
+ <enum name="PARAMETER_BUFFER_ARB" value="0x80EE"/>
+ <enum name="PARAMETER_BUFFER_BINDING_ARB" value="0x80EF"/>
+
+ <function name="MultiDrawArraysIndirectCountARB" exec="dynamic">
+ <param name="mode" type="GLenum"/>
+ <param name="indirect" type="GLintptr"/>
+ <param name="drawcount" type="GLintptr"/>
+ <param name="maxdrawcount" type="GLsizei"/>
+ <param name="stride" type="GLsizei"/>
+ </function>
+
+ <function name="MultiDrawElementsIndirectCountARB" exec="dynamic">
+ <param name="mode" type="GLenum"/>
+ <param name="type" type="GLenum"/>
+ <param name="indirect" type="GLintptr"/>
+ <param name="drawcount" type="GLintptr"/>
+ <param name="maxdrawcount" type="GLsizei"/>
+ <param name="stride" type="GLsizei"/>
+ </function>
+
+</category>
+
+</OpenGLAPI>
ARB_get_texture_sub_image.xml \
ARB_gpu_shader_fp64.xml \
ARB_gpu_shader5.xml \
+ ARB_indirect_parameters.xml \
ARB_instanced_arrays.xml \
ARB_internalformat_query.xml \
ARB_invalidate_subdata.xml \
<xi:include href="ARB_multi_bind.xml" xmlns:xi="http://www.w3.org/2001/XInclude"/>
-<!-- ARB extensions 148 - 159 -->
+<!-- ARB extensions 148 - 153 -->
+
+<xi:include href="ARB_indirect_parameters.xml" xmlns:xi="http://www.w3.org/2001/XInclude"/>
+
+<!-- ARB extensions 155 - 159 -->
<xi:include href="ARB_clip_control.xml" xmlns:xi="http://www.w3.org/2001/XInclude"/>
_mesa_reference_buffer_object(ctx, &decompress->buf_obj, NULL);
}
- if (decompress->Sampler != 0)
- _mesa_DeleteSamplers(1, &decompress->Sampler);
+ _mesa_reference_sampler_object(ctx, &decompress->samp_obj, NULL);
memset(decompress, 0, sizeof(*decompress));
}
GLenum rbFormat;
GLenum faceTarget;
struct vertex verts[4];
- GLuint samplerSave;
+ struct gl_sampler_object *samp_obj_save = NULL;
GLenum status;
const bool use_glsl_version = ctx->Extensions.ARB_vertex_shader &&
ctx->Extensions.ARB_fragment_shader;
_mesa_meta_begin(ctx, MESA_META_ALL & ~(MESA_META_PIXEL_STORE |
MESA_META_DRAW_BUFFERS));
- samplerSave = ctx->Texture.Unit[ctx->Texture.CurrentUnit].Sampler ?
- ctx->Texture.Unit[ctx->Texture.CurrentUnit].Sampler->Name : 0;
+ _mesa_reference_sampler_object(ctx, &samp_obj_save,
+ ctx->Texture.Unit[ctx->Texture.CurrentUnit].Sampler);
/* Create/bind FBO/renderbuffer */
if (decompress_fbo->FBO == 0) {
&decompress->buf_obj, 3);
}
- if (!decompress->Sampler) {
- _mesa_GenSamplers(1, &decompress->Sampler);
- _mesa_BindSampler(ctx->Texture.CurrentUnit, decompress->Sampler);
- /* nearest filtering */
- _mesa_SamplerParameteri(decompress->Sampler, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
- _mesa_SamplerParameteri(decompress->Sampler, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
- /* No sRGB decode or encode.*/
- if (ctx->Extensions.EXT_texture_sRGB_decode) {
- _mesa_SamplerParameteri(decompress->Sampler, GL_TEXTURE_SRGB_DECODE_EXT,
- GL_SKIP_DECODE_EXT);
+ if (decompress->samp_obj == NULL) {
+ decompress->samp_obj = ctx->Driver.NewSamplerObject(ctx, 0xDEADBEEF);
+ if (decompress->samp_obj == NULL) {
+ _mesa_meta_end(ctx);
+
+ /* This is a bit lazy. Flag out of memory, and then don't bother to
+ * clean up. Once out of memory is flagged, the only realistic next
+ * move is to destroy the context. That will trigger all the right
+ * clean up.
+ *
+ * Returning true prevents other GetTexImage methods from attempting
+ * anything since they will likely fail too.
+ */
+ _mesa_error(ctx, GL_OUT_OF_MEMORY, "glGetTexImage");
+ return true;
}
- } else {
- _mesa_BindSampler(ctx->Texture.CurrentUnit, decompress->Sampler);
+ /* nearest filtering */
+ _mesa_set_sampler_filters(ctx, decompress->samp_obj, GL_NEAREST, GL_NEAREST);
+
+ /* We don't want to encode or decode sRGB values; treat them as linear. */
+ _mesa_set_sampler_srgb_decode(ctx, decompress->samp_obj, GL_SKIP_DECODE_EXT);
}
+ _mesa_bind_sampler(ctx, ctx->Texture.CurrentUnit, decompress->samp_obj);
+
/* Silence valgrind warnings about reading uninitialized stack. */
memset(verts, 0, sizeof(verts));
if (!use_glsl_version)
_mesa_set_enable(ctx, target, GL_FALSE);
- _mesa_BindSampler(ctx->Texture.CurrentUnit, samplerSave);
+ _mesa_bind_sampler(ctx, ctx->Texture.CurrentUnit, samp_obj_save);
+ _mesa_reference_sampler_object(ctx, &samp_obj_save, NULL);
_mesa_meta_end(ctx);
struct fb_tex_blit_state
{
GLint baseLevelSave, maxLevelSave;
- GLuint sampler, samplerSave, stencilSamplingSave;
+ struct gl_sampler_object *samp_obj;
+ struct gl_sampler_object *samp_obj_save;
+ GLuint stencilSamplingSave;
GLuint tempTex;
};
GLuint VAO;
struct gl_buffer_object *buf_obj;
GLuint FBO;
- GLuint Sampler;
+ struct gl_sampler_object *samp_obj;
struct blit_shader_table shaders;
};
GLuint VAO;
struct decompress_fbo_state byteFBO, floatFBO;
struct gl_buffer_object *buf_obj;
- GLuint Sampler;
+ struct gl_sampler_object *samp_obj;
struct blit_shader_table shaders;
};
}
extern void
-_mesa_meta_fb_tex_blit_begin(const struct gl_context *ctx,
+_mesa_meta_fb_tex_blit_begin(struct gl_context *ctx,
struct fb_tex_blit_state *blit);
extern void
struct gl_texture_object **texObj,
GLenum *target);
-GLuint
+struct gl_sampler_object *
_mesa_meta_setup_sampler(struct gl_context *ctx,
const struct gl_texture_object *texObj,
GLenum target, GLenum filter, GLuint srcLevel);
printf(" srcTex %p dstText %p\n", texObj, drawAtt->Texture);
*/
- fb_tex_blit.sampler = _mesa_meta_setup_sampler(ctx, texObj, target, filter,
- srcLevel);
+ fb_tex_blit.samp_obj = _mesa_meta_setup_sampler(ctx, texObj, target, filter,
+ srcLevel);
/* Always do our blits with no net sRGB decode or encode.
*
if (ctx->Extensions.EXT_texture_sRGB_decode) {
if (_mesa_get_format_color_encoding(rb->Format) == GL_SRGB &&
drawFb->Visual.sRGBCapable) {
- _mesa_SamplerParameteri(fb_tex_blit.sampler,
- GL_TEXTURE_SRGB_DECODE_EXT, GL_DECODE_EXT);
+ _mesa_set_sampler_srgb_decode(ctx, fb_tex_blit.samp_obj,
+ GL_DECODE_EXT);
_mesa_set_framebuffer_srgb(ctx, GL_TRUE);
} else {
- _mesa_SamplerParameteri(fb_tex_blit.sampler,
- GL_TEXTURE_SRGB_DECODE_EXT,
- GL_SKIP_DECODE_EXT);
+ _mesa_set_sampler_srgb_decode(ctx, fb_tex_blit.samp_obj,
+ GL_SKIP_DECODE_EXT);
/* set_framebuffer_srgb was set by _mesa_meta_begin(). */
}
}
}
void
-_mesa_meta_fb_tex_blit_begin(const struct gl_context *ctx,
+_mesa_meta_fb_tex_blit_begin(struct gl_context *ctx,
struct fb_tex_blit_state *blit)
{
- blit->samplerSave =
- ctx->Texture.Unit[ctx->Texture.CurrentUnit].Sampler ?
- ctx->Texture.Unit[ctx->Texture.CurrentUnit].Sampler->Name : 0;
+ /* None of the existing callers preinitialize fb_tex_blit_state to zeros,
+ * and both use stack variables. If samp_obj_save is not NULL,
+ * _mesa_reference_sampler_object will try to dereference it. Leaving
+ * random garbage in samp_obj_save can only lead to crashes.
+ *
+ * Since the state isn't persistent across calls, we won't catch ref
+ * counting problems.
+ */
+ blit->samp_obj_save = NULL;
+ _mesa_reference_sampler_object(ctx, &blit->samp_obj_save,
+ ctx->Texture.Unit[ctx->Texture.CurrentUnit].Sampler);
blit->tempTex = 0;
}
}
}
- _mesa_BindSampler(ctx->Texture.CurrentUnit, blit->samplerSave);
- _mesa_DeleteSamplers(1, &blit->sampler);
+ _mesa_bind_sampler(ctx, ctx->Texture.CurrentUnit, blit->samp_obj_save);
+ _mesa_reference_sampler_object(ctx, &blit->samp_obj_save, NULL);
+ _mesa_reference_sampler_object(ctx, &blit->samp_obj, NULL);
+
if (blit->tempTex)
_mesa_DeleteTextures(1, &blit->tempTex);
}
return true;
}
-GLuint
+struct gl_sampler_object *
_mesa_meta_setup_sampler(struct gl_context *ctx,
const struct gl_texture_object *texObj,
GLenum target, GLenum filter, GLuint srcLevel)
{
- GLuint sampler;
+ struct gl_sampler_object *samp_obj;
GLenum tex_filter = (filter == GL_SCALED_RESOLVE_FASTEST_EXT ||
filter == GL_SCALED_RESOLVE_NICEST_EXT) ?
GL_NEAREST : filter;
- _mesa_GenSamplers(1, &sampler);
- _mesa_BindSampler(ctx->Texture.CurrentUnit, sampler);
+ samp_obj = ctx->Driver.NewSamplerObject(ctx, 0xDEADBEEF);
+ if (samp_obj == NULL)
+ return NULL;
+
+ _mesa_bind_sampler(ctx, ctx->Texture.CurrentUnit, samp_obj);
+ _mesa_set_sampler_filters(ctx, samp_obj, tex_filter, tex_filter);
+ _mesa_set_sampler_wrap(ctx, samp_obj, GL_CLAMP_TO_EDGE, GL_CLAMP_TO_EDGE,
+ samp_obj->WrapR);
/* Prepare src texture state */
_mesa_BindTexture(target, texObj->Name);
- _mesa_SamplerParameteri(sampler, GL_TEXTURE_MIN_FILTER, tex_filter);
- _mesa_SamplerParameteri(sampler, GL_TEXTURE_MAG_FILTER, tex_filter);
if (target != GL_TEXTURE_RECTANGLE_ARB) {
_mesa_TexParameteri(target, GL_TEXTURE_BASE_LEVEL, srcLevel);
_mesa_TexParameteri(target, GL_TEXTURE_MAX_LEVEL, srcLevel);
}
- _mesa_SamplerParameteri(sampler, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
- _mesa_SamplerParameteri(sampler, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
- return sampler;
+ return samp_obj;
}
/**
_mesa_DeleteVertexArrays(1, &mipmap->VAO);
mipmap->VAO = 0;
_mesa_reference_buffer_object(ctx, &mipmap->buf_obj, NULL);
- _mesa_DeleteSamplers(1, &mipmap->Sampler);
- mipmap->Sampler = 0;
+ _mesa_reference_sampler_object(ctx, &mipmap->samp_obj, NULL);
if (mipmap->FBO != 0) {
_mesa_DeleteFramebuffers(1, &mipmap->FBO);
ctx->Extensions.ARB_fragment_shader;
GLenum faceTarget;
GLuint dstLevel;
- GLuint samplerSave;
+ struct gl_sampler_object *samp_obj_save = NULL;
GLint swizzle[4];
GLboolean swizzleSaved = GL_FALSE;
_mesa_set_enable(ctx, target, GL_TRUE);
}
- samplerSave = ctx->Texture.Unit[ctx->Texture.CurrentUnit].Sampler ?
- ctx->Texture.Unit[ctx->Texture.CurrentUnit].Sampler->Name : 0;
+ _mesa_reference_sampler_object(ctx, &samp_obj_save,
+ ctx->Texture.Unit[ctx->Texture.CurrentUnit].Sampler);
/* We may have been called from glGenerateTextureMipmap with CurrentUnit
* still set to 0, so we don't know when we can skip binding the texture.
*/
_mesa_BindTexture(target, texObj->Name);
- if (!mipmap->Sampler) {
- _mesa_GenSamplers(1, &mipmap->Sampler);
- _mesa_BindSampler(ctx->Texture.CurrentUnit, mipmap->Sampler);
-
- _mesa_SamplerParameteri(mipmap->Sampler,
- GL_TEXTURE_MIN_FILTER,
- GL_LINEAR_MIPMAP_LINEAR);
- _mesa_SamplerParameteri(mipmap->Sampler, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
- _mesa_SamplerParameteri(mipmap->Sampler, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
- _mesa_SamplerParameteri(mipmap->Sampler, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
- _mesa_SamplerParameteri(mipmap->Sampler, GL_TEXTURE_WRAP_R, GL_CLAMP_TO_EDGE);
-
- /* We don't want to encode or decode sRGB values; treat them as linear.
- * This is not technically correct for GLES3 but we don't get any API
- * error at the moment.
- */
- if (ctx->Extensions.EXT_texture_sRGB_decode) {
- _mesa_SamplerParameteri(mipmap->Sampler, GL_TEXTURE_SRGB_DECODE_EXT,
- GL_SKIP_DECODE_EXT);
+ if (mipmap->samp_obj == NULL) {
+ mipmap->samp_obj = ctx->Driver.NewSamplerObject(ctx, 0xDEADBEEF);
+ if (mipmap->samp_obj == NULL) {
+ /* This is a bit lazy. Flag out of memory, and then don't bother to
+ * clean up. Once out of memory is flagged, the only realistic next
+ * move is to destroy the context. That will trigger all the right
+ * clean up.
+ */
+ _mesa_error(ctx, GL_OUT_OF_MEMORY, "glGenerateMipmap");
+ return;
}
- } else {
- _mesa_BindSampler(ctx->Texture.CurrentUnit, mipmap->Sampler);
+
+ _mesa_set_sampler_filters(ctx, mipmap->samp_obj, GL_LINEAR_MIPMAP_LINEAR,
+ GL_LINEAR);
+ _mesa_set_sampler_wrap(ctx, mipmap->samp_obj, GL_CLAMP_TO_EDGE,
+ GL_CLAMP_TO_EDGE, GL_CLAMP_TO_EDGE);
+
+ /* We don't want to encode or decode sRGB values; treat them as linear. */
+ _mesa_set_sampler_srgb_decode(ctx, mipmap->samp_obj, GL_SKIP_DECODE_EXT);
}
+ _mesa_bind_sampler(ctx, ctx->Texture.CurrentUnit, mipmap->samp_obj);
+
assert(mipmap->FBO != 0);
_mesa_BindFramebuffer(GL_FRAMEBUFFER_EXT, mipmap->FBO);
_mesa_lock_texture(ctx, texObj); /* relock */
- _mesa_BindSampler(ctx->Texture.CurrentUnit, samplerSave);
+ _mesa_bind_sampler(ctx, ctx->Texture.CurrentUnit, samp_obj_save);
+ _mesa_reference_sampler_object(ctx, &samp_obj_save, NULL);
_mesa_meta_end(ctx);
<application name="Unigine Heaven (32-bit)" executable="heaven_x86">
<option name="allow_glsl_extension_directive_midshader" value="true" />
+ <!-- remove disable_blend_func_extended if 4.1 ever comes out -->
+ <option name="disable_blend_func_extended" value="true" />
</application>
<application name="Unigine Heaven (64-bit)" executable="heaven_x64">
<option name="allow_glsl_extension_directive_midshader" value="true" />
+ <!-- remove disable_blend_func_extended if 4.1 ever comes out -->
+ <option name="disable_blend_func_extended" value="true" />
</application>
<application name="Unigine Valley (32-bit)" executable="valley_x86">
<option name="allow_glsl_extension_directive_midshader" value="true" />
+ <!-- remove disable_blend_func_extended if 1.1 ever comes out -->
+ <option name="disable_blend_func_extended" value="true" />
</application>
<application name="Unigine Valley (64-bit)" executable="valley_x64">
<option name="allow_glsl_extension_directive_midshader" value="true" />
+ <!-- remove disable_blend_func_extended if 1.1 ever comes out -->
+ <option name="disable_blend_func_extended" value="true" />
</application>
<application name="Unigine OilRush (32-bit)" executable="OilRush_x86">
_mesa_align_free(intel_obj->sys_buffer);
drm_intel_bo_unreference(intel_obj->buffer);
- free(intel_obj);
+ _mesa_delete_buffer_object(ctx, obj);
}
}
if (ctx->Pixel.ZoomX != 1.0F || ctx->Pixel.ZoomY != 1.0F) {
- perf_debug("glCopyPixles(): Unsupported pixel zoom\n");
+ perf_debug("glCopyPixels(): Unsupported pixel zoom\n");
return false;
}
brw_cfg.cpp \
brw_cfg.h \
brw_compiler.h \
- brw_cubemap_normalize.cpp \
brw_dead_control_flow.cpp \
brw_dead_control_flow.h \
brw_defines.h \
brw_eu_util.c \
brw_eu_validate.c \
brw_fs_builder.h \
- brw_fs_channel_expressions.cpp \
brw_fs_cmod_propagation.cpp \
brw_fs_combine_constants.cpp \
brw_fs_copy_propagation.cpp \
brw_fs_surface_builder.cpp \
brw_fs_surface_builder.h \
brw_fs_validate.cpp \
- brw_fs_vector_splitting.cpp \
brw_fs_visitor.cpp \
brw_inst.h \
brw_interpolation_map.c \
brw_ir_allocator.h \
brw_ir_fs.h \
brw_ir_vec4.h \
- brw_lower_texture_gradients.cpp \
- brw_lower_unnormalized_offset.cpp \
brw_nir.h \
brw_nir.c \
brw_nir_analyze_boolean_resolves.c \
brw_shader.cpp \
brw_shader.h \
brw_surface_formats.c \
+ brw_surface_formats.h \
brw_util.c \
brw_util.h \
brw_vec4_builder.h \
brw_vec4_reg_allocate.cpp \
brw_vec4_surface_builder.cpp \
brw_vec4_surface_builder.h \
+ brw_vec4_tcs.cpp \
+ brw_vec4_tes.cpp \
brw_vec4_visitor.cpp \
brw_vec4_vs_visitor.cpp \
brw_vue_map.c \
brw_context.h \
brw_cs.c \
brw_cs.h \
+ brw_cubemap_normalize.cpp \
brw_curbe.c \
brw_draw.c \
brw_draw.h \
brw_ff_gs.c \
brw_ff_gs_emit.c \
brw_ff_gs.h \
+ brw_fs_channel_expressions.cpp \
+ brw_fs_vector_splitting.cpp \
brw_gs.c \
brw_gs.h \
brw_gs_state.c \
brw_gs_surface_state.c \
brw_link.cpp \
+ brw_lower_texture_gradients.cpp \
+ brw_lower_unnormalized_offset.cpp \
brw_meta_fast_clear.c \
brw_meta_stencil_blit.c \
brw_meta_updownsample.c \
brw_state.h \
brw_state_upload.c \
brw_structs.h \
+ brw_tcs.c \
brw_tcs_surface_state.c \
+ brw_tes.c \
brw_tes_surface_state.c \
brw_tex.c \
brw_tex_layout.c \
.emit = brw_upload_wm_binding_table,
};
-/** Upload the TCS binding table (if TCS is active). */
+/** Upload the TCS binding table (if tessellation stages are active). */
static void
brw_tcs_upload_binding_table(struct brw_context *brw)
{
- /* If there's no TCS, skip changing anything. */
- if (brw->tess_ctrl_program == NULL)
+ /* Skip if the tessellation stages are disabled. */
+ if (brw->tess_eval_program == NULL)
return;
/* BRW_NEW_TCS_PROG_DATA */
.dirty = {
.mesa = 0,
.brw = BRW_NEW_BATCH |
+ BRW_NEW_DEFAULT_TESS_LEVELS |
BRW_NEW_SURFACES |
BRW_NEW_TCS_CONSTBUF |
BRW_NEW_TCS_PROG_DATA,
generator(brw->intelScreen->compiler, brw,
mem_ctx, (void *) rzalloc(mem_ctx, struct brw_wm_prog_key),
(struct brw_stage_prog_data *) rzalloc(mem_ctx, struct brw_wm_prog_data),
- 0, false, "BLORP")
+ 0, false, MESA_SHADER_FRAGMENT)
{
if (debug_flag)
generator.enable_debug("blorp");
unsigned msg_length)
{
fs_inst *inst = new (mem_ctx) fs_inst(op, 16, dst, brw_message_reg(base_mrf),
- brw_imm_ud(0u));
+ brw_imm_ud(0u), brw_imm_ud(0u));
inst->base_mrf = base_mrf;
inst->mlen = msg_length;
struct gl_shader_compiler_options glsl_compiler_options[MESA_SHADER_STAGES];
};
+struct brw_compiler *
+brw_compiler_create(void *mem_ctx, const struct brw_device_info *devinfo);
+
/**
* Program key structures.
struct brw_sampler_prog_key_data tex;
};
+/** The program key for Tessellation Control Shaders. */
+struct brw_tcs_prog_key
+{
+ unsigned program_string_id;
+
+ GLenum tes_primitive_mode;
+
+ unsigned input_vertices;
+
+ /** A bitfield of per-patch outputs written. */
+ uint32_t patch_outputs_written;
+
+ /** A bitfield of per-vertex outputs written. */
+ uint64_t outputs_written;
+
+ struct brw_sampler_prog_key_data tex;
+};
+
+/** The program key for Tessellation Evaluation Shaders. */
+struct brw_tes_prog_key
+{
+ unsigned program_string_id;
+
+ /** A bitfield of per-patch inputs read. */
+ uint32_t patch_inputs_read;
+
+ /** A bitfield of per-vertex inputs read. */
+ uint64_t inputs_read;
+
+ struct brw_sampler_prog_key_data tex;
+};
+
/** The program key for Geometry Shaders. */
struct brw_gs_prog_key
{
* additional processing is applied before storing them in the VUE), the
* value is -1.
*/
- signed char varying_to_slot[BRW_VARYING_SLOT_COUNT];
+ signed char varying_to_slot[VARYING_SLOT_TESS_MAX];
/**
* Map from VUE slot to gl_varying_slot value. For slots that do not
*
* For slots that are not in use, the value is BRW_VARYING_SLOT_PAD.
*/
- signed char slot_to_varying[BRW_VARYING_SLOT_COUNT];
+ signed char slot_to_varying[VARYING_SLOT_TESS_MAX];
/**
* Total number of VUE slots in use
*/
int num_slots;
+
+ /**
+ * Number of per-patch VUE slots. Only valid for tessellation control
+ * shader outputs and tessellation evaluation shader inputs.
+ */
+ int num_per_patch_slots;
+
+ /**
+ * Number of per-vertex VUE slots. Only valid for tessellation control
+ * shader outputs and tessellation evaluation shader inputs.
+ */
+ int num_per_vertex_slots;
};
void brw_print_vue_map(FILE *fp, const struct brw_vue_map *vue_map);
GLbitfield64 slots_valid,
bool separate_shader);
+void brw_compute_tess_vue_map(struct brw_vue_map *const vue_map,
+ const GLbitfield64 slots_valid,
+ const GLbitfield is_patch);
+
enum shader_dispatch_mode {
DISPATCH_MODE_4X1_SINGLE = 0,
DISPATCH_MODE_4X2_DUAL_INSTANCE = 1,
bool uses_vertexid;
bool uses_instanceid;
+ bool uses_basevertex;
+ bool uses_baseinstance;
+ bool uses_drawid;
};
struct brw_tcs_prog_data
char **error_str);
/**
+ * Compile a tessellation control shader.
+ *
+ * Returns the final assembly and the program's size.
+ */
+const unsigned *
+brw_compile_tcs(const struct brw_compiler *compiler,
+ void *log_data,
+ void *mem_ctx,
+ const struct brw_tcs_prog_key *key,
+ struct brw_tcs_prog_data *prog_data,
+ const struct nir_shader *nir,
+ int shader_time_index,
+ unsigned *final_assembly_size,
+ char **error_str);
+
+/**
+ * Compile a tessellation evaluation shader.
+ *
+ * Returns the final assembly and the program's size.
+ */
+const unsigned *
+brw_compile_tes(const struct brw_compiler *compiler, void *log_data,
+ void *mem_ctx,
+ const struct brw_tes_prog_key *key,
+ struct brw_tes_prog_data *prog_data,
+ const struct nir_shader *shader,
+ struct gl_shader_program *shader_prog,
+ int shader_time_index,
+ unsigned *final_assembly_size,
+ char **error_str);
+
+/**
* Compile a vertex shader.
*
* Returns the final assembly and the program's size.
__DRIcontext *driContext = brw->driContext;
if (_mesa_is_winsys_fbo(ctx->DrawBuffer)) {
- dri2InvalidateDrawable(driContext->driDrawablePriv);
- dri2InvalidateDrawable(driContext->driReadablePriv);
+ if (driContext->driDrawablePriv)
+ dri2InvalidateDrawable(driContext->driDrawablePriv);
+ if (driContext->driReadablePriv)
+ dri2InvalidateDrawable(driContext->driReadablePriv);
}
}
const bool stage_exists[MESA_SHADER_STAGES] = {
[MESA_SHADER_VERTEX] = true,
- [MESA_SHADER_TESS_CTRL] = brw->gen >= 8,
- [MESA_SHADER_TESS_EVAL] = brw->gen >= 8,
+ [MESA_SHADER_TESS_CTRL] = brw->gen >= 7,
+ [MESA_SHADER_TESS_EVAL] = brw->gen >= 7,
[MESA_SHADER_GEOMETRY] = brw->gen >= 6,
[MESA_SHADER_FRAGMENT] = true,
[MESA_SHADER_COMPUTE] =
- (ctx->Const.MaxComputeWorkGroupSize[0] >= 1024) ||
+ (ctx->API == API_OPENGL_CORE &&
+ ctx->Const.MaxComputeWorkGroupSize[0] >= 1024) ||
+ (ctx->API == API_OPENGLES2 &&
+ ctx->Const.MaxComputeWorkGroupSize[0] >= 128) ||
_mesa_extension_override_enables.ARB_compute_shader,
};
brw->needs_unlit_centroid_workaround =
devinfo->needs_unlit_centroid_workaround;
- brw->must_use_separate_stencil = screen->hw_must_use_separate_stencil;
+ brw->must_use_separate_stencil = devinfo->must_use_separate_stencil;
brw->has_swizzling = screen->hw_has_swizzling;
brw->vs.base.stage = MESA_SHADER_VERTEX;
BRW_STATE_URB_FENCE = BRW_MAX_CACHE,
BRW_STATE_FRAGMENT_PROGRAM,
BRW_STATE_GEOMETRY_PROGRAM,
- BRW_STATE_TESS_CTRL_PROGRAM,
- BRW_STATE_TESS_EVAL_PROGRAM,
+ BRW_STATE_TESS_PROGRAMS,
BRW_STATE_VERTEX_PROGRAM,
BRW_STATE_CURBE_OFFSETS,
BRW_STATE_REDUCED_PRIMITIVE,
BRW_STATE_BINDING_TABLE_POINTERS,
BRW_STATE_INDICES,
BRW_STATE_VERTICES,
+ BRW_STATE_DEFAULT_TESS_LEVELS,
BRW_STATE_BATCH,
BRW_STATE_INDEX_BUFFER,
BRW_STATE_VS_CONSTBUF,
#define BRW_NEW_URB_FENCE (1ull << BRW_STATE_URB_FENCE)
#define BRW_NEW_FRAGMENT_PROGRAM (1ull << BRW_STATE_FRAGMENT_PROGRAM)
#define BRW_NEW_GEOMETRY_PROGRAM (1ull << BRW_STATE_GEOMETRY_PROGRAM)
-#define BRW_NEW_TESS_EVAL_PROGRAM (1ull << BRW_STATE_TESS_EVAL_PROGRAM)
-#define BRW_NEW_TESS_CTRL_PROGRAM (1ull << BRW_STATE_TESS_CTRL_PROGRAM)
+#define BRW_NEW_TESS_PROGRAMS (1ull << BRW_STATE_TESS_PROGRAMS)
#define BRW_NEW_VERTEX_PROGRAM (1ull << BRW_STATE_VERTEX_PROGRAM)
#define BRW_NEW_CURBE_OFFSETS (1ull << BRW_STATE_CURBE_OFFSETS)
#define BRW_NEW_REDUCED_PRIMITIVE (1ull << BRW_STATE_REDUCED_PRIMITIVE)
#define BRW_NEW_BINDING_TABLE_POINTERS (1ull << BRW_STATE_BINDING_TABLE_POINTERS)
#define BRW_NEW_INDICES (1ull << BRW_STATE_INDICES)
#define BRW_NEW_VERTICES (1ull << BRW_STATE_VERTICES)
+#define BRW_NEW_DEFAULT_TESS_LEVELS (1ull << BRW_STATE_DEFAULT_TESS_LEVELS)
/**
* Used for any batch entry with a relocated pointer that will be used
* by any 3D rendering.
uint32_t pma_stall_bits;
struct {
- /** The value of gl_BaseVertex for the current _mesa_prim. */
- int gl_basevertex;
+ struct {
+ /** The value of gl_BaseVertex for the current _mesa_prim. */
+ int gl_basevertex;
+
+ /** The value of gl_BaseInstance for the current _mesa_prim. */
+ int gl_baseinstance;
+ } params;
/**
* Buffer and offset used for GL_ARB_shader_draw_parameters
*/
drm_intel_bo *draw_params_bo;
uint32_t draw_params_offset;
+
+ /**
+ * The value of gl_DrawID for the current _mesa_prim. This always comes
+ * in from it's own vertex buffer since it's not part of the indirect
+ * draw parameters.
+ */
+ int gl_drawid;
+ drm_intel_bo *draw_id_bo;
+ uint32_t draw_id_offset;
} draw;
struct {
struct {
GLuint vsize; /* vertex size plus header in urb registers */
GLuint gsize; /* GS output size in urb registers */
+ GLuint hsize; /* Tessellation control output size in urb registers */
+ GLuint dsize; /* Tessellation evaluation output size in urb registers */
GLuint csize; /* constant buffer size in urb registers */
GLuint sfsize; /* setup data size in urb registers */
GLuint max_gs_entries; /* Maximum number of GS entries */
GLuint nr_vs_entries;
+ GLuint nr_hs_entries;
+ GLuint nr_ds_entries;
GLuint nr_gs_entries;
GLuint nr_clip_entries;
GLuint nr_sf_entries;
GLuint nr_cs_entries;
GLuint vs_start;
+ GLuint hs_start;
+ GLuint ds_start;
GLuint gs_start;
GLuint clip_start;
GLuint sf_start;
* URB space for the GS.
*/
bool gs_present;
+
+ /* True if the most recently sent _3DSTATE_URB message allocated
+ * URB space for the HS and DS.
+ */
+ bool tess_present;
} urb;
/* gen7_urb.c */
void
gen7_emit_push_constant_state(struct brw_context *brw, unsigned vs_size,
+ unsigned hs_size, unsigned ds_size,
unsigned gs_size, unsigned fs_size);
void
gen7_emit_urb_state(struct brw_context *brw,
- unsigned nr_vs_entries, unsigned vs_size,
- unsigned vs_start, unsigned nr_gs_entries,
+ unsigned nr_vs_entries,
+ unsigned vs_size, unsigned vs_start,
+ unsigned nr_hs_entries,
+ unsigned hs_size, unsigned hs_start,
+ unsigned nr_ds_entries,
+ unsigned ds_size, unsigned ds_start,
+ unsigned nr_gs_entries,
unsigned gs_size, unsigned gs_start);
return (const struct brw_vertex_program *) p;
}
+static inline struct brw_tess_ctrl_program *
+brw_tess_ctrl_program(struct gl_tess_ctrl_program *p)
+{
+ return (struct brw_tess_ctrl_program *) p;
+}
+
+static inline struct brw_tess_eval_program *
+brw_tess_eval_program(struct gl_tess_eval_program *p)
+{
+ return (struct brw_tess_eval_program *) p;
+}
+
static inline struct brw_geometry_program *
brw_geometry_program(struct gl_geometry_program *p)
{
# define GEN7_3DPRIM_VERTEXBUFFER_ACCESS_SEQUENTIAL (0 << 8)
# define GEN7_3DPRIM_VERTEXBUFFER_ACCESS_RANDOM (1 << 8)
+#ifndef _3DPRIM_POINTLIST /* FIXME: Avoid clashing with defines from bdw_pack.h */
#define _3DPRIM_POINTLIST 0x01
#define _3DPRIM_LINELIST 0x02
#define _3DPRIM_LINESTRIP 0x03
#define _3DPRIM_TRIFAN_NOSTIPPLE 0x16
#define _3DPRIM_PATCHLIST(n) ({ assert(n > 0 && n <= 32); 0x20 + (n - 1); })
+#endif /* bdw_pack.h */
/* We use this offset to be able to pass native primitive types in struct
* _mesa_prim::mode. Native primitive types are BRW_PRIM_OFFSET +
* UD immediate).
*/
SHADER_OPCODE_MOV_INDIRECT,
+
+ VEC4_OPCODE_URB_READ,
+ TCS_OPCODE_GET_INSTANCE_ID,
+ TCS_OPCODE_URB_WRITE,
+ TCS_OPCODE_SET_INPUT_URB_OFFSETS,
+ TCS_OPCODE_SET_OUTPUT_URB_OFFSETS,
+ TCS_OPCODE_GET_PRIMITIVE_ID,
+ TCS_OPCODE_CREATE_BARRIER_HEADER,
+ TCS_OPCODE_SRC0_010_IS_ZERO,
+ TCS_OPCODE_RELEASE_INPUT,
+ TCS_OPCODE_THREAD_END,
+
+ TES_OPCODE_GET_PRIMITIVE_ID,
+ TES_OPCODE_CREATE_INPUT_READ_HEADER,
+ TES_OPCODE_ADD_INDIRECT_URB_OFFSET,
};
enum brw_urb_write_flags {
}
};
+/*
+ * Note: for all KBL SKUs, the PRM says SKL for GS entries, not SKL+.
+ * There's no KBL entry. Using the default SKL (GEN9) GS entries value.
+ */
+
+/*
+ * Both SKL and KBL support a maximum of 64 threads per
+ * Pixel Shader Dispatch (PSD) unit.
+ */
+#define KBL_MAX_THREADS_PER_PSD 64
+
+static const struct brw_device_info brw_device_info_kbl_gt1 = {
+ GEN9_FEATURES,
+ .gt = 1,
+
+ .max_cs_threads = 7 * 6,
+ .max_wm_threads = KBL_MAX_THREADS_PER_PSD * 2,
+ .urb.size = 192,
+ .num_slices = 1,
+};
+
+static const struct brw_device_info brw_device_info_kbl_gt1_5 = {
+ GEN9_FEATURES,
+ .gt = 1,
+
+ .max_cs_threads = 7 * 6,
+ .max_wm_threads = KBL_MAX_THREADS_PER_PSD * 3,
+ .num_slices = 1,
+};
+
+static const struct brw_device_info brw_device_info_kbl_gt2 = {
+ GEN9_FEATURES,
+ .gt = 2,
+
+ .max_wm_threads = KBL_MAX_THREADS_PER_PSD * 3,
+ .num_slices = 1,
+};
+
+static const struct brw_device_info brw_device_info_kbl_gt3 = {
+ GEN9_FEATURES,
+ .gt = 3,
+
+ .max_wm_threads = KBL_MAX_THREADS_PER_PSD * 6,
+ .num_slices = 2,
+};
+
+static const struct brw_device_info brw_device_info_kbl_gt4 = {
+ GEN9_FEATURES,
+ .gt = 4,
+
+ .max_wm_threads = KBL_MAX_THREADS_PER_PSD * 9,
+ /*
+ * From the "L3 Allocation and Programming" documentation:
+ *
+ * "URB is limited to 1008KB due to programming restrictions. This
+ * is not a restriction of the L3 implementation, but of the FF and
+ * other clients. Therefore, in a GT4 implementation it is
+ * possible for the programmed allocation of the L3 data array to
+ * provide 3*384KB=1152KB for URB, but only 1008KB of this
+ * will be used."
+ */
+ .urb.size = 1008 / 3,
+ .num_slices = 3,
+};
+
const struct brw_device_info *
brw_get_device_info(int devid)
{
return devinfo;
}
+
+const char *
+brw_get_device_name(int devid)
+{
+ switch (devid) {
+#undef CHIPSET
+#define CHIPSET(id, family, name) case id: return name;
+#include "pci_ids/i965_pci_ids.h"
+ default:
+ return NULL;
+ }
+}
};
const struct brw_device_info *brw_get_device_info(int devid);
+const char *brw_get_device_name(int devid);
#define FILE_DEBUG_FLAG DEBUG_PRIMS
-static const GLuint prim_to_hw_prim[GL_TRIANGLE_STRIP_ADJACENCY+1] = {
- [GL_POINTS] =_3DPRIM_POINTLIST,
- [GL_LINES] = _3DPRIM_LINELIST,
- [GL_LINE_LOOP] = _3DPRIM_LINELOOP,
- [GL_LINE_STRIP] = _3DPRIM_LINESTRIP,
- [GL_TRIANGLES] = _3DPRIM_TRILIST,
- [GL_TRIANGLE_STRIP] = _3DPRIM_TRISTRIP,
- [GL_TRIANGLE_FAN] = _3DPRIM_TRIFAN,
- [GL_QUADS] = _3DPRIM_QUADLIST,
- [GL_QUAD_STRIP] = _3DPRIM_QUADSTRIP,
- [GL_POLYGON] = _3DPRIM_POLYGON,
- [GL_LINES_ADJACENCY] = _3DPRIM_LINELIST_ADJ,
- [GL_LINE_STRIP_ADJACENCY] = _3DPRIM_LINESTRIP_ADJ,
- [GL_TRIANGLES_ADJACENCY] = _3DPRIM_TRILIST_ADJ,
- [GL_TRIANGLE_STRIP_ADJACENCY] = _3DPRIM_TRISTRIP_ADJ,
-};
-
static const GLenum reduced_prim[GL_POLYGON+1] = {
[GL_POINTS] = GL_POINTS,
[GL_POLYGON] = GL_TRIANGLES
};
-uint32_t
-get_hw_prim_for_gl_prim(int mode)
-{
- if (mode >= BRW_PRIM_OFFSET)
- return mode - BRW_PRIM_OFFSET;
- else {
- assert(mode < ARRAY_SIZE(prim_to_hw_prim));
- return prim_to_hw_prim[mode];
- }
-}
-
-
/* When the primitive changes, set a state bit and re-validate. Not
* the nicest and would rather deal with this by having all the
* programs be immune to the active primitive (ie. cope with all
}
}
- brw->draw.gl_basevertex =
+ /* Determine if we need to flag BRW_NEW_VERTICES for updating the
+ * gl_BaseVertexARB or gl_BaseInstanceARB values. For indirect draw, we
+ * always flag if the shader uses one of the values. For direct draws,
+ * we only flag if the values change.
+ */
+ const int new_basevertex =
prims[i].indexed ? prims[i].basevertex : prims[i].start;
+ const int new_baseinstance = prims[i].base_instance;
+ if (i > 0) {
+ const bool uses_draw_parameters =
+ brw->vs.prog_data->uses_basevertex ||
+ brw->vs.prog_data->uses_baseinstance;
+
+ if ((uses_draw_parameters && prims[i].is_indirect) ||
+ (brw->vs.prog_data->uses_basevertex &&
+ brw->draw.params.gl_basevertex != new_basevertex) ||
+ (brw->vs.prog_data->uses_baseinstance &&
+ brw->draw.params.gl_baseinstance != new_baseinstance))
+ brw->ctx.NewDriverState |= BRW_NEW_VERTICES;
+ }
+ brw->draw.params.gl_basevertex = new_basevertex;
+ brw->draw.params.gl_baseinstance = new_baseinstance;
drm_intel_bo_unreference(brw->draw.draw_params_bo);
if (prims[i].is_indirect) {
brw->draw.draw_params_offset = 0;
}
+ /* gl_DrawID always needs its own vertex buffer since it's not part of
+ * the indirect parameter buffer. If the program uses gl_DrawID we need
+ * to flag BRW_NEW_VERTICES. For the first iteration, we don't have
+ * valid brw->vs.prog_data, but we always flag BRW_NEW_VERTICES before
+ * the loop.
+ */
+ brw->draw.gl_drawid = prims[i].draw_id;
+ drm_intel_bo_unreference(brw->draw.draw_id_bo);
+ brw->draw.draw_id_bo = NULL;
+ if (i > 0 && brw->vs.prog_data->uses_drawid)
+ brw->ctx.NewDriverState |= BRW_NEW_VERTICES;
+
if (brw->gen < 6)
brw_set_prim(brw, &prims[i]);
else
brw_prepare_shader_draw_parameters(struct brw_context *brw)
{
/* For non-indirect draws, upload gl_BaseVertex. */
- if (brw->vs.prog_data->uses_vertexid && brw->draw.draw_params_bo == NULL) {
- intel_upload_data(brw, &brw->draw.gl_basevertex, 4, 4,
+ if ((brw->vs.prog_data->uses_basevertex ||
+ brw->vs.prog_data->uses_baseinstance) &&
+ brw->draw.draw_params_bo == NULL) {
+ intel_upload_data(brw, &brw->draw.params, sizeof(brw->draw.params), 4,
&brw->draw.draw_params_bo,
&brw->draw.draw_params_offset);
}
+
+ if (brw->vs.prog_data->uses_drawid) {
+ intel_upload_data(brw, &brw->draw.gl_drawid, sizeof(brw->draw.gl_drawid), 4,
+ &brw->draw.draw_id_bo,
+ &brw->draw.draw_id_offset);
+ }
}
/**
brw_emit_query_begin(brw);
unsigned nr_elements = brw->vb.nr_enabled;
- if (brw->vs.prog_data->uses_vertexid || brw->vs.prog_data->uses_instanceid)
+ if (brw->vs.prog_data->uses_vertexid || brw->vs.prog_data->uses_instanceid ||
+ brw->vs.prog_data->uses_basevertex || brw->vs.prog_data->uses_baseinstance)
++nr_elements;
+ if (brw->vs.prog_data->uses_drawid)
+ nr_elements++;
/* If the VS doesn't read any inputs (calculating vertex position from
* a state variable for some reason, for example), emit a single pad
/* Now emit VB and VEP state packets.
*/
- unsigned nr_buffers =
- brw->vb.nr_buffers + brw->vs.prog_data->uses_vertexid;
+ const bool uses_draw_params =
+ brw->vs.prog_data->uses_basevertex ||
+ brw->vs.prog_data->uses_baseinstance;
+ const unsigned nr_buffers = brw->vb.nr_buffers +
+ uses_draw_params + brw->vs.prog_data->uses_drawid;
if (nr_buffers) {
if (brw->gen >= 6) {
}
- if (brw->vs.prog_data->uses_vertexid) {
+ if (uses_draw_params) {
EMIT_VERTEX_BUFFER_STATE(brw, brw->vb.nr_buffers,
brw->draw.draw_params_bo,
brw->draw.draw_params_bo->size - 1,
0, /* stride */
0); /* step rate */
}
+
+ if (brw->vs.prog_data->uses_drawid) {
+ EMIT_VERTEX_BUFFER_STATE(brw, brw->vb.nr_buffers + 1,
+ brw->draw.draw_id_bo,
+ brw->draw.draw_id_bo->size - 1,
+ brw->draw.draw_id_offset,
+ 0, /* stride */
+ 0); /* step rate */
+ }
+
ADVANCE_BATCH();
}
((i * 4) << BRW_VE1_DST_OFFSET_SHIFT));
}
- if (brw->vs.prog_data->uses_vertexid || brw->vs.prog_data->uses_instanceid) {
+ if (brw->vs.prog_data->uses_vertexid || brw->vs.prog_data->uses_instanceid ||
+ brw->vs.prog_data->uses_basevertex || brw->vs.prog_data->uses_baseinstance) {
uint32_t dw0 = 0, dw1 = 0;
uint32_t comp0 = BRW_VE1_COMPONENT_STORE_0;
uint32_t comp1 = BRW_VE1_COMPONENT_STORE_0;
uint32_t comp2 = BRW_VE1_COMPONENT_STORE_0;
uint32_t comp3 = BRW_VE1_COMPONENT_STORE_0;
- if (brw->vs.prog_data->uses_vertexid) {
+ if (brw->vs.prog_data->uses_basevertex)
comp0 = BRW_VE1_COMPONENT_STORE_SRC;
+
+ if (brw->vs.prog_data->uses_baseinstance)
+ comp1 = BRW_VE1_COMPONENT_STORE_SRC;
+
+ if (brw->vs.prog_data->uses_vertexid)
comp2 = BRW_VE1_COMPONENT_STORE_VID;
- }
- if (brw->vs.prog_data->uses_instanceid) {
+ if (brw->vs.prog_data->uses_instanceid)
comp3 = BRW_VE1_COMPONENT_STORE_IID;
- }
dw1 = (comp0 << BRW_VE1_COMPONENT_0_SHIFT) |
(comp1 << BRW_VE1_COMPONENT_1_SHIFT) |
if (brw->gen >= 6) {
dw0 |= GEN6_VE0_VALID |
brw->vb.nr_buffers << GEN6_VE0_INDEX_SHIFT |
- BRW_SURFACEFORMAT_R32_UINT << BRW_VE0_FORMAT_SHIFT;
+ BRW_SURFACEFORMAT_R32G32_UINT << BRW_VE0_FORMAT_SHIFT;
} else {
dw0 |= BRW_VE0_VALID |
brw->vb.nr_buffers << BRW_VE0_INDEX_SHIFT |
- BRW_SURFACEFORMAT_R32_UINT << BRW_VE0_FORMAT_SHIFT;
+ BRW_SURFACEFORMAT_R32G32_UINT << BRW_VE0_FORMAT_SHIFT;
dw1 |= (i * 4) << BRW_VE1_DST_OFFSET_SHIFT;
}
OUT_BATCH(dw1);
}
+ if (brw->vs.prog_data->uses_drawid) {
+ uint32_t dw0 = 0, dw1 = 0;
+
+ dw1 = (BRW_VE1_COMPONENT_STORE_SRC << BRW_VE1_COMPONENT_0_SHIFT) |
+ (BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_1_SHIFT) |
+ (BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) |
+ (BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_3_SHIFT);
+
+ if (brw->gen >= 6) {
+ dw0 |= GEN6_VE0_VALID |
+ ((brw->vb.nr_buffers + 1) << GEN6_VE0_INDEX_SHIFT) |
+ (BRW_SURFACEFORMAT_R32_UINT << BRW_VE0_FORMAT_SHIFT);
+ } else {
+ dw0 |= BRW_VE0_VALID |
+ ((brw->vb.nr_buffers + 1) << BRW_VE0_INDEX_SHIFT) |
+ (BRW_SURFACEFORMAT_R32_UINT << BRW_VE0_FORMAT_SHIFT);
+
+ dw1 |= (i * 4) << BRW_VE1_DST_OFFSET_SHIFT;
+ }
+
+ OUT_BATCH(dw0);
+ OUT_BATCH(dw1);
+ }
+
if (brw->gen >= 6 && gen6_edgeflag_input) {
uint32_t format =
brw_get_vertex_surface_type(brw, gen6_edgeflag_input->glarray);
static int
get_3src_subreg_nr(struct brw_reg reg)
{
- if (reg.vstride == BRW_VERTICAL_STRIDE_0) {
- assert(brw_is_single_value_swizzle(reg.swizzle));
- return reg.subnr / 4 + BRW_GET_SWZ(reg.swizzle, 0);
- } else {
- return reg.subnr / 4;
- }
+ /* Normally, SubRegNum is in bytes (0..31). However, 3-src instructions
+ * use 32-bit units (components 0..7). Since they only support F/D/UD
+ * types, this doesn't lose any flexibility, but uses fewer bits.
+ */
+ return reg.subnr / 4;
}
static brw_inst *
case GLSL_TYPE_ERROR:
case GLSL_TYPE_INTERFACE:
case GLSL_TYPE_DOUBLE:
+ case GLSL_TYPE_FUNCTION:
unreachable("not reached");
}
case SHADER_OPCODE_LOD_LOGICAL:
case SHADER_OPCODE_TG4_LOGICAL:
case SHADER_OPCODE_TG4_OFFSET_LOGICAL:
- assert(src[8].file == IMM && src[9].file == IMM);
+ assert(src[9].file == IMM && src[10].file == IMM);
/* Texture coordinates. */
if (i == 0)
- return src[8].ud;
+ return src[9].ud;
/* Texture derivatives. */
else if ((i == 2 || i == 3) && opcode == SHADER_OPCODE_TXD_LOGICAL)
- return src[9].ud;
+ return src[10].ud;
/* Texture offset. */
- else if (i == 7)
+ else if (i == 8)
return 2;
/* MCS */
else if (i == 5 && opcode == SHADER_OPCODE_TXF_CMS_W_LOGICAL)
assert(stage == MESA_SHADER_VERTEX);
int count = _mesa_bitcount_64(vs_prog_data->inputs_read);
- if (vs_prog_data->uses_vertexid || vs_prog_data->uses_instanceid)
+ if (vs_prog_data->uses_vertexid || vs_prog_data->uses_instanceid ||
+ vs_prog_data->uses_basevertex || vs_prog_data->uses_baseinstance)
+ count++;
+ if (vs_prog_data->uses_drawid)
count++;
/* Each attribute is 4 regs. */
}
void
+fs_visitor::assign_tes_urb_setup()
+{
+ assert(stage == MESA_SHADER_TESS_EVAL);
+
+ brw_vue_prog_data *vue_prog_data = (brw_vue_prog_data *) prog_data;
+
+ first_non_payload_grf += 8 * vue_prog_data->urb_read_length;
+
+ /* Rewrite all ATTR file references to HW_REGs. */
+ foreach_block_and_inst(block, fs_inst, inst, cfg) {
+ convert_attr_sources_to_hw_regs(inst);
+ }
+}
+
+void
fs_visitor::assign_gs_urb_setup()
{
assert(stage == MESA_SHADER_GEOMETRY);
brw_wm_prog_key *key = (brw_wm_prog_key*) this->key;
int base_mrf = 1;
int color_mrf = base_mrf + 2;
+ fs_inst *mov;
- fs_inst *mov = bld.exec_all().group(4, 0)
- .MOV(brw_message_reg(color_mrf),
- fs_reg(UNIFORM, 0, BRW_REGISTER_TYPE_F));
+ if (uniforms == 1) {
+ mov = bld.exec_all().group(4, 0)
+ .MOV(brw_message_reg(color_mrf),
+ fs_reg(UNIFORM, 0, BRW_REGISTER_TYPE_F));
+ } else {
+ struct brw_reg reg =
+ brw_reg(BRW_GENERAL_REGISTER_FILE,
+ 2, 3, 0, 0, BRW_REGISTER_TYPE_F,
+ BRW_VERTICAL_STRIDE_8,
+ BRW_WIDTH_2,
+ BRW_HORIZONTAL_STRIDE_4, BRW_SWIZZLE_XYZW, WRITEMASK_XYZW);
+
+ mov = bld.exec_all().group(4, 0)
+ .MOV(vec4(brw_message_reg(color_mrf)), fs_reg(reg));
+ }
fs_inst *write;
if (key->nr_color_regions == 1) {
assign_curb_setup();
/* Now that we have the uniform assigned, go ahead and force it to a vec4. */
- assert(mov->src[0].file == FIXED_GRF);
- mov->src[0] = brw_vec4_grf(mov->src[0].nr, 0);
+ if (uniforms == 1) {
+ assert(mov->src[0].file == FIXED_GRF);
+ mov->src[0] = brw_vec4_grf(mov->src[0].nr, 0);
+ }
}
/**
*/
assert(mul->src[1].type == BRW_REGISTER_TYPE_D ||
mul->src[1].type == BRW_REGISTER_TYPE_UD);
- mul->src[1].type = (type_is_signed(mul->src[1].type) ?
- BRW_REGISTER_TYPE_W : BRW_REGISTER_TYPE_UW);
+ mul->src[1].type = BRW_REGISTER_TYPE_UW;
mul->src[1].stride *= 2;
} else if (devinfo->gen == 7 && !devinfo->is_haswell &&
const fs_reg &coordinate,
const fs_reg &shadow_c,
const fs_reg &lod, const fs_reg &lod2,
+ const fs_reg &surface,
const fs_reg &sampler,
unsigned coord_components,
unsigned grad_components)
inst->opcode = op;
inst->src[0] = reg_undef;
- inst->src[1] = sampler;
- inst->resize_sources(2);
+ inst->src[1] = surface;
+ inst->src[2] = sampler;
+ inst->resize_sources(3);
inst->base_mrf = msg_begin.nr;
inst->mlen = msg_end.nr - msg_begin.nr;
inst->header_size = 1;
const fs_reg &shadow_c,
fs_reg lod, fs_reg lod2,
const fs_reg &sample_index,
+ const fs_reg &surface,
const fs_reg &sampler,
const fs_reg &offset_value,
unsigned coord_components,
inst->opcode = op;
inst->src[0] = reg_undef;
- inst->src[1] = sampler;
- inst->resize_sources(2);
+ inst->src[1] = surface;
+ inst->src[2] = sampler;
+ inst->resize_sources(3);
inst->base_mrf = message.nr;
inst->mlen = msg_end.nr - message.nr;
inst->header_size = header_size;
const fs_reg &shadow_c,
fs_reg lod, fs_reg lod2,
const fs_reg &sample_index,
- const fs_reg &mcs, const fs_reg &sampler,
+ const fs_reg &mcs,
+ const fs_reg &surface,
+ const fs_reg &sampler,
fs_reg offset_value,
unsigned coord_components,
unsigned grad_components)
/* Generate the SEND. */
inst->opcode = op;
inst->src[0] = src_payload;
- inst->src[1] = sampler;
- inst->resize_sources(2);
+ inst->src[1] = surface;
+ inst->src[2] = sampler;
+ inst->resize_sources(3);
inst->base_mrf = -1;
inst->mlen = mlen;
inst->header_size = header_size;
const fs_reg &lod2 = inst->src[3];
const fs_reg &sample_index = inst->src[4];
const fs_reg &mcs = inst->src[5];
- const fs_reg &sampler = inst->src[6];
- const fs_reg &offset_value = inst->src[7];
- assert(inst->src[8].file == IMM && inst->src[9].file == IMM);
- const unsigned coord_components = inst->src[8].ud;
- const unsigned grad_components = inst->src[9].ud;
+ const fs_reg &surface = inst->src[6];
+ const fs_reg &sampler = inst->src[7];
+ const fs_reg &offset_value = inst->src[8];
+ assert(inst->src[9].file == IMM && inst->src[10].file == IMM);
+ const unsigned coord_components = inst->src[9].ud;
+ const unsigned grad_components = inst->src[10].ud;
if (devinfo->gen >= 7) {
lower_sampler_logical_send_gen7(bld, inst, op, coordinate,
shadow_c, lod, lod2, sample_index,
- mcs, sampler, offset_value,
+ mcs, surface, sampler, offset_value,
coord_components, grad_components);
} else if (devinfo->gen >= 5) {
lower_sampler_logical_send_gen5(bld, inst, op, coordinate,
shadow_c, lod, lod2, sample_index,
- sampler, offset_value,
+ surface, sampler, offset_value,
coord_components, grad_components);
} else {
lower_sampler_logical_send_gen4(bld, inst, op, coordinate,
- shadow_c, lod, lod2, sampler,
+ shadow_c, lod, lod2,
+ surface, sampler,
coord_components, grad_components);
}
}
}
bool
+fs_visitor::run_tes()
+{
+ assert(stage == MESA_SHADER_TESS_EVAL);
+
+ /* R0: thread header, R1-3: gl_TessCoord.xyz, R4: URB handles */
+ payload.num_regs = 5;
+
+ if (shader_time_index >= 0)
+ emit_shader_time_begin();
+
+ emit_nir_code();
+
+ if (failed)
+ return false;
+
+ emit_urb_writes();
+
+ if (shader_time_index >= 0)
+ emit_shader_time_end();
+
+ calculate_cfg();
+
+ optimize();
+
+ assign_curb_setup();
+ assign_tes_urb_setup();
+
+ fixup_3src_null_dest();
+ allocate_registers();
+
+ return !failed;
+}
+
+bool
fs_visitor::run_gs()
{
assert(stage == MESA_SHADER_GEOMETRY);
}
fs_generator g(compiler, log_data, mem_ctx, (void *) key, &prog_data->base,
- v.promoted_constants, v.runtime_check_aads_emit, "FS");
+ v.promoted_constants, v.runtime_check_aads_emit,
+ MESA_SHADER_FRAGMENT);
if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
g.enable_debug(ralloc_asprintf(mem_ctx, "%s fragment shader %s",
}
fs_generator g(compiler, log_data, mem_ctx, (void*) key, &prog_data->base,
- v8.promoted_constants, v8.runtime_check_aads_emit, "CS");
+ v8.promoted_constants, v8.runtime_check_aads_emit,
+ MESA_SHADER_COMPUTE);
if (INTEL_DEBUG & DEBUG_CS) {
char *name = ralloc_asprintf(mem_ctx, "%s compute shader %s",
shader->info.label ? shader->info.label :
struct gl_program *prog,
const nir_shader *shader,
unsigned dispatch_width,
- int shader_time_index);
+ int shader_time_index,
+ const struct brw_vue_map *input_vue_map = NULL);
fs_visitor(const struct brw_compiler *compiler, void *log_data,
void *mem_ctx,
struct brw_gs_compile *gs_compile,
bool run_fs(bool do_rep_send);
bool run_vs(gl_clip_plane *clip_planes);
+ bool run_tes();
bool run_gs();
bool run_cs();
void optimize();
void assign_urb_setup();
void convert_attr_sources_to_hw_regs(fs_inst *inst);
void assign_vs_urb_setup();
+ void assign_tes_urb_setup();
void assign_gs_urb_setup();
bool assign_regs(bool allow_spilling);
void assign_regs_trivial();
fs_reg mcs,
int gather_component,
bool is_cube_array,
+ uint32_t surface,
+ fs_reg surface_reg,
uint32_t sampler,
fs_reg sampler_reg);
fs_reg emit_mcs_fetch(const fs_reg &coordinate, unsigned components,
nir_intrinsic_instr *instr);
void nir_emit_intrinsic(const brw::fs_builder &bld,
nir_intrinsic_instr *instr);
+ void nir_emit_tes_intrinsic(const brw::fs_builder &bld,
+ nir_intrinsic_instr *instr);
void nir_emit_ssbo_atomic(const brw::fs_builder &bld,
int op, nir_intrinsic_instr *instr);
void nir_emit_shared_atomic(const brw::fs_builder &bld,
fs_reg get_nir_src(nir_src src);
fs_reg get_nir_dest(nir_dest dest);
fs_reg get_nir_image_deref(const nir_deref_var *deref);
+ fs_reg get_indirect_offset(nir_intrinsic_instr *instr);
void emit_percomp(const brw::fs_builder &bld, const fs_inst &inst,
unsigned wr_mask);
struct brw_stage_prog_data *prog_data;
struct gl_program *prog;
+ const struct brw_vue_map *input_vue_map;
+
int *virtual_grf_start;
int *virtual_grf_end;
brw::fs_live_variables *live_intervals;
struct brw_stage_prog_data *prog_data,
unsigned promoted_constants,
bool runtime_check_aads_emit,
- const char *stage_abbrev);
+ gl_shader_stage stage);
~fs_generator();
void enable_debug(const char *shader_name);
void generate_linterp(fs_inst *inst, struct brw_reg dst,
struct brw_reg *src);
void generate_tex(fs_inst *inst, struct brw_reg dst, struct brw_reg src,
+ struct brw_reg surface_index,
struct brw_reg sampler_index);
void generate_get_buffer_size(fs_inst *inst, struct brw_reg dst,
struct brw_reg src,
bool runtime_check_aads_emit;
bool debug_flag;
const char *shader_name;
- const char *stage_abbrev;
+ gl_shader_stage stage;
void *mem_ctx;
};
ir_expression *expr = ir->rhs->as_expression();
bool found_vector = false;
unsigned int i, vector_elements = 1;
- ir_variable *op_var[3];
+ ir_variable *op_var[4];
if (!expr)
return visit_continue;
}
break;
- case ir_unop_any: {
- ir_expression *temp;
- temp = new(mem_ctx) ir_expression(ir_binop_logic_or,
- element_type,
- get_element(op_var[0], 0),
- get_element(op_var[0], 1));
-
- for (i = 2; i < vector_elements; i++) {
- temp = new(mem_ctx) ir_expression(ir_binop_logic_or,
- element_type,
- get_element(op_var[0], i),
- temp);
- }
- assign(ir, 0, temp);
- break;
- }
-
case ir_binop_dot: {
ir_expression *last = NULL;
for (i = 0; i < vector_elements; i++) {
case ir_unop_noise:
unreachable("noise should have been broken down to function call");
- case ir_binop_bfm: {
- /* Does not need to be scalarized, since its result will be identical
- * for all channels.
- */
- ir_rvalue *op0 = get_element(op_var[0], 0);
- ir_rvalue *op1 = get_element(op_var[1], 0);
-
- assign(ir, 0, new(mem_ctx) ir_expression(expr->operation,
- element_type,
- op0,
- op1));
- break;
- }
-
case ir_binop_ubo_load:
case ir_unop_get_buffer_size:
unreachable("not yet supported");
}
break;
- case ir_triop_bfi: {
- /* Only a single BFM is needed for multiple BFIs. */
- ir_rvalue *op0 = get_element(op_var[0], 0);
-
+ case ir_quadop_bitfield_insert:
for (i = 0; i < vector_elements; i++) {
+ ir_rvalue *op0 = get_element(op_var[0], i);
ir_rvalue *op1 = get_element(op_var[1], i);
ir_rvalue *op2 = get_element(op_var[2], i);
+ ir_rvalue *op3 = get_element(op_var[3], i);
assign(ir, i, new(mem_ctx) ir_expression(expr->operation,
element_type,
- op0->clone(mem_ctx, NULL),
+ op0,
op1,
- op2));
+ op2,
+ op3));
}
break;
- }
case ir_unop_pack_snorm_2x16:
case ir_unop_pack_snorm_4x8:
case ir_binop_ldexp:
case ir_binop_vector_extract:
case ir_triop_vector_insert:
- case ir_quadop_bitfield_insert:
case ir_quadop_vector:
case ir_unop_ssbo_unsized_array_length:
unreachable("should have been lowered");
using namespace brw;
+static const bool debug = false;
+
/* Returns whether an instruction could co-issue if its immediate source were
* replaced with a GRF source.
*/
if (cfg->num_blocks != 1)
qsort(table.imm, table.len, sizeof(struct imm), compare);
-
/* Insert MOVs to load the constant values into GRFs. */
fs_reg reg(VGRF, alloc.allocate(dispatch_width / 8));
reg.stride = 0;
reg->subreg_offset = table.imm[i].subreg_offset;
reg->stride = 0;
reg->negate = signbit(reg->f) != signbit(table.imm[i].val);
- assert(fabsf(reg->f) == table.imm[i].val);
+ assert((isnan(reg->f) && isnan(table.imm[i].val)) ||
+ fabsf(reg->f) == table.imm[i].val);
+ }
+ }
+
+ if (debug) {
+ for (int i = 0; i < table.len; i++) {
+ struct imm *imm = &table.imm[i];
+
+ printf("%.3fF - block %3d, reg %3d sub %2d, Uses: (%2d, %2d), "
+ "IP: %4d to %4d, length %4d\n",
+ imm->val,
+ imm->block->num,
+ imm->nr,
+ imm->subreg_offset,
+ imm->must_promote,
+ imm->uses_by_coissue,
+ imm->first_use_ip,
+ imm->last_use_ip,
+ imm->last_use_ip - imm->first_use_ip);
}
}
struct brw_stage_prog_data *prog_data,
unsigned promoted_constants,
bool runtime_check_aads_emit,
- const char *stage_abbrev)
+ gl_shader_stage stage)
: compiler(compiler), log_data(log_data),
devinfo(compiler->devinfo), key(key),
prog_data(prog_data),
promoted_constants(promoted_constants),
runtime_check_aads_emit(runtime_check_aads_emit), debug_flag(false),
- stage_abbrev(stage_abbrev), mem_ctx(mem_ctx)
+ stage(stage), mem_ctx(mem_ctx)
{
p = rzalloc(mem_ctx, struct brw_codegen);
brw_init_codegen(devinfo, p, mem_ctx);
void
fs_generator::generate_tex(fs_inst *inst, struct brw_reg dst, struct brw_reg src,
+ struct brw_reg surface_index,
struct brw_reg sampler_index)
{
int msg_type = -1;
/* Set the offset bits in DWord 2. */
brw_MOV(p, get_element_ud(header_reg, 2),
brw_imm_ud(inst->offset));
+ } else if (stage != MESA_SHADER_VERTEX &&
+ stage != MESA_SHADER_FRAGMENT) {
+ /* The vertex and fragment stages have g0.2 set to 0, so
+ * header0.2 is 0 when g0 is copied. Other stages may not, so we
+ * must set it to 0 to avoid setting undesirable bits in the
+ * message.
+ */
+ brw_MOV(p, get_element_ud(header_reg, 2), brw_imm_ud(0));
}
brw_adjust_sampler_state_pointer(p, header_reg, sampler_index);
? prog_data->binding_table.gather_texture_start
: prog_data->binding_table.texture_start;
- if (sampler_index.file == BRW_IMMEDIATE_VALUE) {
+ if (surface_index.file == BRW_IMMEDIATE_VALUE &&
+ sampler_index.file == BRW_IMMEDIATE_VALUE) {
+ uint32_t surface = surface_index.ud;
uint32_t sampler = sampler_index.ud;
brw_SAMPLE(p,
retype(dst, BRW_REGISTER_TYPE_UW),
inst->base_mrf,
src,
- sampler + base_binding_table_index,
+ surface + base_binding_table_index,
sampler % 16,
msg_type,
rlen,
simd_mode,
return_format);
- brw_mark_surface_used(prog_data, sampler + base_binding_table_index);
+ brw_mark_surface_used(prog_data, surface + base_binding_table_index);
} else {
/* Non-const sampler index */
struct brw_reg addr = vec1(retype(brw_address_reg(0), BRW_REGISTER_TYPE_UD));
+ struct brw_reg surface_reg = vec1(retype(surface_index, BRW_REGISTER_TYPE_UD));
struct brw_reg sampler_reg = vec1(retype(sampler_index, BRW_REGISTER_TYPE_UD));
brw_push_insn_state(p);
brw_set_default_mask_control(p, BRW_MASK_DISABLE);
brw_set_default_access_mode(p, BRW_ALIGN_1);
- /* addr = ((sampler * 0x101) + base_binding_table_index) & 0xfff */
- brw_MUL(p, addr, sampler_reg, brw_imm_uw(0x101));
+ if (memcmp(&surface_reg, &sampler_reg, sizeof(surface_reg)) == 0) {
+ brw_MUL(p, addr, sampler_reg, brw_imm_uw(0x101));
+ } else {
+ brw_SHL(p, addr, sampler_reg, brw_imm_ud(8));
+ brw_OR(p, addr, addr, surface_reg);
+ }
if (base_binding_table_index)
brw_ADD(p, addr, addr, brw_imm_ud(base_binding_table_index));
brw_AND(p, addr, addr, brw_imm_ud(0xfff));
case SHADER_OPCODE_TG4:
case SHADER_OPCODE_TG4_OFFSET:
case SHADER_OPCODE_SAMPLEINFO:
- generate_tex(inst, dst, src[0], src[1]);
+ generate_tex(inst, dst, src[0], src[1], src[2]);
break;
case FS_OPCODE_DDX_COARSE:
case FS_OPCODE_DDX_FINE:
compiler->shader_debug_log(log_data,
"%s SIMD%d shader: %d inst, %d loops, %u cycles, "
"%d:%d spills:fills, Promoted %u constants, "
- "compacted %d to %d bytes.\n",
- stage_abbrev, dispatch_width, before_size / 16,
+ "compacted %d to %d bytes.",
+ _mesa_shader_stage_to_abbrev(stage),
+ dispatch_width, before_size / 16,
loop_count, cfg->cycle_count, spill_count,
fill_count, promoted_constants, before_size,
after_size);
nir_emit_system_values();
/* get the main function and emit it */
- nir_foreach_overload(nir, overload) {
- assert(strcmp(overload->function->name, "main") == 0);
- assert(overload->impl);
- nir_emit_impl(overload->impl);
+ nir_foreach_function(nir, function) {
+ assert(strcmp(function->name, "main") == 0);
+ assert(function->impl);
+ nir_emit_impl(function->impl);
}
}
switch (stage) {
case MESA_SHADER_VERTEX:
+ case MESA_SHADER_TESS_EVAL:
case MESA_SHADER_GEOMETRY: {
unsigned location = var->data.location;
nir_setup_single_output_varying(®, var->type, &location);
*reg = *v->emit_vs_system_value(SYSTEM_VALUE_INSTANCE_ID);
break;
+ case nir_intrinsic_load_base_instance:
+ assert(v->stage == MESA_SHADER_VERTEX);
+ reg = &v->nir_system_values[SYSTEM_VALUE_BASE_INSTANCE];
+ if (reg->file == BAD_FILE)
+ *reg = *v->emit_vs_system_value(SYSTEM_VALUE_BASE_INSTANCE);
+ break;
+
+ case nir_intrinsic_load_draw_id:
+ assert(v->stage == MESA_SHADER_VERTEX);
+ reg = &v->nir_system_values[SYSTEM_VALUE_DRAW_ID];
+ if (reg->file == BAD_FILE)
+ *reg = *v->emit_vs_system_value(SYSTEM_VALUE_DRAW_ID);
+ break;
+
case nir_intrinsic_load_invocation_id:
assert(v->stage == MESA_SHADER_GEOMETRY);
reg = &v->nir_system_values[SYSTEM_VALUE_INVOCATION_ID];
nir_system_values[i] = fs_reg();
}
- nir_foreach_overload(nir, overload) {
- assert(strcmp(overload->function->name, "main") == 0);
- assert(overload->impl);
- nir_foreach_block(overload->impl, emit_system_values_block, this);
+ nir_foreach_function(nir, function) {
+ assert(strcmp(function->name, "main") == 0);
+ assert(function->impl);
+ nir_foreach_block(function->impl, emit_system_values_block, this);
}
}
case MESA_SHADER_VERTEX:
nir_emit_vs_intrinsic(abld, nir_instr_as_intrinsic(instr));
break;
+ case MESA_SHADER_TESS_EVAL:
+ nir_emit_tes_intrinsic(abld, nir_instr_as_intrinsic(instr));
+ break;
case MESA_SHADER_GEOMETRY:
nir_emit_gs_intrinsic(abld, nir_instr_as_intrinsic(instr));
break;
unreachable("Should have been lowered by borrow_to_arith().");
case nir_op_umod:
+ case nir_op_irem:
+ /* According to the sign table for INT DIV in the Ivy Bridge PRM, it
+ * appears that our hardware just does the right thing for signed
+ * remainder.
+ */
bld.emit(SHADER_OPCODE_INT_REMAINDER, result, op[0], op[1]);
break;
+ case nir_op_imod: {
+ /* Get a regular C-style remainder. If a % b == 0, set the predicate. */
+ bld.emit(SHADER_OPCODE_INT_REMAINDER, result, op[0], op[1]);
+
+ /* Math instructions don't support conditional mod */
+ inst = bld.MOV(bld.null_reg_d(), result);
+ inst->conditional_mod = BRW_CONDITIONAL_NZ;
+
+ /* Now, we need to determine if signs of the sources are different.
+ * When we XOR the sources, the top bit is 0 if they are the same and 1
+ * if they are different. We can then use a conditional modifier to
+ * turn that into a predicate. This leads us to an XOR.l instruction.
+ */
+ fs_reg tmp = bld.vgrf(BRW_REGISTER_TYPE_D);
+ inst = bld.XOR(tmp, op[0], op[1]);
+ inst->predicate = BRW_PREDICATE_NORMAL;
+ inst->conditional_mod = BRW_CONDITIONAL_L;
+
+ /* If the result of the initial remainder operation is non-zero and the
+ * two sources have different signs, add in a copy of op[1] to get the
+ * final integer modulus value.
+ */
+ inst = bld.ADD(result, result, op[1]);
+ inst->predicate = BRW_PREDICATE_NORMAL;
+ break;
+ }
+
case nir_op_flt:
case nir_op_ilt:
case nir_op_ult:
case nir_op_fdot2:
case nir_op_fdot3:
case nir_op_fdot4:
- case nir_op_bany2:
- case nir_op_bany3:
- case nir_op_bany4:
- case nir_op_ball2:
- case nir_op_ball3:
- case nir_op_ball4:
case nir_op_ball_fequal2:
case nir_op_ball_iequal2:
case nir_op_ball_fequal3:
inst->saturate = instr->dest.saturate;
break;
+ case nir_op_fquantize2f16: {
+ fs_reg tmp = bld.vgrf(BRW_REGISTER_TYPE_D);
+
+ /* The destination stride must be at least as big as the source stride. */
+ tmp.type = BRW_REGISTER_TYPE_W;
+ tmp.stride = 2;
+
+ bld.emit(BRW_OPCODE_F32TO16, tmp, op[0]);
+ inst = bld.emit(BRW_OPCODE_F16TO32, result, tmp);
+ inst->saturate = instr->dest.saturate;
+ break;
+ }
+
case nir_op_fmin:
case nir_op_imin:
case nir_op_umin:
case nir_op_ubitfield_extract:
case nir_op_ibitfield_extract:
+ unreachable("should have been lowered");
+ case nir_op_ubfe:
+ case nir_op_ibfe:
bld.BFE(result, op[2], op[1], op[0]);
break;
case nir_op_bfm:
break;
case nir_op_bitfield_insert:
- unreachable("not reached: should be handled by "
- "lower_instructions::bitfield_insert_to_bfm_bfi");
+ unreachable("not reached: should have been lowered");
case nir_op_ishl:
bld.SHL(result, op[0], op[1]);
}
}
+fs_reg
+fs_visitor::get_indirect_offset(nir_intrinsic_instr *instr)
+{
+ nir_src *offset_src = nir_get_io_offset_src(instr);
+ nir_const_value *const_value = nir_src_as_const_value(*offset_src);
+
+ if (const_value) {
+ /* The only constant offset we should find is 0. brw_nir.c's
+ * add_const_offset_to_base() will fold other constant offsets
+ * into instr->const_index[0].
+ */
+ assert(const_value->u[0] == 0);
+ return fs_reg();
+ }
+
+ return get_nir_src(*offset_src);
+}
+
void
fs_visitor::nir_emit_vs_intrinsic(const fs_builder &bld,
nir_intrinsic_instr *instr)
case nir_intrinsic_load_vertex_id_zero_base:
case nir_intrinsic_load_base_vertex:
- case nir_intrinsic_load_instance_id: {
+ case nir_intrinsic_load_instance_id:
+ case nir_intrinsic_load_base_instance:
+ case nir_intrinsic_load_draw_id: {
gl_system_value sv = nir_system_value_from_intrinsic(instr->intrinsic);
fs_reg val = nir_system_values[sv];
assert(val.file != BAD_FILE);
}
void
+fs_visitor::nir_emit_tes_intrinsic(const fs_builder &bld,
+ nir_intrinsic_instr *instr)
+{
+ assert(stage == MESA_SHADER_TESS_EVAL);
+ struct brw_tes_prog_data *tes_prog_data = (struct brw_tes_prog_data *) prog_data;
+
+ fs_reg dest;
+ if (nir_intrinsic_infos[instr->intrinsic].has_dest)
+ dest = get_nir_dest(instr->dest);
+
+ switch (instr->intrinsic) {
+ case nir_intrinsic_load_primitive_id:
+ bld.MOV(dest, fs_reg(brw_vec1_grf(0, 1)));
+ break;
+ case nir_intrinsic_load_tess_coord:
+ /* gl_TessCoord is part of the payload in g1-3 */
+ for (unsigned i = 0; i < 3; i++) {
+ bld.MOV(offset(dest, bld, i), fs_reg(brw_vec8_grf(1 + i, 0)));
+ }
+ break;
+
+ case nir_intrinsic_load_tess_level_outer:
+ /* When the TES reads gl_TessLevelOuter, we ensure that the patch header
+ * appears as a push-model input. So, we can simply use the ATTR file
+ * rather than issuing URB read messages. The data is stored in the
+ * high DWords in reverse order - DWord 7 contains .x, DWord 6 contains
+ * .y, and so on.
+ */
+ switch (tes_prog_data->domain) {
+ case BRW_TESS_DOMAIN_QUAD:
+ for (unsigned i = 0; i < 4; i++)
+ bld.MOV(offset(dest, bld, i), component(fs_reg(ATTR, 0), 7 - i));
+ break;
+ case BRW_TESS_DOMAIN_TRI:
+ for (unsigned i = 0; i < 3; i++)
+ bld.MOV(offset(dest, bld, i), component(fs_reg(ATTR, 0), 7 - i));
+ break;
+ case BRW_TESS_DOMAIN_ISOLINE:
+ for (unsigned i = 0; i < 2; i++)
+ bld.MOV(offset(dest, bld, i), component(fs_reg(ATTR, 0), 7 - i));
+ break;
+ }
+ break;
+
+ case nir_intrinsic_load_tess_level_inner:
+ /* When the TES reads gl_TessLevelInner, we ensure that the patch header
+ * appears as a push-model input. So, we can simply use the ATTR file
+ * rather than issuing URB read messages.
+ */
+ switch (tes_prog_data->domain) {
+ case BRW_TESS_DOMAIN_QUAD:
+ bld.MOV(dest, component(fs_reg(ATTR, 0), 3));
+ bld.MOV(offset(dest, bld, 1), component(fs_reg(ATTR, 0), 2));
+ break;
+ case BRW_TESS_DOMAIN_TRI:
+ bld.MOV(dest, component(fs_reg(ATTR, 0), 4));
+ break;
+ case BRW_TESS_DOMAIN_ISOLINE:
+ /* ignore - value is undefined */
+ break;
+ }
+ break;
+
+ case nir_intrinsic_load_input:
+ case nir_intrinsic_load_per_vertex_input: {
+ fs_reg indirect_offset = get_indirect_offset(instr);
+ unsigned imm_offset = instr->const_index[0];
+
+ fs_inst *inst;
+ if (indirect_offset.file == BAD_FILE) {
+ /* Arbitrarily only push up to 32 vec4 slots worth of data,
+ * which is 16 registers (since each holds 2 vec4 slots).
+ */
+ const unsigned max_push_slots = 32;
+ if (imm_offset < max_push_slots) {
+ fs_reg src = fs_reg(ATTR, imm_offset / 2, dest.type);
+ for (int i = 0; i < instr->num_components; i++) {
+ bld.MOV(offset(dest, bld, i),
+ component(src, 4 * (imm_offset % 2) + i));
+ }
+ tes_prog_data->base.urb_read_length =
+ MAX2(tes_prog_data->base.urb_read_length,
+ DIV_ROUND_UP(imm_offset + 1, 2));
+ } else {
+ /* Replicate the patch handle to all enabled channels */
+ const fs_reg srcs[] = {
+ retype(brw_vec1_grf(0, 0), BRW_REGISTER_TYPE_UD)
+ };
+ fs_reg patch_handle = bld.vgrf(BRW_REGISTER_TYPE_UD, 1);
+ bld.LOAD_PAYLOAD(patch_handle, srcs, ARRAY_SIZE(srcs), 0);
+
+ inst = bld.emit(SHADER_OPCODE_URB_READ_SIMD8, dest, patch_handle);
+ inst->mlen = 1;
+ inst->offset = imm_offset;
+ inst->base_mrf = -1;
+ inst->regs_written = instr->num_components;
+ }
+ } else {
+ /* Indirect indexing - use per-slot offsets as well. */
+ const fs_reg srcs[] = {
+ retype(brw_vec1_grf(0, 0), BRW_REGISTER_TYPE_UD),
+ indirect_offset
+ };
+ fs_reg payload = bld.vgrf(BRW_REGISTER_TYPE_UD, 2);
+ bld.LOAD_PAYLOAD(payload, srcs, ARRAY_SIZE(srcs), 0);
+
+ inst = bld.emit(SHADER_OPCODE_URB_READ_SIMD8_PER_SLOT, dest, payload);
+ inst->mlen = 2;
+ inst->offset = imm_offset;
+ inst->base_mrf = -1;
+ inst->regs_written = instr->num_components;
+ }
+ break;
+ }
+ default:
+ nir_emit_intrinsic(bld, instr);
+ break;
+ }
+}
+
+void
fs_visitor::nir_emit_gs_intrinsic(const fs_builder &bld,
nir_intrinsic_instr *instr)
{
nir_emit_shared_atomic(bld, BRW_AOP_CMPWR, instr);
break;
+ case nir_intrinsic_load_shared: {
+ assert(devinfo->gen >= 7);
+
+ fs_reg surf_index = brw_imm_ud(GEN7_BTI_SLM);
+
+ /* Get the offset to read from */
+ fs_reg offset_reg;
+ nir_const_value *const_offset = nir_src_as_const_value(instr->src[0]);
+ if (const_offset) {
+ offset_reg = brw_imm_ud(instr->const_index[0] + const_offset->u[0]);
+ } else {
+ offset_reg = vgrf(glsl_type::uint_type);
+ bld.ADD(offset_reg,
+ retype(get_nir_src(instr->src[0]), BRW_REGISTER_TYPE_UD),
+ brw_imm_ud(instr->const_index[0]));
+ }
+
+ /* Read the vector */
+ fs_reg read_result = emit_untyped_read(bld, surf_index, offset_reg,
+ 1 /* dims */,
+ instr->num_components,
+ BRW_PREDICATE_NONE);
+ read_result.type = dest.type;
+ for (int i = 0; i < instr->num_components; i++)
+ bld.MOV(offset(dest, bld, i), offset(read_result, bld, i));
+
+ break;
+ }
+
+ case nir_intrinsic_store_shared: {
+ assert(devinfo->gen >= 7);
+
+ /* Block index */
+ fs_reg surf_index = brw_imm_ud(GEN7_BTI_SLM);
+
+ /* Value */
+ fs_reg val_reg = get_nir_src(instr->src[0]);
+
+ /* Writemask */
+ unsigned writemask = instr->const_index[1];
+
+ /* Combine groups of consecutive enabled channels in one write
+ * message. We use ffs to find the first enabled channel and then ffs on
+ * the bit-inverse, down-shifted writemask to determine the length of
+ * the block of enabled bits.
+ */
+ while (writemask) {
+ unsigned first_component = ffs(writemask) - 1;
+ unsigned length = ffs(~(writemask >> first_component)) - 1;
+ fs_reg offset_reg;
+
+ nir_const_value *const_offset = nir_src_as_const_value(instr->src[1]);
+ if (const_offset) {
+ offset_reg = brw_imm_ud(instr->const_index[0] + const_offset->u[0] +
+ 4 * first_component);
+ } else {
+ offset_reg = vgrf(glsl_type::uint_type);
+ bld.ADD(offset_reg,
+ retype(get_nir_src(instr->src[1]), BRW_REGISTER_TYPE_UD),
+ brw_imm_ud(instr->const_index[0] + 4 * first_component));
+ }
+
+ emit_untyped_write(bld, surf_index, offset_reg,
+ offset(val_reg, bld, first_component),
+ 1 /* dims */, length,
+ BRW_PREDICATE_NONE);
+
+ /* Clear the bits in the writemask that we just wrote, then try
+ * again to see if more channels are left.
+ */
+ writemask &= (15 << (first_component + length));
+ }
+
+ break;
+ }
+
default:
nir_emit_intrinsic(bld, instr);
break;
break;
}
- case nir_intrinsic_load_shared: {
- assert(devinfo->gen >= 7);
-
- fs_reg surf_index = brw_imm_ud(GEN7_BTI_SLM);
-
- /* Get the offset to read from */
- fs_reg offset_reg;
- nir_const_value *const_offset = nir_src_as_const_value(instr->src[0]);
- if (const_offset) {
- offset_reg = brw_imm_ud(instr->const_index[0] + const_offset->u[0]);
- } else {
- offset_reg = vgrf(glsl_type::uint_type);
- bld.ADD(offset_reg,
- retype(get_nir_src(instr->src[0]), BRW_REGISTER_TYPE_UD),
- brw_imm_ud(instr->const_index[0]));
- }
-
- /* Read the vector */
- fs_reg read_result = emit_untyped_read(bld, surf_index, offset_reg,
- 1 /* dims */,
- instr->num_components,
- BRW_PREDICATE_NONE);
- read_result.type = dest.type;
- for (int i = 0; i < instr->num_components; i++)
- bld.MOV(offset(dest, bld, i), offset(read_result, bld, i));
-
- break;
- }
-
- case nir_intrinsic_store_shared: {
- assert(devinfo->gen >= 7);
-
- /* Block index */
- fs_reg surf_index = brw_imm_ud(GEN7_BTI_SLM);
-
- /* Value */
- fs_reg val_reg = get_nir_src(instr->src[0]);
-
- /* Writemask */
- unsigned writemask = instr->const_index[1];
-
- /* Combine groups of consecutive enabled channels in one write
- * message. We use ffs to find the first enabled channel and then ffs on
- * the bit-inverse, down-shifted writemask to determine the length of
- * the block of enabled bits.
- */
- while (writemask) {
- unsigned first_component = ffs(writemask) - 1;
- unsigned length = ffs(~(writemask >> first_component)) - 1;
- fs_reg offset_reg;
-
- nir_const_value *const_offset = nir_src_as_const_value(instr->src[1]);
- if (const_offset) {
- offset_reg = brw_imm_ud(instr->const_index[0] + const_offset->u[0] +
- 4 * first_component);
- } else {
- offset_reg = vgrf(glsl_type::uint_type);
- bld.ADD(offset_reg,
- retype(get_nir_src(instr->src[1]), BRW_REGISTER_TYPE_UD),
- brw_imm_ud(instr->const_index[0] + 4 * first_component));
- }
-
- emit_untyped_write(bld, surf_index, offset_reg,
- offset(val_reg, bld, first_component),
- 1 /* dims */, length,
- BRW_PREDICATE_NONE);
-
- /* Clear the bits in the writemask that we just wrote, then try
- * again to see if more channels are left.
- */
- writemask &= (15 << (first_component + length));
- }
-
- break;
- }
-
case nir_intrinsic_load_input: {
fs_reg src;
if (stage == MESA_SHADER_VERTEX) {
void
fs_visitor::nir_emit_texture(const fs_builder &bld, nir_tex_instr *instr)
{
+ unsigned texture = instr->texture_index;
unsigned sampler = instr->sampler_index;
+ fs_reg texture_reg(brw_imm_ud(texture));
fs_reg sampler_reg(brw_imm_ud(sampler));
int gather_component = instr->component;
fs_reg coordinate, shadow_comparitor, lod, lod2, sample_index, mcs, tex_offset;
+ /* Our hardware requires a LOD for buffer textures */
+ if (instr->sampler_dim == GLSL_SAMPLER_DIM_BUF)
+ lod = brw_imm_d(0);
+
for (unsigned i = 0; i < instr->num_srcs; i++) {
fs_reg src = get_nir_src(instr->src[i].src);
switch (instr->src[i].src_type) {
case nir_tex_src_projector:
unreachable("should be lowered");
- case nir_tex_src_sampler_offset: {
- /* Figure out the highest possible sampler index and mark it as used */
- uint32_t max_used = sampler + instr->sampler_array_size - 1;
+ case nir_tex_src_texture_offset: {
+ /* Figure out the highest possible texture index and mark it as used */
+ uint32_t max_used = texture + instr->texture_array_size - 1;
if (instr->op == nir_texop_tg4 && devinfo->gen < 8) {
max_used += stage_prog_data->binding_table.gather_texture_start;
} else {
brw_mark_surface_used(prog_data, max_used);
/* Emit code to evaluate the actual indexing expression */
+ texture_reg = vgrf(glsl_type::uint_type);
+ bld.ADD(texture_reg, src, brw_imm_ud(texture));
+ texture_reg = bld.emit_uniformize(texture_reg);
+ break;
+ }
+
+ case nir_tex_src_sampler_offset: {
+ /* Emit code to evaluate the actual indexing expression */
sampler_reg = vgrf(glsl_type::uint_type);
bld.ADD(sampler_reg, src, brw_imm_ud(sampler));
sampler_reg = bld.emit_uniformize(sampler_reg);
if (instr->op == nir_texop_txf_ms ||
instr->op == nir_texop_samples_identical) {
if (devinfo->gen >= 7 &&
- key_tex->compressed_multisample_layout_mask & (1 << sampler)) {
- mcs = emit_mcs_fetch(coordinate, instr->coord_components, sampler_reg);
+ key_tex->compressed_multisample_layout_mask & (1 << texture)) {
+ mcs = emit_mcs_fetch(coordinate, instr->coord_components, texture_reg);
} else {
mcs = brw_imm_ud(0u);
}
fs_reg dst = retype(get_nir_dest(instr->dest), BRW_REGISTER_TYPE_D);
fs_inst *inst = bld.emit(SHADER_OPCODE_SAMPLEINFO, dst,
bld.vgrf(BRW_REGISTER_TYPE_D, 1),
- sampler_reg);
+ texture_reg, texture_reg);
inst->mlen = 1;
inst->header_size = 1;
inst->base_mrf = -1;
emit_texture(op, dest_type, coordinate, instr->coord_components,
shadow_comparitor, lod, lod2, lod_components, sample_index,
tex_offset, mcs, gather_component,
- is_cube_array, sampler, sampler_reg);
+ is_cube_array, texture, texture_reg, sampler, sampler_reg);
fs_reg dest = get_nir_dest(instr->dest);
dest.type = this->result.type;
switch (location) {
case SYSTEM_VALUE_BASE_VERTEX:
reg->reg_offset = 0;
- vs_prog_data->uses_vertexid = true;
+ vs_prog_data->uses_basevertex = true;
+ break;
+ case SYSTEM_VALUE_BASE_INSTANCE:
+ reg->reg_offset = 1;
+ vs_prog_data->uses_baseinstance = true;
break;
case SYSTEM_VALUE_VERTEX_ID:
+ unreachable("should have been lowered");
case SYSTEM_VALUE_VERTEX_ID_ZERO_BASE:
reg->reg_offset = 2;
vs_prog_data->uses_vertexid = true;
reg->reg_offset = 3;
vs_prog_data->uses_instanceid = true;
break;
+ case SYSTEM_VALUE_DRAW_ID:
+ if (nir->info.system_values_read &
+ (BITFIELD64_BIT(SYSTEM_VALUE_BASE_VERTEX) |
+ BITFIELD64_BIT(SYSTEM_VALUE_BASE_INSTANCE) |
+ BITFIELD64_BIT(SYSTEM_VALUE_VERTEX_ID_ZERO_BASE) |
+ BITFIELD64_BIT(SYSTEM_VALUE_INSTANCE_ID)))
+ reg->nr += 4;
+ reg->reg_offset = 0;
+ vs_prog_data->uses_drawid = true;
+ break;
default:
unreachable("not reached");
}
/* Sample from the MCS surface attached to this multisample texture. */
fs_reg
fs_visitor::emit_mcs_fetch(const fs_reg &coordinate, unsigned components,
- const fs_reg &sampler)
+ const fs_reg &texture)
{
const fs_reg dest = vgrf(glsl_type::uvec4_type);
const fs_reg srcs[] = {
coordinate, fs_reg(), fs_reg(), fs_reg(), fs_reg(), fs_reg(),
- sampler, fs_reg(), brw_imm_ud(components), brw_imm_d(0)
+ texture, texture, fs_reg(), brw_imm_ud(components), brw_imm_d(0)
};
fs_inst *inst = bld.emit(SHADER_OPCODE_TXF_MCS_LOGICAL, dest, srcs,
ARRAY_SIZE(srcs));
fs_reg mcs,
int gather_component,
bool is_cube_array,
+ uint32_t surface,
+ fs_reg surface_reg,
uint32_t sampler,
fs_reg sampler_reg)
{
fs_reg dst = vgrf(glsl_type::get_instance(dest_type->base_type, 4, 1));
const fs_reg srcs[] = {
coordinate, shadow_c, lod, lod2,
- sample_index, mcs, sampler_reg, offset_value,
+ sample_index, mcs, surface_reg, sampler_reg, offset_value,
brw_imm_d(coord_components), brw_imm_d(grad_components)
};
enum opcode opcode;
if (op == ir_tg4) {
if (gather_component == 1 &&
- key_tex->gather_channel_quirk_mask & (1 << sampler)) {
+ key_tex->gather_channel_quirk_mask & (1 << surface)) {
/* gather4 sampler is broken for green channel on RG32F --
* we must ask for blue instead.
*/
}
if (devinfo->gen == 6)
- emit_gen6_gather_wa(key_tex->gen6_gather_wa[sampler], dst);
+ emit_gen6_gather_wa(key_tex->gen6_gather_wa[surface], dst);
}
/* fixup #layers for cube map arrays */
fs_reg sources[8];
fs_reg urb_handle;
- urb_handle = fs_reg(retype(brw_vec8_grf(1, 0), BRW_REGISTER_TYPE_UD));
+ if (stage == MESA_SHADER_TESS_EVAL)
+ urb_handle = fs_reg(retype(brw_vec8_grf(4, 0), BRW_REGISTER_TYPE_UD));
+ else
+ urb_handle = fs_reg(retype(brw_vec8_grf(1, 0), BRW_REGISTER_TYPE_UD));
/* If we don't have any valid slots to write, just do a minimal urb write
* send to terminate the shader. This includes 1 slot of undefined data,
struct gl_program *prog,
const nir_shader *shader,
unsigned dispatch_width,
- int shader_time_index)
+ int shader_time_index,
+ const struct brw_vue_map *input_vue_map)
: backend_shader(compiler, log_data, mem_ctx, shader, prog_data),
key(key), gs_compile(NULL), prog_data(prog_data), prog(prog),
+ input_vue_map(input_vue_map),
dispatch_width(dispatch_width),
shader_time_index(shader_time_index),
bld(fs_builder(this, dispatch_width).at_end())
case MESA_SHADER_VERTEX:
key_tex = &((const brw_vs_prog_key *) key)->tex;
break;
+ case MESA_SHADER_TESS_EVAL:
+ key_tex = &((const brw_tes_prog_key *) key)->tex;
+ break;
case MESA_SHADER_GEOMETRY:
key_tex = &((const brw_gs_prog_key *) key)->tex;
break;
#include "brw_nir.h"
#include "brw_program.h"
#include "glsl/ir_optimization.h"
-#include "glsl/glsl_parser_extras.h"
#include "program/program.h"
#include "main/shaderapi.h"
#include "main/uniforms.h"
struct gl_shader_program *sh_prog)
{
struct gl_shader *vs = sh_prog->_LinkedShaders[MESA_SHADER_VERTEX];
+ struct gl_shader *tcs = sh_prog->_LinkedShaders[MESA_SHADER_TESS_CTRL];
+ struct gl_shader *tes = sh_prog->_LinkedShaders[MESA_SHADER_TESS_EVAL];
struct gl_shader *gs = sh_prog->_LinkedShaders[MESA_SHADER_GEOMETRY];
struct gl_shader *fs = sh_prog->_LinkedShaders[MESA_SHADER_FRAGMENT];
struct gl_shader *cs = sh_prog->_LinkedShaders[MESA_SHADER_COMPUTE];
if (gs && !brw_gs_precompile(ctx, sh_prog, gs->Program))
return false;
+ if (tes && !brw_tes_precompile(ctx, sh_prog, tes->Program))
+ return false;
+
+ if (tcs && !brw_tcs_precompile(ctx, sh_prog, tcs->Program))
+ return false;
+
if (vs && !brw_vs_precompile(ctx, sh_prog, vs->Program))
return false;
*/
brw_lower_packing_builtins(brw, shader->Stage, shader->ir);
do_mat_op_to_vec(shader->ir);
- const int bitfield_insert = brw->gen >= 7 ? BITFIELD_INSERT_TO_BFM_BFI : 0;
lower_instructions(shader->ir,
MOD_TO_FLOOR |
DIV_TO_MUL_RCP |
SUB_TO_ADD_NEG |
EXP_TO_EXP2 |
LOG_TO_LOG2 |
- bitfield_insert |
LDEXP_TO_ARITH |
CARRY_TO_ARITH |
BORROW_TO_ARITH);
/* 2. quotient rule */
ir_variable *recip = temp(mem_ctx, glsl_type::float_type, "recip");
- EMIT(assign(recip, div(new(mem_ctx) ir_constant(1.0f), swizzle_z(Q))));
+ EMIT(assign(recip, expr(ir_unop_rcp, swizzle_z(Q))));
ir_variable *dx = temp(mem_ctx, glsl_type::vec2_type, "dx");
ir_variable *dy = temp(mem_ctx, glsl_type::vec2_type, "dy");
blit->baseLevelSave = tex_obj->BaseLevel;
blit->maxLevelSave = tex_obj->MaxLevel;
blit->stencilSamplingSave = tex_obj->StencilSampling;
- blit->sampler = _mesa_meta_setup_sampler(ctx, tex_obj, *target,
- GL_NEAREST, level);
+ blit->samp_obj = _mesa_meta_setup_sampler(ctx, tex_obj, *target,
+ GL_NEAREST, level);
return true;
}
#include "glsl/nir/nir_builder.h"
#include "program/prog_to_nir.h"
-struct remap_vs_attrs_state {
+static bool
+is_input(nir_intrinsic_instr *intrin)
+{
+ return intrin->intrinsic == nir_intrinsic_load_input ||
+ intrin->intrinsic == nir_intrinsic_load_per_vertex_input;
+}
+
+static bool
+is_output(nir_intrinsic_instr *intrin)
+{
+ return intrin->intrinsic == nir_intrinsic_load_output ||
+ intrin->intrinsic == nir_intrinsic_load_per_vertex_output ||
+ intrin->intrinsic == nir_intrinsic_store_output ||
+ intrin->intrinsic == nir_intrinsic_store_per_vertex_output;
+}
+
+/**
+ * In many cases, we just add the base and offset together, so there's no
+ * reason to keep them separate. Sometimes, combining them is essential:
+ * if a shader only accesses part of a compound variable (such as a matrix
+ * or array), the variable's base may not actually exist in the VUE map.
+ *
+ * This pass adds constant offsets to instr->const_index[0], and resets
+ * the offset source to 0. Non-constant offsets remain unchanged - since
+ * we don't know what part of a compound variable is accessed, we allocate
+ * storage for the entire thing.
+ */
+struct add_const_offset_to_base_params {
nir_builder b;
- uint64_t inputs_read;
+ nir_variable_mode mode;
};
static bool
-remap_vs_attrs(nir_block *block, void *void_state)
+add_const_offset_to_base(nir_block *block, void *closure)
{
- struct remap_vs_attrs_state *state = void_state;
+ struct add_const_offset_to_base_params *params = closure;
+ nir_builder *b = ¶ms->b;
nir_foreach_instr_safe(block, instr) {
if (instr->type != nir_instr_type_intrinsic)
nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);
+ if ((params->mode == nir_var_shader_in && is_input(intrin)) ||
+ (params->mode == nir_var_shader_out && is_output(intrin))) {
+ nir_src *offset = nir_get_io_offset_src(intrin);
+ nir_const_value *const_offset = nir_src_as_const_value(*offset);
+
+ if (const_offset) {
+ intrin->const_index[0] += const_offset->u[0];
+ b->cursor = nir_before_instr(&intrin->instr);
+ nir_instr_rewrite_src(&intrin->instr, offset,
+ nir_src_for_ssa(nir_imm_int(b, 0)));
+ }
+ }
+ }
+ return true;
+
+}
+
+static bool
+remap_vs_attrs(nir_block *block, void *closure)
+{
+ GLbitfield64 inputs_read = *((GLbitfield64 *) closure);
+
+ nir_foreach_instr(block, instr) {
+ if (instr->type != nir_instr_type_intrinsic)
+ continue;
+
+ nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);
+
if (intrin->intrinsic == nir_intrinsic_load_input) {
/* Attributes come in a contiguous block, ordered by their
* gl_vert_attrib value. That means we can compute the slot
* number for an attribute by masking out the enabled attributes
* before it and counting the bits.
*/
- nir_const_value *const_offset = nir_src_as_const_value(intrin->src[0]);
+ int attr = intrin->const_index[0];
+ int slot = _mesa_bitcount_64(inputs_read & BITFIELD64_MASK(attr));
- /* We set EmitNoIndirect for VS inputs, so there are no indirects. */
- assert(const_offset);
+ intrin->const_index[0] = 4 * slot;
+ }
+ }
+ return true;
+}
- int attr = intrin->const_index[0] + const_offset->u[0];
- int slot = _mesa_bitcount_64(state->inputs_read &
- BITFIELD64_MASK(attr));
+static bool
+remap_inputs_with_vue_map(nir_block *block, void *closure)
+{
+ const struct brw_vue_map *vue_map = closure;
- /* The NIR -> FS pass will just add the base and offset together, so
- * there's no reason to keep them separate. Just put it all in
- * const_index[0] and set the offset src[0] to load_const(0).
- */
- intrin->const_index[0] = 4 * slot;
+ nir_foreach_instr(block, instr) {
+ if (instr->type != nir_instr_type_intrinsic)
+ continue;
- state->b.cursor = nir_before_instr(&intrin->instr);
- nir_instr_rewrite_src(&intrin->instr, &intrin->src[0],
- nir_src_for_ssa(nir_imm_int(&state->b, 0)));
+ nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);
+
+ if (intrin->intrinsic == nir_intrinsic_load_input ||
+ intrin->intrinsic == nir_intrinsic_load_per_vertex_input) {
+ int vue_slot = vue_map->varying_to_slot[intrin->const_index[0]];
+ assert(vue_slot != -1);
+ intrin->const_index[0] = vue_slot;
+ }
+ }
+ return true;
+}
+
+struct remap_patch_urb_offsets_state {
+ nir_builder b;
+ struct brw_vue_map vue_map;
+};
+
+static bool
+remap_patch_urb_offsets(nir_block *block, void *closure)
+{
+ struct remap_patch_urb_offsets_state *state = closure;
+
+ nir_foreach_instr_safe(block, instr) {
+ if (instr->type != nir_instr_type_intrinsic)
+ continue;
+
+ nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);
+
+ gl_shader_stage stage = state->b.shader->stage;
+
+ if ((stage == MESA_SHADER_TESS_CTRL && is_output(intrin)) ||
+ (stage == MESA_SHADER_TESS_EVAL && is_input(intrin))) {
+ int vue_slot = state->vue_map.varying_to_slot[intrin->const_index[0]];
+ assert(vue_slot != -1);
+ intrin->const_index[0] = vue_slot;
+
+ nir_src *vertex = nir_get_io_vertex_index_src(intrin);
+ if (vertex) {
+ nir_const_value *const_vertex = nir_src_as_const_value(*vertex);
+ if (const_vertex) {
+ intrin->const_index[0] += const_vertex->u[0] *
+ state->vue_map.num_per_vertex_slots;
+ } else {
+ state->b.cursor = nir_before_instr(&intrin->instr);
+
+ /* Multiply by the number of per-vertex slots. */
+ nir_ssa_def *vertex_offset =
+ nir_imul(&state->b,
+ nir_ssa_for_src(&state->b, *vertex, 1),
+ nir_imm_int(&state->b,
+ state->vue_map.num_per_vertex_slots));
+
+ /* Add it to the existing offset */
+ nir_src *offset = nir_get_io_offset_src(intrin);
+ nir_ssa_def *total_offset =
+ nir_iadd(&state->b, vertex_offset,
+ nir_ssa_for_src(&state->b, *offset, 1));
+
+ nir_instr_rewrite_src(&intrin->instr, offset,
+ nir_src_for_ssa(total_offset));
+ }
+ }
}
}
return true;
const struct brw_device_info *devinfo,
bool is_scalar)
{
+ struct add_const_offset_to_base_params params = {
+ .mode = nir_var_shader_in
+ };
+
switch (nir->stage) {
case MESA_SHADER_VERTEX:
/* Start with the location of the variable's base. */
* key->inputs_read since the two are identical aside from Gen4-5
* edge flag differences.
*/
- struct remap_vs_attrs_state remap_state = {
- .inputs_read = nir->info.inputs_read,
- };
+ GLbitfield64 inputs_read = nir->info.inputs_read;
/* This pass needs actual constants */
nir_opt_constant_folding(nir);
- nir_foreach_overload(nir, overload) {
- if (overload->impl) {
- nir_builder_init(&remap_state.b, overload->impl);
- nir_foreach_block(overload->impl, remap_vs_attrs, &remap_state);
+ nir_foreach_function(nir, function) {
+ if (function->impl) {
+ nir_builder_init(¶ms.b, function->impl);
+ nir_foreach_block(function->impl, add_const_offset_to_base, ¶ms);
+ nir_foreach_block(function->impl, remap_vs_attrs, &inputs_read);
}
}
}
break;
+ case MESA_SHADER_TESS_CTRL:
case MESA_SHADER_GEOMETRY: {
- if (!is_scalar) {
+ if (!is_scalar && nir->stage == MESA_SHADER_GEOMETRY) {
foreach_list_typed(nir_variable, var, node, &nir->inputs) {
var->data.driver_location = var->data.location;
}
GLbitfield64 inputs_read =
nir->info.inputs_read & ~VARYING_BIT_PRIMITIVE_ID;
brw_compute_vue_map(devinfo, &input_vue_map, inputs_read,
- nir->info.separate_shader);
+ nir->info.separate_shader ||
+ nir->stage == MESA_SHADER_TESS_CTRL);
- /* Start with the slot for the variable's base. */
foreach_list_typed(nir_variable, var, node, &nir->inputs) {
- assert(input_vue_map.varying_to_slot[var->data.location] != -1);
- var->data.driver_location =
- input_vue_map.varying_to_slot[var->data.location];
+ var->data.driver_location = var->data.location;
}
/* Inputs are stored in vec4 slots, so use type_size_vec4(). */
nir_lower_io(nir, nir_var_shader_in, type_size_vec4);
+
+ /* This pass needs actual constants */
+ nir_opt_constant_folding(nir);
+
+ nir_foreach_function(nir, function) {
+ if (function->impl) {
+ nir_builder_init(¶ms.b, function->impl);
+ nir_foreach_block(function->impl, add_const_offset_to_base, ¶ms);
+ nir_foreach_block(function->impl, remap_inputs_with_vue_map,
+ &input_vue_map);
+ }
+ }
+ }
+ break;
+ }
+ case MESA_SHADER_TESS_EVAL: {
+ struct remap_patch_urb_offsets_state state;
+ brw_compute_tess_vue_map(&state.vue_map,
+ nir->info.inputs_read & ~VARYING_BIT_PRIMITIVE_ID,
+ nir->info.patch_inputs_read);
+
+ foreach_list_typed(nir_variable, var, node, &nir->inputs) {
+ var->data.driver_location = var->data.location;
+ }
+
+ nir_lower_io(nir, nir_var_shader_in, type_size_vec4);
+
+ /* This pass needs actual constants */
+ nir_opt_constant_folding(nir);
+
+ nir_foreach_function(nir, function) {
+ if (function->impl) {
+ nir_builder_init(¶ms.b, function->impl);
+ nir_foreach_block(function->impl, add_const_offset_to_base, ¶ms);
+ nir_builder_init(&state.b, function->impl);
+ nir_foreach_block(function->impl, remap_patch_urb_offsets, &state);
+ }
}
break;
}
}
static void
-brw_nir_lower_outputs(nir_shader *nir, bool is_scalar)
+brw_nir_lower_outputs(nir_shader *nir,
+ const struct brw_device_info *devinfo,
+ bool is_scalar)
{
switch (nir->stage) {
case MESA_SHADER_VERTEX:
+ case MESA_SHADER_TESS_EVAL:
case MESA_SHADER_GEOMETRY:
if (is_scalar) {
nir_assign_var_locations(&nir->outputs, &nir->num_outputs,
var->data.driver_location = var->data.location;
}
break;
+ case MESA_SHADER_TESS_CTRL: {
+ struct add_const_offset_to_base_params params = {
+ .mode = nir_var_shader_out
+ };
+
+ struct remap_patch_urb_offsets_state state;
+ brw_compute_tess_vue_map(&state.vue_map, nir->info.outputs_written,
+ nir->info.patch_outputs_written);
+
+ nir_foreach_variable(var, &nir->outputs) {
+ var->data.driver_location = var->data.location;
+ }
+
+ nir_lower_io(nir, nir_var_shader_out, type_size_vec4);
+
+ /* This pass needs actual constants */
+ nir_opt_constant_folding(nir);
+
+ nir_foreach_function(nir, function) {
+ if (function->impl) {
+ nir_builder_init(¶ms.b, function->impl);
+ nir_foreach_block(function->impl, add_const_offset_to_base, ¶ms);
+ nir_builder_init(&state.b, function->impl);
+ nir_foreach_block(function->impl, remap_patch_urb_offsets, &state);
+ }
+ }
+ break;
+ }
case MESA_SHADER_FRAGMENT:
nir_assign_var_locations(&nir->outputs, &nir->num_outputs,
type_size_scalar);
}
}
-#include "util/debug.h"
-
-static bool
-should_clone_nir()
+static void
+brw_nir_lower_shared(nir_shader *nir)
{
- static int should_clone = -1;
- if (should_clone < 1)
- should_clone = env_var_as_boolean("NIR_TEST_CLONE", false);
-
- return should_clone;
+ nir_assign_var_locations(&nir->shared, &nir->num_shared,
+ type_size_scalar_bytes);
+ nir_lower_io(nir, nir_var_shared, type_size_scalar_bytes);
}
-#define _OPT(do_pass) (({ \
- bool this_progress = true; \
- do_pass \
- nir_validate_shader(nir); \
- if (should_clone_nir()) { \
- nir_shader *clone = nir_shader_clone(ralloc_parent(nir), nir); \
- ralloc_free(nir); \
- nir = clone; \
- } \
- this_progress; \
-}))
-
-#define OPT(pass, ...) _OPT( \
- nir_metadata_set_validation_flag(nir); \
- this_progress = pass(nir ,##__VA_ARGS__); \
- if (this_progress) { \
- progress = true; \
- nir_metadata_check_validation_flag(nir); \
- } \
-)
-
-#define OPT_V(pass, ...) _OPT( \
- pass(nir, ##__VA_ARGS__); \
-)
+#define OPT(pass, ...) ({ \
+ bool this_progress = false; \
+ NIR_PASS(this_progress, nir, pass, ##__VA_ARGS__); \
+ if (this_progress) \
+ progress = true; \
+ this_progress; \
+})
+
+#define OPT_V(pass, ...) NIR_PASS_V(nir, pass, ##__VA_ARGS__)
static nir_shader *
nir_optimize(nir_shader *nir, bool is_scalar)
/* Get rid of split copies */
nir = nir_optimize(nir, is_scalar);
- OPT(nir_remove_dead_variables);
+ OPT(nir_remove_dead_variables, nir_var_local);
return nir;
}
-/* Lowers inputs, outputs, uniforms, and samplers for i965
- *
- * This function does all of the standard lowering prior to post-processing.
- * The lowering done is highly gen, stage, and backend-specific. The
- * shader_prog parameter is optional and is used only for lowering sampler
- * derefs and atomics for GLSL shaders.
- */
+/** Lower input and output loads and stores for i965. */
nir_shader *
-brw_lower_nir(nir_shader *nir,
- const struct brw_device_info *devinfo,
- const struct gl_shader_program *shader_prog,
- bool is_scalar)
+brw_nir_lower_io(nir_shader *nir,
+ const struct brw_device_info *devinfo,
+ bool is_scalar)
{
bool progress; /* Written by OPT and OPT_V */
(void)progress;
OPT_V(brw_nir_lower_inputs, devinfo, is_scalar);
- OPT_V(brw_nir_lower_outputs, is_scalar);
- OPT_V(brw_nir_lower_uniforms, is_scalar);
+ OPT_V(brw_nir_lower_outputs, devinfo, is_scalar);
+ if (nir->stage == MESA_SHADER_COMPUTE)
+ OPT_V(brw_nir_lower_shared);
OPT_V(nir_lower_io, nir_var_all, is_scalar ? type_size_scalar : type_size_vec4);
- if (shader_prog) {
- OPT_V(nir_lower_samplers, shader_prog);
- }
-
- OPT(nir_lower_system_values);
-
- if (shader_prog) {
- OPT_V(nir_lower_atomics, shader_prog);
- }
-
return nir_optimize(nir, is_scalar);
}
if (unlikely(debug_enabled)) {
/* Re-index SSA defs so we print more sensible numbers. */
- nir_foreach_overload(nir, overload) {
- if (overload->impl)
- nir_index_ssa_defs(overload->impl);
+ nir_foreach_function(nir, function) {
+ if (function->impl)
+ nir_index_ssa_defs(function->impl);
}
fprintf(stderr, "NIR (SSA form) for %s shader:\n",
(void)progress;
nir = brw_preprocess_nir(nir, is_scalar);
- nir = brw_lower_nir(nir, devinfo, shader_prog, is_scalar);
+
+ OPT(nir_lower_system_values);
+ OPT_V(brw_nir_lower_uniforms, is_scalar);
+
+ if (shader_prog) {
+ OPT_V(nir_lower_samplers, shader_prog);
+ OPT_V(nir_lower_atomics, shader_prog);
+ }
+
+ if (nir->stage != MESA_SHADER_TESS_CTRL &&
+ nir->stage != MESA_SHADER_TESS_EVAL) {
+ nir = brw_nir_lower_io(nir, devinfo, is_scalar);
+ }
return nir;
}
bool is_scalar);
nir_shader *brw_preprocess_nir(nir_shader *nir, bool is_scalar);
-nir_shader *brw_lower_nir(nir_shader *nir,
- const struct brw_device_info *devinfo,
- const struct gl_shader_program *shader_prog,
- bool is_scalar);
+nir_shader *brw_nir_lower_io(nir_shader *nir,
+ const struct brw_device_info *devinfo,
+ bool is_scalar);
nir_shader *brw_postprocess_nir(nir_shader *nir,
const struct brw_device_info *devinfo,
bool is_scalar);
uint8_t resolve_status;
nir_alu_instr *alu = nir_instr_as_alu(instr);
switch (alu->op) {
- case nir_op_bany2:
- case nir_op_bany3:
- case nir_op_bany4:
case nir_op_ball_fequal2:
case nir_op_ball_iequal2:
case nir_op_ball_fequal3:
void
brw_nir_analyze_boolean_resolves(nir_shader *shader)
{
- nir_foreach_overload(shader, overload)
- if (overload->impl)
- analyze_boolean_resolves_impl(overload->impl);
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ analyze_boolean_resolves_impl(function->impl);
+ }
}
{
bool progress = false;
- nir_foreach_overload(shader, overload) {
- if (overload->impl)
- progress |= brw_nir_opt_peephole_ffma_impl(overload->impl);
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ progress |= brw_nir_opt_peephole_ffma_impl(function->impl);
}
return progress;
brw_emit_pipe_control_flush(struct brw_context *brw, uint32_t flags)
{
if (brw->gen >= 8) {
- gen8_add_cs_stall_workaround_bits(&flags);
+ if (brw->gen == 8)
+ gen8_add_cs_stall_workaround_bits(&flags);
BEGIN_BATCH(6);
OUT_BATCH(_3DSTATE_PIPE_CONTROL | (6 - 2));
uint32_t imm_lower, uint32_t imm_upper)
{
if (brw->gen >= 8) {
- gen8_add_cs_stall_workaround_bits(&flags);
+ if (brw->gen == 8)
+ gen8_add_cs_stall_workaround_bits(&flags);
BEGIN_BATCH(6);
OUT_BATCH(_3DSTATE_PIPE_CONTROL | (6 - 2));
void brwInitFragProgFuncs( struct dd_function_table *functions )
{
- assert(functions->ProgramStringNotify == _tnl_program_string);
+ /* assert(functions->ProgramStringNotify == _tnl_program_string); */
functions->NewProgram = brwNewProgram;
functions->DeleteProgram = brwDeleteProgram;
brw_dump_ir(const char *stage, struct gl_shader_program *shader_prog,
struct gl_shader *shader, struct gl_program *prog);
+void brw_upload_tcs_prog(struct brw_context *brw,
+ uint64_t per_vertex_slots, uint32_t per_patch_slots);
+void brw_upload_tes_prog(struct brw_context *brw,
+ uint64_t per_vertex_slots, uint32_t per_patch_slots);
+
#ifdef __cplusplus
} /* extern "C" */
#endif
#define BRW_SWIZZLE_YZXW BRW_SWIZZLE4(1,2,0,3)
#define BRW_SWIZZLE_ZXYW BRW_SWIZZLE4(2,0,1,3)
#define BRW_SWIZZLE_ZWZW BRW_SWIZZLE4(2,3,2,3)
+#define BRW_SWIZZLE_WZYX BRW_SWIZZLE4(3,2,1,0)
static inline bool
brw_is_single_value_swizzle(unsigned swiz)
}
}
-static inline bool
-type_is_signed(unsigned type)
-{
- switch(type) {
- case BRW_REGISTER_TYPE_D:
- case BRW_REGISTER_TYPE_W:
- case BRW_REGISTER_TYPE_F:
- case BRW_REGISTER_TYPE_B:
- case BRW_REGISTER_TYPE_V:
- case BRW_REGISTER_TYPE_VF:
- case BRW_REGISTER_TYPE_DF:
- case BRW_REGISTER_TYPE_HF:
- case BRW_REGISTER_TYPE_Q:
- return true;
-
- case BRW_REGISTER_TYPE_UD:
- case BRW_REGISTER_TYPE_UW:
- case BRW_REGISTER_TYPE_UB:
- case BRW_REGISTER_TYPE_UV:
- case BRW_REGISTER_TYPE_UQ:
- return false;
-
- default:
- unreachable("not reached");
- }
-}
-
/**
* Construct a brw_reg.
* \param file one of the BRW_x_REGISTER_FILE values
static void
brw_upload_tcs_samplers(struct brw_context *brw)
{
- /* BRW_NEW_TESS_CTRL_PROGRAM */
+ /* BRW_NEW_TESS_PROGRAMS */
struct gl_program *tcs = (struct gl_program *) brw->tess_ctrl_program;
if (!tcs)
return;
.dirty = {
.mesa = _NEW_TEXTURE,
.brw = BRW_NEW_BATCH |
- BRW_NEW_TESS_CTRL_PROGRAM,
+ BRW_NEW_TESS_PROGRAMS,
},
.emit = brw_upload_tcs_samplers,
};
static void
brw_upload_tes_samplers(struct brw_context *brw)
{
- /* BRW_NEW_TESS_EVAL_PROGRAM */
+ /* BRW_NEW_TESS_PROGRAMS */
struct gl_program *tes = (struct gl_program *) brw->tess_eval_program;
if (!tes)
return;
.dirty = {
.mesa = _NEW_TEXTURE,
.brw = BRW_NEW_BATCH |
- BRW_NEW_TESS_EVAL_PROGRAM,
+ BRW_NEW_TESS_PROGRAMS,
},
.emit = brw_upload_tes_samplers,
};
#include "brw_context.h"
#include "brw_cfg.h"
#include "brw_eu.h"
+#include "brw_fs.h"
#include "brw_nir.h"
-#include "glsl/glsl_parser_extras.h"
+#include "brw_vec4_tes.h"
#include "main/shaderobj.h"
#include "main/uniforms.h"
#include "util/debug.h"
compiler->scalar_stage[MESA_SHADER_VERTEX] =
devinfo->gen >= 8 && !(INTEL_DEBUG & DEBUG_VEC4VS);
+ compiler->scalar_stage[MESA_SHADER_TESS_CTRL] = false;
+ compiler->scalar_stage[MESA_SHADER_TESS_EVAL] =
+ devinfo->gen >= 8 && env_var_as_boolean("INTEL_SCALAR_TES", true);
compiler->scalar_stage[MESA_SHADER_GEOMETRY] =
devinfo->gen >= 8 && env_var_as_boolean("INTEL_SCALAR_GS", false);
compiler->scalar_stage[MESA_SHADER_FRAGMENT] = true;
nir_shader_compiler_options *nir_options =
rzalloc(compiler, nir_shader_compiler_options);
nir_options->native_integers = true;
+ nir_options->vertex_id_zero_based = true;
+ nir_options->lower_fdiv = true;
/* In order to help allow for better CSE at the NIR level we tell NIR
* to split all ffma instructions during opt_algebraic and we then
* re-combine them as a later step.
*/
nir_options->lower_ffma = true;
nir_options->lower_sub = true;
+ nir_options->lower_fdiv = true;
+ nir_options->lower_scmp = true;
+ nir_options->lower_fmod = true;
+ nir_options->lower_bitfield_extract = true;
+ nir_options->lower_bitfield_insert = true;
+ nir_options->lower_uadd_carry = true;
+ nir_options->lower_usub_borrow = true;
+
/* In the vec4 backend, our dpN instruction replicates its result to all
* the components of a vec4. We would like NIR to give us replicated fdot
* instructions because it can optimize better for us.
compiler->glsl_compiler_options[i].LowerBufferInterfaceBlocks = true;
}
+ compiler->glsl_compiler_options[MESA_SHADER_TESS_CTRL].EmitNoIndirectInput = false;
+ compiler->glsl_compiler_options[MESA_SHADER_TESS_EVAL].EmitNoIndirectInput = false;
+
if (compiler->scalar_stage[MESA_SHADER_GEOMETRY])
compiler->glsl_compiler_options[MESA_SHADER_GEOMETRY].EmitNoIndirectInput = false;
case GLSL_TYPE_ERROR:
case GLSL_TYPE_INTERFACE:
case GLSL_TYPE_DOUBLE:
+ case GLSL_TYPE_FUNCTION:
unreachable("not reached");
}
return "mulh";
case SHADER_OPCODE_MOV_INDIRECT:
return "mov_indirect";
+
+ case VEC4_OPCODE_URB_READ:
+ return "urb_read";
+ case TCS_OPCODE_GET_INSTANCE_ID:
+ return "tcs_get_instance_id";
+ case TCS_OPCODE_URB_WRITE:
+ return "tcs_urb_write";
+ case TCS_OPCODE_SET_INPUT_URB_OFFSETS:
+ return "tcs_set_input_urb_offsets";
+ case TCS_OPCODE_SET_OUTPUT_URB_OFFSETS:
+ return "tcs_set_output_urb_offsets";
+ case TCS_OPCODE_GET_PRIMITIVE_ID:
+ return "tcs_get_primitive_id";
+ case TCS_OPCODE_CREATE_BARRIER_HEADER:
+ return "tcs_create_barrier_header";
+ case TCS_OPCODE_SRC0_010_IS_ZERO:
+ return "tcs_src0<0,1,0>_is_zero";
+ case TCS_OPCODE_RELEASE_INPUT:
+ return "tcs_release_input";
+ case TCS_OPCODE_THREAD_END:
+ return "tcs_thread_end";
+ case TES_OPCODE_CREATE_INPUT_READ_HEADER:
+ return "tes_create_input_read_header";
+ case TES_OPCODE_ADD_INDIRECT_URB_OFFSET:
+ return "tes_add_indirect_urb_offset";
+ case TES_OPCODE_GET_PRIMITIVE_ID:
+ return "tes_get_primitive_id";
}
unreachable("not reached");
case SHADER_OPCODE_URB_WRITE_SIMD8_MASKED_PER_SLOT:
case FS_OPCODE_FB_WRITE:
case SHADER_OPCODE_BARRIER:
+ case TCS_OPCODE_URB_WRITE:
+ case TCS_OPCODE_RELEASE_INPUT:
return true;
default:
return false;
}
}
+extern "C" const unsigned *
+brw_compile_tes(const struct brw_compiler *compiler,
+ void *log_data,
+ void *mem_ctx,
+ const struct brw_tes_prog_key *key,
+ struct brw_tes_prog_data *prog_data,
+ const nir_shader *src_shader,
+ struct gl_shader_program *shader_prog,
+ int shader_time_index,
+ unsigned *final_assembly_size,
+ char **error_str)
+{
+ const struct brw_device_info *devinfo = compiler->devinfo;
+ struct gl_shader *shader =
+ shader_prog->_LinkedShaders[MESA_SHADER_TESS_EVAL];
+ const bool is_scalar = compiler->scalar_stage[MESA_SHADER_TESS_EVAL];
+
+ nir_shader *nir = nir_shader_clone(mem_ctx, src_shader);
+ nir = brw_nir_apply_sampler_key(nir, devinfo, &key->tex, is_scalar);
+ nir->info.inputs_read = key->inputs_read;
+ nir->info.patch_inputs_read = key->patch_inputs_read;
+ nir = brw_nir_lower_io(nir, compiler->devinfo, is_scalar);
+ nir = brw_postprocess_nir(nir, compiler->devinfo, is_scalar);
+
+ brw_compute_vue_map(devinfo, &prog_data->base.vue_map,
+ nir->info.outputs_written,
+ nir->info.separate_shader);
+
+ unsigned output_size_bytes = prog_data->base.vue_map.num_slots * 4 * 4;
+
+ assert(output_size_bytes >= 1);
+ if (output_size_bytes > GEN7_MAX_DS_URB_ENTRY_SIZE_BYTES) {
+ if (error_str)
+ *error_str = ralloc_strdup(mem_ctx, "DS outputs exceed maximum size");
+ return NULL;
+ }
+
+ /* URB entry sizes are stored as a multiple of 64 bytes. */
+ prog_data->base.urb_entry_size = ALIGN(output_size_bytes, 64) / 64;
+
+ struct brw_vue_map input_vue_map;
+ brw_compute_tess_vue_map(&input_vue_map,
+ nir->info.inputs_read & ~VARYING_BIT_PRIMITIVE_ID,
+ nir->info.patch_inputs_read);
+
+ bool need_patch_header = nir->info.system_values_read &
+ (BITFIELD64_BIT(SYSTEM_VALUE_TESS_LEVEL_OUTER) |
+ BITFIELD64_BIT(SYSTEM_VALUE_TESS_LEVEL_INNER));
+
+ /* The TES will pull most inputs using URB read messages.
+ *
+ * However, we push the patch header for TessLevel factors when required,
+ * as it's a tiny amount of extra data.
+ */
+ prog_data->base.urb_read_length = need_patch_header ? 1 : 0;
+
+ if (unlikely(INTEL_DEBUG & DEBUG_TES)) {
+ fprintf(stderr, "TES Input ");
+ brw_print_vue_map(stderr, &input_vue_map);
+ fprintf(stderr, "TES Output ");
+ brw_print_vue_map(stderr, &prog_data->base.vue_map);
+ }
+
+ if (is_scalar) {
+ fs_visitor v(compiler, log_data, mem_ctx, (void *) key,
+ &prog_data->base.base, shader->Program, nir, 8,
+ shader_time_index, &input_vue_map);
+ if (!v.run_tes()) {
+ if (error_str)
+ *error_str = ralloc_strdup(mem_ctx, v.fail_msg);
+ return NULL;
+ }
+
+ prog_data->base.dispatch_mode = DISPATCH_MODE_SIMD8;
+
+ fs_generator g(compiler, log_data, mem_ctx, (void *) key,
+ &prog_data->base.base, v.promoted_constants, false,
+ MESA_SHADER_TESS_EVAL);
+ if (unlikely(INTEL_DEBUG & DEBUG_TES)) {
+ g.enable_debug(ralloc_asprintf(mem_ctx,
+ "%s tessellation evaluation shader %s",
+ nir->info.label ? nir->info.label
+ : "unnamed",
+ nir->info.name));
+ }
+
+ g.generate_code(v.cfg, 8);
+
+ return g.get_assembly(final_assembly_size);
+ } else {
+ brw::vec4_tes_visitor v(compiler, log_data, key, prog_data,
+ nir, mem_ctx, shader_time_index);
+ if (!v.run()) {
+ if (error_str)
+ *error_str = ralloc_strdup(mem_ctx, v.fail_msg);
+ return NULL;
+ }
+
+ if (unlikely(INTEL_DEBUG & DEBUG_TES))
+ v.dump_instructions();
+
+ return brw_vec4_generate_assembly(compiler, log_data, mem_ctx, nir,
+ &prog_data->base, v.cfg,
+ final_assembly_size);
+ }
+}
unsigned control_data_header_size_bits;
};
-struct brw_compiler *
-brw_compiler_create(void *mem_ctx, const struct brw_device_info *devinfo);
-
void
brw_assign_common_binding_table_offsets(gl_shader_stage stage,
const struct brw_device_info *devinfo,
bool brw_vs_precompile(struct gl_context *ctx,
struct gl_shader_program *shader_prog,
struct gl_program *prog);
+bool brw_tcs_precompile(struct gl_context *ctx,
+ struct gl_shader_program *shader_prog,
+ struct gl_program *prog);
+bool brw_tes_precompile(struct gl_context *ctx,
+ struct gl_shader_program *shader_prog,
+ struct gl_program *prog);
bool brw_gs_precompile(struct gl_context *ctx,
struct gl_shader_program *shader_prog,
struct gl_program *prog);
GET_FIELD(surf[4], GEN7_SURFACE_MIN_ARRAY_ELEMENT),
GET_FIELD(surf[4], GEN7_SURFACE_RENDER_TARGET_VIEW_EXTENT) + 1,
1 << GET_BITS(surf[4], 5, 3));
- batch_out(brw, name, offset, 5, "x,y offset: %d,%d, min LOD: %d\n",
+ batch_out(brw, name, offset, 5, "x,y offset: %d,%d, min LOD: %d,"
+ " tr_mode (gen9+): %d, mip tail (gen9+): %d\n",
GET_FIELD(surf[5], BRW_SURFACE_X_OFFSET),
GET_FIELD(surf[5], BRW_SURFACE_Y_OFFSET),
- GET_FIELD(surf[5], GEN7_SURFACE_MIN_LOD));
+ GET_FIELD(surf[5], GEN7_SURFACE_MIN_LOD),
+ GET_FIELD(surf[5], GEN9_SURFACE_TRMODE),
+ GET_FIELD(surf[5], GEN9_SURFACE_MIP_TAIL_START_LOD));
batch_out(brw, name, offset, 6, "AUX pitch: %d qpitch: %d\n",
GET_FIELD(surf[6], GEN8_SURFACE_AUX_QPITCH) << 2,
GET_FIELD(surf[6], GEN8_SURFACE_AUX_PITCH) << 2);
&gen7_hw_binding_tables, /* Enable hw-generated binding tables for Haswell */
&brw_vs_image_surfaces, /* Before vs push/pull constants and binding table */
+ &brw_tcs_image_surfaces, /* Before tcs push/pull constants and binding table */
+ &brw_tes_image_surfaces, /* Before tes push/pull constants and binding table */
&brw_gs_image_surfaces, /* Before gs push/pull constants and binding table */
&brw_wm_image_surfaces, /* Before wm push/pull constants and binding table */
&gen6_vs_push_constants, /* Before vs_state */
+ &gen7_tcs_push_constants,
+ &gen7_tes_push_constants,
&gen6_gs_push_constants, /* Before gs_state */
&gen6_wm_push_constants, /* Before wm_surfaces and constant_buffer */
&brw_vs_pull_constants,
&brw_vs_ubo_surfaces,
&brw_vs_abo_surfaces,
+ &brw_tcs_pull_constants,
+ &brw_tcs_ubo_surfaces,
+ &brw_tcs_abo_surfaces,
+ &brw_tes_pull_constants,
+ &brw_tes_ubo_surfaces,
+ &brw_tes_abo_surfaces,
&brw_gs_pull_constants,
&brw_gs_ubo_surfaces,
&brw_gs_abo_surfaces,
&gen6_renderbuffer_surfaces,
&brw_texture_surfaces,
&brw_vs_binding_table,
+ &brw_tcs_binding_table,
+ &brw_tes_binding_table,
&brw_gs_binding_table,
&brw_wm_binding_table,
&brw_fs_samplers,
&brw_vs_samplers,
+ &brw_tcs_samplers,
+ &brw_tes_samplers,
&brw_gs_samplers,
&gen6_multisample_state,
ctx->DriverFlags.NewTextureBuffer = BRW_NEW_TEXTURE_BUFFER;
ctx->DriverFlags.NewAtomicBuffer = BRW_NEW_ATOMIC_BUFFER;
ctx->DriverFlags.NewImageUnits = BRW_NEW_IMAGE_UNITS;
+ ctx->DriverFlags.NewDefaultTessLevels = BRW_NEW_DEFAULT_TESS_LEVELS;
}
DEFINE_BIT(BRW_NEW_URB_FENCE),
DEFINE_BIT(BRW_NEW_FRAGMENT_PROGRAM),
DEFINE_BIT(BRW_NEW_GEOMETRY_PROGRAM),
- DEFINE_BIT(BRW_NEW_TESS_EVAL_PROGRAM),
- DEFINE_BIT(BRW_NEW_TESS_CTRL_PROGRAM),
+ DEFINE_BIT(BRW_NEW_TESS_PROGRAMS),
DEFINE_BIT(BRW_NEW_VERTEX_PROGRAM),
DEFINE_BIT(BRW_NEW_CURBE_OFFSETS),
DEFINE_BIT(BRW_NEW_REDUCED_PRIMITIVE),
DEFINE_BIT(BRW_NEW_BINDING_TABLE_POINTERS),
DEFINE_BIT(BRW_NEW_INDICES),
DEFINE_BIT(BRW_NEW_VERTICES),
+ DEFINE_BIT(BRW_NEW_DEFAULT_TESS_LEVELS),
DEFINE_BIT(BRW_NEW_BATCH),
DEFINE_BIT(BRW_NEW_INDEX_BUFFER),
DEFINE_BIT(BRW_NEW_VS_CONSTBUF),
}
static inline void
+brw_upload_tess_programs(struct brw_context *brw)
+{
+ if (brw->tess_eval_program) {
+ uint64_t per_vertex_slots = brw->tess_eval_program->Base.InputsRead;
+ uint32_t per_patch_slots =
+ brw->tess_eval_program->Base.PatchInputsRead;
+
+ /* The TCS may have additional outputs which aren't read by the
+ * TES (possibly for cross-thread communication). These need to
+ * be stored in the Patch URB Entry as well.
+ */
+ if (brw->tess_ctrl_program) {
+ per_vertex_slots |= brw->tess_ctrl_program->Base.OutputsWritten;
+ per_patch_slots |=
+ brw->tess_ctrl_program->Base.PatchOutputsWritten;
+ }
+
+ brw_upload_tcs_prog(brw, per_vertex_slots, per_patch_slots);
+ brw_upload_tes_prog(brw, per_vertex_slots, per_patch_slots);
+ } else {
+ brw->tcs.prog_data = NULL;
+ brw->tcs.base.prog_data = NULL;
+ brw->tes.prog_data = NULL;
+ brw->tes.base.prog_data = NULL;
+ }
+}
+
+static inline void
brw_upload_programs(struct brw_context *brw,
enum brw_pipeline pipeline)
{
if (pipeline == BRW_RENDER_PIPELINE) {
brw_upload_vs_prog(brw);
+ brw_upload_tess_programs(brw);
if (brw->gen < 6)
brw_upload_ff_gs_prog(brw);
bool old_separate = brw->vue_map_geom_out.separate;
if (brw->geometry_program)
brw->vue_map_geom_out = brw->gs.prog_data->base.vue_map;
+ else if (brw->tess_eval_program)
+ brw->vue_map_geom_out = brw->tes.prog_data->base.vue_map;
else
brw->vue_map_geom_out = brw->vs.prog_data->base.vue_map;
if (brw->tess_eval_program != ctx->TessEvalProgram._Current) {
brw->tess_eval_program = ctx->TessEvalProgram._Current;
- brw->ctx.NewDriverState |= BRW_NEW_TESS_EVAL_PROGRAM;
+ brw->ctx.NewDriverState |= BRW_NEW_TESS_PROGRAMS;
}
if (brw->tess_ctrl_program != ctx->TessCtrlProgram._Current) {
brw->tess_ctrl_program = ctx->TessCtrlProgram._Current;
- brw->ctx.NewDriverState |= BRW_NEW_TESS_CTRL_PROGRAM;
+ brw->ctx.NewDriverState |= BRW_NEW_TESS_PROGRAMS;
}
if (brw->geometry_program != ctx->GeometryProgram._Current) {
#include "brw_context.h"
#include "brw_state.h"
#include "brw_defines.h"
-
-struct surface_format_info {
- bool exists;
- int sampling;
- int filtering;
- int shadow_compare;
- int chroma_key;
- int render_target;
- int alpha_blend;
- int input_vb;
- int streamed_output_vb;
- int color_processing;
- int lossless_compression;
- const char *name;
-};
+#include "brw_wm.h"
+#include "brw_surface_formats.h"
/* This macro allows us to write the table almost as it appears in the PRM,
* while restructuring it to turn it into the C code we want.
* - VOL4_Part1 section 3.9.11 Render Target Write.
* - Render Target Surface Types [SKL+]
*/
-const struct surface_format_info surface_formats[] = {
+const struct brw_surface_format_info surface_formats[] = {
/* smpl filt shad CK RT AB VB SO color ccs_e */
SF( Y, 50, x, x, Y, Y, Y, Y, x, 90, R32G32B32A32_FLOAT)
SF( Y, x, x, x, Y, x, Y, Y, x, 90, R32G32B32A32_SINT)
for (format = MESA_FORMAT_NONE + 1; format < MESA_FORMAT_COUNT; format++) {
uint32_t texture, render;
- const struct surface_format_info *rinfo, *tinfo;
+ const struct brw_surface_format_info *rinfo, *tinfo;
bool is_integer = _mesa_is_format_integer_color(format);
render = texture = brw_format_for_mesa_format(format);
if (brw->gen >= 8)
ctx->TextureFormatSupported[MESA_FORMAT_Z_UNORM16] = true;
+ /* The RGBX formats are not renderable. Normally these get mapped
+ * internally to RGBA formats when rendering. However on Gen9+ when this
+ * internal override is used fast clears don't work so they are disabled in
+ * brw_meta_fast_clear. To avoid this problem we can just pretend not to
+ * support RGBX formats at all. This will cause the upper layers of Mesa to
+ * pick the RGBA formats instead. This works fine because when it is used
+ * as a texture source the swizzle state is programmed to force the alpha
+ * channel to 1.0 anyway. We could also do this for all gens except that
+ * it's a bit more difficult when the hardware doesn't support texture
+ * swizzling. Gens using the blorp have further problems because that
+ * doesn't implement this swizzle override. We don't need to do this for
+ * BGRX because that actually is supported natively on Gen8+.
+ */
+ if (brw->gen >= 9) {
+ static const mesa_format rgbx_formats[] = {
+ MESA_FORMAT_R8G8B8X8_UNORM,
+ MESA_FORMAT_R8G8B8X8_SRGB,
+ MESA_FORMAT_RGBX_UNORM16,
+ MESA_FORMAT_RGBX_FLOAT16,
+ MESA_FORMAT_RGBX_FLOAT32
+ };
+
+ for (int i = 0; i < ARRAY_SIZE(rgbx_formats); i++) {
+ ctx->TextureFormatSupported[rgbx_formats[i]] = false;
+ brw->format_supported_as_render_target[rgbx_formats[i]] = false;
+ }
+ }
+
/* On hardware that lacks support for ETC1, we map ETC1 to RGBX
* during glCompressedTexImage2D(). See intel_mipmap_tree::wraps_etc1.
*/
brw_losslessly_compressible_format(struct brw_context *brw,
uint32_t brw_format)
{
- const struct surface_format_info * const sinfo =
+ const struct brw_surface_format_info * const sinfo =
&surface_formats[brw_format];
const int gen = brw->gen * 10;
--- /dev/null
+/*
+ * Copyright © 2011 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#pragma once
+
+struct brw_surface_format_info {
+ bool exists;
+ int sampling;
+ int filtering;
+ int shadow_compare;
+ int chroma_key;
+ int render_target;
+ int alpha_blend;
+ int input_vb;
+ int streamed_output_vb;
+ int color_processing;
+ int lossless_compression;
+ const char *name;
+};
+
+extern const struct brw_surface_format_info surface_formats[];
--- /dev/null
+/*
+ * Copyright © 2014 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file brw_tcs.c
+ *
+ * Tessellation control shader state upload code.
+ */
+
+#include "brw_context.h"
+#include "brw_nir.h"
+#include "brw_program.h"
+#include "brw_shader.h"
+#include "brw_state.h"
+#include "program/prog_parameter.h"
+
+static void
+brw_tcs_debug_recompile(struct brw_context *brw,
+ struct gl_shader_program *shader_prog,
+ const struct brw_tcs_prog_key *key)
+{
+ struct brw_cache_item *c = NULL;
+ const struct brw_tcs_prog_key *old_key = NULL;
+ bool found = false;
+
+ perf_debug("Recompiling tessellation control shader for program %d\n",
+ shader_prog->Name);
+
+ for (unsigned int i = 0; i < brw->cache.size; i++) {
+ for (c = brw->cache.items[i]; c; c = c->next) {
+ if (c->cache_id == BRW_CACHE_TCS_PROG) {
+ old_key = c->key;
+
+ if (old_key->program_string_id == key->program_string_id)
+ break;
+ }
+ }
+ if (c)
+ break;
+ }
+
+ if (!c) {
+ perf_debug(" Didn't find previous compile in the shader cache for "
+ "debug\n");
+ return;
+ }
+
+ found |= key_debug(brw, "input vertices", old_key->input_vertices,
+ key->input_vertices);
+ found |= key_debug(brw, "outputs written", old_key->outputs_written,
+ key->outputs_written);
+ found |= key_debug(brw, "patch outputs written", old_key->patch_outputs_written,
+ key->patch_outputs_written);
+ found |= key_debug(brw, "TES primitive mode", old_key->tes_primitive_mode,
+ key->tes_primitive_mode);
+ found |= brw_debug_recompile_sampler_key(brw, &old_key->tex, &key->tex);
+
+ if (!found) {
+ perf_debug(" Something else\n");
+ }
+}
+
+static bool
+brw_codegen_tcs_prog(struct brw_context *brw,
+ struct gl_shader_program *shader_prog,
+ struct brw_tess_ctrl_program *tcp,
+ struct brw_tcs_prog_key *key)
+{
+ struct gl_context *ctx = &brw->ctx;
+ const struct brw_compiler *compiler = brw->intelScreen->compiler;
+ struct brw_stage_state *stage_state = &brw->tcs.base;
+ nir_shader *nir;
+ struct brw_tcs_prog_data prog_data;
+ bool start_busy = false;
+ double start_time = 0;
+
+ if (tcp) {
+ nir = tcp->program.Base.nir;
+ } else {
+ /* Create a dummy nir_shader. We won't actually use NIR code to
+ * generate assembly (it's easier to generate assembly directly),
+ * but the whole compiler assumes one of these exists.
+ */
+ const nir_shader_compiler_options *options =
+ ctx->Const.ShaderCompilerOptions[MESA_SHADER_TESS_CTRL].NirOptions;
+ nir = nir_shader_create(NULL, MESA_SHADER_TESS_CTRL, options);
+ nir->num_uniforms = 2; /* both halves of the patch header */
+ nir->info.outputs_written = key->outputs_written;
+ nir->info.inputs_read = key->outputs_written;
+ nir->info.tcs.vertices_out = key->input_vertices;
+ nir->info.name = ralloc_strdup(nir, "passthrough");
+ }
+
+ memset(&prog_data, 0, sizeof(prog_data));
+
+ /* Allocate the references to the uniforms that will end up in the
+ * prog_data associated with the compiled program, and which will be freed
+ * by the state cache.
+ *
+ * Note: param_count needs to be num_uniform_components * 4, since we add
+ * padding around uniform values below vec4 size, so the worst case is that
+ * every uniform is a float which gets padded to the size of a vec4.
+ */
+ struct gl_shader *tcs = shader_prog ?
+ shader_prog->_LinkedShaders[MESA_SHADER_TESS_CTRL] : NULL;
+ int param_count = nir->num_uniforms;
+ if (!compiler->scalar_stage[MESA_SHADER_TESS_CTRL])
+ param_count *= 4;
+
+ prog_data.base.base.param =
+ rzalloc_array(NULL, const gl_constant_value *, param_count);
+ prog_data.base.base.pull_param =
+ rzalloc_array(NULL, const gl_constant_value *, param_count);
+ prog_data.base.base.nr_params = param_count;
+
+ if (tcs) {
+ prog_data.base.base.image_param =
+ rzalloc_array(NULL, struct brw_image_param, tcs->NumImages);
+ prog_data.base.base.nr_image_params = tcs->NumImages;
+
+ brw_nir_setup_glsl_uniforms(nir, shader_prog, &tcp->program.Base,
+ &prog_data.base.base, false);
+ } else {
+ /* Upload the Patch URB Header as the first two uniforms.
+ * Do the annoying scrambling so the shader doesn't have to.
+ */
+ const float **param = (const float **) prog_data.base.base.param;
+ static float zero = 0.0f;
+ for (int i = 0; i < 4; i++) {
+ param[7 - i] = &ctx->TessCtrlProgram.patch_default_outer_level[i];
+ }
+
+ if (key->tes_primitive_mode == GL_QUADS) {
+ param[3] = &ctx->TessCtrlProgram.patch_default_inner_level[0];
+ param[2] = &ctx->TessCtrlProgram.patch_default_inner_level[1];
+ param[1] = &zero;
+ param[0] = &zero;
+ } else if (key->tes_primitive_mode == GL_TRIANGLES) {
+ param[4] = &ctx->TessCtrlProgram.patch_default_inner_level[0];
+ for (int i = 0; i < 4; i++)
+ param[i] = &zero;
+ }
+ }
+
+ if (unlikely(INTEL_DEBUG & DEBUG_TCS) && tcs)
+ brw_dump_ir("tessellation control", shader_prog, tcs, NULL);
+
+ int st_index = -1;
+ if (unlikely(INTEL_DEBUG & DEBUG_SHADER_TIME))
+ st_index = brw_get_shader_time_index(brw, shader_prog, NULL, ST_TCS);
+
+ if (unlikely(brw->perf_debug)) {
+ start_busy = brw->batch.last_bo && drm_intel_bo_busy(brw->batch.last_bo);
+ start_time = get_time();
+ }
+
+ void *mem_ctx = ralloc_context(NULL);
+ unsigned program_size;
+ char *error_str;
+ const unsigned *program =
+ brw_compile_tcs(compiler, brw, mem_ctx, key, &prog_data, nir, st_index,
+ &program_size, &error_str);
+ if (program == NULL) {
+ if (shader_prog) {
+ shader_prog->LinkStatus = false;
+ ralloc_strcat(&shader_prog->InfoLog, error_str);
+ } else {
+ ralloc_free(nir);
+ }
+
+ _mesa_problem(NULL, "Failed to compile tessellation control shader: "
+ "%s\n", error_str);
+
+ ralloc_free(mem_ctx);
+ return false;
+ }
+
+ if (unlikely(brw->perf_debug)) {
+ struct brw_shader *btcs = (struct brw_shader *) tcs;
+ if (btcs->compiled_once) {
+ brw_tcs_debug_recompile(brw, shader_prog, key);
+ }
+ if (start_busy && !drm_intel_bo_busy(brw->batch.last_bo)) {
+ perf_debug("TCS compile took %.03f ms and stalled the GPU\n",
+ (get_time() - start_time) * 1000);
+ }
+ btcs->compiled_once = true;
+ }
+
+ /* Scratch space is used for register spilling */
+ if (prog_data.base.base.total_scratch) {
+ brw_get_scratch_bo(brw, &stage_state->scratch_bo,
+ prog_data.base.base.total_scratch *
+ brw->max_hs_threads);
+ }
+
+ brw_upload_cache(&brw->cache, BRW_CACHE_TCS_PROG,
+ key, sizeof(*key),
+ program, program_size,
+ &prog_data, sizeof(prog_data),
+ &stage_state->prog_offset, &brw->tcs.prog_data);
+ ralloc_free(mem_ctx);
+ if (!tcs)
+ ralloc_free(nir);
+
+ return true;
+}
+
+
+void
+brw_upload_tcs_prog(struct brw_context *brw,
+ uint64_t per_vertex_slots,
+ uint32_t per_patch_slots)
+{
+ struct gl_context *ctx = &brw->ctx;
+ struct gl_shader_program **current = ctx->_Shader->CurrentProgram;
+ struct brw_stage_state *stage_state = &brw->tcs.base;
+ struct brw_tcs_prog_key key;
+ /* BRW_NEW_TESS_PROGRAMS */
+ struct brw_tess_ctrl_program *tcp =
+ (struct brw_tess_ctrl_program *) brw->tess_ctrl_program;
+ struct brw_tess_eval_program *tep =
+ (struct brw_tess_eval_program *) brw->tess_eval_program;
+ assert(tep);
+
+ if (!brw_state_dirty(brw,
+ _NEW_TEXTURE,
+ BRW_NEW_PATCH_PRIMITIVE |
+ BRW_NEW_TESS_PROGRAMS))
+ return;
+
+ struct gl_program *prog = &tcp->program.Base;
+
+ memset(&key, 0, sizeof(key));
+
+ key.input_vertices = ctx->TessCtrlProgram.patch_vertices;
+ key.outputs_written = per_vertex_slots;
+ key.patch_outputs_written = per_patch_slots;
+
+ /* We need to specialize our code generation for tessellation levels
+ * based on the domain the DS is expecting to tessellate.
+ */
+ key.tes_primitive_mode = tep->program.PrimitiveMode;
+
+ if (tcp) {
+ key.program_string_id = tcp->id;
+
+ /* _NEW_TEXTURE */
+ brw_populate_sampler_prog_key_data(ctx, prog, stage_state->sampler_count,
+ &key.tex);
+ } else {
+ key.outputs_written = tep->program.Base.InputsRead;
+ }
+
+
+ if (!brw_search_cache(&brw->cache, BRW_CACHE_TCS_PROG,
+ &key, sizeof(key),
+ &stage_state->prog_offset, &brw->tcs.prog_data)) {
+ bool success = brw_codegen_tcs_prog(brw, current[MESA_SHADER_TESS_CTRL],
+ tcp, &key);
+ assert(success);
+ (void)success;
+ }
+ brw->tcs.base.prog_data = &brw->tcs.prog_data->base.base;
+}
+
+
+bool
+brw_tcs_precompile(struct gl_context *ctx,
+ struct gl_shader_program *shader_prog,
+ struct gl_program *prog)
+{
+ struct brw_context *brw = brw_context(ctx);
+ struct brw_tcs_prog_key key;
+ uint32_t old_prog_offset = brw->tcs.base.prog_offset;
+ struct brw_tcs_prog_data *old_prog_data = brw->tcs.prog_data;
+ bool success;
+
+ struct gl_tess_ctrl_program *tcp = (struct gl_tess_ctrl_program *)prog;
+ struct brw_tess_ctrl_program *btcp = brw_tess_ctrl_program(tcp);
+
+ memset(&key, 0, sizeof(key));
+
+ key.program_string_id = btcp->id;
+ brw_setup_tex_for_precompile(brw, &key.tex, prog);
+
+ /* Guess that the input and output patches have the same dimensionality. */
+ key.input_vertices = shader_prog->TessCtrl.VerticesOut;
+
+ key.tes_primitive_mode =
+ shader_prog->_LinkedShaders[MESA_SHADER_TESS_EVAL] ?
+ shader_prog->TessEval.PrimitiveMode : GL_TRIANGLES;
+
+ key.outputs_written = prog->OutputsWritten;
+ key.patch_outputs_written = prog->PatchOutputsWritten;
+
+ success = brw_codegen_tcs_prog(brw, shader_prog, btcp, &key);
+
+ brw->tcs.base.prog_offset = old_prog_offset;
+ brw->tcs.prog_data = old_prog_data;
+
+ return success;
+}
{
struct brw_stage_state *stage_state = &brw->tcs.base;
- /* BRW_NEW_TESS_CTRL_PROGRAM */
+ /* BRW_NEW_TESS_PROGRAMS */
struct brw_tess_ctrl_program *tcp =
(struct brw_tess_ctrl_program *) brw->tess_ctrl_program;
.mesa = _NEW_PROGRAM_CONSTANTS,
.brw = BRW_NEW_BATCH |
BRW_NEW_TCS_PROG_DATA |
- BRW_NEW_TESS_CTRL_PROGRAM,
+ BRW_NEW_TESS_PROGRAMS,
},
.emit = brw_upload_tcs_pull_constants,
};
brw_upload_tcs_image_surfaces(struct brw_context *brw)
{
struct gl_context *ctx = &brw->ctx;
- /* BRW_NEW_TESS_CTRL_PROGRAM */
+ /* BRW_NEW_TESS_PROGRAMS */
struct gl_shader_program *prog =
ctx->_Shader->CurrentProgram[MESA_SHADER_TESS_CTRL];
.brw = BRW_NEW_BATCH |
BRW_NEW_TCS_PROG_DATA |
BRW_NEW_IMAGE_UNITS |
- BRW_NEW_TESS_CTRL_PROGRAM,
+ BRW_NEW_TESS_PROGRAMS,
},
.emit = brw_upload_tcs_image_surfaces,
};
--- /dev/null
+/*
+ * Copyright © 2014 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file brw_tes.c
+ *
+ * Tessellation evaluation shader state upload code.
+ */
+
+#include "brw_context.h"
+#include "brw_nir.h"
+#include "brw_program.h"
+#include "brw_shader.h"
+#include "brw_state.h"
+#include "program/prog_parameter.h"
+
+static void
+brw_tes_debug_recompile(struct brw_context *brw,
+ struct gl_shader_program *shader_prog,
+ const struct brw_tes_prog_key *key)
+{
+ struct brw_cache_item *c = NULL;
+ const struct brw_tes_prog_key *old_key = NULL;
+ bool found = false;
+
+ perf_debug("Recompiling tessellation evaluation shader for program %d\n",
+ shader_prog->Name);
+
+ for (unsigned int i = 0; i < brw->cache.size; i++) {
+ for (c = brw->cache.items[i]; c; c = c->next) {
+ if (c->cache_id == BRW_CACHE_TES_PROG) {
+ old_key = c->key;
+
+ if (old_key->program_string_id == key->program_string_id)
+ break;
+ }
+ }
+ if (c)
+ break;
+ }
+
+ if (!c) {
+ perf_debug(" Didn't find previous compile in the shader cache for "
+ "debug\n");
+ return;
+ }
+
+ found |= brw_debug_recompile_sampler_key(brw, &old_key->tex, &key->tex);
+ found |= key_debug(brw, "inputs read", old_key->inputs_read,
+ key->inputs_read);
+ found |= key_debug(brw, "patch inputs read", old_key->patch_inputs_read,
+ key->patch_inputs_read);
+
+ if (!found) {
+ perf_debug(" Something else\n");
+ }
+}
+
+static bool
+brw_codegen_tes_prog(struct brw_context *brw,
+ struct gl_shader_program *shader_prog,
+ struct brw_tess_eval_program *tep,
+ struct brw_tes_prog_key *key)
+{
+ const struct brw_compiler *compiler = brw->intelScreen->compiler;
+ const struct brw_device_info *devinfo = brw->intelScreen->devinfo;
+ struct brw_stage_state *stage_state = &brw->tes.base;
+ nir_shader *nir = tep->program.Base.nir;
+ struct brw_tes_prog_data prog_data;
+ bool start_busy = false;
+ double start_time = 0;
+
+ memset(&prog_data, 0, sizeof(prog_data));
+
+ brw_assign_common_binding_table_offsets(MESA_SHADER_TESS_EVAL, devinfo,
+ shader_prog, &tep->program.Base,
+ &prog_data.base.base, 0);
+
+ switch (tep->program.Spacing) {
+ case GL_EQUAL:
+ prog_data.partitioning = BRW_TESS_PARTITIONING_INTEGER;
+ break;
+ case GL_FRACTIONAL_ODD:
+ prog_data.partitioning = BRW_TESS_PARTITIONING_ODD_FRACTIONAL;
+ break;
+ case GL_FRACTIONAL_EVEN:
+ prog_data.partitioning = BRW_TESS_PARTITIONING_EVEN_FRACTIONAL;
+ break;
+ default:
+ unreachable("invalid domain shader spacing");
+ }
+
+ switch (tep->program.PrimitiveMode) {
+ case GL_QUADS:
+ prog_data.domain = BRW_TESS_DOMAIN_QUAD;
+ break;
+ case GL_TRIANGLES:
+ prog_data.domain = BRW_TESS_DOMAIN_TRI;
+ break;
+ case GL_ISOLINES:
+ prog_data.domain = BRW_TESS_DOMAIN_ISOLINE;
+ break;
+ default:
+ unreachable("invalid domain shader primitive mode");
+ }
+
+ if (tep->program.PointMode) {
+ prog_data.output_topology = BRW_TESS_OUTPUT_TOPOLOGY_POINT;
+ } else if (tep->program.PrimitiveMode == GL_ISOLINES) {
+ prog_data.output_topology = BRW_TESS_OUTPUT_TOPOLOGY_LINE;
+ } else {
+ /* Hardware winding order is backwards from OpenGL */
+ switch (tep->program.VertexOrder) {
+ case GL_CCW:
+ prog_data.output_topology = BRW_TESS_OUTPUT_TOPOLOGY_TRI_CW;
+ break;
+ case GL_CW:
+ prog_data.output_topology = BRW_TESS_OUTPUT_TOPOLOGY_TRI_CCW;
+ break;
+ default:
+ unreachable("invalid domain shader vertex order");
+ }
+ }
+
+ /* Allocate the references to the uniforms that will end up in the
+ * prog_data associated with the compiled program, and which will be freed
+ * by the state cache.
+ *
+ * Note: param_count needs to be num_uniform_components * 4, since we add
+ * padding around uniform values below vec4 size, so the worst case is that
+ * every uniform is a float which gets padded to the size of a vec4.
+ */
+ struct gl_shader *tes = shader_prog->_LinkedShaders[MESA_SHADER_TESS_EVAL];
+ int param_count = nir->num_uniforms;
+ if (!compiler->scalar_stage[MESA_SHADER_TESS_EVAL])
+ param_count *= 4;
+
+ prog_data.base.base.param =
+ rzalloc_array(NULL, const gl_constant_value *, param_count);
+ prog_data.base.base.pull_param =
+ rzalloc_array(NULL, const gl_constant_value *, param_count);
+ prog_data.base.base.image_param =
+ rzalloc_array(NULL, struct brw_image_param, tes->NumImages);
+ prog_data.base.base.nr_params = param_count;
+ prog_data.base.base.nr_image_params = tes->NumImages;
+
+ brw_nir_setup_glsl_uniforms(nir, shader_prog, &tep->program.Base,
+ &prog_data.base.base,
+ compiler->scalar_stage[MESA_SHADER_TESS_EVAL]);
+
+ if (unlikely(INTEL_DEBUG & DEBUG_TES))
+ brw_dump_ir("tessellation evaluation", shader_prog, tes, NULL);
+
+ int st_index = -1;
+ if (unlikely(INTEL_DEBUG & DEBUG_SHADER_TIME))
+ st_index = brw_get_shader_time_index(brw, shader_prog, NULL, ST_TES);
+
+ if (unlikely(brw->perf_debug)) {
+ start_busy = brw->batch.last_bo && drm_intel_bo_busy(brw->batch.last_bo);
+ start_time = get_time();
+ }
+
+ void *mem_ctx = ralloc_context(NULL);
+ unsigned program_size;
+ char *error_str;
+ const unsigned *program =
+ brw_compile_tes(compiler, brw, mem_ctx, key, &prog_data, nir,
+ shader_prog, st_index, &program_size, &error_str);
+ if (program == NULL) {
+ if (shader_prog) {
+ shader_prog->LinkStatus = false;
+ ralloc_strcat(&shader_prog->InfoLog, error_str);
+ }
+
+ _mesa_problem(NULL, "Failed to compile tessellation evaluation shader: "
+ "%s\n", error_str);
+
+ ralloc_free(mem_ctx);
+ return false;
+ }
+
+ if (unlikely(brw->perf_debug)) {
+ struct brw_shader *btes = (struct brw_shader *) tes;
+ if (btes->compiled_once) {
+ brw_tes_debug_recompile(brw, shader_prog, key);
+ }
+ if (start_busy && !drm_intel_bo_busy(brw->batch.last_bo)) {
+ perf_debug("TES compile took %.03f ms and stalled the GPU\n",
+ (get_time() - start_time) * 1000);
+ }
+ btes->compiled_once = true;
+ }
+
+ /* Scratch space is used for register spilling */
+ if (prog_data.base.base.total_scratch) {
+ brw_get_scratch_bo(brw, &stage_state->scratch_bo,
+ prog_data.base.base.total_scratch *
+ brw->max_ds_threads);
+ }
+
+ brw_upload_cache(&brw->cache, BRW_CACHE_TES_PROG,
+ key, sizeof(*key),
+ program, program_size,
+ &prog_data, sizeof(prog_data),
+ &stage_state->prog_offset, &brw->tes.prog_data);
+ ralloc_free(mem_ctx);
+
+ return true;
+}
+
+
+void
+brw_upload_tes_prog(struct brw_context *brw,
+ uint64_t per_vertex_slots,
+ uint32_t per_patch_slots)
+{
+ struct gl_context *ctx = &brw->ctx;
+ struct gl_shader_program **current = ctx->_Shader->CurrentProgram;
+ struct brw_stage_state *stage_state = &brw->tes.base;
+ struct brw_tes_prog_key key;
+ /* BRW_NEW_TESS_PROGRAMS */
+ struct brw_tess_eval_program *tep =
+ (struct brw_tess_eval_program *) brw->tess_eval_program;
+
+ if (!brw_state_dirty(brw,
+ _NEW_TEXTURE,
+ BRW_NEW_TESS_PROGRAMS))
+ return;
+
+ struct gl_program *prog = &tep->program.Base;
+
+ memset(&key, 0, sizeof(key));
+
+ key.program_string_id = tep->id;
+
+ /* Ignore gl_TessLevelInner/Outer - we treat them as system values,
+ * not inputs, and they're always present in the URB entry regardless
+ * of whether or not we read them.
+ */
+ key.inputs_read = per_vertex_slots &
+ ~(VARYING_BIT_TESS_LEVEL_INNER | VARYING_BIT_TESS_LEVEL_OUTER);
+ key.patch_inputs_read = per_patch_slots;
+
+ /* _NEW_TEXTURE */
+ brw_populate_sampler_prog_key_data(ctx, prog, stage_state->sampler_count,
+ &key.tex);
+
+ if (!brw_search_cache(&brw->cache, BRW_CACHE_TES_PROG,
+ &key, sizeof(key),
+ &stage_state->prog_offset, &brw->tes.prog_data)) {
+ bool success = brw_codegen_tes_prog(brw, current[MESA_SHADER_TESS_EVAL],
+ tep, &key);
+ assert(success);
+ (void)success;
+ }
+ brw->tes.base.prog_data = &brw->tes.prog_data->base.base;
+}
+
+
+bool
+brw_tes_precompile(struct gl_context *ctx,
+ struct gl_shader_program *shader_prog,
+ struct gl_program *prog)
+{
+ struct brw_context *brw = brw_context(ctx);
+ struct brw_tes_prog_key key;
+ uint32_t old_prog_offset = brw->tes.base.prog_offset;
+ struct brw_tes_prog_data *old_prog_data = brw->tes.prog_data;
+ bool success;
+
+ struct gl_tess_eval_program *tep = (struct gl_tess_eval_program *)prog;
+ struct brw_tess_eval_program *btep = brw_tess_eval_program(tep);
+
+ memset(&key, 0, sizeof(key));
+
+ key.program_string_id = btep->id;
+ key.inputs_read = prog->InputsRead;
+ key.patch_inputs_read = prog->PatchInputsRead;
+
+ if (shader_prog->_LinkedShaders[MESA_SHADER_TESS_CTRL]) {
+ struct gl_program *tcp =
+ shader_prog->_LinkedShaders[MESA_SHADER_TESS_CTRL]->Program;
+ key.inputs_read |= tcp->OutputsWritten;
+ key.patch_inputs_read |= tcp->PatchOutputsWritten;
+ }
+
+ /* Ignore gl_TessLevelInner/Outer - they're system values. */
+ key.inputs_read &= ~(VARYING_BIT_TESS_LEVEL_INNER |
+ VARYING_BIT_TESS_LEVEL_OUTER);
+
+ brw_setup_tex_for_precompile(brw, &key.tex, prog);
+
+ success = brw_codegen_tes_prog(brw, shader_prog, btep, &key);
+
+ brw->tes.base.prog_offset = old_prog_offset;
+ brw->tes.prog_data = old_prog_data;
+
+ return success;
+}
{
struct brw_stage_state *stage_state = &brw->tes.base;
- /* BRW_NEW_TESS_EVAL_PROGRAM */
+ /* BRW_NEW_TESS_PROGRAMS */
struct brw_tess_eval_program *dp =
(struct brw_tess_eval_program *) brw->tess_eval_program;
.mesa = _NEW_PROGRAM_CONSTANTS,
.brw = BRW_NEW_BATCH |
BRW_NEW_TES_PROG_DATA |
- BRW_NEW_TESS_EVAL_PROGRAM,
+ BRW_NEW_TESS_PROGRAMS,
},
.emit = brw_upload_tes_pull_constants,
};
brw_upload_tes_image_surfaces(struct brw_context *brw)
{
struct gl_context *ctx = &brw->ctx;
- /* BRW_NEW_TESS_EVAL_PROGRAM */
+ /* BRW_NEW_TESS_PROGRAMS */
struct gl_shader_program *prog =
ctx->_Shader->CurrentProgram[MESA_SHADER_TESS_EVAL];
.dirty = {
.brw = BRW_NEW_BATCH |
BRW_NEW_IMAGE_UNITS |
- BRW_NEW_TESS_EVAL_PROGRAM |
+ BRW_NEW_TESS_PROGRAMS |
BRW_NEW_TES_PROG_DATA,
},
.emit = brw_upload_tes_image_surfaces,
unreachable("not reached");
}
}
+
+static const GLuint prim_to_hw_prim[GL_TRIANGLE_STRIP_ADJACENCY+1] = {
+ [GL_POINTS] =_3DPRIM_POINTLIST,
+ [GL_LINES] = _3DPRIM_LINELIST,
+ [GL_LINE_LOOP] = _3DPRIM_LINELOOP,
+ [GL_LINE_STRIP] = _3DPRIM_LINESTRIP,
+ [GL_TRIANGLES] = _3DPRIM_TRILIST,
+ [GL_TRIANGLE_STRIP] = _3DPRIM_TRISTRIP,
+ [GL_TRIANGLE_FAN] = _3DPRIM_TRIFAN,
+ [GL_QUADS] = _3DPRIM_QUADLIST,
+ [GL_QUAD_STRIP] = _3DPRIM_QUADSTRIP,
+ [GL_POLYGON] = _3DPRIM_POLYGON,
+ [GL_LINES_ADJACENCY] = _3DPRIM_LINELIST_ADJ,
+ [GL_LINE_STRIP_ADJACENCY] = _3DPRIM_LINESTRIP_ADJ,
+ [GL_TRIANGLES_ADJACENCY] = _3DPRIM_TRILIST_ADJ,
+ [GL_TRIANGLE_STRIP_ADJACENCY] = _3DPRIM_TRISTRIP_ADJ,
+};
+
+uint32_t
+get_hw_prim_for_gl_prim(int mode)
+{
+ if (mode >= BRW_PRIM_OFFSET)
+ return mode - BRW_PRIM_OFFSET;
+ else {
+ assert(mode < ARRAY_SIZE(prim_to_hw_prim));
+ return prim_to_hw_prim[mode];
+ }
+}
case SHADER_OPCODE_TYPED_ATOMIC:
case SHADER_OPCODE_TYPED_SURFACE_READ:
case SHADER_OPCODE_TYPED_SURFACE_WRITE:
+ case VEC4_OPCODE_URB_READ:
+ case TCS_OPCODE_URB_WRITE:
+ case TCS_OPCODE_RELEASE_INPUT:
+ case SHADER_OPCODE_BARRIER:
return true;
default:
return false;
vec4_instruction::has_source_and_destination_hazard() const
{
switch (opcode) {
- /* Most opcodes in the vec4 world use MRFs. */
+ case TCS_OPCODE_SET_INPUT_URB_OFFSETS:
+ case TCS_OPCODE_SET_OUTPUT_URB_OFFSETS:
+ case TES_OPCODE_ADD_INDIRECT_URB_OFFSET:
+ return true;
default:
return false;
}
case SHADER_OPCODE_TYPED_ATOMIC:
case SHADER_OPCODE_TYPED_SURFACE_READ:
case SHADER_OPCODE_TYPED_SURFACE_WRITE:
+ case TCS_OPCODE_URB_WRITE:
return arg == 0 ? mlen : 1;
case VS_OPCODE_PULL_CONSTANT_LOAD_GEN7:
case SHADER_OPCODE_POW:
return 2;
case VS_OPCODE_URB_WRITE:
+ case TCS_OPCODE_THREAD_END:
return 1;
case VS_OPCODE_PULL_CONSTANT_LOAD:
return 2;
return 0;
case GS_OPCODE_FF_SYNC:
return 1;
+ case TCS_OPCODE_URB_WRITE:
+ return 0;
case SHADER_OPCODE_SHADER_TIME_ADD:
return 0;
case SHADER_OPCODE_TEX:
vec4_vs_visitor::setup_attributes(int payload_reg)
{
int nr_attributes;
- int attribute_map[VERT_ATTRIB_MAX + 1];
+ int attribute_map[VERT_ATTRIB_MAX + 2];
memset(attribute_map, 0, sizeof(attribute_map));
nr_attributes = 0;
}
}
+ if (vs_prog_data->uses_drawid) {
+ attribute_map[VERT_ATTRIB_MAX + 1] = payload_reg + nr_attributes;
+ nr_attributes++;
+ }
+
/* VertexID is stored by the VF as the last vertex element, but we
* don't represent it with a flag in inputs_read, so we call it
* VERT_ATTRIB_MAX.
*/
- if (vs_prog_data->uses_vertexid || vs_prog_data->uses_instanceid) {
+ if (vs_prog_data->uses_vertexid || vs_prog_data->uses_instanceid ||
+ vs_prog_data->uses_basevertex || vs_prog_data->uses_baseinstance) {
attribute_map[VERT_ATTRIB_MAX] = payload_reg + nr_attributes;
+ nr_attributes++;
}
lower_attributes_to_hw_regs(attribute_map, false /* interleaved */);
case ATTR:
unreachable("not reached");
}
+
src = reg;
}
+ if (inst->is_3src()) {
+ /* 3-src instructions with scalar sources support arbitrary subnr,
+ * but don't actually use swizzles. Convert swizzle into subnr.
+ */
+ for (int i = 0; i < 3; i++) {
+ if (inst->src[i].vstride == BRW_VERTICAL_STRIDE_0) {
+ assert(brw_is_single_value_swizzle(inst->src[i].swizzle));
+ inst->src[i].subnr += 4 * BRW_GET_SWZ(inst->src[i].swizzle, 0);
+ }
+ }
+ }
+
dst_reg &dst = inst->dst;
struct brw_reg reg;
* incoming vertex attribute. So, add an extra slot.
*/
if (shader->info.system_values_read &
- (BITFIELD64_BIT(SYSTEM_VALUE_VERTEX_ID_ZERO_BASE) |
+ (BITFIELD64_BIT(SYSTEM_VALUE_BASE_VERTEX) |
+ BITFIELD64_BIT(SYSTEM_VALUE_BASE_INSTANCE) |
+ BITFIELD64_BIT(SYSTEM_VALUE_VERTEX_ID_ZERO_BASE) |
BITFIELD64_BIT(SYSTEM_VALUE_INSTANCE_ID))) {
nr_attributes++;
}
+ /* gl_DrawID has its very own vec4 */
+ if (shader->info.system_values_read & BITFIELD64_BIT(SYSTEM_VALUE_DRAW_ID)) {
+ nr_attributes++;
+ }
+
/* The 3DSTATE_VS documentation lists the lower bound on "Vertex URB Entry
* Read Length" as 1 in vec4 mode, and 0 in SIMD8 mode. Empirically, in
* vec4 mode, the hardware appears to wedge unless we read something.
fs_generator g(compiler, log_data, mem_ctx, (void *) key,
&prog_data->base.base, v.promoted_constants,
- v.runtime_check_aads_emit, "VS");
+ v.runtime_check_aads_emit, MESA_SHADER_VERTEX);
if (INTEL_DEBUG & DEBUG_VS) {
const char *debug_name =
ralloc_asprintf(mem_ctx, "%s vertex shader %s",
src_reg offset_value,
src_reg mcs,
bool is_cube_array,
+ uint32_t surface, src_reg surface_reg,
uint32_t sampler, src_reg sampler_reg);
src_reg emit_mcs_fetch(const glsl_type *coordinate_type, src_reg coordinate,
bool is_high_sampler(src_reg sampler);
+ bool optimize_predicate(nir_alu_instr *instr, enum brw_predicate *predicate);
+
virtual void emit_nir_code();
virtual void nir_setup_uniforms();
virtual void nir_setup_system_value_intrinsic(nir_intrinsic_instr *instr);
unsigned num_components = 4);
src_reg get_nir_src(nir_src src,
unsigned num_components = 4);
+ src_reg get_indirect_offset(nir_intrinsic_instr *instr);
virtual dst_reg *make_reg_for_system_value(int location,
const glsl_type *type) = 0;
case VEC4_OPCODE_UNPACK_UNIFORM:
case SHADER_OPCODE_FIND_LIVE_CHANNEL:
case SHADER_OPCODE_BROADCAST:
+ case TCS_OPCODE_SET_INPUT_URB_OFFSETS:
+ case TCS_OPCODE_SET_OUTPUT_URB_OFFSETS:
return true;
case SHADER_OPCODE_RCP:
case SHADER_OPCODE_RSQ:
case VS_OPCODE_PULL_CONSTANT_LOAD:
case VS_OPCODE_PULL_CONSTANT_LOAD_GEN7:
case VS_OPCODE_SET_SIMD4X2_HEADER_GEN9:
+ case TCS_OPCODE_SET_INPUT_URB_OFFSETS:
+ case TCS_OPCODE_SET_OUTPUT_URB_OFFSETS:
+ case TES_OPCODE_CREATE_INPUT_READ_HEADER:
+ case TES_OPCODE_ADD_INDIRECT_URB_OFFSET:
+ case VEC4_OPCODE_URB_READ:
return false;
default:
/* The MATH instruction on Gen6 only executes in align1 mode, which does
* IN THE SOFTWARE.
*/
-#include "glsl/glsl_parser_extras.h"
#include "brw_vec4.h"
#include "brw_cfg.h"
#include "brw_eu.h"
vec4_instruction *inst,
struct brw_reg dst,
struct brw_reg src,
+ struct brw_reg surface_index,
struct brw_reg sampler_index)
{
const struct brw_device_info *devinfo = p->devinfo;
? prog_data->base.binding_table.gather_texture_start
: prog_data->base.binding_table.texture_start;
- if (sampler_index.file == BRW_IMMEDIATE_VALUE) {
+ if (surface_index.file == BRW_IMMEDIATE_VALUE &&
+ sampler_index.file == BRW_IMMEDIATE_VALUE) {
+ uint32_t surface = surface_index.ud;
uint32_t sampler = sampler_index.ud;
brw_SAMPLE(p,
dst,
inst->base_mrf,
src,
- sampler + base_binding_table_index,
+ surface + base_binding_table_index,
sampler % 16,
msg_type,
1, /* response length */
/* Non-constant sampler index. */
struct brw_reg addr = vec1(retype(brw_address_reg(0), BRW_REGISTER_TYPE_UD));
+ struct brw_reg surface_reg = vec1(retype(surface_index, BRW_REGISTER_TYPE_UD));
struct brw_reg sampler_reg = vec1(retype(sampler_index, BRW_REGISTER_TYPE_UD));
brw_push_insn_state(p);
brw_set_default_mask_control(p, BRW_MASK_DISABLE);
brw_set_default_access_mode(p, BRW_ALIGN_1);
- /* addr = ((sampler * 0x101) + base_binding_table_index) & 0xfff */
- brw_MUL(p, addr, sampler_reg, brw_imm_uw(0x101));
+ if (memcmp(&surface_reg, &sampler_reg, sizeof(surface_reg)) == 0) {
+ brw_MUL(p, addr, sampler_reg, brw_imm_uw(0x101));
+ } else {
+ brw_SHL(p, addr, sampler_reg, brw_imm_ud(8));
+ brw_OR(p, addr, addr, surface_reg);
+ }
if (base_binding_table_index)
brw_ADD(p, addr, addr, brw_imm_ud(base_binding_table_index));
brw_AND(p, addr, addr, brw_imm_ud(0xfff));
}
static void
+generate_tcs_get_instance_id(struct brw_codegen *p, struct brw_reg dst)
+{
+ const struct brw_device_info *devinfo = p->devinfo;
+ const bool ivb = devinfo->is_ivybridge || devinfo->is_baytrail;
+
+ /* "Instance Count" comes as part of the payload in r0.2 bits 23:17.
+ *
+ * Since we operate in SIMD4x2 mode, we need run half as many threads
+ * as necessary. So we assign (2i + 1, 2i) as the thread counts. We
+ * shift right by one less to accomplish the multiplication by two.
+ */
+ dst = retype(dst, BRW_REGISTER_TYPE_UD);
+ struct brw_reg r0(retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UD));
+
+ brw_push_insn_state(p);
+ brw_set_default_access_mode(p, BRW_ALIGN_1);
+
+ const int mask = ivb ? INTEL_MASK(22, 16) : INTEL_MASK(23, 17);
+ const int shift = ivb ? 16 : 17;
+
+ brw_AND(p, get_element_ud(dst, 0), get_element_ud(r0, 2), brw_imm_ud(mask));
+ brw_SHR(p, get_element_ud(dst, 0), get_element_ud(dst, 0),
+ brw_imm_ud(shift - 1));
+ brw_ADD(p, get_element_ud(dst, 4), get_element_ud(dst, 0), brw_imm_ud(1));
+
+ brw_pop_insn_state(p);
+}
+
+static void
+generate_tcs_urb_write(struct brw_codegen *p,
+ vec4_instruction *inst,
+ struct brw_reg urb_header)
+{
+ const struct brw_device_info *devinfo = p->devinfo;
+
+ brw_inst *send = brw_next_insn(p, BRW_OPCODE_SEND);
+ brw_set_dest(p, send, brw_null_reg());
+ brw_set_src0(p, send, urb_header);
+
+ brw_set_message_descriptor(p, send, BRW_SFID_URB,
+ inst->mlen /* mlen */, 0 /* rlen */,
+ true /* header */, false /* eot */);
+ brw_inst_set_urb_opcode(devinfo, send, BRW_URB_OPCODE_WRITE_OWORD);
+ brw_inst_set_urb_global_offset(devinfo, send, inst->offset);
+ if (inst->urb_write_flags & BRW_URB_WRITE_EOT) {
+ brw_inst_set_eot(devinfo, send, 1);
+ } else {
+ brw_inst_set_urb_per_slot_offset(devinfo, send, 1);
+ brw_inst_set_urb_swizzle_control(devinfo, send, BRW_URB_SWIZZLE_INTERLEAVE);
+ }
+
+ /* what happens to swizzles? */
+}
+
+
+static void
+generate_tcs_input_urb_offsets(struct brw_codegen *p,
+ struct brw_reg dst,
+ struct brw_reg vertex,
+ struct brw_reg offset)
+{
+ /* Generates an URB read/write message header for HS/DS operation.
+ * Inputs are a vertex index, and a byte offset from the beginning of
+ * the vertex. */
+
+ /* If `vertex` is not an immediate, we clobber a0.0 */
+
+ assert(vertex.file == BRW_IMMEDIATE_VALUE || vertex.file == BRW_GENERAL_REGISTER_FILE);
+ assert(vertex.type == BRW_REGISTER_TYPE_UD || vertex.type == BRW_REGISTER_TYPE_D);
+
+ assert(dst.file == BRW_GENERAL_REGISTER_FILE);
+
+ brw_push_insn_state(p);
+ brw_set_default_access_mode(p, BRW_ALIGN_1);
+ brw_set_default_mask_control(p, BRW_MASK_DISABLE);
+ brw_MOV(p, dst, brw_imm_ud(0));
+
+ /* m0.5 bits 8-15 are channel enables */
+ brw_MOV(p, get_element_ud(dst, 5), brw_imm_ud(0xff00));
+
+ /* m0.0-0.1: URB handles */
+ if (vertex.file == BRW_IMMEDIATE_VALUE) {
+ uint32_t vertex_index = vertex.ud;
+ struct brw_reg index_reg = brw_vec1_grf(
+ 1 + (vertex_index >> 3), vertex_index & 7);
+
+ brw_MOV(p, vec2(get_element_ud(dst, 0)),
+ retype(index_reg, BRW_REGISTER_TYPE_UD));
+ } else {
+ /* Use indirect addressing. ICP Handles are DWords (single channels
+ * of a register) and start at g1.0.
+ *
+ * In order to start our region at g1.0, we add 8 to the vertex index,
+ * effectively skipping over the 8 channels in g0.0. This gives us a
+ * DWord offset to the ICP Handle.
+ *
+ * Indirect addressing works in terms of bytes, so we then multiply
+ * the DWord offset by 4 (by shifting left by 2).
+ */
+ struct brw_reg addr = brw_address_reg(0);
+
+ /* bottom half: m0.0 = g[1.0 + vertex.0]UD */
+ brw_ADD(p, addr, get_element_ud(vertex, 0), brw_imm_uw(0x8));
+ brw_SHL(p, addr, addr, brw_imm_ud(2));
+ brw_MOV(p, get_element_ud(dst, 0), deref_1ud(brw_indirect(0, 0), 0));
+
+ /* top half: m0.1 = g[1.0 + vertex.4]UD */
+ brw_ADD(p, addr, get_element_ud(vertex, 4), brw_imm_uw(0x8));
+ brw_SHL(p, addr, addr, brw_imm_ud(2));
+ brw_MOV(p, get_element_ud(dst, 1), deref_1ud(brw_indirect(0, 0), 0));
+ }
+
+ /* m0.3-0.4: 128bit-granular offsets into the URB from the handles */
+ if (offset.file != ARF)
+ brw_MOV(p, vec2(get_element_ud(dst, 3)), stride(offset, 4, 1, 0));
+
+ brw_pop_insn_state(p);
+}
+
+
+static void
+generate_tcs_output_urb_offsets(struct brw_codegen *p,
+ struct brw_reg dst,
+ struct brw_reg write_mask,
+ struct brw_reg offset)
+{
+ /* Generates an URB read/write message header for HS/DS operation, for the patch URB entry. */
+ assert(dst.file == BRW_GENERAL_REGISTER_FILE || dst.file == BRW_MESSAGE_REGISTER_FILE);
+
+ assert(write_mask.file == BRW_IMMEDIATE_VALUE);
+ assert(write_mask.type == BRW_REGISTER_TYPE_UD);
+
+ brw_push_insn_state(p);
+
+ brw_set_default_access_mode(p, BRW_ALIGN_1);
+ brw_set_default_mask_control(p, BRW_MASK_DISABLE);
+ brw_MOV(p, dst, brw_imm_ud(0));
+
+ unsigned mask = write_mask.ud;
+
+ /* m0.5 bits 15:12 and 11:8 are channel enables */
+ brw_MOV(p, get_element_ud(dst, 5), brw_imm_ud((mask << 8) | (mask << 12)));
+
+ /* HS patch URB handle is delivered in r0.0 */
+ struct brw_reg urb_handle = brw_vec1_grf(0, 0);
+
+ /* m0.0-0.1: URB handles */
+ brw_MOV(p, vec2(get_element_ud(dst, 0)),
+ retype(urb_handle, BRW_REGISTER_TYPE_UD));
+
+ /* m0.3-0.4: 128bit-granular offsets into the URB from the handles */
+ if (offset.file != ARF)
+ brw_MOV(p, vec2(get_element_ud(dst, 3)), stride(offset, 4, 1, 0));
+
+ brw_pop_insn_state(p);
+}
+
+static void
+generate_tes_create_input_read_header(struct brw_codegen *p,
+ struct brw_reg dst)
+{
+ brw_push_insn_state(p);
+ brw_set_default_access_mode(p, BRW_ALIGN_1);
+ brw_set_default_mask_control(p, BRW_MASK_DISABLE);
+
+ /* Initialize the register to 0 */
+ brw_MOV(p, dst, brw_imm_ud(0));
+
+ /* Enable all the channels in m0.5 bits 15:8 */
+ brw_MOV(p, get_element_ud(dst, 5), brw_imm_ud(0xff00));
+
+ /* Copy g1.3 (the patch URB handle) to m0.0 and m0.1. For safety,
+ * mask out irrelevant "Reserved" bits, as they're not marked MBZ.
+ */
+ brw_AND(p, vec2(get_element_ud(dst, 0)),
+ retype(brw_vec1_grf(1, 3), BRW_REGISTER_TYPE_UD),
+ brw_imm_ud(0x1fff));
+ brw_pop_insn_state(p);
+}
+
+static void
+generate_tes_add_indirect_urb_offset(struct brw_codegen *p,
+ struct brw_reg dst,
+ struct brw_reg header,
+ struct brw_reg offset)
+{
+ brw_push_insn_state(p);
+ brw_set_default_access_mode(p, BRW_ALIGN_1);
+ brw_set_default_mask_control(p, BRW_MASK_DISABLE);
+
+ brw_MOV(p, dst, header);
+ /* m0.3-0.4: 128-bit-granular offsets into the URB from the handles */
+ brw_MOV(p, vec2(get_element_ud(dst, 3)), stride(offset, 4, 1, 0));
+
+ brw_pop_insn_state(p);
+}
+
+static void
+generate_vec4_urb_read(struct brw_codegen *p,
+ vec4_instruction *inst,
+ struct brw_reg dst,
+ struct brw_reg header)
+{
+ const struct brw_device_info *devinfo = p->devinfo;
+
+ assert(header.file == BRW_GENERAL_REGISTER_FILE);
+ assert(header.type == BRW_REGISTER_TYPE_UD);
+
+ brw_inst *send = brw_next_insn(p, BRW_OPCODE_SEND);
+ brw_set_dest(p, send, dst);
+ brw_set_src0(p, send, header);
+
+ brw_set_message_descriptor(p, send, BRW_SFID_URB,
+ 1 /* mlen */, 1 /* rlen */,
+ true /* header */, false /* eot */);
+ brw_inst_set_urb_opcode(devinfo, send, BRW_URB_OPCODE_READ_OWORD);
+ brw_inst_set_urb_swizzle_control(devinfo, send, BRW_URB_SWIZZLE_INTERLEAVE);
+ brw_inst_set_urb_per_slot_offset(devinfo, send, 1);
+
+ brw_inst_set_urb_global_offset(devinfo, send, inst->offset);
+}
+
+static void
+generate_tcs_release_input(struct brw_codegen *p,
+ struct brw_reg header,
+ struct brw_reg vertex,
+ struct brw_reg is_unpaired)
+{
+ const struct brw_device_info *devinfo = p->devinfo;
+
+ assert(vertex.file == BRW_IMMEDIATE_VALUE);
+ assert(vertex.type == BRW_REGISTER_TYPE_UD);
+
+ /* m0.0-0.1: URB handles */
+ struct brw_reg urb_handles =
+ retype(brw_vec2_grf(1 + (vertex.ud >> 3), vertex.ud & 7),
+ BRW_REGISTER_TYPE_UD);
+
+ brw_push_insn_state(p);
+ brw_set_default_access_mode(p, BRW_ALIGN_1);
+ brw_set_default_mask_control(p, BRW_MASK_DISABLE);
+ brw_MOV(p, header, brw_imm_ud(0));
+ brw_MOV(p, vec2(get_element_ud(header, 0)), urb_handles);
+ brw_pop_insn_state(p);
+
+ brw_inst *send = brw_next_insn(p, BRW_OPCODE_SEND);
+ brw_set_dest(p, send, brw_null_reg());
+ brw_set_src0(p, send, header);
+ brw_set_message_descriptor(p, send, BRW_SFID_URB,
+ 1 /* mlen */, 0 /* rlen */,
+ true /* header */, false /* eot */);
+ brw_inst_set_urb_opcode(devinfo, send, BRW_URB_OPCODE_READ_OWORD);
+ brw_inst_set_urb_complete(devinfo, send, 1);
+ brw_inst_set_urb_swizzle_control(devinfo, send, is_unpaired.ud ?
+ BRW_URB_SWIZZLE_NONE :
+ BRW_URB_SWIZZLE_INTERLEAVE);
+}
+
+static void
+generate_tcs_thread_end(struct brw_codegen *p, vec4_instruction *inst)
+{
+ struct brw_reg header = brw_message_reg(inst->base_mrf);
+
+ brw_push_insn_state(p);
+ brw_set_default_access_mode(p, BRW_ALIGN_1);
+ brw_set_default_mask_control(p, BRW_MASK_DISABLE);
+ brw_MOV(p, header, brw_imm_ud(0));
+ brw_MOV(p, get_element_ud(header, 0),
+ retype(brw_vec1_grf(0, 0), BRW_REGISTER_TYPE_UD));
+ brw_pop_insn_state(p);
+
+ brw_urb_WRITE(p,
+ brw_null_reg(), /* dest */
+ inst->base_mrf, /* starting mrf reg nr */
+ header,
+ BRW_URB_WRITE_EOT | inst->urb_write_flags,
+ inst->mlen,
+ 0, /* response len */
+ 0, /* urb destination offset */
+ 0);
+}
+
+static void
+generate_tes_get_primitive_id(struct brw_codegen *p, struct brw_reg dst)
+{
+ brw_push_insn_state(p);
+ brw_set_default_access_mode(p, BRW_ALIGN_1);
+ brw_MOV(p, dst, retype(brw_vec1_grf(1, 7), BRW_REGISTER_TYPE_D));
+ brw_pop_insn_state(p);
+}
+
+static void
+generate_tcs_get_primitive_id(struct brw_codegen *p, struct brw_reg dst)
+{
+ brw_push_insn_state(p);
+ brw_set_default_access_mode(p, BRW_ALIGN_1);
+ brw_MOV(p, dst, retype(brw_vec1_grf(0, 1), BRW_REGISTER_TYPE_UD));
+ brw_pop_insn_state(p);
+}
+
+static void
+generate_tcs_create_barrier_header(struct brw_codegen *p,
+ struct brw_vue_prog_data *prog_data,
+ struct brw_reg dst)
+{
+ const struct brw_device_info *devinfo = p->devinfo;
+ const bool ivb = devinfo->is_ivybridge || devinfo->is_baytrail;
+ struct brw_reg m0_2 = get_element_ud(dst, 2);
+ unsigned instances = ((struct brw_tcs_prog_data *) prog_data)->instances;
+
+ brw_push_insn_state(p);
+ brw_set_default_access_mode(p, BRW_ALIGN_1);
+ brw_set_default_mask_control(p, BRW_MASK_DISABLE);
+
+ /* Zero the message header */
+ brw_MOV(p, retype(dst, BRW_REGISTER_TYPE_UD), brw_imm_ud(0u));
+
+ /* Copy "Barrier ID" from r0.2, bits 16:13 (Gen7.5+) or 15:12 (Gen7) */
+ brw_AND(p, m0_2,
+ retype(brw_vec1_grf(0, 2), BRW_REGISTER_TYPE_UD),
+ brw_imm_ud(ivb ? INTEL_MASK(15, 12) : INTEL_MASK(16, 13)));
+
+ /* Shift it up to bits 27:24. */
+ brw_SHL(p, m0_2, get_element_ud(dst, 2), brw_imm_ud(ivb ? 12 : 11));
+
+ /* Set the Barrier Count and the enable bit */
+ brw_OR(p, m0_2, m0_2, brw_imm_ud(instances << 9 | (1 << 15)));
+
+ brw_pop_insn_state(p);
+}
+
+static void
generate_oword_dual_block_offsets(struct brw_codegen *p,
struct brw_reg m1,
struct brw_reg index)
case SHADER_OPCODE_TG4:
case SHADER_OPCODE_TG4_OFFSET:
case SHADER_OPCODE_SAMPLEINFO:
- generate_tex(p, prog_data, inst, dst, src[0], src[1]);
+ generate_tex(p, prog_data, inst, dst, src[0], src[1], src[2]);
break;
case VS_OPCODE_URB_WRITE:
break;
}
+ case TCS_OPCODE_URB_WRITE:
+ generate_tcs_urb_write(p, inst, src[0]);
+ break;
+
+ case VEC4_OPCODE_URB_READ:
+ generate_vec4_urb_read(p, inst, dst, src[0]);
+ break;
+
+ case TCS_OPCODE_SET_INPUT_URB_OFFSETS:
+ generate_tcs_input_urb_offsets(p, dst, src[0], src[1]);
+ break;
+
+ case TCS_OPCODE_SET_OUTPUT_URB_OFFSETS:
+ generate_tcs_output_urb_offsets(p, dst, src[0], src[1]);
+ break;
+
+ case TCS_OPCODE_GET_INSTANCE_ID:
+ generate_tcs_get_instance_id(p, dst);
+ break;
+
+ case TCS_OPCODE_GET_PRIMITIVE_ID:
+ generate_tcs_get_primitive_id(p, dst);
+ break;
+
+ case TCS_OPCODE_CREATE_BARRIER_HEADER:
+ generate_tcs_create_barrier_header(p, prog_data, dst);
+ break;
+
+ case TES_OPCODE_CREATE_INPUT_READ_HEADER:
+ generate_tes_create_input_read_header(p, dst);
+ break;
+
+ case TES_OPCODE_ADD_INDIRECT_URB_OFFSET:
+ generate_tes_add_indirect_urb_offset(p, dst, src[0], src[1]);
+ break;
+
+ case TES_OPCODE_GET_PRIMITIVE_ID:
+ generate_tes_get_primitive_id(p, dst);
+ break;
+
+ case TCS_OPCODE_SRC0_010_IS_ZERO:
+ /* If src_reg had stride like fs_reg, we wouldn't need this. */
+ brw_MOV(p, brw_null_reg(), stride(src[0], 0, 1, 0));
+ brw_inst_set_cond_modifier(devinfo, brw_last_inst, BRW_CONDITIONAL_Z);
+ break;
+
+ case TCS_OPCODE_RELEASE_INPUT:
+ generate_tcs_release_input(p, dst, src[0], src[1]);
+ break;
+
+ case TCS_OPCODE_THREAD_END:
+ generate_tcs_thread_end(p, inst);
+ break;
+
+ case SHADER_OPCODE_BARRIER:
+ brw_barrier(p, src[0]);
+ brw_WAIT(p);
+ break;
+
case SHADER_OPCODE_MOV_INDIRECT:
generate_mov_indirect(p, inst, dst, src[0], src[1], src[2]);
compiler->shader_debug_log(log_data,
"%s vec4 shader: %d inst, %d loops, %u cycles, "
- "compacted %d to %d bytes.\n",
+ "compacted %d to %d bytes.",
stage_abbrev, before_size / 16,
loop_count, cfg->cycle_count,
before_size, after_size);
fs_generator g(compiler, log_data, mem_ctx, &c.key,
&prog_data->base.base, v.promoted_constants,
- false, "GS");
+ false, MESA_SHADER_GEOMETRY);
if (unlikely(INTEL_DEBUG & DEBUG_GS)) {
const char *label =
shader->info.label ? shader->info.label : "unnamed";
nir_setup_system_values();
/* get the main function and emit it */
- nir_foreach_overload(nir, overload) {
- assert(strcmp(overload->function->name, "main") == 0);
- assert(overload->impl);
- nir_emit_impl(overload->impl);
+ nir_foreach_function(nir, function) {
+ assert(strcmp(function->name, "main") == 0);
+ assert(function->impl);
+ nir_emit_impl(function->impl);
}
}
glsl_type::int_type);
break;
+ case nir_intrinsic_load_base_instance:
+ reg = &nir_system_values[SYSTEM_VALUE_BASE_INSTANCE];
+ if (reg->file == BAD_FILE)
+ *reg = *make_reg_for_system_value(SYSTEM_VALUE_BASE_INSTANCE,
+ glsl_type::int_type);
+ break;
+
+ case nir_intrinsic_load_draw_id:
+ reg = &nir_system_values[SYSTEM_VALUE_DRAW_ID];
+ if (reg->file == BAD_FILE)
+ *reg = *make_reg_for_system_value(SYSTEM_VALUE_DRAW_ID,
+ glsl_type::int_type);
+ break;
+
default:
break;
}
nir_system_values[i] = dst_reg();
}
- nir_foreach_overload(nir, overload) {
- assert(strcmp(overload->function->name, "main") == 0);
- assert(overload->impl);
- nir_foreach_block(overload->impl, setup_system_values_block, this);
+ nir_foreach_function(nir, function) {
+ assert(strcmp(function->name, "main") == 0);
+ assert(function->impl);
+ nir_foreach_block(function->impl, setup_system_values_block, this);
}
}
return get_nir_src(src, nir_type_int, num_components);
}
+src_reg
+vec4_visitor::get_indirect_offset(nir_intrinsic_instr *instr)
+{
+ nir_src *offset_src = nir_get_io_offset_src(instr);
+ nir_const_value *const_value = nir_src_as_const_value(*offset_src);
+
+ if (const_value) {
+ /* The only constant offset we should find is 0. brw_nir.c's
+ * add_const_offset_to_base() will fold other constant offsets
+ * into instr->const_index[0].
+ */
+ assert(const_value->u[0] == 0);
+ return src_reg();
+ }
+
+ return get_nir_src(*offset_src, BRW_REGISTER_TYPE_UD, 1);
+}
+
void
vec4_visitor::nir_emit_load_const(nir_load_const_instr *instr)
{
case nir_intrinsic_load_vertex_id_zero_base:
case nir_intrinsic_load_base_vertex:
- case nir_intrinsic_load_instance_id: {
+ case nir_intrinsic_load_instance_id:
+ case nir_intrinsic_load_base_instance:
+ case nir_intrinsic_load_draw_id:
+ case nir_intrinsic_load_invocation_id:
+ case nir_intrinsic_load_tess_level_inner:
+ case nir_intrinsic_load_tess_level_outer: {
gl_system_value sv = nir_system_value_from_intrinsic(instr->intrinsic);
src_reg val = src_reg(nir_system_values[sv]);
assert(val.file != BAD_FILE);
}
}
+bool
+vec4_visitor::optimize_predicate(nir_alu_instr *instr,
+ enum brw_predicate *predicate)
+{
+ if (!instr->src[0].src.is_ssa ||
+ instr->src[0].src.ssa->parent_instr->type != nir_instr_type_alu)
+ return false;
+
+ nir_alu_instr *cmp_instr =
+ nir_instr_as_alu(instr->src[0].src.ssa->parent_instr);
+
+ switch (cmp_instr->op) {
+ case nir_op_bany_fnequal2:
+ case nir_op_bany_inequal2:
+ case nir_op_bany_fnequal3:
+ case nir_op_bany_inequal3:
+ case nir_op_bany_fnequal4:
+ case nir_op_bany_inequal4:
+ *predicate = BRW_PREDICATE_ALIGN16_ANY4H;
+ break;
+ case nir_op_ball_fequal2:
+ case nir_op_ball_iequal2:
+ case nir_op_ball_fequal3:
+ case nir_op_ball_iequal3:
+ case nir_op_ball_fequal4:
+ case nir_op_ball_iequal4:
+ *predicate = BRW_PREDICATE_ALIGN16_ALL4H;
+ break;
+ default:
+ return false;
+ }
+
+ unsigned size_swizzle =
+ brw_swizzle_for_size(nir_op_infos[cmp_instr->op].input_sizes[0]);
+
+ src_reg op[2];
+ assert(nir_op_infos[cmp_instr->op].num_inputs == 2);
+ for (unsigned i = 0; i < 2; i++) {
+ op[i] = get_nir_src(cmp_instr->src[i].src,
+ nir_op_infos[cmp_instr->op].input_types[i], 4);
+ unsigned base_swizzle =
+ brw_swizzle_for_nir_swizzle(cmp_instr->src[i].swizzle);
+ op[i].swizzle = brw_compose_swizzle(size_swizzle, base_swizzle);
+ op[i].abs = cmp_instr->src[i].abs;
+ op[i].negate = cmp_instr->src[i].negate;
+ }
+
+ emit(CMP(dst_null_d(), op[0], op[1],
+ brw_conditional_for_nir_comparison(cmp_instr->op)));
+
+ return true;
+}
+
void
vec4_visitor::nir_emit_alu(nir_alu_instr *instr)
{
case nir_op_umul_high: {
struct brw_reg acc = retype(brw_acc_reg(8), dst.type);
- emit(MUL(acc, op[0], op[1]));
+ if (devinfo->gen >=8)
+ emit(MUL(acc, op[0], retype(op[1], BRW_REGISTER_TYPE_UW)));
+ else
+ emit(MUL(acc, op[0], op[1]));
+
emit(MACH(dst, op[0], op[1]));
break;
}
break;
case nir_op_umod:
+ case nir_op_irem:
+ /* According to the sign table for INT DIV in the Ivy Bridge PRM, it
+ * appears that our hardware just does the right thing for signed
+ * remainder.
+ */
emit_math(SHADER_OPCODE_INT_REMAINDER, dst, op[0], op[1]);
break;
+ case nir_op_imod: {
+ /* Get a regular C-style remainder. If a % b == 0, set the predicate. */
+ inst = emit_math(SHADER_OPCODE_INT_REMAINDER, dst, op[0], op[1]);
+
+ /* Math instructions don't support conditional mod */
+ inst = emit(MOV(dst_null_d(), src_reg(dst)));
+ inst->conditional_mod = BRW_CONDITIONAL_NZ;
+
+ /* Now, we need to determine if signs of the sources are different.
+ * When we XOR the sources, the top bit is 0 if they are the same and 1
+ * if they are different. We can then use a conditional modifier to
+ * turn that into a predicate. This leads us to an XOR.l instruction.
+ */
+ src_reg tmp = src_reg(this, glsl_type::ivec4_type);
+ inst = emit(XOR(dst_reg(tmp), op[0], op[1]));
+ inst->predicate = BRW_PREDICATE_NORMAL;
+ inst->conditional_mod = BRW_CONDITIONAL_L;
+
+ /* If the result of the initial remainder operation is non-zero and the
+ * two sources have different signs, add in a copy of op[1] to get the
+ * final integer modulus value.
+ */
+ inst = emit(ADD(dst, src_reg(dst), op[1]));
+ inst->predicate = BRW_PREDICATE_NORMAL;
+ break;
+ }
+
case nir_op_ldexp:
unreachable("not reached: should be handled by ldexp_to_arith()");
inst->saturate = instr->dest.saturate;
break;
+ case nir_op_fquantize2f16: {
+ /* See also vec4_visitor::emit_pack_half_2x16() */
+ src_reg tmp = src_reg(this, glsl_type::uvec4_type);
+
+ emit(F32TO16(dst_reg(tmp), op[0]));
+ inst = emit(F16TO32(dst, tmp));
+ inst->saturate = instr->dest.saturate;
+ break;
+ }
+
case nir_op_fmin:
case nir_op_imin:
case nir_op_umin:
case nir_op_ubitfield_extract:
case nir_op_ibitfield_extract:
+ unreachable("should have been lowered");
+ case nir_op_ubfe:
+ case nir_op_ibfe:
op[0] = fix_3src_operand(op[0]);
op[1] = fix_3src_operand(op[1]);
op[2] = fix_3src_operand(op[2]);
break;
case nir_op_bitfield_insert:
- unreachable("not reached: should be handled by "
- "lower_instructions::bitfield_insert_to_bfm_bfi");
+ unreachable("not reached: should have been lowered");
case nir_op_fsign:
/* AND(val, 0x80000000) gives the sign bit.
break;
case nir_op_bcsel:
- emit(CMP(dst_null_d(), op[0], brw_imm_d(0), BRW_CONDITIONAL_NZ));
- inst = emit(BRW_OPCODE_SEL, dst, op[1], op[2]);
- switch (dst.writemask) {
- case WRITEMASK_X:
- inst->predicate = BRW_PREDICATE_ALIGN16_REPLICATE_X;
- break;
- case WRITEMASK_Y:
- inst->predicate = BRW_PREDICATE_ALIGN16_REPLICATE_Y;
- break;
- case WRITEMASK_Z:
- inst->predicate = BRW_PREDICATE_ALIGN16_REPLICATE_Z;
- break;
- case WRITEMASK_W:
- inst->predicate = BRW_PREDICATE_ALIGN16_REPLICATE_W;
- break;
- default:
- inst->predicate = BRW_PREDICATE_NORMAL;
- break;
+ enum brw_predicate predicate;
+ if (!optimize_predicate(instr, &predicate)) {
+ emit(CMP(dst_null_d(), op[0], brw_imm_d(0), BRW_CONDITIONAL_NZ));
+ switch (dst.writemask) {
+ case WRITEMASK_X:
+ predicate = BRW_PREDICATE_ALIGN16_REPLICATE_X;
+ break;
+ case WRITEMASK_Y:
+ predicate = BRW_PREDICATE_ALIGN16_REPLICATE_Y;
+ break;
+ case WRITEMASK_Z:
+ predicate = BRW_PREDICATE_ALIGN16_REPLICATE_Z;
+ break;
+ case WRITEMASK_W:
+ predicate = BRW_PREDICATE_ALIGN16_REPLICATE_W;
+ break;
+ default:
+ predicate = BRW_PREDICATE_NORMAL;
+ break;
+ }
}
+ inst = emit(BRW_OPCODE_SEL, dst, op[1], op[2]);
+ inst->predicate = predicate;
break;
case nir_op_fdot_replicated2:
inst->saturate = instr->dest.saturate;
break;
- case nir_op_bany2:
- case nir_op_bany3:
- case nir_op_bany4: {
- unsigned swiz =
- brw_swizzle_for_size(nir_op_infos[instr->op].input_sizes[0]);
-
- emit(CMP(dst_null_d(), swizzle(op[0], swiz), brw_imm_d(0),
- BRW_CONDITIONAL_NZ));
- emit(MOV(dst, brw_imm_d(0)));
- inst = emit(MOV(dst, brw_imm_d(~0)));
- inst->predicate = BRW_PREDICATE_ALIGN16_ANY4H;
- break;
- }
-
case nir_op_fabs:
case nir_op_iabs:
case nir_op_fneg:
break;
case nir_jump_return:
- /* fall through */
default:
unreachable("unknown jump");
}
void
vec4_visitor::nir_emit_texture(nir_tex_instr *instr)
{
+ unsigned texture = instr->texture_index;
unsigned sampler = instr->sampler_index;
+ src_reg texture_reg = brw_imm_ud(texture);
src_reg sampler_reg = brw_imm_ud(sampler);
src_reg coordinate;
const glsl_type *coord_type = NULL;
nir_tex_instr_dest_size(instr));
dst_reg dest = get_nir_dest(instr->dest, instr->dest_type);
+ /* Our hardware requires a LOD for buffer textures */
+ if (instr->sampler_dim == GLSL_SAMPLER_DIM_BUF)
+ lod = brw_imm_d(0);
+
/* Load the texture operation sources */
for (unsigned i = 0; i < instr->num_srcs; i++) {
switch (instr->src[i].src_type) {
offset_value = get_nir_src(instr->src[i].src, BRW_REGISTER_TYPE_D, 2);
break;
- case nir_tex_src_sampler_offset: {
- /* The highest sampler which may be used by this operation is
+ case nir_tex_src_texture_offset: {
+ /* The highest texture which may be used by this operation is
* the last element of the array. Mark it here, because the generator
* doesn't have enough information to determine the bound.
*/
- uint32_t array_size = instr->sampler_array_size;
- uint32_t max_used = sampler + array_size - 1;
+ uint32_t max_used = texture + instr->texture_array_size - 1;
if (instr->op == nir_texop_tg4) {
max_used += prog_data->base.binding_table.gather_texture_start;
} else {
/* Emit code to evaluate the actual indexing expression */
src_reg src = get_nir_src(instr->src[i].src, 1);
src_reg temp(this, glsl_type::uint_type);
+ emit(ADD(dst_reg(temp), src, brw_imm_ud(texture)));
+ texture_reg = emit_uniformize(temp);
+ break;
+ }
+
+ case nir_tex_src_sampler_offset: {
+ /* Emit code to evaluate the actual indexing expression */
+ src_reg src = get_nir_src(instr->src[i].src, 1);
+ src_reg temp(this, glsl_type::uint_type);
emit(ADD(dst_reg(temp), src, brw_imm_ud(sampler)));
sampler_reg = emit_uniformize(temp);
break;
/* Stuff the channel select bits in the top of the texture offset */
if (instr->op == nir_texop_tg4) {
if (instr->component == 1 &&
- (key_tex->gather_channel_quirk_mask & (1 << sampler))) {
+ (key_tex->gather_channel_quirk_mask & (1 << texture))) {
/* gather4 sampler is broken for green channel on RG32F --
* we must ask for blue instead.
*/
shadow_comparitor,
lod, lod2, sample_index,
constant_offset, offset_value,
- mcs, is_cube_array, sampler, sampler_reg);
+ mcs, is_cube_array,
+ texture, texture_reg, sampler, sampler_reg);
}
void
--- /dev/null
+/*
+ * Copyright © 2013 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file brw_vec4_tcs.cpp
+ *
+ * Tessellaton control shader specific code derived from the vec4_visitor class.
+ */
+
+#include "brw_nir.h"
+#include "brw_vec4_tcs.h"
+
+namespace brw {
+
+vec4_tcs_visitor::vec4_tcs_visitor(const struct brw_compiler *compiler,
+ void *log_data,
+ const struct brw_tcs_prog_key *key,
+ struct brw_tcs_prog_data *prog_data,
+ const nir_shader *nir,
+ void *mem_ctx,
+ int shader_time_index,
+ const struct brw_vue_map *input_vue_map)
+ : vec4_visitor(compiler, log_data, &key->tex, &prog_data->base,
+ nir, mem_ctx, false, shader_time_index),
+ input_vue_map(input_vue_map), key(key)
+{
+}
+
+
+void
+vec4_tcs_visitor::emit_nir_code()
+{
+ if (key->program_string_id != 0) {
+ /* We have a real application-supplied TCS, emit real code. */
+ vec4_visitor::emit_nir_code();
+ } else {
+ /* There is no TCS; automatically generate a passthrough shader
+ * that writes the API-specified default tessellation levels and
+ * copies VS outputs to TES inputs.
+ */
+ uniforms = 2;
+
+ uint64_t varyings = key->outputs_written;
+
+ src_reg vertex_offset(this, glsl_type::uint_type);
+ emit(MUL(dst_reg(vertex_offset), invocation_id,
+ brw_imm_ud(prog_data->vue_map.num_per_vertex_slots)));
+
+ while (varyings != 0) {
+ const int varying = ffsll(varyings) - 1;
+
+ unsigned in_offset = input_vue_map->varying_to_slot[varying];
+ unsigned out_offset = prog_data->vue_map.varying_to_slot[varying];
+ assert(out_offset >= 2);
+
+ dst_reg val(this, glsl_type::vec4_type);
+ emit_input_urb_read(val, invocation_id, in_offset, src_reg());
+ emit_urb_write(src_reg(val), WRITEMASK_XYZW, out_offset,
+ vertex_offset);
+
+ varyings &= ~BITFIELD64_BIT(varying);
+ }
+
+ /* Only write the tessellation factors from invocation 0.
+ * There's no point in making other threads do redundant work.
+ */
+ emit(CMP(dst_null_d(), invocation_id, brw_imm_ud(0),
+ BRW_CONDITIONAL_EQ));
+ emit(IF(BRW_PREDICATE_NORMAL));
+ emit_urb_write(src_reg(UNIFORM, 0, glsl_type::vec4_type),
+ WRITEMASK_XYZW, 0, src_reg());
+ emit_urb_write(src_reg(UNIFORM, 1, glsl_type::vec4_type),
+ WRITEMASK_XYZW, 1, src_reg());
+ emit(BRW_OPCODE_ENDIF);
+ }
+}
+
+void
+vec4_tcs_visitor::nir_setup_system_value_intrinsic(nir_intrinsic_instr *instr)
+{
+}
+
+dst_reg *
+vec4_tcs_visitor::make_reg_for_system_value(int location, const glsl_type *type)
+{
+ return NULL;
+}
+
+
+void
+vec4_tcs_visitor::setup_payload()
+{
+ int reg = 0;
+
+ /* The payload always contains important data in r0, which contains
+ * the URB handles that are passed on to the URB write at the end
+ * of the thread.
+ */
+ reg++;
+
+ /* r1.0 - r4.7 may contain the input control point URB handles,
+ * which we use to pull vertex data.
+ */
+ reg += 4;
+
+ /* Push constants may start at r5.0 */
+ reg = setup_uniforms(reg);
+
+ this->first_non_payload_grf = reg;
+}
+
+
+void
+vec4_tcs_visitor::emit_prolog()
+{
+ invocation_id = src_reg(this, glsl_type::uint_type);
+ emit(TCS_OPCODE_GET_INSTANCE_ID, dst_reg(invocation_id));
+
+ /* HS threads are dispatched with the dispatch mask set to 0xFF.
+ * If there are an odd number of output vertices, then the final
+ * HS instance dispatched will only have its bottom half doing real
+ * work, and so we need to disable the upper half:
+ */
+ if (nir->info.tcs.vertices_out % 2) {
+ emit(CMP(dst_null_d(), invocation_id,
+ brw_imm_ud(nir->info.tcs.vertices_out), BRW_CONDITIONAL_L));
+
+ /* Matching ENDIF is in emit_thread_end() */
+ emit(IF(BRW_PREDICATE_NORMAL));
+ }
+}
+
+
+void
+vec4_tcs_visitor::emit_thread_end()
+{
+ vec4_instruction *inst;
+ current_annotation = "thread end";
+
+ if (nir->info.tcs.vertices_out % 2) {
+ emit(BRW_OPCODE_ENDIF);
+ }
+
+ if (devinfo->gen == 7) {
+ struct brw_tcs_prog_data *tcs_prog_data =
+ (struct brw_tcs_prog_data *) prog_data;
+
+ current_annotation = "release input vertices";
+
+ /* Synchronize all threads, so we know that no one is still
+ * using the input URB handles.
+ */
+ if (tcs_prog_data->instances > 1) {
+ dst_reg header = dst_reg(this, glsl_type::uvec4_type);
+ emit(TCS_OPCODE_CREATE_BARRIER_HEADER, header);
+ emit(SHADER_OPCODE_BARRIER, dst_null_ud(), src_reg(header));
+ }
+
+ /* Make thread 0 (invocations <1, 0>) release pairs of ICP handles.
+ * We want to compare the bottom half of invocation_id with 0, but
+ * use that truth value for the top half as well. Unfortunately,
+ * we don't have stride in the vec4 world, nor UV immediates in
+ * align16, so we need an opcode to get invocation_id<0,4,0>.
+ */
+ emit(TCS_OPCODE_SRC0_010_IS_ZERO, dst_null_d(), invocation_id);
+ emit(IF(BRW_PREDICATE_NORMAL));
+ for (unsigned i = 0; i < key->input_vertices; i += 2) {
+ /* If we have an odd number of input vertices, the last will be
+ * unpaired. We don't want to use an interleaved URB write in
+ * that case.
+ */
+ const bool is_unpaired = i == key->input_vertices - 1;
+
+ dst_reg header(this, glsl_type::uvec4_type);
+ emit(TCS_OPCODE_RELEASE_INPUT, header, brw_imm_ud(i),
+ brw_imm_ud(is_unpaired));
+ }
+ emit(BRW_OPCODE_ENDIF);
+ }
+
+ if (unlikely(INTEL_DEBUG & DEBUG_SHADER_TIME))
+ emit_shader_time_end();
+
+ inst = emit(TCS_OPCODE_THREAD_END);
+ inst->base_mrf = 14;
+ inst->mlen = 1;
+}
+
+
+void
+vec4_tcs_visitor::emit_input_urb_read(const dst_reg &dst,
+ const src_reg &vertex_index,
+ unsigned base_offset,
+ const src_reg &indirect_offset)
+{
+ vec4_instruction *inst;
+ dst_reg temp(this, glsl_type::ivec4_type);
+ temp.type = dst.type;
+
+ /* Set up the message header to reference the proper parts of the URB */
+ dst_reg header = dst_reg(this, glsl_type::uvec4_type);
+ inst = emit(TCS_OPCODE_SET_INPUT_URB_OFFSETS, header, vertex_index,
+ indirect_offset);
+ inst->force_writemask_all = true;
+
+ /* Read into a temporary, ignoring writemasking. */
+ inst = emit(VEC4_OPCODE_URB_READ, temp, src_reg(header));
+ inst->offset = base_offset;
+ inst->mlen = 1;
+ inst->base_mrf = -1;
+
+ /* Copy the temporary to the destination to deal with writemasking.
+ *
+ * Also attempt to deal with gl_PointSize being in the .w component.
+ */
+ if (inst->offset == 0 && indirect_offset.file == BAD_FILE) {
+ emit(MOV(dst, swizzle(src_reg(temp), BRW_SWIZZLE_WWWW)));
+ } else {
+ emit(MOV(dst, src_reg(temp)));
+ }
+}
+
+void
+vec4_tcs_visitor::emit_output_urb_read(const dst_reg &dst,
+ unsigned base_offset,
+ const src_reg &indirect_offset)
+{
+ vec4_instruction *inst;
+
+ /* Set up the message header to reference the proper parts of the URB */
+ dst_reg header = dst_reg(this, glsl_type::uvec4_type);
+ inst = emit(TCS_OPCODE_SET_OUTPUT_URB_OFFSETS, header,
+ brw_imm_ud(dst.writemask), indirect_offset);
+ inst->force_writemask_all = true;
+
+ /* Read into a temporary, ignoring writemasking. */
+ vec4_instruction *read = emit(VEC4_OPCODE_URB_READ, dst, src_reg(header));
+ read->offset = base_offset;
+ read->mlen = 1;
+ read->base_mrf = -1;
+}
+
+void
+vec4_tcs_visitor::emit_urb_write(const src_reg &value,
+ unsigned writemask,
+ unsigned base_offset,
+ const src_reg &indirect_offset)
+{
+ if (writemask == 0)
+ return;
+
+ src_reg message(this, glsl_type::uvec4_type, 2);
+ vec4_instruction *inst;
+
+ inst = emit(TCS_OPCODE_SET_OUTPUT_URB_OFFSETS, dst_reg(message),
+ brw_imm_ud(writemask), indirect_offset);
+ inst->force_writemask_all = true;
+ inst = emit(MOV(offset(dst_reg(retype(message, value.type)), 1), value));
+ inst->force_writemask_all = true;
+
+ inst = emit(TCS_OPCODE_URB_WRITE, dst_null_f(), message);
+ inst->offset = base_offset;
+ inst->mlen = 2;
+ inst->base_mrf = -1;
+}
+
+static unsigned
+tesslevel_outer_components(GLenum tes_primitive_mode)
+{
+ switch (tes_primitive_mode) {
+ case GL_QUADS:
+ return 4;
+ case GL_TRIANGLES:
+ return 3;
+ case GL_ISOLINES:
+ return 2;
+ default:
+ unreachable("Bogus tessellation domain");
+ }
+ return 0;
+}
+
+static unsigned
+tesslevel_inner_components(GLenum tes_primitive_mode)
+{
+ switch (tes_primitive_mode) {
+ case GL_QUADS:
+ return 2;
+ case GL_TRIANGLES:
+ return 1;
+ case GL_ISOLINES:
+ return 0;
+ default:
+ unreachable("Bogus tessellation domain");
+ }
+ return 0;
+}
+
+/**
+ * Given a normal .xyzw writemask, convert it to a writemask for a vector
+ * that's stored backwards, i.e. .wzyx.
+ */
+static unsigned
+writemask_for_backwards_vector(unsigned mask)
+{
+ unsigned new_mask = 0;
+
+ for (int i = 0; i < 4; i++)
+ new_mask |= ((mask >> i) & 1) << (3 - i);
+
+ return new_mask;
+}
+
+void
+vec4_tcs_visitor::nir_emit_intrinsic(nir_intrinsic_instr *instr)
+{
+ switch (instr->intrinsic) {
+ case nir_intrinsic_load_invocation_id:
+ emit(MOV(get_nir_dest(instr->dest, BRW_REGISTER_TYPE_UD),
+ invocation_id));
+ break;
+ case nir_intrinsic_load_primitive_id:
+ emit(TCS_OPCODE_GET_PRIMITIVE_ID,
+ get_nir_dest(instr->dest, BRW_REGISTER_TYPE_UD));
+ break;
+ case nir_intrinsic_load_patch_vertices_in:
+ emit(MOV(get_nir_dest(instr->dest, BRW_REGISTER_TYPE_D),
+ brw_imm_d(key->input_vertices)));
+ break;
+ case nir_intrinsic_load_per_vertex_input: {
+ src_reg indirect_offset = get_indirect_offset(instr);
+ unsigned imm_offset = instr->const_index[0];
+
+ nir_const_value *vertex_const = nir_src_as_const_value(instr->src[0]);
+ src_reg vertex_index =
+ vertex_const ? src_reg(brw_imm_ud(vertex_const->u[0]))
+ : get_nir_src(instr->src[0], BRW_REGISTER_TYPE_UD, 1);
+
+ dst_reg dst = get_nir_dest(instr->dest, BRW_REGISTER_TYPE_D);
+ dst.writemask = brw_writemask_for_size(instr->num_components);
+
+ emit_input_urb_read(dst, vertex_index, imm_offset, indirect_offset);
+ break;
+ }
+ case nir_intrinsic_load_input:
+ unreachable("nir_lower_io should use load_per_vertex_input intrinsics");
+ break;
+ case nir_intrinsic_load_output:
+ case nir_intrinsic_load_per_vertex_output: {
+ src_reg indirect_offset = get_indirect_offset(instr);
+ unsigned imm_offset = instr->const_index[0];;
+
+ dst_reg dst = get_nir_dest(instr->dest, BRW_REGISTER_TYPE_D);
+ dst.writemask = brw_writemask_for_size(instr->num_components);
+
+ if (imm_offset == 0 && indirect_offset.file == BAD_FILE) {
+ dst.type = BRW_REGISTER_TYPE_F;
+
+ /* This is a read of gl_TessLevelInner[], which lives in the
+ * Patch URB header. The layout depends on the domain.
+ */
+ switch (key->tes_primitive_mode) {
+ case GL_QUADS: {
+ /* DWords 3-2 (reversed); use offset 0 and WZYX swizzle. */
+ dst_reg tmp(this, glsl_type::vec4_type);
+ emit_output_urb_read(tmp, 0, src_reg());
+ emit(MOV(writemask(dst, WRITEMASK_XY),
+ swizzle(src_reg(tmp), BRW_SWIZZLE_WZYX)));
+ break;
+ }
+ case GL_TRIANGLES:
+ /* DWord 4; use offset 1 but normal swizzle/writemask. */
+ emit_output_urb_read(writemask(dst, WRITEMASK_X), 1, src_reg());
+ break;
+ case GL_ISOLINES:
+ /* All channels are undefined. */
+ return;
+ default:
+ unreachable("Bogus tessellation domain");
+ }
+ } else if (imm_offset == 1 && indirect_offset.file == BAD_FILE) {
+ dst.type = BRW_REGISTER_TYPE_F;
+
+ /* This is a read of gl_TessLevelOuter[], which lives in the
+ * high 4 DWords of the Patch URB header, in reverse order.
+ */
+ switch (key->tes_primitive_mode) {
+ case GL_QUADS:
+ dst.writemask = WRITEMASK_XYZW;
+ break;
+ case GL_TRIANGLES:
+ dst.writemask = WRITEMASK_XYZ;
+ break;
+ case GL_ISOLINES:
+ dst.writemask = WRITEMASK_XY;
+ return;
+ default:
+ unreachable("Bogus tessellation domain");
+ }
+
+ dst_reg tmp(this, glsl_type::vec4_type);
+ emit_output_urb_read(tmp, 1, src_reg());
+ emit(MOV(dst, swizzle(src_reg(tmp), BRW_SWIZZLE_WZYX)));
+ } else {
+ emit_output_urb_read(dst, imm_offset, indirect_offset);
+ }
+ break;
+ }
+ case nir_intrinsic_store_output:
+ case nir_intrinsic_store_per_vertex_output: {
+ src_reg value = get_nir_src(instr->src[0]);
+ unsigned mask = instr->const_index[1];
+ unsigned swiz = BRW_SWIZZLE_XYZW;
+
+ src_reg indirect_offset = get_indirect_offset(instr);
+ unsigned imm_offset = instr->const_index[0];
+
+ if (imm_offset == 0 && indirect_offset.file == BAD_FILE) {
+ value.type = BRW_REGISTER_TYPE_F;
+
+ mask &= (1 << tesslevel_inner_components(key->tes_primitive_mode)) - 1;
+
+ /* This is a write to gl_TessLevelInner[], which lives in the
+ * Patch URB header. The layout depends on the domain.
+ */
+ switch (key->tes_primitive_mode) {
+ case GL_QUADS:
+ /* gl_TessLevelInner[].xy lives at DWords 3-2 (reversed).
+ * We use an XXYX swizzle to reverse put .xy in the .wz
+ * channels, and use a .zw writemask.
+ */
+ swiz = BRW_SWIZZLE4(0, 0, 1, 0);
+ mask = writemask_for_backwards_vector(mask);
+ break;
+ case GL_TRIANGLES:
+ /* gl_TessLevelInner[].x lives at DWord 4, so we set the
+ * writemask to X and bump the URB offset by 1.
+ */
+ imm_offset = 1;
+ break;
+ case GL_ISOLINES:
+ /* Skip; gl_TessLevelInner[] doesn't exist for isolines. */
+ return;
+ default:
+ unreachable("Bogus tessellation domain");
+ }
+ } else if (imm_offset == 1 && indirect_offset.file == BAD_FILE) {
+ value.type = BRW_REGISTER_TYPE_F;
+
+ mask &= (1 << tesslevel_outer_components(key->tes_primitive_mode)) - 1;
+
+ /* This is a write to gl_TessLevelOuter[] which lives in the
+ * Patch URB Header at DWords 4-7. However, it's reversed, so
+ * instead of .xyzw we have .wzyx.
+ */
+ swiz = BRW_SWIZZLE_WZYX;
+ mask = writemask_for_backwards_vector(mask);
+ }
+
+ emit_urb_write(swizzle(value, swiz), mask,
+ imm_offset, indirect_offset);
+ break;
+ }
+
+ case nir_intrinsic_barrier: {
+ dst_reg header = dst_reg(this, glsl_type::uvec4_type);
+ emit(TCS_OPCODE_CREATE_BARRIER_HEADER, header);
+ emit(SHADER_OPCODE_BARRIER, dst_null_ud(), src_reg(header));
+ break;
+ }
+
+ default:
+ vec4_visitor::nir_emit_intrinsic(instr);
+ }
+}
+
+
+extern "C" const unsigned *
+brw_compile_tcs(const struct brw_compiler *compiler,
+ void *log_data,
+ void *mem_ctx,
+ const struct brw_tcs_prog_key *key,
+ struct brw_tcs_prog_data *prog_data,
+ const nir_shader *src_shader,
+ int shader_time_index,
+ unsigned *final_assembly_size,
+ char **error_str)
+{
+ const struct brw_device_info *devinfo = compiler->devinfo;
+ struct brw_vue_prog_data *vue_prog_data = &prog_data->base;
+ const bool is_scalar = compiler->scalar_stage[MESA_SHADER_TESS_CTRL];
+
+ nir_shader *nir = nir_shader_clone(mem_ctx, src_shader);
+ nir = brw_nir_apply_sampler_key(nir, devinfo, &key->tex, is_scalar);
+ nir->info.outputs_written = key->outputs_written;
+ nir->info.patch_outputs_written = key->patch_outputs_written;
+ nir = brw_nir_lower_io(nir, compiler->devinfo, is_scalar);
+ nir = brw_postprocess_nir(nir, compiler->devinfo, is_scalar);
+
+ prog_data->instances = DIV_ROUND_UP(nir->info.tcs.vertices_out, 2);
+
+ brw_compute_tess_vue_map(&vue_prog_data->vue_map,
+ nir->info.outputs_written,
+ nir->info.patch_outputs_written);
+
+ /* Compute URB entry size. The maximum allowed URB entry size is 32k.
+ * That divides up as follows:
+ *
+ * 32 bytes for the patch header (tessellation factors)
+ * 480 bytes for per-patch varyings (a varying component is 4 bytes and
+ * gl_MaxTessPatchComponents = 120)
+ * 16384 bytes for per-vertex varyings (a varying component is 4 bytes,
+ * gl_MaxPatchVertices = 32 and
+ * gl_MaxTessControlOutputComponents = 128)
+ *
+ * 15808 bytes left for varying packing overhead
+ */
+ const int num_per_patch_slots = vue_prog_data->vue_map.num_per_patch_slots;
+ const int num_per_vertex_slots = vue_prog_data->vue_map.num_per_vertex_slots;
+ unsigned output_size_bytes = 0;
+ /* Note that the patch header is counted in num_per_patch_slots. */
+ output_size_bytes += num_per_patch_slots * 16;
+ output_size_bytes += nir->info.tcs.vertices_out * num_per_vertex_slots * 16;
+
+ assert(output_size_bytes >= 1);
+ if (output_size_bytes > GEN7_MAX_HS_URB_ENTRY_SIZE_BYTES)
+ return false;
+
+ /* URB entry sizes are stored as a multiple of 64 bytes. */
+ vue_prog_data->urb_entry_size = ALIGN(output_size_bytes, 64) / 64;
+
+ struct brw_vue_map input_vue_map;
+ brw_compute_vue_map(devinfo, &input_vue_map,
+ nir->info.inputs_read & ~VARYING_BIT_PRIMITIVE_ID,
+ true);
+
+ /* HS does not use the usual payload pushing from URB to GRFs,
+ * because we don't have enough registers for a full-size payload, and
+ * the hardware is broken on Haswell anyway.
+ */
+ vue_prog_data->urb_read_length = 0;
+
+ if (unlikely(INTEL_DEBUG & DEBUG_TCS)) {
+ fprintf(stderr, "TCS Input ");
+ brw_print_vue_map(stderr, &input_vue_map);
+ fprintf(stderr, "TCS Output ");
+ brw_print_vue_map(stderr, &vue_prog_data->vue_map);
+ }
+
+ vec4_tcs_visitor v(compiler, log_data, key, prog_data,
+ nir, mem_ctx, shader_time_index, &input_vue_map);
+ if (!v.run()) {
+ if (error_str)
+ *error_str = ralloc_strdup(mem_ctx, v.fail_msg);
+ return NULL;
+ }
+
+ if (unlikely(INTEL_DEBUG & DEBUG_TCS))
+ v.dump_instructions();
+
+ return brw_vec4_generate_assembly(compiler, log_data, mem_ctx, nir,
+ &prog_data->base, v.cfg,
+ final_assembly_size);
+}
+
+
+} /* namespace brw */
--- /dev/null
+/*
+ * Copyright © 2013 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file brw_vec4_tcs.h
+ *
+ * The vec4-mode tessellation control shader compiler backend.
+ */
+
+#ifndef BRW_VEC4_TCS_H
+#define BRW_VEC4_TCS_H
+
+#include "brw_compiler.h"
+#include "brw_vec4.h"
+
+#ifdef __cplusplus
+namespace brw {
+
+class vec4_tcs_visitor : public vec4_visitor
+{
+public:
+ vec4_tcs_visitor(const struct brw_compiler *compiler,
+ void *log_data,
+ const struct brw_tcs_prog_key *key,
+ struct brw_tcs_prog_data *prog_data,
+ const nir_shader *nir,
+ void *mem_ctx,
+ int shader_time_index,
+ const struct brw_vue_map *input_vue_map);
+
+protected:
+ virtual void emit_nir_code();
+ virtual dst_reg *make_reg_for_system_value(int location,
+ const glsl_type *type);
+ virtual void nir_setup_system_value_intrinsic(nir_intrinsic_instr *instr);
+ virtual void setup_payload();
+ virtual void emit_prolog();
+ virtual void emit_thread_end();
+
+ virtual void nir_emit_intrinsic(nir_intrinsic_instr *instr);
+
+ void emit_input_urb_read(const dst_reg &dst,
+ const src_reg &vertex_index,
+ unsigned base_offset,
+ const src_reg &indirect_offset);
+ void emit_output_urb_read(const dst_reg &dst,
+ unsigned base_offset,
+ const src_reg &indirect_offset);
+
+ void emit_urb_write(const src_reg &value, unsigned writemask,
+ unsigned base_offset, const src_reg &indirect_offset);
+
+ /* we do not use the normal end-of-shader URB write mechanism -- but every vec4 stage
+ * must provide implementations of these:
+ */
+ virtual void emit_urb_write_header(int mrf) {}
+ virtual vec4_instruction *emit_urb_write_opcode(bool complete) { return NULL; }
+
+ const struct brw_vue_map *input_vue_map;
+
+ const struct brw_tcs_prog_key *key;
+ src_reg invocation_id;
+};
+
+} /* namespace brw */
+#endif /* __cplusplus */
+
+#endif /* BRW_VEC4_TCS_H */
--- /dev/null
+/*
+ * Copyright © 2013 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file brw_vec4_tes.cpp
+ *
+ * Tessellaton evaluation shader specific code derived from the vec4_visitor class.
+ */
+
+#include "brw_vec4_tes.h"
+
+namespace brw {
+
+vec4_tes_visitor::vec4_tes_visitor(const struct brw_compiler *compiler,
+ void *log_data,
+ const struct brw_tes_prog_key *key,
+ struct brw_tes_prog_data *prog_data,
+ const nir_shader *shader,
+ void *mem_ctx,
+ int shader_time_index)
+ : vec4_visitor(compiler, log_data, &key->tex, &prog_data->base,
+ shader, mem_ctx, false, shader_time_index)
+{
+}
+
+
+dst_reg *
+vec4_tes_visitor::make_reg_for_system_value(int location, const glsl_type *type)
+{
+ return NULL;
+}
+
+void
+vec4_tes_visitor::nir_setup_system_value_intrinsic(nir_intrinsic_instr *instr)
+{
+ const struct brw_tes_prog_data *tes_prog_data =
+ (const struct brw_tes_prog_data *) prog_data;
+
+ switch (instr->intrinsic) {
+ case nir_intrinsic_load_tess_level_outer: {
+ dst_reg dst(this, glsl_type::vec4_type);
+ nir_system_values[SYSTEM_VALUE_TESS_LEVEL_OUTER] = dst;
+
+ dst_reg temp(this, glsl_type::vec4_type);
+ vec4_instruction *read =
+ emit(VEC4_OPCODE_URB_READ, temp, input_read_header);
+ read->offset = 1;
+ read->urb_write_flags = BRW_URB_WRITE_PER_SLOT_OFFSET;
+ emit(MOV(dst, swizzle(src_reg(temp), BRW_SWIZZLE_WZYX)));
+ break;
+ }
+ case nir_intrinsic_load_tess_level_inner: {
+ dst_reg dst(this, glsl_type::vec2_type);
+ nir_system_values[SYSTEM_VALUE_TESS_LEVEL_INNER] = dst;
+
+ /* Set up the message header to reference the proper parts of the URB */
+ dst_reg temp(this, glsl_type::vec4_type);
+ vec4_instruction *read =
+ emit(VEC4_OPCODE_URB_READ, temp, input_read_header);
+ read->urb_write_flags = BRW_URB_WRITE_PER_SLOT_OFFSET;
+ if (tes_prog_data->domain == BRW_TESS_DOMAIN_QUAD) {
+ emit(MOV(dst, swizzle(src_reg(temp), BRW_SWIZZLE_WZYX)));
+ } else {
+ read->offset = 1;
+ emit(MOV(dst, src_reg(temp)));
+ }
+ break;
+ }
+ default:
+ vec4_visitor::nir_setup_system_value_intrinsic(instr);
+ }
+}
+
+
+void
+vec4_tes_visitor::setup_payload()
+{
+ int reg = 0;
+
+ /* The payload always contains important data in r0 and r1, which contains
+ * the URB handles that are passed on to the URB write at the end
+ * of the thread.
+ */
+ reg += 2;
+
+ reg = setup_uniforms(reg);
+
+ this->first_non_payload_grf = reg;
+}
+
+
+void
+vec4_tes_visitor::emit_prolog()
+{
+ input_read_header = src_reg(this, glsl_type::uvec4_type);
+ emit(TES_OPCODE_CREATE_INPUT_READ_HEADER, dst_reg(input_read_header));
+
+ this->current_annotation = NULL;
+}
+
+
+void
+vec4_tes_visitor::emit_urb_write_header(int mrf)
+{
+ /* No need to do anything for DS; an implied write to this MRF will be
+ * performed by VS_OPCODE_URB_WRITE.
+ */
+ (void) mrf;
+}
+
+
+vec4_instruction *
+vec4_tes_visitor::emit_urb_write_opcode(bool complete)
+{
+ /* For DS, the URB writes end the thread. */
+ if (complete) {
+ if (INTEL_DEBUG & DEBUG_SHADER_TIME)
+ emit_shader_time_end();
+ }
+
+ vec4_instruction *inst = emit(VS_OPCODE_URB_WRITE);
+ inst->urb_write_flags = complete ?
+ BRW_URB_WRITE_EOT_COMPLETE : BRW_URB_WRITE_NO_FLAGS;
+
+ return inst;
+}
+
+void
+vec4_tes_visitor::nir_emit_intrinsic(nir_intrinsic_instr *instr)
+{
+ switch (instr->intrinsic) {
+ case nir_intrinsic_load_tess_coord:
+ /* gl_TessCoord is part of the payload in g1 channels 0-2 and 4-6. */
+ emit(MOV(get_nir_dest(instr->dest, BRW_REGISTER_TYPE_F),
+ src_reg(brw_vec8_grf(1, 0))));
+ break;
+ case nir_intrinsic_load_primitive_id:
+ emit(TES_OPCODE_GET_PRIMITIVE_ID,
+ get_nir_dest(instr->dest, BRW_REGISTER_TYPE_UD));
+ break;
+
+ case nir_intrinsic_load_input:
+ case nir_intrinsic_load_per_vertex_input: {
+ src_reg indirect_offset = get_indirect_offset(instr);
+ unsigned imm_offset = instr->const_index[0];
+ src_reg header = input_read_header;
+
+ if (indirect_offset.file != BAD_FILE) {
+ header = src_reg(this, glsl_type::uvec4_type);
+ emit(TES_OPCODE_ADD_INDIRECT_URB_OFFSET, dst_reg(header),
+ input_read_header, indirect_offset);
+ }
+
+ dst_reg temp(this, glsl_type::ivec4_type);
+ vec4_instruction *read =
+ emit(VEC4_OPCODE_URB_READ, temp, src_reg(header));
+ read->offset = imm_offset;
+ read->urb_write_flags = BRW_URB_WRITE_PER_SLOT_OFFSET;
+
+ /* Copy to target. We might end up with some funky writemasks landing
+ * in here, but we really don't want them in the above pseudo-ops.
+ */
+ dst_reg dst = get_nir_dest(instr->dest, BRW_REGISTER_TYPE_D);
+ dst.writemask = brw_writemask_for_size(instr->num_components);
+ emit(MOV(dst, src_reg(temp)));
+ break;
+ }
+ default:
+ vec4_visitor::nir_emit_intrinsic(instr);
+ }
+}
+
+
+void
+vec4_tes_visitor::emit_thread_end()
+{
+ /* For DS, we always end the thread by emitting a single vertex.
+ * emit_urb_write_opcode() will take care of setting the eot flag on the
+ * SEND instruction.
+ */
+ emit_vertex();
+}
+
+} /* namespace brw */
--- /dev/null
+/*
+ * Copyright © 2013 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file brw_vec4_tes.h
+ *
+ * The vec4 mode tessellation evaluation shader compiler backend.
+ */
+
+#ifndef BRW_VEC4_TES_H
+#define BRW_VEC4_TES_H
+
+#include "brw_vec4.h"
+
+#ifdef __cplusplus
+namespace brw {
+
+class vec4_tes_visitor : public vec4_visitor
+{
+public:
+ vec4_tes_visitor(const struct brw_compiler *compiler,
+ void *log_data,
+ const struct brw_tes_prog_key *key,
+ struct brw_tes_prog_data *prog_data,
+ const nir_shader *nir,
+ void *mem_ctx,
+ int shader_time_index);
+
+protected:
+ virtual dst_reg *make_reg_for_system_value(int location,
+ const glsl_type *type);
+ virtual void nir_setup_system_value_intrinsic(nir_intrinsic_instr *instr);
+ virtual void nir_emit_intrinsic(nir_intrinsic_instr *instr);
+
+ virtual void setup_payload();
+ virtual void emit_prolog();
+ virtual void emit_thread_end();
+
+ virtual void emit_urb_write_header(int mrf);
+ virtual vec4_instruction *emit_urb_write_opcode(bool complete);
+
+private:
+ src_reg input_read_header;
+};
+
+} /* namespace brw */
+#endif /* __cplusplus */
+
+#endif /* BRW_VEC4_TES_H */
case GLSL_TYPE_DOUBLE:
case GLSL_TYPE_ERROR:
case GLSL_TYPE_INTERFACE:
+ case GLSL_TYPE_FUNCTION:
unreachable("not reached");
}
src_reg offset_value,
src_reg mcs,
bool is_cube_array,
+ uint32_t surface,
+ src_reg surface_reg,
uint32_t sampler,
src_reg sampler_reg)
{
inst->dst.writemask = WRITEMASK_XYZW;
inst->shadow_compare = shadow_comparitor.file != BAD_FILE;
- inst->src[1] = sampler_reg;
+ inst->src[1] = surface_reg;
+ inst->src[2] = sampler_reg;
/* MRF for the first parameter */
int param_base = inst->base_mrf + inst->header_size;
}
if (devinfo->gen == 6 && op == ir_tg4) {
- emit_gen6_gather_wa(key_tex->gen6_gather_wa[sampler], inst->dst);
+ emit_gen6_gather_wa(key_tex->gen6_gather_wa[surface], inst->dst);
}
if (op == ir_query_levels) {
switch (location) {
case SYSTEM_VALUE_BASE_VERTEX:
reg->writemask = WRITEMASK_X;
- vs_prog_data->uses_vertexid = true;
+ vs_prog_data->uses_basevertex = true;
+ break;
+ case SYSTEM_VALUE_BASE_INSTANCE:
+ reg->writemask = WRITEMASK_Y;
+ vs_prog_data->uses_baseinstance = true;
break;
case SYSTEM_VALUE_VERTEX_ID:
case SYSTEM_VALUE_VERTEX_ID_ZERO_BASE:
reg->writemask = WRITEMASK_W;
vs_prog_data->uses_instanceid = true;
break;
+ case SYSTEM_VALUE_DRAW_ID:
+ reg = new(mem_ctx) dst_reg(ATTR, VERT_ATTRIB_MAX + 1);
+ reg->writemask = WRITEMASK_X;
+ vs_prog_data->uses_drawid = true;
+ break;
default:
unreachable("not reached");
}
brw_compute_vue_map(brw->intelScreen->devinfo,
&prog_data.base.vue_map, outputs_written,
- prog ? prog->SeparateShader : false);
+ prog ? prog->SeparateShader ||
+ prog->_LinkedShaders[MESA_SHADER_TESS_EVAL]
+ : false);
if (0) {
_mesa_fprint_program_opt(stderr, &vp->program.Base, PROG_PRINT_DEBUG,
}
vue_map->num_slots = separate ? slot + 1 : slot;
+ vue_map->num_per_vertex_slots = 0;
+ vue_map->num_per_patch_slots = 0;
+}
+
+/**
+ * Compute the VUE map for tessellation control shader outputs and
+ * tessellation evaluation shader inputs.
+ */
+void
+brw_compute_tess_vue_map(struct brw_vue_map *vue_map,
+ GLbitfield64 vertex_slots,
+ GLbitfield patch_slots)
+{
+ /* I don't think anything actually uses this... */
+ vue_map->slots_valid = vertex_slots;
+
+ vertex_slots &= ~(VARYING_BIT_TESS_LEVEL_OUTER |
+ VARYING_BIT_TESS_LEVEL_INNER);
+
+ /* Make sure that the values we store in vue_map->varying_to_slot and
+ * vue_map->slot_to_varying won't overflow the signed chars that are used
+ * to store them. Note that since vue_map->slot_to_varying sometimes holds
+ * values equal to VARYING_SLOT_TESS_MAX , we need to ensure that
+ * VARYING_SLOT_TESS_MAX is <= 127, not 128.
+ */
+ STATIC_ASSERT(VARYING_SLOT_TESS_MAX <= 127);
+
+ for (int i = 0; i < VARYING_SLOT_TESS_MAX ; ++i) {
+ vue_map->varying_to_slot[i] = -1;
+ vue_map->slot_to_varying[i] = BRW_VARYING_SLOT_PAD;
+ }
+
+ int slot = 0;
+
+ /* The first 8 DWords are reserved for the "Patch Header".
+ *
+ * VARYING_SLOT_TESS_LEVEL_OUTER / INNER live here, but the exact layout
+ * depends on the domain type. They might not be in slots 0 and 1 as
+ * described here, but pretending they're separate allows us to uniquely
+ * identify them by distinct slot locations.
+ */
+ assign_vue_slot(vue_map, VARYING_SLOT_TESS_LEVEL_INNER, slot++);
+ assign_vue_slot(vue_map, VARYING_SLOT_TESS_LEVEL_OUTER, slot++);
+
+ /* first assign per-patch varyings */
+ while (patch_slots != 0) {
+ const int varying = ffsll(patch_slots) - 1;
+ if (vue_map->varying_to_slot[varying + VARYING_SLOT_PATCH0] == -1) {
+ assign_vue_slot(vue_map, varying + VARYING_SLOT_PATCH0, slot++);
+ }
+ patch_slots &= ~BITFIELD64_BIT(varying);
+ }
+
+ /* apparently, including the patch header... */
+ vue_map->num_per_patch_slots = slot;
+
+ /* then assign per-vertex varyings for each vertex in our patch */
+ while (vertex_slots != 0) {
+ const int varying = ffsll(vertex_slots) - 1;
+ if (vue_map->varying_to_slot[varying] == -1) {
+ assign_vue_slot(vue_map, varying, slot++);
+ }
+ vertex_slots &= ~BITFIELD64_BIT(varying);
+ }
+
+ vue_map->num_per_vertex_slots = slot - vue_map->num_per_patch_slots;
+ vue_map->num_slots = slot;
}
static const char *
[BRW_VARYING_SLOT_PNTC - VARYING_SLOT_MAX] = "BRW_VARYING_SLOT_PNTC",
};
+ assert(slot < BRW_VARYING_SLOT_COUNT);
return brw_names[slot - VARYING_SLOT_MAX];
}
void
brw_print_vue_map(FILE *fp, const struct brw_vue_map *vue_map)
{
- fprintf(fp, "VUE map (%d slots, %s)\n",
- vue_map->num_slots, vue_map->separate ? "SSO" : "non-SSO");
- for (int i = 0; i < vue_map->num_slots; i++) {
- fprintf(fp, " [%d] %s\n", i,
- varying_name(vue_map->slot_to_varying[i]));
+ if (vue_map->num_per_vertex_slots > 0 || vue_map->num_per_patch_slots > 0) {
+ fprintf(fp, "PUE map (%d slots, %d/patch, %d/vertex, %s)\n",
+ vue_map->num_slots,
+ vue_map->num_per_patch_slots,
+ vue_map->num_per_vertex_slots,
+ vue_map->separate ? "SSO" : "non-SSO");
+ for (int i = 0; i < vue_map->num_slots; i++) {
+ if (vue_map->slot_to_varying[i] >= VARYING_SLOT_PATCH0) {
+ fprintf(fp, " [%d] VARYING_SLOT_PATCH%d\n", i,
+ vue_map->slot_to_varying[i] - VARYING_SLOT_PATCH0);
+ } else {
+ fprintf(fp, " [%d] %s\n", i,
+ varying_name(vue_map->slot_to_varying[i]));
+ }
+ }
+ } else {
+ fprintf(fp, "VUE map (%d slots, %s)\n",
+ vue_map->num_slots, vue_map->separate ? "SSO" : "non-SSO");
+ for (int i = 0; i < vue_map->num_slots; i++) {
+ fprintf(fp, " [%d] %s\n", i,
+ varying_name(vue_map->slot_to_varying[i]));
+ }
}
fprintf(fp, "\n");
}
/* BRW_NEW_VERTEX_PROGRAM */
struct gl_program *vs = (struct gl_program *) brw->vertex_program;
- /* BRW_NEW_TESS_CTRL_PROGRAM */
+ /* BRW_NEW_TESS_PROGRAMS */
struct gl_program *tcs = (struct gl_program *) brw->tess_ctrl_program;
-
- /* BRW_NEW_TESS_EVAL_PROGRAM */
struct gl_program *tes = (struct gl_program *) brw->tess_eval_program;
/* BRW_NEW_GEOMETRY_PROGRAM */
BRW_NEW_FS_PROG_DATA |
BRW_NEW_GEOMETRY_PROGRAM |
BRW_NEW_GS_PROG_DATA |
- BRW_NEW_TESS_CTRL_PROGRAM |
- BRW_NEW_TESS_EVAL_PROGRAM |
+ BRW_NEW_TESS_PROGRAMS |
BRW_NEW_TCS_PROG_DATA |
BRW_NEW_TES_PROG_DATA |
BRW_NEW_TEXTURE_BUFFER |
} else {
struct intel_buffer_object *intel_bo =
intel_buffer_object(binding->BufferObject);
+ GLsizeiptr size = binding->BufferObject->Size - binding->Offset;
+ if (!binding->AutomaticSize)
+ size = MIN2(size, binding->Size);
drm_intel_bo *bo =
intel_bufferobj_buffer(brw, intel_bo,
binding->Offset,
- binding->BufferObject->Size - binding->Offset);
+ size);
brw_create_constant_surface(brw, bo, binding->Offset,
- binding->BufferObject->Size - binding->Offset,
+ size,
&ubo_surf_offsets[i]);
}
}
} else {
struct intel_buffer_object *intel_bo =
intel_buffer_object(binding->BufferObject);
+ GLsizeiptr size = binding->BufferObject->Size - binding->Offset;
+ if (!binding->AutomaticSize)
+ size = MIN2(size, binding->Size);
drm_intel_bo *bo =
intel_bufferobj_buffer(brw, intel_bo,
binding->Offset,
- binding->BufferObject->Size - binding->Offset);
+ size);
brw_create_buffer_surface(brw, bo, binding->Offset,
- binding->BufferObject->Size - binding->Offset,
+ size,
&ssbo_surf_offsets[i]);
}
}
#include "program/prog_parameter.h"
#include "program/prog_statevars.h"
#include "intel_batchbuffer.h"
-#include "glsl/glsl_parser_extras.h"
/**
* Creates a streamed BO containing the push constants for the VS or GS on
* basic type of PROGRAM_STATE_VAR.
*/
/* XXX: Should this happen somewhere before to get our state flag set? */
- _mesa_load_state_parameters(ctx, prog->Parameters);
+ if (prog)
+ _mesa_load_state_parameters(ctx, prog->Parameters);
gl_constant_value *param;
unsigned i;
unsigned urb_size = (brw->is_haswell && brw->gt == 3) ? 32 : 16;
gen7_emit_push_constant_state(brw,
urb_size / 2 /* vs_size */,
+ 0 /* hs_size */,
+ 0 /* ds_size */,
0 /* gs_size */,
urb_size / 2 /* fs_size */);
32 /* num_vs_entries */,
2 /* vs_size */,
2 /* vs_start */,
+ 0 /* num_hs_entries */,
+ 1 /* hs_size */,
+ 2 /* hs_start */,
+ 0 /* num_ds_entries */,
+ 1 /* ds_size */,
+ 2 /* ds_start */,
0 /* num_gs_entries */,
1 /* gs_size */,
2 /* gs_start */);
BEGIN_BATCH(4);
OUT_BATCH(MEDIA_CURBE_LOAD << 16 | (4 - 2));
OUT_BATCH(0);
- OUT_BATCH(reg_aligned_constant_size * threads);
+ OUT_BATCH(ALIGN(reg_aligned_constant_size * threads, 64));
OUT_BATCH(stage_state->push_const_offset);
ADVANCE_BATCH();
}
param = (gl_constant_value*)
brw_state_batch(brw, type,
- reg_aligned_constant_size * threads,
- 32, &stage_state->push_const_offset);
+ ALIGN(reg_aligned_constant_size * threads, 64),
+ 64, &stage_state->push_const_offset);
assert(param);
STATIC_ASSERT(sizeof(gl_constant_value) == sizeof(float));
gen7_upload_tes_push_constants(struct brw_context *brw)
{
struct brw_stage_state *stage_state = &brw->tes.base;
- /* BRW_NEW_TESS_EVAL_PROGRAM */
+ /* BRW_NEW_TESS_PROGRAMS */
const struct brw_tess_eval_program *tep =
(struct brw_tess_eval_program *) brw->tess_eval_program;
.mesa = _NEW_PROGRAM_CONSTANTS,
.brw = BRW_NEW_BATCH |
BRW_NEW_PUSH_CONSTANT_ALLOCATION |
- BRW_NEW_TESS_EVAL_PROGRAM |
+ BRW_NEW_TESS_PROGRAMS |
BRW_NEW_TES_PROG_DATA,
},
.emit = gen7_upload_tes_push_constants,
static void
gen7_upload_ds_state(struct brw_context *brw)
{
- /* Disable the DS Unit */
- BEGIN_BATCH(7);
- OUT_BATCH(_3DSTATE_CONSTANT_DS << 16 | (7 - 2));
- OUT_BATCH(0);
- OUT_BATCH(0);
- OUT_BATCH(0);
- OUT_BATCH(0);
- OUT_BATCH(0);
- OUT_BATCH(0);
- ADVANCE_BATCH();
+ const struct brw_stage_state *stage_state = &brw->tes.base;
+ /* BRW_NEW_TESS_PROGRAMS */
+ bool active = brw->tess_eval_program;
- BEGIN_BATCH(6);
- OUT_BATCH(_3DSTATE_DS << 16 | (6 - 2));
- OUT_BATCH(0);
- OUT_BATCH(0);
- OUT_BATCH(0);
- OUT_BATCH(0);
- OUT_BATCH(0);
- ADVANCE_BATCH();
+ /* BRW_NEW_TES_PROG_DATA */
+ const struct brw_tes_prog_data *tes_prog_data = brw->tes.prog_data;
+ const struct brw_vue_prog_data *vue_prog_data = &tes_prog_data->base;
+ const struct brw_stage_prog_data *prog_data = &vue_prog_data->base;
- BEGIN_BATCH(2);
- OUT_BATCH(_3DSTATE_BINDING_TABLE_POINTERS_DS << 16 | (2 - 2));
- OUT_BATCH(brw->hw_bt_pool.next_offset);
- ADVANCE_BATCH();
+ const unsigned thread_count = (brw->max_ds_threads - 1) <<
+ (brw->is_haswell ? HSW_DS_MAX_THREADS_SHIFT : GEN7_DS_MAX_THREADS_SHIFT);
+
+ if (active) {
+ BEGIN_BATCH(6);
+ OUT_BATCH(_3DSTATE_DS << 16 | (6 - 2));
+ OUT_BATCH(stage_state->prog_offset);
+ OUT_BATCH(SET_FIELD(DIV_ROUND_UP(stage_state->sampler_count, 4),
+ GEN7_DS_SAMPLER_COUNT) |
+ SET_FIELD(prog_data->binding_table.size_bytes / 4,
+ GEN7_DS_BINDING_TABLE_ENTRY_COUNT));
+ if (prog_data->total_scratch) {
+ OUT_RELOC(stage_state->scratch_bo,
+ I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
+ ffs(prog_data->total_scratch) - 11);
+ } else {
+ OUT_BATCH(0);
+ }
+ OUT_BATCH(SET_FIELD(prog_data->dispatch_grf_start_reg,
+ GEN7_DS_DISPATCH_START_GRF) |
+ SET_FIELD(vue_prog_data->urb_read_length,
+ GEN7_DS_URB_READ_LENGTH));
+
+ OUT_BATCH(GEN7_DS_ENABLE |
+ GEN7_DS_STATISTICS_ENABLE |
+ thread_count |
+ (tes_prog_data->domain == BRW_TESS_DOMAIN_TRI ?
+ GEN7_DS_COMPUTE_W_COORDINATE_ENABLE : 0));
+ ADVANCE_BATCH();
+ } else {
+ BEGIN_BATCH(6);
+ OUT_BATCH(_3DSTATE_DS << 16 | (6 - 2));
+ OUT_BATCH(0);
+ OUT_BATCH(0);
+ OUT_BATCH(0);
+ OUT_BATCH(0);
+ OUT_BATCH(0);
+ ADVANCE_BATCH();
+ }
+ brw->tes.enabled = active;
}
const struct brw_tracked_state gen7_ds_state = {
.dirty = {
- .mesa = 0,
- .brw = BRW_NEW_CONTEXT,
+ .mesa = _NEW_TRANSFORM,
+ .brw = BRW_NEW_BATCH |
+ BRW_NEW_CONTEXT |
+ BRW_NEW_TESS_PROGRAMS |
+ BRW_NEW_TES_PROG_DATA,
},
.emit = gen7_upload_ds_state,
};
gen7_upload_tcs_push_constants(struct brw_context *brw)
{
struct brw_stage_state *stage_state = &brw->tcs.base;
- /* BRW_NEW_TESS_CTRL_PROGRAM */
+ /* BRW_NEW_TESS_PROGRAMS */
const struct brw_tess_ctrl_program *tcp =
(struct brw_tess_ctrl_program *) brw->tess_ctrl_program;
+ bool active = brw->tess_eval_program;
- if (tcp) {
+ if (active) {
/* BRW_NEW_TCS_PROG_DATA */
const struct brw_stage_prog_data *prog_data = &brw->tcs.prog_data->base.base;
gen6_upload_push_constants(brw, &tcp->program.Base, prog_data,
stage_state, AUB_TRACE_VS_CONSTANTS);
}
- gen7_upload_constant_state(brw, stage_state, tcp, _3DSTATE_CONSTANT_HS);
+ gen7_upload_constant_state(brw, stage_state, active, _3DSTATE_CONSTANT_HS);
}
const struct brw_tracked_state gen7_tcs_push_constants = {
.dirty = {
.mesa = _NEW_PROGRAM_CONSTANTS,
.brw = BRW_NEW_BATCH |
+ BRW_NEW_DEFAULT_TESS_LEVELS |
BRW_NEW_PUSH_CONSTANT_ALLOCATION |
- BRW_NEW_TESS_CTRL_PROGRAM |
+ BRW_NEW_TESS_PROGRAMS |
BRW_NEW_TCS_PROG_DATA,
},
.emit = gen7_upload_tcs_push_constants,
static void
gen7_upload_hs_state(struct brw_context *brw)
{
- /* Disable the HS Unit */
- BEGIN_BATCH(7);
- OUT_BATCH(_3DSTATE_CONSTANT_HS << 16 | (7 - 2));
- OUT_BATCH(0);
- OUT_BATCH(0);
- OUT_BATCH(0);
- OUT_BATCH(0);
- OUT_BATCH(0);
- OUT_BATCH(0);
- ADVANCE_BATCH();
+ const struct brw_stage_state *stage_state = &brw->tcs.base;
+ /* BRW_NEW_TESS_PROGRAMS */
+ bool active = brw->tess_eval_program;
+ /* BRW_NEW_TCS_PROG_DATA */
+ const struct brw_vue_prog_data *prog_data = &brw->tcs.prog_data->base;
- BEGIN_BATCH(7);
- OUT_BATCH(_3DSTATE_HS << 16 | (7 - 2));
- OUT_BATCH(0);
- OUT_BATCH(0);
- OUT_BATCH(0);
- OUT_BATCH(0);
- OUT_BATCH(0);
- OUT_BATCH(0);
- ADVANCE_BATCH();
-
- BEGIN_BATCH(2);
- OUT_BATCH(_3DSTATE_BINDING_TABLE_POINTERS_HS << 16 | (2 - 2));
- OUT_BATCH(brw->hw_bt_pool.next_offset);
- ADVANCE_BATCH();
+ if (active) {
+ BEGIN_BATCH(7);
+ OUT_BATCH(_3DSTATE_HS << 16 | (7 - 2));
+ OUT_BATCH(SET_FIELD(DIV_ROUND_UP(stage_state->sampler_count, 4),
+ GEN7_HS_SAMPLER_COUNT) |
+ SET_FIELD(prog_data->base.binding_table.size_bytes / 4,
+ GEN7_HS_BINDING_TABLE_ENTRY_COUNT) |
+ (brw->max_hs_threads - 1));
+ OUT_BATCH(GEN7_HS_ENABLE |
+ GEN7_HS_STATISTICS_ENABLE |
+ SET_FIELD(brw->tcs.prog_data->instances - 1,
+ GEN7_HS_INSTANCE_COUNT));
+ OUT_BATCH(stage_state->prog_offset);
+ if (prog_data->base.total_scratch) {
+ OUT_RELOC(stage_state->scratch_bo,
+ I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
+ ffs(prog_data->base.total_scratch) - 11);
+ } else {
+ OUT_BATCH(0);
+ }
+ OUT_BATCH(GEN7_HS_INCLUDE_VERTEX_HANDLES |
+ SET_FIELD(prog_data->base.dispatch_grf_start_reg,
+ GEN7_HS_DISPATCH_START_GRF));
+ /* Ignore URB semaphores */
+ OUT_BATCH(0);
+ ADVANCE_BATCH();
+ } else {
+ BEGIN_BATCH(7);
+ OUT_BATCH(_3DSTATE_HS << 16 | (7 - 2));
+ OUT_BATCH(0);
+ OUT_BATCH(0);
+ OUT_BATCH(0);
+ OUT_BATCH(0);
+ OUT_BATCH(0);
+ OUT_BATCH(0);
+ ADVANCE_BATCH();
+ }
+ brw->tcs.enabled = active;
}
const struct brw_tracked_state gen7_hs_state = {
.dirty = {
.mesa = 0,
- .brw = BRW_NEW_CONTEXT,
+ .brw = BRW_NEW_BATCH |
+ BRW_NEW_TCS_PROG_DATA |
+ BRW_NEW_TESS_PROGRAMS,
},
.emit = gen7_upload_hs_state,
};
static void
upload_te_state(struct brw_context *brw)
{
- /* BRW_NEW_TESS_EVAL_PROGRAM */
+ /* BRW_NEW_TESS_PROGRAMS */
bool active = brw->tess_eval_program;
- if (active)
- assert(brw->tess_ctrl_program);
const struct brw_tes_prog_data *tes_prog_data = brw->tes.prog_data;
.mesa = 0,
.brw = BRW_NEW_CONTEXT |
BRW_NEW_TES_PROG_DATA |
- BRW_NEW_TESS_EVAL_PROGRAM,
+ BRW_NEW_TESS_PROGRAMS,
},
.emit = upload_te_state,
};
* __________-__________ _________________-_________________
* / \ / \
* +-------------------------------------------------------------+
- * | VS/FS/GS Push | VS/GS URB |
+ * | VS/HS/DS/GS/FS Push | VS/HS/DS/GS URB |
* | Constants | Entries |
* +-------------------------------------------------------------+
*
static void
gen7_allocate_push_constants(struct brw_context *brw)
{
+ /* BRW_NEW_GEOMETRY_PROGRAM */
+ bool gs_present = brw->geometry_program;
+
+ /* BRW_NEW_TESS_PROGRAMS */
+ bool tess_present = brw->tess_eval_program;
+
unsigned avail_size = 16;
unsigned multiplier =
(brw->gen >= 8 || (brw->is_haswell && brw->gt == 3)) ? 2 : 1;
- /* BRW_NEW_GEOMETRY_PROGRAM */
- bool gs_present = brw->geometry_program;
+ int stages = 2 + gs_present + 2 * tess_present;
- unsigned vs_size, gs_size;
- if (gs_present) {
- vs_size = avail_size / 3;
- avail_size -= vs_size;
- gs_size = avail_size / 2;
- avail_size -= gs_size;
- } else {
- vs_size = avail_size / 2;
- avail_size -= vs_size;
- gs_size = 0;
- }
- unsigned fs_size = avail_size;
+ /* Divide up the available space equally between stages. Because we
+ * round down (using floor division), there may be some left over
+ * space. We allocate that to the pixel shader stage.
+ */
+ unsigned size_per_stage = avail_size / stages;
+
+ unsigned vs_size = size_per_stage;
+ unsigned hs_size = tess_present ? size_per_stage : 0;
+ unsigned ds_size = tess_present ? size_per_stage : 0;
+ unsigned gs_size = gs_present ? size_per_stage : 0;
+ unsigned fs_size = avail_size - size_per_stage * (stages - 1);
gen7_emit_push_constant_state(brw, multiplier * vs_size,
+ multiplier * hs_size, multiplier * ds_size,
multiplier * gs_size, multiplier * fs_size);
/* From p115 of the Ivy Bridge PRM (3.2.1.4 3DSTATE_PUSH_CONSTANT_ALLOC_VS):
void
gen7_emit_push_constant_state(struct brw_context *brw, unsigned vs_size,
+ unsigned hs_size, unsigned ds_size,
unsigned gs_size, unsigned fs_size)
{
unsigned offset = 0;
- BEGIN_BATCH(6);
+ BEGIN_BATCH(10);
OUT_BATCH(_3DSTATE_PUSH_CONSTANT_ALLOC_VS << 16 | (2 - 2));
OUT_BATCH(vs_size | offset << GEN7_PUSH_CONSTANT_BUFFER_OFFSET_SHIFT);
offset += vs_size;
+ OUT_BATCH(_3DSTATE_PUSH_CONSTANT_ALLOC_HS << 16 | (2 - 2));
+ OUT_BATCH(hs_size | offset << GEN7_PUSH_CONSTANT_BUFFER_OFFSET_SHIFT);
+ offset += hs_size;
+
+ OUT_BATCH(_3DSTATE_PUSH_CONSTANT_ALLOC_DS << 16 | (2 - 2));
+ OUT_BATCH(ds_size | offset << GEN7_PUSH_CONSTANT_BUFFER_OFFSET_SHIFT);
+ offset += ds_size;
+
OUT_BATCH(_3DSTATE_PUSH_CONSTANT_ALLOC_GS << 16 | (2 - 2));
OUT_BATCH(gs_size | offset << GEN7_PUSH_CONSTANT_BUFFER_OFFSET_SHIFT);
offset += gs_size;
const struct brw_tracked_state gen7_push_constant_space = {
.dirty = {
.mesa = 0,
- .brw = BRW_NEW_CONTEXT | BRW_NEW_GEOMETRY_PROGRAM,
+ .brw = BRW_NEW_CONTEXT |
+ BRW_NEW_GEOMETRY_PROGRAM |
+ BRW_NEW_TESS_PROGRAMS,
},
.emit = gen7_allocate_push_constants,
};
static void
gen7_upload_urb(struct brw_context *brw)
{
+ const struct brw_device_info *devinfo = brw->intelScreen->devinfo;
const int push_size_kB =
(brw->gen >= 8 || (brw->is_haswell && brw->gt == 3)) ? 32 : 16;
unsigned gs_size = gs_present ? brw->gs.prog_data->base.urb_entry_size : 1;
unsigned gs_entry_size_bytes = gs_size * 64;
+ /* BRW_NEW_TESS_PROGRAMS */
+ const bool tess_present = brw->tess_eval_program;
+ /* BRW_NEW_TCS_PROG_DATA */
+ unsigned hs_size = tess_present ? brw->tcs.prog_data->base.urb_entry_size : 1;
+ unsigned hs_entry_size_bytes = hs_size * 64;
+ /* BRW_NEW_TES_PROG_DATA */
+ unsigned ds_size = tess_present ? brw->tes.prog_data->base.urb_entry_size : 1;
+ unsigned ds_entry_size_bytes = ds_size * 64;
+
/* If we're just switching between programs with the same URB requirements,
* skip the rest of the logic.
*/
!(brw->ctx.NewDriverState & BRW_NEW_URB_SIZE) &&
brw->urb.vsize == vs_size &&
brw->urb.gs_present == gs_present &&
- brw->urb.gsize == gs_size) {
+ brw->urb.gsize == gs_size &&
+ brw->urb.tess_present == tess_present &&
+ brw->urb.hsize == hs_size &&
+ brw->urb.dsize == ds_size) {
return;
}
brw->urb.vsize = vs_size;
brw->urb.gs_present = gs_present;
brw->urb.gsize = gs_size;
+ brw->urb.tess_present = tess_present;
+ brw->urb.hsize = hs_size;
+ brw->urb.dsize = ds_size;
/* From p35 of the Ivy Bridge PRM (section 1.7.1: 3DSTATE_URB_GS):
*
* VS Number of URB Entries must be divisible by 8 if the VS URB Entry
* Allocation Size is less than 9 512-bit URB entries.
*
- * Similar text exists for GS.
+ * Similar text exists for HS, DS and GS.
*/
unsigned vs_granularity = (vs_size < 9) ? 8 : 1;
+ unsigned hs_granularity = (hs_size < 9) ? 8 : 1;
+ unsigned ds_granularity = (ds_size < 9) ? 8 : 1;
unsigned gs_granularity = (gs_size < 9) ? 8 : 1;
/* URB allocations must be done in 8k chunks. */
* additional space it could actually make use of).
*/
- /* VS has a lower limit on the number of URB entries */
+ /* VS has a lower limit on the number of URB entries.
+ *
+ * From the Broadwell PRM, 3DSTATE_URB_VS instruction:
+ * "When tessellation is enabled, the VS Number of URB Entries must be
+ * greater than or equal to 192."
+ */
+ unsigned vs_min_entries =
+ tess_present && brw->gen == 8 ? 192 : brw->urb.min_vs_entries;
+
unsigned vs_chunks =
- ALIGN(brw->urb.min_vs_entries * vs_entry_size_bytes, chunk_size_bytes) /
- chunk_size_bytes;
+ DIV_ROUND_UP(vs_min_entries * vs_entry_size_bytes, chunk_size_bytes);
unsigned vs_wants =
- ALIGN(brw->urb.max_vs_entries * vs_entry_size_bytes,
- chunk_size_bytes) / chunk_size_bytes - vs_chunks;
+ DIV_ROUND_UP(brw->urb.max_vs_entries * vs_entry_size_bytes,
+ chunk_size_bytes) - vs_chunks;
unsigned gs_chunks = 0;
unsigned gs_wants = 0;
*
* (2) We can't allocate less than nr_gs_entries_granularity.
*/
- gs_chunks = ALIGN(MAX2(gs_granularity, 2) * gs_entry_size_bytes,
- chunk_size_bytes) / chunk_size_bytes;
- gs_wants =
- ALIGN(brw->urb.max_gs_entries * gs_entry_size_bytes,
- chunk_size_bytes) / chunk_size_bytes - gs_chunks;
+ gs_chunks = DIV_ROUND_UP(MAX2(gs_granularity, 2) * gs_entry_size_bytes,
+ chunk_size_bytes);
+ gs_wants = DIV_ROUND_UP(brw->urb.max_gs_entries * gs_entry_size_bytes,
+ chunk_size_bytes) - gs_chunks;
+ }
+
+ unsigned hs_chunks = 0;
+ unsigned hs_wants = 0;
+ unsigned ds_chunks = 0;
+ unsigned ds_wants = 0;
+
+ if (tess_present) {
+ hs_chunks =
+ DIV_ROUND_UP(hs_granularity * hs_entry_size_bytes,
+ chunk_size_bytes);
+ hs_wants =
+ DIV_ROUND_UP(devinfo->urb.max_hs_entries * hs_entry_size_bytes,
+ chunk_size_bytes) - hs_chunks;
+
+ ds_chunks =
+ DIV_ROUND_UP(devinfo->urb.min_ds_entries * ds_entry_size_bytes,
+ chunk_size_bytes);
+ ds_wants =
+ DIV_ROUND_UP(brw->urb.max_ds_entries * ds_entry_size_bytes,
+ chunk_size_bytes) - ds_chunks;
}
/* There should always be enough URB space to satisfy the minimum
* requirements of each stage.
*/
- unsigned total_needs = push_constant_chunks + vs_chunks + gs_chunks;
+ unsigned total_needs = push_constant_chunks +
+ vs_chunks + hs_chunks + ds_chunks + gs_chunks;
assert(total_needs <= urb_chunks);
/* Mete out remaining space (if any) in proportion to "wants". */
- unsigned total_wants = vs_wants + gs_wants;
+ unsigned total_wants = vs_wants + hs_wants + ds_wants + gs_wants;
unsigned remaining_space = urb_chunks - total_needs;
if (remaining_space > total_wants)
remaining_space = total_wants;
roundf(vs_wants * (((float) remaining_space) / total_wants));
vs_chunks += vs_additional;
remaining_space -= vs_additional;
+ total_wants -= vs_wants;
+
+ unsigned hs_additional = (unsigned)
+ round(hs_wants * (((double) remaining_space) / total_wants));
+ hs_chunks += hs_additional;
+ remaining_space -= hs_additional;
+ total_wants -= hs_wants;
+
+ unsigned ds_additional = (unsigned)
+ round(ds_wants * (((double) remaining_space) / total_wants));
+ ds_chunks += ds_additional;
+ remaining_space -= ds_additional;
+ total_wants -= ds_wants;
+
gs_chunks += remaining_space;
}
/* Sanity check that we haven't over-allocated. */
- assert(push_constant_chunks + vs_chunks + gs_chunks <= urb_chunks);
+ assert(push_constant_chunks +
+ vs_chunks + hs_chunks + ds_chunks + gs_chunks <= urb_chunks);
/* Finally, compute the number of entries that can fit in the space
* allocated to each stage.
*/
unsigned nr_vs_entries = vs_chunks * chunk_size_bytes / vs_entry_size_bytes;
+ unsigned nr_hs_entries = hs_chunks * chunk_size_bytes / hs_entry_size_bytes;
+ unsigned nr_ds_entries = ds_chunks * chunk_size_bytes / ds_entry_size_bytes;
unsigned nr_gs_entries = gs_chunks * chunk_size_bytes / gs_entry_size_bytes;
/* Since we rounded up when computing *_wants, this may be slightly more
* than the maximum allowed amount, so correct for that.
*/
nr_vs_entries = MIN2(nr_vs_entries, brw->urb.max_vs_entries);
+ nr_hs_entries = MIN2(nr_hs_entries, brw->urb.max_hs_entries);
+ nr_ds_entries = MIN2(nr_ds_entries, brw->urb.max_ds_entries);
nr_gs_entries = MIN2(nr_gs_entries, brw->urb.max_gs_entries);
/* Ensure that we program a multiple of the granularity. */
nr_vs_entries = ROUND_DOWN_TO(nr_vs_entries, vs_granularity);
+ nr_hs_entries = ROUND_DOWN_TO(nr_hs_entries, hs_granularity);
+ nr_ds_entries = ROUND_DOWN_TO(nr_ds_entries, ds_granularity);
nr_gs_entries = ROUND_DOWN_TO(nr_gs_entries, gs_granularity);
/* Finally, sanity check to make sure we have at least the minimum number
* of entries needed for each stage.
*/
- assert(nr_vs_entries >= brw->urb.min_vs_entries);
+ assert(nr_vs_entries >= vs_min_entries);
if (gs_present)
assert(nr_gs_entries >= 2);
+ if (tess_present) {
+ assert(nr_hs_entries >= 1);
+ assert(nr_ds_entries >= devinfo->urb.min_ds_entries);
+ }
/* Gen7 doesn't actually use brw->urb.nr_{vs,gs}_entries, but it seems
* better to put reasonable data in there rather than leave them
* uninitialized.
*/
brw->urb.nr_vs_entries = nr_vs_entries;
+ brw->urb.nr_hs_entries = nr_hs_entries;
+ brw->urb.nr_ds_entries = nr_ds_entries;
brw->urb.nr_gs_entries = nr_gs_entries;
/* Lay out the URB in the following order:
* - push constants
* - VS
+ * - HS
+ * - DS
* - GS
*/
brw->urb.vs_start = push_constant_chunks;
- brw->urb.gs_start = push_constant_chunks + vs_chunks;
+ brw->urb.hs_start = push_constant_chunks + vs_chunks;
+ brw->urb.ds_start = push_constant_chunks + vs_chunks + hs_chunks;
+ brw->urb.gs_start = push_constant_chunks + vs_chunks + hs_chunks +
+ ds_chunks;
if (brw->gen == 7 && !brw->is_haswell && !brw->is_baytrail)
gen7_emit_vs_workaround_flush(brw);
gen7_emit_urb_state(brw,
brw->urb.nr_vs_entries, vs_size, brw->urb.vs_start,
+ brw->urb.nr_hs_entries, hs_size, brw->urb.hs_start,
+ brw->urb.nr_ds_entries, ds_size, brw->urb.ds_start,
brw->urb.nr_gs_entries, gs_size, brw->urb.gs_start);
}
void
gen7_emit_urb_state(struct brw_context *brw,
- unsigned nr_vs_entries, unsigned vs_size,
- unsigned vs_start, unsigned nr_gs_entries,
+ unsigned nr_vs_entries,
+ unsigned vs_size, unsigned vs_start,
+ unsigned nr_hs_entries,
+ unsigned hs_size, unsigned hs_start,
+ unsigned nr_ds_entries,
+ unsigned ds_size, unsigned ds_start,
+ unsigned nr_gs_entries,
unsigned gs_size, unsigned gs_start)
{
BEGIN_BATCH(8);
((gs_size - 1) << GEN7_URB_ENTRY_SIZE_SHIFT) |
(gs_start << GEN7_URB_STARTING_ADDRESS_SHIFT));
- /* Allocate the HS and DS zero space - we don't use them. */
OUT_BATCH(_3DSTATE_URB_HS << 16 | (2 - 2));
- OUT_BATCH((0 << GEN7_URB_ENTRY_SIZE_SHIFT) |
- (vs_start << GEN7_URB_STARTING_ADDRESS_SHIFT));
+ OUT_BATCH(nr_hs_entries |
+ ((hs_size - 1) << GEN7_URB_ENTRY_SIZE_SHIFT) |
+ (hs_start << GEN7_URB_STARTING_ADDRESS_SHIFT));
OUT_BATCH(_3DSTATE_URB_DS << 16 | (2 - 2));
- OUT_BATCH((0 << GEN7_URB_ENTRY_SIZE_SHIFT) |
- (vs_start << GEN7_URB_STARTING_ADDRESS_SHIFT));
+ OUT_BATCH(nr_ds_entries |
+ ((ds_size - 1) << GEN7_URB_ENTRY_SIZE_SHIFT) |
+ (ds_start << GEN7_URB_STARTING_ADDRESS_SHIFT));
ADVANCE_BATCH();
}
.brw = BRW_NEW_CONTEXT |
BRW_NEW_URB_SIZE |
BRW_NEW_GEOMETRY_PROGRAM |
+ BRW_NEW_TESS_PROGRAMS |
BRW_NEW_GS_PROG_DATA |
+ BRW_NEW_TCS_PROG_DATA |
+ BRW_NEW_TES_PROG_DATA |
BRW_NEW_VS_PROG_DATA,
},
.emit = gen7_upload_urb,
dw1 |= GEN7_WM_KILL_ENABLE;
}
- if (_mesa_active_fragment_shader_has_atomic_ops(&brw->ctx)) {
- dw1 |= GEN7_WM_DISPATCH_ENABLE;
- }
-
/* _NEW_BUFFERS | _NEW_COLOR */
+ const bool active_fs_has_side_effects =
+ _mesa_active_fragment_shader_has_side_effects(&brw->ctx);
if (brw_color_buffer_write_enabled(brw) || writes_depth ||
- prog_data->base.nr_image_params ||
- dw1 & GEN7_WM_KILL_ENABLE) {
+ active_fs_has_side_effects || dw1 & GEN7_WM_KILL_ENABLE) {
dw1 |= GEN7_WM_DISPATCH_ENABLE;
}
if (multisampled_fbo) {
/* BRW_NEW_FS_PROG_DATA */
if (prog_data->early_fragment_tests)
dw1 |= GEN7_WM_EARLY_DS_CONTROL_PREPS;
- else if (prog_data->base.nr_image_params)
+ else if (active_fs_has_side_effects)
dw1 |= GEN7_WM_EARLY_DS_CONTROL_PSEXEC;
/* The "UAV access enable" bits are unnecessary on HSW because they only
*/
if (brw->is_haswell &&
!(brw_color_buffer_write_enabled(brw) || writes_depth) &&
- prog_data->base.nr_image_params)
+ active_fs_has_side_effects)
dw2 |= HSW_WM_UAV_ONLY;
BEGIN_BATCH(3);
}
/* Now emit 3DSTATE_VERTEX_BUFFERS and 3DSTATE_VERTEX_ELEMENTS packets. */
- unsigned nr_buffers = brw->vb.nr_buffers + brw->vs.prog_data->uses_vertexid;
+ const bool uses_draw_params =
+ brw->vs.prog_data->uses_basevertex ||
+ brw->vs.prog_data->uses_baseinstance;
+ const unsigned nr_buffers = brw->vb.nr_buffers +
+ uses_draw_params + brw->vs.prog_data->uses_drawid;
+
if (nr_buffers) {
assert(nr_buffers <= 33);
OUT_BATCH(buffer->bo->size);
}
- if (brw->vs.prog_data->uses_vertexid) {
+ if (uses_draw_params) {
OUT_BATCH(brw->vb.nr_buffers << GEN6_VB0_INDEX_SHIFT |
GEN7_VB0_ADDRESS_MODIFYENABLE |
mocs_wb << 16);
brw->draw.draw_params_offset);
OUT_BATCH(brw->draw.draw_params_bo->size);
}
+
+ if (brw->vs.prog_data->uses_drawid) {
+ OUT_BATCH((brw->vb.nr_buffers + 1) << GEN6_VB0_INDEX_SHIFT |
+ GEN7_VB0_ADDRESS_MODIFYENABLE |
+ mocs_wb << 16);
+ OUT_RELOC64(brw->draw.draw_id_bo, I915_GEM_DOMAIN_VERTEX, 0,
+ brw->draw.draw_id_offset);
+ OUT_BATCH(brw->draw.draw_id_bo->size);
+ }
ADVANCE_BATCH();
}
/* Normally we don't need an element for the SGVS attribute because the
* 3DSTATE_VF_SGVS instruction lets you store the generated attribute in an
- * element that is past the list in 3DSTATE_VERTEX_ELEMENTS. However if the
- * vertex ID is used then it needs an element for the base vertex buffer.
- * Additionally if there is an edge flag element then the SGVS can't be
- * inserted past that so we need a dummy element to ensure that the edge
- * flag is the last one.
+ * element that is past the list in 3DSTATE_VERTEX_ELEMENTS. However if
+ * we're using draw parameters then we need an element for the those
+ * values. Additionally if there is an edge flag element then the SGVS
+ * can't be inserted past that so we need a dummy element to ensure that
+ * the edge flag is the last one.
*/
- bool needs_sgvs_element = (brw->vs.prog_data->uses_vertexid ||
- (brw->vs.prog_data->uses_instanceid &&
- uses_edge_flag));
- unsigned nr_elements = brw->vb.nr_enabled + needs_sgvs_element;
+ const bool needs_sgvs_element = (brw->vs.prog_data->uses_basevertex ||
+ brw->vs.prog_data->uses_baseinstance ||
+ ((brw->vs.prog_data->uses_instanceid ||
+ brw->vs.prog_data->uses_vertexid) &&
+ uses_edge_flag));
+ const unsigned nr_elements =
+ brw->vb.nr_enabled + needs_sgvs_element + brw->vs.prog_data->uses_drawid;
/* The hardware allows one more VERTEX_ELEMENTS than VERTEX_BUFFERS,
* presumably for VertexID/InstanceID.
}
if (needs_sgvs_element) {
- if (brw->vs.prog_data->uses_vertexid) {
+ if (brw->vs.prog_data->uses_basevertex ||
+ brw->vs.prog_data->uses_baseinstance) {
OUT_BATCH(GEN6_VE0_VALID |
brw->vb.nr_buffers << GEN6_VE0_INDEX_SHIFT |
- BRW_SURFACEFORMAT_R32_UINT << BRW_VE0_FORMAT_SHIFT);
+ BRW_SURFACEFORMAT_R32G32_UINT << BRW_VE0_FORMAT_SHIFT);
OUT_BATCH((BRW_VE1_COMPONENT_STORE_SRC << BRW_VE1_COMPONENT_0_SHIFT) |
- (BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_1_SHIFT) |
+ (BRW_VE1_COMPONENT_STORE_SRC << BRW_VE1_COMPONENT_1_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_3_SHIFT));
} else {
}
}
+ if (brw->vs.prog_data->uses_drawid) {
+ OUT_BATCH(GEN6_VE0_VALID |
+ ((brw->vb.nr_buffers + 1) << GEN6_VE0_INDEX_SHIFT) |
+ (BRW_SURFACEFORMAT_R32_UINT << BRW_VE0_FORMAT_SHIFT));
+ OUT_BATCH((BRW_VE1_COMPONENT_STORE_SRC << BRW_VE1_COMPONENT_0_SHIFT) |
+ (BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_1_SHIFT) |
+ (BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) |
+ (BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_3_SHIFT));
+ }
+
if (gen6_edgeflag_input) {
uint32_t format =
brw_get_vertex_surface_type(brw, gen6_edgeflag_input->glarray);
OUT_BATCH(buffer->step_rate);
ADVANCE_BATCH();
}
+
+ if (brw->vs.prog_data->uses_drawid) {
+ const unsigned element = brw->vb.nr_enabled + needs_sgvs_element;
+ BEGIN_BATCH(3);
+ OUT_BATCH(_3DSTATE_VF_INSTANCING << 16 | (3 - 2));
+ OUT_BATCH(element);
+ OUT_BATCH(0);
+ ADVANCE_BATCH();
+ }
}
const struct brw_tracked_state gen8_vertices = {
{
struct gl_context *ctx = &brw->ctx;
const struct brw_stage_state *stage_state = &brw->tes.base;
- /* BRW_NEW_TESS_EVAL_PROGRAM */
+ /* BRW_NEW_TESS_PROGRAMS */
bool active = brw->tess_eval_program;
- assert(!active || brw->tess_ctrl_program);
/* BRW_NEW_TES_PROG_DATA */
const struct brw_tes_prog_data *tes_prog_data = brw->tes.prog_data;
.dirty = {
.mesa = 0,
.brw = BRW_NEW_BATCH |
- BRW_NEW_TESS_EVAL_PROGRAM |
+ BRW_NEW_TESS_PROGRAMS |
BRW_NEW_TES_PROG_DATA,
},
.emit = gen8_upload_ds_state,
gen8_upload_hs_state(struct brw_context *brw)
{
const struct brw_stage_state *stage_state = &brw->tcs.base;
- /* BRW_NEW_TESS_CTRL_PROGRAM */
- bool active = brw->tess_ctrl_program;
- assert(!active || brw->tess_eval_program);
+ /* BRW_NEW_TESS_PROGRAMS */
+ bool active = brw->tess_eval_program;
/* BRW_NEW_HS_PROG_DATA */
const struct brw_vue_prog_data *prog_data = &brw->tcs.prog_data->base;
.mesa = 0,
.brw = BRW_NEW_BATCH |
BRW_NEW_TCS_PROG_DATA |
- BRW_NEW_TESS_CTRL_PROGRAM,
+ BRW_NEW_TESS_PROGRAMS,
},
.emit = gen8_upload_hs_state,
};
*
* BRW_NEW_FS_PROG_DATA | BRW_NEW_FRAGMENT_PROGRAM | _NEW_BUFFERS | _NEW_COLOR
*/
- if ((_mesa_active_fragment_shader_has_atomic_ops(&brw->ctx) ||
- prog_data->base.nr_image_params) &&
+ if (_mesa_active_fragment_shader_has_side_effects(&brw->ctx) &&
!brw_color_buffer_write_enabled(brw))
dw1 |= GEN8_PSX_SHADER_HAS_UAV;
/* BRW_NEW_FS_PROG_DATA */
if (brw->wm.prog_data->early_fragment_tests)
dw1 |= GEN7_WM_EARLY_DS_CONTROL_PREPS;
- else if (brw->wm.prog_data->base.nr_image_params)
+ else if (_mesa_active_fragment_shader_has_side_effects(&brw->ctx))
dw1 |= GEN7_WM_EARLY_DS_CONTROL_PSEXEC;
BEGIN_BATCH(2);
if (brw->gen < 9)
return false;
- if (src_buffer->handle == dst_buffer->handle &&
- _mesa_regions_overlap(src_x, src_y, src_x + w, src_y + h,
- dst_x, dst_y, dst_x + w, dst_y + h))
- return false;
-
/* Enable fast copy blit only if the surfaces are Yf/Ys tiled.
* FIXME: Based on performance data, remove this condition later to
* enable for all types of surfaces.
if ((dst_offset | src_offset) & 63)
return false;
- /* Color depth greater than 128 bits not supported. */
- if (cpp > 16)
+ /* Color depths which are not power of 2 or greater than 128 bits are
+ * not supported.
+ */
+ if (!_mesa_is_pow_two(cpp) || cpp > 16)
return false;
/* For Fast Copy Blits the pitch cannot be a negative number. So, bit 15
dst_offset, dst_pitch,
dst_tiling, dst_tr_mode,
w, h, cpp);
- assert(use_fast_copy_blit ||
- (src_tr_mode == INTEL_MIPTREE_TRMODE_NONE &&
- dst_tr_mode == INTEL_MIPTREE_TRMODE_NONE));
+ if (!use_fast_copy_blit &&
+ (src_tr_mode != INTEL_MIPTREE_TRMODE_NONE ||
+ dst_tr_mode != INTEL_MIPTREE_TRMODE_NONE))
+ return false;
if (use_fast_copy_blit) {
/* When two sequential fast copy blits have different source surfaces,
_mesa_buffer_unmap_all_mappings(ctx, obj);
drm_intel_bo_unreference(intel_obj->buffer);
- free(intel_obj);
+ _mesa_delete_buffer_object(ctx, obj);
}
ctx->Extensions.ARB_point_sprite = true;
ctx->Extensions.ARB_seamless_cube_map = true;
ctx->Extensions.ARB_shader_bit_encoding = true;
+ ctx->Extensions.ARB_shader_draw_parameters = true;
ctx->Extensions.ARB_shader_texture_lod = true;
ctx->Extensions.ARB_shadow = true;
ctx->Extensions.ARB_sync = true;
ctx->Extensions.ARB_shader_image_load_store = true;
ctx->Extensions.ARB_shader_image_size = true;
ctx->Extensions.ARB_shader_texture_image_samples = true;
+ ctx->Extensions.ARB_tessellation_shader = true;
ctx->Extensions.ARB_texture_compression_bptc = true;
ctx->Extensions.ARB_texture_view = true;
ctx->Extensions.ARB_shader_storage_buffer_object = true;
ctx->Extensions.ARB_transform_feedback3 = true;
ctx->Extensions.ARB_transform_feedback_instanced = true;
+ if (ctx->Const.MaxComputeWorkGroupSize[0] >= 1024)
+ ctx->Extensions.ARB_compute_shader = true;
+
if (brw->intelScreen->cmd_parser_version >= 2)
brw->predicate.supported = true;
}
ctx->Extensions.ARB_viewport_array = true;
ctx->Extensions.AMD_vertex_shader_viewport_index = true;
ctx->Extensions.ARB_shader_subroutine = true;
- if (ctx->Const.MaxComputeWorkGroupSize[0] >= 1024)
- ctx->Extensions.ARB_compute_shader = true;
}
}
{
if (brw->has_llc &&
/* It's probably not worth swapping to the blit ring because of
- * all the overhead involved.
+ * all the overhead involved. But, we must use blitter for the
+ * surfaces with INTEL_MIPTREE_TRMODE_{YF,YS}.
*/
- !(mode & GL_MAP_WRITE_BIT) &&
+ (!(mode & GL_MAP_WRITE_BIT) ||
+ mt->tr_mode != INTEL_MIPTREE_TRMODE_NONE) &&
!mt->compressed &&
(mt->tiling == I915_TILING_X ||
/* Prior to Sandybridge, the blitter can't handle Y tiling */
- (brw->gen >= 6 && mt->tiling == I915_TILING_Y)) &&
+ (brw->gen >= 6 && mt->tiling == I915_TILING_Y) ||
+ /* Fast copy blit on skl+ supports all tiling formats. */
+ brw->gen >= 9) &&
can_blit_slice(mt, level, slice))
return true;
intel_miptree_map_movntdqa(brw, mt, map, level, slice);
#endif
} else {
+ /* intel_miptree_map_gtt() doesn't support surfaces with Yf/Ys tiling. */
+ assert(mt->tr_mode == INTEL_MIPTREE_TRMODE_NONE);
intel_miptree_map_gtt(brw, mt, map, level, slice);
}
}
if (ctx->Pixel.ZoomX != 1.0F || ctx->Pixel.ZoomY != 1.0F) {
- perf_debug("glCopyPixles(): Unsupported pixel zoom\n");
+ perf_debug("glCopyPixels(): Unsupported pixel zoom\n");
return false;
}
fb->Visual.samples = num_samples;
}
- if (mesaVis->redBits == 5) {
+ if (mesaVis->redBits == 5)
rgbFormat = MESA_FORMAT_B5G6R5_UNORM;
- } else {
- if (mesaVis->alphaBits == 0)
- rgbFormat = MESA_FORMAT_B8G8R8X8_SRGB;
- else
- rgbFormat = MESA_FORMAT_B8G8R8A8_SRGB;
+ else if (mesaVis->sRGBCapable)
+ rgbFormat = MESA_FORMAT_B8G8R8A8_SRGB;
+ else if (mesaVis->alphaBits == 0)
+ rgbFormat = MESA_FORMAT_B8G8R8X8_UNORM;
+ else {
+ rgbFormat = MESA_FORMAT_B8G8R8A8_SRGB;
fb->Visual.sRGBCapable = true;
}
switch (screen->devinfo->gen) {
case 9:
case 8:
+ psp->max_gl_core_version = 33;
+ psp->max_gl_compat_version = 30;
+ psp->max_gl_es1_version = 11;
+ psp->max_gl_es2_version = 31;
+ break;
case 7:
case 6:
psp->max_gl_core_version = 33;
if (INTEL_DEBUG & DEBUG_AUB)
drm_intel_bufmgr_gem_set_aub_dump(intelScreen->bufmgr, true);
- intelScreen->hw_must_use_separate_stencil = intelScreen->devinfo->gen >= 7;
-
intelScreen->hw_has_swizzling = intel_detect_swizzling(intelScreen);
intelScreen->hw_has_timestamp = intel_detect_timestamp(intelScreen);
intelScreen->compiler = brw_compiler_create(intelScreen,
intelScreen->devinfo);
+ intelScreen->program_id = 1;
if (intelScreen->devinfo->has_resource_streamer) {
int val = -1;
bool no_hw;
- bool hw_must_use_separate_stencil;
-
bool hw_has_swizzling;
int hw_has_timestamp;
#include "drivers/common/meta.h"
-const char const *nouveau_vendor_string = "Nouveau";
+const char * const nouveau_vendor_string = "Nouveau";
const char *
nouveau_get_renderer_string(unsigned chipset)
#define nouveau_error(format, ...) \
fprintf(stderr, "%s: " format, __func__, ## __VA_ARGS__)
-extern const char const *nouveau_vendor_string;
+extern const char * const nouveau_vendor_string;
const char *
nouveau_get_renderer_string(unsigned chipset);
#include "main/renderbuffer.h"
#include "swrast/s_renderbuffer.h"
+#include <nvif/class.h>
+#include <nvif/cl0080.h>
+
static const __DRIextension *nouveau_screen_extensions[];
static void
dri_screen->driverPrivate = screen;
/* Open the DRM device. */
- ret = nouveau_device_wrap(dri_screen->fd, 0, &screen->device);
+ ret = nouveau_drm_new(dri_screen->fd, &screen->drm);
if (ret) {
nouveau_error("Error opening the DRM device.\n");
goto fail;
}
+ ret = nouveau_device_new(&screen->drm->client, NV_DEVICE,
+ &(struct nv_device_v0) {
+ .device = ~0ULL,
+ }, sizeof(struct nv_device_v0),
+ &screen->device);
+ if (ret) {
+ nouveau_error("Error creating device object.\n");
+ goto fail;
+ }
+
/* Choose the card specific function pointers. */
switch (screen->device->chipset & 0xf0) {
case 0x00:
return;
nouveau_device_del(&screen->device);
+ nouveau_drm_del(&screen->drm);
free(screen);
dri_screen->driverPrivate = NULL;
struct nouveau_screen {
__DRIscreen *dri_screen;
+ struct nouveau_drm *drm;
struct nouveau_device *device;
const struct nouveau_driver *driver;
};
radeon_bo_unref(radeon_obj->bo);
}
- free(radeon_obj);
+ _mesa_delete_buffer_object(ctx, obj);
}
OSMesaCreateContextExt( GLenum format, GLint depthBits, GLint stencilBits,
GLint accumBits, OSMesaContext sharelist )
{
+ int attribs[100], n = 0;
+
+ attribs[n++] = OSMESA_FORMAT;
+ attribs[n++] = format;
+ attribs[n++] = OSMESA_DEPTH_BITS;
+ attribs[n++] = depthBits;
+ attribs[n++] = OSMESA_STENCIL_BITS;
+ attribs[n++] = stencilBits;
+ attribs[n++] = OSMESA_ACCUM_BITS;
+ attribs[n++] = accumBits;
+ attribs[n++] = 0;
+
+ return OSMesaCreateContextAttribs(attribs, sharelist);
+}
+
+
+/**
+ * New in Mesa 11.2
+ *
+ * Create context with attribute list.
+ */
+GLAPI OSMesaContext GLAPIENTRY
+OSMesaCreateContextAttribs(const int *attribList, OSMesaContext sharelist)
+{
OSMesaContext osmesa;
struct dd_function_table functions;
GLint rind, gind, bind, aind;
GLint redBits = 0, greenBits = 0, blueBits = 0, alphaBits =0;
+ GLenum format = OSMESA_RGBA;
+ GLint depthBits = 0, stencilBits = 0, accumBits = 0;
+ int profile = OSMESA_COMPAT_PROFILE, version_major = 1, version_minor = 0;
+ gl_api api_profile = API_OPENGL_COMPAT;
+ int i;
+
+ for (i = 0; attribList[i]; i += 2) {
+ switch (attribList[i]) {
+ case OSMESA_FORMAT:
+ format = attribList[i+1];
+ switch (format) {
+ case OSMESA_COLOR_INDEX:
+ case OSMESA_RGBA:
+ case OSMESA_BGRA:
+ case OSMESA_ARGB:
+ case OSMESA_RGB:
+ case OSMESA_BGR:
+ case OSMESA_RGB_565:
+ /* legal */
+ break;
+ default:
+ return NULL;
+ }
+ break;
+ case OSMESA_DEPTH_BITS:
+ depthBits = attribList[i+1];
+ if (depthBits < 0)
+ return NULL;
+ break;
+ case OSMESA_STENCIL_BITS:
+ stencilBits = attribList[i+1];
+ if (stencilBits < 0)
+ return NULL;
+ break;
+ case OSMESA_ACCUM_BITS:
+ accumBits = attribList[i+1];
+ if (accumBits < 0)
+ return NULL;
+ break;
+ case OSMESA_PROFILE:
+ profile = attribList[i+1];
+ if (profile == OSMESA_COMPAT_PROFILE)
+ api_profile = API_OPENGL_COMPAT;
+ else if (profile == OSMESA_CORE_PROFILE)
+ api_profile = API_OPENGL_CORE;
+ else
+ return NULL;
+ break;
+ case OSMESA_CONTEXT_MAJOR_VERSION:
+ version_major = attribList[i+1];
+ if (version_major < 1)
+ return NULL;
+ break;
+ case OSMESA_CONTEXT_MINOR_VERSION:
+ version_minor = attribList[i+1];
+ if (version_minor < 0)
+ return NULL;
+ break;
+ case 0:
+ /* end of list */
+ break;
+ default:
+ fprintf(stderr, "Bad attribute in OSMesaCreateContextAttribs()\n");
+ return NULL;
+ }
+ }
rind = gind = bind = aind = 0;
if (format==OSMESA_RGBA) {
functions.UpdateState = osmesa_update_state;
if (!_mesa_initialize_context(&osmesa->mesa,
- API_OPENGL_COMPAT,
+ api_profile,
osmesa->gl_visual,
sharelist ? &sharelist->mesa
: (struct gl_context *) NULL,
_mesa_compute_version(ctx);
+ if (ctx->Version < version_major * 10 + version_minor) {
+ _mesa_destroy_visual(osmesa->gl_visual);
+ _mesa_free_context_data(ctx);
+ free(osmesa);
+ return NULL;
+ }
+
/* Exec table initialization requires the version to be computed */
_mesa_initialize_dispatch_tables(ctx);
_mesa_initialize_vbo_vtxfmt(ctx);
static struct name_function functions[] = {
{ "OSMesaCreateContext", (OSMESAproc) OSMesaCreateContext },
{ "OSMesaCreateContextExt", (OSMESAproc) OSMesaCreateContextExt },
+ { "OSMesaCreateContextAttribs", (OSMESAproc) OSMesaCreateContextAttribs },
{ "OSMesaDestroyContext", (OSMESAproc) OSMesaDestroyContext },
{ "OSMesaMakeCurrent", (OSMESAproc) OSMesaMakeCurrent },
{ "OSMesaGetCurrentContext", (OSMESAproc) OSMesaGetCurrentContext },
typedef struct {
AEarray arrays[32];
AEattrib attribs[VERT_ATTRIB_MAX + 1];
- GLuint NewState;
+ GLbitfield NewState;
/* List of VBOs we need to map before executing ArrayElements */
struct gl_buffer_object *vbo[VERT_ATTRIB_MAX];
void
-_ae_invalidate_state(struct gl_context *ctx, GLuint new_state)
+_ae_invalidate_state(struct gl_context *ctx, GLbitfield new_state)
{
AEcontext *actx = AE_CONTEXT(ctx);
extern GLboolean _ae_create_context( struct gl_context *ctx );
extern void _ae_destroy_context( struct gl_context *ctx );
-extern void _ae_invalidate_state( struct gl_context *ctx, GLuint new_state );
+extern void _ae_invalidate_state( struct gl_context *ctx, GLbitfield new_state );
extern void GLAPIENTRY _ae_ArrayElement( GLint elt );
/* May optionally be called before a batch of element calls:
void GLAPIENTRY
_mesa_Vertex3dv( const GLdouble *v )
{
- VERTEX3( (GLfloat) v[0], (GLfloat) v[1], (GLfloat) v[2] );
+ if (v[2] == 0.0)
+ VERTEX2( (GLfloat) v[0], (GLfloat) v[1] );
+ else
+ VERTEX3( (GLfloat) v[0], (GLfloat) v[1], (GLfloat) v[2] );
}
void GLAPIENTRY
return GL_TRUE;
}
+static GLboolean
+valid_draw_indirect_parameters(struct gl_context *ctx,
+ const char *name,
+ GLintptr drawcount)
+{
+ /* From the ARB_indirect_parameters specification:
+ * "INVALID_VALUE is generated by MultiDrawArraysIndirectCountARB or
+ * MultiDrawElementsIndirectCountARB if <drawcount> is not a multiple of
+ * four."
+ */
+ if (drawcount & 3) {
+ _mesa_error(ctx, GL_INVALID_VALUE,
+ "%s(drawcount is not a multiple of 4)", name);
+ return GL_FALSE;
+ }
+
+ /* From the ARB_indirect_parameters specification:
+ * "INVALID_OPERATION is generated by MultiDrawArraysIndirectCountARB or
+ * MultiDrawElementsIndirectCountARB if no buffer is bound to the
+ * PARAMETER_BUFFER_ARB binding point."
+ */
+ if (!_mesa_is_bufferobj(ctx->ParameterBuffer)) {
+ _mesa_error(ctx, GL_INVALID_OPERATION,
+ "%s: no buffer bound to PARAMETER_BUFFER", name);
+ return GL_FALSE;
+ }
+
+ if (_mesa_check_disallowed_mapping(ctx->ParameterBuffer)) {
+ _mesa_error(ctx, GL_INVALID_OPERATION,
+ "%s(PARAMETER_BUFFER is mapped)", name);
+ return GL_FALSE;
+ }
+
+ /* From the ARB_indirect_parameters specification:
+ * "INVALID_OPERATION is generated by MultiDrawArraysIndirectCountARB or
+ * MultiDrawElementsIndirectCountARB if reading a <sizei> typed value
+ * from the buffer bound to the PARAMETER_BUFFER_ARB target at the offset
+ * specified by <drawcount> would result in an out-of-bounds access."
+ */
+ if (ctx->ParameterBuffer->Size < drawcount + sizeof(GLsizei)) {
+ _mesa_error(ctx, GL_INVALID_OPERATION,
+ "%s(PARAMETER_BUFFER too small)", name);
+ return GL_FALSE;
+ }
+
+ return GL_TRUE;
+}
+
+GLboolean
+_mesa_validate_MultiDrawArraysIndirectCount(struct gl_context *ctx,
+ GLenum mode,
+ GLintptr indirect,
+ GLintptr drawcount,
+ GLsizei maxdrawcount,
+ GLsizei stride)
+{
+ GLsizeiptr size = 0;
+ const unsigned drawArraysNumParams = 4;
+
+ FLUSH_CURRENT(ctx, 0);
+
+ /* caller has converted stride==0 to drawArraysNumParams * sizeof(GLuint) */
+ assert(stride != 0);
+
+ if (!valid_draw_indirect_multi(ctx, maxdrawcount, stride,
+ "glMultiDrawArraysIndirectCountARB"))
+ return GL_FALSE;
+
+ /* number of bytes of the indirect buffer which will be read */
+ size = maxdrawcount
+ ? (maxdrawcount - 1) * stride + drawArraysNumParams * sizeof(GLuint)
+ : 0;
+
+ if (!valid_draw_indirect(ctx, mode, (void *)indirect, size,
+ "glMultiDrawArraysIndirectCountARB"))
+ return GL_FALSE;
+
+ return valid_draw_indirect_parameters(
+ ctx, "glMultiDrawArraysIndirectCountARB", drawcount);
+}
+
+GLboolean
+_mesa_validate_MultiDrawElementsIndirectCount(struct gl_context *ctx,
+ GLenum mode, GLenum type,
+ GLintptr indirect,
+ GLintptr drawcount,
+ GLsizei maxdrawcount,
+ GLsizei stride)
+{
+ GLsizeiptr size = 0;
+ const unsigned drawElementsNumParams = 5;
+
+ FLUSH_CURRENT(ctx, 0);
+
+ /* caller has converted stride==0 to drawElementsNumParams * sizeof(GLuint) */
+ assert(stride != 0);
+
+ if (!valid_draw_indirect_multi(ctx, maxdrawcount, stride,
+ "glMultiDrawElementsIndirectCountARB"))
+ return GL_FALSE;
+
+ /* number of bytes of the indirect buffer which will be read */
+ size = maxdrawcount
+ ? (maxdrawcount - 1) * stride + drawElementsNumParams * sizeof(GLuint)
+ : 0;
+
+ if (!valid_draw_indirect_elements(ctx, mode, type,
+ (void *)indirect, size,
+ "glMultiDrawElementsIndirectCountARB"))
+ return GL_FALSE;
+
+ return valid_draw_indirect_parameters(
+ ctx, "glMultiDrawElementsIndirectCountARB", drawcount);
+}
+
static bool
check_valid_to_compute(struct gl_context *ctx, const char *function)
{
GLsizei stride);
extern GLboolean
+_mesa_validate_MultiDrawArraysIndirectCount(struct gl_context *ctx,
+ GLenum mode,
+ GLintptr indirect,
+ GLintptr drawcount,
+ GLsizei maxdrawcount,
+ GLsizei stride);
+
+extern GLboolean
+_mesa_validate_MultiDrawElementsIndirectCount(struct gl_context *ctx,
+ GLenum mode, GLenum type,
+ GLintptr indirect,
+ GLintptr drawcount,
+ GLsizei maxdrawcount,
+ GLsizei stride);
+
+extern GLboolean
_mesa_validate_DispatchCompute(struct gl_context *ctx,
const GLuint *num_groups);
prog->RefCount--;
if (prog->RefCount <= 0) {
assert(prog != &DummyShader);
- free(prog);
+ _mesa_delete_ati_fragment_shader(ctx, prog);
}
}
}
ctx->ATIFragmentShader.Current->isValid = GL_FALSE;
ctx->ATIFragmentShader.Current->swizzlerq = 0;
ctx->ATIFragmentShader.Compiling = 1;
+#if MESA_DEBUG_ATI_FS
+ _mesa_debug(ctx, "%s %u\n", __func__, ctx->ATIFragmentShader.Current->Id);
+#endif
}
void GLAPIENTRY
}
/* extra checks for multisample copies... */
if (readFb->Visual.samples > 0 || drawFb->Visual.samples > 0) {
- /* color formats must match */
- if (!compatible_resolve_formats(colorReadRb, colorDrawRb)) {
+ /* color formats must match on GLES. This isn't checked on
+ * desktop GL because the GL 4.4 spec was changed to allow it.
+ * In the section entitled “Changes in the released
+ * Specification of July 22, 2013” it says:
+ *
+ * “Relax BlitFramebuffer in section 18.3.1 so that format
+ * conversion can take place during multisample blits, since
+ * drivers already allow this and some apps depend on it.”
+ */
+ if (_mesa_is_gles(ctx) &&
+ !compatible_resolve_formats(colorReadRb, colorDrawRb)) {
_mesa_error(ctx, GL_INVALID_OPERATION,
"%s(bad src/dst multisample pixel formats)", func);
return;
return &ctx->DrawIndirectBuffer;
}
break;
+ case GL_PARAMETER_BUFFER_ARB:
+ if (_mesa_has_ARB_indirect_parameters(ctx)) {
+ return &ctx->ParameterBuffer;
+ }
+ break;
case GL_DISPATCH_INDIRECT_BUFFER:
if (_mesa_has_compute_shaders(ctx)) {
return &ctx->DispatchIndirectBuffer;
*
* Default callback for the \c dd_function_table::DeleteBuffer() hook.
*/
-static void
+void
_mesa_delete_buffer_object(struct gl_context *ctx,
struct gl_buffer_object *bufObj)
{
_mesa_reference_buffer_object(ctx, &ctx->DrawIndirectBuffer,
ctx->Shared->NullBufferObj);
+ _mesa_reference_buffer_object(ctx, &ctx->ParameterBuffer,
+ ctx->Shared->NullBufferObj);
+
_mesa_reference_buffer_object(ctx, &ctx->DispatchIndirectBuffer,
ctx->Shared->NullBufferObj);
_mesa_reference_buffer_object(ctx, &ctx->DrawIndirectBuffer, NULL);
+ _mesa_reference_buffer_object(ctx, &ctx->ParameterBuffer, NULL);
+
_mesa_reference_buffer_object(ctx, &ctx->DispatchIndirectBuffer, NULL);
for (i = 0; i < MAX_COMBINED_UNIFORM_BUFFERS; i++) {
{
GET_CURRENT_CONTEXT(ctx);
- if (MESA_VERBOSE & VERBOSE_API)
+ if (MESA_VERBOSE & VERBOSE_API) {
_mesa_debug(ctx, "glBindBuffer(%s, %u)\n",
_mesa_enum_to_string(target), buffer);
+ }
bind_buffer_object(ctx, target, buffer);
}
_mesa_BindBuffer( GL_DRAW_INDIRECT_BUFFER, 0 );
}
+ /* unbind ARB_indirect_parameters binding point */
+ if (ctx->ParameterBuffer == bufObj) {
+ _mesa_BindBuffer(GL_PARAMETER_BUFFER_ARB, 0);
+ }
+
/* unbind ARB_compute_shader binding point */
if (ctx->DispatchIndirectBuffer == bufObj) {
_mesa_BindBuffer(GL_DISPATCH_INDIRECT_BUFFER, 0);
{
bool valid_usage;
- if (MESA_VERBOSE & VERBOSE_API)
+ if (MESA_VERBOSE & VERBOSE_API) {
_mesa_debug(ctx, "%s(%s, %ld, %p, %s)\n",
func,
_mesa_enum_to_string(target),
(long int) size, data,
_mesa_enum_to_string(usage));
+ }
if (size < 0) {
_mesa_error(ctx, GL_INVALID_VALUE, "%s(size < 0)", func);
}
static void
-bind_uniform_buffers_base(struct gl_context *ctx, GLuint first, GLsizei count,
- const GLuint *buffers)
-{
- GLint i;
-
- if (!error_check_bind_uniform_buffers(ctx, first, count, "glBindBuffersBase"))
- return;
-
- /* Assume that at least one binding will be changed */
- FLUSH_VERTICES(ctx, 0);
- ctx->NewDriverState |= ctx->DriverFlags.NewUniformBuffer;
-
- if (!buffers) {
- /* The ARB_multi_bind spec says:
- *
- * "If <buffers> is NULL, all bindings from <first> through
- * <first>+<count>-1 are reset to their unbound (zero) state."
- */
- unbind_uniform_buffers(ctx, first, count);
- return;
- }
-
- /* Note that the error semantics for multi-bind commands differ from
- * those of other GL commands.
- *
- * The Issues section in the ARB_multi_bind spec says:
- *
- * "(11) Typically, OpenGL specifies that if an error is generated by a
- * command, that command has no effect. This is somewhat
- * unfortunate for multi-bind commands, because it would require a
- * first pass to scan the entire list of bound objects for errors
- * and then a second pass to actually perform the bindings.
- * Should we have different error semantics?
- *
- * RESOLVED: Yes. In this specification, when the parameters for
- * one of the <count> binding points are invalid, that binding point
- * is not updated and an error will be generated. However, other
- * binding points in the same command will be updated if their
- * parameters are valid and no other error occurs."
- */
-
- _mesa_begin_bufferobj_lookups(ctx);
-
- for (i = 0; i < count; i++) {
- struct gl_uniform_buffer_binding *binding =
- &ctx->UniformBufferBindings[first + i];
- struct gl_buffer_object *bufObj;
-
- if (binding->BufferObject && binding->BufferObject->Name == buffers[i])
- bufObj = binding->BufferObject;
- else
- bufObj = _mesa_multi_bind_lookup_bufferobj(ctx, buffers, i,
- "glBindBuffersBase");
-
- if (bufObj) {
- if (bufObj == ctx->Shared->NullBufferObj)
- set_ubo_binding(ctx, binding, bufObj, -1, -1, GL_TRUE);
- else
- set_ubo_binding(ctx, binding, bufObj, 0, 0, GL_TRUE);
- }
- }
-
- _mesa_end_bufferobj_lookups(ctx);
-}
-
-static void
-bind_shader_storage_buffers_base(struct gl_context *ctx, GLuint first,
- GLsizei count, const GLuint *buffers)
-{
- GLint i;
-
- if (!error_check_bind_shader_storage_buffers(ctx, first, count,
- "glBindBuffersBase"))
- return;
-
- /* Assume that at least one binding will be changed */
- FLUSH_VERTICES(ctx, 0);
- ctx->NewDriverState |= ctx->DriverFlags.NewShaderStorageBuffer;
-
- if (!buffers) {
- /* The ARB_multi_bind spec says:
- *
- * "If <buffers> is NULL, all bindings from <first> through
- * <first>+<count>-1 are reset to their unbound (zero) state."
- */
- unbind_shader_storage_buffers(ctx, first, count);
- return;
- }
-
- /* Note that the error semantics for multi-bind commands differ from
- * those of other GL commands.
- *
- * The Issues section in the ARB_multi_bind spec says:
- *
- * "(11) Typically, OpenGL specifies that if an error is generated by a
- * command, that command has no effect. This is somewhat
- * unfortunate for multi-bind commands, because it would require a
- * first pass to scan the entire list of bound objects for errors
- * and then a second pass to actually perform the bindings.
- * Should we have different error semantics?
- *
- * RESOLVED: Yes. In this specification, when the parameters for
- * one of the <count> binding points are invalid, that binding point
- * is not updated and an error will be generated. However, other
- * binding points in the same command will be updated if their
- * parameters are valid and no other error occurs."
- */
-
- _mesa_begin_bufferobj_lookups(ctx);
-
- for (i = 0; i < count; i++) {
- struct gl_shader_storage_buffer_binding *binding =
- &ctx->ShaderStorageBufferBindings[first + i];
- struct gl_buffer_object *bufObj;
-
- if (binding->BufferObject && binding->BufferObject->Name == buffers[i])
- bufObj = binding->BufferObject;
- else
- bufObj = _mesa_multi_bind_lookup_bufferobj(ctx, buffers, i,
- "glBindBuffersBase");
-
- if (bufObj) {
- if (bufObj == ctx->Shared->NullBufferObj)
- set_ssbo_binding(ctx, binding, bufObj, -1, -1, GL_TRUE);
- else
- set_ssbo_binding(ctx, binding, bufObj, 0, 0, GL_TRUE);
- }
- }
-
- _mesa_end_bufferobj_lookups(ctx);
-}
-
-static void
-bind_uniform_buffers_range(struct gl_context *ctx, GLuint first, GLsizei count,
- const GLuint *buffers,
- const GLintptr *offsets, const GLsizeiptr *sizes)
+bind_uniform_buffers(struct gl_context *ctx, GLuint first, GLsizei count,
+ const GLuint *buffers,
+ bool range,
+ const GLintptr *offsets, const GLsizeiptr *sizes,
+ const char *caller)
{
GLint i;
- if (!error_check_bind_uniform_buffers(ctx, first, count,
- "glBindBuffersRange"))
+ if (!error_check_bind_uniform_buffers(ctx, first, count, caller))
return;
/* Assume that at least one binding will be changed */
struct gl_uniform_buffer_binding *binding =
&ctx->UniformBufferBindings[first + i];
struct gl_buffer_object *bufObj;
+ GLintptr offset = 0;
+ GLsizeiptr size = 0;
- if (!bind_buffers_check_offset_and_size(ctx, i, offsets, sizes))
- continue;
+ if (range) {
+ if (!bind_buffers_check_offset_and_size(ctx, i, offsets, sizes))
+ continue;
- /* The ARB_multi_bind spec says:
- *
- * "An INVALID_VALUE error is generated by BindBuffersRange if any
- * pair of values in <offsets> and <sizes> does not respectively
- * satisfy the constraints described for those parameters for the
- * specified target, as described in section 6.7.1 (per binding)."
- *
- * Section 6.7.1 refers to table 6.5, which says:
- *
- * "┌───────────────────────────────────────────────────────────────┐
- * │ Uniform buffer array bindings (see sec. 7.6) │
- * ├─────────────────────┬─────────────────────────────────────────┤
- * │ ... │ ... │
- * │ offset restriction │ multiple of value of UNIFORM_BUFFER_- │
- * │ │ OFFSET_ALIGNMENT │
- * │ ... │ ... │
- * │ size restriction │ none │
- * └─────────────────────┴─────────────────────────────────────────┘"
- */
- if (offsets[i] & (ctx->Const.UniformBufferOffsetAlignment - 1)) {
- _mesa_error(ctx, GL_INVALID_VALUE,
- "glBindBuffersRange(offsets[%u]=%" PRId64
- " is misaligned; it must be a multiple of the value of "
- "GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT=%u when "
- "target=GL_UNIFORM_BUFFER)",
- i, (int64_t) offsets[i],
- ctx->Const.UniformBufferOffsetAlignment);
- continue;
+ /* The ARB_multi_bind spec says:
+ *
+ * "An INVALID_VALUE error is generated by BindBuffersRange if any
+ * pair of values in <offsets> and <sizes> does not respectively
+ * satisfy the constraints described for those parameters for the
+ * specified target, as described in section 6.7.1 (per binding)."
+ *
+ * Section 6.7.1 refers to table 6.5, which says:
+ *
+ * "┌───────────────────────────────────────────────────────────────┐
+ * │ Uniform buffer array bindings (see sec. 7.6) │
+ * ├─────────────────────┬─────────────────────────────────────────┤
+ * │ ... │ ... │
+ * │ offset restriction │ multiple of value of UNIFORM_BUFFER_- │
+ * │ │ OFFSET_ALIGNMENT │
+ * │ ... │ ... │
+ * │ size restriction │ none │
+ * └─────────────────────┴─────────────────────────────────────────┘"
+ */
+ if (offsets[i] & (ctx->Const.UniformBufferOffsetAlignment - 1)) {
+ _mesa_error(ctx, GL_INVALID_VALUE,
+ "glBindBuffersRange(offsets[%u]=%" PRId64
+ " is misaligned; it must be a multiple of the value of "
+ "GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT=%u when "
+ "target=GL_UNIFORM_BUFFER)",
+ i, (int64_t) offsets[i],
+ ctx->Const.UniformBufferOffsetAlignment);
+ continue;
+ }
+
+ offset = offsets[i];
+ size = sizes[i];
}
if (binding->BufferObject && binding->BufferObject->Name == buffers[i])
bufObj = binding->BufferObject;
else
- bufObj = _mesa_multi_bind_lookup_bufferobj(ctx, buffers, i,
- "glBindBuffersRange");
+ bufObj = _mesa_multi_bind_lookup_bufferobj(ctx, buffers, i, caller);
if (bufObj) {
if (bufObj == ctx->Shared->NullBufferObj)
- set_ubo_binding(ctx, binding, bufObj, -1, -1, GL_FALSE);
+ set_ubo_binding(ctx, binding, bufObj, -1, -1, !range);
else
- set_ubo_binding(ctx, binding, bufObj,
- offsets[i], sizes[i], GL_FALSE);
+ set_ubo_binding(ctx, binding, bufObj, offset, size, !range);
}
}
}
static void
-bind_shader_storage_buffers_range(struct gl_context *ctx, GLuint first,
- GLsizei count, const GLuint *buffers,
- const GLintptr *offsets,
- const GLsizeiptr *sizes)
+bind_shader_storage_buffers(struct gl_context *ctx, GLuint first,
+ GLsizei count, const GLuint *buffers,
+ bool range,
+ const GLintptr *offsets,
+ const GLsizeiptr *sizes,
+ const char *caller)
{
GLint i;
- if (!error_check_bind_shader_storage_buffers(ctx, first, count,
- "glBindBuffersRange"))
+ if (!error_check_bind_shader_storage_buffers(ctx, first, count, caller))
return;
/* Assume that at least one binding will be changed */
struct gl_shader_storage_buffer_binding *binding =
&ctx->ShaderStorageBufferBindings[first + i];
struct gl_buffer_object *bufObj;
+ GLintptr offset = 0;
+ GLsizeiptr size = 0;
+
+ if (range) {
+ if (!bind_buffers_check_offset_and_size(ctx, i, offsets, sizes))
+ continue;
+
+ /* The ARB_multi_bind spec says:
+ *
+ * "An INVALID_VALUE error is generated by BindBuffersRange if any
+ * pair of values in <offsets> and <sizes> does not respectively
+ * satisfy the constraints described for those parameters for the
+ * specified target, as described in section 6.7.1 (per binding)."
+ *
+ * Section 6.7.1 refers to table 6.5, which says:
+ *
+ * "┌───────────────────────────────────────────────────────────────┐
+ * │ Shader storage buffer array bindings (see sec. 7.8) │
+ * ├─────────────────────┬─────────────────────────────────────────┤
+ * │ ... │ ... │
+ * │ offset restriction │ multiple of value of SHADER_STORAGE_- │
+ * │ │ BUFFER_OFFSET_ALIGNMENT │
+ * │ ... │ ... │
+ * │ size restriction │ none │
+ * └─────────────────────┴─────────────────────────────────────────┘"
+ */
+ if (offsets[i] & (ctx->Const.ShaderStorageBufferOffsetAlignment - 1)) {
+ _mesa_error(ctx, GL_INVALID_VALUE,
+ "glBindBuffersRange(offsets[%u]=%" PRId64
+ " is misaligned; it must be a multiple of the value of "
+ "GL_SHADER_STORAGE_BUFFER_OFFSET_ALIGNMENT=%u when "
+ "target=GL_SHADER_STORAGE_BUFFER)",
+ i, (int64_t) offsets[i],
+ ctx->Const.ShaderStorageBufferOffsetAlignment);
+ continue;
+ }
- if (!bind_buffers_check_offset_and_size(ctx, i, offsets, sizes))
- continue;
-
- /* The ARB_multi_bind spec says:
- *
- * "An INVALID_VALUE error is generated by BindBuffersRange if any
- * pair of values in <offsets> and <sizes> does not respectively
- * satisfy the constraints described for those parameters for the
- * specified target, as described in section 6.7.1 (per binding)."
- *
- * Section 6.7.1 refers to table 6.5, which says:
- *
- * "┌───────────────────────────────────────────────────────────────┐
- * │ Shader storage buffer array bindings (see sec. 7.8) │
- * ├─────────────────────┬─────────────────────────────────────────┤
- * │ ... │ ... │
- * │ offset restriction │ multiple of value of SHADER_STORAGE_- │
- * │ │ BUFFER_OFFSET_ALIGNMENT │
- * │ ... │ ... │
- * │ size restriction │ none │
- * └─────────────────────┴─────────────────────────────────────────┘"
- */
- if (offsets[i] & (ctx->Const.ShaderStorageBufferOffsetAlignment - 1)) {
- _mesa_error(ctx, GL_INVALID_VALUE,
- "glBindBuffersRange(offsets[%u]=%" PRId64
- " is misaligned; it must be a multiple of the value of "
- "GL_SHADER_STORAGE_BUFFER_OFFSET_ALIGNMENT=%u when "
- "target=GL_SHADER_STORAGE_BUFFER)",
- i, (int64_t) offsets[i],
- ctx->Const.ShaderStorageBufferOffsetAlignment);
- continue;
+ offset = offsets[i];
+ size = sizes[i];
}
if (binding->BufferObject && binding->BufferObject->Name == buffers[i])
bufObj = binding->BufferObject;
else
- bufObj = _mesa_multi_bind_lookup_bufferobj(ctx, buffers, i,
- "glBindBuffersRange");
+ bufObj = _mesa_multi_bind_lookup_bufferobj(ctx, buffers, i, caller);
if (bufObj) {
if (bufObj == ctx->Shared->NullBufferObj)
- set_ssbo_binding(ctx, binding, bufObj, -1, -1, GL_FALSE);
+ set_ssbo_binding(ctx, binding, bufObj, -1, -1, !range);
else
- set_ssbo_binding(ctx, binding, bufObj,
- offsets[i], sizes[i], GL_FALSE);
+ set_ssbo_binding(ctx, binding, bufObj, offset, size, !range);
}
}
}
static void
-bind_xfb_buffers_base(struct gl_context *ctx,
- GLuint first, GLsizei count,
- const GLuint *buffers)
-{
- struct gl_transform_feedback_object *tfObj =
- ctx->TransformFeedback.CurrentObject;
- GLint i;
-
- if (!error_check_bind_xfb_buffers(ctx, tfObj, first, count,
- "glBindBuffersBase"))
- return;
-
- /* Assume that at least one binding will be changed */
- FLUSH_VERTICES(ctx, 0);
- ctx->NewDriverState |= ctx->DriverFlags.NewTransformFeedback;
-
- if (!buffers) {
- /* The ARB_multi_bind spec says:
- *
- * "If <buffers> is NULL, all bindings from <first> through
- * <first>+<count>-1 are reset to their unbound (zero) state."
- */
- unbind_xfb_buffers(ctx, tfObj, first, count);
- return;
- }
-
- /* Note that the error semantics for multi-bind commands differ from
- * those of other GL commands.
- *
- * The Issues section in the ARB_multi_bind spec says:
- *
- * "(11) Typically, OpenGL specifies that if an error is generated by a
- * command, that command has no effect. This is somewhat
- * unfortunate for multi-bind commands, because it would require a
- * first pass to scan the entire list of bound objects for errors
- * and then a second pass to actually perform the bindings.
- * Should we have different error semantics?
- *
- * RESOLVED: Yes. In this specification, when the parameters for
- * one of the <count> binding points are invalid, that binding point
- * is not updated and an error will be generated. However, other
- * binding points in the same command will be updated if their
- * parameters are valid and no other error occurs."
- */
-
- _mesa_begin_bufferobj_lookups(ctx);
-
- for (i = 0; i < count; i++) {
- struct gl_buffer_object * const boundBufObj = tfObj->Buffers[first + i];
- struct gl_buffer_object *bufObj;
-
- if (boundBufObj && boundBufObj->Name == buffers[i])
- bufObj = boundBufObj;
- else
- bufObj = _mesa_multi_bind_lookup_bufferobj(ctx, buffers, i,
- "glBindBuffersBase");
-
- if (bufObj)
- _mesa_set_transform_feedback_binding(ctx, tfObj, first + i,
- bufObj, 0, 0);
- }
-
- _mesa_end_bufferobj_lookups(ctx);
-}
-
-static void
-bind_xfb_buffers_range(struct gl_context *ctx,
- GLuint first, GLsizei count,
- const GLuint *buffers,
- const GLintptr *offsets,
- const GLsizeiptr *sizes)
+bind_xfb_buffers(struct gl_context *ctx,
+ GLuint first, GLsizei count,
+ const GLuint *buffers,
+ bool range,
+ const GLintptr *offsets,
+ const GLsizeiptr *sizes,
+ const char *caller)
{
struct gl_transform_feedback_object *tfObj =
ctx->TransformFeedback.CurrentObject;
GLint i;
- if (!error_check_bind_xfb_buffers(ctx, tfObj, first, count,
- "glBindBuffersRange"))
+ if (!error_check_bind_xfb_buffers(ctx, tfObj, first, count, caller))
return;
/* Assume that at least one binding will be changed */
const GLuint index = first + i;
struct gl_buffer_object * const boundBufObj = tfObj->Buffers[index];
struct gl_buffer_object *bufObj;
+ GLintptr offset = 0;
+ GLsizeiptr size = 0;
- if (!bind_buffers_check_offset_and_size(ctx, i, offsets, sizes))
- continue;
+ if (range) {
+ offset = offsets[i];
+ size = sizes[i];
- /* The ARB_multi_bind spec says:
- *
- * "An INVALID_VALUE error is generated by BindBuffersRange if any
- * pair of values in <offsets> and <sizes> does not respectively
- * satisfy the constraints described for those parameters for the
- * specified target, as described in section 6.7.1 (per binding)."
- *
- * Section 6.7.1 refers to table 6.5, which says:
- *
- * "┌───────────────────────────────────────────────────────────────┐
- * │ Transform feedback array bindings (see sec. 13.2.2) │
- * ├───────────────────────┬───────────────────────────────────────┤
- * │ ... │ ... │
- * │ offset restriction │ multiple of 4 │
- * │ ... │ ... │
- * │ size restriction │ multiple of 4 │
- * └───────────────────────┴───────────────────────────────────────┘"
- */
- if (offsets[i] & 0x3) {
- _mesa_error(ctx, GL_INVALID_VALUE,
- "glBindBuffersRange(offsets[%u]=%" PRId64
- " is misaligned; it must be a multiple of 4 when "
- "target=GL_TRANSFORM_FEEDBACK_BUFFER)",
- i, (int64_t) offsets[i]);
- continue;
- }
+ if (!bind_buffers_check_offset_and_size(ctx, i, offsets, sizes))
+ continue;
- if (sizes[i] & 0x3) {
- _mesa_error(ctx, GL_INVALID_VALUE,
- "glBindBuffersRange(sizes[%u]=%" PRId64
- " is misaligned; it must be a multiple of 4 when "
- "target=GL_TRANSFORM_FEEDBACK_BUFFER)",
- i, (int64_t) sizes[i]);
- continue;
+ /* The ARB_multi_bind spec says:
+ *
+ * "An INVALID_VALUE error is generated by BindBuffersRange if any
+ * pair of values in <offsets> and <sizes> does not respectively
+ * satisfy the constraints described for those parameters for the
+ * specified target, as described in section 6.7.1 (per binding)."
+ *
+ * Section 6.7.1 refers to table 6.5, which says:
+ *
+ * "┌───────────────────────────────────────────────────────────────┐
+ * │ Transform feedback array bindings (see sec. 13.2.2) │
+ * ├───────────────────────┬───────────────────────────────────────┤
+ * │ ... │ ... │
+ * │ offset restriction │ multiple of 4 │
+ * │ ... │ ... │
+ * │ size restriction │ multiple of 4 │
+ * └───────────────────────┴───────────────────────────────────────┘"
+ */
+ if (offsets[i] & 0x3) {
+ _mesa_error(ctx, GL_INVALID_VALUE,
+ "glBindBuffersRange(offsets[%u]=%" PRId64
+ " is misaligned; it must be a multiple of 4 when "
+ "target=GL_TRANSFORM_FEEDBACK_BUFFER)",
+ i, (int64_t) offsets[i]);
+ continue;
+ }
+
+ if (sizes[i] & 0x3) {
+ _mesa_error(ctx, GL_INVALID_VALUE,
+ "glBindBuffersRange(sizes[%u]=%" PRId64
+ " is misaligned; it must be a multiple of 4 when "
+ "target=GL_TRANSFORM_FEEDBACK_BUFFER)",
+ i, (int64_t) sizes[i]);
+ continue;
+ }
+
+ offset = offsets[i];
+ size = sizes[i];
}
if (boundBufObj && boundBufObj->Name == buffers[i])
bufObj = boundBufObj;
else
- bufObj = _mesa_multi_bind_lookup_bufferobj(ctx, buffers, i,
- "glBindBuffersRange");
+ bufObj = _mesa_multi_bind_lookup_bufferobj(ctx, buffers, i, caller);
if (bufObj)
_mesa_set_transform_feedback_binding(ctx, tfObj, index, bufObj,
- offsets[i], sizes[i]);
+ offset, size);
}
_mesa_end_bufferobj_lookups(ctx);
}
static void
-bind_atomic_buffers_base(struct gl_context *ctx,
- GLuint first,
- GLsizei count,
- const GLuint *buffers)
-{
- GLint i;
-
- if (!error_check_bind_atomic_buffers(ctx, first, count,
- "glBindBuffersBase"))
- return;
-
- /* Assume that at least one binding will be changed */
- FLUSH_VERTICES(ctx, 0);
- ctx->NewDriverState |= ctx->DriverFlags.NewAtomicBuffer;
-
- if (!buffers) {
- /* The ARB_multi_bind spec says:
- *
- * "If <buffers> is NULL, all bindings from <first> through
- * <first>+<count>-1 are reset to their unbound (zero) state."
- */
- unbind_atomic_buffers(ctx, first, count);
- return;
- }
-
- /* Note that the error semantics for multi-bind commands differ from
- * those of other GL commands.
- *
- * The Issues section in the ARB_multi_bind spec says:
- *
- * "(11) Typically, OpenGL specifies that if an error is generated by a
- * command, that command has no effect. This is somewhat
- * unfortunate for multi-bind commands, because it would require a
- * first pass to scan the entire list of bound objects for errors
- * and then a second pass to actually perform the bindings.
- * Should we have different error semantics?
- *
- * RESOLVED: Yes. In this specification, when the parameters for
- * one of the <count> binding points are invalid, that binding point
- * is not updated and an error will be generated. However, other
- * binding points in the same command will be updated if their
- * parameters are valid and no other error occurs."
- */
-
- _mesa_begin_bufferobj_lookups(ctx);
-
- for (i = 0; i < count; i++) {
- struct gl_atomic_buffer_binding *binding =
- &ctx->AtomicBufferBindings[first + i];
- struct gl_buffer_object *bufObj;
-
- if (binding->BufferObject && binding->BufferObject->Name == buffers[i])
- bufObj = binding->BufferObject;
- else
- bufObj = _mesa_multi_bind_lookup_bufferobj(ctx, buffers, i,
- "glBindBuffersBase");
-
- if (bufObj)
- set_atomic_buffer_binding(ctx, binding, bufObj, 0, 0);
- }
-
- _mesa_end_bufferobj_lookups(ctx);
-}
-
-static void
-bind_atomic_buffers_range(struct gl_context *ctx,
- GLuint first,
- GLsizei count,
- const GLuint *buffers,
- const GLintptr *offsets,
- const GLsizeiptr *sizes)
+bind_atomic_buffers(struct gl_context *ctx,
+ GLuint first,
+ GLsizei count,
+ const GLuint *buffers,
+ bool range,
+ const GLintptr *offsets,
+ const GLsizeiptr *sizes,
+ const char *caller)
{
GLint i;
- if (!error_check_bind_atomic_buffers(ctx, first, count,
- "glBindBuffersRange"))
+ if (!error_check_bind_atomic_buffers(ctx, first, count, caller))
return;
/* Assume that at least one binding will be changed */
struct gl_atomic_buffer_binding *binding =
&ctx->AtomicBufferBindings[first + i];
struct gl_buffer_object *bufObj;
+ GLintptr offset = 0;
+ GLsizeiptr size = 0;
- if (!bind_buffers_check_offset_and_size(ctx, i, offsets, sizes))
- continue;
+ if (range) {
+ if (!bind_buffers_check_offset_and_size(ctx, i, offsets, sizes))
+ continue;
- /* The ARB_multi_bind spec says:
- *
- * "An INVALID_VALUE error is generated by BindBuffersRange if any
- * pair of values in <offsets> and <sizes> does not respectively
- * satisfy the constraints described for those parameters for the
- * specified target, as described in section 6.7.1 (per binding)."
- *
- * Section 6.7.1 refers to table 6.5, which says:
- *
- * "┌───────────────────────────────────────────────────────────────┐
- * │ Atomic counter array bindings (see sec. 7.7.2) │
- * ├───────────────────────┬───────────────────────────────────────┤
- * │ ... │ ... │
- * │ offset restriction │ multiple of 4 │
- * │ ... │ ... │
- * │ size restriction │ none │
- * └───────────────────────┴───────────────────────────────────────┘"
- */
- if (offsets[i] & (ATOMIC_COUNTER_SIZE - 1)) {
- _mesa_error(ctx, GL_INVALID_VALUE,
- "glBindBuffersRange(offsets[%u]=%" PRId64
- " is misaligned; it must be a multiple of %d when "
- "target=GL_ATOMIC_COUNTER_BUFFER)",
- i, (int64_t) offsets[i], ATOMIC_COUNTER_SIZE);
- continue;
+ /* The ARB_multi_bind spec says:
+ *
+ * "An INVALID_VALUE error is generated by BindBuffersRange if any
+ * pair of values in <offsets> and <sizes> does not respectively
+ * satisfy the constraints described for those parameters for the
+ * specified target, as described in section 6.7.1 (per binding)."
+ *
+ * Section 6.7.1 refers to table 6.5, which says:
+ *
+ * "┌───────────────────────────────────────────────────────────────┐
+ * │ Atomic counter array bindings (see sec. 7.7.2) │
+ * ├───────────────────────┬───────────────────────────────────────┤
+ * │ ... │ ... │
+ * │ offset restriction │ multiple of 4 │
+ * │ ... │ ... │
+ * │ size restriction │ none │
+ * └───────────────────────┴───────────────────────────────────────┘"
+ */
+ if (offsets[i] & (ATOMIC_COUNTER_SIZE - 1)) {
+ _mesa_error(ctx, GL_INVALID_VALUE,
+ "glBindBuffersRange(offsets[%u]=%" PRId64
+ " is misaligned; it must be a multiple of %d when "
+ "target=GL_ATOMIC_COUNTER_BUFFER)",
+ i, (int64_t) offsets[i], ATOMIC_COUNTER_SIZE);
+ continue;
+ }
+
+ offset = offsets[i];
+ size = sizes[i];
}
if (binding->BufferObject && binding->BufferObject->Name == buffers[i])
bufObj = binding->BufferObject;
else
- bufObj = _mesa_multi_bind_lookup_bufferobj(ctx, buffers, i,
- "glBindBuffersRange");
+ bufObj = _mesa_multi_bind_lookup_bufferobj(ctx, buffers, i, caller);
if (bufObj)
- set_atomic_buffer_binding(ctx, binding, bufObj, offsets[i], sizes[i]);
+ set_atomic_buffer_binding(ctx, binding, bufObj, offset, size);
}
_mesa_end_bufferobj_lookups(ctx);
GET_CURRENT_CONTEXT(ctx);
struct gl_buffer_object *bufObj;
+ if (MESA_VERBOSE & VERBOSE_API) {
+ _mesa_debug(ctx, "glBindBufferRange(%s, %u, %u, %ld, %ld)\n",
+ _mesa_enum_to_string(target), index, buffer, offset, size);
+ }
+
if (buffer == 0) {
bufObj = ctx->Shared->NullBufferObj;
} else {
GET_CURRENT_CONTEXT(ctx);
struct gl_buffer_object *bufObj;
+ if (MESA_VERBOSE & VERBOSE_API) {
+ _mesa_debug(ctx, "glBindBufferBase(%s, %u, %u)\n",
+ _mesa_enum_to_string(target), index, buffer);
+ }
+
if (buffer == 0) {
bufObj = ctx->Shared->NullBufferObj;
} else {
{
GET_CURRENT_CONTEXT(ctx);
+ if (MESA_VERBOSE & VERBOSE_API) {
+ _mesa_debug(ctx, "glBindBuffersRange(%s, %u, %d, %p, %p, %p)\n",
+ _mesa_enum_to_string(target), first, count,
+ buffers, offsets, sizes);
+ }
+
switch (target) {
case GL_TRANSFORM_FEEDBACK_BUFFER:
- bind_xfb_buffers_range(ctx, first, count, buffers, offsets, sizes);
+ bind_xfb_buffers(ctx, first, count, buffers, true, offsets, sizes,
+ "glBindBuffersRange");
return;
case GL_UNIFORM_BUFFER:
- bind_uniform_buffers_range(ctx, first, count, buffers, offsets, sizes);
+ bind_uniform_buffers(ctx, first, count, buffers, true, offsets, sizes,
+ "glBindBuffersRange");
return;
case GL_SHADER_STORAGE_BUFFER:
- bind_shader_storage_buffers_range(ctx, first, count, buffers, offsets,
- sizes);
+ bind_shader_storage_buffers(ctx, first, count, buffers, true, offsets, sizes,
+ "glBindBuffersRange");
return;
case GL_ATOMIC_COUNTER_BUFFER:
- bind_atomic_buffers_range(ctx, first, count, buffers,
- offsets, sizes);
+ bind_atomic_buffers(ctx, first, count, buffers, true, offsets, sizes,
+ "glBindBuffersRange");
return;
default:
_mesa_error(ctx, GL_INVALID_ENUM, "glBindBuffersRange(target=%s)",
{
GET_CURRENT_CONTEXT(ctx);
+ if (MESA_VERBOSE & VERBOSE_API) {
+ _mesa_debug(ctx, "glBindBuffersBase(%s, %u, %d, %p)\n",
+ _mesa_enum_to_string(target), first, count, buffers);
+ }
+
switch (target) {
case GL_TRANSFORM_FEEDBACK_BUFFER:
- bind_xfb_buffers_base(ctx, first, count, buffers);
+ bind_xfb_buffers(ctx, first, count, buffers, false, NULL, NULL,
+ "glBindBuffersBase");
return;
case GL_UNIFORM_BUFFER:
- bind_uniform_buffers_base(ctx, first, count, buffers);
+ bind_uniform_buffers(ctx, first, count, buffers, false, NULL, NULL,
+ "glBindBuffersBase");
return;
case GL_SHADER_STORAGE_BUFFER:
- bind_shader_storage_buffers_base(ctx, first, count, buffers);
+ bind_shader_storage_buffers(ctx, first, count, buffers, false, NULL, NULL,
+ "glBindBuffersBase");
return;
case GL_ATOMIC_COUNTER_BUFFER:
- bind_atomic_buffers_base(ctx, first, count, buffers);
+ bind_atomic_buffers(ctx, first, count, buffers, false, NULL, NULL,
+ "glBindBuffersBase");
return;
default:
_mesa_error(ctx, GL_INVALID_ENUM, "glBindBuffersBase(target=%s)",
struct gl_buffer_object *bufObj;
const GLintptr end = offset + length;
+ /* Section 6.5 (Invalidating Buffer Data) of the OpenGL 4.5 (Compatibility
+ * Profile) spec says:
+ *
+ * "An INVALID_VALUE error is generated if buffer is zero or is not the
+ * name of an existing buffer object."
+ */
bufObj = _mesa_lookup_bufferobj(ctx, buffer);
- if (!bufObj) {
+ if (!bufObj || bufObj == &DummyBufferObject) {
_mesa_error(ctx, GL_INVALID_VALUE,
"glInvalidateBufferSubData(name = 0x%x) invalid object",
buffer);
* negative, or if <offset> + <length> is greater than the value of
* BUFFER_SIZE."
*/
- if (end < 0 || end > bufObj->Size) {
+ if (offset < 0 || length < 0 || end > bufObj->Size) {
_mesa_error(ctx, GL_INVALID_VALUE,
"glInvalidateBufferSubData(invalid offset or length)");
return;
return;
}
- /* We don't actually do anything for this yet. Just return after
- * validating the parameters and generating the required errors.
- */
- return;
+ if (ctx->Driver.InvalidateBufferSubData)
+ ctx->Driver.InvalidateBufferSubData(ctx, bufObj, offset, length);
}
void GLAPIENTRY
GET_CURRENT_CONTEXT(ctx);
struct gl_buffer_object *bufObj;
+ /* Section 6.5 (Invalidating Buffer Data) of the OpenGL 4.5 (Compatibility
+ * Profile) spec says:
+ *
+ * "An INVALID_VALUE error is generated if buffer is zero or is not the
+ * name of an existing buffer object."
+ */
bufObj = _mesa_lookup_bufferobj(ctx, buffer);
- if (!bufObj) {
+ if (!bufObj || bufObj == &DummyBufferObject) {
_mesa_error(ctx, GL_INVALID_VALUE,
"glInvalidateBufferData(name = 0x%x) invalid object",
buffer);
return;
}
- /* We don't actually do anything for this yet. Just return after
- * validating the parameters and generating the required errors.
- */
- return;
+ if (ctx->Driver.InvalidateBufferSubData)
+ ctx->Driver.InvalidateBufferSubData(ctx, bufObj, 0, bufObj->Size);
}
GLuint name);
extern void
+_mesa_delete_buffer_object(struct gl_context *ctx,
+ struct gl_buffer_object *bufObj);
+
+extern void
_mesa_reference_buffer_object_(struct gl_context *ctx,
struct gl_buffer_object **ptr,
struct gl_buffer_object *bufObj);
GLintptr readOffset, GLintptr writeOffset,
GLsizeiptr size );
+ void (*InvalidateBufferSubData)( struct gl_context *ctx,
+ struct gl_buffer_object *obj,
+ GLintptr offset,
+ GLsizeiptr length );
+
/* Returns pointer to the start of the mapped range.
* May return NULL if MESA_MAP_NOWAIT_BIT is set in access:
*/
EXT(ARB_gpu_shader_fp64 , ARB_gpu_shader_fp64 , x , GLC, x , x , 2010)
EXT(ARB_half_float_pixel , dummy_true , GLL, GLC, x , x , 2003)
EXT(ARB_half_float_vertex , ARB_half_float_vertex , GLL, GLC, x , x , 2008)
+EXT(ARB_indirect_parameters , ARB_indirect_parameters , x , GLC, x , x , 2013)
EXT(ARB_instanced_arrays , ARB_instanced_arrays , GLL, GLC, x , x , 2008)
EXT(ARB_internalformat_query , ARB_internalformat_query , GLL, GLC, x , x , 2011)
EXT(ARB_invalidate_subdata , dummy_true , GLL, GLC, x , x , 2012)
EXT(ARB_shader_atomic_counters , ARB_shader_atomic_counters , GLL, GLC, x , x , 2011)
EXT(ARB_shader_bit_encoding , ARB_shader_bit_encoding , GLL, GLC, x , x , 2010)
EXT(ARB_shader_clock , ARB_shader_clock , GLL, GLC, x , x , 2015)
+EXT(ARB_shader_draw_parameters , ARB_shader_draw_parameters , GLL, GLC, x , x , 2013)
EXT(ARB_shader_image_load_store , ARB_shader_image_load_store , GLL, GLC, x , x , 2011)
EXT(ARB_shader_image_size , ARB_shader_image_size , GLL, GLC, x , x , 2012)
EXT(ARB_shader_objects , dummy_true , GLL, GLC, x , x , 2002)
ctx->Driver.ValidateFramebuffer(ctx, fb);
if (fb->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) {
fbo_incomplete(ctx, "driver marked FBO as incomplete", -1);
+ return;
}
}
- if (fb->_Status == GL_FRAMEBUFFER_COMPLETE_EXT) {
- /*
- * Note that if ARB_framebuffer_object is supported and the attached
- * renderbuffers/textures are different sizes, the framebuffer
- * width/height will be set to the smallest width/height.
- */
- if (numImages != 0) {
- fb->Width = minWidth;
- fb->Height = minHeight;
- }
-
- /* finally, update the visual info for the framebuffer */
- _mesa_update_framebuffer_visual(ctx, fb);
+ /*
+ * Note that if ARB_framebuffer_object is supported and the attached
+ * renderbuffers/textures are different sizes, the framebuffer
+ * width/height will be set to the smallest width/height.
+ */
+ if (numImages != 0) {
+ fb->Width = minWidth;
+ fb->Height = minHeight;
}
+
+ /* finally, update the visual info for the framebuffer */
+ _mesa_update_framebuffer_visual(ctx, fb);
}
EXTRA_EXT(ARB_tessellation_shader);
EXTRA_EXT(ARB_shader_subroutine);
EXTRA_EXT(ARB_shader_storage_buffer_object);
+EXTRA_EXT(ARB_indirect_parameters);
static const int
extra_ARB_color_buffer_float_or_glcore[] = {
case GL_DRAW_INDIRECT_BUFFER_BINDING:
v->value_int = ctx->DrawIndirectBuffer->Name;
break;
+ /* GL_ARB_indirect_parameters */
+ case GL_PARAMETER_BUFFER_BINDING_ARB:
+ v->value_int = ctx->ParameterBuffer->Name;
+ break;
/* GL_ARB_separate_shader_objects */
case GL_PROGRAM_PIPELINE_BINDING:
if (ctx->Pipeline.Current) {
# GL_ARB_shader_subroutine
[ "MAX_SUBROUTINES", "CONST(MAX_SUBROUTINES), extra_ARB_shader_subroutine" ],
[ "MAX_SUBROUTINE_UNIFORM_LOCATIONS", "CONST(MAX_SUBROUTINE_UNIFORM_LOCATIONS), extra_ARB_shader_subroutine" ],
+
+# GL_ARB_indirect_parameters
+ [ "PARAMETER_BUFFER_BINDING_ARB", "LOC_CUSTOM, TYPE_INT, 0, extra_ARB_indirect_parameters" ],
+
]}
]
return (int) ((f >= 0.0F) ? (f + 0.5F) : (f - 0.5F));
}
+/**
+ * Convert double to int by rounding to nearest integer, away from zero.
+ */
+static inline int IROUNDD(double d)
+{
+ return (int) ((d >= 0.0) ? (d + 0.5) : (d - 0.5));
+}
/**
* Convert float to int64 by rounding to nearest integer.
/**
* ATI_fragment_shader runtime state
*/
-#define ATI_FS_INPUT_PRIMARY 0
-#define ATI_FS_INPUT_SECONDARY 1
struct atifs_instruction;
struct atifs_setupinst;
GLuint Binding;
/**
+ * Vulkan descriptor set qualifier for this block.
+ */
+ GLuint Set;
+
+ /**
* Minimum size (in bytes) of a buffer object to back this uniform buffer
* (GL_UNIFORM_BLOCK_DATA_SIZE).
*/
};
/**
+ * Data container for shader queries. This holds only the minimal
+ * amount of required information for resource queries to work.
+ */
+struct gl_shader_variable
+{
+ /**
+ * Declared type of the variable
+ */
+ const struct glsl_type *type;
+
+ /**
+ * Declared name of the variable
+ */
+ char *name;
+
+ /**
+ * Storage location of the base of this variable
+ *
+ * The precise meaning of this field depends on the nature of the variable.
+ *
+ * - Vertex shader input: one of the values from \c gl_vert_attrib.
+ * - Vertex shader output: one of the values from \c gl_varying_slot.
+ * - Geometry shader input: one of the values from \c gl_varying_slot.
+ * - Geometry shader output: one of the values from \c gl_varying_slot.
+ * - Fragment shader input: one of the values from \c gl_varying_slot.
+ * - Fragment shader output: one of the values from \c gl_frag_result.
+ * - Uniforms: Per-stage uniform slot number for default uniform block.
+ * - Uniforms: Index within the uniform block definition for UBO members.
+ * - Non-UBO Uniforms: explicit location until linking then reused to
+ * store uniform slot number.
+ * - Other: This field is not currently used.
+ *
+ * If the variable is a uniform, shader input, or shader output, and the
+ * slot has not been assigned, the value will be -1.
+ */
+ int location;
+
+ /**
+ * Output index for dual source blending.
+ *
+ * \note
+ * The GLSL spec only allows the values 0 or 1 for the index in \b dual
+ * source blending.
+ */
+ unsigned index:1;
+
+ /**
+ * Specifies whether a shader input/output is per-patch in tessellation
+ * shader stages.
+ */
+ unsigned patch:1;
+
+ /**
+ * Storage class of the variable.
+ *
+ * \sa (n)ir_variable_mode
+ */
+ unsigned mode:4;
+};
+
+/**
* Active resource in a gl_shader_program
*/
struct gl_program_resource
int *InterfaceBlockStageIndex[MESA_SHADER_STAGES];
/**
+ * Indices into the BufferInterfaceBlocks[] array for Uniform Buffer
+ * Objects and Shader Storage Buffer Objects.
+ */
+ unsigned *UboInterfaceBlockIndex;
+ unsigned *SsboInterfaceBlockIndex;
+
+ /**
* Map of active uniform names to locations
*
* Maps any active uniform that is not an array element to a location.
*/
GLboolean GLSLSkipStrictMaxUniformLimitCheck;
+ /** Whether gl_FragCoord and gl_FrontFacing are system values. */
+ bool GLSLFragCoordIsSysVal;
+ bool GLSLFrontFacingIsSysVal;
+
/**
* Always use the GetTransformFeedbackVertexCount() driver hook, rather
* than passing the transform feedback object to the drawing function.
GLboolean ARB_gpu_shader5;
GLboolean ARB_gpu_shader_fp64;
GLboolean ARB_half_float_vertex;
+ GLboolean ARB_indirect_parameters;
GLboolean ARB_instanced_arrays;
GLboolean ARB_internalformat_query;
GLboolean ARB_map_buffer_range;
GLboolean ARB_shader_atomic_counters;
GLboolean ARB_shader_bit_encoding;
GLboolean ARB_shader_clock;
+ GLboolean ARB_shader_draw_parameters;
GLboolean ARB_shader_image_load_store;
GLboolean ARB_shader_image_size;
GLboolean ARB_shader_precision;
struct gl_perf_monitor_state PerfMonitor;
struct gl_buffer_object *DrawIndirectBuffer; /** < GL_ARB_draw_indirect */
+ struct gl_buffer_object *ParameterBuffer; /** < GL_ARB_indirect_parameters */
struct gl_buffer_object *DispatchIndirectBuffer; /** < GL_ARB_compute_shader */
struct gl_buffer_object *CopyReadBuffer; /**< GL_ARB_copy_buffer */
DEBUG_INCOMPLETE_FBO = (1 << 3)
};
+/**
+ * Checks if the active fragment shader program can have side effects due
+ * to use of things like atomic buffers or images
+ */
static inline bool
-_mesa_active_fragment_shader_has_atomic_ops(const struct gl_context *ctx)
+_mesa_active_fragment_shader_has_side_effects(const struct gl_context *ctx)
{
- return ctx->Shader._CurrentFragmentProgram != NULL &&
- ctx->Shader._CurrentFragmentProgram->_LinkedShaders[MESA_SHADER_FRAGMENT]->NumAtomicBuffers > 0;
+ const struct gl_shader *sh;
+
+ if (!ctx->_Shader->_CurrentFragmentProgram)
+ return false;
+
+ sh = ctx->_Shader->_CurrentFragmentProgram->_LinkedShaders[MESA_SHADER_FRAGMENT];
+ return sh->NumAtomicBuffers > 0 ||
+ sh->NumImages > 0 ||
+ sh->NumShaderStorageBlocks > 0;
}
#ifdef __cplusplus
for (group = 0; group < ctx->PerfMonitor.NumGroups; group++) {
const struct gl_perf_monitor_group *g = &ctx->PerfMonitor.Groups[group];
- for (counter = 0; counter < g->NumCounters; counter++) {
- const struct gl_perf_monitor_counter *c = &g->Counters[counter];
+ BITSET_WORD tmp;
- if (!BITSET_TEST(m->ActiveCounters[group], counter))
- continue;
+ BITSET_FOREACH_SET(counter, tmp, m->ActiveCounters[group], g->NumCounters) {
+ const struct gl_perf_monitor_counter *c = &g->Counters[counter];
size += sizeof(uint32_t); /* Group ID */
size += sizeof(uint32_t); /* Counter ID */
GLenum pname, GLint *params)
{
GET_CURRENT_CONTEXT(ctx);
+
+ if (MESA_VERBOSE & VERBOSE_API) {
+ _mesa_debug(ctx, "glGetProgramInterfaceiv(%u, %s, %s, %p)\n",
+ program, _mesa_enum_to_string(programInterface),
+ _mesa_enum_to_string(pname), params);
+ }
+
unsigned i;
struct gl_shader_program *shProg =
_mesa_lookup_shader_program_err(ctx, program,
const GLchar *name)
{
GET_CURRENT_CONTEXT(ctx);
+
+ if (MESA_VERBOSE & VERBOSE_API) {
+ _mesa_debug(ctx, "glGetProgramResourceIndex(%u, %s, %s)\n",
+ program, _mesa_enum_to_string(programInterface), name);
+ }
+
unsigned array_index = 0;
struct gl_program_resource *res;
struct gl_shader_program *shProg =
GLchar *name)
{
GET_CURRENT_CONTEXT(ctx);
+
+ if (MESA_VERBOSE & VERBOSE_API) {
+ _mesa_debug(ctx, "glGetProgramResourceName(%u, %s, %u, %d, %p, %p)\n",
+ program, _mesa_enum_to_string(programInterface), index,
+ bufSize, length, name);
+ }
+
struct gl_shader_program *shProg =
_mesa_lookup_shader_program_err(ctx, program,
"glGetProgramResourceName");
GLsizei *length, GLint *params)
{
GET_CURRENT_CONTEXT(ctx);
+
+ if (MESA_VERBOSE & VERBOSE_API) {
+ _mesa_debug(ctx, "glGetProgramResourceiv(%u, %s, %u, %d, %p, %d, %p, %p)\n",
+ program, _mesa_enum_to_string(programInterface), index,
+ propCount, props, bufSize, length, params);
+ }
+
struct gl_shader_program *shProg =
_mesa_lookup_shader_program_err(ctx, program, "glGetProgramResourceiv");
const GLchar *name)
{
GET_CURRENT_CONTEXT(ctx);
+
+ if (MESA_VERBOSE & VERBOSE_API) {
+ _mesa_debug(ctx, "glGetProgramResourceLocation(%u, %s, %s)\n",
+ program, _mesa_enum_to_string(programInterface), name);
+ }
+
struct gl_shader_program *shProg =
lookup_linked_program(program, "glGetProgramResourceLocation");
const GLchar *name)
{
GET_CURRENT_CONTEXT(ctx);
+
+ if (MESA_VERBOSE & VERBOSE_API) {
+ _mesa_debug(ctx, "glGetProgramResourceLocationIndex(%u, %s, %s)\n",
+ program, _mesa_enum_to_string(programInterface), name);
+ }
+
struct gl_shader_program *shProg =
lookup_linked_program(program, "glGetProgramResourceLocationIndex");
return sampObj != NULL;
}
+void
+_mesa_bind_sampler(struct gl_context *ctx, GLuint unit,
+ struct gl_sampler_object *sampObj)
+{
+ if (ctx->Texture.Unit[unit].Sampler != sampObj) {
+ FLUSH_VERTICES(ctx, _NEW_TEXTURE);
+ }
+
+ _mesa_reference_sampler_object(ctx, &ctx->Texture.Unit[unit].Sampler,
+ sampObj);
+}
void GLAPIENTRY
_mesa_BindSampler(GLuint unit, GLuint sampler)
}
}
- if (ctx->Texture.Unit[unit].Sampler != sampObj) {
- FLUSH_VERTICES(ctx, _NEW_TEXTURE);
- }
-
/* bind new sampler */
- _mesa_reference_sampler_object(ctx, &ctx->Texture.Unit[unit].Sampler,
- sampObj);
+ _mesa_bind_sampler(ctx, unit, sampObj);
}
FLUSH_VERTICES(ctx, _NEW_TEXTURE);
}
+void
+_mesa_set_sampler_wrap(struct gl_context *ctx, struct gl_sampler_object *samp,
+ GLenum s, GLenum t, GLenum r)
+{
+ assert(validate_texture_wrap_mode(ctx, s));
+ assert(validate_texture_wrap_mode(ctx, t));
+ assert(validate_texture_wrap_mode(ctx, r));
+
+ if (samp->WrapS == s && samp->WrapT == t && samp->WrapR == r)
+ return;
+
+ flush(ctx);
+ samp->WrapS = s;
+ samp->WrapT = t;
+ samp->WrapR = r;
+}
#define INVALID_PARAM 0x100
#define INVALID_PNAME 0x101
return INVALID_PARAM;
}
+void
+_mesa_set_sampler_filters(struct gl_context *ctx,
+ struct gl_sampler_object *samp,
+ GLenum min_filter, GLenum mag_filter)
+{
+ assert(min_filter == GL_NEAREST ||
+ min_filter == GL_LINEAR ||
+ min_filter == GL_NEAREST_MIPMAP_NEAREST ||
+ min_filter == GL_LINEAR_MIPMAP_NEAREST ||
+ min_filter == GL_NEAREST_MIPMAP_LINEAR ||
+ min_filter == GL_LINEAR_MIPMAP_LINEAR);
+ assert(mag_filter == GL_NEAREST ||
+ mag_filter == GL_LINEAR);
+
+ if (samp->MinFilter == min_filter && samp->MagFilter == mag_filter)
+ return;
+
+ flush(ctx);
+ samp->MinFilter = min_filter;
+ samp->MagFilter = mag_filter;
+}
static GLuint
set_sampler_min_filter(struct gl_context *ctx, struct gl_sampler_object *samp,
return GL_TRUE;
}
+void
+_mesa_set_sampler_srgb_decode(struct gl_context *ctx,
+ struct gl_sampler_object *samp, GLenum param)
+{
+ assert(param == GL_DECODE_EXT || param == GL_SKIP_DECODE_EXT);
+
+ flush(ctx);
+ samp->sRGBDecode = param;
+}
+
static GLuint
set_sampler_srgb_decode(struct gl_context *ctx,
struct gl_sampler_object *samp, GLenum param)
extern void
_mesa_init_sampler_object_functions(struct dd_function_table *driver);
+extern void
+_mesa_set_sampler_wrap(struct gl_context *ctx, struct gl_sampler_object *samp,
+ GLenum s, GLenum t, GLenum r);
+
+extern void
+_mesa_set_sampler_filters(struct gl_context *ctx,
+ struct gl_sampler_object *samp,
+ GLenum min_filter, GLenum mag_filter);
+
+extern void
+_mesa_set_sampler_srgb_decode(struct gl_context *ctx,
+ struct gl_sampler_object *samp, GLenum param);
+
+extern void
+_mesa_bind_sampler(struct gl_context *ctx, GLuint unit,
+ struct gl_sampler_object *sampObj);
+
void GLAPIENTRY
_mesa_GenSamplers(GLsizei count, GLuint *samplers);
void GLAPIENTRY
return (type *) res->Data; \
}
-DECL_RESOURCE_FUNC(VAR, ir_variable);
+DECL_RESOURCE_FUNC(VAR, gl_shader_variable);
DECL_RESOURCE_FUNC(UBO, gl_uniform_block);
DECL_RESOURCE_FUNC(UNI, gl_uniform_storage);
DECL_RESOURCE_FUNC(ATC, gl_active_atomic_buffer);
}
static bool
-is_active_attrib(const ir_variable *var)
+is_active_attrib(const gl_shader_variable *var)
{
if (!var)
return false;
- switch (var->data.mode) {
+ switch (var->mode) {
case ir_var_shader_in:
- return var->data.location != -1;
+ return var->location != -1;
case ir_var_system_value:
/* From GL 4.3 core spec, section 11.1.1 (Vertex Attributes):
* are enumerated, including the special built-in inputs gl_VertexID
* and gl_InstanceID."
*/
- return var->data.location == SYSTEM_VALUE_VERTEX_ID ||
- var->data.location == SYSTEM_VALUE_VERTEX_ID_ZERO_BASE ||
- var->data.location == SYSTEM_VALUE_INSTANCE_ID;
+ return var->location == SYSTEM_VALUE_VERTEX_ID ||
+ var->location == SYSTEM_VALUE_VERTEX_ID_ZERO_BASE ||
+ var->location == SYSTEM_VALUE_INSTANCE_ID;
default:
return false;
return;
}
- const ir_variable *const var = RESOURCE_VAR(res);
+ const gl_shader_variable *const var = RESOURCE_VAR(res);
if (!is_active_attrib(var))
return;
* consider gl_VertexIDMESA as gl_VertexID for purposes of checking
* active attributes.
*/
- if (var->data.mode == ir_var_system_value &&
- var->data.location == SYSTEM_VALUE_VERTEX_ID_ZERO_BASE) {
+ if (var->mode == ir_var_system_value &&
+ var->location == SYSTEM_VALUE_VERTEX_ID_ZERO_BASE) {
var_name = "gl_VertexID";
}
const char*
_mesa_program_resource_name(struct gl_program_resource *res)
{
- const ir_variable *var;
+ const gl_shader_variable *var;
switch (res->Type) {
case GL_UNIFORM_BLOCK:
case GL_SHADER_STORAGE_BLOCK:
case GL_PROGRAM_INPUT:
var = RESOURCE_VAR(res);
/* Special case gl_VertexIDMESA -> gl_VertexID. */
- if (var->data.mode == ir_var_system_value &&
- var->data.location == SYSTEM_VALUE_VERTEX_ID_ZERO_BASE) {
+ if (var->mode == ir_var_system_value &&
+ var->location == SYSTEM_VALUE_VERTEX_ID_ZERO_BASE) {
return "gl_VertexID";
}
/* fallthrough */
static bool
add_index_to_name(struct gl_program_resource *res)
{
- bool add_index = !(((res->Type == GL_PROGRAM_INPUT) &&
- res->StageReferences & (1 << MESA_SHADER_GEOMETRY)));
+ bool add_index = !((res->Type == GL_PROGRAM_INPUT &&
+ res->StageReferences & (1 << MESA_SHADER_GEOMETRY |
+ 1 << MESA_SHADER_TESS_CTRL |
+ 1 << MESA_SHADER_TESS_EVAL)) ||
+ (res->Type == GL_PROGRAM_OUTPUT &&
+ res->StageReferences & 1 << MESA_SHADER_TESS_CTRL));
/* Transform feedback varyings have array index already appended
* in their names.
*/
switch (res->Type) {
case GL_PROGRAM_INPUT: {
- const ir_variable *var = RESOURCE_VAR(res);
+ const gl_shader_variable *var = RESOURCE_VAR(res);
/* If the input is an array, fail if the index is out of bounds. */
if (array_index > 0
&& array_index >= var->type->length) {
return -1;
}
- return (var->data.location +
+ return (var->location +
(array_index * var->type->without_array()->matrix_columns) -
VERT_ATTRIB_GENERIC0);
}
&& array_index >= RESOURCE_VAR(res)->type->length) {
return -1;
}
- return RESOURCE_VAR(res)->data.location + array_index - FRAG_RESULT_DATA0;
+ return RESOURCE_VAR(res)->location + array_index - FRAG_RESULT_DATA0;
case GL_UNIFORM:
/* If the uniform is built-in, fail. */
if (RESOURCE_UNI(res)->builtin)
if (!res || !(res->StageReferences & (1 << MESA_SHADER_FRAGMENT)))
return -1;
- return RESOURCE_VAR(res)->data.index;
+ return RESOURCE_VAR(res)->index;
}
static uint8_t
case GL_LOCATION_INDEX:
if (res->Type != GL_PROGRAM_OUTPUT)
goto invalid_operation;
- *val = RESOURCE_VAR(res)->data.index;
+ *val = RESOURCE_VAR(res)->index;
return 1;
case GL_NUM_COMPATIBLE_SUBROUTINES:
switch (res->Type) {
case GL_PROGRAM_INPUT:
case GL_PROGRAM_OUTPUT:
- *val = RESOURCE_VAR(res)->data.patch;
+ *val = RESOURCE_VAR(res)->patch;
return 1;
default:
goto invalid_operation;
}
static bool
-validate_io(const struct gl_shader *input_stage,
- const struct gl_shader *output_stage, bool isES)
+validate_io(const struct gl_shader *producer,
+ const struct gl_shader *consumer, bool isES)
{
- assert(input_stage && output_stage);
+ assert(producer && consumer);
+ unsigned inputs = 0, outputs = 0;
+
+ /* From OpenGL ES 3.1 spec (Interface matching):
+ *
+ * "An output variable is considered to match an input variable in the
+ * subsequent shader if:
+ *
+ * - the two variables match in name, type, and qualification; or
+ * - the two variables are declared with the same location qualifier and
+ * match in type and qualification.
+ *
+ * ...
+ *
+ * At an interface between program objects, the set of inputs and outputs
+ * are considered to match exactly if and only if:
+ *
+ * - Every declared input variable has a matching output, as described
+ * above.
+ *
+ * - There are no user-defined output variables declared without a
+ * matching input variable declaration.
+ *
+ * - All matched input and output variables have identical precision
+ * qualification.
+ *
+ * When the set of inputs and outputs on an interface between programs
+ * matches exactly, all inputs are well-defined except when the
+ * corresponding outputs were not written in the previous shader. However,
+ * any mismatch between inputs and outputs will result in a validation
+ * failure."
+ *
+ * OpenGL Core 4.5 spec includes same paragraph as above but without check
+ * for precision and the last 'validation failure' clause. Therefore
+ * behaviour is more relaxed, input and output amount is not required by the
+ * spec to be validated.
+ *
+ * FIXME: Update once Khronos spec bug #15331 is resolved.
+ * FIXME: Add validation by type, currently information loss during varying
+ * packing makes this challenging.
+ */
+
+ /* Currently no matching done for desktop. */
+ if (!isES)
+ return true;
/* For each output in a, find input in b and do any required checks. */
- foreach_in_list(ir_instruction, out, input_stage->ir) {
+ foreach_in_list(ir_instruction, out, producer->ir) {
ir_variable *out_var = out->as_variable();
- if (!out_var || out_var->data.mode != ir_var_shader_out)
+ if (!out_var || out_var->data.mode != ir_var_shader_out ||
+ is_gl_identifier(out_var->name))
continue;
- foreach_in_list(ir_instruction, in, output_stage->ir) {
+ outputs++;
+
+ inputs = 0;
+ foreach_in_list(ir_instruction, in, consumer->ir) {
ir_variable *in_var = in->as_variable();
- if (!in_var || in_var->data.mode != ir_var_shader_in)
+ if (!in_var || in_var->data.mode != ir_var_shader_in ||
+ is_gl_identifier(in_var->name))
+ continue;
+
+ inputs++;
+
+ /* Match by location qualifier and precision.
+ *
+ * FIXME: Add explicit location matching validation here. Be careful
+ * not to match varyings with explicit locations to varyings without
+ * explicit locations.
+ */
+ if ((in_var->data.explicit_location &&
+ out_var->data.explicit_location) &&
+ in_var->data.location == out_var->data.location &&
+ in_var->data.precision == out_var->data.precision)
continue;
- if (strcmp(in_var->name, out_var->name) == 0) {
- /* Since we now only validate precision, we can skip this step for
- * desktop GLSL shaders, there precision qualifier is ignored.
- *
- * From OpenGL 4.50 Shading Language spec, section 4.7:
- * "For the purposes of determining if an output from one
- * shader stage matches an input of the next stage, the
- * precision qualifier need not match."
+ unsigned len = strlen(in_var->name);
+
+ /* Handle input swizzle in variable name. */
+ const char *dot = strchr(in_var->name, '.');
+ if (dot)
+ len = dot - in_var->name;
+
+ /* Match by name and precision. */
+ if (strncmp(in_var->name, out_var->name, len) == 0) {
+ /* From OpenGL ES 3.1 spec:
+ * "When both shaders are in separate programs, mismatched
+ * precision qualifiers will result in a program interface
+ * mismatch that will result in program pipeline validation
+ * failures, as described in section 7.4.1 (“Shader Interface
+ * Matching”) of the OpenGL ES 3.1 Specification."
*/
- if (isES) {
- /* From OpenGL ES 3.1 spec:
- * "When both shaders are in separate programs, mismatched
- * precision qualifiers will result in a program interface
- * mismatch that will result in program pipeline validation
- * failures, as described in section 7.4.1 (“Shader Interface
- * Matching”) of the OpenGL ES 3.1 Specification."
- */
- if (in_var->data.precision != out_var->data.precision)
- return false;
- }
+ if (in_var->data.precision != out_var->data.precision)
+ return false;
}
}
}
- return true;
+ return inputs == outputs;
}
/**
for (idx = prev + 1; idx < ARRAY_SIZE(pipeline->CurrentProgram); idx++) {
if (shProg[idx]) {
+ /* Pipeline might include both non-compute and a compute program, do
+ * not attempt to validate varyings between non-compute and compute
+ * stage.
+ */
+ if (shProg[idx]->_LinkedShaders[idx]->Stage == MESA_SHADER_COMPUTE)
+ break;
+
if (!validate_io(shProg[prev]->_LinkedShaders[prev],
shProg[idx]->_LinkedShaders[idx],
shProg[prev]->IsES || shProg[idx]->IsES))
case GL_TESS_EVALUATION_SHADER:
return ctx == NULL || _mesa_has_tessellation(ctx);
case GL_COMPUTE_SHADER:
- return ctx == NULL || ctx->Extensions.ARB_compute_shader;
+ return ctx == NULL || _mesa_has_compute_shaders(ctx);
default:
return false;
}
GLuint name;
if (!_mesa_validate_shader_target(ctx, type)) {
- _mesa_error(ctx, GL_INVALID_ENUM, "CreateShader(type)");
+ _mesa_error(ctx, GL_INVALID_ENUM, "CreateShader(%s)",
+ _mesa_enum_to_string(type));
return 0;
}
_mesa_LinkProgram(GLhandleARB programObj)
{
GET_CURRENT_CONTEXT(ctx);
+ if (MESA_VERBOSE & VERBOSE_API)
+ _mesa_debug(ctx, "glLinkProgram %u\n", programObj);
link_program(ctx, programObj);
}
GET_CURRENT_CONTEXT(ctx);
struct gl_shader_program *shProg;
+ if (MESA_VERBOSE & VERBOSE_API)
+ _mesa_debug(ctx, "glUseProgram %u\n", program);
+
if (_mesa_is_xfb_active_and_unpaused(ctx)) {
_mesa_error(ctx, GL_INVALID_OPERATION,
"glUseProgram(transform feedback active)");
* That is, if barriers is the special value GL_ALL_BARRIER_BITS, then all
* barriers allowed by glMemoryBarrierByRegion should be activated."
*/
- if (barriers == GL_ALL_BARRIER_BITS)
- return ctx->Driver.MemoryBarrier(ctx, all_allowed_bits);
+ if (barriers == GL_ALL_BARRIER_BITS) {
+ ctx->Driver.MemoryBarrier(ctx, all_allowed_bits);
+ return;
+ }
/* From section 7.11.2 of the OpenGL ES 3.1 specification:
*
* \sa glStencilFunc().
*
* Verifies the parameters and updates the respective values in
- * __struct gl_contextRec::Stencil. On change flushes the vertices and notifies the
- * driver via the dd_function_table::StencilFunc callback.
+ * __struct gl_contextRec::Stencil. On change flushes the vertices and notifies
+ * the driver via the dd_function_table::StencilFunc callback.
*/
void GLAPIENTRY
_mesa_StencilFuncSeparateATI( GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask )
* \sa glStencilFunc().
*
* Verifies the parameters and updates the respective values in
- * __struct gl_contextRec::Stencil. On change flushes the vertices and notifies the
- * driver via the dd_function_table::StencilFunc callback.
+ * __struct gl_contextRec::Stencil. On change flushes the vertices and notifies
+ * the driver via the dd_function_table::StencilFunc callback.
*/
void GLAPIENTRY
_mesa_StencilFunc( GLenum func, GLint ref, GLuint mask )
* \sa glStencilOp().
*
* Verifies the parameters and updates the respective fields in
- * __struct gl_contextRec::Stencil. On change flushes the vertices and notifies the
- * driver via the dd_function_table::StencilOp callback.
+ * __struct gl_contextRec::Stencil. On change flushes the vertices and notifies
+ * the driver via the dd_function_table::StencilOp callback.
*/
void GLAPIENTRY
_mesa_StencilOp(GLenum fail, GLenum zfail, GLenum zpass)
-/**
- * OpenGL 2.0 function.
- * \todo Make StencilOp() call this function. And eventually remove the
- * ctx->Driver.StencilOp function and use ctx->Driver.StencilOpSeparate
- * instead.
- */
void GLAPIENTRY
_mesa_StencilOpSeparate(GLenum face, GLenum sfail, GLenum zfail, GLenum zpass)
{
{ "glGetQueryBufferObjecti64v", 45, -1 },
{ "glGetQueryBufferObjectui64v", 45, -1 },
+ /* GL_ARB_indirect_parameters */
+ { "glMultiDrawArraysIndirectCountARB", 31, -1 },
+ { "glMultiDrawElementsIndirectCountARB", 31, -1 },
+
{ NULL, 0, -1 }
};
_mesa_enum_to_string(internalFormat));
return GL_TRUE;
}
+ } else {
+ /*
+ * Section 8.6 (Alternate Texture Image Specification Commands) of the
+ * OpenGL 4.5 (Compatibility Profile) spec says:
+ *
+ * "Parameters level, internalformat, and border are specified using
+ * the same values, with the same meanings, as the corresponding
+ * arguments of TexImage2D, except that internalformat may not be
+ * specified as 1, 2, 3, or 4."
+ */
+ if (internalFormat >= 1 && internalFormat <= 4) {
+ _mesa_error(ctx, GL_INVALID_ENUM,
+ "glCopyTexImage%dD(internalFormat=%d)", dimensions,
+ internalFormat);
+ return GL_TRUE;
+ }
}
baseFormat = _mesa_base_tex_format(ctx, internalFormat);
incomplete(t, MIPMAP, "TexImage[%d] is missing", i);
return;
}
- if (img->TexFormat != baseImage->TexFormat) {
+ if (img->InternalFormat != baseImage->InternalFormat) {
incomplete(t, MIPMAP, "Format[i] != Format[baseLevel]");
return;
}
dst[didx].i = src[sidx].i ? 1 : 0;
break;
case GLSL_TYPE_DOUBLE:
- dst[didx].i = *(double *)&src[sidx].f;
+ dst[didx].i = IROUNDD(*(double *)&src[sidx].f);
break;
default:
assert(!"Should not get here.");
if (!shProg)
return;
- if (uniformBlockIndex >= shProg->NumBufferInterfaceBlocks) {
+ if (uniformBlockIndex >= shProg->NumUniformBlocks) {
_mesa_error(ctx, GL_INVALID_VALUE,
"glUniformBlockBinding(block index %u >= %u)",
- uniformBlockIndex, shProg->NumBufferInterfaceBlocks);
+ uniformBlockIndex, shProg->NumUniformBlocks);
return;
}
return;
}
- if (shProg->BufferInterfaceBlocks[uniformBlockIndex].Binding !=
+ if (shProg->UniformBlocks[uniformBlockIndex]->Binding !=
uniformBlockBinding) {
int i;
FLUSH_VERTICES(ctx, 0);
ctx->NewDriverState |= ctx->DriverFlags.NewUniformBuffer;
- shProg->BufferInterfaceBlocks[uniformBlockIndex].Binding = uniformBlockBinding;
+ const int interface_block_index =
+ shProg->UboInterfaceBlockIndex[uniformBlockIndex];
+
+ shProg->BufferInterfaceBlocks[interface_block_index].Binding =
+ uniformBlockBinding;
for (i = 0; i < MESA_SHADER_STAGES; i++) {
- int stage_index = shProg->InterfaceBlockStageIndex[i][uniformBlockIndex];
+ int stage_index =
+ shProg->InterfaceBlockStageIndex[i][interface_block_index];
if (stage_index != -1) {
struct gl_shader *sh = shProg->_LinkedShaders[i];
if (!shProg)
return;
- if (shaderStorageBlockIndex >= shProg->NumBufferInterfaceBlocks) {
+ if (shaderStorageBlockIndex >= shProg->NumShaderStorageBlocks) {
_mesa_error(ctx, GL_INVALID_VALUE,
"glShaderStorageBlockBinding(block index %u >= %u)",
- shaderStorageBlockIndex, shProg->NumBufferInterfaceBlocks);
+ shaderStorageBlockIndex, shProg->NumShaderStorageBlocks);
return;
}
return;
}
- if (shProg->BufferInterfaceBlocks[shaderStorageBlockIndex].Binding !=
+ if (shProg->ShaderStorageBlocks[shaderStorageBlockIndex]->Binding !=
shaderStorageBlockBinding) {
int i;
FLUSH_VERTICES(ctx, 0);
ctx->NewDriverState |= ctx->DriverFlags.NewShaderStorageBuffer;
- shProg->BufferInterfaceBlocks[shaderStorageBlockIndex].Binding = shaderStorageBlockBinding;
+ const int interface_block_index =
+ shProg->SsboInterfaceBlockIndex[shaderStorageBlockIndex];
+
+ shProg->BufferInterfaceBlocks[interface_block_index].Binding =
+ shaderStorageBlockBinding;
for (i = 0; i < MESA_SHADER_STAGES; i++) {
- int stage_index = shProg->InterfaceBlockStageIndex[i][shaderStorageBlockIndex];
+ int stage_index =
+ shProg->InterfaceBlockStageIndex[i][interface_block_index];
if (stage_index != -1) {
struct gl_shader *sh = shProg->_LinkedShaders[i];
}
static GLuint
-compute_version_es2(const struct gl_extensions *extensions)
+compute_version_es2(const struct gl_extensions *extensions,
+ const struct gl_constants *consts)
{
/* OpenGL ES 2.0 is derived from OpenGL 2.0 */
const bool ver_2_0 = (extensions->ARB_texture_cube_map &&
extensions->EXT_texture_snorm &&
extensions->NV_primitive_restart &&
extensions->OES_depth_texture_cube_map);
+ const bool es31_compute_shader =
+ consts->MaxComputeWorkGroupInvocations >= 128;
const bool ver_3_1 = (ver_3_0 &&
extensions->ARB_arrays_of_arrays &&
- extensions->ARB_compute_shader &&
+ es31_compute_shader &&
extensions->ARB_draw_indirect &&
extensions->ARB_explicit_uniform_location &&
extensions->ARB_framebuffer_no_attachments &&
case API_OPENGLES:
return compute_version_es1(extensions);
case API_OPENGLES2:
- return compute_version_es2(extensions);
+ return compute_version_es2(extensions, consts);
}
return 0;
}
/**
* Identity matrix.
*/
-static GLfloat Identity[16] = {
+static const GLfloat Identity[16] = {
1.0, 0.0, 0.0, 0.0,
0.0, 1.0, 0.0, 0.0,
0.0, 0.0, 1.0, 0.0,
if (MAT(in,0,0) == 0 || MAT(in,1,1) == 0 || MAT(in,2,2) == 0 )
return GL_FALSE;
- memcpy( out, Identity, 16 * sizeof(GLfloat) );
+ memcpy( out, Identity, sizeof(Identity) );
MAT(out,0,0) = 1.0F / MAT(in,0,0);
MAT(out,1,1) = 1.0F / MAT(in,1,1);
MAT(out,2,2) = 1.0F / MAT(in,2,2);
if (MAT(in,0,0) == 0 || MAT(in,1,1) == 0)
return GL_FALSE;
- memcpy( out, Identity, 16 * sizeof(GLfloat) );
+ memcpy( out, Identity, sizeof(Identity) );
MAT(out,0,0) = 1.0F / MAT(in,0,0);
MAT(out,1,1) = 1.0F / MAT(in,1,1);
if (MAT(in,2,3) == 0)
return GL_FALSE;
- memcpy( out, Identity, 16 * sizeof(GLfloat) );
+ memcpy( out, Identity, sizeof(Identity) );
MAT(out,0,0) = 1.0F / MAT(in,0,0);
MAT(out,1,1) = 1.0F / MAT(in,1,1);
s = sinf( angle * M_PI / 180.0 );
c = cosf( angle * M_PI / 180.0 );
- memcpy(m, Identity, sizeof(GLfloat)*16);
+ memcpy(m, Identity, sizeof(Identity));
optimized = GL_FALSE;
#define M(row,col) m[col*4+row]
void
_math_matrix_set_identity( GLmatrix *mat )
{
- memcpy( mat->m, Identity, 16*sizeof(GLfloat) );
- memcpy( mat->inv, Identity, 16*sizeof(GLfloat) );
+ memcpy( mat->m, Identity, sizeof(Identity) );
+ memcpy( mat->inv, Identity, sizeof(Identity) );
mat->type = MATRIX_IDENTITY;
mat->flags &= ~(MAT_DIRTY_FLAGS|
void
_math_matrix_copy( GLmatrix *to, const GLmatrix *from )
{
- memcpy( to->m, from->m, sizeof(Identity) );
+ memcpy(to->m, from->m, 16 * sizeof(GLfloat));
memcpy(to->inv, from->inv, 16 * sizeof(GLfloat));
to->flags = from->flags;
to->type = from->type;
case GLSL_TYPE_VOID:
case GLSL_TYPE_ERROR:
case GLSL_TYPE_INTERFACE:
+ case GLSL_TYPE_FUNCTION:
assert(!"Invalid type in type_size");
break;
}
if (ir->operands[0]->type->is_vector() ||
ir->operands[1]->type->is_vector()) {
src_reg temp = get_temp(glsl_type::vec4_type);
- emit(ir, OPCODE_SNE, dst_reg(temp), op[0], op[1]);
+ if (ir->operands[0]->type->is_boolean() &&
+ ir->operands[1]->as_constant() &&
+ ir->operands[1]->as_constant()->is_zero()) {
+ temp = op[0];
+ } else {
+ emit(ir, OPCODE_SNE, dst_reg(temp), op[0], op[1]);
+ }
/* After the dot-product, the value will be an integer on the
* range [0,4]. Zero stays zero, and positive values become 1.0.
}
break;
- case ir_unop_any: {
- assert(ir->operands[0]->type->is_vector());
-
- /* After the dot-product, the value will be an integer on the
- * range [0,4]. Zero stays zero, and positive values become 1.0.
- */
- ir_to_mesa_instruction *const dp =
- emit_dp(ir, result_dst, op[0], op[0],
- ir->operands[0]->type->vector_elements);
- if (this->prog->Target == GL_FRAGMENT_PROGRAM_ARB) {
- /* The clamping to [0,1] can be done for free in the fragment
- * shader with a saturate.
- */
- dp->saturate = true;
- } else {
- /* Negating the result of the dot-product gives values on the range
- * [-4, 0]. Zero stays zero, and negative values become 1.0. This
- * is achieved using SLT.
- */
- src_reg slt_src = result_src;
- slt_src.negate = ~slt_src.negate;
- emit(ir, OPCODE_SLT, result_dst, slt_src, src_reg_for_float(0.0));
- }
- break;
- }
-
case ir_binop_logic_xor:
emit(ir, OPCODE_SNE, result_dst, op[0], op[1]);
break;
break;
case ir_binop_vector_extract:
- case ir_binop_bfm:
case ir_triop_fma:
- case ir_triop_bfi:
case ir_triop_bitfield_extract:
case ir_triop_vector_insert:
case ir_quadop_bitfield_insert:
case GLSL_TYPE_STRUCT:
case GLSL_TYPE_ERROR:
case GLSL_TYPE_INTERFACE:
+ case GLSL_TYPE_FUNCTION:
assert(!"Should not get here.");
break;
}
/**
+ * Make sure there are enough unused parameter slots. Reallocate the list
+ * if needed.
+ *
+ * \param paramList where to reserve parameter slots
+ * \param reserve_slots number of slots to reserve
+ */
+void
+_mesa_reserve_parameter_storage(struct gl_program_parameter_list *paramList,
+ unsigned reserve_slots)
+{
+ const GLuint oldNum = paramList->NumParameters;
+
+ if (oldNum + reserve_slots > paramList->Size) {
+ /* Need to grow the parameter list array (alloc some extra) */
+ paramList->Size = paramList->Size + 4 * reserve_slots;
+
+ /* realloc arrays */
+ paramList->Parameters =
+ realloc(paramList->Parameters,
+ paramList->Size * sizeof(struct gl_program_parameter));
+
+ paramList->ParameterValues = (gl_constant_value (*)[4])
+ _mesa_align_realloc(paramList->ParameterValues, /* old buf */
+ oldNum * 4 * sizeof(gl_constant_value),/* old sz */
+ paramList->Size*4*sizeof(gl_constant_value),/*new*/
+ 16);
+ }
+}
+
+
+/**
* Add a new parameter to a parameter list.
* Note that parameter values are usually 4-element GLfloat vectors.
* When size > 4 we'll allocate a sequential block of parameters to
assert(size > 0);
- if (oldNum + sz4 > paramList->Size) {
- /* Need to grow the parameter list array (alloc some extra) */
- paramList->Size = paramList->Size + 4 * sz4;
-
- /* realloc arrays */
- paramList->Parameters =
- realloc(paramList->Parameters,
- paramList->Size * sizeof(struct gl_program_parameter));
-
- paramList->ParameterValues = (gl_constant_value (*)[4])
- _mesa_align_realloc(paramList->ParameterValues, /* old buf */
- oldNum * 4 * sizeof(gl_constant_value),/* old sz */
- paramList->Size*4*sizeof(gl_constant_value),/*new*/
- 16);
- }
+ _mesa_reserve_parameter_storage(paramList, sz4);
if (!paramList->Parameters ||
!paramList->ParameterValues) {
return list ? list->NumParameters : 0;
}
+extern void
+_mesa_reserve_parameter_storage(struct gl_program_parameter_list *paramList,
+ unsigned reserve_slots);
+
extern GLint
_mesa_add_parameter(struct gl_program_parameter_list *paramList,
gl_register_file type, const char *name,
* single MAD.
* linear: fogcoord * -1/(end-start) + end/(end-start)
* exp: 2^-(density/ln(2) * fogcoord)
- * exp2: 2^-((density/(ln(2)^2) * fogcoord)^2)
+ * exp2: 2^-((density/(sqrt(ln(2))) * fogcoord)^2)
*/
value[0] = (ctx->Fog.End == ctx->Fog.Start)
? 1.0f : (GLfloat)(-1.0F / (ctx->Fog.End - ctx->Fog.Start));
ptn_kil(nir_builder *b, nir_alu_dest dest, nir_ssa_def **src)
{
nir_ssa_def *cmp = b->shader->options->native_integers ?
- nir_bany4(b, nir_flt(b, src[0], nir_imm_float(b, 0.0))) :
- nir_fany4(b, nir_slt(b, src[0], nir_imm_float(b, 0.0)));
+ nir_bany_inequal4(b, nir_flt(b, src[0], nir_imm_float(b, 0.0)), nir_imm_int(b, 0)) :
+ nir_fany_nequal4(b, nir_slt(b, src[0], nir_imm_float(b, 0.0)), nir_imm_float(b, 0.0));
nir_intrinsic_instr *discard =
nir_intrinsic_instr_create(b->shader, nir_intrinsic_discard_if);
instr->op = op;
instr->dest_type = nir_type_float;
instr->is_shadow = prog_inst->TexShadow;
+ instr->texture_index = prog_inst->TexSrcUnit;
instr->sampler_index = prog_inst->TexSrcUnit;
switch (prog_inst->TexSrcTarget) {
nir_intrinsic_instr *store =
nir_intrinsic_instr_create(b->shader, nir_intrinsic_store_var);
store->num_components = glsl_get_vector_elements(var->type);
+ store->const_index[0] = (1 << store->num_components) - 1;
store->variables[0] =
nir_deref_var_create(store, c->output_vars[var->data.location]);
nir_intrinsic_instr *store =
nir_intrinsic_instr_create(shader, nir_intrinsic_store_var);
store->num_components = 4;
+ store->const_index[0] = WRITEMASK_XYZW;
store->variables[0] = nir_deref_var_create(store, fullvar);
store->src[0] = nir_src_for_ssa(f001);
nir_builder_instr_insert(b, &store->instr);
c = rzalloc(NULL, struct ptn_compile);
if (!c)
return NULL;
- s = nir_shader_create(NULL, stage, options);
- if (!s)
- goto fail;
c->prog = prog;
+ nir_builder_init_simple_shader(&c->build, NULL, stage, options);
+ s = c->build.shader;
+
if (prog->Parameters->NumParameters > 0) {
c->parameters = rzalloc(s, nir_variable);
c->parameters->type =
exec_list_push_tail(&s->uniforms, &c->parameters->node);
}
- nir_function *func = nir_function_create(s, "main");
- nir_function_overload *overload = nir_function_overload_create(func);
- nir_function_impl *impl = nir_function_impl_create(overload);
-
- c->build.shader = s;
- c->build.impl = impl;
- c->build.cursor = nir_after_cf_list(&impl->body);
-
setup_registers_and_variables(c);
if (unlikely(c->error))
goto fail;
}
}
}
+
+void
+_mesa_program_fragment_position_to_sysval(struct gl_program *prog)
+{
+ GLuint i;
+
+ if (prog->Target != GL_FRAGMENT_PROGRAM_ARB ||
+ !(prog->InputsRead & BITFIELD64_BIT(VARYING_SLOT_POS)))
+ return;
+
+ prog->InputsRead &= ~BITFIELD64_BIT(VARYING_SLOT_POS);
+ prog->SystemValuesRead |= 1 << SYSTEM_VALUE_FRAG_COORD;
+
+ for (i = 0; i < prog->NumInstructions; i++) {
+ struct prog_instruction *inst = prog->Instructions + i;
+ const GLuint numSrc = _mesa_num_inst_src_regs(inst->Opcode);
+ GLuint j;
+
+ for (j = 0; j < numSrc; j++) {
+ if (inst->SrcReg[j].File == PROGRAM_INPUT &&
+ inst->SrcReg[j].Index == VARYING_SLOT_POS) {
+ inst->SrcReg[j].File = PROGRAM_SYSTEM_VALUE;
+ inst->SrcReg[j].Index = SYSTEM_VALUE_FRAG_COORD;
+ }
+ }
+ }
+}
extern void
_mesa_remove_output_reads(struct gl_program *prog, gl_register_file type);
+extern void
+_mesa_program_fragment_position_to_sysval(struct gl_program *prog);
#ifdef __cplusplus
}
#include "pipe/p_defines.h"
#include "st_context.h"
#include "st_atom.h"
-#include "st_cb_bitmap.h"
#include "st_program.h"
#include "st_manager.h"
}
-/***********************************************************************
- */
-static GLboolean check_state( const struct st_state_flags *a,
- const struct st_state_flags *b )
+static bool
+check_state(const struct st_state_flags *a, const struct st_state_flags *b)
{
- return ((a->mesa & b->mesa) ||
- (a->st & b->st));
+ return (a->mesa & b->mesa) || (a->st & b->st);
}
-static void accumulate_state( struct st_state_flags *a,
- const struct st_state_flags *b )
+
+static void
+accumulate_state(struct st_state_flags *a, const struct st_state_flags *b)
{
a->mesa |= b->mesa;
a->st |= b->st;
}
-static void xor_states( struct st_state_flags *result,
- const struct st_state_flags *a,
- const struct st_state_flags *b )
+static void
+xor_states(struct st_state_flags *result,
+ const struct st_state_flags *a,
+ const struct st_state_flags *b)
{
result->mesa = a->mesa ^ b->mesa;
result->st = a->st ^ b->st;
check_attrib_edgeflag(st);
- if (state->mesa)
- st_flush_bitmap_cache(st);
-
check_program_state( st );
st_manager_validate_framebuffers(st);
- if (state->st == 0)
+ if (state->st == 0 && state->mesa == 0)
return;
/*printf("%s %x/%x\n", __func__, state->mesa, state->st);*/
memset(state, 0, sizeof(*state));
}
-
-
-
cb.buffer = NULL;
cb.user_buffer = NULL;
u_upload_data(st->constbuf_uploader, 0, paramBytes,
+ st->ctx->Const.UniformBufferOffsetAlignment,
params->ParameterValues, &cb.buffer_offset, &cb.buffer);
u_upload_unmap(st->constbuf_uploader);
} else {
raster->line_smooth = ctx->Line.SmoothFlag;
if (ctx->Line.SmoothFlag) {
raster->line_width = CLAMP(ctx->Line.Width,
- ctx->Const.MinLineWidthAA,
- ctx->Const.MaxLineWidthAA);
+ ctx->Const.MinLineWidthAA,
+ ctx->Const.MaxLineWidthAA);
}
else {
raster->line_width = CLAMP(ctx->Line.Width,
- ctx->Const.MinLineWidth,
- ctx->Const.MaxLineWidth);
+ ctx->Const.MinLineWidth,
+ ctx->Const.MaxLineWidth);
}
raster->line_stipple_enable = ctx->Line.StippleFlag;
tBot = (GLfloat) height;
}
- u_upload_alloc(st->uploader, 0, 4 * sizeof(vertices[0]),
+ u_upload_alloc(st->uploader, 0, 4 * sizeof(vertices[0]), 4,
vbuf_offset, vbuf, (void **) &vertices);
if (!*vbuf) {
return;
GLfloat colorSave[4];
COPY_4V(colorSave, ctx->Current.Attrib[VERT_ATTRIB_COLOR0]);
COPY_4V(ctx->Current.Attrib[VERT_ATTRIB_COLOR0], color);
- st_upload_constants(st, fpv->parameters, PIPE_SHADER_FRAGMENT);
+ st_upload_constants(st, st->fp->Base.Base.Parameters,
+ PIPE_SHADER_FRAGMENT);
COPY_4V(ctx->Current.Attrib[VERT_ATTRIB_COLOR0], colorSave);
}
cso_restore_stream_outputs(cso);
pipe_resource_reference(&vbuf, NULL);
+
+ /* We uploaded modified constants, need to invalidate them. */
+ st->dirty.mesa |= _NEW_PROGRAM_CONSTANTS;
}
assert(cache->xmin <= cache->xmax);
-/* printf("flush size %d x %d at %d, %d\n",
- cache->xmax - cache->xmin,
- cache->ymax - cache->ymin,
- cache->xpos, cache->ypos);
-*/
+ if (0)
+ printf("flush bitmap, size %d x %d at %d, %d\n",
+ cache->xmax - cache->xmin,
+ cache->ymax - cache->ymin,
+ cache->xpos, cache->ypos);
/* The texture transfer has been mapped until now.
- * So unmap and release the texture transfer before drawing.
- */
+ * So unmap and release the texture transfer before drawing.
+ */
if (cache->trans && cache->buffer) {
if (0)
print_cache(cache);
struct st_context *st = st_context(ctx);
struct pipe_resource *pt;
- if (width == 0 || height == 0)
- return;
+ assert(width > 0);
+ assert(height > 0);
- st_validate_state(st);
+ /* We only need to validate state of the st dirty flags are set or
+ * any non-_NEW_PROGRAM_CONSTANTS mesa flags are set. The VS we use
+ * for bitmap drawing uses no constants and the FS constants are
+ * explicitly uploaded in the draw_bitmap_quad() function.
+ */
+ if ((st->dirty.mesa & ~_NEW_PROGRAM_CONSTANTS) || st->dirty.st) {
+ st_validate_state(st);
+ }
if (!st->bitmap.vs) {
/* create pass-through vertex shader now */
if (st_obj->buffer)
pipe_resource_reference(&st_obj->buffer, NULL);
- mtx_destroy(&st_obj->Base.Mutex);
- free(st_obj->Base.Label);
- free(st_obj);
+ _mesa_delete_buffer_object(ctx, obj);
}
{
struct st_context *st = st_context(ctx);
struct pipe_context *pipe = st->pipe;
+ struct pipe_screen *screen = pipe->screen;
struct st_buffer_object *st_obj = st_buffer_object(obj);
unsigned bind, pipe_usage, pipe_flags = 0;
if (target != GL_EXTERNAL_VIRTUAL_MEMORY_BUFFER_AMD &&
- size && data && st_obj->buffer &&
+ size && st_obj->buffer &&
st_obj->Base.Size == size &&
st_obj->Base.Usage == usage &&
st_obj->Base.StorageFlags == storageFlags) {
- /* Just discard the old contents and write new data.
- * This should be the same as creating a new buffer, but we avoid
- * a lot of validation in Mesa.
- */
- struct pipe_box box;
-
- u_box_1d(0, size, &box);
- pipe->transfer_inline_write(pipe, st_obj->buffer, 0,
- PIPE_TRANSFER_DISCARD_WHOLE_RESOURCE,
- &box, data, 0, 0);
- return GL_TRUE;
+ if (data) {
+ /* Just discard the old contents and write new data.
+ * This should be the same as creating a new buffer, but we avoid
+ * a lot of validation in Mesa.
+ */
+ struct pipe_box box;
+
+ u_box_1d(0, size, &box);
+ pipe->transfer_inline_write(pipe, st_obj->buffer, 0,
+ PIPE_TRANSFER_DISCARD_WHOLE_RESOURCE,
+ &box, data, 0, 0);
+ return GL_TRUE;
+ } else if (screen->get_param(screen, PIPE_CAP_INVALIDATE_BUFFER)) {
+ pipe->invalidate_resource(pipe, st_obj->buffer);
+ return GL_TRUE;
+ }
}
st_obj->Base.Size = size;
bind = PIPE_BIND_CONSTANT_BUFFER;
break;
case GL_DRAW_INDIRECT_BUFFER:
+ case GL_PARAMETER_BUFFER_ARB:
bind = PIPE_BIND_COMMAND_ARGS_BUFFER;
break;
default:
}
if (size != 0) {
- struct pipe_screen *screen = pipe->screen;
struct pipe_resource buffer;
memset(&buffer, 0, sizeof buffer);
/**
+ * Called via glInvalidateBuffer(Sub)Data.
+ */
+static void
+st_bufferobj_invalidate(struct gl_context *ctx,
+ struct gl_buffer_object *obj,
+ GLintptr offset,
+ GLsizeiptr size)
+{
+ struct st_context *st = st_context(ctx);
+ struct pipe_context *pipe = st->pipe;
+ struct st_buffer_object *st_obj = st_buffer_object(obj);
+
+ /* We ignore partial invalidates. */
+ if (offset != 0 || size != obj->Size)
+ return;
+
+ /* Nothing to invalidate. */
+ if (!st_obj->buffer)
+ return;
+
+ pipe->invalidate_resource(pipe, st_obj->buffer);
+}
+
+
+/**
* Called via glMapBufferRange().
*/
static void *
void
-st_init_bufferobject_functions(struct dd_function_table *functions)
+st_init_bufferobject_functions(struct pipe_screen *screen,
+ struct dd_function_table *functions)
{
/* plug in default driver fallbacks (such as for ClearBufferSubData) */
_mesa_init_buffer_object_functions(functions);
functions->UnmapBuffer = st_bufferobj_unmap;
functions->CopyBufferSubData = st_copy_buffer_subdata;
functions->ClearBufferSubData = st_clear_buffer_subdata;
+
+ if (screen->get_param(screen, PIPE_CAP_INVALIDATE_BUFFER))
+ functions->InvalidateBufferSubData = st_bufferobj_invalidate;
}
struct dd_function_table;
struct pipe_resource;
+struct pipe_screen;
struct st_context;
/**
extern void
-st_init_bufferobject_functions(struct dd_function_table *functions);
+st_init_bufferobject_functions(struct pipe_screen *screen,
+ struct dd_function_table *functions);
#endif
#include "program/prog_instruction.h"
#include "st_context.h"
#include "st_atom.h"
+#include "st_cb_bitmap.h"
#include "st_cb_clear.h"
#include "st_cb_fbo.h"
#include "st_format.h"
vb.stride = 8 * sizeof(float);
- u_upload_alloc(st->uploader, 0, 4 * sizeof(vertices[0]),
+ u_upload_alloc(st->uploader, 0, 4 * sizeof(vertices[0]), 4,
&vb.buffer_offset, &vb.buffer,
(void **) &vertices);
if (!vb.buffer) {
GLbitfield clear_buffers = 0x0;
GLuint i;
+ st_flush_bitmap_cache(st);
+
/* This makes sure the pipe has the latest scissor, etc values */
st_validate_state( st );
#include "st_atom.h"
#include "st_atom_constbuf.h"
+#include "st_cb_bitmap.h"
#include "st_cb_drawpixels.h"
#include "st_cb_readpixels.h"
#include "st_cb_fbo.h"
struct pipe_resource *buf = NULL;
unsigned offset;
- u_upload_alloc(st->uploader, 0, 4 * sizeof(verts[0]), &offset,
+ u_upload_alloc(st->uploader, 0, 4 * sizeof(verts[0]), 4, &offset,
&buf, (void **) &verts);
if (!buf) {
return;
/* Mesa state should be up to date by now */
assert(ctx->NewState == 0x0);
+ st_flush_bitmap_cache(st);
+
st_validate_state(st);
/* Limit the size of the glDrawPixels to the max texture size.
num_sampler_view++;
}
- /* update fragment program constants */
- st_upload_constants(st, fpv->parameters, PIPE_SHADER_FRAGMENT);
+ /* compiling a new fragment shader variant added new state constants
+ * into the constant buffer, we need to update them
+ */
+ st_upload_constants(st, st->fp->Base.Base.Parameters,
+ PIPE_SHADER_FRAGMENT);
}
/* Put glDrawPixels image into a texture */
ctx->_ImageTransferState == 0x0 &&
!ctx->Color.BlendEnabled &&
!ctx->Color.AlphaEnabled &&
+ (!ctx->Color.ColorLogicOpEnabled || ctx->Color.LogicOp == GL_COPY) &&
!ctx->Depth.Test &&
!ctx->Fog.Enabled &&
!ctx->Stencil.Enabled &&
GLint readX, readY, readW, readH;
struct gl_pixelstore_attrib pack = ctx->DefaultPacking;
+ st_flush_bitmap_cache(st);
+
st_validate_state(st);
if (type == GL_DEPTH_STENCIL) {
num_sampler_view++;
}
- /* update fragment program constants */
- st_upload_constants(st, fpv->parameters, PIPE_SHADER_FRAGMENT);
+ /* compiling a new fragment shader variant added new state constants
+ * into the constant buffer, we need to update them
+ */
+ st_upload_constants(st, st->fp->Base.Base.Parameters,
+ PIPE_SHADER_FRAGMENT);
}
else {
assert(type == GL_DEPTH);
#include "st_context.h"
#include "st_atom.h"
+#include "st_cb_bitmap.h"
#include "st_cb_drawtex.h"
#include "pipe/p_context.h"
struct pipe_vertex_element velements[2 + MAX_TEXTURE_UNITS];
unsigned offset;
+ st_flush_bitmap_cache(st);
+
st_validate_state(st);
/* determine if we need vertex color */
GLuint attr;
u_upload_alloc(st->uploader, 0,
- numAttribs * 4 * 4 * sizeof(GLfloat),
+ numAttribs * 4 * 4 * sizeof(GLfloat), 4,
&offset, &vbuffer, (void **) &vbuf);
if (!vbuffer) {
return;
/**
+ * Try to determine whether we should allocate memory for a full texture
+ * mipmap. The problem is when we get a glTexImage(level=0) call, we
+ * can't immediately know if other mipmap levels are coming next. Here
+ * we try to guess whether to allocate memory for a mipmap or just the
+ * 0th level.
+ *
+ * If we guess incorrectly here we'll later reallocate the right amount of
+ * memory either in st_AllocTextureImageBuffer() or st_finalize_texture().
+ *
+ * \param stObj the texture object we're going to allocate memory for.
+ * \param stImage describes the incoming image which we need to store.
+ */
+static boolean
+allocate_full_mipmap(const struct st_texture_object *stObj,
+ const struct st_texture_image *stImage)
+{
+ switch (stObj->base.Target) {
+ case GL_TEXTURE_RECTANGLE_NV:
+ case GL_TEXTURE_BUFFER:
+ case GL_TEXTURE_EXTERNAL_OES:
+ case GL_TEXTURE_2D_MULTISAMPLE:
+ case GL_TEXTURE_2D_MULTISAMPLE_ARRAY:
+ /* these texture types cannot be mipmapped */
+ return FALSE;
+ }
+
+ if (stImage->base.Level > 0 || stObj->base.GenerateMipmap)
+ return TRUE;
+
+ if (stImage->base._BaseFormat == GL_DEPTH_COMPONENT ||
+ stImage->base._BaseFormat == GL_DEPTH_STENCIL_EXT)
+ /* depth/stencil textures are seldom mipmapped */
+ return FALSE;
+
+ if (stObj->base.BaseLevel == 0 && stObj->base.MaxLevel == 0)
+ return FALSE;
+
+ if (stObj->base.Sampler.MinFilter == GL_NEAREST ||
+ stObj->base.Sampler.MinFilter == GL_LINEAR)
+ /* not a mipmap minification filter */
+ return FALSE;
+
+ if (stObj->base.Target == GL_TEXTURE_3D)
+ /* 3D textures are seldom mipmapped */
+ return FALSE;
+
+ return TRUE;
+}
+
+
+/**
* Try to allocate a pipe_resource object for the given st_texture_object.
*
* We use the given st_texture_image as a clue to determine the size of the
* to re-allocating a texture buffer with space for more (or fewer)
* mipmap levels later.
*/
- if ((stObj->base.Sampler.MinFilter == GL_NEAREST ||
- stObj->base.Sampler.MinFilter == GL_LINEAR ||
- (stObj->base.BaseLevel == 0 &&
- stObj->base.MaxLevel == 0) ||
- stImage->base._BaseFormat == GL_DEPTH_COMPONENT ||
- stImage->base._BaseFormat == GL_DEPTH_STENCIL_EXT) &&
- !stObj->base.GenerateMipmap &&
- stImage->base.Level == 0) {
- /* only alloc space for a single mipmap level */
- lastLevel = 0;
- }
- else {
+ if (allocate_full_mipmap(stObj, stImage)) {
/* alloc space for a full mipmap */
lastLevel = _mesa_get_tex_max_num_levels(stObj->base.Target,
width, height, depth) - 1;
}
+ else {
+ /* only alloc space for a single mipmap level */
+ lastLevel = 0;
+ }
/* Save the level=0 dimensions */
stObj->width0 = width;
/**
+ * Called via ctx->Driver.Enable()
+ */
+static void st_Enable(struct gl_context * ctx, GLenum cap, GLboolean state)
+{
+ struct st_context *st = st_context(ctx);
+
+ switch (cap) {
+ case GL_DEBUG_OUTPUT:
+ st_enable_debug_output(st, state);
+ break;
+ default:
+ break;
+ }
+}
+
+
+/**
* Called via ctx->Driver.UpdateState()
*/
-void st_invalidate_state(struct gl_context * ctx, GLuint new_state)
+void st_invalidate_state(struct gl_context * ctx, GLbitfield new_state)
{
struct st_context *st = st_context(ctx);
/* Create upload manager for vertex data for glBitmap, glDrawPixels,
* glClear, etc.
*/
- st->uploader = u_upload_create(st->pipe, 65536, 4, PIPE_BIND_VERTEX_BUFFER);
+ st->uploader = u_upload_create(st->pipe, 65536, PIPE_BIND_VERTEX_BUFFER,
+ PIPE_USAGE_STREAM);
if (!screen->get_param(screen, PIPE_CAP_USER_INDEX_BUFFERS)) {
- st->indexbuf_uploader = u_upload_create(st->pipe, 128 * 1024, 4,
- PIPE_BIND_INDEX_BUFFER);
+ st->indexbuf_uploader = u_upload_create(st->pipe, 128 * 1024,
+ PIPE_BIND_INDEX_BUFFER,
+ PIPE_USAGE_STREAM);
}
- if (!screen->get_param(screen, PIPE_CAP_USER_CONSTANT_BUFFERS)) {
- unsigned alignment =
- screen->get_param(screen, PIPE_CAP_CONSTANT_BUFFER_OFFSET_ALIGNMENT);
-
- st->constbuf_uploader = u_upload_create(pipe, 128 * 1024, alignment,
- PIPE_BIND_CONSTANT_BUFFER);
- }
+ if (!screen->get_param(screen, PIPE_CAP_USER_CONSTANT_BUFFERS))
+ st->constbuf_uploader = u_upload_create(pipe, 128 * 1024,
+ PIPE_BIND_CONSTANT_BUFFER,
+ PIPE_USAGE_STREAM);
st->cso_context = cso_create_context(pipe);
PIPE_QUIRK_TEXTURE_BORDER_COLOR_SWIZZLE_R600));
st->has_time_elapsed =
screen->get_param(screen, PIPE_CAP_QUERY_TIME_ELAPSED);
+ st->has_half_float_packing =
+ screen->get_param(screen, PIPE_CAP_TGSI_PACK_HALF_FLOAT);
+ st->has_multi_draw_indirect =
+ screen->get_param(screen, PIPE_CAP_MULTI_DRAW_INDIRECT);
/* GL limits and extensions */
st_init_limits(st->pipe->screen, &ctx->Const, &ctx->Extensions);
_mesa_init_sampler_object_functions(functions);
st_init_blit_functions(functions);
- st_init_bufferobject_functions(functions);
+ st_init_bufferobject_functions(screen, functions);
st_init_clear_functions(functions);
st_init_bitmap_functions(functions);
st_init_copy_image_functions(functions);
st_init_vdpau_functions(functions);
+ functions->Enable = st_Enable;
functions->UpdateState = st_invalidate_state;
}
struct st_state_flags {
- GLuint mesa;
- uint64_t st;
+ GLbitfield mesa; /**< Mask of _NEW_x flags */
+ uint64_t st; /**< Mask of ST_NEW_x flags */
};
struct st_tracked_state {
boolean prefer_blit_based_texture_transfer;
boolean force_persample_in_shader;
boolean has_shareable_shaders;
+ boolean has_half_float_packing;
+ boolean has_multi_draw_indirect;
/**
* If a shader can be created when we get its source.
extern void st_init_driver_functions(struct pipe_screen *screen,
struct dd_function_table *functions);
-void st_invalidate_state(struct gl_context * ctx, GLuint new_state);
+void st_invalidate_state(struct gl_context * ctx, GLbitfield new_state);
_mesa_GetIntegerv(GL_READ_BUFFER, &readBufSave);
/* Read from the winsys buffer */
- _mesa_BindFramebuffer(GL_READ_BUFFER, 0);
+ _mesa_BindFramebuffer(GL_READ_FRAMEBUFFER, 0);
_mesa_ReadBuffer(srcBuffer);
/* copy image from pbuffer to texture */
/* restore readbuffer */
_mesa_ReadBuffer(readBufSave);
- _mesa_BindFramebuffer(GL_READ_BUFFER, readFBOSave);
+ _mesa_BindFramebuffer(GL_READ_FRAMEBUFFER, readFBOSave);
}
}
+/**
+ * Installed as pipe_debug_callback when GL_DEBUG_OUTPUT is enabled.
+ */
+static void
+st_debug_message(void *data,
+ unsigned *id,
+ enum pipe_debug_type ptype,
+ const char *fmt,
+ va_list args)
+{
+ struct st_context *st = data;
+ enum mesa_debug_source source;
+ enum mesa_debug_type type;
+ enum mesa_debug_severity severity;
+
+ switch (ptype) {
+ case PIPE_DEBUG_TYPE_OUT_OF_MEMORY:
+ source = MESA_DEBUG_SOURCE_API;
+ type = MESA_DEBUG_TYPE_ERROR;
+ severity = MESA_DEBUG_SEVERITY_MEDIUM;
+ break;
+ case PIPE_DEBUG_TYPE_ERROR:
+ source = MESA_DEBUG_SOURCE_API;
+ type = MESA_DEBUG_TYPE_ERROR;
+ severity = MESA_DEBUG_SEVERITY_MEDIUM;
+ break;
+ case PIPE_DEBUG_TYPE_SHADER_INFO:
+ source = MESA_DEBUG_SOURCE_SHADER_COMPILER;
+ type = MESA_DEBUG_TYPE_OTHER;
+ severity = MESA_DEBUG_SEVERITY_NOTIFICATION;
+ break;
+ case PIPE_DEBUG_TYPE_PERF_INFO:
+ source = MESA_DEBUG_SOURCE_API;
+ type = MESA_DEBUG_TYPE_PERFORMANCE;
+ severity = MESA_DEBUG_SEVERITY_NOTIFICATION;
+ break;
+ case PIPE_DEBUG_TYPE_INFO:
+ source = MESA_DEBUG_SOURCE_API;
+ type = MESA_DEBUG_TYPE_OTHER;
+ severity = MESA_DEBUG_SEVERITY_NOTIFICATION;
+ break;
+ case PIPE_DEBUG_TYPE_FALLBACK:
+ source = MESA_DEBUG_SOURCE_API;
+ type = MESA_DEBUG_TYPE_PERFORMANCE;
+ severity = MESA_DEBUG_SEVERITY_NOTIFICATION;
+ break;
+ case PIPE_DEBUG_TYPE_CONFORMANCE:
+ source = MESA_DEBUG_SOURCE_API;
+ type = MESA_DEBUG_TYPE_OTHER;
+ severity = MESA_DEBUG_SEVERITY_NOTIFICATION;
+ break;
+ default:
+ unreachable("invalid debug type");
+ }
+ _mesa_gl_vdebug(st->ctx, id, source, type, severity, fmt, args);
+}
+
+void
+st_enable_debug_output(struct st_context *st, boolean enable)
+{
+ struct pipe_context *pipe = st->pipe;
+
+ if (!pipe->set_debug_callback)
+ return;
+
+ if (enable) {
+ struct pipe_debug_callback cb = { st_debug_message, st };
+ pipe->set_debug_callback(pipe, &cb);
+ } else {
+ pipe->set_debug_callback(pipe, NULL);
+ }
+}
#include "pipe/p_compiler.h"
#include "util/u_debug.h"
+struct st_context;
+
extern void
st_print_current(void);
void st_debug_init( void );
+void st_enable_debug_output(struct st_context *st, boolean enable);
+
static inline void
ST_DBG( unsigned flag, const char *fmt, ... )
{
#include "st_context.h"
#include "st_atom.h"
+#include "st_cb_bitmap.h"
#include "st_cb_bufferobjects.h"
#include "st_cb_xformfb.h"
#include "st_debug.h"
else if (st->indexbuf_uploader) {
/* upload indexes from user memory into a real buffer */
u_upload_data(st->indexbuf_uploader, 0,
- ib->count * ibuffer->index_size, ib->ptr,
+ ib->count * ibuffer->index_size, 4, ib->ptr,
&ibuffer->offset, &ibuffer->buffer);
if (!ibuffer->buffer) {
/* out of memory */
/* Mesa core state should have been validated already */
assert(ctx->NewState == 0x0);
+ st_flush_bitmap_cache(st);
+
/* Validate state. */
if (st->dirty.st || ctx->NewDriverState) {
st_validate_state(st);
}
}
- if (indirect) {
- info.indirect = st_buffer_object(indirect)->buffer;
-
- /* Primitive restart is not handled by the VBO module in this case. */
- info.primitive_restart = ctx->Array._PrimitiveRestart;
- info.restart_index = ctx->Array.RestartIndex;
- }
+ assert(!indirect);
/* do actual drawing */
for (i = 0; i < nr_prims; i++) {
info.instance_count = prims[i].num_instances;
info.vertices_per_patch = ctx->TessCtrlProgram.patch_vertices;
info.index_bias = prims[i].basevertex;
+ info.drawid = prims[i].draw_id;
if (!ib) {
info.min_index = info.start;
info.max_index = info.start + info.count - 1;
}
- info.indirect_offset = prims[i].indirect_offset;
if (ST_DEBUG & DEBUG_DRAW) {
debug_printf("st/draw: mode %s start %u count %u indexed %d\n",
info.indexed);
}
- if (info.count_from_stream_output || info.indirect) {
+ if (info.count_from_stream_output) {
cso_draw_vbo(st->cso_context, &info);
}
else if (info.primitive_restart) {
}
}
+static void
+st_indirect_draw_vbo(struct gl_context *ctx,
+ GLuint mode,
+ struct gl_buffer_object *indirect_data,
+ GLsizeiptr indirect_offset,
+ unsigned draw_count,
+ unsigned stride,
+ struct gl_buffer_object *indirect_params,
+ GLsizeiptr indirect_params_offset,
+ const struct _mesa_index_buffer *ib)
+{
+ struct st_context *st = st_context(ctx);
+ struct pipe_index_buffer ibuffer = {0};
+ struct pipe_draw_info info;
+
+ /* Mesa core state should have been validated already */
+ assert(ctx->NewState == 0x0);
+ assert(stride);
+
+ /* Validate state. */
+ if (st->dirty.st || ctx->NewDriverState) {
+ st_validate_state(st);
+ }
+
+ if (st->vertex_array_out_of_memory) {
+ return;
+ }
+
+ util_draw_init_info(&info);
+
+ if (ib) {
+ if (!setup_index_buffer(st, ib, &ibuffer)) {
+ _mesa_error(ctx, GL_OUT_OF_MEMORY, "gl%sDrawElementsIndirect%s",
+ (draw_count > 1) ? "Multi" : "",
+ indirect_params ? "CountARB" : "");
+ return;
+ }
+
+ info.indexed = TRUE;
+ }
+
+ info.mode = translate_prim(ctx, mode);
+ info.vertices_per_patch = ctx->TessCtrlProgram.patch_vertices;
+ info.indirect = st_buffer_object(indirect_data)->buffer;
+ info.indirect_offset = indirect_offset;
+
+ /* Primitive restart is not handled by the VBO module in this case. */
+ info.primitive_restart = ctx->Array._PrimitiveRestart;
+ info.restart_index = ctx->Array.RestartIndex;
+
+ if (ST_DEBUG & DEBUG_DRAW) {
+ debug_printf("st/draw indirect: mode %s drawcount %d indexed %d\n",
+ u_prim_name(info.mode),
+ draw_count,
+ info.indexed);
+ }
+
+ if (!st->has_multi_draw_indirect) {
+ int i;
+
+ assert(!indirect_params);
+ info.indirect_count = 1;
+ for (i = 0; i < draw_count; i++) {
+ info.drawid = i;
+ cso_draw_vbo(st->cso_context, &info);
+ info.indirect_offset += stride;
+ }
+ } else {
+ info.indirect_count = draw_count;
+ info.indirect_stride = stride;
+ if (indirect_params) {
+ info.indirect_params = st_buffer_object(indirect_params)->buffer;
+ info.indirect_params_offset = indirect_params_offset;
+ }
+ cso_draw_vbo(st->cso_context, &info);
+ }
+}
+
void
st_init_draw(struct st_context *st)
struct gl_context *ctx = st->ctx;
vbo_set_draw_func(ctx, st_draw_vbo);
+ vbo_set_indirect_draw_func(ctx, st_indirect_draw_vbo);
st->draw = draw_create(st->pipe); /* for selection/feedback */
#include "st_context.h"
#include "st_atom.h"
+#include "st_cb_bitmap.h"
#include "st_cb_bufferobjects.h"
#include "st_draw.h"
#include "st_program.h"
assert(draw);
+ st_flush_bitmap_cache(st);
+
st_validate_state(st);
if (!index_bounds_valid)
c->GLSLSkipStrictMaxUniformLimitCheck =
screen->get_param(screen, PIPE_CAP_TGSI_CAN_COMPACT_CONSTANTS);
+ c->UniformBufferOffsetAlignment =
+ screen->get_param(screen, PIPE_CAP_CONSTANT_BUFFER_OFFSET_ALIGNMENT);
+
if (can_ubo) {
extensions->ARB_uniform_buffer_object = GL_TRUE;
- c->UniformBufferOffsetAlignment =
- screen->get_param(screen, PIPE_CAP_CONSTANT_BUFFER_OFFSET_ALIGNMENT);
c->MaxCombinedUniformBlocks = c->MaxUniformBufferBindings =
c->Program[MESA_SHADER_VERTEX].MaxUniformBlocks +
c->Program[MESA_SHADER_TESS_CTRL].MaxUniformBlocks +
c->Program[MESA_SHADER_FRAGMENT].MaxUniformBlocks;
assert(c->MaxCombinedUniformBlocks <= MAX_COMBINED_UNIFORM_BUFFERS);
}
+
+ c->GLSLFragCoordIsSysVal =
+ screen->get_param(screen, PIPE_CAP_TGSI_FS_POSITION_IS_SYSVAL);
+ c->GLSLFrontFacingIsSysVal =
+ screen->get_param(screen, PIPE_CAP_TGSI_FS_FACE_IS_INTEGER_SYSVAL);
}
{ o(ARB_base_instance), PIPE_CAP_START_INSTANCE },
{ o(ARB_buffer_storage), PIPE_CAP_BUFFER_MAP_PERSISTENT_COHERENT },
{ o(ARB_clear_texture), PIPE_CAP_CLEAR_TEXTURE },
+ { o(ARB_clip_control), PIPE_CAP_CLIP_HALFZ },
{ o(ARB_color_buffer_float), PIPE_CAP_VERTEX_COLOR_UNCLAMPED },
+ { o(ARB_conditional_render_inverted), PIPE_CAP_CONDITIONAL_RENDER_INVERTED },
{ o(ARB_copy_image), PIPE_CAP_COPY_BETWEEN_COMPRESSED_AND_PLAIN_FORMATS },
{ o(ARB_depth_clamp), PIPE_CAP_DEPTH_CLIP_DISABLE },
{ o(ARB_depth_texture), PIPE_CAP_TEXTURE_SHADOW_MAP },
+ { o(ARB_derivative_control), PIPE_CAP_TGSI_FS_FINE_DERIVATIVE },
{ o(ARB_draw_buffers_blend), PIPE_CAP_INDEP_BLEND_FUNC },
+ { o(ARB_draw_indirect), PIPE_CAP_DRAW_INDIRECT },
{ o(ARB_draw_instanced), PIPE_CAP_TGSI_INSTANCEID },
{ o(ARB_fragment_program_shadow), PIPE_CAP_TEXTURE_SHADOW_MAP },
{ o(ARB_framebuffer_object), PIPE_CAP_MIXED_FRAMEBUFFER_SIZES },
+ { o(ARB_indirect_parameters), PIPE_CAP_MULTI_DRAW_INDIRECT_PARAMS },
{ o(ARB_instanced_arrays), PIPE_CAP_VERTEX_ELEMENT_INSTANCE_DIVISOR },
{ o(ARB_occlusion_query), PIPE_CAP_OCCLUSION_QUERY },
{ o(ARB_occlusion_query2), PIPE_CAP_OCCLUSION_QUERY },
{ o(ARB_pipeline_statistics_query), PIPE_CAP_QUERY_PIPELINE_STATISTICS },
{ o(ARB_point_sprite), PIPE_CAP_POINT_SPRITE },
+ { o(ARB_sample_shading), PIPE_CAP_SAMPLE_SHADING },
{ o(ARB_seamless_cube_map), PIPE_CAP_SEAMLESS_CUBE_MAP },
+ { o(ARB_shader_draw_parameters), PIPE_CAP_DRAW_PARAMETERS },
{ o(ARB_shader_stencil_export), PIPE_CAP_SHADER_STENCIL_EXPORT },
{ o(ARB_shader_texture_image_samples), PIPE_CAP_TGSI_TXQS },
{ o(ARB_shader_texture_lod), PIPE_CAP_SM3 },
{ o(ARB_shadow), PIPE_CAP_TEXTURE_SHADOW_MAP },
{ o(ARB_texture_buffer_object), PIPE_CAP_TEXTURE_BUFFER_OBJECTS },
+ { o(ARB_texture_cube_map_array), PIPE_CAP_CUBE_MAP_ARRAY },
{ o(ARB_texture_gather), PIPE_CAP_MAX_TEXTURE_GATHER_COMPONENTS },
{ o(ARB_texture_mirror_clamp_to_edge), PIPE_CAP_TEXTURE_MIRROR_CLAMP },
+ { o(ARB_texture_multisample), PIPE_CAP_TEXTURE_MULTISAMPLE },
{ o(ARB_texture_non_power_of_two), PIPE_CAP_NPOT_TEXTURES },
+ { o(ARB_texture_query_lod), PIPE_CAP_TEXTURE_QUERY_LOD },
+ { o(ARB_texture_view), PIPE_CAP_SAMPLER_VIEW_TARGET },
{ o(ARB_timer_query), PIPE_CAP_QUERY_TIMESTAMP },
{ o(ARB_transform_feedback2), PIPE_CAP_STREAM_OUTPUT_PAUSE_RESUME },
{ o(ARB_transform_feedback3), PIPE_CAP_STREAM_OUTPUT_PAUSE_RESUME },
{ o(EXT_blend_equation_separate), PIPE_CAP_BLEND_EQUATION_SEPARATE },
+ { o(EXT_depth_bounds_test), PIPE_CAP_DEPTH_BOUNDS_TEST },
{ o(EXT_draw_buffers2), PIPE_CAP_INDEP_BLEND_ENABLE },
+ { o(EXT_polygon_offset_clamp), PIPE_CAP_POLYGON_OFFSET_CLAMP },
{ o(EXT_stencil_two_side), PIPE_CAP_TWO_SIDED_STENCIL },
{ o(EXT_texture_array), PIPE_CAP_MAX_TEXTURE_ARRAY_LAYERS },
{ o(EXT_texture_filter_anisotropic), PIPE_CAP_ANISOTROPIC_FILTER },
{ o(OES_standard_derivatives), PIPE_CAP_SM3 },
{ o(OES_texture_float_linear), PIPE_CAP_TEXTURE_FLOAT_LINEAR },
{ o(OES_texture_half_float_linear), PIPE_CAP_TEXTURE_HALF_FLOAT_LINEAR },
- { o(ARB_texture_cube_map_array), PIPE_CAP_CUBE_MAP_ARRAY },
- { o(ARB_texture_multisample), PIPE_CAP_TEXTURE_MULTISAMPLE },
- { o(ARB_texture_query_lod), PIPE_CAP_TEXTURE_QUERY_LOD },
- { o(ARB_sample_shading), PIPE_CAP_SAMPLE_SHADING },
- { o(ARB_draw_indirect), PIPE_CAP_DRAW_INDIRECT },
- { o(ARB_derivative_control), PIPE_CAP_TGSI_FS_FINE_DERIVATIVE },
- { o(ARB_conditional_render_inverted), PIPE_CAP_CONDITIONAL_RENDER_INVERTED },
- { o(ARB_texture_view), PIPE_CAP_SAMPLER_VIEW_TARGET },
- { o(ARB_clip_control), PIPE_CAP_CLIP_HALFZ },
- { o(EXT_polygon_offset_clamp), PIPE_CAP_POLYGON_OFFSET_CLAMP },
- { o(EXT_depth_bounds_test), PIPE_CAP_DEPTH_BOUNDS_TEST },
};
/* Required: render target and sampler support */
struct st_texture_object *stObj = st_texture_object(texObj);
struct pipe_resource *pt = st_get_texobj_resource(texObj);
const uint baseLevel = texObj->BaseLevel;
+ enum pipe_format format;
uint lastLevel, first_layer, last_layer;
uint dstLevel;
last_layer = util_max_layer(pt, baseLevel);
}
- /* Try to generate the mipmap by rendering/texturing. If that fails,
- * use the software fallback.
+ if (stObj->surface_based)
+ format = stObj->surface_format;
+ else
+ format = pt->format;
+
+ /* First see if the driver supports hardware mipmap generation,
+ * if not then generate the mipmap by rendering/texturing.
+ * If that fails, use the software fallback.
*/
- if (!util_gen_mipmap(st->pipe, pt, pt->format, baseLevel, lastLevel,
- first_layer, last_layer, PIPE_TEX_FILTER_LINEAR)) {
- _mesa_generate_mipmap(ctx, target, texObj);
+ if (!st->pipe->screen->get_param(st->pipe->screen,
+ PIPE_CAP_GENERATE_MIPMAP) ||
+ !st->pipe->generate_mipmap(st->pipe, pt, format, baseLevel,
+ lastLevel, first_layer, last_layer)) {
+
+ if (!util_gen_mipmap(st->pipe, pt, format, baseLevel, lastLevel,
+ first_layer, last_layer, PIPE_TEX_FILTER_LINEAR)) {
+ _mesa_generate_mipmap(ctx, target, texObj);
+ }
}
/* Fill in the Mesa gl_texture_image fields */
this->has_index2 = false;
this->double_reg2 = false;
this->array_id = 0;
+ this->is_double_vertex_input = false;
}
st_src_reg(gl_register_file file, int index, int type)
this->has_index2 = false;
this->double_reg2 = false;
this->array_id = 0;
+ this->is_double_vertex_input = false;
}
st_src_reg(gl_register_file file, int index, int type, int index2D)
this->has_index2 = false;
this->double_reg2 = false;
this->array_id = 0;
+ this->is_double_vertex_input = false;
}
st_src_reg()
this->has_index2 = false;
this->double_reg2 = false;
this->array_id = 0;
+ this->is_double_vertex_input = false;
}
explicit st_src_reg(st_dst_reg reg);
*/
bool double_reg2;
unsigned array_id;
+ bool is_double_vertex_input;
};
class st_dst_reg {
this->has_index2 = reg.has_index2;
this->double_reg2 = false;
this->array_id = reg.array_id;
+ this->is_double_vertex_input = false;
}
st_dst_reg::st_dst_reg(st_src_reg reg)
unsigned mesa_index;
unsigned array_id;
unsigned array_size;
+ unsigned array_type;
};
+static unsigned
+find_array_type(struct array_decl *arrays, unsigned count, unsigned array_id)
+{
+ unsigned i;
+
+ for (i = 0; i < count; i++) {
+ struct array_decl *decl = &arrays[i];
+
+ if (array_id == decl->array_id) {
+ return decl->array_type;
+ }
+ }
+ return GLSL_TYPE_ERROR;
+}
+
struct rename_reg_pair {
int old_reg;
int new_reg;
{
glsl_to_tgsi_instruction *inst = new(mem_ctx) glsl_to_tgsi_instruction();
int num_reladdr = 0, i, j;
+ bool dst_is_double[2];
op = get_opcode(ir, op, dst, src0, src1);
* GLSL [0].z -> TGSI [1].xy
* GLSL [0].w -> TGSI [1].zw
*/
- if (inst->dst[0].type == GLSL_TYPE_DOUBLE || inst->dst[1].type == GLSL_TYPE_DOUBLE ||
+ for (j = 0; j < 2; j++) {
+ dst_is_double[j] = false;
+ if (inst->dst[j].type == GLSL_TYPE_DOUBLE)
+ dst_is_double[j] = true;
+ else if (inst->dst[j].file == PROGRAM_OUTPUT && inst->dst[j].type == GLSL_TYPE_ARRAY) {
+ unsigned type = find_array_type(this->output_arrays, this->num_output_arrays, inst->dst[j].array_id);
+ if (type == GLSL_TYPE_DOUBLE)
+ dst_is_double[j] = true;
+ }
+ }
+
+ if (dst_is_double[0] || dst_is_double[1] ||
inst->src[0].type == GLSL_TYPE_DOUBLE) {
glsl_to_tgsi_instruction *dinst = NULL;
int initial_src_swz[4], initial_src_idx[4];
/* modify the destination if we are splitting */
for (j = 0; j < 2; j++) {
- if (dinst->dst[j].type == GLSL_TYPE_DOUBLE) {
+ if (dst_is_double[j]) {
dinst->dst[j].writemask = (i & 1) ? WRITEMASK_ZW : WRITEMASK_XY;
dinst->dst[j].index = initial_dst_idx[j];
if (i > 1)
- F2D is a float src0, DLDEXP is integer src1 */
if (op == TGSI_OPCODE_F2D ||
op == TGSI_OPCODE_DLDEXP ||
- (op == TGSI_OPCODE_UCMP && dinst->dst[0].type == GLSL_TYPE_DOUBLE)) {
+ (op == TGSI_OPCODE_UCMP && dst_is_double[0])) {
dinst->src[j].swizzle = MAKE_SWIZZLE4(swz, swz, swz, swz);
}
}
}
static int
-type_size(const struct glsl_type *type)
+attrib_type_size(const struct glsl_type *type, bool is_vs_input)
{
unsigned int i;
int size;
break;
case GLSL_TYPE_DOUBLE:
if (type->is_matrix()) {
- if (type->vector_elements <= 2)
+ if (type->vector_elements <= 2 || is_vs_input)
return type->matrix_columns;
else
return type->matrix_columns * 2;
/* For doubles if we have a double or dvec2 they fit in one
* vec4, else they need 2 vec4s.
*/
- if (type->vector_elements <= 2)
+ if (type->vector_elements <= 2 || is_vs_input)
return 1;
else
return 2;
break;
case GLSL_TYPE_ARRAY:
assert(type->length > 0);
- return type_size(type->fields.array) * type->length;
+ return attrib_type_size(type->fields.array, is_vs_input) * type->length;
case GLSL_TYPE_STRUCT:
size = 0;
for (i = 0; i < type->length; i++) {
- size += type_size(type->fields.structure[i].type);
+ size += attrib_type_size(type->fields.structure[i].type, is_vs_input);
}
return size;
case GLSL_TYPE_SAMPLER:
return 0;
}
+static int
+type_size(const struct glsl_type *type)
+{
+ return attrib_type_size(type, false);
+}
/**
* If the given GLSL type is an array or matrix or a structure containing
if (reg->reladdr2) emit_arl(ir, address_reg2, *reg->reladdr2);
if (*num_reladdr != 1) {
- st_src_reg temp = get_temp(glsl_type::vec4_type);
+ st_src_reg temp = get_temp(reg->type == GLSL_TYPE_DOUBLE ? glsl_type::dvec4_type : glsl_type::vec4_type);
emit_asm(ir, TGSI_OPCODE_MOV, st_dst_reg(temp), *reg);
*reg = temp;
}
break;
- case ir_unop_any: {
- assert(ir->operands[0]->type->is_vector());
-
- if (native_integers) {
- int dst_swizzle = 0, op0_swizzle, i;
- st_src_reg accum = op[0];
-
- op0_swizzle = op[0].swizzle;
- accum.swizzle = MAKE_SWIZZLE4(GET_SWZ(op0_swizzle, 0),
- GET_SWZ(op0_swizzle, 0),
- GET_SWZ(op0_swizzle, 0),
- GET_SWZ(op0_swizzle, 0));
- for (i = 0; i < 4; i++) {
- if (result_dst.writemask & (1 << i)) {
- dst_swizzle = MAKE_SWIZZLE4(i, i, i, i);
- break;
- }
- }
- assert(i != 4);
- assert(ir->operands[0]->type->is_boolean());
-
- /* OR all the components together, since they should be either 0 or ~0
- */
- switch (ir->operands[0]->type->vector_elements) {
- case 4:
- op[0].swizzle = MAKE_SWIZZLE4(GET_SWZ(op0_swizzle, 3),
- GET_SWZ(op0_swizzle, 3),
- GET_SWZ(op0_swizzle, 3),
- GET_SWZ(op0_swizzle, 3));
- emit_asm(ir, TGSI_OPCODE_OR, result_dst, accum, op[0]);
- accum = st_src_reg(result_dst);
- accum.swizzle = dst_swizzle;
- /* fallthrough */
- case 3:
- op[0].swizzle = MAKE_SWIZZLE4(GET_SWZ(op0_swizzle, 2),
- GET_SWZ(op0_swizzle, 2),
- GET_SWZ(op0_swizzle, 2),
- GET_SWZ(op0_swizzle, 2));
- emit_asm(ir, TGSI_OPCODE_OR, result_dst, accum, op[0]);
- accum = st_src_reg(result_dst);
- accum.swizzle = dst_swizzle;
- /* fallthrough */
- case 2:
- op[0].swizzle = MAKE_SWIZZLE4(GET_SWZ(op0_swizzle, 1),
- GET_SWZ(op0_swizzle, 1),
- GET_SWZ(op0_swizzle, 1),
- GET_SWZ(op0_swizzle, 1));
- emit_asm(ir, TGSI_OPCODE_OR, result_dst, accum, op[0]);
- break;
- default:
- assert(!"Unexpected vector size");
- break;
- }
- } else {
- /* After the dot-product, the value will be an integer on the
- * range [0,4]. Zero stays zero, and positive values become 1.0.
- */
- glsl_to_tgsi_instruction *const dp =
- emit_dp(ir, result_dst, op[0], op[0],
- ir->operands[0]->type->vector_elements);
- if (this->prog->Target == GL_FRAGMENT_PROGRAM_ARB &&
- result_dst.type == GLSL_TYPE_FLOAT) {
- /* The clamping to [0,1] can be done for free in the fragment
- * shader with a saturate.
- */
- dp->saturate = true;
- } else if (result_dst.type == GLSL_TYPE_FLOAT) {
- /* Negating the result of the dot-product gives values on the range
- * [-4, 0]. Zero stays zero, and negative values become 1.0. This
- * is achieved using SLT.
- */
- st_src_reg slt_src = result_src;
- slt_src.negate = ~slt_src.negate;
- emit_asm(ir, TGSI_OPCODE_SLT, result_dst, slt_src, st_src_reg_for_float(0.0));
- }
- else {
- /* Use SNE 0 if integers are being used as boolean values. */
- emit_asm(ir, TGSI_OPCODE_SNE, result_dst, result_src, st_src_reg_for_int(0));
- }
- }
- break;
- }
-
case ir_binop_logic_xor:
if (native_integers)
emit_asm(ir, TGSI_OPCODE_XOR, result_dst, op[0], op[1]);
}
break;
+ case ir_unop_pack_half_2x16:
+ emit_asm(ir, TGSI_OPCODE_PK2H, result_dst, op[0]);
+ break;
+ case ir_unop_unpack_half_2x16:
+ emit_asm(ir, TGSI_OPCODE_UP2H, result_dst, op[0]);
+ break;
+
case ir_unop_pack_snorm_2x16:
case ir_unop_pack_unorm_2x16:
- case ir_unop_pack_half_2x16:
case ir_unop_pack_snorm_4x8:
case ir_unop_pack_unorm_4x8:
case ir_unop_unpack_snorm_2x16:
case ir_unop_unpack_unorm_2x16:
- case ir_unop_unpack_half_2x16:
case ir_unop_unpack_half_2x16_split_x:
case ir_unop_unpack_half_2x16_split_y:
case ir_unop_unpack_snorm_4x8:
case ir_unop_unpack_unorm_4x8:
case ir_binop_pack_half_2x16_split:
- case ir_binop_bfm:
- case ir_triop_bfi:
case ir_quadop_vector:
case ir_binop_vector_extract:
case ir_triop_vector_insert:
decl->mesa_index = var->data.location;
decl->array_id = num_input_arrays + 1;
- if (is_2d)
+ if (is_2d) {
decl->array_size = type_size(var->type->fields.array);
- else
+ decl->array_type = var->type->fields.array->without_array()->base_type;
+ } else {
decl->array_size = type_size(var->type);
+ decl->array_type = var->type->without_array()->base_type;
+ }
num_input_arrays++;
entry = new(mem_ctx) variable_storage(var,
decl->mesa_index = var->data.location;
decl->array_id = num_output_arrays + 1;
- if (is_2d)
+ if (is_2d) {
decl->array_size = type_size(var->type->fields.array);
- else
+ decl->array_type = var->type->fields.array->without_array()->base_type;
+ } else {
decl->array_size = type_size(var->type);
+ decl->array_type = var->type->without_array()->base_type;
+ }
num_output_arrays++;
entry = new(mem_ctx) variable_storage(var,
this->result = st_src_reg(entry->file, entry->index, var->type);
this->result.array_id = entry->array_id;
+ if (this->shader->Stage == MESA_SHADER_VERTEX && var->data.mode == ir_var_shader_in && var->type->is_double())
+ this->result.is_double_vertex_input = true;
if (!native_integers)
this->result.type = GLSL_TYPE_FLOAT;
}
static void
shrink_array_declarations(struct array_decl *arrays, unsigned count,
GLbitfield64 usage_mask,
+ GLbitfield64 double_usage_mask,
GLbitfield patch_usage_mask)
{
unsigned i, j;
else {
if (usage_mask & BITFIELD64_BIT(decl->mesa_index+j))
break;
+ if (double_usage_mask & BITFIELD64_BIT(decl->mesa_index+j-1))
+ break;
}
decl->mesa_index++;
else {
if (usage_mask & BITFIELD64_BIT(decl->mesa_index+j))
break;
+ if (double_usage_mask & BITFIELD64_BIT(decl->mesa_index+j-1))
+ break;
}
decl->array_size--;
element_size = 1;
if (index) {
+
+ if (this->prog->Target == GL_VERTEX_PROGRAM_ARB &&
+ src.file == PROGRAM_INPUT)
+ element_size = attrib_type_size(ir->type, true);
if (is_2D) {
src.index2D = index->value.i[0];
src.has_index2 = true;
if (type->is_matrix()) {
const struct glsl_type *vec_type;
- vec_type = glsl_type::get_instance(GLSL_TYPE_FLOAT,
+ vec_type = glsl_type::get_instance(type->is_double() ? GLSL_TYPE_DOUBLE : GLSL_TYPE_FLOAT,
type->vector_elements, 1);
for (int i = 0; i < type->matrix_columns; i++) {
}
l->index++;
r->index++;
+ if (type->is_dual_slot_double()) {
+ l->index++;
+ if (r->is_double_vertex_input == false)
+ r->index++;
+ }
}
void
*/
if (ir->write_mask == 0) {
assert(!ir->lhs->type->is_scalar() && !ir->lhs->type->is_vector());
- l.writemask = WRITEMASK_XYZW;
+
+ if (ir->lhs->type->is_array() || ir->lhs->type->without_array()->is_matrix()) {
+ if (ir->lhs->type->without_array()->is_double()) {
+ switch (ir->lhs->type->without_array()->vector_elements) {
+ case 1:
+ l.writemask = WRITEMASK_X;
+ break;
+ case 2:
+ l.writemask = WRITEMASK_XY;
+ break;
+ case 3:
+ l.writemask = WRITEMASK_XYZ;
+ break;
+ case 4:
+ l.writemask = WRITEMASK_XYZW;
+ break;
+ }
+ } else
+ l.writemask = WRITEMASK_XYZW;
+ }
} else if (ir->lhs->type->is_scalar() &&
!ir->lhs->type->is_double() &&
ir->lhs->variable_referenced()->data.mode == ir_var_shader_out) {
st_dst_reg mat_column = st_dst_reg(mat);
for (i = 0; i < ir->type->matrix_columns; i++) {
- assert(ir->type->base_type == GLSL_TYPE_FLOAT);
- values = (gl_constant_value *) &ir->value.f[i * ir->type->vector_elements];
-
- src = st_src_reg(file, -1, ir->type->base_type);
- src.index = add_constant(file,
- values,
- ir->type->vector_elements,
- GL_FLOAT,
- &src.swizzle);
- emit_asm(ir, TGSI_OPCODE_MOV, mat_column, src);
+ switch (ir->type->base_type) {
+ case GLSL_TYPE_FLOAT:
+ values = (gl_constant_value *) &ir->value.f[i * ir->type->vector_elements];
+ src = st_src_reg(file, -1, ir->type->base_type);
+ src.index = add_constant(file,
+ values,
+ ir->type->vector_elements,
+ GL_FLOAT,
+ &src.swizzle);
+ emit_asm(ir, TGSI_OPCODE_MOV, mat_column, src);
+ break;
+ case GLSL_TYPE_DOUBLE:
+ values = (gl_constant_value *) &ir->value.d[i * ir->type->vector_elements];
+ src = st_src_reg(file, -1, ir->type->base_type);
+ src.index = add_constant(file,
+ values,
+ ir->type->vector_elements,
+ GL_DOUBLE,
+ &src.swizzle);
+ if (ir->type->vector_elements >= 2) {
+ mat_column.writemask = WRITEMASK_XY;
+ src.swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_X, SWIZZLE_Y);
+ emit_asm(ir, TGSI_OPCODE_MOV, mat_column, src);
+ } else {
+ mat_column.writemask = WRITEMASK_X;
+ src.swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X);
+ emit_asm(ir, TGSI_OPCODE_MOV, mat_column, src);
+ }
+ src.index++;
+ if (ir->type->vector_elements > 2) {
+ if (ir->type->vector_elements == 4) {
+ mat_column.writemask = WRITEMASK_ZW;
+ src.swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_X, SWIZZLE_Y);
+ emit_asm(ir, TGSI_OPCODE_MOV, mat_column, src);
+ } else {
+ mat_column.writemask = WRITEMASK_Z;
+ src.swizzle = MAKE_SWIZZLE4(SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y);
+ emit_asm(ir, TGSI_OPCODE_MOV, mat_column, src);
+ mat_column.writemask = WRITEMASK_XYZW;
+ src.swizzle = SWIZZLE_XYZW;
+ }
+ mat_column.index++;
+ }
+ break;
+ default:
+ unreachable("Illegal matrix constant type.\n");
+ break;
+ }
mat_column.index++;
}
-
this->result = mat;
return;
}
TGSI_SEMANTIC_INSTANCEID,
TGSI_SEMANTIC_VERTEXID_NOBASE,
TGSI_SEMANTIC_BASEVERTEX,
+ TGSI_SEMANTIC_BASEINSTANCE,
+ TGSI_SEMANTIC_DRAWID,
/* Geometry shader
*/
/* Fragment shader
*/
+ TGSI_SEMANTIC_POSITION,
TGSI_SEMANTIC_FACE,
TGSI_SEMANTIC_SAMPLEID,
TGSI_SEMANTIC_SAMPLEPOS,
if (!reg->array_id) {
assert(t->inputMapping[index] < ARRAY_SIZE(t->inputs));
assert(t->inputs[t->inputMapping[index]].File != TGSI_FILE_NULL);
- return t->inputs[t->inputMapping[index]];
+ return t->inputs[t->inputMapping[index] + double_reg2];
}
else {
struct array_decl *decl = &t->input_arrays[reg->array_id-1];
assert(slot != -1 && t->inputs[slot].File == TGSI_FILE_INPUT);
assert(t->inputs[slot].ArrayID == reg->array_id);
- return ureg_src_array_offset(t->inputs[slot], index - mesa_index);
+ return ureg_src_array_offset(t->inputs[slot], index + double_reg2 - mesa_index);
}
case PROGRAM_ADDRESS:
* a FBO is bound (STATE_FB_WPOS_Y_TRANSFORM).
*/
static void
-emit_wpos_adjustment( struct st_translate *t,
- int wpos_transform_const,
- boolean invert,
- GLfloat adjX, GLfloat adjY[2])
+emit_wpos_adjustment(struct gl_context *ctx,
+ struct st_translate *t,
+ int wpos_transform_const,
+ boolean invert,
+ GLfloat adjX, GLfloat adjY[2])
{
struct ureg_program *ureg = t->ureg;
*/
struct ureg_src wpostrans = ureg_DECL_constant(ureg, wpos_transform_const);
struct ureg_dst wpos_temp = ureg_DECL_temporary( ureg );
- struct ureg_src wpos_input = t->inputs[t->inputMapping[VARYING_SLOT_POS]];
+ struct ureg_src *wpos =
+ ctx->Const.GLSLFragCoordIsSysVal ?
+ &t->systemValues[SYSTEM_VALUE_FRAG_COORD] :
+ &t->inputs[t->inputMapping[VARYING_SLOT_POS]];
+ struct ureg_src wpos_input = *wpos;
/* First, apply the coordinate shift: */
if (adjX || adjY[0] || adjY[1]) {
/* Use wpos_temp as position input from here on:
*/
- t->inputs[t->inputMapping[VARYING_SLOT_POS]] = ureg_src(wpos_temp);
+ *wpos = ureg_src(wpos_temp);
}
/* we invert after adjustment so that we avoid the MOV to temporary,
* and reuse the adjustment ADD instead */
- emit_wpos_adjustment(t, wpos_transform_const, invert, adjX, adjY);
+ emit_wpos_adjustment(st->ctx, t, wpos_transform_const, invert, adjX, adjY);
}
/**
*/
{
GLbitfield sysInputs = proginfo->SystemValuesRead;
- unsigned numSys = 0;
+
for (i = 0; sysInputs; i++) {
if (sysInputs & (1 << i)) {
unsigned semName = _mesa_sysval_to_semantic[i];
- t->systemValues[i] = ureg_DECL_system_value(ureg, numSys, semName, 0);
+
+ t->systemValues[i] = ureg_DECL_system_value(ureg, semName, 0);
+
if (semName == TGSI_SEMANTIC_INSTANCEID ||
semName == TGSI_SEMANTIC_VERTEXID) {
/* From Gallium perspective, these system values are always
t->systemValues[i] = ureg_scalar(ureg_src(temp), 0);
}
}
- numSys++;
+
+ if (procType == TGSI_PROCESSOR_FRAGMENT &&
+ semName == TGSI_SEMANTIC_POSITION)
+ emit_wpos(st_context(ctx), t, proginfo, ureg,
+ program->wpos_transform_const);
+
sysInputs &= ~(1 << i);
}
}
do_set_program_inouts(shader->ir, prog, shader->Stage);
shrink_array_declarations(v->input_arrays, v->num_input_arrays,
- prog->InputsRead, prog->PatchInputsRead);
+ prog->InputsRead, prog->DoubleInputsRead, prog->PatchInputsRead);
shrink_array_declarations(v->output_arrays, v->num_output_arrays,
- prog->OutputsWritten, prog->PatchOutputsWritten);
+ prog->OutputsWritten, 0ULL, prog->PatchOutputsWritten);
count_resources(v, prog);
/* This must be done before the uniform storage is associated. */
if (shader->Type == GL_FRAGMENT_SHADER &&
- prog->InputsRead & VARYING_BIT_POS){
+ (prog->InputsRead & VARYING_BIT_POS ||
+ prog->SystemValuesRead & (1 << SYSTEM_VALUE_FRAG_COORD))) {
static const gl_state_index wposTransformState[STATE_LENGTH] = {
STATE_INTERNAL, STATE_FB_WPOS_Y_TRANSFORM
};
_mesa_reference_program(ctx, &shader->Program, prog);
+ /* Avoid reallocation of the program parameter list, because the uniform
+ * storage is only associated with the original parameter list.
+ * This should be enough for Bitmap and DrawPixels constants.
+ */
+ _mesa_reserve_parameter_storage(prog->Parameters, 8);
+
/* This has to be done last. Any operation the can cause
* prog->ParameterValues to get reallocated (e.g., anything that adds a
* program constant) has to happen before creating this linkage.
LOWER_PACK_SNORM_4x8 |
LOWER_UNPACK_SNORM_4x8 |
LOWER_UNPACK_UNORM_4x8 |
- LOWER_PACK_UNORM_4x8 |
- LOWER_PACK_HALF_2x16 |
- LOWER_UNPACK_HALF_2x16;
+ LOWER_PACK_UNORM_4x8;
if (ctx->Extensions.ARB_gpu_shader5)
lower_inst |= LOWER_PACK_USE_BFI |
LOWER_PACK_USE_BFE;
+ if (!ctx->st->has_half_float_packing)
+ lower_inst |= LOWER_PACK_HALF_2x16 |
+ LOWER_UNPACK_HALF_2x16;
lower_packing_builtins(ir, lower_inst);
}
#include "st_texture.h"
#include "st_context.h"
+#include "st_debug.h"
#include "st_extensions.h"
#include "st_format.h"
#include "st_cb_fbo.h"
st_destroy_context(st);
}
-static void
-st_debug_message(void *data,
- unsigned *id,
- enum pipe_debug_type ptype,
- const char *fmt,
- va_list args)
-{
- struct st_context *st = data;
- enum mesa_debug_source source;
- enum mesa_debug_type type;
- enum mesa_debug_severity severity;
-
- switch (ptype) {
- case PIPE_DEBUG_TYPE_OUT_OF_MEMORY:
- source = MESA_DEBUG_SOURCE_API;
- type = MESA_DEBUG_TYPE_ERROR;
- severity = MESA_DEBUG_SEVERITY_MEDIUM;
- break;
- case PIPE_DEBUG_TYPE_ERROR:
- source = MESA_DEBUG_SOURCE_API;
- type = MESA_DEBUG_TYPE_ERROR;
- severity = MESA_DEBUG_SEVERITY_MEDIUM;
- break;
- case PIPE_DEBUG_TYPE_SHADER_INFO:
- source = MESA_DEBUG_SOURCE_SHADER_COMPILER;
- type = MESA_DEBUG_TYPE_OTHER;
- severity = MESA_DEBUG_SEVERITY_NOTIFICATION;
- break;
- case PIPE_DEBUG_TYPE_PERF_INFO:
- source = MESA_DEBUG_SOURCE_API;
- type = MESA_DEBUG_TYPE_PERFORMANCE;
- severity = MESA_DEBUG_SEVERITY_NOTIFICATION;
- break;
- case PIPE_DEBUG_TYPE_INFO:
- source = MESA_DEBUG_SOURCE_API;
- type = MESA_DEBUG_TYPE_OTHER;
- severity = MESA_DEBUG_SEVERITY_NOTIFICATION;
- break;
- case PIPE_DEBUG_TYPE_FALLBACK:
- source = MESA_DEBUG_SOURCE_API;
- type = MESA_DEBUG_TYPE_PERFORMANCE;
- severity = MESA_DEBUG_SEVERITY_NOTIFICATION;
- break;
- case PIPE_DEBUG_TYPE_CONFORMANCE:
- source = MESA_DEBUG_SOURCE_API;
- type = MESA_DEBUG_TYPE_OTHER;
- severity = MESA_DEBUG_SEVERITY_NOTIFICATION;
- break;
- }
- _mesa_gl_vdebug(st->ctx, id, source, type, severity, fmt, args);
-}
-
static struct st_context_iface *
st_api_create_context(struct st_api *stapi, struct st_manager *smapi,
const struct st_context_attribs *attribs,
return NULL;
}
- if (attribs->flags & ST_CONTEXT_FLAG_DEBUG){
+ if (attribs->flags & ST_CONTEXT_FLAG_DEBUG) {
if (!_mesa_set_debug_state_int(st->ctx, GL_DEBUG_OUTPUT, GL_TRUE)) {
*error = ST_CONTEXT_ERROR_NO_MEMORY;
return NULL;
}
+
st->ctx->Const.ContextFlags |= GL_CONTEXT_FLAG_DEBUG_BIT;
- if (pipe->set_debug_callback) {
- struct pipe_debug_callback cb = { st_debug_message, st };
- pipe->set_debug_callback(pipe, &cb);
- }
+ st_enable_debug_output(st, TRUE);
}
if (attribs->flags & ST_CONTEXT_FLAG_FORWARD_COMPATIBLE)
}
-/**
- * Negate the value of DDY to match GL semantics where (0,0) is the
- * lower-left corner of the window.
- * Note that the GL_ARB_fragment_coord_conventions extension will
- * effect this someday.
- */
-static void emit_ddy( struct st_translate *t,
- struct ureg_dst dst,
- const struct prog_src_register *SrcReg )
-{
- struct ureg_program *ureg = t->ureg;
- struct ureg_src src = translate_src( t, SrcReg );
- src = ureg_negate( src );
- ureg_DDY( ureg, dst, src );
-}
-
-
-
static unsigned
translate_opcode( unsigned op )
{
*/
ureg_MOV( ureg, dst[0], ureg_imm1f(ureg, 0.5) );
break;
-
- case OPCODE_DDY:
- emit_ddy( t, dst[0], &inst->SrcReg[0] );
- break;
case OPCODE_RSQ:
ureg_RSQ( ureg, dst[0], ureg_abs(src[0]) );
* a FBO is bound (STATE_FB_WPOS_Y_TRANSFORM).
*/
static void
-emit_wpos_adjustment( struct st_translate *t,
- const struct gl_program *program,
- boolean invert,
- GLfloat adjX, GLfloat adjY[2])
+emit_wpos_adjustment(struct gl_context *ctx,
+ struct st_translate *t,
+ const struct gl_program *program,
+ boolean invert,
+ GLfloat adjX, GLfloat adjY[2])
{
struct ureg_program *ureg = t->ureg;
struct ureg_src wpostrans = ureg_DECL_constant( ureg, wposTransConst );
struct ureg_dst wpos_temp = ureg_DECL_temporary( ureg );
- struct ureg_src wpos_input = t->inputs[t->inputMapping[VARYING_SLOT_POS]];
+ struct ureg_src *wpos =
+ ctx->Const.GLSLFragCoordIsSysVal ?
+ &t->systemValues[SYSTEM_VALUE_FRAG_COORD] :
+ &t->inputs[t->inputMapping[VARYING_SLOT_POS]];
+ struct ureg_src wpos_input = *wpos;
/* First, apply the coordinate shift: */
if (adjX || adjY[0] || adjY[1]) {
/* Use wpos_temp as position input from here on:
*/
- t->inputs[t->inputMapping[VARYING_SLOT_POS]] = ureg_src(wpos_temp);
+ *wpos = ureg_src(wpos_temp);
}
/* we invert after adjustment so that we avoid the MOV to temporary,
* and reuse the adjustment ADD instead */
- emit_wpos_adjustment(t, program, invert, adjX, adjY);
-}
-
-
-/**
- * OpenGL's fragment gl_FrontFace input is 1 for front-facing, 0 for back.
- * TGSI uses +1 for front, -1 for back.
- * This function converts the TGSI value to the GL value. Simply clamping/
- * saturating the value to [0,1] does the job.
- */
-static void
-emit_face_var( struct st_translate *t,
- const struct gl_program *program )
-{
- struct ureg_program *ureg = t->ureg;
- struct ureg_dst face_temp = ureg_DECL_temporary( ureg );
- struct ureg_src face_input = t->inputs[t->inputMapping[VARYING_SLOT_FACE]];
-
- /* MOV_SAT face_temp, input[face]
- */
- face_temp = ureg_saturate( face_temp );
- ureg_MOV( ureg, face_temp, face_input );
-
- /* Use face_temp as face input from here on:
- */
- t->inputs[t->inputMapping[VARYING_SLOT_FACE]] = ureg_src(face_temp);
+ emit_wpos_adjustment(st->ctx, t, program, invert, adjX, adjY);
}
emit_wpos(st_context(ctx), t, program, ureg);
}
- if (program->InputsRead & VARYING_BIT_FACE) {
- emit_face_var( t, program );
- }
-
/*
* Declare output attributes.
*/
*/
{
GLbitfield sysInputs = program->SystemValuesRead;
- unsigned numSys = 0;
+
for (i = 0; sysInputs; i++) {
if (sysInputs & (1 << i)) {
unsigned semName = _mesa_sysval_to_semantic[i];
- t->systemValues[i] = ureg_DECL_system_value(ureg, numSys, semName, 0);
+
+ t->systemValues[i] = ureg_DECL_system_value(ureg, semName, 0);
+
if (semName == TGSI_SEMANTIC_INSTANCEID ||
semName == TGSI_SEMANTIC_VERTEXID) {
/* From Gallium perspective, these system values are always
t->systemValues[i] = ureg_scalar(ureg_src(temp), 0);
}
}
- numSys++;
+
+ if (procType == TGSI_PROCESSOR_FRAGMENT &&
+ semName == TGSI_SEMANTIC_POSITION)
+ emit_wpos(st_context(ctx), t, program, ureg);
+
sysInputs &= ~(1 << i);
}
}
{
if (fpv->driver_shader)
cso_delete_fragment_shader(st->cso_context, fpv->driver_shader);
- if (fpv->parameters)
- _mesa_free_parameter_list(fpv->parameters);
free(fpv);
}
memset(inputSlotToAttr, ~0, sizeof(inputSlotToAttr));
- if (!stfp->glsl_to_tgsi)
+ if (!stfp->glsl_to_tgsi) {
_mesa_remove_output_reads(&stfp->Base.Base, PROGRAM_OUTPUT);
+ if (st->ctx->Const.GLSLFragCoordIsSysVal)
+ _mesa_program_fragment_position_to_sysval(&stfp->Base.Base);
+ }
/*
* Convert Mesa program inputs to TGSI input register semantics.
if (tgsi.tokens != stfp->tgsi.tokens)
tgsi_free_tokens(tgsi.tokens);
tgsi.tokens = tokens;
- variant->parameters =
- _mesa_clone_parameter_list(stfp->Base.Base.Parameters);
} else
fprintf(stderr, "mesa: cannot create a shader for glBitmap\n");
}
if (key->drawpixels) {
const struct tgsi_token *tokens;
unsigned scale_const = 0, bias_const = 0, texcoord_const = 0;
+ struct gl_program_parameter_list *params = stfp->Base.Base.Parameters;
/* Find the first unused slot. */
variant->drawpix_sampler = ffs(~stfp->Base.Base.SamplersUsed) - 1;
variant->pixelmap_sampler = ffs(~samplers_used) - 1;
}
- variant->parameters =
- _mesa_clone_parameter_list(stfp->Base.Base.Parameters);
-
if (key->scaleAndBias) {
static const gl_state_index scale_state[STATE_LENGTH] =
{ STATE_INTERNAL, STATE_PT_SCALE };
static const gl_state_index bias_state[STATE_LENGTH] =
{ STATE_INTERNAL, STATE_PT_BIAS };
- scale_const = _mesa_add_state_reference(variant->parameters,
- scale_state);
- bias_const = _mesa_add_state_reference(variant->parameters,
- bias_state);
+ scale_const = _mesa_add_state_reference(params, scale_state);
+ bias_const = _mesa_add_state_reference(params, bias_state);
}
{
static const gl_state_index state[STATE_LENGTH] =
{ STATE_INTERNAL, STATE_CURRENT_ATTRIB, VERT_ATTRIB_TEX0 };
- texcoord_const = _mesa_add_state_reference(variant->parameters,
- state);
+ texcoord_const = _mesa_add_state_reference(params, state);
}
tokens = st_get_drawpix_shader(tgsi.tokens,
void *driver_shader;
/** For glBitmap variants */
- struct gl_program_parameter_list *parameters;
uint bitmap_sampler;
/** For glDrawPixels variants */
extern struct st_tcp_variant *
st_get_tcp_variant(struct st_context *st,
- struct st_tessctrl_program *stgp,
+ struct st_tessctrl_program *sttcp,
const struct st_tcp_variant_key *key);
extern struct st_tep_variant *
st_get_tep_variant(struct st_context *st,
- struct st_tesseval_program *stgp,
+ struct st_tesseval_program *sttep,
const struct st_tep_variant_key *key);
extern void
extern void
st_release_tcp_variants(struct st_context *st,
- struct st_tessctrl_program *stgp);
+ struct st_tessctrl_program *sttcp);
extern void
st_release_tep_variants(struct st_context *st,
- struct st_tesseval_program *stgp);
+ struct st_tesseval_program *sttep);
extern void
st_destroy_program_variants(struct st_context *st);
#include "swrast/s_atifragshader.h"
#include "swrast/s_context.h"
+#define ATI_FS_INPUT_PRIMARY 0
+#define ATI_FS_INPUT_SECONDARY 1
/**
* State for executing ATI fragment shader.
GLint basevertex;
GLuint num_instances;
GLuint base_instance;
+ GLuint draw_id;
GLsizeiptr indirect_offset;
};
GLboolean _vbo_CreateContext( struct gl_context *ctx );
void _vbo_DestroyContext( struct gl_context *ctx );
-void _vbo_InvalidateState( struct gl_context *ctx, GLuint new_state );
+void _vbo_InvalidateState( struct gl_context *ctx, GLbitfield new_state );
void
struct gl_buffer_object *indirect);
+typedef void (*vbo_indirect_draw_func)(
+ struct gl_context *ctx,
+ GLuint mode,
+ struct gl_buffer_object *indirect_data,
+ GLsizeiptr indirect_offset,
+ unsigned draw_count,
+ unsigned stride,
+ struct gl_buffer_object *indirect_params,
+ GLsizeiptr indirect_params_offset,
+ const struct _mesa_index_buffer *ib);
+
+
/* Utility function to cope with various constraints on tnl modules or
void vbo_set_draw_func(struct gl_context *ctx, vbo_draw_func func);
+void vbo_set_indirect_draw_func(struct gl_context *ctx,
+ vbo_indirect_draw_func func);
+
void vbo_check_buffers_are_unmapped(struct gl_context *ctx);
void vbo_bind_arrays(struct gl_context *ctx);
}
}
+static void
+vbo_draw_indirect_prims(struct gl_context *ctx,
+ GLuint mode,
+ struct gl_buffer_object *indirect_data,
+ GLsizeiptr indirect_offset,
+ unsigned draw_count,
+ unsigned stride,
+ struct gl_buffer_object *indirect_params,
+ GLsizeiptr indirect_params_offset,
+ const struct _mesa_index_buffer *ib)
+{
+ struct vbo_context *vbo = vbo_context(ctx);
+ struct _mesa_prim *prim;
+ GLsizei i;
+
+ prim = calloc(draw_count, sizeof(*prim));
+ if (prim == NULL) {
+ _mesa_error(ctx, GL_OUT_OF_MEMORY, "gl%sDraw%sIndirect%s",
+ (draw_count > 1) ? "Multi" : "",
+ ib ? "Elements" : "Arrays",
+ indirect_params ? "CountARB" : "");
+ return;
+ }
+
+ prim[0].begin = 1;
+ prim[draw_count - 1].end = 1;
+ for (i = 0; i < draw_count; ++i, indirect_offset += stride) {
+ prim[i].mode = mode;
+ prim[i].indexed = !!ib;
+ prim[i].indirect_offset = indirect_offset;
+ prim[i].is_indirect = 1;
+ prim[i].draw_id = i;
+ }
+
+ vbo->draw_prims(ctx, prim, draw_count,
+ ib, GL_TRUE, 0, ~0,
+ NULL, 0,
+ ctx->DrawIndirectBuffer);
+
+ free(prim);
+}
+
GLboolean _vbo_CreateContext( struct gl_context *ctx )
{
init_legacy_currval( ctx );
init_generic_currval( ctx );
init_mat_currval( ctx );
+ vbo_set_indirect_draw_func(ctx, vbo_draw_indirect_prims);
/* Build mappings from VERT_ATTRIB -> VBO_ATTRIB depending on type
* of vertex program active.
}
-void _vbo_InvalidateState( struct gl_context *ctx, GLuint new_state )
+void _vbo_InvalidateState( struct gl_context *ctx, GLbitfield new_state )
{
vbo_exec_invalidate_state(ctx, new_state);
}
vbo->draw_prims = func;
}
+
+void vbo_set_indirect_draw_func(struct gl_context *ctx,
+ vbo_indirect_draw_func func)
+{
+ struct vbo_context *vbo = vbo_context(ctx);
+ vbo->draw_indirect_prims = func;
+}
* is responsible for initiating any fallback actions required:
*/
vbo_draw_func draw_prims;
+
+ /* Optional callback for indirect draws. This allows multidraws to not be
+ * broken up, as well as for the actual count to be passed in as a separate
+ * indirect parameter.
+ */
+ vbo_indirect_draw_func draw_indirect_prims;
};
* invoked according to the state flags. That will have to wait for a
* mesa rework:
*/
-void vbo_exec_invalidate_state( struct gl_context *ctx, GLuint new_state )
+void vbo_exec_invalidate_state( struct gl_context *ctx, GLbitfield new_state )
{
struct vbo_context *vbo = vbo_context(ctx);
struct vbo_exec_context *exec = &vbo->exec;
*/
void vbo_exec_init( struct gl_context *ctx );
void vbo_exec_destroy( struct gl_context *ctx );
-void vbo_exec_invalidate_state( struct gl_context *ctx, GLuint new_state );
+void vbo_exec_invalidate_state( struct gl_context *ctx, GLbitfield new_state );
/* Internal functions:
prim[i].indexed = 1;
prim[i].num_instances = 1;
prim[i].base_instance = 0;
+ prim[i].draw_id = i;
prim[i].is_indirect = 0;
if (basevertex != NULL)
prim[i].basevertex = basevertex[i];
prim[0].indexed = 1;
prim[0].num_instances = 1;
prim[0].base_instance = 0;
+ prim[0].draw_id = i;
prim[0].is_indirect = 0;
if (basevertex != NULL)
prim[0].basevertex = basevertex[i];
{
struct vbo_context *vbo = vbo_context(ctx);
struct vbo_exec_context *exec = &vbo->exec;
- struct _mesa_prim prim[1];
vbo_bind_arrays(ctx);
- memset(prim, 0, sizeof(prim));
- prim[0].begin = 1;
- prim[0].end = 1;
- prim[0].mode = mode;
- prim[0].is_indirect = 1;
- prim[0].indirect_offset = (GLsizeiptr)indirect;
-
- /* NOTE: We do NOT want to handle primitive restart here, nor perform any
- * other checks that require knowledge of the values in the command buffer.
- * That would defeat the whole purpose of this function.
- */
-
check_buffers_are_unmapped(exec->array.inputs);
- vbo->draw_prims(ctx, prim, 1,
- NULL, GL_TRUE, 0, ~0,
- NULL, 0,
- ctx->DrawIndirectBuffer);
+ vbo->draw_indirect_prims(ctx, mode,
+ ctx->DrawIndirectBuffer, (GLsizeiptr)indirect,
+ 1 /* draw_count */, 16 /* stride */,
+ NULL, 0, NULL);
if (MESA_DEBUG_FLAGS & DEBUG_ALWAYS_FLUSH)
_mesa_flush(ctx);
{
struct vbo_context *vbo = vbo_context(ctx);
struct vbo_exec_context *exec = &vbo->exec;
- struct _mesa_prim *prim;
- GLsizei i;
GLsizeiptr offset = (GLsizeiptr)indirect;
if (primcount == 0)
return;
- prim = calloc(primcount, sizeof(*prim));
- if (prim == NULL) {
- _mesa_error(ctx, GL_OUT_OF_MEMORY, "glMultiDrawArraysIndirect");
- return;
- }
vbo_bind_arrays(ctx);
- prim[0].begin = 1;
- prim[primcount - 1].end = 1;
- for (i = 0; i < primcount; ++i, offset += stride) {
- prim[i].mode = mode;
- prim[i].indirect_offset = offset;
- prim[i].is_indirect = 1;
- }
-
check_buffers_are_unmapped(exec->array.inputs);
- vbo->draw_prims(ctx, prim, primcount,
- NULL, GL_TRUE, 0, ~0,
- NULL, 0,
- ctx->DrawIndirectBuffer);
-
- free(prim);
+ vbo->draw_indirect_prims(ctx, mode,
+ ctx->DrawIndirectBuffer, offset,
+ primcount, stride,
+ NULL, 0, NULL);
if (MESA_DEBUG_FLAGS & DEBUG_ALWAYS_FLUSH)
_mesa_flush(ctx);
struct vbo_context *vbo = vbo_context(ctx);
struct vbo_exec_context *exec = &vbo->exec;
struct _mesa_index_buffer ib;
- struct _mesa_prim prim[1];
vbo_bind_arrays(ctx);
ib.obj = ctx->Array.VAO->IndexBufferObj;
ib.ptr = NULL;
- memset(prim, 0, sizeof(prim));
- prim[0].begin = 1;
- prim[0].end = 1;
- prim[0].mode = mode;
- prim[0].indexed = 1;
- prim[0].indirect_offset = (GLsizeiptr)indirect;
- prim[0].is_indirect = 1;
-
check_buffers_are_unmapped(exec->array.inputs);
- vbo->draw_prims(ctx, prim, 1,
- &ib, GL_TRUE, 0, ~0,
- NULL, 0,
- ctx->DrawIndirectBuffer);
+ vbo->draw_indirect_prims(ctx, mode,
+ ctx->DrawIndirectBuffer, (GLsizeiptr)indirect,
+ 1 /* draw_count */, 20 /* stride */,
+ NULL, 0,
+ &ib);
if (MESA_DEBUG_FLAGS & DEBUG_ALWAYS_FLUSH)
_mesa_flush(ctx);
struct vbo_context *vbo = vbo_context(ctx);
struct vbo_exec_context *exec = &vbo->exec;
struct _mesa_index_buffer ib;
- struct _mesa_prim *prim;
- GLsizei i;
GLsizeiptr offset = (GLsizeiptr)indirect;
if (primcount == 0)
return;
- prim = calloc(primcount, sizeof(*prim));
- if (prim == NULL) {
- _mesa_error(ctx, GL_OUT_OF_MEMORY, "glMultiDrawElementsIndirect");
- return;
- }
vbo_bind_arrays(ctx);
ib.obj = ctx->Array.VAO->IndexBufferObj;
ib.ptr = NULL;
- prim[0].begin = 1;
- prim[primcount - 1].end = 1;
- for (i = 0; i < primcount; ++i, offset += stride) {
- prim[i].mode = mode;
- prim[i].indexed = 1;
- prim[i].indirect_offset = offset;
- prim[i].is_indirect = 1;
- }
-
check_buffers_are_unmapped(exec->array.inputs);
- vbo->draw_prims(ctx, prim, primcount,
- &ib, GL_TRUE, 0, ~0,
- NULL, 0,
- ctx->DrawIndirectBuffer);
-
- free(prim);
+ vbo->draw_indirect_prims(ctx, mode,
+ ctx->DrawIndirectBuffer, offset,
+ primcount, stride,
+ NULL, 0,
+ &ib);
if (MESA_DEBUG_FLAGS & DEBUG_ALWAYS_FLUSH)
_mesa_flush(ctx);
primcount, stride);
}
+static void
+vbo_validated_multidrawarraysindirectcount(struct gl_context *ctx,
+ GLenum mode,
+ GLintptr indirect,
+ GLintptr drawcount,
+ GLsizei maxdrawcount,
+ GLsizei stride)
+{
+ struct vbo_context *vbo = vbo_context(ctx);
+ struct vbo_exec_context *exec = &vbo->exec;
+ GLsizeiptr offset = indirect;
+
+ if (maxdrawcount == 0)
+ return;
+
+ vbo_bind_arrays(ctx);
+
+ check_buffers_are_unmapped(exec->array.inputs);
+ vbo->draw_indirect_prims(ctx, mode,
+ ctx->DrawIndirectBuffer, offset,
+ maxdrawcount, stride,
+ ctx->ParameterBuffer, drawcount,
+ NULL);
+
+ if (MESA_DEBUG_FLAGS & DEBUG_ALWAYS_FLUSH)
+ _mesa_flush(ctx);
+}
+
+static void
+vbo_validated_multidrawelementsindirectcount(struct gl_context *ctx,
+ GLenum mode, GLenum type,
+ GLintptr indirect,
+ GLintptr drawcount,
+ GLsizei maxdrawcount,
+ GLsizei stride)
+{
+ struct vbo_context *vbo = vbo_context(ctx);
+ struct vbo_exec_context *exec = &vbo->exec;
+ struct _mesa_index_buffer ib;
+ GLsizeiptr offset = (GLsizeiptr)indirect;
+
+ if (maxdrawcount == 0)
+ return;
+
+ vbo_bind_arrays(ctx);
+
+ /* NOTE: IndexBufferObj is guaranteed to be a VBO. */
+
+ ib.count = 0; /* unknown */
+ ib.type = type;
+ ib.obj = ctx->Array.VAO->IndexBufferObj;
+ ib.ptr = NULL;
+
+ check_buffers_are_unmapped(exec->array.inputs);
+ vbo->draw_indirect_prims(ctx, mode,
+ ctx->DrawIndirectBuffer, offset,
+ maxdrawcount, stride,
+ ctx->ParameterBuffer, drawcount,
+ &ib);
+
+ if (MESA_DEBUG_FLAGS & DEBUG_ALWAYS_FLUSH)
+ _mesa_flush(ctx);
+}
+
+static void GLAPIENTRY
+vbo_exec_MultiDrawArraysIndirectCount(GLenum mode,
+ GLintptr indirect,
+ GLintptr drawcount,
+ GLsizei maxdrawcount, GLsizei stride)
+{
+ GET_CURRENT_CONTEXT(ctx);
+
+ if (MESA_VERBOSE & VERBOSE_DRAW)
+ _mesa_debug(ctx, "glMultiDrawArraysIndirectCountARB"
+ "(%s, %lx, %lx, %i, %i)\n",
+ _mesa_enum_to_string(mode), indirect,
+ drawcount, maxdrawcount, stride);
+
+ /* If <stride> is zero, the array elements are treated as tightly packed. */
+ if (stride == 0)
+ stride = 4 * sizeof(GLuint); /* sizeof(DrawArraysIndirectCommand) */
+
+ if (!_mesa_validate_MultiDrawArraysIndirectCount(ctx, mode,
+ indirect, drawcount,
+ maxdrawcount, stride))
+ return;
+
+ vbo_validated_multidrawarraysindirectcount(ctx, mode,
+ indirect, drawcount,
+ maxdrawcount, stride);
+}
+
+static void GLAPIENTRY
+vbo_exec_MultiDrawElementsIndirectCount(GLenum mode, GLenum type,
+ GLintptr indirect,
+ GLintptr drawcount,
+ GLsizei maxdrawcount, GLsizei stride)
+{
+ GET_CURRENT_CONTEXT(ctx);
+
+ if (MESA_VERBOSE & VERBOSE_DRAW)
+ _mesa_debug(ctx, "glMultiDrawElementsIndirectCountARB"
+ "(%s, %s, %lx, %lx, %i, %i)\n",
+ _mesa_enum_to_string(mode),
+ _mesa_enum_to_string(type), indirect,
+ drawcount, maxdrawcount, stride);
+
+ /* If <stride> is zero, the array elements are treated as tightly packed. */
+ if (stride == 0)
+ stride = 5 * sizeof(GLuint); /* sizeof(DrawElementsIndirectCommand) */
+
+ if (!_mesa_validate_MultiDrawElementsIndirectCount(ctx, mode, type,
+ indirect, drawcount,
+ maxdrawcount, stride))
+ return;
+
+ vbo_validated_multidrawelementsindirectcount(ctx, mode, type,
+ indirect, drawcount,
+ maxdrawcount, stride);
+}
+
+
/**
* Initialize the dispatch table with the VBO functions for drawing.
*/
if (ctx->API == API_OPENGL_CORE) {
SET_MultiDrawArraysIndirect(exec, vbo_exec_MultiDrawArraysIndirect);
SET_MultiDrawElementsIndirect(exec, vbo_exec_MultiDrawElementsIndirect);
+ SET_MultiDrawArraysIndirectCountARB(exec, vbo_exec_MultiDrawArraysIndirectCount);
+ SET_MultiDrawElementsIndirectCountARB(exec, vbo_exec_MultiDrawElementsIndirectCount);
}
if (_mesa_is_desktop_gl(ctx) || _mesa_is_gles3(ctx)) {
static inline unsigned
__bitset_next_set(unsigned i, BITSET_WORD *tmp,
- BITSET_WORD *set, unsigned size)
+ const BITSET_WORD *set, unsigned size)
{
unsigned bit, word;
return length;
}
+static inline void list_splice(struct list_head *src, struct list_head *dst)
+{
+ if (list_empty(src))
+ return;
+
+ src->next->prev = dst;
+ src->prev->next = dst->next;
+ dst->next->prev = src->prev;
+ dst->next = src->next;
+}
+
+static inline void list_splicetail(struct list_head *src, struct list_head *dst)
+{
+ if (list_empty(src))
+ return;
+
+ src->prev->next = dst;
+ src->next->prev = dst->prev;
+ dst->prev->next = src->next;
+ dst->prev = src->prev;
+}
+
static inline void list_validate(struct list_head *list)
{
struct list_head *node;
/* Connect the two lists together; parent them to new_ctx; make old_ctx empty. */
child->next = new_info->child;
+ child->parent = new_info;
new_info->child = old_info->child;
old_info->child = NULL;
}
--- /dev/null
+# Generated source files
+/*_spirv_autogen.h
+/anv_entrypoints.c
+/anv_entrypoints.h
+/wayland-drm-protocol.c
+/wayland-drm-client-protocol.h
+/anv_icd.json
--- /dev/null
+# Copyright © 2015 Intel Corporation
+#
+# Permission is hereby granted, free of charge, to any person obtaining a
+# copy of this software and associated documentation files (the "Software"),
+# to deal in the Software without restriction, including without limitation
+# the rights to use, copy, modify, merge, publish, distribute, sublicense,
+# and/or sell copies of the Software, and to permit persons to whom the
+# Software is furnished to do so, subject to the following conditions:
+#
+# The above copyright notice and this permission notice (including the next
+# paragraph) shall be included in all copies or substantial portions of the
+# Software.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+# THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+# IN THE SOFTWARE.
+
+SUBDIRS = . tests
+
+vulkan_includedir = $(includedir)/vulkan
+
+vulkan_include_HEADERS = \
+ $(top_srcdir)/include/vulkan/vk_platform.h \
+ $(top_srcdir)/include/vulkan/vulkan.h \
+ $(top_srcdir)/include/vulkan/vulkan_intel.h
+
+# Used when generating entrypoints to filter out unwanted extensions
+VULKAN_ENTRYPOINT_CPPFLAGS = \
+ -I$(top_srcdir)/include/vulkan \
+ -DVK_USE_PLATFORM_XCB_KHR \
+ -DVK_USE_PLATFORM_WAYLAND_KHR
+
+lib_LTLIBRARIES = libvulkan.la
+
+check_LTLIBRARIES = libvulkan-test.la
+
+PER_GEN_LIBS = \
+ libanv-gen7.la \
+ libanv-gen75.la \
+ libanv-gen8.la \
+ libanv-gen9.la
+
+noinst_LTLIBRARIES = $(PER_GEN_LIBS)
+
+# The gallium includes are for the util/u_math.h include from main/macros.h
+
+AM_CPPFLAGS = \
+ $(INTEL_CFLAGS) \
+ $(VALGRIND_CFLAGS) \
+ $(DEFINES) \
+ -I$(top_srcdir)/include \
+ -I$(top_srcdir)/src \
+ -I$(top_srcdir)/src/glsl/nir \
+ -I$(top_srcdir)/src/mapi \
+ -I$(top_srcdir)/src/mesa \
+ -I$(top_srcdir)/src/mesa/drivers/dri/common \
+ -I$(top_srcdir)/src/mesa/drivers/dri/i965 \
+ -I$(top_srcdir)/src/gallium/auxiliary \
+ -I$(top_srcdir)/src/gallium/include \
+ -I$(top_srcdir)/src/isl/ \
+ -I$(top_builddir)/src \
+ -I$(top_builddir)/src/glsl/nir \
+ -I$(top_builddir)/src/vulkan
+
+libvulkan_la_CFLAGS = $(CFLAGS) -Wno-override-init
+
+VULKAN_SOURCES = \
+ anv_allocator.c \
+ anv_cmd_buffer.c \
+ anv_batch_chain.c \
+ anv_descriptor_set.c \
+ anv_device.c \
+ anv_dump.c \
+ anv_entrypoints.c \
+ anv_entrypoints.h \
+ anv_formats.c \
+ anv_image.c \
+ anv_intel.c \
+ anv_meta.c \
+ anv_meta_clear.c \
+ anv_nir_apply_dynamic_offsets.c \
+ anv_nir_apply_pipeline_layout.c \
+ anv_nir_lower_push_constants.c \
+ anv_pass.c \
+ anv_pipeline.c \
+ anv_private.h \
+ anv_query.c \
+ anv_util.c \
+ anv_wsi.c \
+ anv_wsi_x11.c
+
+BUILT_SOURCES = \
+ anv_entrypoints.h \
+ anv_entrypoints.c
+
+libanv_gen7_la_SOURCES = \
+ genX_cmd_buffer.c \
+ gen7_cmd_buffer.c \
+ gen7_pipeline.c \
+ gen7_state.c
+libanv_gen7_la_CFLAGS = $(libvulkan_la_CFLAGS) -DANV_GENx10=70
+
+libanv_gen75_la_SOURCES = \
+ genX_cmd_buffer.c \
+ gen7_cmd_buffer.c \
+ gen7_pipeline.c \
+ gen7_state.c
+libanv_gen75_la_CFLAGS = $(libvulkan_la_CFLAGS) -DANV_GENx10=75
+
+libanv_gen8_la_SOURCES = \
+ genX_cmd_buffer.c \
+ gen8_cmd_buffer.c \
+ gen8_pipeline.c \
+ gen8_state.c
+libanv_gen8_la_CFLAGS = $(libvulkan_la_CFLAGS) -DANV_GENx10=80
+
+libanv_gen9_la_SOURCES = \
+ genX_cmd_buffer.c \
+ gen8_cmd_buffer.c \
+ gen8_pipeline.c \
+ gen8_state.c
+libanv_gen9_la_CFLAGS = $(libvulkan_la_CFLAGS) -DANV_GENx10=90
+
+if HAVE_EGL_PLATFORM_WAYLAND
+BUILT_SOURCES += \
+ wayland-drm-protocol.c \
+ wayland-drm-client-protocol.h
+
+%-protocol.c : $(top_srcdir)/src/egl/wayland/wayland-drm/%.xml
+ $(AM_V_GEN)$(WAYLAND_SCANNER) code < $< > $@
+
+%-client-protocol.h : $(top_srcdir)/src/egl/wayland/wayland-drm/%.xml
+ $(AM_V_GEN)$(WAYLAND_SCANNER) client-header < $< > $@
+
+AM_CPPFLAGS += -I$(top_srcdir)/src/egl/wayland/wayland-drm
+VULKAN_SOURCES += \
+ wayland-drm-protocol.c \
+ anv_wsi_wayland.c
+libvulkan_la_CFLAGS += -DHAVE_WAYLAND_PLATFORM
+endif
+
+libvulkan_la_SOURCES = \
+ $(VULKAN_SOURCES) \
+ anv_gem.c
+
+anv_entrypoints.h : anv_entrypoints_gen.py $(vulkan_include_HEADERS)
+ $(AM_V_GEN) cat $(vulkan_include_HEADERS) | $(CPP) $(VULKAN_ENTRYPOINT_CPPFLAGS) - | $(PYTHON2) $< header > $@
+
+anv_entrypoints.c : anv_entrypoints_gen.py $(vulkan_include_HEADERS)
+ $(AM_V_GEN) cat $(vulkan_include_HEADERS) | $(CPP) $(VULKAN_ENTRYPOINT_CPPFLAGS) - | $(PYTHON2) $< code > $@
+
+CLEANFILES = $(BUILT_SOURCES)
+
+libvulkan_la_LIBADD = $(WAYLAND_LIBS) -lxcb -lxcb-dri3 \
+ $(top_builddir)/src/isl/libisl.la \
+ $(top_builddir)/src/mesa/drivers/dri/i965/libi965_compiler.la \
+ ../mesa/libmesa.la \
+ ../mesa/drivers/dri/common/libdri_test_stubs.la \
+ -lpthread -ldl -lstdc++ \
+ $(PER_GEN_LIBS)
+
+# Libvulkan with dummy gem. Used for unit tests.
+
+libvulkan_test_la_SOURCES = \
+ $(VULKAN_SOURCES) \
+ anv_gem_stubs.c
+
+libvulkan_test_la_CFLAGS = $(libvulkan_la_CFLAGS)
+libvulkan_test_la_LIBADD = $(libvulkan_la_LIBADD)
+
+include $(top_srcdir)/install-lib-links.mk
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#define _DEFAULT_SOURCE
+
+#include <stdint.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <values.h>
+#include <assert.h>
+#include <linux/futex.h>
+#include <linux/memfd.h>
+#include <sys/time.h>
+#include <sys/mman.h>
+#include <sys/syscall.h>
+
+#include "anv_private.h"
+
+#ifdef HAVE_VALGRIND
+#define VG_NOACCESS_READ(__ptr) ({ \
+ VALGRIND_MAKE_MEM_DEFINED((__ptr), sizeof(*(__ptr))); \
+ __typeof(*(__ptr)) __val = *(__ptr); \
+ VALGRIND_MAKE_MEM_NOACCESS((__ptr), sizeof(*(__ptr)));\
+ __val; \
+})
+#define VG_NOACCESS_WRITE(__ptr, __val) ({ \
+ VALGRIND_MAKE_MEM_UNDEFINED((__ptr), sizeof(*(__ptr))); \
+ *(__ptr) = (__val); \
+ VALGRIND_MAKE_MEM_NOACCESS((__ptr), sizeof(*(__ptr))); \
+})
+#else
+#define VG_NOACCESS_READ(__ptr) (*(__ptr))
+#define VG_NOACCESS_WRITE(__ptr, __val) (*(__ptr) = (__val))
+#endif
+
+/* Design goals:
+ *
+ * - Lock free (except when resizing underlying bos)
+ *
+ * - Constant time allocation with typically only one atomic
+ *
+ * - Multiple allocation sizes without fragmentation
+ *
+ * - Can grow while keeping addresses and offset of contents stable
+ *
+ * - All allocations within one bo so we can point one of the
+ * STATE_BASE_ADDRESS pointers at it.
+ *
+ * The overall design is a two-level allocator: top level is a fixed size, big
+ * block (8k) allocator, which operates out of a bo. Allocation is done by
+ * either pulling a block from the free list or growing the used range of the
+ * bo. Growing the range may run out of space in the bo which we then need to
+ * grow. Growing the bo is tricky in a multi-threaded, lockless environment:
+ * we need to keep all pointers and contents in the old map valid. GEM bos in
+ * general can't grow, but we use a trick: we create a memfd and use ftruncate
+ * to grow it as necessary. We mmap the new size and then create a gem bo for
+ * it using the new gem userptr ioctl. Without heavy-handed locking around
+ * our allocation fast-path, there isn't really a way to munmap the old mmap,
+ * so we just keep it around until garbage collection time. While the block
+ * allocator is lockless for normal operations, we block other threads trying
+ * to allocate while we're growing the map. It sholdn't happen often, and
+ * growing is fast anyway.
+ *
+ * At the next level we can use various sub-allocators. The state pool is a
+ * pool of smaller, fixed size objects, which operates much like the block
+ * pool. It uses a free list for freeing objects, but when it runs out of
+ * space it just allocates a new block from the block pool. This allocator is
+ * intended for longer lived state objects such as SURFACE_STATE and most
+ * other persistent state objects in the API. We may need to track more info
+ * with these object and a pointer back to the CPU object (eg VkImage). In
+ * those cases we just allocate a slightly bigger object and put the extra
+ * state after the GPU state object.
+ *
+ * The state stream allocator works similar to how the i965 DRI driver streams
+ * all its state. Even with Vulkan, we need to emit transient state (whether
+ * surface state base or dynamic state base), and for that we can just get a
+ * block and fill it up. These cases are local to a command buffer and the
+ * sub-allocator need not be thread safe. The streaming allocator gets a new
+ * block when it runs out of space and chains them together so they can be
+ * easily freed.
+ */
+
+/* Allocations are always at least 64 byte aligned, so 1 is an invalid value.
+ * We use it to indicate the free list is empty. */
+#define EMPTY 1
+
+struct anv_mmap_cleanup {
+ void *map;
+ size_t size;
+ uint32_t gem_handle;
+};
+
+#define ANV_MMAP_CLEANUP_INIT ((struct anv_mmap_cleanup){0})
+
+static inline long
+sys_futex(void *addr1, int op, int val1,
+ struct timespec *timeout, void *addr2, int val3)
+{
+ return syscall(SYS_futex, addr1, op, val1, timeout, addr2, val3);
+}
+
+static inline int
+futex_wake(uint32_t *addr, int count)
+{
+ return sys_futex(addr, FUTEX_WAKE, count, NULL, NULL, 0);
+}
+
+static inline int
+futex_wait(uint32_t *addr, int32_t value)
+{
+ return sys_futex(addr, FUTEX_WAIT, value, NULL, NULL, 0);
+}
+
+static inline int
+memfd_create(const char *name, unsigned int flags)
+{
+ return syscall(SYS_memfd_create, name, flags);
+}
+
+static inline uint32_t
+ilog2_round_up(uint32_t value)
+{
+ assert(value != 0);
+ return 32 - __builtin_clz(value - 1);
+}
+
+static inline uint32_t
+round_to_power_of_two(uint32_t value)
+{
+ return 1 << ilog2_round_up(value);
+}
+
+static bool
+anv_free_list_pop(union anv_free_list *list, void **map, int32_t *offset)
+{
+ union anv_free_list current, new, old;
+
+ current.u64 = list->u64;
+ while (current.offset != EMPTY) {
+ /* We have to add a memory barrier here so that the list head (and
+ * offset) gets read before we read the map pointer. This way we
+ * know that the map pointer is valid for the given offset at the
+ * point where we read it.
+ */
+ __sync_synchronize();
+
+ int32_t *next_ptr = *map + current.offset;
+ new.offset = VG_NOACCESS_READ(next_ptr);
+ new.count = current.count + 1;
+ old.u64 = __sync_val_compare_and_swap(&list->u64, current.u64, new.u64);
+ if (old.u64 == current.u64) {
+ *offset = current.offset;
+ return true;
+ }
+ current = old;
+ }
+
+ return false;
+}
+
+static void
+anv_free_list_push(union anv_free_list *list, void *map, int32_t offset)
+{
+ union anv_free_list current, old, new;
+ int32_t *next_ptr = map + offset;
+
+ old = *list;
+ do {
+ current = old;
+ VG_NOACCESS_WRITE(next_ptr, current.offset);
+ new.offset = offset;
+ new.count = current.count + 1;
+ old.u64 = __sync_val_compare_and_swap(&list->u64, current.u64, new.u64);
+ } while (old.u64 != current.u64);
+}
+
+/* All pointers in the ptr_free_list are assumed to be page-aligned. This
+ * means that the bottom 12 bits should all be zero.
+ */
+#define PFL_COUNT(x) ((uintptr_t)(x) & 0xfff)
+#define PFL_PTR(x) ((void *)((uintptr_t)(x) & ~0xfff))
+#define PFL_PACK(ptr, count) ({ \
+ assert(((uintptr_t)(ptr) & 0xfff) == 0); \
+ (void *)((uintptr_t)(ptr) | (uintptr_t)((count) & 0xfff)); \
+})
+
+static bool
+anv_ptr_free_list_pop(void **list, void **elem)
+{
+ void *current = *list;
+ while (PFL_PTR(current) != NULL) {
+ void **next_ptr = PFL_PTR(current);
+ void *new_ptr = VG_NOACCESS_READ(next_ptr);
+ unsigned new_count = PFL_COUNT(current) + 1;
+ void *new = PFL_PACK(new_ptr, new_count);
+ void *old = __sync_val_compare_and_swap(list, current, new);
+ if (old == current) {
+ *elem = PFL_PTR(current);
+ return true;
+ }
+ current = old;
+ }
+
+ return false;
+}
+
+static void
+anv_ptr_free_list_push(void **list, void *elem)
+{
+ void *old, *current;
+ void **next_ptr = elem;
+
+ old = *list;
+ do {
+ current = old;
+ VG_NOACCESS_WRITE(next_ptr, PFL_PTR(current));
+ unsigned new_count = PFL_COUNT(current) + 1;
+ void *new = PFL_PACK(elem, new_count);
+ old = __sync_val_compare_and_swap(list, current, new);
+ } while (old != current);
+}
+
+static uint32_t
+anv_block_pool_grow(struct anv_block_pool *pool, struct anv_block_state *state);
+
+void
+anv_block_pool_init(struct anv_block_pool *pool,
+ struct anv_device *device, uint32_t block_size)
+{
+ assert(util_is_power_of_two(block_size));
+
+ pool->device = device;
+ pool->bo.gem_handle = 0;
+ pool->bo.offset = 0;
+ pool->bo.size = 0;
+ pool->block_size = block_size;
+ pool->free_list = ANV_FREE_LIST_EMPTY;
+ pool->back_free_list = ANV_FREE_LIST_EMPTY;
+
+ pool->fd = memfd_create("block pool", MFD_CLOEXEC);
+ if (pool->fd == -1)
+ return;
+
+ /* Just make it 2GB up-front. The Linux kernel won't actually back it
+ * with pages until we either map and fault on one of them or we use
+ * userptr and send a chunk of it off to the GPU.
+ */
+ if (ftruncate(pool->fd, BLOCK_POOL_MEMFD_SIZE) == -1)
+ return;
+
+ anv_vector_init(&pool->mmap_cleanups,
+ round_to_power_of_two(sizeof(struct anv_mmap_cleanup)), 128);
+
+ pool->state.next = 0;
+ pool->state.end = 0;
+ pool->back_state.next = 0;
+ pool->back_state.end = 0;
+
+ /* Immediately grow the pool so we'll have a backing bo. */
+ pool->state.end = anv_block_pool_grow(pool, &pool->state);
+}
+
+void
+anv_block_pool_finish(struct anv_block_pool *pool)
+{
+ struct anv_mmap_cleanup *cleanup;
+
+ anv_vector_foreach(cleanup, &pool->mmap_cleanups) {
+ if (cleanup->map)
+ munmap(cleanup->map, cleanup->size);
+ if (cleanup->gem_handle)
+ anv_gem_close(pool->device, cleanup->gem_handle);
+ }
+
+ anv_vector_finish(&pool->mmap_cleanups);
+
+ close(pool->fd);
+}
+
+#define PAGE_SIZE 4096
+
+/** Grows and re-centers the block pool.
+ *
+ * We grow the block pool in one or both directions in such a way that the
+ * following conditions are met:
+ *
+ * 1) The size of the entire pool is always a power of two.
+ *
+ * 2) The pool only grows on both ends. Neither end can get
+ * shortened.
+ *
+ * 3) At the end of the allocation, we have about twice as much space
+ * allocated for each end as we have used. This way the pool doesn't
+ * grow too far in one direction or the other.
+ *
+ * 4) If the _alloc_back() has never been called, then the back portion of
+ * the pool retains a size of zero. (This makes it easier for users of
+ * the block pool that only want a one-sided pool.)
+ *
+ * 5) We have enough space allocated for at least one more block in
+ * whichever side `state` points to.
+ *
+ * 6) The center of the pool is always aligned to both the block_size of
+ * the pool and a 4K CPU page.
+ */
+static uint32_t
+anv_block_pool_grow(struct anv_block_pool *pool, struct anv_block_state *state)
+{
+ size_t size;
+ void *map;
+ uint32_t gem_handle;
+ struct anv_mmap_cleanup *cleanup;
+
+ pthread_mutex_lock(&pool->device->mutex);
+
+ assert(state == &pool->state || state == &pool->back_state);
+
+ /* Gather a little usage information on the pool. Since we may have
+ * threadsd waiting in queue to get some storage while we resize, it's
+ * actually possible that total_used will be larger than old_size. In
+ * particular, block_pool_alloc() increments state->next prior to
+ * calling block_pool_grow, so this ensures that we get enough space for
+ * which ever side tries to grow the pool.
+ *
+ * We align to a page size because it makes it easier to do our
+ * calculations later in such a way that we state page-aigned.
+ */
+ uint32_t back_used = align_u32(pool->back_state.next, PAGE_SIZE);
+ uint32_t front_used = align_u32(pool->state.next, PAGE_SIZE);
+ uint32_t total_used = front_used + back_used;
+
+ assert(state == &pool->state || back_used > 0);
+
+ size_t old_size = pool->bo.size;
+
+ if (old_size != 0 &&
+ back_used * 2 <= pool->center_bo_offset &&
+ front_used * 2 <= (old_size - pool->center_bo_offset)) {
+ /* If we're in this case then this isn't the firsta allocation and we
+ * already have enough space on both sides to hold double what we
+ * have allocated. There's nothing for us to do.
+ */
+ goto done;
+ }
+
+ if (old_size == 0) {
+ /* This is the first allocation */
+ size = MAX2(32 * pool->block_size, PAGE_SIZE);
+ } else {
+ size = old_size * 2;
+ }
+
+ /* We can't have a block pool bigger than 1GB because we use signed
+ * 32-bit offsets in the free list and we don't want overflow. We
+ * should never need a block pool bigger than 1GB anyway.
+ */
+ assert(size <= (1u << 31));
+
+ /* We compute a new center_bo_offset such that, when we double the size
+ * of the pool, we maintain the ratio of how much is used by each side.
+ * This way things should remain more-or-less balanced.
+ */
+ uint32_t center_bo_offset;
+ if (back_used == 0) {
+ /* If we're in this case then we have never called alloc_back(). In
+ * this case, we want keep the offset at 0 to make things as simple
+ * as possible for users that don't care about back allocations.
+ */
+ center_bo_offset = 0;
+ } else {
+ /* Try to "center" the allocation based on how much is currently in
+ * use on each side of the center line.
+ */
+ center_bo_offset = ((uint64_t)size * back_used) / total_used;
+
+ /* Align down to a multiple of both the block size and page size */
+ uint32_t granularity = MAX2(pool->block_size, PAGE_SIZE);
+ assert(util_is_power_of_two(granularity));
+ center_bo_offset &= ~(granularity - 1);
+
+ assert(center_bo_offset >= back_used);
+
+ /* Make sure we don't shrink the back end of the pool */
+ if (center_bo_offset < pool->back_state.end)
+ center_bo_offset = pool->back_state.end;
+
+ /* Make sure that we don't shrink the front end of the pool */
+ if (size - center_bo_offset < pool->state.end)
+ center_bo_offset = size - pool->state.end;
+ }
+
+ assert(center_bo_offset % pool->block_size == 0);
+ assert(center_bo_offset % PAGE_SIZE == 0);
+
+ /* Assert that we only ever grow the pool */
+ assert(center_bo_offset >= pool->back_state.end);
+ assert(size - center_bo_offset >= pool->state.end);
+
+ cleanup = anv_vector_add(&pool->mmap_cleanups);
+ if (!cleanup)
+ goto fail;
+ *cleanup = ANV_MMAP_CLEANUP_INIT;
+
+ /* Just leak the old map until we destroy the pool. We can't munmap it
+ * without races or imposing locking on the block allocate fast path. On
+ * the whole the leaked maps adds up to less than the size of the
+ * current map. MAP_POPULATE seems like the right thing to do, but we
+ * should try to get some numbers.
+ */
+ map = mmap(NULL, size, PROT_READ | PROT_WRITE,
+ MAP_SHARED | MAP_POPULATE, pool->fd,
+ BLOCK_POOL_MEMFD_CENTER - center_bo_offset);
+ cleanup->map = map;
+ cleanup->size = size;
+
+ if (map == MAP_FAILED)
+ goto fail;
+
+ gem_handle = anv_gem_userptr(pool->device, map, size);
+ if (gem_handle == 0)
+ goto fail;
+ cleanup->gem_handle = gem_handle;
+
+ /* Regular objects are created I915_CACHING_CACHED on LLC platforms and
+ * I915_CACHING_NONE on non-LLC platforms. However, userptr objects are
+ * always created as I915_CACHING_CACHED, which on non-LLC means
+ * snooped. That can be useful but comes with a bit of overheard. Since
+ * we're eplicitly clflushing and don't want the overhead we need to turn
+ * it off. */
+ if (!pool->device->info.has_llc) {
+ anv_gem_set_caching(pool->device, gem_handle, I915_CACHING_NONE);
+ anv_gem_set_domain(pool->device, gem_handle,
+ I915_GEM_DOMAIN_GTT, I915_GEM_DOMAIN_GTT);
+ }
+
+ /* Now that we successfull allocated everything, we can write the new
+ * values back into pool. */
+ pool->map = map + center_bo_offset;
+ pool->center_bo_offset = center_bo_offset;
+ pool->bo.gem_handle = gem_handle;
+ pool->bo.size = size;
+ pool->bo.map = map;
+ pool->bo.index = 0;
+
+done:
+ pthread_mutex_unlock(&pool->device->mutex);
+
+ /* Return the appropreate new size. This function never actually
+ * updates state->next. Instead, we let the caller do that because it
+ * needs to do so in order to maintain its concurrency model.
+ */
+ if (state == &pool->state) {
+ return pool->bo.size - pool->center_bo_offset;
+ } else {
+ assert(pool->center_bo_offset > 0);
+ return pool->center_bo_offset;
+ }
+
+fail:
+ pthread_mutex_unlock(&pool->device->mutex);
+
+ return 0;
+}
+
+static uint32_t
+anv_block_pool_alloc_new(struct anv_block_pool *pool,
+ struct anv_block_state *pool_state)
+{
+ struct anv_block_state state, old, new;
+
+ while (1) {
+ state.u64 = __sync_fetch_and_add(&pool_state->u64, pool->block_size);
+ if (state.next < state.end) {
+ assert(pool->map);
+ return state.next;
+ } else if (state.next == state.end) {
+ /* We allocated the first block outside the pool, we have to grow it.
+ * pool_state->next acts a mutex: threads who try to allocate now will
+ * get block indexes above the current limit and hit futex_wait
+ * below. */
+ new.next = state.next + pool->block_size;
+ new.end = anv_block_pool_grow(pool, pool_state);
+ assert(new.end >= new.next && new.end % pool->block_size == 0);
+ old.u64 = __sync_lock_test_and_set(&pool_state->u64, new.u64);
+ if (old.next != state.next)
+ futex_wake(&pool_state->end, INT_MAX);
+ return state.next;
+ } else {
+ futex_wait(&pool_state->end, state.end);
+ continue;
+ }
+ }
+}
+
+int32_t
+anv_block_pool_alloc(struct anv_block_pool *pool)
+{
+ int32_t offset;
+
+ /* Try free list first. */
+ if (anv_free_list_pop(&pool->free_list, &pool->map, &offset)) {
+ assert(offset >= 0);
+ assert(pool->map);
+ return offset;
+ }
+
+ return anv_block_pool_alloc_new(pool, &pool->state);
+}
+
+/* Allocates a block out of the back of the block pool.
+ *
+ * This will allocated a block earlier than the "start" of the block pool.
+ * The offsets returned from this function will be negative but will still
+ * be correct relative to the block pool's map pointer.
+ *
+ * If you ever use anv_block_pool_alloc_back, then you will have to do
+ * gymnastics with the block pool's BO when doing relocations.
+ */
+int32_t
+anv_block_pool_alloc_back(struct anv_block_pool *pool)
+{
+ int32_t offset;
+
+ /* Try free list first. */
+ if (anv_free_list_pop(&pool->back_free_list, &pool->map, &offset)) {
+ assert(offset < 0);
+ assert(pool->map);
+ return offset;
+ }
+
+ offset = anv_block_pool_alloc_new(pool, &pool->back_state);
+
+ /* The offset we get out of anv_block_pool_alloc_new() is actually the
+ * number of bytes downwards from the middle to the end of the block.
+ * We need to turn it into a (negative) offset from the middle to the
+ * start of the block.
+ */
+ assert(offset >= 0);
+ return -(offset + pool->block_size);
+}
+
+void
+anv_block_pool_free(struct anv_block_pool *pool, int32_t offset)
+{
+ if (offset < 0) {
+ anv_free_list_push(&pool->back_free_list, pool->map, offset);
+ } else {
+ anv_free_list_push(&pool->free_list, pool->map, offset);
+ }
+}
+
+static void
+anv_fixed_size_state_pool_init(struct anv_fixed_size_state_pool *pool,
+ size_t state_size)
+{
+ /* At least a cache line and must divide the block size. */
+ assert(state_size >= 64 && util_is_power_of_two(state_size));
+
+ pool->state_size = state_size;
+ pool->free_list = ANV_FREE_LIST_EMPTY;
+ pool->block.next = 0;
+ pool->block.end = 0;
+}
+
+static uint32_t
+anv_fixed_size_state_pool_alloc(struct anv_fixed_size_state_pool *pool,
+ struct anv_block_pool *block_pool)
+{
+ int32_t offset;
+ struct anv_block_state block, old, new;
+
+ /* Try free list first. */
+ if (anv_free_list_pop(&pool->free_list, &block_pool->map, &offset)) {
+ assert(offset >= 0);
+ return offset;
+ }
+
+ /* If free list was empty (or somebody raced us and took the items) we
+ * allocate a new item from the end of the block */
+ restart:
+ block.u64 = __sync_fetch_and_add(&pool->block.u64, pool->state_size);
+
+ if (block.next < block.end) {
+ return block.next;
+ } else if (block.next == block.end) {
+ offset = anv_block_pool_alloc(block_pool);
+ new.next = offset + pool->state_size;
+ new.end = offset + block_pool->block_size;
+ old.u64 = __sync_lock_test_and_set(&pool->block.u64, new.u64);
+ if (old.next != block.next)
+ futex_wake(&pool->block.end, INT_MAX);
+ return offset;
+ } else {
+ futex_wait(&pool->block.end, block.end);
+ goto restart;
+ }
+}
+
+static void
+anv_fixed_size_state_pool_free(struct anv_fixed_size_state_pool *pool,
+ struct anv_block_pool *block_pool,
+ uint32_t offset)
+{
+ anv_free_list_push(&pool->free_list, block_pool->map, offset);
+}
+
+void
+anv_state_pool_init(struct anv_state_pool *pool,
+ struct anv_block_pool *block_pool)
+{
+ pool->block_pool = block_pool;
+ for (unsigned i = 0; i < ANV_STATE_BUCKETS; i++) {
+ size_t size = 1 << (ANV_MIN_STATE_SIZE_LOG2 + i);
+ anv_fixed_size_state_pool_init(&pool->buckets[i], size);
+ }
+ VG(VALGRIND_CREATE_MEMPOOL(pool, 0, false));
+}
+
+void
+anv_state_pool_finish(struct anv_state_pool *pool)
+{
+ VG(VALGRIND_DESTROY_MEMPOOL(pool));
+}
+
+struct anv_state
+anv_state_pool_alloc(struct anv_state_pool *pool, size_t size, size_t align)
+{
+ unsigned size_log2 = ilog2_round_up(size < align ? align : size);
+ assert(size_log2 <= ANV_MAX_STATE_SIZE_LOG2);
+ if (size_log2 < ANV_MIN_STATE_SIZE_LOG2)
+ size_log2 = ANV_MIN_STATE_SIZE_LOG2;
+ unsigned bucket = size_log2 - ANV_MIN_STATE_SIZE_LOG2;
+
+ struct anv_state state;
+ state.alloc_size = 1 << size_log2;
+ state.offset = anv_fixed_size_state_pool_alloc(&pool->buckets[bucket],
+ pool->block_pool);
+ state.map = pool->block_pool->map + state.offset;
+ VG(VALGRIND_MEMPOOL_ALLOC(pool, state.map, size));
+ return state;
+}
+
+void
+anv_state_pool_free(struct anv_state_pool *pool, struct anv_state state)
+{
+ assert(util_is_power_of_two(state.alloc_size));
+ unsigned size_log2 = ilog2_round_up(state.alloc_size);
+ assert(size_log2 >= ANV_MIN_STATE_SIZE_LOG2 &&
+ size_log2 <= ANV_MAX_STATE_SIZE_LOG2);
+ unsigned bucket = size_log2 - ANV_MIN_STATE_SIZE_LOG2;
+
+ VG(VALGRIND_MEMPOOL_FREE(pool, state.map));
+ anv_fixed_size_state_pool_free(&pool->buckets[bucket],
+ pool->block_pool, state.offset);
+}
+
+#define NULL_BLOCK 1
+struct anv_state_stream_block {
+ /* The next block */
+ struct anv_state_stream_block *next;
+
+ /* The offset into the block pool at which this block starts */
+ uint32_t offset;
+
+#ifdef HAVE_VALGRIND
+ /* A pointer to the first user-allocated thing in this block. This is
+ * what valgrind sees as the start of the block.
+ */
+ void *_vg_ptr;
+#endif
+};
+
+/* The state stream allocator is a one-shot, single threaded allocator for
+ * variable sized blocks. We use it for allocating dynamic state.
+ */
+void
+anv_state_stream_init(struct anv_state_stream *stream,
+ struct anv_block_pool *block_pool)
+{
+ stream->block_pool = block_pool;
+ stream->block = NULL;
+
+ /* Ensure that next + whatever > end. This way the first call to
+ * state_stream_alloc fetches a new block.
+ */
+ stream->next = 1;
+ stream->end = 0;
+
+ VG(VALGRIND_CREATE_MEMPOOL(stream, 0, false));
+}
+
+void
+anv_state_stream_finish(struct anv_state_stream *stream)
+{
+ const uint32_t block_size = stream->block_pool->block_size;
+
+ struct anv_state_stream_block *next = stream->block;
+ while (next != NULL) {
+ VG(VALGRIND_MAKE_MEM_DEFINED(next, sizeof(*next)));
+ struct anv_state_stream_block sb = VG_NOACCESS_READ(next);
+ VG(VALGRIND_MEMPOOL_FREE(stream, sb._vg_ptr));
+ VG(VALGRIND_MAKE_MEM_UNDEFINED(next, block_size));
+ anv_block_pool_free(stream->block_pool, sb.offset);
+ next = sb.next;
+ }
+
+ VG(VALGRIND_DESTROY_MEMPOOL(stream));
+}
+
+struct anv_state
+anv_state_stream_alloc(struct anv_state_stream *stream,
+ uint32_t size, uint32_t alignment)
+{
+ struct anv_state_stream_block *sb = stream->block;
+
+ struct anv_state state;
+
+ state.offset = align_u32(stream->next, alignment);
+ if (state.offset + size > stream->end) {
+ uint32_t block = anv_block_pool_alloc(stream->block_pool);
+ sb = stream->block_pool->map + block;
+
+ VG(VALGRIND_MAKE_MEM_UNDEFINED(sb, sizeof(*sb)));
+ sb->next = stream->block;
+ sb->offset = block;
+ VG(sb->_vg_ptr = NULL);
+ VG(VALGRIND_MAKE_MEM_NOACCESS(sb, stream->block_pool->block_size));
+
+ stream->block = sb;
+ stream->start = block;
+ stream->next = block + sizeof(*sb);
+ stream->end = block + stream->block_pool->block_size;
+
+ state.offset = align_u32(stream->next, alignment);
+ assert(state.offset + size <= stream->end);
+ }
+
+ assert(state.offset > stream->start);
+ state.map = (void *)sb + (state.offset - stream->start);
+ state.alloc_size = size;
+
+#ifdef HAVE_VALGRIND
+ void *vg_ptr = VG_NOACCESS_READ(&sb->_vg_ptr);
+ if (vg_ptr == NULL) {
+ vg_ptr = state.map;
+ VG_NOACCESS_WRITE(&sb->_vg_ptr, vg_ptr);
+ VALGRIND_MEMPOOL_ALLOC(stream, vg_ptr, size);
+ } else {
+ void *state_end = state.map + state.alloc_size;
+ /* This only updates the mempool. The newly allocated chunk is still
+ * marked as NOACCESS. */
+ VALGRIND_MEMPOOL_CHANGE(stream, vg_ptr, vg_ptr, state_end - vg_ptr);
+ /* Mark the newly allocated chunk as undefined */
+ VALGRIND_MAKE_MEM_UNDEFINED(state.map, state.alloc_size);
+ }
+#endif
+
+ stream->next = state.offset + size;
+
+ return state;
+}
+
+struct bo_pool_bo_link {
+ struct bo_pool_bo_link *next;
+ struct anv_bo bo;
+};
+
+void
+anv_bo_pool_init(struct anv_bo_pool *pool,
+ struct anv_device *device, uint32_t bo_size)
+{
+ pool->device = device;
+ pool->bo_size = bo_size;
+ pool->free_list = NULL;
+
+ VG(VALGRIND_CREATE_MEMPOOL(pool, 0, false));
+}
+
+void
+anv_bo_pool_finish(struct anv_bo_pool *pool)
+{
+ struct bo_pool_bo_link *link = PFL_PTR(pool->free_list);
+ while (link != NULL) {
+ struct bo_pool_bo_link link_copy = VG_NOACCESS_READ(link);
+
+ anv_gem_munmap(link_copy.bo.map, pool->bo_size);
+ anv_gem_close(pool->device, link_copy.bo.gem_handle);
+ link = link_copy.next;
+ }
+
+ VG(VALGRIND_DESTROY_MEMPOOL(pool));
+}
+
+VkResult
+anv_bo_pool_alloc(struct anv_bo_pool *pool, struct anv_bo *bo)
+{
+ VkResult result;
+
+ void *next_free_void;
+ if (anv_ptr_free_list_pop(&pool->free_list, &next_free_void)) {
+ struct bo_pool_bo_link *next_free = next_free_void;
+ *bo = VG_NOACCESS_READ(&next_free->bo);
+ assert(bo->map == next_free);
+ assert(bo->size == pool->bo_size);
+
+ VG(VALGRIND_MEMPOOL_ALLOC(pool, bo->map, pool->bo_size));
+
+ return VK_SUCCESS;
+ }
+
+ struct anv_bo new_bo;
+
+ result = anv_bo_init_new(&new_bo, pool->device, pool->bo_size);
+ if (result != VK_SUCCESS)
+ return result;
+
+ assert(new_bo.size == pool->bo_size);
+
+ new_bo.map = anv_gem_mmap(pool->device, new_bo.gem_handle, 0, pool->bo_size, 0);
+ if (new_bo.map == NULL) {
+ anv_gem_close(pool->device, new_bo.gem_handle);
+ return vk_error(VK_ERROR_MEMORY_MAP_FAILED);
+ }
+
+ *bo = new_bo;
+
+ VG(VALGRIND_MEMPOOL_ALLOC(pool, bo->map, pool->bo_size));
+
+ return VK_SUCCESS;
+}
+
+void
+anv_bo_pool_free(struct anv_bo_pool *pool, const struct anv_bo *bo)
+{
+ struct bo_pool_bo_link *link = bo->map;
+ link->bo = *bo;
+
+ VG(VALGRIND_MEMPOOL_FREE(pool, bo->map));
+ anv_ptr_free_list_push(&pool->free_list, link);
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ * Authors:
+ * Eric Anholt <eric@anholt.net>
+ *
+ */
+
+/** @file intel_aub.h
+ *
+ * The AUB file is a file format used by Intel's internal simulation
+ * and other validation tools. It can be used at various levels by a
+ * driver to input state to the simulated hardware or a replaying
+ * debugger.
+ *
+ * We choose to dump AUB files using the trace block format for ease
+ * of implementation -- dump out the blocks of memory as plain blobs
+ * and insert ring commands to execute the batchbuffer blob.
+ */
+
+#ifndef _INTEL_AUB_H
+#define _INTEL_AUB_H
+
+#define AUB_MI_NOOP (0)
+#define AUB_MI_BATCH_BUFFER_START (0x31 << 23)
+#define AUB_PIPE_CONTROL (0x7a000002)
+
+/* DW0: instruction type. */
+
+#define CMD_AUB (7 << 29)
+
+#define CMD_AUB_HEADER (CMD_AUB | (1 << 23) | (0x05 << 16))
+/* DW1 */
+# define AUB_HEADER_MAJOR_SHIFT 24
+# define AUB_HEADER_MINOR_SHIFT 16
+
+#define CMD_AUB_TRACE_HEADER_BLOCK (CMD_AUB | (1 << 23) | (0x41 << 16))
+#define CMD_AUB_DUMP_BMP (CMD_AUB | (1 << 23) | (0x9e << 16))
+
+/* DW1 */
+#define AUB_TRACE_OPERATION_MASK 0x000000ff
+#define AUB_TRACE_OP_COMMENT 0x00000000
+#define AUB_TRACE_OP_DATA_WRITE 0x00000001
+#define AUB_TRACE_OP_COMMAND_WRITE 0x00000002
+#define AUB_TRACE_OP_MMIO_WRITE 0x00000003
+// operation = TRACE_DATA_WRITE, Type
+#define AUB_TRACE_TYPE_MASK 0x0000ff00
+#define AUB_TRACE_TYPE_NOTYPE (0 << 8)
+#define AUB_TRACE_TYPE_BATCH (1 << 8)
+#define AUB_TRACE_TYPE_VERTEX_BUFFER (5 << 8)
+#define AUB_TRACE_TYPE_2D_MAP (6 << 8)
+#define AUB_TRACE_TYPE_CUBE_MAP (7 << 8)
+#define AUB_TRACE_TYPE_VOLUME_MAP (9 << 8)
+#define AUB_TRACE_TYPE_1D_MAP (10 << 8)
+#define AUB_TRACE_TYPE_CONSTANT_BUFFER (11 << 8)
+#define AUB_TRACE_TYPE_CONSTANT_URB (12 << 8)
+#define AUB_TRACE_TYPE_INDEX_BUFFER (13 << 8)
+#define AUB_TRACE_TYPE_GENERAL (14 << 8)
+#define AUB_TRACE_TYPE_SURFACE (15 << 8)
+
+
+// operation = TRACE_COMMAND_WRITE, Type =
+#define AUB_TRACE_TYPE_RING_HWB (1 << 8)
+#define AUB_TRACE_TYPE_RING_PRB0 (2 << 8)
+#define AUB_TRACE_TYPE_RING_PRB1 (3 << 8)
+#define AUB_TRACE_TYPE_RING_PRB2 (4 << 8)
+
+// Address space
+#define AUB_TRACE_ADDRESS_SPACE_MASK 0x00ff0000
+#define AUB_TRACE_MEMTYPE_GTT (0 << 16)
+#define AUB_TRACE_MEMTYPE_LOCAL (1 << 16)
+#define AUB_TRACE_MEMTYPE_NONLOCAL (2 << 16)
+#define AUB_TRACE_MEMTYPE_PCI (3 << 16)
+#define AUB_TRACE_MEMTYPE_GTT_ENTRY (4 << 16)
+
+/* DW2 */
+
+/**
+ * aub_state_struct_type enum values are encoded with the top 16 bits
+ * representing the type to be delivered to the .aub file, and the bottom 16
+ * bits representing the subtype. This macro performs the encoding.
+ */
+#define ENCODE_SS_TYPE(type, subtype) (((type) << 16) | (subtype))
+
+enum aub_state_struct_type {
+ AUB_TRACE_VS_STATE = ENCODE_SS_TYPE(AUB_TRACE_TYPE_GENERAL, 1),
+ AUB_TRACE_GS_STATE = ENCODE_SS_TYPE(AUB_TRACE_TYPE_GENERAL, 2),
+ AUB_TRACE_CLIP_STATE = ENCODE_SS_TYPE(AUB_TRACE_TYPE_GENERAL, 3),
+ AUB_TRACE_SF_STATE = ENCODE_SS_TYPE(AUB_TRACE_TYPE_GENERAL, 4),
+ AUB_TRACE_WM_STATE = ENCODE_SS_TYPE(AUB_TRACE_TYPE_GENERAL, 5),
+ AUB_TRACE_CC_STATE = ENCODE_SS_TYPE(AUB_TRACE_TYPE_GENERAL, 6),
+ AUB_TRACE_CLIP_VP_STATE = ENCODE_SS_TYPE(AUB_TRACE_TYPE_GENERAL, 7),
+ AUB_TRACE_SF_VP_STATE = ENCODE_SS_TYPE(AUB_TRACE_TYPE_GENERAL, 8),
+ AUB_TRACE_CC_VP_STATE = ENCODE_SS_TYPE(AUB_TRACE_TYPE_GENERAL, 0x9),
+ AUB_TRACE_SAMPLER_STATE = ENCODE_SS_TYPE(AUB_TRACE_TYPE_GENERAL, 0xa),
+ AUB_TRACE_KERNEL_INSTRUCTIONS = ENCODE_SS_TYPE(AUB_TRACE_TYPE_GENERAL, 0xb),
+ AUB_TRACE_SCRATCH_SPACE = ENCODE_SS_TYPE(AUB_TRACE_TYPE_GENERAL, 0xc),
+ AUB_TRACE_SAMPLER_DEFAULT_COLOR = ENCODE_SS_TYPE(AUB_TRACE_TYPE_GENERAL, 0xd),
+
+ AUB_TRACE_SCISSOR_STATE = ENCODE_SS_TYPE(AUB_TRACE_TYPE_GENERAL, 0x15),
+ AUB_TRACE_BLEND_STATE = ENCODE_SS_TYPE(AUB_TRACE_TYPE_GENERAL, 0x16),
+ AUB_TRACE_DEPTH_STENCIL_STATE = ENCODE_SS_TYPE(AUB_TRACE_TYPE_GENERAL, 0x17),
+
+ AUB_TRACE_VERTEX_BUFFER = ENCODE_SS_TYPE(AUB_TRACE_TYPE_VERTEX_BUFFER, 0),
+ AUB_TRACE_BINDING_TABLE = ENCODE_SS_TYPE(AUB_TRACE_TYPE_SURFACE, 0x100),
+ AUB_TRACE_SURFACE_STATE = ENCODE_SS_TYPE(AUB_TRACE_TYPE_SURFACE, 0x200),
+ AUB_TRACE_VS_CONSTANTS = ENCODE_SS_TYPE(AUB_TRACE_TYPE_CONSTANT_BUFFER, 0),
+ AUB_TRACE_WM_CONSTANTS = ENCODE_SS_TYPE(AUB_TRACE_TYPE_CONSTANT_BUFFER, 1),
+};
+
+#undef ENCODE_SS_TYPE
+
+/**
+ * Decode a aub_state_struct_type value to determine the type that should be
+ * stored in the .aub file.
+ */
+static inline uint32_t AUB_TRACE_TYPE(enum aub_state_struct_type ss_type)
+{
+ return (ss_type & 0xFFFF0000) >> 16;
+}
+
+/**
+ * Decode a state_struct_type value to determine the subtype that should be
+ * stored in the .aub file.
+ */
+static inline uint32_t AUB_TRACE_SUBTYPE(enum aub_state_struct_type ss_type)
+{
+ return ss_type & 0xFFFF;
+}
+
+/* DW3: address */
+/* DW4: len */
+
+#endif /* _INTEL_AUB_H */
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <assert.h>
+#include <stdbool.h>
+#include <string.h>
+#include <unistd.h>
+#include <fcntl.h>
+
+#include "anv_private.h"
+
+#include "gen7_pack.h"
+#include "gen8_pack.h"
+
+/** \file anv_batch_chain.c
+ *
+ * This file contains functions related to anv_cmd_buffer as a data
+ * structure. This involves everything required to create and destroy
+ * the actual batch buffers as well as link them together and handle
+ * relocations and surface state. It specifically does *not* contain any
+ * handling of actual vkCmd calls beyond vkCmdExecuteCommands.
+ */
+
+/*-----------------------------------------------------------------------*
+ * Functions related to anv_reloc_list
+ *-----------------------------------------------------------------------*/
+
+static VkResult
+anv_reloc_list_init_clone(struct anv_reloc_list *list,
+ const VkAllocationCallbacks *alloc,
+ const struct anv_reloc_list *other_list)
+{
+ if (other_list) {
+ list->num_relocs = other_list->num_relocs;
+ list->array_length = other_list->array_length;
+ } else {
+ list->num_relocs = 0;
+ list->array_length = 256;
+ }
+
+ list->relocs =
+ anv_alloc(alloc, list->array_length * sizeof(*list->relocs), 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+
+ if (list->relocs == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ list->reloc_bos =
+ anv_alloc(alloc, list->array_length * sizeof(*list->reloc_bos), 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+
+ if (list->reloc_bos == NULL) {
+ anv_free(alloc, list->relocs);
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+ }
+
+ if (other_list) {
+ memcpy(list->relocs, other_list->relocs,
+ list->array_length * sizeof(*list->relocs));
+ memcpy(list->reloc_bos, other_list->reloc_bos,
+ list->array_length * sizeof(*list->reloc_bos));
+ }
+
+ return VK_SUCCESS;
+}
+
+VkResult
+anv_reloc_list_init(struct anv_reloc_list *list,
+ const VkAllocationCallbacks *alloc)
+{
+ return anv_reloc_list_init_clone(list, alloc, NULL);
+}
+
+void
+anv_reloc_list_finish(struct anv_reloc_list *list,
+ const VkAllocationCallbacks *alloc)
+{
+ anv_free(alloc, list->relocs);
+ anv_free(alloc, list->reloc_bos);
+}
+
+static VkResult
+anv_reloc_list_grow(struct anv_reloc_list *list,
+ const VkAllocationCallbacks *alloc,
+ size_t num_additional_relocs)
+{
+ if (list->num_relocs + num_additional_relocs <= list->array_length)
+ return VK_SUCCESS;
+
+ size_t new_length = list->array_length * 2;
+ while (new_length < list->num_relocs + num_additional_relocs)
+ new_length *= 2;
+
+ struct drm_i915_gem_relocation_entry *new_relocs =
+ anv_alloc(alloc, new_length * sizeof(*list->relocs), 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (new_relocs == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ struct anv_bo **new_reloc_bos =
+ anv_alloc(alloc, new_length * sizeof(*list->reloc_bos), 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (new_relocs == NULL) {
+ anv_free(alloc, new_relocs);
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+ }
+
+ memcpy(new_relocs, list->relocs, list->num_relocs * sizeof(*list->relocs));
+ memcpy(new_reloc_bos, list->reloc_bos,
+ list->num_relocs * sizeof(*list->reloc_bos));
+
+ anv_free(alloc, list->relocs);
+ anv_free(alloc, list->reloc_bos);
+
+ list->array_length = new_length;
+ list->relocs = new_relocs;
+ list->reloc_bos = new_reloc_bos;
+
+ return VK_SUCCESS;
+}
+
+uint64_t
+anv_reloc_list_add(struct anv_reloc_list *list,
+ const VkAllocationCallbacks *alloc,
+ uint32_t offset, struct anv_bo *target_bo, uint32_t delta)
+{
+ struct drm_i915_gem_relocation_entry *entry;
+ int index;
+
+ anv_reloc_list_grow(list, alloc, 1);
+ /* TODO: Handle failure */
+
+ /* XXX: Can we use I915_EXEC_HANDLE_LUT? */
+ index = list->num_relocs++;
+ list->reloc_bos[index] = target_bo;
+ entry = &list->relocs[index];
+ entry->target_handle = target_bo->gem_handle;
+ entry->delta = delta;
+ entry->offset = offset;
+ entry->presumed_offset = target_bo->offset;
+ entry->read_domains = 0;
+ entry->write_domain = 0;
+ VG(VALGRIND_CHECK_MEM_IS_DEFINED(entry, sizeof(*entry)));
+
+ return target_bo->offset + delta;
+}
+
+static void
+anv_reloc_list_append(struct anv_reloc_list *list,
+ const VkAllocationCallbacks *alloc,
+ struct anv_reloc_list *other, uint32_t offset)
+{
+ anv_reloc_list_grow(list, alloc, other->num_relocs);
+ /* TODO: Handle failure */
+
+ memcpy(&list->relocs[list->num_relocs], &other->relocs[0],
+ other->num_relocs * sizeof(other->relocs[0]));
+ memcpy(&list->reloc_bos[list->num_relocs], &other->reloc_bos[0],
+ other->num_relocs * sizeof(other->reloc_bos[0]));
+
+ for (uint32_t i = 0; i < other->num_relocs; i++)
+ list->relocs[i + list->num_relocs].offset += offset;
+
+ list->num_relocs += other->num_relocs;
+}
+
+/*-----------------------------------------------------------------------*
+ * Functions related to anv_batch
+ *-----------------------------------------------------------------------*/
+
+void *
+anv_batch_emit_dwords(struct anv_batch *batch, int num_dwords)
+{
+ if (batch->next + num_dwords * 4 > batch->end)
+ batch->extend_cb(batch, batch->user_data);
+
+ void *p = batch->next;
+
+ batch->next += num_dwords * 4;
+ assert(batch->next <= batch->end);
+
+ return p;
+}
+
+uint64_t
+anv_batch_emit_reloc(struct anv_batch *batch,
+ void *location, struct anv_bo *bo, uint32_t delta)
+{
+ return anv_reloc_list_add(batch->relocs, batch->alloc,
+ location - batch->start, bo, delta);
+}
+
+void
+anv_batch_emit_batch(struct anv_batch *batch, struct anv_batch *other)
+{
+ uint32_t size, offset;
+
+ size = other->next - other->start;
+ assert(size % 4 == 0);
+
+ if (batch->next + size > batch->end)
+ batch->extend_cb(batch, batch->user_data);
+
+ assert(batch->next + size <= batch->end);
+
+ VG(VALGRIND_CHECK_MEM_IS_DEFINED(other->start, size));
+ memcpy(batch->next, other->start, size);
+
+ offset = batch->next - batch->start;
+ anv_reloc_list_append(batch->relocs, batch->alloc,
+ other->relocs, offset);
+
+ batch->next += size;
+}
+
+/*-----------------------------------------------------------------------*
+ * Functions related to anv_batch_bo
+ *-----------------------------------------------------------------------*/
+
+static VkResult
+anv_batch_bo_create(struct anv_cmd_buffer *cmd_buffer,
+ struct anv_batch_bo **bbo_out)
+{
+ VkResult result;
+
+ struct anv_batch_bo *bbo = anv_alloc(&cmd_buffer->pool->alloc, sizeof(*bbo),
+ 8, VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (bbo == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ result = anv_bo_pool_alloc(&cmd_buffer->device->batch_bo_pool, &bbo->bo);
+ if (result != VK_SUCCESS)
+ goto fail_alloc;
+
+ result = anv_reloc_list_init(&bbo->relocs, &cmd_buffer->pool->alloc);
+ if (result != VK_SUCCESS)
+ goto fail_bo_alloc;
+
+ *bbo_out = bbo;
+
+ return VK_SUCCESS;
+
+ fail_bo_alloc:
+ anv_bo_pool_free(&cmd_buffer->device->batch_bo_pool, &bbo->bo);
+ fail_alloc:
+ anv_free(&cmd_buffer->pool->alloc, bbo);
+
+ return result;
+}
+
+static VkResult
+anv_batch_bo_clone(struct anv_cmd_buffer *cmd_buffer,
+ const struct anv_batch_bo *other_bbo,
+ struct anv_batch_bo **bbo_out)
+{
+ VkResult result;
+
+ struct anv_batch_bo *bbo = anv_alloc(&cmd_buffer->pool->alloc, sizeof(*bbo),
+ 8, VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (bbo == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ result = anv_bo_pool_alloc(&cmd_buffer->device->batch_bo_pool, &bbo->bo);
+ if (result != VK_SUCCESS)
+ goto fail_alloc;
+
+ result = anv_reloc_list_init_clone(&bbo->relocs, &cmd_buffer->pool->alloc,
+ &other_bbo->relocs);
+ if (result != VK_SUCCESS)
+ goto fail_bo_alloc;
+
+ bbo->length = other_bbo->length;
+ memcpy(bbo->bo.map, other_bbo->bo.map, other_bbo->length);
+
+ bbo->last_ss_pool_bo_offset = other_bbo->last_ss_pool_bo_offset;
+
+ *bbo_out = bbo;
+
+ return VK_SUCCESS;
+
+ fail_bo_alloc:
+ anv_bo_pool_free(&cmd_buffer->device->batch_bo_pool, &bbo->bo);
+ fail_alloc:
+ anv_free(&cmd_buffer->pool->alloc, bbo);
+
+ return result;
+}
+
+static void
+anv_batch_bo_start(struct anv_batch_bo *bbo, struct anv_batch *batch,
+ size_t batch_padding)
+{
+ batch->next = batch->start = bbo->bo.map;
+ batch->end = bbo->bo.map + bbo->bo.size - batch_padding;
+ batch->relocs = &bbo->relocs;
+ bbo->last_ss_pool_bo_offset = 0;
+ bbo->relocs.num_relocs = 0;
+}
+
+static void
+anv_batch_bo_continue(struct anv_batch_bo *bbo, struct anv_batch *batch,
+ size_t batch_padding)
+{
+ batch->start = bbo->bo.map;
+ batch->next = bbo->bo.map + bbo->length;
+ batch->end = bbo->bo.map + bbo->bo.size - batch_padding;
+ batch->relocs = &bbo->relocs;
+}
+
+static void
+anv_batch_bo_finish(struct anv_batch_bo *bbo, struct anv_batch *batch)
+{
+ assert(batch->start == bbo->bo.map);
+ bbo->length = batch->next - batch->start;
+ VG(VALGRIND_CHECK_MEM_IS_DEFINED(batch->start, bbo->length));
+}
+
+static void
+anv_batch_bo_destroy(struct anv_batch_bo *bbo,
+ struct anv_cmd_buffer *cmd_buffer)
+{
+ anv_reloc_list_finish(&bbo->relocs, &cmd_buffer->pool->alloc);
+ anv_bo_pool_free(&cmd_buffer->device->batch_bo_pool, &bbo->bo);
+ anv_free(&cmd_buffer->pool->alloc, bbo);
+}
+
+static VkResult
+anv_batch_bo_list_clone(const struct list_head *list,
+ struct anv_cmd_buffer *cmd_buffer,
+ struct list_head *new_list)
+{
+ VkResult result = VK_SUCCESS;
+
+ list_inithead(new_list);
+
+ struct anv_batch_bo *prev_bbo = NULL;
+ list_for_each_entry(struct anv_batch_bo, bbo, list, link) {
+ struct anv_batch_bo *new_bbo;
+ result = anv_batch_bo_clone(cmd_buffer, bbo, &new_bbo);
+ if (result != VK_SUCCESS)
+ break;
+ list_addtail(&new_bbo->link, new_list);
+
+ if (prev_bbo) {
+ /* As we clone this list of batch_bo's, they chain one to the
+ * other using MI_BATCH_BUFFER_START commands. We need to fix up
+ * those relocations as we go. Fortunately, this is pretty easy
+ * as it will always be the last relocation in the list.
+ */
+ uint32_t last_idx = prev_bbo->relocs.num_relocs - 1;
+ assert(prev_bbo->relocs.reloc_bos[last_idx] == &bbo->bo);
+ prev_bbo->relocs.reloc_bos[last_idx] = &new_bbo->bo;
+ }
+
+ prev_bbo = new_bbo;
+ }
+
+ if (result != VK_SUCCESS) {
+ list_for_each_entry_safe(struct anv_batch_bo, bbo, new_list, link)
+ anv_batch_bo_destroy(bbo, cmd_buffer);
+ }
+
+ return result;
+}
+
+/*-----------------------------------------------------------------------*
+ * Functions related to anv_batch_bo
+ *-----------------------------------------------------------------------*/
+
+static inline struct anv_batch_bo *
+anv_cmd_buffer_current_batch_bo(struct anv_cmd_buffer *cmd_buffer)
+{
+ return LIST_ENTRY(struct anv_batch_bo, cmd_buffer->batch_bos.prev, link);
+}
+
+struct anv_address
+anv_cmd_buffer_surface_base_address(struct anv_cmd_buffer *cmd_buffer)
+{
+ return (struct anv_address) {
+ .bo = &cmd_buffer->device->surface_state_block_pool.bo,
+ .offset = *(int32_t *)anv_vector_head(&cmd_buffer->bt_blocks),
+ };
+}
+
+static void
+emit_batch_buffer_start(struct anv_cmd_buffer *cmd_buffer,
+ struct anv_bo *bo, uint32_t offset)
+{
+ /* In gen8+ the address field grew to two dwords to accomodate 48 bit
+ * offsets. The high 16 bits are in the last dword, so we can use the gen8
+ * version in either case, as long as we set the instruction length in the
+ * header accordingly. This means that we always emit three dwords here
+ * and all the padding and adjustment we do in this file works for all
+ * gens.
+ */
+
+ const uint32_t gen7_length =
+ GEN7_MI_BATCH_BUFFER_START_length - GEN7_MI_BATCH_BUFFER_START_length_bias;
+ const uint32_t gen8_length =
+ GEN8_MI_BATCH_BUFFER_START_length - GEN8_MI_BATCH_BUFFER_START_length_bias;
+
+ anv_batch_emit(&cmd_buffer->batch, GEN8_MI_BATCH_BUFFER_START,
+ .DwordLength = cmd_buffer->device->info.gen < 8 ?
+ gen7_length : gen8_length,
+ ._2ndLevelBatchBuffer = _1stlevelbatch,
+ .AddressSpaceIndicator = ASI_PPGTT,
+ .BatchBufferStartAddress = { bo, offset });
+}
+
+static void
+cmd_buffer_chain_to_batch_bo(struct anv_cmd_buffer *cmd_buffer,
+ struct anv_batch_bo *bbo)
+{
+ struct anv_batch *batch = &cmd_buffer->batch;
+ struct anv_batch_bo *current_bbo =
+ anv_cmd_buffer_current_batch_bo(cmd_buffer);
+
+ /* We set the end of the batch a little short so we would be sure we
+ * have room for the chaining command. Since we're about to emit the
+ * chaining command, let's set it back where it should go.
+ */
+ batch->end += GEN8_MI_BATCH_BUFFER_START_length * 4;
+ assert(batch->end == current_bbo->bo.map + current_bbo->bo.size);
+
+ emit_batch_buffer_start(cmd_buffer, &bbo->bo, 0);
+
+ anv_batch_bo_finish(current_bbo, batch);
+}
+
+static VkResult
+anv_cmd_buffer_chain_batch(struct anv_batch *batch, void *_data)
+{
+ struct anv_cmd_buffer *cmd_buffer = _data;
+ struct anv_batch_bo *new_bbo;
+
+ VkResult result = anv_batch_bo_create(cmd_buffer, &new_bbo);
+ if (result != VK_SUCCESS)
+ return result;
+
+ struct anv_batch_bo **seen_bbo = anv_vector_add(&cmd_buffer->seen_bbos);
+ if (seen_bbo == NULL) {
+ anv_batch_bo_destroy(new_bbo, cmd_buffer);
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+ }
+ *seen_bbo = new_bbo;
+
+ cmd_buffer_chain_to_batch_bo(cmd_buffer, new_bbo);
+
+ list_addtail(&new_bbo->link, &cmd_buffer->batch_bos);
+
+ anv_batch_bo_start(new_bbo, batch, GEN8_MI_BATCH_BUFFER_START_length * 4);
+
+ return VK_SUCCESS;
+}
+
+struct anv_state
+anv_cmd_buffer_alloc_binding_table(struct anv_cmd_buffer *cmd_buffer,
+ uint32_t entries, uint32_t *state_offset)
+{
+ struct anv_block_pool *block_pool =
+ &cmd_buffer->device->surface_state_block_pool;
+ int32_t *bt_block = anv_vector_head(&cmd_buffer->bt_blocks);
+ struct anv_state state;
+
+ state.alloc_size = align_u32(entries * 4, 32);
+
+ if (cmd_buffer->bt_next + state.alloc_size > block_pool->block_size)
+ return (struct anv_state) { 0 };
+
+ state.offset = cmd_buffer->bt_next;
+ state.map = block_pool->map + *bt_block + state.offset;
+
+ cmd_buffer->bt_next += state.alloc_size;
+
+ assert(*bt_block < 0);
+ *state_offset = -(*bt_block);
+
+ return state;
+}
+
+struct anv_state
+anv_cmd_buffer_alloc_surface_state(struct anv_cmd_buffer *cmd_buffer)
+{
+ return anv_state_stream_alloc(&cmd_buffer->surface_state_stream, 64, 64);
+}
+
+struct anv_state
+anv_cmd_buffer_alloc_dynamic_state(struct anv_cmd_buffer *cmd_buffer,
+ uint32_t size, uint32_t alignment)
+{
+ return anv_state_stream_alloc(&cmd_buffer->dynamic_state_stream,
+ size, alignment);
+}
+
+VkResult
+anv_cmd_buffer_new_binding_table_block(struct anv_cmd_buffer *cmd_buffer)
+{
+ struct anv_block_pool *block_pool =
+ &cmd_buffer->device->surface_state_block_pool;
+
+ int32_t *offset = anv_vector_add(&cmd_buffer->bt_blocks);
+ if (offset == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ *offset = anv_block_pool_alloc_back(block_pool);
+ cmd_buffer->bt_next = 0;
+
+ return VK_SUCCESS;
+}
+
+VkResult
+anv_cmd_buffer_init_batch_bo_chain(struct anv_cmd_buffer *cmd_buffer)
+{
+ struct anv_batch_bo *batch_bo;
+ VkResult result;
+
+ list_inithead(&cmd_buffer->batch_bos);
+
+ result = anv_batch_bo_create(cmd_buffer, &batch_bo);
+ if (result != VK_SUCCESS)
+ return result;
+
+ list_addtail(&batch_bo->link, &cmd_buffer->batch_bos);
+
+ cmd_buffer->batch.alloc = &cmd_buffer->pool->alloc;
+ cmd_buffer->batch.extend_cb = anv_cmd_buffer_chain_batch;
+ cmd_buffer->batch.user_data = cmd_buffer;
+
+ anv_batch_bo_start(batch_bo, &cmd_buffer->batch,
+ GEN8_MI_BATCH_BUFFER_START_length * 4);
+
+ int success = anv_vector_init(&cmd_buffer->seen_bbos,
+ sizeof(struct anv_bo *),
+ 8 * sizeof(struct anv_bo *));
+ if (!success)
+ goto fail_batch_bo;
+
+ *(struct anv_batch_bo **)anv_vector_add(&cmd_buffer->seen_bbos) = batch_bo;
+
+ success = anv_vector_init(&cmd_buffer->bt_blocks, sizeof(int32_t),
+ 8 * sizeof(int32_t));
+ if (!success)
+ goto fail_seen_bbos;
+
+ result = anv_reloc_list_init(&cmd_buffer->surface_relocs,
+ &cmd_buffer->pool->alloc);
+ if (result != VK_SUCCESS)
+ goto fail_bt_blocks;
+
+ anv_cmd_buffer_new_binding_table_block(cmd_buffer);
+
+ cmd_buffer->execbuf2.objects = NULL;
+ cmd_buffer->execbuf2.bos = NULL;
+ cmd_buffer->execbuf2.array_length = 0;
+
+ return VK_SUCCESS;
+
+ fail_bt_blocks:
+ anv_vector_finish(&cmd_buffer->bt_blocks);
+ fail_seen_bbos:
+ anv_vector_finish(&cmd_buffer->seen_bbos);
+ fail_batch_bo:
+ anv_batch_bo_destroy(batch_bo, cmd_buffer);
+
+ return result;
+}
+
+void
+anv_cmd_buffer_fini_batch_bo_chain(struct anv_cmd_buffer *cmd_buffer)
+{
+ int32_t *bt_block;
+ anv_vector_foreach(bt_block, &cmd_buffer->bt_blocks) {
+ anv_block_pool_free(&cmd_buffer->device->surface_state_block_pool,
+ *bt_block);
+ }
+ anv_vector_finish(&cmd_buffer->bt_blocks);
+
+ anv_reloc_list_finish(&cmd_buffer->surface_relocs, &cmd_buffer->pool->alloc);
+
+ anv_vector_finish(&cmd_buffer->seen_bbos);
+
+ /* Destroy all of the batch buffers */
+ list_for_each_entry_safe(struct anv_batch_bo, bbo,
+ &cmd_buffer->batch_bos, link) {
+ anv_batch_bo_destroy(bbo, cmd_buffer);
+ }
+
+ anv_free(&cmd_buffer->pool->alloc, cmd_buffer->execbuf2.objects);
+ anv_free(&cmd_buffer->pool->alloc, cmd_buffer->execbuf2.bos);
+}
+
+void
+anv_cmd_buffer_reset_batch_bo_chain(struct anv_cmd_buffer *cmd_buffer)
+{
+ /* Delete all but the first batch bo */
+ assert(!list_empty(&cmd_buffer->batch_bos));
+ while (cmd_buffer->batch_bos.next != cmd_buffer->batch_bos.prev) {
+ struct anv_batch_bo *bbo = anv_cmd_buffer_current_batch_bo(cmd_buffer);
+ list_del(&bbo->link);
+ anv_batch_bo_destroy(bbo, cmd_buffer);
+ }
+ assert(!list_empty(&cmd_buffer->batch_bos));
+
+ anv_batch_bo_start(anv_cmd_buffer_current_batch_bo(cmd_buffer),
+ &cmd_buffer->batch,
+ GEN8_MI_BATCH_BUFFER_START_length * 4);
+
+ while (anv_vector_length(&cmd_buffer->bt_blocks) > 1) {
+ int32_t *bt_block = anv_vector_remove(&cmd_buffer->bt_blocks);
+ anv_block_pool_free(&cmd_buffer->device->surface_state_block_pool,
+ *bt_block);
+ }
+ assert(anv_vector_length(&cmd_buffer->bt_blocks) == 1);
+ cmd_buffer->bt_next = 0;
+
+ cmd_buffer->surface_relocs.num_relocs = 0;
+
+ /* Reset the list of seen buffers */
+ cmd_buffer->seen_bbos.head = 0;
+ cmd_buffer->seen_bbos.tail = 0;
+
+ *(struct anv_batch_bo **)anv_vector_add(&cmd_buffer->seen_bbos) =
+ anv_cmd_buffer_current_batch_bo(cmd_buffer);
+}
+
+void
+anv_cmd_buffer_end_batch_buffer(struct anv_cmd_buffer *cmd_buffer)
+{
+ struct anv_batch_bo *batch_bo = anv_cmd_buffer_current_batch_bo(cmd_buffer);
+
+ if (cmd_buffer->level == VK_COMMAND_BUFFER_LEVEL_PRIMARY) {
+ /* When we start a batch buffer, we subtract a certain amount of
+ * padding from the end to ensure that we always have room to emit a
+ * BATCH_BUFFER_START to chain to the next BO. We need to remove
+ * that padding before we end the batch; otherwise, we may end up
+ * with our BATCH_BUFFER_END in another BO.
+ */
+ cmd_buffer->batch.end += GEN8_MI_BATCH_BUFFER_START_length * 4;
+ assert(cmd_buffer->batch.end == batch_bo->bo.map + batch_bo->bo.size);
+
+ anv_batch_emit(&cmd_buffer->batch, GEN7_MI_BATCH_BUFFER_END);
+
+ /* Round batch up to an even number of dwords. */
+ if ((cmd_buffer->batch.next - cmd_buffer->batch.start) & 4)
+ anv_batch_emit(&cmd_buffer->batch, GEN7_MI_NOOP);
+
+ cmd_buffer->exec_mode = ANV_CMD_BUFFER_EXEC_MODE_PRIMARY;
+ }
+
+ anv_batch_bo_finish(batch_bo, &cmd_buffer->batch);
+
+ if (cmd_buffer->level == VK_COMMAND_BUFFER_LEVEL_SECONDARY) {
+ /* If this is a secondary command buffer, we need to determine the
+ * mode in which it will be executed with vkExecuteCommands. We
+ * determine this statically here so that this stays in sync with the
+ * actual ExecuteCommands implementation.
+ */
+ if ((cmd_buffer->batch_bos.next == cmd_buffer->batch_bos.prev) &&
+ (batch_bo->length < ANV_CMD_BUFFER_BATCH_SIZE / 2)) {
+ /* If the secondary has exactly one batch buffer in its list *and*
+ * that batch buffer is less than half of the maximum size, we're
+ * probably better of simply copying it into our batch.
+ */
+ cmd_buffer->exec_mode = ANV_CMD_BUFFER_EXEC_MODE_EMIT;
+ } else if (!(cmd_buffer->usage_flags &
+ VK_COMMAND_BUFFER_USAGE_SIMULTANEOUS_USE_BIT)) {
+ cmd_buffer->exec_mode = ANV_CMD_BUFFER_EXEC_MODE_CHAIN;
+
+ /* When we chain, we need to add an MI_BATCH_BUFFER_START command
+ * with its relocation. In order to handle this we'll increment here
+ * so we can unconditionally decrement right before adding the
+ * MI_BATCH_BUFFER_START command.
+ */
+ batch_bo->relocs.num_relocs++;
+ cmd_buffer->batch.next += GEN8_MI_BATCH_BUFFER_START_length * 4;
+ } else {
+ cmd_buffer->exec_mode = ANV_CMD_BUFFER_EXEC_MODE_COPY_AND_CHAIN;
+ }
+ }
+}
+
+static inline VkResult
+anv_cmd_buffer_add_seen_bbos(struct anv_cmd_buffer *cmd_buffer,
+ struct list_head *list)
+{
+ list_for_each_entry(struct anv_batch_bo, bbo, list, link) {
+ struct anv_batch_bo **bbo_ptr = anv_vector_add(&cmd_buffer->seen_bbos);
+ if (bbo_ptr == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ *bbo_ptr = bbo;
+ }
+
+ return VK_SUCCESS;
+}
+
+void
+anv_cmd_buffer_add_secondary(struct anv_cmd_buffer *primary,
+ struct anv_cmd_buffer *secondary)
+{
+ switch (secondary->exec_mode) {
+ case ANV_CMD_BUFFER_EXEC_MODE_EMIT:
+ anv_batch_emit_batch(&primary->batch, &secondary->batch);
+ break;
+ case ANV_CMD_BUFFER_EXEC_MODE_CHAIN: {
+ struct anv_batch_bo *first_bbo =
+ list_first_entry(&secondary->batch_bos, struct anv_batch_bo, link);
+ struct anv_batch_bo *last_bbo =
+ list_last_entry(&secondary->batch_bos, struct anv_batch_bo, link);
+
+ emit_batch_buffer_start(primary, &first_bbo->bo, 0);
+
+ struct anv_batch_bo *this_bbo = anv_cmd_buffer_current_batch_bo(primary);
+ assert(primary->batch.start == this_bbo->bo.map);
+ uint32_t offset = primary->batch.next - primary->batch.start;
+ const uint32_t inst_size = GEN8_MI_BATCH_BUFFER_START_length * 4;
+
+ /* Roll back the previous MI_BATCH_BUFFER_START and its relocation so we
+ * can emit a new command and relocation for the current splice. In
+ * order to handle the initial-use case, we incremented next and
+ * num_relocs in end_batch_buffer() so we can alyways just subtract
+ * here.
+ */
+ last_bbo->relocs.num_relocs--;
+ secondary->batch.next -= inst_size;
+ emit_batch_buffer_start(secondary, &this_bbo->bo, offset);
+ anv_cmd_buffer_add_seen_bbos(primary, &secondary->batch_bos);
+
+ /* After patching up the secondary buffer, we need to clflush the
+ * modified instruction in case we're on a !llc platform. We use a
+ * little loop to handle the case where the instruction crosses a cache
+ * line boundary.
+ */
+ if (!primary->device->info.has_llc) {
+ void *inst = secondary->batch.next - inst_size;
+ void *p = (void *) (((uintptr_t) inst) & ~CACHELINE_MASK);
+ __builtin_ia32_sfence();
+ while (p < secondary->batch.next) {
+ __builtin_ia32_clflush(p);
+ p += CACHELINE_SIZE;
+ }
+ }
+
+ break;
+ }
+ case ANV_CMD_BUFFER_EXEC_MODE_COPY_AND_CHAIN: {
+ struct list_head copy_list;
+ VkResult result = anv_batch_bo_list_clone(&secondary->batch_bos,
+ secondary,
+ ©_list);
+ if (result != VK_SUCCESS)
+ return; /* FIXME */
+
+ anv_cmd_buffer_add_seen_bbos(primary, ©_list);
+
+ struct anv_batch_bo *first_bbo =
+ list_first_entry(©_list, struct anv_batch_bo, link);
+ struct anv_batch_bo *last_bbo =
+ list_last_entry(©_list, struct anv_batch_bo, link);
+
+ cmd_buffer_chain_to_batch_bo(primary, first_bbo);
+
+ list_splicetail(©_list, &primary->batch_bos);
+
+ anv_batch_bo_continue(last_bbo, &primary->batch,
+ GEN8_MI_BATCH_BUFFER_START_length * 4);
+
+ anv_cmd_buffer_emit_state_base_address(primary);
+ break;
+ }
+ default:
+ assert(!"Invalid execution mode");
+ }
+
+ anv_reloc_list_append(&primary->surface_relocs, &primary->pool->alloc,
+ &secondary->surface_relocs, 0);
+}
+
+static VkResult
+anv_cmd_buffer_add_bo(struct anv_cmd_buffer *cmd_buffer,
+ struct anv_bo *bo,
+ struct anv_reloc_list *relocs)
+{
+ struct drm_i915_gem_exec_object2 *obj = NULL;
+
+ if (bo->index < cmd_buffer->execbuf2.bo_count &&
+ cmd_buffer->execbuf2.bos[bo->index] == bo)
+ obj = &cmd_buffer->execbuf2.objects[bo->index];
+
+ if (obj == NULL) {
+ /* We've never seen this one before. Add it to the list and assign
+ * an id that we can use later.
+ */
+ if (cmd_buffer->execbuf2.bo_count >= cmd_buffer->execbuf2.array_length) {
+ uint32_t new_len = cmd_buffer->execbuf2.objects ?
+ cmd_buffer->execbuf2.array_length * 2 : 64;
+
+ struct drm_i915_gem_exec_object2 *new_objects =
+ anv_alloc(&cmd_buffer->pool->alloc, new_len * sizeof(*new_objects),
+ 8, VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (new_objects == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ struct anv_bo **new_bos =
+ anv_alloc(&cmd_buffer->pool->alloc, new_len * sizeof(*new_bos),
+ 8, VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (new_objects == NULL) {
+ anv_free(&cmd_buffer->pool->alloc, new_objects);
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+ }
+
+ if (cmd_buffer->execbuf2.objects) {
+ memcpy(new_objects, cmd_buffer->execbuf2.objects,
+ cmd_buffer->execbuf2.bo_count * sizeof(*new_objects));
+ memcpy(new_bos, cmd_buffer->execbuf2.bos,
+ cmd_buffer->execbuf2.bo_count * sizeof(*new_bos));
+ }
+
+ cmd_buffer->execbuf2.objects = new_objects;
+ cmd_buffer->execbuf2.bos = new_bos;
+ cmd_buffer->execbuf2.array_length = new_len;
+ }
+
+ assert(cmd_buffer->execbuf2.bo_count < cmd_buffer->execbuf2.array_length);
+
+ bo->index = cmd_buffer->execbuf2.bo_count++;
+ obj = &cmd_buffer->execbuf2.objects[bo->index];
+ cmd_buffer->execbuf2.bos[bo->index] = bo;
+
+ obj->handle = bo->gem_handle;
+ obj->relocation_count = 0;
+ obj->relocs_ptr = 0;
+ obj->alignment = 0;
+ obj->offset = bo->offset;
+ obj->flags = 0;
+ obj->rsvd1 = 0;
+ obj->rsvd2 = 0;
+ }
+
+ if (relocs != NULL && obj->relocation_count == 0) {
+ /* This is the first time we've ever seen a list of relocations for
+ * this BO. Go ahead and set the relocations and then walk the list
+ * of relocations and add them all.
+ */
+ obj->relocation_count = relocs->num_relocs;
+ obj->relocs_ptr = (uintptr_t) relocs->relocs;
+
+ for (size_t i = 0; i < relocs->num_relocs; i++) {
+ /* A quick sanity check on relocations */
+ assert(relocs->relocs[i].offset < bo->size);
+ anv_cmd_buffer_add_bo(cmd_buffer, relocs->reloc_bos[i], NULL);
+ }
+ }
+
+ return VK_SUCCESS;
+}
+
+static void
+anv_cmd_buffer_process_relocs(struct anv_cmd_buffer *cmd_buffer,
+ struct anv_reloc_list *list)
+{
+ struct anv_bo *bo;
+
+ /* If the kernel supports I915_EXEC_NO_RELOC, it will compare offset in
+ * struct drm_i915_gem_exec_object2 against the bos current offset and if
+ * all bos haven't moved it will skip relocation processing alltogether.
+ * If I915_EXEC_NO_RELOC is not supported, the kernel ignores the incoming
+ * value of offset so we can set it either way. For that to work we need
+ * to make sure all relocs use the same presumed offset.
+ */
+
+ for (size_t i = 0; i < list->num_relocs; i++) {
+ bo = list->reloc_bos[i];
+ if (bo->offset != list->relocs[i].presumed_offset)
+ cmd_buffer->execbuf2.need_reloc = true;
+
+ list->relocs[i].target_handle = bo->index;
+ }
+}
+
+static void
+adjust_relocations_from_block_pool(struct anv_block_pool *pool,
+ struct anv_reloc_list *relocs)
+{
+ for (size_t i = 0; i < relocs->num_relocs; i++) {
+ /* In general, we don't know how stale the relocated value is. It
+ * may have been used last time or it may not. Since we don't want
+ * to stomp it while the GPU may be accessing it, we haven't updated
+ * it anywhere else in the code. Instead, we just set the presumed
+ * offset to what it is now based on the delta and the data in the
+ * block pool. Then the kernel will update it for us if needed.
+ */
+ assert(relocs->relocs[i].offset < pool->state.end);
+ uint32_t *reloc_data = pool->map + relocs->relocs[i].offset;
+
+ /* We're reading back the relocated value from potentially incoherent
+ * memory here. However, any change to the value will be from the kernel
+ * writing out relocations, which will keep the CPU cache up to date.
+ */
+ relocs->relocs[i].presumed_offset = *reloc_data - relocs->relocs[i].delta;
+
+ /* All of the relocations from this block pool to other BO's should
+ * have been emitted relative to the surface block pool center. We
+ * need to add the center offset to make them relative to the
+ * beginning of the actual GEM bo.
+ */
+ relocs->relocs[i].offset += pool->center_bo_offset;
+ }
+}
+
+static void
+adjust_relocations_to_block_pool(struct anv_block_pool *pool,
+ struct anv_bo *from_bo,
+ struct anv_reloc_list *relocs,
+ uint32_t *last_pool_center_bo_offset)
+{
+ assert(*last_pool_center_bo_offset <= pool->center_bo_offset);
+ uint32_t delta = pool->center_bo_offset - *last_pool_center_bo_offset;
+
+ /* When we initially emit relocations into a block pool, we don't
+ * actually know what the final center_bo_offset will be so we just emit
+ * it as if center_bo_offset == 0. Now that we know what the center
+ * offset is, we need to walk the list of relocations and adjust any
+ * relocations that point to the pool bo with the correct offset.
+ */
+ for (size_t i = 0; i < relocs->num_relocs; i++) {
+ if (relocs->reloc_bos[i] == &pool->bo) {
+ /* Adjust the delta value in the relocation to correctly
+ * correspond to the new delta. Initially, this value may have
+ * been negative (if treated as unsigned), but we trust in
+ * uint32_t roll-over to fix that for us at this point.
+ */
+ relocs->relocs[i].delta += delta;
+
+ /* Since the delta has changed, we need to update the actual
+ * relocated value with the new presumed value. This function
+ * should only be called on batch buffers, so we know it isn't in
+ * use by the GPU at the moment.
+ */
+ assert(relocs->relocs[i].offset < from_bo->size);
+ uint32_t *reloc_data = from_bo->map + relocs->relocs[i].offset;
+ *reloc_data = relocs->relocs[i].presumed_offset +
+ relocs->relocs[i].delta;
+ }
+ }
+
+ *last_pool_center_bo_offset = pool->center_bo_offset;
+}
+
+void
+anv_cmd_buffer_prepare_execbuf(struct anv_cmd_buffer *cmd_buffer)
+{
+ struct anv_batch *batch = &cmd_buffer->batch;
+ struct anv_block_pool *ss_pool =
+ &cmd_buffer->device->surface_state_block_pool;
+
+ cmd_buffer->execbuf2.bo_count = 0;
+ cmd_buffer->execbuf2.need_reloc = false;
+
+ adjust_relocations_from_block_pool(ss_pool, &cmd_buffer->surface_relocs);
+ anv_cmd_buffer_add_bo(cmd_buffer, &ss_pool->bo, &cmd_buffer->surface_relocs);
+
+ /* First, we walk over all of the bos we've seen and add them and their
+ * relocations to the validate list.
+ */
+ struct anv_batch_bo **bbo;
+ anv_vector_foreach(bbo, &cmd_buffer->seen_bbos) {
+ adjust_relocations_to_block_pool(ss_pool, &(*bbo)->bo, &(*bbo)->relocs,
+ &(*bbo)->last_ss_pool_bo_offset);
+
+ anv_cmd_buffer_add_bo(cmd_buffer, &(*bbo)->bo, &(*bbo)->relocs);
+ }
+
+ struct anv_batch_bo *first_batch_bo =
+ list_first_entry(&cmd_buffer->batch_bos, struct anv_batch_bo, link);
+
+ /* The kernel requires that the last entry in the validation list be the
+ * batch buffer to execute. We can simply swap the element
+ * corresponding to the first batch_bo in the chain with the last
+ * element in the list.
+ */
+ if (first_batch_bo->bo.index != cmd_buffer->execbuf2.bo_count - 1) {
+ uint32_t idx = first_batch_bo->bo.index;
+ uint32_t last_idx = cmd_buffer->execbuf2.bo_count - 1;
+
+ struct drm_i915_gem_exec_object2 tmp_obj =
+ cmd_buffer->execbuf2.objects[idx];
+ assert(cmd_buffer->execbuf2.bos[idx] == &first_batch_bo->bo);
+
+ cmd_buffer->execbuf2.objects[idx] = cmd_buffer->execbuf2.objects[last_idx];
+ cmd_buffer->execbuf2.bos[idx] = cmd_buffer->execbuf2.bos[last_idx];
+ cmd_buffer->execbuf2.bos[idx]->index = idx;
+
+ cmd_buffer->execbuf2.objects[last_idx] = tmp_obj;
+ cmd_buffer->execbuf2.bos[last_idx] = &first_batch_bo->bo;
+ first_batch_bo->bo.index = last_idx;
+ }
+
+ /* Now we go through and fixup all of the relocation lists to point to
+ * the correct indices in the object array. We have to do this after we
+ * reorder the list above as some of the indices may have changed.
+ */
+ anv_vector_foreach(bbo, &cmd_buffer->seen_bbos)
+ anv_cmd_buffer_process_relocs(cmd_buffer, &(*bbo)->relocs);
+
+ anv_cmd_buffer_process_relocs(cmd_buffer, &cmd_buffer->surface_relocs);
+
+ if (!cmd_buffer->device->info.has_llc) {
+ __builtin_ia32_sfence();
+ anv_vector_foreach(bbo, &cmd_buffer->seen_bbos) {
+ for (uint32_t i = 0; i < (*bbo)->length; i += CACHELINE_SIZE)
+ __builtin_ia32_clflush((*bbo)->bo.map + i);
+ }
+ }
+
+ cmd_buffer->execbuf2.execbuf = (struct drm_i915_gem_execbuffer2) {
+ .buffers_ptr = (uintptr_t) cmd_buffer->execbuf2.objects,
+ .buffer_count = cmd_buffer->execbuf2.bo_count,
+ .batch_start_offset = 0,
+ .batch_len = batch->next - batch->start,
+ .cliprects_ptr = 0,
+ .num_cliprects = 0,
+ .DR1 = 0,
+ .DR4 = 0,
+ .flags = I915_EXEC_HANDLE_LUT | I915_EXEC_RENDER |
+ I915_EXEC_CONSTANTS_REL_GENERAL,
+ .rsvd1 = cmd_buffer->device->context_id,
+ .rsvd2 = 0,
+ };
+
+ if (!cmd_buffer->execbuf2.need_reloc)
+ cmd_buffer->execbuf2.execbuf.flags |= I915_EXEC_NO_RELOC;
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <assert.h>
+#include <stdbool.h>
+#include <string.h>
+#include <unistd.h>
+#include <fcntl.h>
+
+#include "anv_private.h"
+
+/** \file anv_cmd_buffer.c
+ *
+ * This file contains all of the stuff for emitting commands into a command
+ * buffer. This includes implementations of most of the vkCmd*
+ * entrypoints. This file is concerned entirely with state emission and
+ * not with the command buffer data structure itself. As far as this file
+ * is concerned, most of anv_cmd_buffer is magic.
+ */
+
+/* TODO: These are taken from GLES. We should check the Vulkan spec */
+const struct anv_dynamic_state default_dynamic_state = {
+ .viewport = {
+ .count = 0,
+ },
+ .scissor = {
+ .count = 0,
+ },
+ .line_width = 1.0f,
+ .depth_bias = {
+ .bias = 0.0f,
+ .clamp = 0.0f,
+ .slope = 0.0f,
+ },
+ .blend_constants = { 0.0f, 0.0f, 0.0f, 0.0f },
+ .depth_bounds = {
+ .min = 0.0f,
+ .max = 1.0f,
+ },
+ .stencil_compare_mask = {
+ .front = ~0u,
+ .back = ~0u,
+ },
+ .stencil_write_mask = {
+ .front = ~0u,
+ .back = ~0u,
+ },
+ .stencil_reference = {
+ .front = 0u,
+ .back = 0u,
+ },
+};
+
+void
+anv_dynamic_state_copy(struct anv_dynamic_state *dest,
+ const struct anv_dynamic_state *src,
+ uint32_t copy_mask)
+{
+ if (copy_mask & (1 << VK_DYNAMIC_STATE_VIEWPORT)) {
+ dest->viewport.count = src->viewport.count;
+ typed_memcpy(dest->viewport.viewports, src->viewport.viewports,
+ src->viewport.count);
+ }
+
+ if (copy_mask & (1 << VK_DYNAMIC_STATE_SCISSOR)) {
+ dest->scissor.count = src->scissor.count;
+ typed_memcpy(dest->scissor.scissors, src->scissor.scissors,
+ src->scissor.count);
+ }
+
+ if (copy_mask & (1 << VK_DYNAMIC_STATE_LINE_WIDTH))
+ dest->line_width = src->line_width;
+
+ if (copy_mask & (1 << VK_DYNAMIC_STATE_DEPTH_BIAS))
+ dest->depth_bias = src->depth_bias;
+
+ if (copy_mask & (1 << VK_DYNAMIC_STATE_BLEND_CONSTANTS))
+ typed_memcpy(dest->blend_constants, src->blend_constants, 4);
+
+ if (copy_mask & (1 << VK_DYNAMIC_STATE_DEPTH_BOUNDS))
+ dest->depth_bounds = src->depth_bounds;
+
+ if (copy_mask & (1 << VK_DYNAMIC_STATE_STENCIL_COMPARE_MASK))
+ dest->stencil_compare_mask = src->stencil_compare_mask;
+
+ if (copy_mask & (1 << VK_DYNAMIC_STATE_STENCIL_WRITE_MASK))
+ dest->stencil_write_mask = src->stencil_write_mask;
+
+ if (copy_mask & (1 << VK_DYNAMIC_STATE_STENCIL_REFERENCE))
+ dest->stencil_reference = src->stencil_reference;
+}
+
+static void
+anv_cmd_state_reset(struct anv_cmd_buffer *cmd_buffer)
+{
+ struct anv_cmd_state *state = &cmd_buffer->state;
+
+ memset(&state->descriptors, 0, sizeof(state->descriptors));
+ memset(&state->push_constants, 0, sizeof(state->push_constants));
+
+ state->dirty = ~0;
+ state->vb_dirty = 0;
+ state->descriptors_dirty = 0;
+ state->push_constants_dirty = 0;
+ state->pipeline = NULL;
+ state->restart_index = UINT32_MAX;
+ state->dynamic = default_dynamic_state;
+ state->need_query_wa = true;
+
+ if (state->attachments != NULL) {
+ anv_free(&cmd_buffer->pool->alloc, state->attachments);
+ state->attachments = NULL;
+ }
+
+ state->gen7.index_buffer = NULL;
+}
+
+/**
+ * Setup anv_cmd_state::attachments for vkCmdBeginRenderPass.
+ */
+void
+anv_cmd_state_setup_attachments(struct anv_cmd_buffer *cmd_buffer,
+ const VkRenderPassBeginInfo *info)
+{
+ struct anv_cmd_state *state = &cmd_buffer->state;
+ ANV_FROM_HANDLE(anv_render_pass, pass, info->renderPass);
+
+ anv_free(&cmd_buffer->pool->alloc, state->attachments);
+
+ if (pass->attachment_count == 0) {
+ state->attachments = NULL;
+ return;
+ }
+
+ state->attachments = anv_alloc(&cmd_buffer->pool->alloc,
+ pass->attachment_count *
+ sizeof(state->attachments[0]),
+ 8, VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (state->attachments == NULL) {
+ /* FIXME: Propagate VK_ERROR_OUT_OF_HOST_MEMORY to vkEndCommandBuffer */
+ abort();
+ }
+
+ for (uint32_t i = 0; i < pass->attachment_count; ++i) {
+ struct anv_render_pass_attachment *att = &pass->attachments[i];
+ VkImageAspectFlags clear_aspects = 0;
+
+ if (anv_format_is_color(att->format)) {
+ /* color attachment */
+ if (att->load_op == VK_ATTACHMENT_LOAD_OP_CLEAR) {
+ clear_aspects |= VK_IMAGE_ASPECT_COLOR_BIT;
+ }
+ } else {
+ /* depthstencil attachment */
+ if (att->format->depth_format &&
+ att->load_op == VK_ATTACHMENT_LOAD_OP_CLEAR) {
+ clear_aspects |= VK_IMAGE_ASPECT_DEPTH_BIT;
+ }
+ if (att->format->has_stencil &&
+ att->stencil_load_op == VK_ATTACHMENT_LOAD_OP_CLEAR) {
+ clear_aspects |= VK_IMAGE_ASPECT_STENCIL_BIT;
+ }
+ }
+
+ state->attachments[i].pending_clear_aspects = clear_aspects;
+ if (clear_aspects) {
+ assert(info->clearValueCount > i);
+ state->attachments[i].clear_value = info->pClearValues[i];
+ }
+ }
+}
+
+static VkResult
+anv_cmd_buffer_ensure_push_constants_size(struct anv_cmd_buffer *cmd_buffer,
+ gl_shader_stage stage, uint32_t size)
+{
+ struct anv_push_constants **ptr = &cmd_buffer->state.push_constants[stage];
+
+ if (*ptr == NULL) {
+ *ptr = anv_alloc(&cmd_buffer->pool->alloc, size, 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (*ptr == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+ } else if ((*ptr)->size < size) {
+ *ptr = anv_realloc(&cmd_buffer->pool->alloc, *ptr, size, 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (*ptr == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+ }
+ (*ptr)->size = size;
+
+ return VK_SUCCESS;
+}
+
+#define anv_cmd_buffer_ensure_push_constant_field(cmd_buffer, stage, field) \
+ anv_cmd_buffer_ensure_push_constants_size(cmd_buffer, stage, \
+ (offsetof(struct anv_push_constants, field) + \
+ sizeof(cmd_buffer->state.push_constants[0]->field)))
+
+static VkResult anv_create_cmd_buffer(
+ struct anv_device * device,
+ struct anv_cmd_pool * pool,
+ VkCommandBufferLevel level,
+ VkCommandBuffer* pCommandBuffer)
+{
+ struct anv_cmd_buffer *cmd_buffer;
+ VkResult result;
+
+ cmd_buffer = anv_alloc(&pool->alloc, sizeof(*cmd_buffer), 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (cmd_buffer == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ cmd_buffer->_loader_data.loaderMagic = ICD_LOADER_MAGIC;
+ cmd_buffer->device = device;
+ cmd_buffer->pool = pool;
+ cmd_buffer->level = level;
+ cmd_buffer->state.attachments = NULL;
+
+ result = anv_cmd_buffer_init_batch_bo_chain(cmd_buffer);
+ if (result != VK_SUCCESS)
+ goto fail;
+
+ anv_state_stream_init(&cmd_buffer->surface_state_stream,
+ &device->surface_state_block_pool);
+ anv_state_stream_init(&cmd_buffer->dynamic_state_stream,
+ &device->dynamic_state_block_pool);
+
+ if (pool) {
+ list_addtail(&cmd_buffer->pool_link, &pool->cmd_buffers);
+ } else {
+ /* Init the pool_link so we can safefly call list_del when we destroy
+ * the command buffer
+ */
+ list_inithead(&cmd_buffer->pool_link);
+ }
+
+ *pCommandBuffer = anv_cmd_buffer_to_handle(cmd_buffer);
+
+ return VK_SUCCESS;
+
+ fail:
+ anv_free(&cmd_buffer->pool->alloc, cmd_buffer);
+
+ return result;
+}
+
+VkResult anv_AllocateCommandBuffers(
+ VkDevice _device,
+ const VkCommandBufferAllocateInfo* pAllocateInfo,
+ VkCommandBuffer* pCommandBuffers)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ ANV_FROM_HANDLE(anv_cmd_pool, pool, pAllocateInfo->commandPool);
+
+ VkResult result = VK_SUCCESS;
+ uint32_t i;
+
+ for (i = 0; i < pAllocateInfo->commandBufferCount; i++) {
+ result = anv_create_cmd_buffer(device, pool, pAllocateInfo->level,
+ &pCommandBuffers[i]);
+ if (result != VK_SUCCESS)
+ break;
+ }
+
+ if (result != VK_SUCCESS)
+ anv_FreeCommandBuffers(_device, pAllocateInfo->commandPool,
+ i, pCommandBuffers);
+
+ return result;
+}
+
+static void
+anv_cmd_buffer_destroy(struct anv_cmd_buffer *cmd_buffer)
+{
+ list_del(&cmd_buffer->pool_link);
+
+ anv_cmd_buffer_fini_batch_bo_chain(cmd_buffer);
+
+ anv_state_stream_finish(&cmd_buffer->surface_state_stream);
+ anv_state_stream_finish(&cmd_buffer->dynamic_state_stream);
+
+ anv_free(&cmd_buffer->pool->alloc, cmd_buffer->state.attachments);
+ anv_free(&cmd_buffer->pool->alloc, cmd_buffer);
+}
+
+void anv_FreeCommandBuffers(
+ VkDevice device,
+ VkCommandPool commandPool,
+ uint32_t commandBufferCount,
+ const VkCommandBuffer* pCommandBuffers)
+{
+ for (uint32_t i = 0; i < commandBufferCount; i++) {
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, pCommandBuffers[i]);
+
+ anv_cmd_buffer_destroy(cmd_buffer);
+ }
+}
+
+VkResult anv_ResetCommandBuffer(
+ VkCommandBuffer commandBuffer,
+ VkCommandBufferResetFlags flags)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+
+ cmd_buffer->usage_flags = 0;
+ cmd_buffer->state.current_pipeline = UINT32_MAX;
+ anv_cmd_buffer_reset_batch_bo_chain(cmd_buffer);
+ anv_cmd_state_reset(cmd_buffer);
+
+ return VK_SUCCESS;
+}
+
+void
+anv_cmd_buffer_emit_state_base_address(struct anv_cmd_buffer *cmd_buffer)
+{
+ switch (cmd_buffer->device->info.gen) {
+ case 7:
+ if (cmd_buffer->device->info.is_haswell)
+ return gen7_cmd_buffer_emit_state_base_address(cmd_buffer);
+ else
+ return gen7_cmd_buffer_emit_state_base_address(cmd_buffer);
+ case 8:
+ return gen8_cmd_buffer_emit_state_base_address(cmd_buffer);
+ case 9:
+ return gen9_cmd_buffer_emit_state_base_address(cmd_buffer);
+ default:
+ unreachable("unsupported gen\n");
+ }
+}
+
+VkResult anv_BeginCommandBuffer(
+ VkCommandBuffer commandBuffer,
+ const VkCommandBufferBeginInfo* pBeginInfo)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+
+ /* If this is the first vkBeginCommandBuffer, we must *initialize* the
+ * command buffer's state. Otherwise, we must *reset* its state. In both
+ * cases we reset it.
+ *
+ * From the Vulkan 1.0 spec:
+ *
+ * If a command buffer is in the executable state and the command buffer
+ * was allocated from a command pool with the
+ * VK_COMMAND_POOL_CREATE_RESET_COMMAND_BUFFER_BIT flag set, then
+ * vkBeginCommandBuffer implicitly resets the command buffer, behaving
+ * as if vkResetCommandBuffer had been called with
+ * VK_COMMAND_BUFFER_RESET_RELEASE_RESOURCES_BIT not set. It then puts
+ * the command buffer in the recording state.
+ */
+ anv_ResetCommandBuffer(commandBuffer, /*flags*/ 0);
+
+ cmd_buffer->usage_flags = pBeginInfo->flags;
+
+ assert(cmd_buffer->level == VK_COMMAND_BUFFER_LEVEL_SECONDARY ||
+ !(cmd_buffer->usage_flags & VK_COMMAND_BUFFER_USAGE_RENDER_PASS_CONTINUE_BIT));
+
+ if (cmd_buffer->usage_flags &
+ VK_COMMAND_BUFFER_USAGE_RENDER_PASS_CONTINUE_BIT) {
+ cmd_buffer->state.framebuffer =
+ anv_framebuffer_from_handle(pBeginInfo->pInheritanceInfo->framebuffer);
+ cmd_buffer->state.pass =
+ anv_render_pass_from_handle(pBeginInfo->pInheritanceInfo->renderPass);
+
+ struct anv_subpass *subpass =
+ &cmd_buffer->state.pass->subpasses[pBeginInfo->pInheritanceInfo->subpass];
+
+ anv_cmd_buffer_set_subpass(cmd_buffer, subpass);
+ }
+
+ anv_cmd_buffer_emit_state_base_address(cmd_buffer);
+
+ return VK_SUCCESS;
+}
+
+VkResult anv_EndCommandBuffer(
+ VkCommandBuffer commandBuffer)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ struct anv_device *device = cmd_buffer->device;
+
+ anv_cmd_buffer_end_batch_buffer(cmd_buffer);
+
+ if (cmd_buffer->level == VK_COMMAND_BUFFER_LEVEL_PRIMARY) {
+ /* The algorithm used to compute the validate list is not threadsafe as
+ * it uses the bo->index field. We have to lock the device around it.
+ * Fortunately, the chances for contention here are probably very low.
+ */
+ pthread_mutex_lock(&device->mutex);
+ anv_cmd_buffer_prepare_execbuf(cmd_buffer);
+ pthread_mutex_unlock(&device->mutex);
+ }
+
+ return VK_SUCCESS;
+}
+
+void anv_CmdBindPipeline(
+ VkCommandBuffer commandBuffer,
+ VkPipelineBindPoint pipelineBindPoint,
+ VkPipeline _pipeline)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_pipeline, pipeline, _pipeline);
+
+ switch (pipelineBindPoint) {
+ case VK_PIPELINE_BIND_POINT_COMPUTE:
+ cmd_buffer->state.compute_pipeline = pipeline;
+ cmd_buffer->state.compute_dirty |= ANV_CMD_DIRTY_PIPELINE;
+ cmd_buffer->state.push_constants_dirty |= VK_SHADER_STAGE_COMPUTE_BIT;
+ break;
+
+ case VK_PIPELINE_BIND_POINT_GRAPHICS:
+ cmd_buffer->state.pipeline = pipeline;
+ cmd_buffer->state.vb_dirty |= pipeline->vb_used;
+ cmd_buffer->state.dirty |= ANV_CMD_DIRTY_PIPELINE;
+ cmd_buffer->state.push_constants_dirty |= pipeline->active_stages;
+
+ /* Apply the dynamic state from the pipeline */
+ cmd_buffer->state.dirty |= pipeline->dynamic_state_mask;
+ anv_dynamic_state_copy(&cmd_buffer->state.dynamic,
+ &pipeline->dynamic_state,
+ pipeline->dynamic_state_mask);
+ break;
+
+ default:
+ assert(!"invalid bind point");
+ break;
+ }
+}
+
+void anv_CmdSetViewport(
+ VkCommandBuffer commandBuffer,
+ uint32_t firstViewport,
+ uint32_t viewportCount,
+ const VkViewport* pViewports)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+
+ const uint32_t total_count = firstViewport + viewportCount;
+ if (cmd_buffer->state.dynamic.viewport.count < total_count);
+ cmd_buffer->state.dynamic.viewport.count = total_count;
+
+ memcpy(cmd_buffer->state.dynamic.viewport.viewports + firstViewport,
+ pViewports, viewportCount * sizeof(*pViewports));
+
+ cmd_buffer->state.dirty |= ANV_CMD_DIRTY_DYNAMIC_VIEWPORT;
+}
+
+void anv_CmdSetScissor(
+ VkCommandBuffer commandBuffer,
+ uint32_t firstScissor,
+ uint32_t scissorCount,
+ const VkRect2D* pScissors)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+
+ const uint32_t total_count = firstScissor + scissorCount;
+ if (cmd_buffer->state.dynamic.scissor.count < total_count);
+ cmd_buffer->state.dynamic.scissor.count = total_count;
+
+ memcpy(cmd_buffer->state.dynamic.scissor.scissors + firstScissor,
+ pScissors, scissorCount * sizeof(*pScissors));
+
+ cmd_buffer->state.dirty |= ANV_CMD_DIRTY_DYNAMIC_SCISSOR;
+}
+
+void anv_CmdSetLineWidth(
+ VkCommandBuffer commandBuffer,
+ float lineWidth)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+
+ cmd_buffer->state.dynamic.line_width = lineWidth;
+ cmd_buffer->state.dirty |= ANV_CMD_DIRTY_DYNAMIC_LINE_WIDTH;
+}
+
+void anv_CmdSetDepthBias(
+ VkCommandBuffer commandBuffer,
+ float depthBiasConstantFactor,
+ float depthBiasClamp,
+ float depthBiasSlopeFactor)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+
+ cmd_buffer->state.dynamic.depth_bias.bias = depthBiasConstantFactor;
+ cmd_buffer->state.dynamic.depth_bias.clamp = depthBiasClamp;
+ cmd_buffer->state.dynamic.depth_bias.slope = depthBiasSlopeFactor;
+
+ cmd_buffer->state.dirty |= ANV_CMD_DIRTY_DYNAMIC_DEPTH_BIAS;
+}
+
+void anv_CmdSetBlendConstants(
+ VkCommandBuffer commandBuffer,
+ const float blendConstants[4])
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+
+ memcpy(cmd_buffer->state.dynamic.blend_constants,
+ blendConstants, sizeof(float) * 4);
+
+ cmd_buffer->state.dirty |= ANV_CMD_DIRTY_DYNAMIC_BLEND_CONSTANTS;
+}
+
+void anv_CmdSetDepthBounds(
+ VkCommandBuffer commandBuffer,
+ float minDepthBounds,
+ float maxDepthBounds)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+
+ cmd_buffer->state.dynamic.depth_bounds.min = minDepthBounds;
+ cmd_buffer->state.dynamic.depth_bounds.max = maxDepthBounds;
+
+ cmd_buffer->state.dirty |= ANV_CMD_DIRTY_DYNAMIC_DEPTH_BOUNDS;
+}
+
+void anv_CmdSetStencilCompareMask(
+ VkCommandBuffer commandBuffer,
+ VkStencilFaceFlags faceMask,
+ uint32_t compareMask)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+
+ if (faceMask & VK_STENCIL_FACE_FRONT_BIT)
+ cmd_buffer->state.dynamic.stencil_compare_mask.front = compareMask;
+ if (faceMask & VK_STENCIL_FACE_BACK_BIT)
+ cmd_buffer->state.dynamic.stencil_compare_mask.back = compareMask;
+
+ cmd_buffer->state.dirty |= ANV_CMD_DIRTY_DYNAMIC_STENCIL_COMPARE_MASK;
+}
+
+void anv_CmdSetStencilWriteMask(
+ VkCommandBuffer commandBuffer,
+ VkStencilFaceFlags faceMask,
+ uint32_t writeMask)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+
+ if (faceMask & VK_STENCIL_FACE_FRONT_BIT)
+ cmd_buffer->state.dynamic.stencil_write_mask.front = writeMask;
+ if (faceMask & VK_STENCIL_FACE_BACK_BIT)
+ cmd_buffer->state.dynamic.stencil_write_mask.back = writeMask;
+
+ cmd_buffer->state.dirty |= ANV_CMD_DIRTY_DYNAMIC_STENCIL_WRITE_MASK;
+}
+
+void anv_CmdSetStencilReference(
+ VkCommandBuffer commandBuffer,
+ VkStencilFaceFlags faceMask,
+ uint32_t reference)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+
+ if (faceMask & VK_STENCIL_FACE_FRONT_BIT)
+ cmd_buffer->state.dynamic.stencil_reference.front = reference;
+ if (faceMask & VK_STENCIL_FACE_BACK_BIT)
+ cmd_buffer->state.dynamic.stencil_reference.back = reference;
+
+ cmd_buffer->state.dirty |= ANV_CMD_DIRTY_DYNAMIC_STENCIL_REFERENCE;
+}
+
+void anv_CmdBindDescriptorSets(
+ VkCommandBuffer commandBuffer,
+ VkPipelineBindPoint pipelineBindPoint,
+ VkPipelineLayout _layout,
+ uint32_t firstSet,
+ uint32_t descriptorSetCount,
+ const VkDescriptorSet* pDescriptorSets,
+ uint32_t dynamicOffsetCount,
+ const uint32_t* pDynamicOffsets)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_pipeline_layout, layout, _layout);
+ struct anv_descriptor_set_layout *set_layout;
+
+ assert(firstSet + descriptorSetCount < MAX_SETS);
+
+ uint32_t dynamic_slot = 0;
+ for (uint32_t i = 0; i < descriptorSetCount; i++) {
+ ANV_FROM_HANDLE(anv_descriptor_set, set, pDescriptorSets[i]);
+ set_layout = layout->set[firstSet + i].layout;
+
+ if (cmd_buffer->state.descriptors[firstSet + i] != set) {
+ cmd_buffer->state.descriptors[firstSet + i] = set;
+ cmd_buffer->state.descriptors_dirty |= set_layout->shader_stages;
+ }
+
+ if (set_layout->dynamic_offset_count > 0) {
+ anv_foreach_stage(s, set_layout->shader_stages) {
+ anv_cmd_buffer_ensure_push_constant_field(cmd_buffer, s, dynamic);
+
+ struct anv_push_constants *push =
+ cmd_buffer->state.push_constants[s];
+
+ unsigned d = layout->set[firstSet + i].dynamic_offset_start;
+ const uint32_t *offsets = pDynamicOffsets + dynamic_slot;
+ struct anv_descriptor *desc = set->descriptors;
+
+ for (unsigned b = 0; b < set_layout->binding_count; b++) {
+ if (set_layout->binding[b].dynamic_offset_index < 0)
+ continue;
+
+ unsigned array_size = set_layout->binding[b].array_size;
+ for (unsigned j = 0; j < array_size; j++) {
+ uint32_t range = 0;
+ if (desc->buffer_view)
+ range = desc->buffer_view->range;
+ push->dynamic[d].offset = *(offsets++);
+ push->dynamic[d].range = range;
+ desc++;
+ d++;
+ }
+ }
+ }
+ cmd_buffer->state.push_constants_dirty |= set_layout->shader_stages;
+ }
+ }
+}
+
+void anv_CmdBindVertexBuffers(
+ VkCommandBuffer commandBuffer,
+ uint32_t firstBinding,
+ uint32_t bindingCount,
+ const VkBuffer* pBuffers,
+ const VkDeviceSize* pOffsets)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ struct anv_vertex_binding *vb = cmd_buffer->state.vertex_bindings;
+
+ /* We have to defer setting up vertex buffer since we need the buffer
+ * stride from the pipeline. */
+
+ assert(firstBinding + bindingCount < MAX_VBS);
+ for (uint32_t i = 0; i < bindingCount; i++) {
+ vb[firstBinding + i].buffer = anv_buffer_from_handle(pBuffers[i]);
+ vb[firstBinding + i].offset = pOffsets[i];
+ cmd_buffer->state.vb_dirty |= 1 << (firstBinding + i);
+ }
+}
+
+static void
+add_surface_state_reloc(struct anv_cmd_buffer *cmd_buffer,
+ struct anv_state state, struct anv_bo *bo, uint32_t offset)
+{
+ /* The address goes in SURFACE_STATE dword 1 for gens < 8 and dwords 8 and
+ * 9 for gen8+. We only write the first dword for gen8+ here and rely on
+ * the initial state to set the high bits to 0. */
+
+ const uint32_t dword = cmd_buffer->device->info.gen < 8 ? 1 : 8;
+
+ anv_reloc_list_add(&cmd_buffer->surface_relocs, &cmd_buffer->pool->alloc,
+ state.offset + dword * 4, bo, offset);
+}
+
+const struct anv_format *
+anv_format_for_descriptor_type(VkDescriptorType type)
+{
+ switch (type) {
+ case VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER:
+ case VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC:
+ return anv_format_for_vk_format(VK_FORMAT_R32G32B32A32_SFLOAT);
+
+ case VK_DESCRIPTOR_TYPE_STORAGE_BUFFER:
+ case VK_DESCRIPTOR_TYPE_STORAGE_BUFFER_DYNAMIC:
+ return anv_format_for_vk_format(VK_FORMAT_UNDEFINED);
+
+ default:
+ unreachable("Invalid descriptor type");
+ }
+}
+
+VkResult
+anv_cmd_buffer_emit_binding_table(struct anv_cmd_buffer *cmd_buffer,
+ gl_shader_stage stage,
+ struct anv_state *bt_state)
+{
+ struct anv_framebuffer *fb = cmd_buffer->state.framebuffer;
+ struct anv_subpass *subpass = cmd_buffer->state.subpass;
+ struct anv_pipeline_layout *layout;
+ uint32_t color_count, bias, state_offset;
+
+ switch (stage) {
+ case MESA_SHADER_FRAGMENT:
+ layout = cmd_buffer->state.pipeline->layout;
+ bias = MAX_RTS;
+ color_count = subpass->color_count;
+ break;
+ case MESA_SHADER_COMPUTE:
+ layout = cmd_buffer->state.compute_pipeline->layout;
+ bias = 1;
+ color_count = 0;
+ break;
+ default:
+ layout = cmd_buffer->state.pipeline->layout;
+ bias = 0;
+ color_count = 0;
+ break;
+ }
+
+ /* This is a little awkward: layout can be NULL but we still have to
+ * allocate and set a binding table for the PS stage for render
+ * targets. */
+ uint32_t surface_count = layout ? layout->stage[stage].surface_count : 0;
+
+ if (color_count + surface_count == 0) {
+ *bt_state = (struct anv_state) { 0, };
+ return VK_SUCCESS;
+ }
+
+ *bt_state = anv_cmd_buffer_alloc_binding_table(cmd_buffer,
+ bias + surface_count,
+ &state_offset);
+ uint32_t *bt_map = bt_state->map;
+
+ if (bt_state->map == NULL)
+ return VK_ERROR_OUT_OF_DEVICE_MEMORY;
+
+ for (uint32_t a = 0; a < color_count; a++) {
+ const struct anv_image_view *iview =
+ fb->attachments[subpass->color_attachments[a]];
+
+ assert(iview->color_rt_surface_state.alloc_size);
+ bt_map[a] = iview->color_rt_surface_state.offset + state_offset;
+ add_surface_state_reloc(cmd_buffer, iview->color_rt_surface_state,
+ iview->bo, iview->offset);
+ }
+
+ if (stage == MESA_SHADER_COMPUTE &&
+ cmd_buffer->state.compute_pipeline->cs_prog_data.uses_num_work_groups) {
+ struct anv_bo *bo = cmd_buffer->state.num_workgroups_bo;
+ uint32_t bo_offset = cmd_buffer->state.num_workgroups_offset;
+
+ struct anv_state surface_state;
+ surface_state =
+ anv_cmd_buffer_alloc_surface_state(cmd_buffer);
+
+ const struct anv_format *format =
+ anv_format_for_descriptor_type(VK_DESCRIPTOR_TYPE_STORAGE_BUFFER);
+ anv_fill_buffer_surface_state(cmd_buffer->device, surface_state.map,
+ format->surface_format, bo_offset, 12, 1);
+
+ if (!cmd_buffer->device->info.has_llc)
+ anv_state_clflush(surface_state);
+
+ bt_map[0] = surface_state.offset + state_offset;
+ add_surface_state_reloc(cmd_buffer, surface_state, bo, bo_offset);
+ }
+
+ if (layout == NULL)
+ goto out;
+
+ if (layout->stage[stage].image_count > 0) {
+ VkResult result =
+ anv_cmd_buffer_ensure_push_constant_field(cmd_buffer, stage, images);
+ if (result != VK_SUCCESS)
+ return result;
+
+ cmd_buffer->state.push_constants_dirty |= 1 << stage;
+ }
+
+ uint32_t image = 0;
+ for (uint32_t s = 0; s < layout->stage[stage].surface_count; s++) {
+ struct anv_pipeline_binding *binding =
+ &layout->stage[stage].surface_to_descriptor[s];
+ struct anv_descriptor_set *set =
+ cmd_buffer->state.descriptors[binding->set];
+ struct anv_descriptor *desc = &set->descriptors[binding->offset];
+
+ struct anv_state surface_state;
+ struct anv_bo *bo;
+ uint32_t bo_offset;
+
+ switch (desc->type) {
+ case VK_DESCRIPTOR_TYPE_SAMPLER:
+ /* Nothing for us to do here */
+ continue;
+
+ case VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER:
+ case VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE:
+ case VK_DESCRIPTOR_TYPE_INPUT_ATTACHMENT:
+ surface_state = desc->image_view->nonrt_surface_state;
+ assert(surface_state.alloc_size);
+ bo = desc->image_view->bo;
+ bo_offset = desc->image_view->offset;
+ break;
+
+ case VK_DESCRIPTOR_TYPE_STORAGE_IMAGE: {
+ surface_state = desc->image_view->storage_surface_state;
+ assert(surface_state.alloc_size);
+ bo = desc->image_view->bo;
+ bo_offset = desc->image_view->offset;
+
+ struct brw_image_param *image_param =
+ &cmd_buffer->state.push_constants[stage]->images[image++];
+
+ anv_image_view_fill_image_param(cmd_buffer->device, desc->image_view,
+ image_param);
+ image_param->surface_idx = bias + s;
+ break;
+ }
+
+ case VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER:
+ case VK_DESCRIPTOR_TYPE_STORAGE_BUFFER:
+ case VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC:
+ case VK_DESCRIPTOR_TYPE_STORAGE_BUFFER_DYNAMIC:
+ case VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER:
+ surface_state = desc->buffer_view->surface_state;
+ assert(surface_state.alloc_size);
+ bo = desc->buffer_view->bo;
+ bo_offset = desc->buffer_view->offset;
+ break;
+
+ case VK_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER:
+ surface_state = desc->buffer_view->storage_surface_state;
+ assert(surface_state.alloc_size);
+ bo = desc->buffer_view->bo;
+ bo_offset = desc->buffer_view->offset;
+
+ struct brw_image_param *image_param =
+ &cmd_buffer->state.push_constants[stage]->images[image++];
+
+ anv_buffer_view_fill_image_param(cmd_buffer->device, desc->buffer_view,
+ image_param);
+ image_param->surface_idx = bias + s;
+ break;
+
+ default:
+ assert(!"Invalid descriptor type");
+ continue;
+ }
+
+ bt_map[bias + s] = surface_state.offset + state_offset;
+ add_surface_state_reloc(cmd_buffer, surface_state, bo, bo_offset);
+ }
+ assert(image == layout->stage[stage].image_count);
+
+ out:
+ if (!cmd_buffer->device->info.has_llc)
+ anv_state_clflush(*bt_state);
+
+ return VK_SUCCESS;
+}
+
+VkResult
+anv_cmd_buffer_emit_samplers(struct anv_cmd_buffer *cmd_buffer,
+ gl_shader_stage stage, struct anv_state *state)
+{
+ struct anv_pipeline_layout *layout;
+ uint32_t sampler_count;
+
+ if (stage == MESA_SHADER_COMPUTE)
+ layout = cmd_buffer->state.compute_pipeline->layout;
+ else
+ layout = cmd_buffer->state.pipeline->layout;
+
+ sampler_count = layout ? layout->stage[stage].sampler_count : 0;
+ if (sampler_count == 0) {
+ *state = (struct anv_state) { 0, };
+ return VK_SUCCESS;
+ }
+
+ uint32_t size = sampler_count * 16;
+ *state = anv_cmd_buffer_alloc_dynamic_state(cmd_buffer, size, 32);
+
+ if (state->map == NULL)
+ return VK_ERROR_OUT_OF_DEVICE_MEMORY;
+
+ for (uint32_t s = 0; s < layout->stage[stage].sampler_count; s++) {
+ struct anv_pipeline_binding *binding =
+ &layout->stage[stage].sampler_to_descriptor[s];
+ struct anv_descriptor_set *set =
+ cmd_buffer->state.descriptors[binding->set];
+ struct anv_descriptor *desc = &set->descriptors[binding->offset];
+
+ if (desc->type != VK_DESCRIPTOR_TYPE_SAMPLER &&
+ desc->type != VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER)
+ continue;
+
+ struct anv_sampler *sampler = desc->sampler;
+
+ /* This can happen if we have an unfilled slot since TYPE_SAMPLER
+ * happens to be zero.
+ */
+ if (sampler == NULL)
+ continue;
+
+ memcpy(state->map + (s * 16),
+ sampler->state, sizeof(sampler->state));
+ }
+
+ if (!cmd_buffer->device->info.has_llc)
+ anv_state_clflush(*state);
+
+ return VK_SUCCESS;
+}
+
+struct anv_state
+anv_cmd_buffer_emit_dynamic(struct anv_cmd_buffer *cmd_buffer,
+ const void *data, uint32_t size, uint32_t alignment)
+{
+ struct anv_state state;
+
+ state = anv_cmd_buffer_alloc_dynamic_state(cmd_buffer, size, alignment);
+ memcpy(state.map, data, size);
+
+ if (!cmd_buffer->device->info.has_llc)
+ anv_state_clflush(state);
+
+ VG(VALGRIND_CHECK_MEM_IS_DEFINED(state.map, size));
+
+ return state;
+}
+
+struct anv_state
+anv_cmd_buffer_merge_dynamic(struct anv_cmd_buffer *cmd_buffer,
+ uint32_t *a, uint32_t *b,
+ uint32_t dwords, uint32_t alignment)
+{
+ struct anv_state state;
+ uint32_t *p;
+
+ state = anv_cmd_buffer_alloc_dynamic_state(cmd_buffer,
+ dwords * 4, alignment);
+ p = state.map;
+ for (uint32_t i = 0; i < dwords; i++)
+ p[i] = a[i] | b[i];
+
+ if (!cmd_buffer->device->info.has_llc)
+ anv_state_clflush(state);
+
+ VG(VALGRIND_CHECK_MEM_IS_DEFINED(p, dwords * 4));
+
+ return state;
+}
+
+/**
+ * @brief Setup the command buffer for recording commands inside the given
+ * subpass.
+ *
+ * This does not record all commands needed for starting the subpass.
+ * Starting the subpass may require additional commands.
+ *
+ * Note that vkCmdBeginRenderPass, vkCmdNextSubpass, and vkBeginCommandBuffer
+ * with VK_COMMAND_BUFFER_USAGE_RENDER_PASS_CONTINUE_BIT, all setup the
+ * command buffer for recording commands for some subpass. But only the first
+ * two, vkCmdBeginRenderPass and vkCmdNextSubpass, can start a subpass.
+ */
+void
+anv_cmd_buffer_set_subpass(struct anv_cmd_buffer *cmd_buffer,
+ struct anv_subpass *subpass)
+{
+ switch (cmd_buffer->device->info.gen) {
+ case 7:
+ gen7_cmd_buffer_set_subpass(cmd_buffer, subpass);
+ break;
+ case 8:
+ gen8_cmd_buffer_set_subpass(cmd_buffer, subpass);
+ break;
+ case 9:
+ gen9_cmd_buffer_set_subpass(cmd_buffer, subpass);
+ break;
+ default:
+ unreachable("unsupported gen\n");
+ }
+}
+
+struct anv_state
+anv_cmd_buffer_push_constants(struct anv_cmd_buffer *cmd_buffer,
+ gl_shader_stage stage)
+{
+ struct anv_push_constants *data =
+ cmd_buffer->state.push_constants[stage];
+ struct brw_stage_prog_data *prog_data =
+ cmd_buffer->state.pipeline->prog_data[stage];
+
+ /* If we don't actually have any push constants, bail. */
+ if (data == NULL || prog_data->nr_params == 0)
+ return (struct anv_state) { .offset = 0 };
+
+ struct anv_state state =
+ anv_cmd_buffer_alloc_dynamic_state(cmd_buffer,
+ prog_data->nr_params * sizeof(float),
+ 32 /* bottom 5 bits MBZ */);
+
+ /* Walk through the param array and fill the buffer with data */
+ uint32_t *u32_map = state.map;
+ for (unsigned i = 0; i < prog_data->nr_params; i++) {
+ uint32_t offset = (uintptr_t)prog_data->param[i];
+ u32_map[i] = *(uint32_t *)((uint8_t *)data + offset);
+ }
+
+ if (!cmd_buffer->device->info.has_llc)
+ anv_state_clflush(state);
+
+ return state;
+}
+
+struct anv_state
+anv_cmd_buffer_cs_push_constants(struct anv_cmd_buffer *cmd_buffer)
+{
+ struct anv_push_constants *data =
+ cmd_buffer->state.push_constants[MESA_SHADER_COMPUTE];
+ struct anv_pipeline *pipeline = cmd_buffer->state.compute_pipeline;
+ const struct brw_cs_prog_data *cs_prog_data = &pipeline->cs_prog_data;
+ const struct brw_stage_prog_data *prog_data = &cs_prog_data->base;
+
+ const unsigned local_id_dwords = cs_prog_data->local_invocation_id_regs * 8;
+ const unsigned push_constant_data_size =
+ (local_id_dwords + prog_data->nr_params) * 4;
+ const unsigned reg_aligned_constant_size = ALIGN(push_constant_data_size, 32);
+ const unsigned param_aligned_count =
+ reg_aligned_constant_size / sizeof(uint32_t);
+
+ /* If we don't actually have any push constants, bail. */
+ if (reg_aligned_constant_size == 0)
+ return (struct anv_state) { .offset = 0 };
+
+ const unsigned threads = pipeline->cs_thread_width_max;
+ const unsigned total_push_constants_size =
+ reg_aligned_constant_size * threads;
+ const unsigned push_constant_alignment =
+ cmd_buffer->device->info.gen < 8 ? 32 : 64;
+ const unsigned aligned_total_push_constants_size =
+ ALIGN(total_push_constants_size, push_constant_alignment);
+ struct anv_state state =
+ anv_cmd_buffer_alloc_dynamic_state(cmd_buffer,
+ aligned_total_push_constants_size,
+ push_constant_alignment);
+
+ /* Walk through the param array and fill the buffer with data */
+ uint32_t *u32_map = state.map;
+
+ brw_cs_fill_local_id_payload(cs_prog_data, u32_map, threads,
+ reg_aligned_constant_size);
+
+ /* Setup uniform data for the first thread */
+ for (unsigned i = 0; i < prog_data->nr_params; i++) {
+ uint32_t offset = (uintptr_t)prog_data->param[i];
+ u32_map[local_id_dwords + i] = *(uint32_t *)((uint8_t *)data + offset);
+ }
+
+ /* Copy uniform data from the first thread to every other thread */
+ const size_t uniform_data_size = prog_data->nr_params * sizeof(uint32_t);
+ for (unsigned t = 1; t < threads; t++) {
+ memcpy(&u32_map[t * param_aligned_count + local_id_dwords],
+ &u32_map[local_id_dwords],
+ uniform_data_size);
+ }
+
+ if (!cmd_buffer->device->info.has_llc)
+ anv_state_clflush(state);
+
+ return state;
+}
+
+void anv_CmdPushConstants(
+ VkCommandBuffer commandBuffer,
+ VkPipelineLayout layout,
+ VkShaderStageFlags stageFlags,
+ uint32_t offset,
+ uint32_t size,
+ const void* pValues)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+
+ anv_foreach_stage(stage, stageFlags) {
+ anv_cmd_buffer_ensure_push_constant_field(cmd_buffer, stage, client_data);
+
+ memcpy(cmd_buffer->state.push_constants[stage]->client_data + offset,
+ pValues, size);
+ }
+
+ cmd_buffer->state.push_constants_dirty |= stageFlags;
+}
+
+void anv_CmdExecuteCommands(
+ VkCommandBuffer commandBuffer,
+ uint32_t commandBufferCount,
+ const VkCommandBuffer* pCmdBuffers)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, primary, commandBuffer);
+
+ assert(primary->level == VK_COMMAND_BUFFER_LEVEL_PRIMARY);
+
+ anv_assert(primary->state.subpass == &primary->state.pass->subpasses[0]);
+
+ for (uint32_t i = 0; i < commandBufferCount; i++) {
+ ANV_FROM_HANDLE(anv_cmd_buffer, secondary, pCmdBuffers[i]);
+
+ assert(secondary->level == VK_COMMAND_BUFFER_LEVEL_SECONDARY);
+
+ anv_cmd_buffer_add_secondary(primary, secondary);
+ }
+}
+
+VkResult anv_CreateCommandPool(
+ VkDevice _device,
+ const VkCommandPoolCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkCommandPool* pCmdPool)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ struct anv_cmd_pool *pool;
+
+ pool = anv_alloc2(&device->alloc, pAllocator, sizeof(*pool), 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (pool == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ if (pAllocator)
+ pool->alloc = *pAllocator;
+ else
+ pool->alloc = device->alloc;
+
+ list_inithead(&pool->cmd_buffers);
+
+ *pCmdPool = anv_cmd_pool_to_handle(pool);
+
+ return VK_SUCCESS;
+}
+
+void anv_DestroyCommandPool(
+ VkDevice _device,
+ VkCommandPool commandPool,
+ const VkAllocationCallbacks* pAllocator)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ ANV_FROM_HANDLE(anv_cmd_pool, pool, commandPool);
+
+ anv_ResetCommandPool(_device, commandPool, 0);
+
+ anv_free2(&device->alloc, pAllocator, pool);
+}
+
+VkResult anv_ResetCommandPool(
+ VkDevice device,
+ VkCommandPool commandPool,
+ VkCommandPoolResetFlags flags)
+{
+ ANV_FROM_HANDLE(anv_cmd_pool, pool, commandPool);
+
+ /* FIXME: vkResetCommandPool must not destroy its command buffers. The
+ * Vulkan 1.0 spec requires that it only reset them:
+ *
+ * Resetting a command pool recycles all of the resources from all of
+ * the command buffers allocated from the command pool back to the
+ * command pool. All command buffers that have been allocated from the
+ * command pool are put in the initial state.
+ */
+ list_for_each_entry_safe(struct anv_cmd_buffer, cmd_buffer,
+ &pool->cmd_buffers, pool_link) {
+ anv_cmd_buffer_destroy(cmd_buffer);
+ }
+
+ return VK_SUCCESS;
+}
+
+/**
+ * Return NULL if the current subpass has no depthstencil attachment.
+ */
+const struct anv_image_view *
+anv_cmd_buffer_get_depth_stencil_view(const struct anv_cmd_buffer *cmd_buffer)
+{
+ const struct anv_subpass *subpass = cmd_buffer->state.subpass;
+ const struct anv_framebuffer *fb = cmd_buffer->state.framebuffer;
+
+ if (subpass->depth_stencil_attachment == VK_ATTACHMENT_UNUSED)
+ return NULL;
+
+ const struct anv_image_view *iview =
+ fb->attachments[subpass->depth_stencil_attachment];
+
+ assert(iview->aspect_mask & (VK_IMAGE_ASPECT_DEPTH_BIT |
+ VK_IMAGE_ASPECT_STENCIL_BIT));
+
+ return iview;
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <assert.h>
+#include <stdbool.h>
+#include <string.h>
+#include <unistd.h>
+#include <fcntl.h>
+
+#include "anv_private.h"
+
+/*
+ * Descriptor set layouts.
+ */
+
+VkResult anv_CreateDescriptorSetLayout(
+ VkDevice _device,
+ const VkDescriptorSetLayoutCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkDescriptorSetLayout* pSetLayout)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ struct anv_descriptor_set_layout *set_layout;
+
+ assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO);
+
+ uint32_t max_binding = 0;
+ uint32_t immutable_sampler_count = 0;
+ for (uint32_t j = 0; j < pCreateInfo->bindingCount; j++) {
+ max_binding = MAX2(max_binding, pCreateInfo->pBindings[j].binding);
+ if (pCreateInfo->pBindings[j].pImmutableSamplers)
+ immutable_sampler_count += pCreateInfo->pBindings[j].descriptorCount;
+ }
+
+ size_t size = sizeof(struct anv_descriptor_set_layout) +
+ (max_binding + 1) * sizeof(set_layout->binding[0]) +
+ immutable_sampler_count * sizeof(struct anv_sampler *);
+
+ set_layout = anv_alloc2(&device->alloc, pAllocator, size, 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (!set_layout)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ /* We just allocate all the samplers at the end of the struct */
+ struct anv_sampler **samplers =
+ (struct anv_sampler **)&set_layout->binding[max_binding + 1];
+
+ set_layout->binding_count = max_binding + 1;
+ set_layout->shader_stages = 0;
+ set_layout->size = 0;
+
+ for (uint32_t b = 0; b <= max_binding; b++) {
+ /* Initialize all binding_layout entries to -1 */
+ memset(&set_layout->binding[b], -1, sizeof(set_layout->binding[b]));
+
+ set_layout->binding[b].immutable_samplers = NULL;
+ }
+
+ /* Initialize all samplers to 0 */
+ memset(samplers, 0, immutable_sampler_count * sizeof(*samplers));
+
+ uint32_t sampler_count[MESA_SHADER_STAGES] = { 0, };
+ uint32_t surface_count[MESA_SHADER_STAGES] = { 0, };
+ uint32_t image_count[MESA_SHADER_STAGES] = { 0, };
+ uint32_t buffer_count = 0;
+ uint32_t dynamic_offset_count = 0;
+
+ for (uint32_t j = 0; j < pCreateInfo->bindingCount; j++) {
+ const VkDescriptorSetLayoutBinding *binding = &pCreateInfo->pBindings[j];
+ uint32_t b = binding->binding;
+
+ assert(binding->descriptorCount > 0);
+ set_layout->binding[b].array_size = binding->descriptorCount;
+ set_layout->binding[b].descriptor_index = set_layout->size;
+ set_layout->size += binding->descriptorCount;
+
+ switch (binding->descriptorType) {
+ case VK_DESCRIPTOR_TYPE_SAMPLER:
+ case VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER:
+ anv_foreach_stage(s, binding->stageFlags) {
+ set_layout->binding[b].stage[s].sampler_index = sampler_count[s];
+ sampler_count[s] += binding->descriptorCount;
+ }
+ break;
+ default:
+ break;
+ }
+
+ switch (binding->descriptorType) {
+ case VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER:
+ case VK_DESCRIPTOR_TYPE_STORAGE_BUFFER:
+ case VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC:
+ case VK_DESCRIPTOR_TYPE_STORAGE_BUFFER_DYNAMIC:
+ set_layout->binding[b].buffer_index = buffer_count;
+ buffer_count += binding->descriptorCount;
+ /* fall through */
+
+ case VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER:
+ case VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE:
+ case VK_DESCRIPTOR_TYPE_STORAGE_IMAGE:
+ case VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER:
+ case VK_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER:
+ case VK_DESCRIPTOR_TYPE_INPUT_ATTACHMENT:
+ anv_foreach_stage(s, binding->stageFlags) {
+ set_layout->binding[b].stage[s].surface_index = surface_count[s];
+ surface_count[s] += binding->descriptorCount;
+ }
+ break;
+ default:
+ break;
+ }
+
+ switch (binding->descriptorType) {
+ case VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC:
+ case VK_DESCRIPTOR_TYPE_STORAGE_BUFFER_DYNAMIC:
+ set_layout->binding[b].dynamic_offset_index = dynamic_offset_count;
+ dynamic_offset_count += binding->descriptorCount;
+ break;
+ default:
+ break;
+ }
+
+ switch (binding->descriptorType) {
+ case VK_DESCRIPTOR_TYPE_STORAGE_IMAGE:
+ case VK_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER:
+ anv_foreach_stage(s, binding->stageFlags) {
+ set_layout->binding[b].stage[s].image_index = image_count[s];
+ image_count[s] += binding->descriptorCount;
+ }
+ break;
+ default:
+ break;
+ }
+
+ if (binding->pImmutableSamplers) {
+ set_layout->binding[b].immutable_samplers = samplers;
+ samplers += binding->descriptorCount;
+
+ for (uint32_t i = 0; i < binding->descriptorCount; i++)
+ set_layout->binding[b].immutable_samplers[i] =
+ anv_sampler_from_handle(binding->pImmutableSamplers[i]);
+ } else {
+ set_layout->binding[b].immutable_samplers = NULL;
+ }
+
+ set_layout->shader_stages |= binding->stageFlags;
+ }
+
+ set_layout->buffer_count = buffer_count;
+ set_layout->dynamic_offset_count = dynamic_offset_count;
+
+ *pSetLayout = anv_descriptor_set_layout_to_handle(set_layout);
+
+ return VK_SUCCESS;
+}
+
+void anv_DestroyDescriptorSetLayout(
+ VkDevice _device,
+ VkDescriptorSetLayout _set_layout,
+ const VkAllocationCallbacks* pAllocator)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ ANV_FROM_HANDLE(anv_descriptor_set_layout, set_layout, _set_layout);
+
+ anv_free2(&device->alloc, pAllocator, set_layout);
+}
+
+/*
+ * Pipeline layouts. These have nothing to do with the pipeline. They are
+ * just muttiple descriptor set layouts pasted together
+ */
+
+VkResult anv_CreatePipelineLayout(
+ VkDevice _device,
+ const VkPipelineLayoutCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkPipelineLayout* pPipelineLayout)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ struct anv_pipeline_layout l, *layout;
+
+ assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO);
+
+ l.num_sets = pCreateInfo->setLayoutCount;
+
+ unsigned dynamic_offset_count = 0;
+
+ memset(l.stage, 0, sizeof(l.stage));
+ for (uint32_t set = 0; set < pCreateInfo->setLayoutCount; set++) {
+ ANV_FROM_HANDLE(anv_descriptor_set_layout, set_layout,
+ pCreateInfo->pSetLayouts[set]);
+ l.set[set].layout = set_layout;
+
+ l.set[set].dynamic_offset_start = dynamic_offset_count;
+ for (uint32_t b = 0; b < set_layout->binding_count; b++) {
+ if (set_layout->binding[b].dynamic_offset_index >= 0)
+ dynamic_offset_count += set_layout->binding[b].array_size;
+ }
+
+ for (gl_shader_stage s = 0; s < MESA_SHADER_STAGES; s++) {
+ l.set[set].stage[s].surface_start = l.stage[s].surface_count;
+ l.set[set].stage[s].sampler_start = l.stage[s].sampler_count;
+ l.set[set].stage[s].image_start = l.stage[s].image_count;
+
+ for (uint32_t b = 0; b < set_layout->binding_count; b++) {
+ unsigned array_size = set_layout->binding[b].array_size;
+
+ if (set_layout->binding[b].stage[s].surface_index >= 0) {
+ l.stage[s].surface_count += array_size;
+
+ if (set_layout->binding[b].dynamic_offset_index >= 0)
+ l.stage[s].has_dynamic_offsets = true;
+ }
+
+ if (set_layout->binding[b].stage[s].sampler_index >= 0)
+ l.stage[s].sampler_count += array_size;
+
+ if (set_layout->binding[b].stage[s].image_index >= 0)
+ l.stage[s].image_count += array_size;
+ }
+ }
+ }
+
+ unsigned num_bindings = 0;
+ for (gl_shader_stage s = 0; s < MESA_SHADER_STAGES; s++) {
+ num_bindings += l.stage[s].surface_count +
+ l.stage[s].sampler_count +
+ l.stage[s].image_count;
+ }
+
+ size_t size = sizeof(*layout) + num_bindings * sizeof(layout->entries[0]);
+
+ layout = anv_alloc2(&device->alloc, pAllocator, size, 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (layout == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ /* Now we can actually build our surface and sampler maps */
+ struct anv_pipeline_binding *entry = layout->entries;
+ for (gl_shader_stage s = 0; s < MESA_SHADER_STAGES; s++) {
+ l.stage[s].surface_to_descriptor = entry;
+ entry += l.stage[s].surface_count;
+ l.stage[s].sampler_to_descriptor = entry;
+ entry += l.stage[s].sampler_count;
+ entry += l.stage[s].image_count;
+
+ int surface = 0;
+ int sampler = 0;
+ for (uint32_t set = 0; set < pCreateInfo->setLayoutCount; set++) {
+ struct anv_descriptor_set_layout *set_layout = l.set[set].layout;
+
+ for (uint32_t b = 0; b < set_layout->binding_count; b++) {
+ unsigned array_size = set_layout->binding[b].array_size;
+ unsigned set_offset = set_layout->binding[b].descriptor_index;
+
+ if (set_layout->binding[b].stage[s].surface_index >= 0) {
+ assert(surface == l.set[set].stage[s].surface_start +
+ set_layout->binding[b].stage[s].surface_index);
+ for (unsigned i = 0; i < array_size; i++) {
+ l.stage[s].surface_to_descriptor[surface + i].set = set;
+ l.stage[s].surface_to_descriptor[surface + i].offset = set_offset + i;
+ }
+ surface += array_size;
+ }
+
+ if (set_layout->binding[b].stage[s].sampler_index >= 0) {
+ assert(sampler == l.set[set].stage[s].sampler_start +
+ set_layout->binding[b].stage[s].sampler_index);
+ for (unsigned i = 0; i < array_size; i++) {
+ l.stage[s].sampler_to_descriptor[sampler + i].set = set;
+ l.stage[s].sampler_to_descriptor[sampler + i].offset = set_offset + i;
+ }
+ sampler += array_size;
+ }
+ }
+ }
+ }
+
+ /* Finally, we're done setting it up, copy into the allocated version */
+ *layout = l;
+
+ *pPipelineLayout = anv_pipeline_layout_to_handle(layout);
+
+ return VK_SUCCESS;
+}
+
+void anv_DestroyPipelineLayout(
+ VkDevice _device,
+ VkPipelineLayout _pipelineLayout,
+ const VkAllocationCallbacks* pAllocator)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ ANV_FROM_HANDLE(anv_pipeline_layout, pipeline_layout, _pipelineLayout);
+
+ anv_free2(&device->alloc, pAllocator, pipeline_layout);
+}
+
+/*
+ * Descriptor pools. These are a no-op for now.
+ */
+
+VkResult anv_CreateDescriptorPool(
+ VkDevice device,
+ const VkDescriptorPoolCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkDescriptorPool* pDescriptorPool)
+{
+ anv_finishme("VkDescriptorPool is a stub");
+ *pDescriptorPool = (VkDescriptorPool)1;
+ return VK_SUCCESS;
+}
+
+void anv_DestroyDescriptorPool(
+ VkDevice _device,
+ VkDescriptorPool _pool,
+ const VkAllocationCallbacks* pAllocator)
+{
+ anv_finishme("VkDescriptorPool is a stub: free the pool's descriptor sets");
+}
+
+VkResult anv_ResetDescriptorPool(
+ VkDevice device,
+ VkDescriptorPool descriptorPool,
+ VkDescriptorPoolResetFlags flags)
+{
+ anv_finishme("VkDescriptorPool is a stub: free the pool's descriptor sets");
+ return VK_SUCCESS;
+}
+
+VkResult
+anv_descriptor_set_create(struct anv_device *device,
+ const struct anv_descriptor_set_layout *layout,
+ struct anv_descriptor_set **out_set)
+{
+ struct anv_descriptor_set *set;
+ size_t size = sizeof(*set) + layout->size * sizeof(set->descriptors[0]);
+
+ set = anv_alloc(&device->alloc /* XXX: Use the pool */, size, 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (!set)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ /* A descriptor set may not be 100% filled. Clear the set so we can can
+ * later detect holes in it.
+ */
+ memset(set, 0, size);
+
+ set->layout = layout;
+
+ /* Go through and fill out immutable samplers if we have any */
+ struct anv_descriptor *desc = set->descriptors;
+ for (uint32_t b = 0; b < layout->binding_count; b++) {
+ if (layout->binding[b].immutable_samplers) {
+ for (uint32_t i = 0; i < layout->binding[b].array_size; i++)
+ desc[i].sampler = layout->binding[b].immutable_samplers[i];
+ }
+ desc += layout->binding[b].array_size;
+ }
+
+ /* XXX: Use the pool */
+ set->buffer_views =
+ anv_alloc(&device->alloc,
+ sizeof(set->buffer_views[0]) * layout->buffer_count, 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (!set->buffer_views) {
+ anv_free(&device->alloc, set);
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+ }
+
+ for (uint32_t b = 0; b < layout->buffer_count; b++) {
+ set->buffer_views[b].surface_state =
+ anv_state_pool_alloc(&device->surface_state_pool, 64, 64);
+ }
+
+ *out_set = set;
+
+ return VK_SUCCESS;
+}
+
+void
+anv_descriptor_set_destroy(struct anv_device *device,
+ struct anv_descriptor_set *set)
+{
+ /* XXX: Use the pool */
+ for (uint32_t b = 0; b < set->layout->buffer_count; b++)
+ anv_state_pool_free(&device->surface_state_pool,
+ set->buffer_views[b].surface_state);
+
+ anv_free(&device->alloc, set->buffer_views);
+ anv_free(&device->alloc, set);
+}
+
+VkResult anv_AllocateDescriptorSets(
+ VkDevice _device,
+ const VkDescriptorSetAllocateInfo* pAllocateInfo,
+ VkDescriptorSet* pDescriptorSets)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+
+ VkResult result = VK_SUCCESS;
+ struct anv_descriptor_set *set;
+ uint32_t i;
+
+ for (i = 0; i < pAllocateInfo->descriptorSetCount; i++) {
+ ANV_FROM_HANDLE(anv_descriptor_set_layout, layout,
+ pAllocateInfo->pSetLayouts[i]);
+
+ result = anv_descriptor_set_create(device, layout, &set);
+ if (result != VK_SUCCESS)
+ break;
+
+ pDescriptorSets[i] = anv_descriptor_set_to_handle(set);
+ }
+
+ if (result != VK_SUCCESS)
+ anv_FreeDescriptorSets(_device, pAllocateInfo->descriptorPool,
+ i, pDescriptorSets);
+
+ return result;
+}
+
+VkResult anv_FreeDescriptorSets(
+ VkDevice _device,
+ VkDescriptorPool descriptorPool,
+ uint32_t count,
+ const VkDescriptorSet* pDescriptorSets)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+
+ for (uint32_t i = 0; i < count; i++) {
+ ANV_FROM_HANDLE(anv_descriptor_set, set, pDescriptorSets[i]);
+
+ anv_descriptor_set_destroy(device, set);
+ }
+
+ return VK_SUCCESS;
+}
+
+void anv_UpdateDescriptorSets(
+ VkDevice _device,
+ uint32_t descriptorWriteCount,
+ const VkWriteDescriptorSet* pDescriptorWrites,
+ uint32_t descriptorCopyCount,
+ const VkCopyDescriptorSet* pDescriptorCopies)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+
+ for (uint32_t i = 0; i < descriptorWriteCount; i++) {
+ const VkWriteDescriptorSet *write = &pDescriptorWrites[i];
+ ANV_FROM_HANDLE(anv_descriptor_set, set, write->dstSet);
+ const struct anv_descriptor_set_binding_layout *bind_layout =
+ &set->layout->binding[write->dstBinding];
+ struct anv_descriptor *desc =
+ &set->descriptors[bind_layout->descriptor_index];
+ desc += write->dstArrayElement;
+
+ switch (write->descriptorType) {
+ case VK_DESCRIPTOR_TYPE_SAMPLER:
+ for (uint32_t j = 0; j < write->descriptorCount; j++) {
+ ANV_FROM_HANDLE(anv_sampler, sampler,
+ write->pImageInfo[j].sampler);
+
+ desc[j] = (struct anv_descriptor) {
+ .type = VK_DESCRIPTOR_TYPE_SAMPLER,
+ .sampler = sampler,
+ };
+ }
+ break;
+
+ case VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER:
+ for (uint32_t j = 0; j < write->descriptorCount; j++) {
+ ANV_FROM_HANDLE(anv_image_view, iview,
+ write->pImageInfo[j].imageView);
+ ANV_FROM_HANDLE(anv_sampler, sampler,
+ write->pImageInfo[j].sampler);
+
+ desc[j].type = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER;
+ desc[j].image_view = iview;
+
+ /* If this descriptor has an immutable sampler, we don't want
+ * to stomp on it.
+ */
+ if (sampler)
+ desc[j].sampler = sampler;
+ }
+ break;
+
+ case VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE:
+ case VK_DESCRIPTOR_TYPE_STORAGE_IMAGE:
+ for (uint32_t j = 0; j < write->descriptorCount; j++) {
+ ANV_FROM_HANDLE(anv_image_view, iview,
+ write->pImageInfo[j].imageView);
+
+ desc[j] = (struct anv_descriptor) {
+ .type = write->descriptorType,
+ .image_view = iview,
+ };
+ }
+ break;
+
+ case VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER:
+ case VK_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER:
+ for (uint32_t j = 0; j < write->descriptorCount; j++) {
+ ANV_FROM_HANDLE(anv_buffer_view, bview,
+ write->pTexelBufferView[j]);
+
+ desc[j] = (struct anv_descriptor) {
+ .type = write->descriptorType,
+ .buffer_view = bview,
+ };
+ }
+ break;
+
+ case VK_DESCRIPTOR_TYPE_INPUT_ATTACHMENT:
+ anv_finishme("input attachments not implemented");
+ break;
+
+ case VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER:
+ case VK_DESCRIPTOR_TYPE_STORAGE_BUFFER:
+ case VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC:
+ case VK_DESCRIPTOR_TYPE_STORAGE_BUFFER_DYNAMIC:
+ for (uint32_t j = 0; j < write->descriptorCount; j++) {
+ assert(write->pBufferInfo[j].buffer);
+ ANV_FROM_HANDLE(anv_buffer, buffer, write->pBufferInfo[j].buffer);
+ assert(buffer);
+
+ struct anv_buffer_view *view =
+ &set->buffer_views[bind_layout->buffer_index];
+ view += write->dstArrayElement + j;
+
+ const struct anv_format *format =
+ anv_format_for_descriptor_type(write->descriptorType);
+
+ view->format = format->surface_format;
+ view->bo = buffer->bo;
+ view->offset = buffer->offset + write->pBufferInfo[j].offset;
+
+ /* For buffers with dynamic offsets, we use the full possible
+ * range in the surface state and do the actual range-checking
+ * in the shader.
+ */
+ if (bind_layout->dynamic_offset_index >= 0 ||
+ write->pBufferInfo[j].range == VK_WHOLE_SIZE)
+ view->range = buffer->size - write->pBufferInfo[j].offset;
+ else
+ view->range = write->pBufferInfo[j].range;
+
+ anv_fill_buffer_surface_state(device, view->surface_state.map,
+ view->format,
+ view->offset, view->range, 1);
+
+ if (!device->info.has_llc)
+ anv_state_clflush(view->surface_state);
+
+ desc[j] = (struct anv_descriptor) {
+ .type = write->descriptorType,
+ .buffer_view = view,
+ };
+
+ }
+
+ default:
+ break;
+ }
+ }
+
+ for (uint32_t i = 0; i < descriptorCopyCount; i++) {
+ const VkCopyDescriptorSet *copy = &pDescriptorCopies[i];
+ ANV_FROM_HANDLE(anv_descriptor_set, src, copy->dstSet);
+ ANV_FROM_HANDLE(anv_descriptor_set, dest, copy->dstSet);
+ for (uint32_t j = 0; j < copy->descriptorCount; j++) {
+ dest->descriptors[copy->dstBinding + j] =
+ src->descriptors[copy->srcBinding + j];
+ }
+ }
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <assert.h>
+#include <stdbool.h>
+#include <string.h>
+#include <unistd.h>
+#include <fcntl.h>
+
+#include "anv_private.h"
+#include "mesa/main/git_sha1.h"
+#include "util/strtod.h"
+#include "util/debug.h"
+
+#include "gen7_pack.h"
+
+struct anv_dispatch_table dtable;
+
+static void
+compiler_debug_log(void *data, const char *fmt, ...)
+{ }
+
+static void
+compiler_perf_log(void *data, const char *fmt, ...)
+{
+ va_list args;
+ va_start(args, fmt);
+
+ if (unlikely(INTEL_DEBUG & DEBUG_PERF))
+ vfprintf(stderr, fmt, args);
+
+ va_end(args);
+}
+
+static VkResult
+anv_physical_device_init(struct anv_physical_device *device,
+ struct anv_instance *instance,
+ const char *path)
+{
+ VkResult result;
+ int fd;
+
+ fd = open(path, O_RDWR | O_CLOEXEC);
+ if (fd < 0)
+ return vk_errorf(VK_ERROR_INITIALIZATION_FAILED,
+ "failed to open %s: %m", path);
+
+ device->_loader_data.loaderMagic = ICD_LOADER_MAGIC;
+ device->instance = instance;
+ device->path = path;
+
+ device->chipset_id = anv_gem_get_param(fd, I915_PARAM_CHIPSET_ID);
+ if (!device->chipset_id) {
+ result = vk_errorf(VK_ERROR_INITIALIZATION_FAILED,
+ "failed to get chipset id: %m");
+ goto fail;
+ }
+
+ device->name = brw_get_device_name(device->chipset_id);
+ device->info = brw_get_device_info(device->chipset_id);
+ if (!device->info) {
+ result = vk_errorf(VK_ERROR_INITIALIZATION_FAILED,
+ "failed to get device info");
+ goto fail;
+ }
+
+ if (device->info->is_haswell) {
+ fprintf(stderr, "WARNING: Haswell Vulkan support is incomplete\n");
+ } else if (device->info->gen == 7 && !device->info->is_baytrail) {
+ fprintf(stderr, "WARNING: Ivy Bridge Vulkan support is incomplete\n");
+ } else if (device->info->gen == 7 && device->info->is_baytrail) {
+ fprintf(stderr, "WARNING: Bay Trail Vulkan support is incomplete\n");
+ } else if (device->info->gen >= 8) {
+ /* Broadwell, Cherryview, Skylake, Broxton, Kabylake is as fully
+ * supported as anything */
+ } else {
+ result = vk_errorf(VK_ERROR_INCOMPATIBLE_DRIVER,
+ "Vulkan not yet supported on %s", device->name);
+ goto fail;
+ }
+
+ if (anv_gem_get_aperture(fd, &device->aperture_size) == -1) {
+ result = vk_errorf(VK_ERROR_INITIALIZATION_FAILED,
+ "failed to get aperture size: %m");
+ goto fail;
+ }
+
+ if (!anv_gem_get_param(fd, I915_PARAM_HAS_WAIT_TIMEOUT)) {
+ result = vk_errorf(VK_ERROR_INITIALIZATION_FAILED,
+ "kernel missing gem wait");
+ goto fail;
+ }
+
+ if (!anv_gem_get_param(fd, I915_PARAM_HAS_EXECBUF2)) {
+ result = vk_errorf(VK_ERROR_INITIALIZATION_FAILED,
+ "kernel missing execbuf2");
+ goto fail;
+ }
+
+ if (!device->info->has_llc &&
+ anv_gem_get_param(fd, I915_PARAM_MMAP_VERSION) < 1) {
+ result = vk_errorf(VK_ERROR_INITIALIZATION_FAILED,
+ "kernel missing wc mmap");
+ goto fail;
+ }
+
+ bool swizzled = anv_gem_get_bit6_swizzle(fd, I915_TILING_X);
+
+ close(fd);
+
+ brw_process_intel_debug_variable();
+
+ device->compiler = brw_compiler_create(NULL, device->info);
+ if (device->compiler == NULL) {
+ result = vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+ goto fail;
+ }
+ device->compiler->shader_debug_log = compiler_debug_log;
+ device->compiler->shader_perf_log = compiler_perf_log;
+
+ /* Default to use scalar GS on BDW+ */
+ device->compiler->scalar_stage[MESA_SHADER_GEOMETRY] =
+ device->info->gen >= 8 && env_var_as_boolean("INTEL_SCALAR_GS", true);
+
+ /* XXX: Actually detect bit6 swizzling */
+ isl_device_init(&device->isl_dev, device->info, swizzled);
+
+ return VK_SUCCESS;
+
+fail:
+ close(fd);
+ return result;
+}
+
+static void
+anv_physical_device_finish(struct anv_physical_device *device)
+{
+ ralloc_free(device->compiler);
+}
+
+static const VkExtensionProperties global_extensions[] = {
+ {
+ .extensionName = VK_KHR_SURFACE_EXTENSION_NAME,
+ .specVersion = 24,
+ },
+ {
+ .extensionName = VK_KHR_XCB_SURFACE_EXTENSION_NAME,
+ .specVersion = 5,
+ },
+#ifdef HAVE_WAYLAND_PLATFORM
+ {
+ .extensionName = VK_KHR_WAYLAND_SURFACE_EXTENSION_NAME,
+ .specVersion = 4,
+ },
+#endif
+};
+
+static const VkExtensionProperties device_extensions[] = {
+ {
+ .extensionName = VK_KHR_SWAPCHAIN_EXTENSION_NAME,
+ .specVersion = 67,
+ },
+};
+
+static void *
+default_alloc_func(void *pUserData, size_t size, size_t align,
+ VkSystemAllocationScope allocationScope)
+{
+ return malloc(size);
+}
+
+static void *
+default_realloc_func(void *pUserData, void *pOriginal, size_t size,
+ size_t align, VkSystemAllocationScope allocationScope)
+{
+ return realloc(pOriginal, size);
+}
+
+static void
+default_free_func(void *pUserData, void *pMemory)
+{
+ free(pMemory);
+}
+
+static const VkAllocationCallbacks default_alloc = {
+ .pUserData = NULL,
+ .pfnAllocation = default_alloc_func,
+ .pfnReallocation = default_realloc_func,
+ .pfnFree = default_free_func,
+};
+
+VkResult anv_CreateInstance(
+ const VkInstanceCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkInstance* pInstance)
+{
+ struct anv_instance *instance;
+
+ assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_INSTANCE_CREATE_INFO);
+
+ if (pCreateInfo->pApplicationInfo->apiVersion != VK_MAKE_VERSION(1, 0, 0))
+ return vk_error(VK_ERROR_INCOMPATIBLE_DRIVER);
+
+ for (uint32_t i = 0; i < pCreateInfo->enabledExtensionCount; i++) {
+ bool found = false;
+ for (uint32_t j = 0; j < ARRAY_SIZE(global_extensions); j++) {
+ if (strcmp(pCreateInfo->ppEnabledExtensionNames[i],
+ global_extensions[j].extensionName) == 0) {
+ found = true;
+ break;
+ }
+ }
+ if (!found)
+ return vk_error(VK_ERROR_EXTENSION_NOT_PRESENT);
+ }
+
+ instance = anv_alloc2(&default_alloc, pAllocator, sizeof(*instance), 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_INSTANCE);
+ if (!instance)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ instance->_loader_data.loaderMagic = ICD_LOADER_MAGIC;
+
+ if (pAllocator)
+ instance->alloc = *pAllocator;
+ else
+ instance->alloc = default_alloc;
+
+ instance->apiVersion = pCreateInfo->pApplicationInfo->apiVersion;
+ instance->physicalDeviceCount = -1;
+
+ _mesa_locale_init();
+
+ VG(VALGRIND_CREATE_MEMPOOL(instance, 0, false));
+
+ anv_init_wsi(instance);
+
+ *pInstance = anv_instance_to_handle(instance);
+
+ return VK_SUCCESS;
+}
+
+void anv_DestroyInstance(
+ VkInstance _instance,
+ const VkAllocationCallbacks* pAllocator)
+{
+ ANV_FROM_HANDLE(anv_instance, instance, _instance);
+
+ if (instance->physicalDeviceCount > 0) {
+ /* We support at most one physical device. */
+ assert(instance->physicalDeviceCount == 1);
+ anv_physical_device_finish(&instance->physicalDevice);
+ }
+
+ anv_finish_wsi(instance);
+
+ VG(VALGRIND_DESTROY_MEMPOOL(instance));
+
+ _mesa_locale_fini();
+
+ anv_free(&instance->alloc, instance);
+}
+
+VkResult anv_EnumeratePhysicalDevices(
+ VkInstance _instance,
+ uint32_t* pPhysicalDeviceCount,
+ VkPhysicalDevice* pPhysicalDevices)
+{
+ ANV_FROM_HANDLE(anv_instance, instance, _instance);
+ VkResult result;
+
+ if (instance->physicalDeviceCount < 0) {
+ result = anv_physical_device_init(&instance->physicalDevice,
+ instance, "/dev/dri/renderD128");
+ if (result == VK_ERROR_INCOMPATIBLE_DRIVER) {
+ instance->physicalDeviceCount = 0;
+ } else if (result == VK_SUCCESS) {
+ instance->physicalDeviceCount = 1;
+ } else {
+ return result;
+ }
+ }
+
+ /* pPhysicalDeviceCount is an out parameter if pPhysicalDevices is NULL;
+ * otherwise it's an inout parameter.
+ *
+ * The Vulkan spec (git aaed022) says:
+ *
+ * pPhysicalDeviceCount is a pointer to an unsigned integer variable
+ * that is initialized with the number of devices the application is
+ * prepared to receive handles to. pname:pPhysicalDevices is pointer to
+ * an array of at least this many VkPhysicalDevice handles [...].
+ *
+ * Upon success, if pPhysicalDevices is NULL, vkEnumeratePhysicalDevices
+ * overwrites the contents of the variable pointed to by
+ * pPhysicalDeviceCount with the number of physical devices in in the
+ * instance; otherwise, vkEnumeratePhysicalDevices overwrites
+ * pPhysicalDeviceCount with the number of physical handles written to
+ * pPhysicalDevices.
+ */
+ if (!pPhysicalDevices) {
+ *pPhysicalDeviceCount = instance->physicalDeviceCount;
+ } else if (*pPhysicalDeviceCount >= 1) {
+ pPhysicalDevices[0] = anv_physical_device_to_handle(&instance->physicalDevice);
+ *pPhysicalDeviceCount = 1;
+ } else {
+ *pPhysicalDeviceCount = 0;
+ }
+
+ return VK_SUCCESS;
+}
+
+void anv_GetPhysicalDeviceFeatures(
+ VkPhysicalDevice physicalDevice,
+ VkPhysicalDeviceFeatures* pFeatures)
+{
+ anv_finishme("Get correct values for PhysicalDeviceFeatures");
+
+ *pFeatures = (VkPhysicalDeviceFeatures) {
+ .robustBufferAccess = false,
+ .fullDrawIndexUint32 = false,
+ .imageCubeArray = false,
+ .independentBlend = false,
+ .geometryShader = true,
+ .tessellationShader = false,
+ .sampleRateShading = false,
+ .dualSrcBlend = true,
+ .logicOp = true,
+ .multiDrawIndirect = false,
+ .drawIndirectFirstInstance = false,
+ .depthClamp = false,
+ .depthBiasClamp = false,
+ .fillModeNonSolid = true,
+ .depthBounds = false,
+ .wideLines = true,
+ .largePoints = true,
+ .alphaToOne = true,
+ .multiViewport = true,
+ .samplerAnisotropy = false, /* FINISHME */
+ .textureCompressionETC2 = true,
+ .textureCompressionASTC_LDR = true,
+ .textureCompressionBC = true,
+ .occlusionQueryPrecise = false, /* FINISHME */
+ .pipelineStatisticsQuery = true,
+ .vertexPipelineStoresAndAtomics = false,
+ .fragmentStoresAndAtomics = true,
+ .shaderTessellationAndGeometryPointSize = true,
+ .shaderImageGatherExtended = true,
+ .shaderStorageImageExtendedFormats = false,
+ .shaderStorageImageMultisample = false,
+ .shaderUniformBufferArrayDynamicIndexing = true,
+ .shaderSampledImageArrayDynamicIndexing = false,
+ .shaderStorageBufferArrayDynamicIndexing = false,
+ .shaderStorageImageArrayDynamicIndexing = false,
+ .shaderStorageImageReadWithoutFormat = false,
+ .shaderStorageImageWriteWithoutFormat = true,
+ .shaderClipDistance = false,
+ .shaderCullDistance = false,
+ .shaderFloat64 = false,
+ .shaderInt64 = false,
+ .shaderInt16 = false,
+ .alphaToOne = true,
+ .variableMultisampleRate = false,
+ .inheritedQueries = false,
+ };
+}
+
+void anv_GetPhysicalDeviceProperties(
+ VkPhysicalDevice physicalDevice,
+ VkPhysicalDeviceProperties* pProperties)
+{
+ ANV_FROM_HANDLE(anv_physical_device, pdevice, physicalDevice);
+ const struct brw_device_info *devinfo = pdevice->info;
+
+ anv_finishme("Get correct values for VkPhysicalDeviceLimits");
+
+ const float time_stamp_base = devinfo->gen >= 9 ? 83.333 : 80.0;
+
+ VkSampleCountFlags sample_counts =
+ VK_SAMPLE_COUNT_1_BIT |
+ VK_SAMPLE_COUNT_2_BIT |
+ VK_SAMPLE_COUNT_4_BIT |
+ VK_SAMPLE_COUNT_8_BIT;
+
+ VkPhysicalDeviceLimits limits = {
+ .maxImageDimension1D = (1 << 14),
+ .maxImageDimension2D = (1 << 14),
+ .maxImageDimension3D = (1 << 10),
+ .maxImageDimensionCube = (1 << 14),
+ .maxImageArrayLayers = (1 << 10),
+ .maxTexelBufferElements = (1 << 14),
+ .maxUniformBufferRange = UINT32_MAX,
+ .maxStorageBufferRange = UINT32_MAX,
+ .maxPushConstantsSize = MAX_PUSH_CONSTANTS_SIZE,
+ .maxMemoryAllocationCount = UINT32_MAX,
+ .maxSamplerAllocationCount = 64 * 1024,
+ .bufferImageGranularity = 64, /* A cache line */
+ .sparseAddressSpaceSize = 0,
+ .maxBoundDescriptorSets = MAX_SETS,
+ .maxPerStageDescriptorSamplers = 64,
+ .maxPerStageDescriptorUniformBuffers = 64,
+ .maxPerStageDescriptorStorageBuffers = 64,
+ .maxPerStageDescriptorSampledImages = 64,
+ .maxPerStageDescriptorStorageImages = 64,
+ .maxPerStageDescriptorInputAttachments = 64,
+ .maxPerStageResources = 128,
+ .maxDescriptorSetSamplers = 256,
+ .maxDescriptorSetUniformBuffers = 256,
+ .maxDescriptorSetUniformBuffersDynamic = 256,
+ .maxDescriptorSetStorageBuffers = 256,
+ .maxDescriptorSetStorageBuffersDynamic = 256,
+ .maxDescriptorSetSampledImages = 256,
+ .maxDescriptorSetStorageImages = 256,
+ .maxDescriptorSetInputAttachments = 256,
+ .maxVertexInputAttributes = 32,
+ .maxVertexInputBindings = 32,
+ .maxVertexInputAttributeOffset = 256,
+ .maxVertexInputBindingStride = 256,
+ .maxVertexOutputComponents = 32,
+ .maxTessellationGenerationLevel = 0,
+ .maxTessellationPatchSize = 0,
+ .maxTessellationControlPerVertexInputComponents = 0,
+ .maxTessellationControlPerVertexOutputComponents = 0,
+ .maxTessellationControlPerPatchOutputComponents = 0,
+ .maxTessellationControlTotalOutputComponents = 0,
+ .maxTessellationEvaluationInputComponents = 0,
+ .maxTessellationEvaluationOutputComponents = 0,
+ .maxGeometryShaderInvocations = 6,
+ .maxGeometryInputComponents = 16,
+ .maxGeometryOutputComponents = 16,
+ .maxGeometryOutputVertices = 16,
+ .maxGeometryTotalOutputComponents = 16,
+ .maxFragmentInputComponents = 16,
+ .maxFragmentOutputAttachments = 8,
+ .maxFragmentDualSrcAttachments = 2,
+ .maxFragmentCombinedOutputResources = 8,
+ .maxComputeSharedMemorySize = 1024,
+ .maxComputeWorkGroupCount = {
+ 16 * devinfo->max_cs_threads,
+ 16 * devinfo->max_cs_threads,
+ 16 * devinfo->max_cs_threads,
+ },
+ .maxComputeWorkGroupInvocations = 16 * devinfo->max_cs_threads,
+ .maxComputeWorkGroupSize = {
+ 16 * devinfo->max_cs_threads,
+ 16 * devinfo->max_cs_threads,
+ 16 * devinfo->max_cs_threads,
+ },
+ .subPixelPrecisionBits = 4 /* FIXME */,
+ .subTexelPrecisionBits = 4 /* FIXME */,
+ .mipmapPrecisionBits = 4 /* FIXME */,
+ .maxDrawIndexedIndexValue = UINT32_MAX,
+ .maxDrawIndirectCount = UINT32_MAX,
+ .maxSamplerLodBias = 16,
+ .maxSamplerAnisotropy = 16,
+ .maxViewports = MAX_VIEWPORTS,
+ .maxViewportDimensions = { (1 << 14), (1 << 14) },
+ .viewportBoundsRange = { -1.0, 1.0 }, /* FIXME */
+ .viewportSubPixelBits = 13, /* We take a float? */
+ .minMemoryMapAlignment = 4096, /* A page */
+ .minTexelBufferOffsetAlignment = 1,
+ .minUniformBufferOffsetAlignment = 1,
+ .minStorageBufferOffsetAlignment = 1,
+ .minTexelOffset = 0, /* FIXME */
+ .maxTexelOffset = 0, /* FIXME */
+ .minTexelGatherOffset = 0, /* FIXME */
+ .maxTexelGatherOffset = 0, /* FIXME */
+ .minInterpolationOffset = 0, /* FIXME */
+ .maxInterpolationOffset = 0, /* FIXME */
+ .subPixelInterpolationOffsetBits = 0, /* FIXME */
+ .maxFramebufferWidth = (1 << 14),
+ .maxFramebufferHeight = (1 << 14),
+ .maxFramebufferLayers = (1 << 10),
+ .framebufferColorSampleCounts = sample_counts,
+ .framebufferDepthSampleCounts = sample_counts,
+ .framebufferStencilSampleCounts = sample_counts,
+ .framebufferNoAttachmentsSampleCounts = sample_counts,
+ .maxColorAttachments = MAX_RTS,
+ .sampledImageColorSampleCounts = sample_counts,
+ .sampledImageIntegerSampleCounts = VK_SAMPLE_COUNT_1_BIT,
+ .sampledImageDepthSampleCounts = sample_counts,
+ .sampledImageStencilSampleCounts = sample_counts,
+ .storageImageSampleCounts = VK_SAMPLE_COUNT_1_BIT,
+ .maxSampleMaskWords = 1,
+ .timestampComputeAndGraphics = false,
+ .timestampPeriod = time_stamp_base / (1000 * 1000 * 1000),
+ .maxClipDistances = 0 /* FIXME */,
+ .maxCullDistances = 0 /* FIXME */,
+ .maxCombinedClipAndCullDistances = 0 /* FIXME */,
+ .discreteQueuePriorities = 1,
+ .pointSizeRange = { 0.125, 255.875 },
+ .lineWidthRange = { 0.0, 7.9921875 },
+ .pointSizeGranularity = (1.0 / 8.0),
+ .lineWidthGranularity = (1.0 / 128.0),
+ .strictLines = false, /* FINISHME */
+ .standardSampleLocations = true, /* FINISHME */
+ .optimalBufferCopyOffsetAlignment = 128,
+ .optimalBufferCopyRowPitchAlignment = 128,
+ .nonCoherentAtomSize = 64,
+ };
+
+ *pProperties = (VkPhysicalDeviceProperties) {
+ .apiVersion = VK_MAKE_VERSION(1, 0, 0),
+ .driverVersion = 1,
+ .vendorID = 0x8086,
+ .deviceID = pdevice->chipset_id,
+ .deviceType = VK_PHYSICAL_DEVICE_TYPE_INTEGRATED_GPU,
+ .limits = limits,
+ .sparseProperties = {0}, /* Broadwell doesn't do sparse. */
+ };
+
+ strcpy(pProperties->deviceName, pdevice->name);
+ snprintf((char *)pProperties->pipelineCacheUUID, VK_UUID_SIZE,
+ "anv-%s", MESA_GIT_SHA1 + 4);
+}
+
+void anv_GetPhysicalDeviceQueueFamilyProperties(
+ VkPhysicalDevice physicalDevice,
+ uint32_t* pCount,
+ VkQueueFamilyProperties* pQueueFamilyProperties)
+{
+ if (pQueueFamilyProperties == NULL) {
+ *pCount = 1;
+ return;
+ }
+
+ assert(*pCount >= 1);
+
+ *pQueueFamilyProperties = (VkQueueFamilyProperties) {
+ .queueFlags = VK_QUEUE_GRAPHICS_BIT |
+ VK_QUEUE_COMPUTE_BIT |
+ VK_QUEUE_TRANSFER_BIT,
+ .queueCount = 1,
+ .timestampValidBits = 36, /* XXX: Real value here */
+ .minImageTransferGranularity = (VkExtent3D) { 1, 1, 1 },
+ };
+}
+
+void anv_GetPhysicalDeviceMemoryProperties(
+ VkPhysicalDevice physicalDevice,
+ VkPhysicalDeviceMemoryProperties* pMemoryProperties)
+{
+ ANV_FROM_HANDLE(anv_physical_device, physical_device, physicalDevice);
+ VkDeviceSize heap_size;
+
+ /* Reserve some wiggle room for the driver by exposing only 75% of the
+ * aperture to the heap.
+ */
+ heap_size = 3 * physical_device->aperture_size / 4;
+
+ if (physical_device->info->has_llc) {
+ /* Big core GPUs share LLC with the CPU and thus one memory type can be
+ * both cached and coherent at the same time.
+ */
+ pMemoryProperties->memoryTypeCount = 1;
+ pMemoryProperties->memoryTypes[0] = (VkMemoryType) {
+ .propertyFlags = VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT |
+ VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT |
+ VK_MEMORY_PROPERTY_HOST_COHERENT_BIT |
+ VK_MEMORY_PROPERTY_HOST_CACHED_BIT,
+ .heapIndex = 0,
+ };
+ } else {
+ /* The spec requires that we expose a host-visible, coherent memory
+ * type, but Atom GPUs don't share LLC. Thus we offer two memory types
+ * to give the application a choice between cached, but not coherent and
+ * coherent but uncached (WC though).
+ */
+ pMemoryProperties->memoryTypeCount = 2;
+ pMemoryProperties->memoryTypes[0] = (VkMemoryType) {
+ .propertyFlags = VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT |
+ VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT |
+ VK_MEMORY_PROPERTY_HOST_COHERENT_BIT,
+ .heapIndex = 0,
+ };
+ pMemoryProperties->memoryTypes[1] = (VkMemoryType) {
+ .propertyFlags = VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT |
+ VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT |
+ VK_MEMORY_PROPERTY_HOST_CACHED_BIT,
+ .heapIndex = 0,
+ };
+ }
+
+ pMemoryProperties->memoryHeapCount = 1;
+ pMemoryProperties->memoryHeaps[0] = (VkMemoryHeap) {
+ .size = heap_size,
+ .flags = VK_MEMORY_HEAP_DEVICE_LOCAL_BIT,
+ };
+}
+
+PFN_vkVoidFunction anv_GetInstanceProcAddr(
+ VkInstance instance,
+ const char* pName)
+{
+ return anv_lookup_entrypoint(pName);
+}
+
+PFN_vkVoidFunction anv_GetDeviceProcAddr(
+ VkDevice device,
+ const char* pName)
+{
+ return anv_lookup_entrypoint(pName);
+}
+
+static VkResult
+anv_queue_init(struct anv_device *device, struct anv_queue *queue)
+{
+ queue->_loader_data.loaderMagic = ICD_LOADER_MAGIC;
+ queue->device = device;
+ queue->pool = &device->surface_state_pool;
+
+ return VK_SUCCESS;
+}
+
+static void
+anv_queue_finish(struct anv_queue *queue)
+{
+}
+
+static struct anv_state
+anv_state_pool_emit_data(struct anv_state_pool *pool, size_t size, size_t align, const void *p)
+{
+ struct anv_state state;
+
+ state = anv_state_pool_alloc(pool, size, align);
+ memcpy(state.map, p, size);
+
+ if (!pool->block_pool->device->info.has_llc)
+ anv_state_clflush(state);
+
+ return state;
+}
+
+struct gen8_border_color {
+ union {
+ float float32[4];
+ uint32_t uint32[4];
+ };
+ /* Pad out to 64 bytes */
+ uint32_t _pad[12];
+};
+
+static void
+anv_device_init_border_colors(struct anv_device *device)
+{
+ static const struct gen8_border_color border_colors[] = {
+ [VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK] = { .float32 = { 0.0, 0.0, 0.0, 0.0 } },
+ [VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK] = { .float32 = { 0.0, 0.0, 0.0, 1.0 } },
+ [VK_BORDER_COLOR_FLOAT_OPAQUE_WHITE] = { .float32 = { 1.0, 1.0, 1.0, 1.0 } },
+ [VK_BORDER_COLOR_INT_TRANSPARENT_BLACK] = { .uint32 = { 0, 0, 0, 0 } },
+ [VK_BORDER_COLOR_INT_OPAQUE_BLACK] = { .uint32 = { 0, 0, 0, 1 } },
+ [VK_BORDER_COLOR_INT_OPAQUE_WHITE] = { .uint32 = { 1, 1, 1, 1 } },
+ };
+
+ device->border_colors = anv_state_pool_emit_data(&device->dynamic_state_pool,
+ sizeof(border_colors), 64,
+ border_colors);
+}
+
+VkResult anv_CreateDevice(
+ VkPhysicalDevice physicalDevice,
+ const VkDeviceCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkDevice* pDevice)
+{
+ ANV_FROM_HANDLE(anv_physical_device, physical_device, physicalDevice);
+ VkResult result;
+ struct anv_device *device;
+
+ assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_DEVICE_CREATE_INFO);
+
+ for (uint32_t i = 0; i < pCreateInfo->enabledExtensionCount; i++) {
+ bool found = false;
+ for (uint32_t j = 0; j < ARRAY_SIZE(device_extensions); j++) {
+ if (strcmp(pCreateInfo->ppEnabledExtensionNames[i],
+ device_extensions[j].extensionName) == 0) {
+ found = true;
+ break;
+ }
+ }
+ if (!found)
+ return vk_error(VK_ERROR_EXTENSION_NOT_PRESENT);
+ }
+
+ anv_set_dispatch_devinfo(physical_device->info);
+
+ device = anv_alloc2(&physical_device->instance->alloc, pAllocator,
+ sizeof(*device), 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_DEVICE);
+ if (!device)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ device->_loader_data.loaderMagic = ICD_LOADER_MAGIC;
+ device->instance = physical_device->instance;
+
+ if (pAllocator)
+ device->alloc = *pAllocator;
+ else
+ device->alloc = physical_device->instance->alloc;
+
+ /* XXX(chadv): Can we dup() physicalDevice->fd here? */
+ device->fd = open(physical_device->path, O_RDWR | O_CLOEXEC);
+ if (device->fd == -1) {
+ result = vk_error(VK_ERROR_INITIALIZATION_FAILED);
+ goto fail_device;
+ }
+
+ device->context_id = anv_gem_create_context(device);
+ if (device->context_id == -1) {
+ result = vk_error(VK_ERROR_INITIALIZATION_FAILED);
+ goto fail_fd;
+ }
+
+ device->info = *physical_device->info;
+ device->isl_dev = physical_device->isl_dev;
+
+ pthread_mutex_init(&device->mutex, NULL);
+
+ anv_bo_pool_init(&device->batch_bo_pool, device, ANV_CMD_BUFFER_BATCH_SIZE);
+
+ anv_block_pool_init(&device->dynamic_state_block_pool, device, 16384);
+
+ anv_state_pool_init(&device->dynamic_state_pool,
+ &device->dynamic_state_block_pool);
+
+ anv_block_pool_init(&device->instruction_block_pool, device, 128 * 1024);
+ anv_pipeline_cache_init(&device->default_pipeline_cache, device);
+
+ anv_block_pool_init(&device->surface_state_block_pool, device, 4096);
+
+ anv_state_pool_init(&device->surface_state_pool,
+ &device->surface_state_block_pool);
+
+ anv_bo_init_new(&device->workaround_bo, device, 1024);
+
+ anv_block_pool_init(&device->scratch_block_pool, device, 0x10000);
+
+ anv_queue_init(device, &device->queue);
+
+ result = anv_device_init_meta(device);
+ if (result != VK_SUCCESS)
+ goto fail_fd;
+
+ anv_device_init_border_colors(device);
+
+ *pDevice = anv_device_to_handle(device);
+
+ return VK_SUCCESS;
+
+ fail_fd:
+ close(device->fd);
+ fail_device:
+ anv_free(&device->alloc, device);
+
+ return result;
+}
+
+void anv_DestroyDevice(
+ VkDevice _device,
+ const VkAllocationCallbacks* pAllocator)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+
+ anv_queue_finish(&device->queue);
+
+ anv_device_finish_meta(device);
+
+#ifdef HAVE_VALGRIND
+ /* We only need to free these to prevent valgrind errors. The backing
+ * BO will go away in a couple of lines so we don't actually leak.
+ */
+ anv_state_pool_free(&device->dynamic_state_pool, device->border_colors);
+#endif
+
+ anv_gem_munmap(device->workaround_bo.map, device->workaround_bo.size);
+ anv_gem_close(device, device->workaround_bo.gem_handle);
+
+ anv_bo_pool_finish(&device->batch_bo_pool);
+ anv_state_pool_finish(&device->dynamic_state_pool);
+ anv_block_pool_finish(&device->dynamic_state_block_pool);
+ anv_block_pool_finish(&device->instruction_block_pool);
+ anv_state_pool_finish(&device->surface_state_pool);
+ anv_block_pool_finish(&device->surface_state_block_pool);
+ anv_block_pool_finish(&device->scratch_block_pool);
+
+ close(device->fd);
+
+ pthread_mutex_destroy(&device->mutex);
+
+ anv_free(&device->alloc, device);
+}
+
+VkResult anv_EnumerateInstanceExtensionProperties(
+ const char* pLayerName,
+ uint32_t* pPropertyCount,
+ VkExtensionProperties* pProperties)
+{
+ if (pProperties == NULL) {
+ *pPropertyCount = ARRAY_SIZE(global_extensions);
+ return VK_SUCCESS;
+ }
+
+ assert(*pPropertyCount >= ARRAY_SIZE(global_extensions));
+
+ *pPropertyCount = ARRAY_SIZE(global_extensions);
+ memcpy(pProperties, global_extensions, sizeof(global_extensions));
+
+ return VK_SUCCESS;
+}
+
+VkResult anv_EnumerateDeviceExtensionProperties(
+ VkPhysicalDevice physicalDevice,
+ const char* pLayerName,
+ uint32_t* pPropertyCount,
+ VkExtensionProperties* pProperties)
+{
+ if (pProperties == NULL) {
+ *pPropertyCount = ARRAY_SIZE(device_extensions);
+ return VK_SUCCESS;
+ }
+
+ assert(*pPropertyCount >= ARRAY_SIZE(device_extensions));
+
+ *pPropertyCount = ARRAY_SIZE(device_extensions);
+ memcpy(pProperties, device_extensions, sizeof(device_extensions));
+
+ return VK_SUCCESS;
+}
+
+VkResult anv_EnumerateInstanceLayerProperties(
+ uint32_t* pPropertyCount,
+ VkLayerProperties* pProperties)
+{
+ if (pProperties == NULL) {
+ *pPropertyCount = 0;
+ return VK_SUCCESS;
+ }
+
+ /* None supported at this time */
+ return vk_error(VK_ERROR_LAYER_NOT_PRESENT);
+}
+
+VkResult anv_EnumerateDeviceLayerProperties(
+ VkPhysicalDevice physicalDevice,
+ uint32_t* pPropertyCount,
+ VkLayerProperties* pProperties)
+{
+ if (pProperties == NULL) {
+ *pPropertyCount = 0;
+ return VK_SUCCESS;
+ }
+
+ /* None supported at this time */
+ return vk_error(VK_ERROR_LAYER_NOT_PRESENT);
+}
+
+void anv_GetDeviceQueue(
+ VkDevice _device,
+ uint32_t queueNodeIndex,
+ uint32_t queueIndex,
+ VkQueue* pQueue)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+
+ assert(queueIndex == 0);
+
+ *pQueue = anv_queue_to_handle(&device->queue);
+}
+
+VkResult anv_QueueSubmit(
+ VkQueue _queue,
+ uint32_t submitCount,
+ const VkSubmitInfo* pSubmits,
+ VkFence _fence)
+{
+ ANV_FROM_HANDLE(anv_queue, queue, _queue);
+ ANV_FROM_HANDLE(anv_fence, fence, _fence);
+ struct anv_device *device = queue->device;
+ int ret;
+
+ for (uint32_t i = 0; i < submitCount; i++) {
+ for (uint32_t j = 0; j < pSubmits[i].commandBufferCount; j++) {
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer,
+ pSubmits[i].pCommandBuffers[j]);
+ assert(cmd_buffer->level == VK_COMMAND_BUFFER_LEVEL_PRIMARY);
+
+ ret = anv_gem_execbuffer(device, &cmd_buffer->execbuf2.execbuf);
+ if (ret != 0) {
+ /* We don't know the real error. */
+ return vk_errorf(VK_ERROR_OUT_OF_DEVICE_MEMORY,
+ "execbuf2 failed: %m");
+ }
+
+ if (fence) {
+ ret = anv_gem_execbuffer(device, &fence->execbuf);
+ if (ret != 0) {
+ /* We don't know the real error. */
+ return vk_errorf(VK_ERROR_OUT_OF_DEVICE_MEMORY,
+ "execbuf2 failed: %m");
+ }
+ }
+
+ for (uint32_t k = 0; k < cmd_buffer->execbuf2.bo_count; k++)
+ cmd_buffer->execbuf2.bos[k]->offset = cmd_buffer->execbuf2.objects[k].offset;
+ }
+ }
+
+ return VK_SUCCESS;
+}
+
+VkResult anv_QueueWaitIdle(
+ VkQueue _queue)
+{
+ ANV_FROM_HANDLE(anv_queue, queue, _queue);
+
+ return ANV_CALL(DeviceWaitIdle)(anv_device_to_handle(queue->device));
+}
+
+VkResult anv_DeviceWaitIdle(
+ VkDevice _device)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ struct anv_state state;
+ struct anv_batch batch;
+ struct drm_i915_gem_execbuffer2 execbuf;
+ struct drm_i915_gem_exec_object2 exec2_objects[1];
+ struct anv_bo *bo = NULL;
+ VkResult result;
+ int64_t timeout;
+ int ret;
+
+ state = anv_state_pool_alloc(&device->dynamic_state_pool, 32, 32);
+ bo = &device->dynamic_state_pool.block_pool->bo;
+ batch.start = batch.next = state.map;
+ batch.end = state.map + 32;
+ anv_batch_emit(&batch, GEN7_MI_BATCH_BUFFER_END);
+ anv_batch_emit(&batch, GEN7_MI_NOOP);
+
+ if (!device->info.has_llc)
+ anv_state_clflush(state);
+
+ exec2_objects[0].handle = bo->gem_handle;
+ exec2_objects[0].relocation_count = 0;
+ exec2_objects[0].relocs_ptr = 0;
+ exec2_objects[0].alignment = 0;
+ exec2_objects[0].offset = bo->offset;
+ exec2_objects[0].flags = 0;
+ exec2_objects[0].rsvd1 = 0;
+ exec2_objects[0].rsvd2 = 0;
+
+ execbuf.buffers_ptr = (uintptr_t) exec2_objects;
+ execbuf.buffer_count = 1;
+ execbuf.batch_start_offset = state.offset;
+ execbuf.batch_len = batch.next - state.map;
+ execbuf.cliprects_ptr = 0;
+ execbuf.num_cliprects = 0;
+ execbuf.DR1 = 0;
+ execbuf.DR4 = 0;
+
+ execbuf.flags =
+ I915_EXEC_HANDLE_LUT | I915_EXEC_NO_RELOC | I915_EXEC_RENDER;
+ execbuf.rsvd1 = device->context_id;
+ execbuf.rsvd2 = 0;
+
+ ret = anv_gem_execbuffer(device, &execbuf);
+ if (ret != 0) {
+ /* We don't know the real error. */
+ result = vk_errorf(VK_ERROR_OUT_OF_DEVICE_MEMORY, "execbuf2 failed: %m");
+ goto fail;
+ }
+
+ timeout = INT64_MAX;
+ ret = anv_gem_wait(device, bo->gem_handle, &timeout);
+ if (ret != 0) {
+ /* We don't know the real error. */
+ result = vk_errorf(VK_ERROR_OUT_OF_DEVICE_MEMORY, "execbuf2 failed: %m");
+ goto fail;
+ }
+
+ anv_state_pool_free(&device->dynamic_state_pool, state);
+
+ return VK_SUCCESS;
+
+ fail:
+ anv_state_pool_free(&device->dynamic_state_pool, state);
+
+ return result;
+}
+
+VkResult
+anv_bo_init_new(struct anv_bo *bo, struct anv_device *device, uint64_t size)
+{
+ bo->gem_handle = anv_gem_create(device, size);
+ if (!bo->gem_handle)
+ return vk_error(VK_ERROR_OUT_OF_DEVICE_MEMORY);
+
+ bo->map = NULL;
+ bo->index = 0;
+ bo->offset = 0;
+ bo->size = size;
+
+ return VK_SUCCESS;
+}
+
+VkResult anv_AllocateMemory(
+ VkDevice _device,
+ const VkMemoryAllocateInfo* pAllocateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkDeviceMemory* pMem)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ struct anv_device_memory *mem;
+ VkResult result;
+
+ assert(pAllocateInfo->sType == VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO);
+
+ if (pAllocateInfo->allocationSize == 0) {
+ /* Apparently, this is allowed */
+ *pMem = VK_NULL_HANDLE;
+ return VK_SUCCESS;
+ }
+
+ /* We support exactly one memory heap. */
+ assert(pAllocateInfo->memoryTypeIndex == 0 ||
+ (!device->info.has_llc && pAllocateInfo->memoryTypeIndex < 2));
+
+ /* FINISHME: Fail if allocation request exceeds heap size. */
+
+ mem = anv_alloc2(&device->alloc, pAllocator, sizeof(*mem), 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (mem == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ /* The kernel is going to give us whole pages anyway */
+ uint64_t alloc_size = align_u64(pAllocateInfo->allocationSize, 4096);
+
+ result = anv_bo_init_new(&mem->bo, device, alloc_size);
+ if (result != VK_SUCCESS)
+ goto fail;
+
+ mem->type_index = pAllocateInfo->memoryTypeIndex;
+
+ *pMem = anv_device_memory_to_handle(mem);
+
+ return VK_SUCCESS;
+
+ fail:
+ anv_free2(&device->alloc, pAllocator, mem);
+
+ return result;
+}
+
+void anv_FreeMemory(
+ VkDevice _device,
+ VkDeviceMemory _mem,
+ const VkAllocationCallbacks* pAllocator)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ ANV_FROM_HANDLE(anv_device_memory, mem, _mem);
+
+ if (mem == NULL)
+ return;
+
+ if (mem->bo.map)
+ anv_gem_munmap(mem->bo.map, mem->bo.size);
+
+ if (mem->bo.gem_handle != 0)
+ anv_gem_close(device, mem->bo.gem_handle);
+
+ anv_free2(&device->alloc, pAllocator, mem);
+}
+
+VkResult anv_MapMemory(
+ VkDevice _device,
+ VkDeviceMemory _memory,
+ VkDeviceSize offset,
+ VkDeviceSize size,
+ VkMemoryMapFlags flags,
+ void** ppData)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ ANV_FROM_HANDLE(anv_device_memory, mem, _memory);
+
+ if (mem == NULL) {
+ *ppData = NULL;
+ return VK_SUCCESS;
+ }
+
+ if (size == VK_WHOLE_SIZE)
+ size = mem->bo.size - offset;
+
+ /* FIXME: Is this supposed to be thread safe? Since vkUnmapMemory() only
+ * takes a VkDeviceMemory pointer, it seems like only one map of the memory
+ * at a time is valid. We could just mmap up front and return an offset
+ * pointer here, but that may exhaust virtual memory on 32 bit
+ * userspace. */
+
+ uint32_t gem_flags = 0;
+ if (!device->info.has_llc && mem->type_index == 0)
+ gem_flags |= I915_MMAP_WC;
+
+ /* GEM will fail to map if the offset isn't 4k-aligned. Round down. */
+ uint64_t map_offset = offset & ~4095ull;
+ assert(offset >= map_offset);
+ uint64_t map_size = (offset + size) - map_offset;
+
+ /* Let's map whole pages */
+ map_size = align_u64(map_size, 4096);
+
+ mem->map = anv_gem_mmap(device, mem->bo.gem_handle,
+ map_offset, map_size, gem_flags);
+ mem->map_size = map_size;
+
+ *ppData = mem->map + (offset - map_offset);
+
+ return VK_SUCCESS;
+}
+
+void anv_UnmapMemory(
+ VkDevice _device,
+ VkDeviceMemory _memory)
+{
+ ANV_FROM_HANDLE(anv_device_memory, mem, _memory);
+
+ if (mem == NULL)
+ return;
+
+ anv_gem_munmap(mem->map, mem->map_size);
+}
+
+static void
+clflush_mapped_ranges(struct anv_device *device,
+ uint32_t count,
+ const VkMappedMemoryRange *ranges)
+{
+ for (uint32_t i = 0; i < count; i++) {
+ ANV_FROM_HANDLE(anv_device_memory, mem, ranges[i].memory);
+ void *p = mem->map + (ranges[i].offset & ~CACHELINE_MASK);
+ void *end = mem->map + ranges[i].offset + ranges[i].size;
+
+ while (p < end) {
+ __builtin_ia32_clflush(p);
+ p += CACHELINE_SIZE;
+ }
+ }
+}
+
+VkResult anv_FlushMappedMemoryRanges(
+ VkDevice _device,
+ uint32_t memoryRangeCount,
+ const VkMappedMemoryRange* pMemoryRanges)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+
+ if (device->info.has_llc)
+ return VK_SUCCESS;
+
+ /* Make sure the writes we're flushing have landed. */
+ __builtin_ia32_sfence();
+
+ clflush_mapped_ranges(device, memoryRangeCount, pMemoryRanges);
+
+ return VK_SUCCESS;
+}
+
+VkResult anv_InvalidateMappedMemoryRanges(
+ VkDevice _device,
+ uint32_t memoryRangeCount,
+ const VkMappedMemoryRange* pMemoryRanges)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+
+ if (device->info.has_llc)
+ return VK_SUCCESS;
+
+ clflush_mapped_ranges(device, memoryRangeCount, pMemoryRanges);
+
+ /* Make sure no reads get moved up above the invalidate. */
+ __builtin_ia32_lfence();
+
+ return VK_SUCCESS;
+}
+
+void anv_GetBufferMemoryRequirements(
+ VkDevice device,
+ VkBuffer _buffer,
+ VkMemoryRequirements* pMemoryRequirements)
+{
+ ANV_FROM_HANDLE(anv_buffer, buffer, _buffer);
+
+ /* The Vulkan spec (git aaed022) says:
+ *
+ * memoryTypeBits is a bitfield and contains one bit set for every
+ * supported memory type for the resource. The bit `1<<i` is set if and
+ * only if the memory type `i` in the VkPhysicalDeviceMemoryProperties
+ * structure for the physical device is supported.
+ *
+ * We support exactly one memory type.
+ */
+ pMemoryRequirements->memoryTypeBits = 1;
+
+ pMemoryRequirements->size = buffer->size;
+ pMemoryRequirements->alignment = 16;
+}
+
+void anv_GetImageMemoryRequirements(
+ VkDevice device,
+ VkImage _image,
+ VkMemoryRequirements* pMemoryRequirements)
+{
+ ANV_FROM_HANDLE(anv_image, image, _image);
+
+ /* The Vulkan spec (git aaed022) says:
+ *
+ * memoryTypeBits is a bitfield and contains one bit set for every
+ * supported memory type for the resource. The bit `1<<i` is set if and
+ * only if the memory type `i` in the VkPhysicalDeviceMemoryProperties
+ * structure for the physical device is supported.
+ *
+ * We support exactly one memory type.
+ */
+ pMemoryRequirements->memoryTypeBits = 1;
+
+ pMemoryRequirements->size = image->size;
+ pMemoryRequirements->alignment = image->alignment;
+}
+
+void anv_GetImageSparseMemoryRequirements(
+ VkDevice device,
+ VkImage image,
+ uint32_t* pSparseMemoryRequirementCount,
+ VkSparseImageMemoryRequirements* pSparseMemoryRequirements)
+{
+ stub();
+}
+
+void anv_GetDeviceMemoryCommitment(
+ VkDevice device,
+ VkDeviceMemory memory,
+ VkDeviceSize* pCommittedMemoryInBytes)
+{
+ *pCommittedMemoryInBytes = 0;
+}
+
+VkResult anv_BindBufferMemory(
+ VkDevice device,
+ VkBuffer _buffer,
+ VkDeviceMemory _memory,
+ VkDeviceSize memoryOffset)
+{
+ ANV_FROM_HANDLE(anv_device_memory, mem, _memory);
+ ANV_FROM_HANDLE(anv_buffer, buffer, _buffer);
+
+ if (mem) {
+ buffer->bo = &mem->bo;
+ buffer->offset = memoryOffset;
+ } else {
+ buffer->bo = NULL;
+ buffer->offset = 0;
+ }
+
+ return VK_SUCCESS;
+}
+
+VkResult anv_BindImageMemory(
+ VkDevice device,
+ VkImage _image,
+ VkDeviceMemory _memory,
+ VkDeviceSize memoryOffset)
+{
+ ANV_FROM_HANDLE(anv_device_memory, mem, _memory);
+ ANV_FROM_HANDLE(anv_image, image, _image);
+
+ if (mem) {
+ image->bo = &mem->bo;
+ image->offset = memoryOffset;
+ } else {
+ image->bo = NULL;
+ image->offset = 0;
+ }
+
+ return VK_SUCCESS;
+}
+
+VkResult anv_QueueBindSparse(
+ VkQueue queue,
+ uint32_t bindInfoCount,
+ const VkBindSparseInfo* pBindInfo,
+ VkFence fence)
+{
+ stub_return(VK_ERROR_INCOMPATIBLE_DRIVER);
+}
+
+VkResult anv_CreateFence(
+ VkDevice _device,
+ const VkFenceCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkFence* pFence)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ struct anv_fence *fence;
+ struct anv_batch batch;
+ VkResult result;
+
+ const uint32_t fence_size = 128;
+
+ assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_FENCE_CREATE_INFO);
+
+ fence = anv_alloc2(&device->alloc, pAllocator, sizeof(*fence), 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (fence == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ result = anv_bo_init_new(&fence->bo, device, fence_size);
+ if (result != VK_SUCCESS)
+ goto fail;
+
+ fence->bo.map =
+ anv_gem_mmap(device, fence->bo.gem_handle, 0, fence->bo.size, 0);
+ batch.next = batch.start = fence->bo.map;
+ batch.end = fence->bo.map + fence->bo.size;
+ anv_batch_emit(&batch, GEN7_MI_BATCH_BUFFER_END);
+ anv_batch_emit(&batch, GEN7_MI_NOOP);
+
+ if (!device->info.has_llc) {
+ assert(((uintptr_t) fence->bo.map & CACHELINE_MASK) == 0);
+ assert(batch.next - fence->bo.map <= CACHELINE_SIZE);
+ __builtin_ia32_sfence();
+ __builtin_ia32_clflush(fence->bo.map);
+ }
+
+ fence->exec2_objects[0].handle = fence->bo.gem_handle;
+ fence->exec2_objects[0].relocation_count = 0;
+ fence->exec2_objects[0].relocs_ptr = 0;
+ fence->exec2_objects[0].alignment = 0;
+ fence->exec2_objects[0].offset = fence->bo.offset;
+ fence->exec2_objects[0].flags = 0;
+ fence->exec2_objects[0].rsvd1 = 0;
+ fence->exec2_objects[0].rsvd2 = 0;
+
+ fence->execbuf.buffers_ptr = (uintptr_t) fence->exec2_objects;
+ fence->execbuf.buffer_count = 1;
+ fence->execbuf.batch_start_offset = 0;
+ fence->execbuf.batch_len = batch.next - fence->bo.map;
+ fence->execbuf.cliprects_ptr = 0;
+ fence->execbuf.num_cliprects = 0;
+ fence->execbuf.DR1 = 0;
+ fence->execbuf.DR4 = 0;
+
+ fence->execbuf.flags =
+ I915_EXEC_HANDLE_LUT | I915_EXEC_NO_RELOC | I915_EXEC_RENDER;
+ fence->execbuf.rsvd1 = device->context_id;
+ fence->execbuf.rsvd2 = 0;
+
+ *pFence = anv_fence_to_handle(fence);
+
+ return VK_SUCCESS;
+
+ fail:
+ anv_free2(&device->alloc, pAllocator, fence);
+
+ return result;
+}
+
+void anv_DestroyFence(
+ VkDevice _device,
+ VkFence _fence,
+ const VkAllocationCallbacks* pAllocator)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ ANV_FROM_HANDLE(anv_fence, fence, _fence);
+
+ anv_gem_munmap(fence->bo.map, fence->bo.size);
+ anv_gem_close(device, fence->bo.gem_handle);
+ anv_free2(&device->alloc, pAllocator, fence);
+}
+
+VkResult anv_ResetFences(
+ VkDevice _device,
+ uint32_t fenceCount,
+ const VkFence* pFences)
+{
+ for (uint32_t i = 0; i < fenceCount; i++) {
+ ANV_FROM_HANDLE(anv_fence, fence, pFences[i]);
+ fence->ready = false;
+ }
+
+ return VK_SUCCESS;
+}
+
+VkResult anv_GetFenceStatus(
+ VkDevice _device,
+ VkFence _fence)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ ANV_FROM_HANDLE(anv_fence, fence, _fence);
+ int64_t t = 0;
+ int ret;
+
+ if (fence->ready)
+ return VK_SUCCESS;
+
+ ret = anv_gem_wait(device, fence->bo.gem_handle, &t);
+ if (ret == 0) {
+ fence->ready = true;
+ return VK_SUCCESS;
+ }
+
+ return VK_NOT_READY;
+}
+
+VkResult anv_WaitForFences(
+ VkDevice _device,
+ uint32_t fenceCount,
+ const VkFence* pFences,
+ VkBool32 waitAll,
+ uint64_t timeout)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+
+ /* DRM_IOCTL_I915_GEM_WAIT uses a signed 64 bit timeout and is supposed
+ * to block indefinitely timeouts <= 0. Unfortunately, this was broken
+ * for a couple of kernel releases. Since there's no way to know
+ * whether or not the kernel we're using is one of the broken ones, the
+ * best we can do is to clamp the timeout to INT64_MAX. This limits the
+ * maximum timeout from 584 years to 292 years - likely not a big deal.
+ */
+ if (timeout > INT64_MAX)
+ timeout = INT64_MAX;
+
+ int64_t t = timeout;
+
+ /* FIXME: handle !waitAll */
+
+ for (uint32_t i = 0; i < fenceCount; i++) {
+ ANV_FROM_HANDLE(anv_fence, fence, pFences[i]);
+ int ret = anv_gem_wait(device, fence->bo.gem_handle, &t);
+ if (ret == -1 && errno == ETIME) {
+ return VK_TIMEOUT;
+ } else if (ret == -1) {
+ /* We don't know the real error. */
+ return vk_errorf(VK_ERROR_OUT_OF_DEVICE_MEMORY,
+ "gem wait failed: %m");
+ }
+ }
+
+ return VK_SUCCESS;
+}
+
+// Queue semaphore functions
+
+VkResult anv_CreateSemaphore(
+ VkDevice device,
+ const VkSemaphoreCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkSemaphore* pSemaphore)
+{
+ /* The DRM execbuffer ioctl always execute in-oder, even between different
+ * rings. As such, there's nothing to do for the user space semaphore.
+ */
+
+ *pSemaphore = (VkSemaphore)1;
+
+ return VK_SUCCESS;
+}
+
+void anv_DestroySemaphore(
+ VkDevice device,
+ VkSemaphore semaphore,
+ const VkAllocationCallbacks* pAllocator)
+{
+}
+
+// Event functions
+
+VkResult anv_CreateEvent(
+ VkDevice _device,
+ const VkEventCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkEvent* pEvent)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ struct anv_state state;
+ struct anv_event *event;
+
+ assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_EVENT_CREATE_INFO);
+
+ state = anv_state_pool_alloc(&device->dynamic_state_pool,
+ sizeof(*event), 4);
+ event = state.map;
+ event->state = state;
+ event->semaphore = VK_EVENT_RESET;
+
+ if (!device->info.has_llc) {
+ /* Make sure the writes we're flushing have landed. */
+ __builtin_ia32_sfence();
+ __builtin_ia32_clflush(event);
+ }
+
+ *pEvent = anv_event_to_handle(event);
+
+ return VK_SUCCESS;
+}
+
+void anv_DestroyEvent(
+ VkDevice _device,
+ VkEvent _event,
+ const VkAllocationCallbacks* pAllocator)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ ANV_FROM_HANDLE(anv_event, event, _event);
+
+ anv_state_pool_free(&device->dynamic_state_pool, event->state);
+}
+
+VkResult anv_GetEventStatus(
+ VkDevice _device,
+ VkEvent _event)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ ANV_FROM_HANDLE(anv_event, event, _event);
+
+ if (!device->info.has_llc) {
+ /* Make sure the writes we're flushing have landed. */
+ __builtin_ia32_clflush(event);
+ __builtin_ia32_lfence();
+ }
+
+ return event->semaphore;
+}
+
+VkResult anv_SetEvent(
+ VkDevice _device,
+ VkEvent _event)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ ANV_FROM_HANDLE(anv_event, event, _event);
+
+ event->semaphore = VK_EVENT_SET;
+
+ if (!device->info.has_llc) {
+ /* Make sure the writes we're flushing have landed. */
+ __builtin_ia32_sfence();
+ __builtin_ia32_clflush(event);
+ }
+
+ return VK_SUCCESS;
+}
+
+VkResult anv_ResetEvent(
+ VkDevice _device,
+ VkEvent _event)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ ANV_FROM_HANDLE(anv_event, event, _event);
+
+ event->semaphore = VK_EVENT_RESET;
+
+ if (!device->info.has_llc) {
+ /* Make sure the writes we're flushing have landed. */
+ __builtin_ia32_sfence();
+ __builtin_ia32_clflush(event);
+ }
+
+ return VK_SUCCESS;
+}
+
+// Buffer functions
+
+VkResult anv_CreateBuffer(
+ VkDevice _device,
+ const VkBufferCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkBuffer* pBuffer)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ struct anv_buffer *buffer;
+
+ assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO);
+
+ buffer = anv_alloc2(&device->alloc, pAllocator, sizeof(*buffer), 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (buffer == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ buffer->size = pCreateInfo->size;
+ buffer->usage = pCreateInfo->usage;
+ buffer->bo = NULL;
+ buffer->offset = 0;
+
+ *pBuffer = anv_buffer_to_handle(buffer);
+
+ return VK_SUCCESS;
+}
+
+void anv_DestroyBuffer(
+ VkDevice _device,
+ VkBuffer _buffer,
+ const VkAllocationCallbacks* pAllocator)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ ANV_FROM_HANDLE(anv_buffer, buffer, _buffer);
+
+ anv_free2(&device->alloc, pAllocator, buffer);
+}
+
+void
+anv_fill_buffer_surface_state(struct anv_device *device, void *state,
+ enum isl_format format,
+ uint32_t offset, uint32_t range, uint32_t stride)
+{
+ switch (device->info.gen) {
+ case 7:
+ if (device->info.is_haswell)
+ gen75_fill_buffer_surface_state(state, format, offset, range, stride);
+ else
+ gen7_fill_buffer_surface_state(state, format, offset, range, stride);
+ break;
+ case 8:
+ gen8_fill_buffer_surface_state(state, format, offset, range, stride);
+ break;
+ case 9:
+ gen9_fill_buffer_surface_state(state, format, offset, range, stride);
+ break;
+ default:
+ unreachable("unsupported gen\n");
+ }
+}
+
+void anv_DestroySampler(
+ VkDevice _device,
+ VkSampler _sampler,
+ const VkAllocationCallbacks* pAllocator)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ ANV_FROM_HANDLE(anv_sampler, sampler, _sampler);
+
+ anv_free2(&device->alloc, pAllocator, sampler);
+}
+
+VkResult anv_CreateFramebuffer(
+ VkDevice _device,
+ const VkFramebufferCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkFramebuffer* pFramebuffer)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ struct anv_framebuffer *framebuffer;
+
+ assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO);
+
+ size_t size = sizeof(*framebuffer) +
+ sizeof(struct anv_image_view *) * pCreateInfo->attachmentCount;
+ framebuffer = anv_alloc2(&device->alloc, pAllocator, size, 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (framebuffer == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ framebuffer->attachment_count = pCreateInfo->attachmentCount;
+ for (uint32_t i = 0; i < pCreateInfo->attachmentCount; i++) {
+ VkImageView _iview = pCreateInfo->pAttachments[i];
+ framebuffer->attachments[i] = anv_image_view_from_handle(_iview);
+ }
+
+ framebuffer->width = pCreateInfo->width;
+ framebuffer->height = pCreateInfo->height;
+ framebuffer->layers = pCreateInfo->layers;
+
+ *pFramebuffer = anv_framebuffer_to_handle(framebuffer);
+
+ return VK_SUCCESS;
+}
+
+void anv_DestroyFramebuffer(
+ VkDevice _device,
+ VkFramebuffer _fb,
+ const VkAllocationCallbacks* pAllocator)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ ANV_FROM_HANDLE(anv_framebuffer, fb, _fb);
+
+ anv_free2(&device->alloc, pAllocator, fb);
+}
+
+void vkCmdDbgMarkerBegin(
+ VkCommandBuffer commandBuffer,
+ const char* pMarker)
+ __attribute__ ((visibility ("default")));
+
+void vkCmdDbgMarkerEnd(
+ VkCommandBuffer commandBuffer)
+ __attribute__ ((visibility ("default")));
+
+void vkCmdDbgMarkerBegin(
+ VkCommandBuffer commandBuffer,
+ const char* pMarker)
+{
+}
+
+void vkCmdDbgMarkerEnd(
+ VkCommandBuffer commandBuffer)
+{
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include "anv_private.h"
+
+/* This file contains utility functions for help debugging. They can be
+ * called from GDB or similar to help inspect images and buffers.
+ */
+
+void
+anv_dump_image_to_ppm(struct anv_device *device,
+ struct anv_image *image, unsigned miplevel,
+ unsigned array_layer, const char *filename)
+{
+ VkDevice vk_device = anv_device_to_handle(device);
+ VkResult result;
+
+ VkExtent2D extent = { image->extent.width, image->extent.height };
+ for (unsigned i = 0; i < miplevel; i++) {
+ extent.width = MAX2(1, extent.width / 2);
+ extent.height = MAX2(1, extent.height / 2);
+ }
+
+ VkImage copy_image;
+ result = anv_CreateImage(vk_device,
+ &(VkImageCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO,
+ .imageType = VK_IMAGE_TYPE_2D,
+ .format = VK_FORMAT_R8G8B8A8_UNORM,
+ .extent = (VkExtent3D) { extent.width, extent.height, 1 },
+ .mipLevels = 1,
+ .arrayLayers = 1,
+ .samples = 1,
+ .tiling = VK_IMAGE_TILING_LINEAR,
+ .usage = VK_IMAGE_USAGE_TRANSFER_DST_BIT,
+ .flags = 0,
+ }, NULL, ©_image);
+ assert(result == VK_SUCCESS);
+
+ VkMemoryRequirements reqs;
+ anv_GetImageMemoryRequirements(vk_device, copy_image, &reqs);
+
+ VkDeviceMemory memory;
+ result = anv_AllocateMemory(vk_device,
+ &(VkMemoryAllocateInfo) {
+ .sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO,
+ .allocationSize = reqs.size,
+ .memoryTypeIndex = 0,
+ }, NULL, &memory);
+ assert(result == VK_SUCCESS);
+
+ result = anv_BindImageMemory(vk_device, copy_image, memory, 0);
+ assert(result == VK_SUCCESS);
+
+ VkCommandPool commandPool;
+ result = anv_CreateCommandPool(vk_device,
+ &(VkCommandPoolCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_COMMAND_POOL_CREATE_INFO,
+ .queueFamilyIndex = 0,
+ .flags = 0,
+ }, NULL, &commandPool);
+ assert(result == VK_SUCCESS);
+
+ VkCommandBuffer cmd;
+ result = anv_AllocateCommandBuffers(vk_device,
+ &(VkCommandBufferAllocateInfo) {
+ .sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_ALLOCATE_INFO,
+ .commandPool = commandPool,
+ .level = VK_COMMAND_BUFFER_LEVEL_PRIMARY,
+ .commandBufferCount = 1,
+ }, &cmd);
+ assert(result == VK_SUCCESS);
+
+ result = anv_BeginCommandBuffer(cmd,
+ &(VkCommandBufferBeginInfo) {
+ .sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO,
+ .flags = VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT,
+ });
+ assert(result == VK_SUCCESS);
+
+ anv_CmdBlitImage(cmd,
+ anv_image_to_handle(image), VK_IMAGE_LAYOUT_GENERAL,
+ copy_image, VK_IMAGE_LAYOUT_GENERAL, 1,
+ &(VkImageBlit) {
+ .srcSubresource = {
+ .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
+ .mipLevel = miplevel,
+ .baseArrayLayer = array_layer,
+ .layerCount = 1,
+ },
+ .srcOffsets = {
+ { 0, 0, 0 },
+ { extent.width, extent.height, 1 },
+ },
+ .dstSubresource = {
+ .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
+ .mipLevel = 0,
+ .baseArrayLayer = 0,
+ .layerCount = 1,
+ },
+ .dstOffsets = {
+ { 0, 0, 0 },
+ { extent.width, extent.height, 1 },
+ },
+ }, VK_FILTER_NEAREST);
+
+ ANV_CALL(CmdPipelineBarrier)(cmd,
+ VK_PIPELINE_STAGE_TRANSFER_BIT,
+ VK_PIPELINE_STAGE_TRANSFER_BIT,
+ true, 0, NULL, 0, NULL, 1,
+ &(VkImageMemoryBarrier) {
+ .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
+ .srcAccessMask = VK_ACCESS_HOST_READ_BIT,
+ .dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
+ .oldLayout = VK_IMAGE_LAYOUT_GENERAL,
+ .newLayout = VK_IMAGE_LAYOUT_GENERAL,
+ .srcQueueFamilyIndex = 0,
+ .dstQueueFamilyIndex = 0,
+ .image = copy_image,
+ .subresourceRange = (VkImageSubresourceRange) {
+ .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
+ .baseMipLevel = 0,
+ .levelCount = 1,
+ .baseArrayLayer = 0,
+ .layerCount = 1,
+ },
+ });
+
+ result = anv_EndCommandBuffer(cmd);
+ assert(result == VK_SUCCESS);
+
+ VkFence fence;
+ result = anv_CreateFence(vk_device,
+ &(VkFenceCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_FENCE_CREATE_INFO,
+ .flags = 0,
+ }, NULL, &fence);
+ assert(result == VK_SUCCESS);
+
+ result = anv_QueueSubmit(anv_queue_to_handle(&device->queue), 1,
+ &(VkSubmitInfo) {
+ .sType = VK_STRUCTURE_TYPE_SUBMIT_INFO,
+ .commandBufferCount = 1,
+ .pCommandBuffers = &cmd,
+ }, fence);
+ assert(result == VK_SUCCESS);
+
+ result = anv_WaitForFences(vk_device, 1, &fence, true, UINT64_MAX);
+ assert(result == VK_SUCCESS);
+
+ anv_DestroyFence(vk_device, fence, NULL);
+ anv_DestroyCommandPool(vk_device, commandPool, NULL);
+
+ uint8_t *map;
+ result = anv_MapMemory(vk_device, memory, 0, reqs.size, 0, (void **)&map);
+ assert(result == VK_SUCCESS);
+
+ VkSubresourceLayout layout;
+ anv_GetImageSubresourceLayout(vk_device, copy_image,
+ &(VkImageSubresource) {
+ .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
+ .mipLevel = 0,
+ .arrayLayer = 0,
+ }, &layout);
+
+ map += layout.offset;
+
+ /* Now we can finally write the PPM file */
+ FILE *file = fopen(filename, "wb");
+ assert(file);
+
+ fprintf(file, "P6\n%d %d\n255\n", extent.width, extent.height);
+ for (unsigned y = 0; y < extent.height; y++) {
+ uint8_t row[extent.width * 3];
+ for (unsigned x = 0; x < extent.width; x++) {
+ row[x * 3 + 0] = map[x * 4 + 0];
+ row[x * 3 + 1] = map[x * 4 + 1];
+ row[x * 3 + 2] = map[x * 4 + 2];
+ }
+ fwrite(row, 3, extent.width, file);
+
+ map += layout.rowPitch;
+ }
+ fclose(file);
+
+ anv_UnmapMemory(vk_device, memory);
+ anv_DestroyImage(vk_device, copy_image, NULL);
+ anv_FreeMemory(vk_device, memory, NULL);
+}
--- /dev/null
+# coding=utf-8
+#
+# Copyright © 2015 Intel Corporation
+#
+# Permission is hereby granted, free of charge, to any person obtaining a
+# copy of this software and associated documentation files (the "Software"),
+# to deal in the Software without restriction, including without limitation
+# the rights to use, copy, modify, merge, publish, distribute, sublicense,
+# and/or sell copies of the Software, and to permit persons to whom the
+# Software is furnished to do so, subject to the following conditions:
+#
+# The above copyright notice and this permission notice (including the next
+# paragraph) shall be included in all copies or substantial portions of the
+# Software.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+# THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+# IN THE SOFTWARE.
+#
+
+import fileinput, re, sys
+
+# Each function typedef in the vulkan.h header is all on one line and matches
+# this regepx. We hope that won't change.
+
+p = re.compile('typedef ([^ ]*) *\((?:VKAPI_PTR)? *\*PFN_vk([^(]*)\)(.*);')
+
+entrypoints = []
+
+# We generate a static hash table for entry point lookup
+# (vkGetProcAddress). We use a linear congruential generator for our hash
+# function and a power-of-two size table. The prime numbers are determined
+# experimentally.
+
+none = 0xffff
+hash_size = 256
+u32_mask = 2**32 - 1
+hash_mask = hash_size - 1
+
+prime_factor = 5024183
+prime_step = 19
+
+def hash(name):
+ h = 0;
+ for c in name:
+ h = (h * prime_factor + ord(c)) & u32_mask
+
+ return h
+
+opt_header = False
+opt_code = False
+
+if (sys.argv[1] == "header"):
+ opt_header = True
+ sys.argv.pop()
+elif (sys.argv[1] == "code"):
+ opt_code = True
+ sys.argv.pop()
+
+# Parse the entry points in the header
+
+i = 0
+for line in fileinput.input():
+ m = p.match(line)
+ if (m):
+ if m.group(2) == 'VoidFunction':
+ continue
+ fullname = "vk" + m.group(2)
+ h = hash(fullname)
+ entrypoints.append((m.group(1), m.group(2), m.group(3), i, h))
+ i = i + 1
+
+# For outputting entrypoints.h we generate a anv_EntryPoint() prototype
+# per entry point.
+
+if opt_header:
+ print "/* This file generated from vk_gen.py, don't edit directly. */\n"
+
+ print "struct anv_dispatch_table {"
+ print " union {"
+ print " void *entrypoints[%d];" % len(entrypoints)
+ print " struct {"
+
+ for type, name, args, num, h in entrypoints:
+ print " %s (*%s)%s;" % (type, name, args)
+ print " };\n"
+ print " };\n"
+ print "};\n"
+
+ print "void anv_set_dispatch_devinfo(const struct brw_device_info *info);\n"
+
+ for type, name, args, num, h in entrypoints:
+ print "%s anv_%s%s;" % (type, name, args)
+ print "%s gen7_%s%s;" % (type, name, args)
+ print "%s gen75_%s%s;" % (type, name, args)
+ print "%s gen8_%s%s;" % (type, name, args)
+ print "%s gen9_%s%s;" % (type, name, args)
+ print "%s anv_validate_%s%s;" % (type, name, args)
+ exit()
+
+
+
+print """/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+/* DO NOT EDIT! This is a generated file. */
+
+#include "anv_private.h"
+
+struct anv_entrypoint {
+ uint32_t name;
+ uint32_t hash;
+};
+
+/* We use a big string constant to avoid lots of reloctions from the entry
+ * point table to lots of little strings. The entries in the entry point table
+ * store the index into this big string.
+ */
+
+static const char strings[] ="""
+
+offsets = []
+i = 0;
+for type, name, args, num, h in entrypoints:
+ print " \"vk%s\\0\"" % name
+ offsets.append(i)
+ i += 2 + len(name) + 1
+print """ ;
+
+/* Weak aliases for all potential validate functions. These will resolve to
+ * NULL if they're not defined, which lets the resolve_entrypoint() function
+ * either pick a validate wrapper if available or just plug in the actual
+ * entry point.
+ */
+"""
+
+# Now generate the table of all entry points and their validation functions
+
+print "\nstatic const struct anv_entrypoint entrypoints[] = {"
+for type, name, args, num, h in entrypoints:
+ print " { %5d, 0x%08x }," % (offsets[num], h)
+print "};\n"
+
+for layer in [ "anv", "validate", "gen7", "gen75", "gen8", "gen9" ]:
+ for type, name, args, num, h in entrypoints:
+ print "%s %s_%s%s __attribute__ ((weak));" % (type, layer, name, args)
+ print "\nconst struct anv_dispatch_table %s_layer = {" % layer
+ for type, name, args, num, h in entrypoints:
+ print " .%s = %s_%s," % (name, layer, name)
+ print "};\n"
+
+print """
+#ifdef DEBUG
+static bool enable_validate = true;
+#else
+static bool enable_validate = false;
+#endif
+
+/* We can't use symbols that need resolving (like, oh, getenv) in the resolve
+ * function. This means that we have to determine whether or not to use the
+ * validation layer sometime before that. The constructor function attribute asks
+ * the dynamic linker to invoke determine_validate() at dlopen() time which
+ * works.
+ */
+static void __attribute__ ((constructor))
+determine_validate(void)
+{
+ const char *s = getenv("ANV_VALIDATE");
+
+ if (s)
+ enable_validate = atoi(s);
+}
+
+static const struct brw_device_info *dispatch_devinfo;
+
+void
+anv_set_dispatch_devinfo(const struct brw_device_info *devinfo)
+{
+ dispatch_devinfo = devinfo;
+}
+
+void * __attribute__ ((noinline))
+anv_resolve_entrypoint(uint32_t index)
+{
+ if (enable_validate && validate_layer.entrypoints[index])
+ return validate_layer.entrypoints[index];
+
+ if (dispatch_devinfo == NULL) {
+ assert(anv_layer.entrypoints[index]);
+ return anv_layer.entrypoints[index];
+ }
+
+ switch (dispatch_devinfo->gen) {
+ case 9:
+ if (gen9_layer.entrypoints[index])
+ return gen9_layer.entrypoints[index];
+ /* fall through */
+ case 8:
+ if (gen8_layer.entrypoints[index])
+ return gen8_layer.entrypoints[index];
+ /* fall through */
+ case 7:
+ if (dispatch_devinfo->is_haswell && gen75_layer.entrypoints[index])
+ return gen75_layer.entrypoints[index];
+
+ if (gen7_layer.entrypoints[index])
+ return gen7_layer.entrypoints[index];
+ /* fall through */
+ case 0:
+ return anv_layer.entrypoints[index];
+ default:
+ unreachable("unsupported gen\\n");
+ }
+}
+"""
+
+# Now output ifuncs and their resolve helpers for all entry points. The
+# resolve helper calls resolve_entrypoint() with the entry point index, which
+# lets the resolver look it up in the table.
+
+for type, name, args, num, h in entrypoints:
+ print "static void *resolve_%s(void) { return anv_resolve_entrypoint(%d); }" % (name, num)
+ print "%s vk%s%s\n __attribute__ ((ifunc (\"resolve_%s\"), visibility (\"default\")));\n" % (type, name, args, name)
+
+
+# Now generate the hash table used for entry point look up. This is a
+# uint16_t table of entry point indices. We use 0xffff to indicate an entry
+# in the hash table is empty.
+
+map = [none for f in xrange(hash_size)]
+collisions = [0 for f in xrange(10)]
+for type, name, args, num, h in entrypoints:
+ level = 0
+ while map[h & hash_mask] != none:
+ h = h + prime_step
+ level = level + 1
+ if level > 9:
+ collisions[9] += 1
+ else:
+ collisions[level] += 1
+ map[h & hash_mask] = num
+
+print "/* Hash table stats:"
+print " * size %d entries" % hash_size
+print " * collisions entries"
+for i in xrange(10):
+ if (i == 9):
+ plus = "+"
+ else:
+ plus = " "
+
+ print " * %2d%s %4d" % (i, plus, collisions[i])
+print " */\n"
+
+print "#define none 0x%04x\n" % none
+
+print "static const uint16_t map[] = {"
+for i in xrange(0, hash_size, 8):
+ print " ",
+ for j in xrange(i, i + 8):
+ if map[j] & 0xffff == 0xffff:
+ print " none,",
+ else:
+ print "0x%04x," % (map[j] & 0xffff),
+ print
+
+print "};"
+
+# Finally we generate the hash table lookup function. The hash function and
+# linear probing algorithm matches the hash table generated above.
+
+print """
+void *
+anv_lookup_entrypoint(const char *name)
+{
+ static const uint32_t prime_factor = %d;
+ static const uint32_t prime_step = %d;
+ const struct anv_entrypoint *e;
+ uint32_t hash, h, i;
+ const char *p;
+
+ hash = 0;
+ for (p = name; *p; p++)
+ hash = hash * prime_factor + *p;
+
+ h = hash;
+ do {
+ i = map[h & %d];
+ if (i == none)
+ return NULL;
+ e = &entrypoints[i];
+ h += prime_step;
+ } while (e->hash != hash);
+
+ if (strcmp(name, strings + e->name) != 0)
+ return NULL;
+
+ return anv_resolve_entrypoint(i);
+}
+""" % (prime_factor, prime_step, hash_mask)
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include "anv_private.h"
+#include "brw_surface_formats.h"
+
+#include "gen7_pack.h"
+
+#define fmt(__vk_fmt, __hw_fmt, ...) \
+ [__vk_fmt] = { \
+ .vk_format = __vk_fmt, \
+ .name = #__vk_fmt, \
+ .surface_format = __hw_fmt, \
+ .isl_layout = &isl_format_layouts[__hw_fmt], \
+ __VA_ARGS__ \
+ }
+
+static const struct anv_format anv_formats[] = {
+ fmt(VK_FORMAT_UNDEFINED, ISL_FORMAT_RAW),
+ fmt(VK_FORMAT_R4G4_UNORM_PACK8, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_R4G4B4A4_UNORM_PACK16, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_B4G4R4A4_UNORM_PACK16, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_R5G6B5_UNORM_PACK16, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_B5G6R5_UNORM_PACK16, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_R5G5B5A1_UNORM_PACK16, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_B5G5R5A1_UNORM_PACK16, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_A1R5G5B5_UNORM_PACK16, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_R8_UNORM, ISL_FORMAT_R8_UNORM),
+ fmt(VK_FORMAT_R8_SNORM, ISL_FORMAT_R8_SNORM),
+ fmt(VK_FORMAT_R8_USCALED, ISL_FORMAT_R8_USCALED),
+ fmt(VK_FORMAT_R8_SSCALED, ISL_FORMAT_R8_SSCALED),
+ fmt(VK_FORMAT_R8_UINT, ISL_FORMAT_R8_UINT),
+ fmt(VK_FORMAT_R8_SINT, ISL_FORMAT_R8_SINT),
+ fmt(VK_FORMAT_R8_SRGB, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_R8G8_UNORM, ISL_FORMAT_R8G8_UNORM),
+ fmt(VK_FORMAT_R8G8_SNORM, ISL_FORMAT_R8G8_SNORM),
+ fmt(VK_FORMAT_R8G8_USCALED, ISL_FORMAT_R8G8_USCALED),
+ fmt(VK_FORMAT_R8G8_SSCALED, ISL_FORMAT_R8G8_SSCALED),
+ fmt(VK_FORMAT_R8G8_UINT, ISL_FORMAT_R8G8_UINT),
+ fmt(VK_FORMAT_R8G8_SINT, ISL_FORMAT_R8G8_SINT),
+ fmt(VK_FORMAT_R8G8_SRGB, ISL_FORMAT_UNSUPPORTED), /* L8A8_UNORM_SRGB */
+ fmt(VK_FORMAT_R8G8B8_UNORM, ISL_FORMAT_R8G8B8_UNORM),
+ fmt(VK_FORMAT_R8G8B8_SNORM, ISL_FORMAT_R8G8B8_SNORM),
+ fmt(VK_FORMAT_R8G8B8_USCALED, ISL_FORMAT_R8G8B8_USCALED),
+ fmt(VK_FORMAT_R8G8B8_SSCALED, ISL_FORMAT_R8G8B8_SSCALED),
+ fmt(VK_FORMAT_R8G8B8_UINT, ISL_FORMAT_R8G8B8_UINT),
+ fmt(VK_FORMAT_R8G8B8_SINT, ISL_FORMAT_R8G8B8_SINT),
+ fmt(VK_FORMAT_R8G8B8_SRGB, ISL_FORMAT_UNSUPPORTED), /* B8G8R8A8_UNORM_SRGB */
+ fmt(VK_FORMAT_R8G8B8A8_UNORM, ISL_FORMAT_R8G8B8A8_UNORM),
+ fmt(VK_FORMAT_R8G8B8A8_SNORM, ISL_FORMAT_R8G8B8A8_SNORM),
+ fmt(VK_FORMAT_R8G8B8A8_USCALED, ISL_FORMAT_R8G8B8A8_USCALED),
+ fmt(VK_FORMAT_R8G8B8A8_SSCALED, ISL_FORMAT_R8G8B8A8_SSCALED),
+ fmt(VK_FORMAT_R8G8B8A8_UINT, ISL_FORMAT_R8G8B8A8_UINT),
+ fmt(VK_FORMAT_R8G8B8A8_SINT, ISL_FORMAT_R8G8B8A8_SINT),
+ fmt(VK_FORMAT_R8G8B8A8_SRGB, ISL_FORMAT_R8G8B8A8_UNORM_SRGB),
+ fmt(VK_FORMAT_A8B8G8R8_UNORM_PACK32, ISL_FORMAT_R8G8B8A8_UNORM),
+ fmt(VK_FORMAT_A8B8G8R8_SNORM_PACK32, ISL_FORMAT_R8G8B8A8_SNORM),
+ fmt(VK_FORMAT_A8B8G8R8_USCALED_PACK32, ISL_FORMAT_R8G8B8A8_USCALED),
+ fmt(VK_FORMAT_A8B8G8R8_SSCALED_PACK32, ISL_FORMAT_R8G8B8A8_SSCALED),
+ fmt(VK_FORMAT_A8B8G8R8_UINT_PACK32, ISL_FORMAT_R8G8B8A8_UINT),
+ fmt(VK_FORMAT_A8B8G8R8_SINT_PACK32, ISL_FORMAT_R8G8B8A8_SINT),
+ fmt(VK_FORMAT_A8B8G8R8_SRGB_PACK32, ISL_FORMAT_R8G8B8A8_UNORM_SRGB),
+ fmt(VK_FORMAT_A2R10G10B10_UNORM_PACK32, ISL_FORMAT_B10G10R10A2_UNORM),
+ fmt(VK_FORMAT_A2R10G10B10_SNORM_PACK32, ISL_FORMAT_B10G10R10A2_SNORM),
+ fmt(VK_FORMAT_A2R10G10B10_USCALED_PACK32, ISL_FORMAT_B10G10R10A2_USCALED),
+ fmt(VK_FORMAT_A2R10G10B10_SSCALED_PACK32, ISL_FORMAT_B10G10R10A2_SSCALED),
+ fmt(VK_FORMAT_A2R10G10B10_UINT_PACK32, ISL_FORMAT_B10G10R10A2_UINT),
+ fmt(VK_FORMAT_A2R10G10B10_SINT_PACK32, ISL_FORMAT_B10G10R10A2_SINT),
+ fmt(VK_FORMAT_A2B10G10R10_UNORM_PACK32, ISL_FORMAT_R10G10B10A2_UNORM),
+ fmt(VK_FORMAT_A2B10G10R10_SNORM_PACK32, ISL_FORMAT_R10G10B10A2_SNORM),
+ fmt(VK_FORMAT_A2B10G10R10_USCALED_PACK32, ISL_FORMAT_R10G10B10A2_USCALED),
+ fmt(VK_FORMAT_A2B10G10R10_SSCALED_PACK32, ISL_FORMAT_R10G10B10A2_SSCALED),
+ fmt(VK_FORMAT_A2B10G10R10_UINT_PACK32, ISL_FORMAT_R10G10B10A2_UINT),
+ fmt(VK_FORMAT_A2B10G10R10_SINT_PACK32, ISL_FORMAT_R10G10B10A2_SINT),
+ fmt(VK_FORMAT_R16_UNORM, ISL_FORMAT_R16_UNORM),
+ fmt(VK_FORMAT_R16_SNORM, ISL_FORMAT_R16_SNORM),
+ fmt(VK_FORMAT_R16_USCALED, ISL_FORMAT_R16_USCALED),
+ fmt(VK_FORMAT_R16_SSCALED, ISL_FORMAT_R16_SSCALED),
+ fmt(VK_FORMAT_R16_UINT, ISL_FORMAT_R16_UINT),
+ fmt(VK_FORMAT_R16_SINT, ISL_FORMAT_R16_SINT),
+ fmt(VK_FORMAT_R16_SFLOAT, ISL_FORMAT_R16_FLOAT),
+ fmt(VK_FORMAT_R16G16_UNORM, ISL_FORMAT_R16G16_UNORM),
+ fmt(VK_FORMAT_R16G16_SNORM, ISL_FORMAT_R16G16_SNORM),
+ fmt(VK_FORMAT_R16G16_USCALED, ISL_FORMAT_R16G16_USCALED),
+ fmt(VK_FORMAT_R16G16_SSCALED, ISL_FORMAT_R16G16_SSCALED),
+ fmt(VK_FORMAT_R16G16_UINT, ISL_FORMAT_R16G16_UINT),
+ fmt(VK_FORMAT_R16G16_SINT, ISL_FORMAT_R16G16_SINT),
+ fmt(VK_FORMAT_R16G16_SFLOAT, ISL_FORMAT_R16G16_FLOAT),
+ fmt(VK_FORMAT_R16G16B16_UNORM, ISL_FORMAT_R16G16B16_UNORM),
+ fmt(VK_FORMAT_R16G16B16_SNORM, ISL_FORMAT_R16G16B16_SNORM),
+ fmt(VK_FORMAT_R16G16B16_USCALED, ISL_FORMAT_R16G16B16_USCALED),
+ fmt(VK_FORMAT_R16G16B16_SSCALED, ISL_FORMAT_R16G16B16_SSCALED),
+ fmt(VK_FORMAT_R16G16B16_UINT, ISL_FORMAT_R16G16B16_UINT),
+ fmt(VK_FORMAT_R16G16B16_SINT, ISL_FORMAT_R16G16B16_SINT),
+ fmt(VK_FORMAT_R16G16B16_SFLOAT, ISL_FORMAT_R16G16B16_FLOAT),
+ fmt(VK_FORMAT_R16G16B16A16_UNORM, ISL_FORMAT_R16G16B16A16_UNORM),
+ fmt(VK_FORMAT_R16G16B16A16_SNORM, ISL_FORMAT_R16G16B16A16_SNORM),
+ fmt(VK_FORMAT_R16G16B16A16_USCALED, ISL_FORMAT_R16G16B16A16_USCALED),
+ fmt(VK_FORMAT_R16G16B16A16_SSCALED, ISL_FORMAT_R16G16B16A16_SSCALED),
+ fmt(VK_FORMAT_R16G16B16A16_UINT, ISL_FORMAT_R16G16B16A16_UINT),
+ fmt(VK_FORMAT_R16G16B16A16_SINT, ISL_FORMAT_R16G16B16A16_SINT),
+ fmt(VK_FORMAT_R16G16B16A16_SFLOAT, ISL_FORMAT_R16G16B16A16_FLOAT),
+ fmt(VK_FORMAT_R32_UINT, ISL_FORMAT_R32_UINT,),
+ fmt(VK_FORMAT_R32_SINT, ISL_FORMAT_R32_SINT,),
+ fmt(VK_FORMAT_R32_SFLOAT, ISL_FORMAT_R32_FLOAT,),
+ fmt(VK_FORMAT_R32G32_UINT, ISL_FORMAT_R32G32_UINT,),
+ fmt(VK_FORMAT_R32G32_SINT, ISL_FORMAT_R32G32_SINT,),
+ fmt(VK_FORMAT_R32G32_SFLOAT, ISL_FORMAT_R32G32_FLOAT,),
+ fmt(VK_FORMAT_R32G32B32_UINT, ISL_FORMAT_R32G32B32_UINT,),
+ fmt(VK_FORMAT_R32G32B32_SINT, ISL_FORMAT_R32G32B32_SINT,),
+ fmt(VK_FORMAT_R32G32B32_SFLOAT, ISL_FORMAT_R32G32B32_FLOAT,),
+ fmt(VK_FORMAT_R32G32B32A32_UINT, ISL_FORMAT_R32G32B32A32_UINT,),
+ fmt(VK_FORMAT_R32G32B32A32_SINT, ISL_FORMAT_R32G32B32A32_SINT,),
+ fmt(VK_FORMAT_R32G32B32A32_SFLOAT, ISL_FORMAT_R32G32B32A32_FLOAT,),
+ fmt(VK_FORMAT_R64_UINT, ISL_FORMAT_R64_PASSTHRU),
+ fmt(VK_FORMAT_R64_SINT, ISL_FORMAT_R64_PASSTHRU),
+ fmt(VK_FORMAT_R64_SFLOAT, ISL_FORMAT_R64_FLOAT),
+ fmt(VK_FORMAT_R64G64_UINT, ISL_FORMAT_R64G64_PASSTHRU),
+ fmt(VK_FORMAT_R64G64_SINT, ISL_FORMAT_R64G64_PASSTHRU),
+ fmt(VK_FORMAT_R64G64_SFLOAT, ISL_FORMAT_R64G64_FLOAT),
+ fmt(VK_FORMAT_R64G64B64_UINT, ISL_FORMAT_R64G64B64_PASSTHRU),
+ fmt(VK_FORMAT_R64G64B64_SINT, ISL_FORMAT_R64G64B64_PASSTHRU),
+ fmt(VK_FORMAT_R64G64B64_SFLOAT, ISL_FORMAT_R64G64B64_FLOAT),
+ fmt(VK_FORMAT_R64G64B64A64_UINT, ISL_FORMAT_R64G64B64A64_PASSTHRU),
+ fmt(VK_FORMAT_R64G64B64A64_SINT, ISL_FORMAT_R64G64B64A64_PASSTHRU),
+ fmt(VK_FORMAT_R64G64B64A64_SFLOAT, ISL_FORMAT_R64G64B64A64_FLOAT),
+ fmt(VK_FORMAT_B10G11R11_UFLOAT_PACK32, ISL_FORMAT_R11G11B10_FLOAT),
+ fmt(VK_FORMAT_E5B9G9R9_UFLOAT_PACK32, ISL_FORMAT_R9G9B9E5_SHAREDEXP),
+
+ fmt(VK_FORMAT_D16_UNORM, ISL_FORMAT_R16_UNORM, .depth_format = D16_UNORM),
+ fmt(VK_FORMAT_X8_D24_UNORM_PACK32, ISL_FORMAT_R24_UNORM_X8_TYPELESS, .depth_format = D24_UNORM_X8_UINT),
+ fmt(VK_FORMAT_D32_SFLOAT, ISL_FORMAT_R32_FLOAT, .depth_format = D32_FLOAT),
+ fmt(VK_FORMAT_S8_UINT, ISL_FORMAT_R8_UINT, .has_stencil = true),
+ fmt(VK_FORMAT_D16_UNORM_S8_UINT, ISL_FORMAT_R16_UNORM, .depth_format = D16_UNORM, .has_stencil = true),
+ fmt(VK_FORMAT_D24_UNORM_S8_UINT, ISL_FORMAT_R24_UNORM_X8_TYPELESS, .depth_format = D24_UNORM_X8_UINT, .has_stencil = true),
+ fmt(VK_FORMAT_D32_SFLOAT_S8_UINT, ISL_FORMAT_R32_FLOAT, .depth_format = D32_FLOAT, .has_stencil = true),
+
+ fmt(VK_FORMAT_BC1_RGB_UNORM_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_BC1_RGB_SRGB_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_BC1_RGBA_UNORM_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_BC1_RGBA_SRGB_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_BC2_UNORM_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_BC2_SRGB_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_BC3_UNORM_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_BC3_SRGB_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_BC4_UNORM_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_BC4_SNORM_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_BC5_UNORM_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_BC5_SNORM_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_BC6H_UFLOAT_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_BC6H_SFLOAT_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_BC7_UNORM_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_BC7_SRGB_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ETC2_R8G8B8_UNORM_BLOCK, ISL_FORMAT_ETC2_RGB8),
+ fmt(VK_FORMAT_ETC2_R8G8B8_SRGB_BLOCK, ISL_FORMAT_ETC2_SRGB8),
+ fmt(VK_FORMAT_ETC2_R8G8B8A1_UNORM_BLOCK, ISL_FORMAT_ETC2_RGB8_PTA),
+ fmt(VK_FORMAT_ETC2_R8G8B8A1_SRGB_BLOCK, ISL_FORMAT_ETC2_SRGB8_PTA),
+ fmt(VK_FORMAT_ETC2_R8G8B8A8_UNORM_BLOCK, ISL_FORMAT_ETC2_EAC_RGBA8),
+ fmt(VK_FORMAT_ETC2_R8G8B8A8_SRGB_BLOCK, ISL_FORMAT_ETC2_EAC_SRGB8_A8),
+ fmt(VK_FORMAT_EAC_R11_UNORM_BLOCK, ISL_FORMAT_EAC_R11),
+ fmt(VK_FORMAT_EAC_R11_SNORM_BLOCK, ISL_FORMAT_EAC_SIGNED_R11),
+ fmt(VK_FORMAT_EAC_R11G11_UNORM_BLOCK, ISL_FORMAT_EAC_RG11),
+ fmt(VK_FORMAT_EAC_R11G11_SNORM_BLOCK, ISL_FORMAT_EAC_SIGNED_RG11),
+ fmt(VK_FORMAT_ASTC_4x4_UNORM_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ASTC_4x4_SRGB_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ASTC_5x4_UNORM_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ASTC_5x4_SRGB_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ASTC_5x5_UNORM_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ASTC_5x5_SRGB_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ASTC_6x5_UNORM_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ASTC_6x5_SRGB_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ASTC_6x6_UNORM_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ASTC_6x6_SRGB_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ASTC_8x5_UNORM_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ASTC_8x5_SRGB_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ASTC_8x6_UNORM_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ASTC_8x6_SRGB_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ASTC_8x8_UNORM_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ASTC_8x8_SRGB_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ASTC_10x5_UNORM_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ASTC_10x5_SRGB_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ASTC_10x6_UNORM_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ASTC_10x6_SRGB_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ASTC_10x8_UNORM_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ASTC_10x8_SRGB_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ASTC_10x10_UNORM_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ASTC_10x10_SRGB_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ASTC_12x10_UNORM_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ASTC_12x10_SRGB_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ASTC_12x12_UNORM_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_ASTC_12x12_SRGB_BLOCK, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_B8G8R8_UNORM, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_B8G8R8_SNORM, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_B8G8R8_USCALED, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_B8G8R8_SSCALED, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_B8G8R8_UINT, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_B8G8R8_SINT, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_B8G8R8_SRGB, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_B8G8R8A8_UNORM, ISL_FORMAT_B8G8R8A8_UNORM),
+ fmt(VK_FORMAT_B8G8R8A8_SNORM, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_B8G8R8A8_USCALED, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_B8G8R8A8_SSCALED, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_B8G8R8A8_UINT, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_B8G8R8A8_SINT, ISL_FORMAT_UNSUPPORTED),
+ fmt(VK_FORMAT_B8G8R8A8_SRGB, ISL_FORMAT_B8G8R8A8_UNORM_SRGB),
+};
+
+#undef fmt
+
+const struct anv_format *
+anv_format_for_vk_format(VkFormat format)
+{
+ return &anv_formats[format];
+}
+
+/**
+ * Exactly one bit must be set in \a aspect.
+ */
+enum isl_format
+anv_get_isl_format(VkFormat format, VkImageAspectFlags aspect,
+ VkImageTiling tiling)
+{
+ const struct anv_format *anv_fmt = &anv_formats[format];
+
+ switch (aspect) {
+ case VK_IMAGE_ASPECT_COLOR_BIT:
+ if (anv_fmt->surface_format == ISL_FORMAT_UNSUPPORTED) {
+ return ISL_FORMAT_UNSUPPORTED;
+ } else if (tiling == VK_IMAGE_TILING_OPTIMAL &&
+ !util_is_power_of_two(anv_fmt->isl_layout->bs)) {
+ /* Tiled formats *must* be power-of-two because we need up upload
+ * them with the render pipeline. For 3-channel formats, we fix
+ * this by switching them over to RGBX or RGBA formats under the
+ * hood.
+ */
+ enum isl_format rgbx = isl_format_rgb_to_rgbx(anv_fmt->surface_format);
+ if (rgbx != ISL_FORMAT_UNSUPPORTED)
+ return rgbx;
+ else
+ return isl_format_rgb_to_rgba(anv_fmt->surface_format);
+ } else {
+ return anv_fmt->surface_format;
+ }
+
+ case VK_IMAGE_ASPECT_DEPTH_BIT:
+ case (VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT):
+ assert(anv_fmt->depth_format != 0);
+ return anv_fmt->surface_format;
+
+ case VK_IMAGE_ASPECT_STENCIL_BIT:
+ assert(anv_fmt->has_stencil);
+ return ISL_FORMAT_R8_UINT;
+
+ default:
+ unreachable("bad VkImageAspect");
+ return ISL_FORMAT_UNSUPPORTED;
+ }
+}
+
+// Format capabilities
+
+void anv_validate_GetPhysicalDeviceFormatProperties(
+ VkPhysicalDevice physicalDevice,
+ VkFormat _format,
+ VkFormatProperties* pFormatProperties)
+{
+ const struct anv_format *format = anv_format_for_vk_format(_format);
+ fprintf(stderr, "vkGetFormatProperties(%s)\n", format->name);
+ anv_GetPhysicalDeviceFormatProperties(physicalDevice, _format, pFormatProperties);
+}
+
+static VkFormatFeatureFlags
+get_image_format_properties(int gen, enum isl_format base,
+ enum isl_format actual)
+{
+ const struct brw_surface_format_info *info = &surface_formats[actual];
+
+ if (actual == ISL_FORMAT_UNSUPPORTED || !info->exists)
+ return 0;
+
+ VkFormatFeatureFlags flags = 0;
+ if (info->sampling <= gen) {
+ flags |= VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT |
+ VK_FORMAT_FEATURE_BLIT_SRC_BIT;
+ }
+
+ if (info->render_target <= gen) {
+ flags |= VK_FORMAT_FEATURE_COLOR_ATTACHMENT_BIT |
+ VK_FORMAT_FEATURE_BLIT_DST_BIT;
+ }
+
+ if (info->alpha_blend <= gen)
+ flags |= VK_FORMAT_FEATURE_COLOR_ATTACHMENT_BLEND_BIT;
+
+ /* Load/store is determined based on base format. This prevents RGB
+ * formats from showing up as load/store capable.
+ */
+ if (isl_is_storage_image_format(base))
+ flags |= VK_FORMAT_FEATURE_STORAGE_IMAGE_BIT;
+
+ if (base == ISL_FORMAT_R32_SINT || base == ISL_FORMAT_R32_UINT)
+ flags |= VK_FORMAT_FEATURE_STORAGE_IMAGE_ATOMIC_BIT;
+
+ return flags;
+}
+
+static VkFormatFeatureFlags
+get_buffer_format_properties(int gen, enum isl_format format)
+{
+ const struct brw_surface_format_info *info = &surface_formats[format];
+
+ if (format == ISL_FORMAT_UNSUPPORTED || !info->exists)
+ return 0;
+
+ VkFormatFeatureFlags flags = 0;
+ if (info->sampling <= gen && !isl_format_is_compressed(format))
+ flags |= VK_FORMAT_FEATURE_UNIFORM_TEXEL_BUFFER_BIT;
+
+ if (info->input_vb <= gen)
+ flags |= VK_FORMAT_FEATURE_VERTEX_BUFFER_BIT;
+
+ if (isl_is_storage_image_format(format))
+ flags |= VK_FORMAT_FEATURE_STORAGE_TEXEL_BUFFER_BIT;
+
+ if (format == ISL_FORMAT_R32_SINT || format == ISL_FORMAT_R32_UINT)
+ flags |= VK_FORMAT_FEATURE_STORAGE_TEXEL_BUFFER_ATOMIC_BIT;
+
+ return flags;
+}
+
+static void
+anv_physical_device_get_format_properties(struct anv_physical_device *physical_device,
+ VkFormat format,
+ VkFormatProperties *out_properties)
+{
+ int gen = physical_device->info->gen * 10;
+ if (physical_device->info->is_haswell)
+ gen += 5;
+
+ VkFormatFeatureFlags linear = 0, tiled = 0, buffer = 0;
+ if (anv_format_is_depth_or_stencil(&anv_formats[format])) {
+ tiled |= VK_FORMAT_FEATURE_DEPTH_STENCIL_ATTACHMENT_BIT;
+ if (physical_device->info->gen >= 8) {
+ tiled |= VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT;
+ tiled |= VK_FORMAT_FEATURE_BLIT_SRC_BIT;
+ }
+ if (anv_formats[format].depth_format) {
+ tiled |= VK_FORMAT_FEATURE_BLIT_DST_BIT;
+ }
+ } else {
+ enum isl_format linear_fmt, tiled_fmt;
+ linear_fmt = anv_get_isl_format(format, VK_IMAGE_ASPECT_COLOR_BIT,
+ VK_IMAGE_TILING_LINEAR);
+ tiled_fmt = anv_get_isl_format(format, VK_IMAGE_ASPECT_COLOR_BIT,
+ VK_IMAGE_TILING_OPTIMAL);
+
+ linear = get_image_format_properties(gen, linear_fmt, linear_fmt);
+ tiled = get_image_format_properties(gen, linear_fmt, tiled_fmt);
+ buffer = get_buffer_format_properties(gen, linear_fmt);
+
+ /* XXX: We handle 3-channel formats by switching them out for RGBX or
+ * RGBA formats behind-the-scenes. This works fine for textures
+ * because the upload process will fill in the extra channel.
+ * We could also support it for render targets, but it will take
+ * substantially more work and we have enough RGBX formats to handle
+ * what most clients will want.
+ */
+ if (linear_fmt != ISL_FORMAT_UNSUPPORTED &&
+ isl_format_is_rgb(linear_fmt) &&
+ isl_format_rgb_to_rgbx(linear_fmt) == ISL_FORMAT_UNSUPPORTED) {
+ tiled &= ~VK_FORMAT_FEATURE_COLOR_ATTACHMENT_BIT &
+ ~VK_FORMAT_FEATURE_BLIT_DST_BIT;
+ }
+ }
+
+ out_properties->linearTilingFeatures = linear;
+ out_properties->optimalTilingFeatures = tiled;
+ out_properties->bufferFeatures = buffer;
+
+ return;
+}
+
+
+void anv_GetPhysicalDeviceFormatProperties(
+ VkPhysicalDevice physicalDevice,
+ VkFormat format,
+ VkFormatProperties* pFormatProperties)
+{
+ ANV_FROM_HANDLE(anv_physical_device, physical_device, physicalDevice);
+
+ anv_physical_device_get_format_properties(
+ physical_device,
+ format,
+ pFormatProperties);
+}
+
+VkResult anv_GetPhysicalDeviceImageFormatProperties(
+ VkPhysicalDevice physicalDevice,
+ VkFormat format,
+ VkImageType type,
+ VkImageTiling tiling,
+ VkImageUsageFlags usage,
+ VkImageCreateFlags flags,
+ VkImageFormatProperties* pImageFormatProperties)
+{
+ ANV_FROM_HANDLE(anv_physical_device, physical_device, physicalDevice);
+ VkFormatProperties format_props;
+ VkFormatFeatureFlags format_feature_flags;
+ VkExtent3D maxExtent;
+ uint32_t maxMipLevels;
+ uint32_t maxArraySize;
+
+ anv_physical_device_get_format_properties(physical_device, format,
+ &format_props);
+
+ /* Extract the VkFormatFeatureFlags that are relevant for the queried
+ * tiling.
+ */
+ if (tiling == VK_IMAGE_TILING_LINEAR) {
+ format_feature_flags = format_props.linearTilingFeatures;
+ } else if (tiling == VK_IMAGE_TILING_OPTIMAL) {
+ format_feature_flags = format_props.optimalTilingFeatures;
+ } else {
+ unreachable("bad VkImageTiling");
+ }
+
+ switch (type) {
+ default:
+ unreachable("bad VkImageType");
+ case VK_IMAGE_TYPE_1D:
+ maxExtent.width = 16384;
+ maxExtent.height = 1;
+ maxExtent.depth = 1;
+ maxMipLevels = 15; /* log2(maxWidth) + 1 */
+ maxArraySize = 2048;
+ break;
+ case VK_IMAGE_TYPE_2D:
+ /* FINISHME: Does this really differ for cube maps? The documentation
+ * for RENDER_SURFACE_STATE suggests so.
+ */
+ maxExtent.width = 16384;
+ maxExtent.height = 16384;
+ maxExtent.depth = 1;
+ maxMipLevels = 15; /* log2(maxWidth) + 1 */
+ maxArraySize = 2048;
+ break;
+ case VK_IMAGE_TYPE_3D:
+ maxExtent.width = 2048;
+ maxExtent.height = 2048;
+ maxExtent.depth = 2048;
+ maxMipLevels = 12; /* log2(maxWidth) + 1 */
+ maxArraySize = 1;
+ break;
+ }
+
+ if (usage & VK_IMAGE_USAGE_TRANSFER_SRC_BIT) {
+ /* Meta implements transfers by sampling from the source image. */
+ if (!(format_feature_flags & VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT)) {
+ goto unsupported;
+ }
+ }
+
+ if (usage & VK_IMAGE_USAGE_TRANSFER_DST_BIT) {
+ if (anv_format_for_vk_format(format)->has_stencil) {
+ /* Not yet implemented because copying to a W-tiled surface is crazy
+ * hard.
+ */
+ anv_finishme("support VK_IMAGE_USAGE_TRANSFER_DST_BIT for "
+ "stencil format");
+ goto unsupported;
+ }
+ }
+
+ if (usage & VK_IMAGE_USAGE_SAMPLED_BIT) {
+ if (!(format_feature_flags & VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT)) {
+ goto unsupported;
+ }
+ }
+
+ if (usage & VK_IMAGE_USAGE_STORAGE_BIT) {
+ if (!(format_feature_flags & VK_FORMAT_FEATURE_STORAGE_IMAGE_BIT)) {
+ goto unsupported;
+ }
+ }
+
+ if (usage & VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT) {
+ if (!(format_feature_flags & VK_FORMAT_FEATURE_COLOR_ATTACHMENT_BIT)) {
+ goto unsupported;
+ }
+ }
+
+ if (usage & VK_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
+ if (!(format_feature_flags & VK_FORMAT_FEATURE_DEPTH_STENCIL_ATTACHMENT_BIT)) {
+ goto unsupported;
+ }
+ }
+
+ if (usage & VK_IMAGE_USAGE_TRANSIENT_ATTACHMENT_BIT) {
+ /* Nothing to check. */
+ }
+
+ if (usage & VK_IMAGE_USAGE_INPUT_ATTACHMENT_BIT) {
+ /* Ignore this flag because it was removed from the
+ * provisional_I_20150910 header.
+ */
+ }
+
+ *pImageFormatProperties = (VkImageFormatProperties) {
+ .maxExtent = maxExtent,
+ .maxMipLevels = maxMipLevels,
+ .maxArrayLayers = maxArraySize,
+
+ /* FINISHME: Support multisampling */
+ .sampleCounts = VK_SAMPLE_COUNT_1_BIT,
+
+ /* FINISHME: Accurately calculate
+ * VkImageFormatProperties::maxResourceSize.
+ */
+ .maxResourceSize = UINT32_MAX,
+ };
+
+ return VK_SUCCESS;
+
+unsupported:
+ *pImageFormatProperties = (VkImageFormatProperties) {
+ .maxExtent = { 0, 0, 0 },
+ .maxMipLevels = 0,
+ .maxArrayLayers = 0,
+ .sampleCounts = 0,
+ .maxResourceSize = 0,
+ };
+
+ return VK_SUCCESS;
+}
+
+void anv_GetPhysicalDeviceSparseImageFormatProperties(
+ VkPhysicalDevice physicalDevice,
+ VkFormat format,
+ VkImageType type,
+ uint32_t samples,
+ VkImageUsageFlags usage,
+ VkImageTiling tiling,
+ uint32_t* pNumProperties,
+ VkSparseImageFormatProperties* pProperties)
+{
+ /* Sparse images are not yet supported. */
+ *pNumProperties = 0;
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#define _DEFAULT_SOURCE
+
+#include <sys/ioctl.h>
+#include <sys/mman.h>
+#include <string.h>
+#include <errno.h>
+#include <unistd.h>
+#include <fcntl.h>
+
+#include "anv_private.h"
+
+#define VG_CLEAR(s) VG(memset(&s, 0, sizeof(s)))
+
+static int
+anv_ioctl(int fd, unsigned long request, void *arg)
+{
+ int ret;
+
+ do {
+ ret = ioctl(fd, request, arg);
+ } while (ret == -1 && (errno == EINTR || errno == EAGAIN));
+
+ return ret;
+}
+
+/**
+ * Wrapper around DRM_IOCTL_I915_GEM_CREATE.
+ *
+ * Return gem handle, or 0 on failure. Gem handles are never 0.
+ */
+uint32_t
+anv_gem_create(struct anv_device *device, size_t size)
+{
+ struct drm_i915_gem_create gem_create;
+ int ret;
+
+ VG_CLEAR(gem_create);
+ gem_create.size = size;
+
+ ret = anv_ioctl(device->fd, DRM_IOCTL_I915_GEM_CREATE, &gem_create);
+ if (ret != 0) {
+ /* FIXME: What do we do if this fails? */
+ return 0;
+ }
+
+ return gem_create.handle;
+}
+
+void
+anv_gem_close(struct anv_device *device, uint32_t gem_handle)
+{
+ struct drm_gem_close close;
+
+ VG_CLEAR(close);
+ close.handle = gem_handle;
+ anv_ioctl(device->fd, DRM_IOCTL_GEM_CLOSE, &close);
+}
+
+/**
+ * Wrapper around DRM_IOCTL_I915_GEM_MMAP.
+ */
+void*
+anv_gem_mmap(struct anv_device *device, uint32_t gem_handle,
+ uint64_t offset, uint64_t size, uint32_t flags)
+{
+ struct drm_i915_gem_mmap gem_mmap;
+ int ret;
+
+ gem_mmap.handle = gem_handle;
+ VG_CLEAR(gem_mmap.pad);
+ gem_mmap.offset = offset;
+ gem_mmap.size = size;
+ VG_CLEAR(gem_mmap.addr_ptr);
+ gem_mmap.flags = flags;
+
+ ret = anv_ioctl(device->fd, DRM_IOCTL_I915_GEM_MMAP, &gem_mmap);
+ if (ret != 0) {
+ /* FIXME: Is NULL the right error return? Cf MAP_INVALID */
+ return NULL;
+ }
+
+ VG(VALGRIND_MALLOCLIKE_BLOCK(gem_mmap.addr_ptr, gem_mmap.size, 0, 1));
+ return (void *)(uintptr_t) gem_mmap.addr_ptr;
+}
+
+/* This is just a wrapper around munmap, but it also notifies valgrind that
+ * this map is no longer valid. Pair this with anv_gem_mmap().
+ */
+void
+anv_gem_munmap(void *p, uint64_t size)
+{
+ VG(VALGRIND_FREELIKE_BLOCK(p, 0));
+ munmap(p, size);
+}
+
+uint32_t
+anv_gem_userptr(struct anv_device *device, void *mem, size_t size)
+{
+ struct drm_i915_gem_userptr userptr;
+ int ret;
+
+ VG_CLEAR(userptr);
+ userptr.user_ptr = (__u64)((unsigned long) mem);
+ userptr.user_size = size;
+ userptr.flags = 0;
+
+ ret = anv_ioctl(device->fd, DRM_IOCTL_I915_GEM_USERPTR, &userptr);
+ if (ret == -1)
+ return 0;
+
+ return userptr.handle;
+}
+
+int
+anv_gem_set_caching(struct anv_device *device,
+ uint32_t gem_handle, uint32_t caching)
+{
+ struct drm_i915_gem_caching gem_caching;
+
+ VG_CLEAR(gem_caching);
+ gem_caching.handle = gem_handle;
+ gem_caching.caching = caching;
+
+ return anv_ioctl(device->fd, DRM_IOCTL_I915_GEM_SET_CACHING, &gem_caching);
+}
+
+int
+anv_gem_set_domain(struct anv_device *device, uint32_t gem_handle,
+ uint32_t read_domains, uint32_t write_domain)
+{
+ struct drm_i915_gem_set_domain gem_set_domain;
+
+ VG_CLEAR(gem_set_domain);
+ gem_set_domain.handle = gem_handle;
+ gem_set_domain.read_domains = read_domains;
+ gem_set_domain.write_domain = write_domain;
+
+ return anv_ioctl(device->fd, DRM_IOCTL_I915_GEM_SET_DOMAIN, &gem_set_domain);
+}
+
+/**
+ * On error, \a timeout_ns holds the remaining time.
+ */
+int
+anv_gem_wait(struct anv_device *device, uint32_t gem_handle, int64_t *timeout_ns)
+{
+ struct drm_i915_gem_wait wait;
+ int ret;
+
+ VG_CLEAR(wait);
+ wait.bo_handle = gem_handle;
+ wait.timeout_ns = *timeout_ns;
+ wait.flags = 0;
+
+ ret = anv_ioctl(device->fd, DRM_IOCTL_I915_GEM_WAIT, &wait);
+ *timeout_ns = wait.timeout_ns;
+
+ return ret;
+}
+
+int
+anv_gem_execbuffer(struct anv_device *device,
+ struct drm_i915_gem_execbuffer2 *execbuf)
+{
+ return anv_ioctl(device->fd, DRM_IOCTL_I915_GEM_EXECBUFFER2, execbuf);
+}
+
+int
+anv_gem_set_tiling(struct anv_device *device,
+ uint32_t gem_handle, uint32_t stride, uint32_t tiling)
+{
+ struct drm_i915_gem_set_tiling set_tiling;
+ int ret;
+
+ /* set_tiling overwrites the input on the error path, so we have to open
+ * code anv_ioctl.
+ */
+
+ do {
+ VG_CLEAR(set_tiling);
+ set_tiling.handle = gem_handle;
+ set_tiling.tiling_mode = tiling;
+ set_tiling.stride = stride;
+
+ ret = ioctl(device->fd, DRM_IOCTL_I915_GEM_SET_TILING, &set_tiling);
+ } while (ret == -1 && (errno == EINTR || errno == EAGAIN));
+
+ return ret;
+}
+
+int
+anv_gem_get_param(int fd, uint32_t param)
+{
+ drm_i915_getparam_t gp;
+ int ret, tmp;
+
+ VG_CLEAR(gp);
+ gp.param = param;
+ gp.value = &tmp;
+ ret = anv_ioctl(fd, DRM_IOCTL_I915_GETPARAM, &gp);
+ if (ret == 0)
+ return tmp;
+
+ return 0;
+}
+
+bool
+anv_gem_get_bit6_swizzle(int fd, uint32_t tiling)
+{
+ struct drm_gem_close close;
+ int ret;
+
+ struct drm_i915_gem_create gem_create;
+ VG_CLEAR(gem_create);
+ gem_create.size = 4096;
+
+ if (anv_ioctl(fd, DRM_IOCTL_I915_GEM_CREATE, &gem_create)) {
+ assert(!"Failed to create GEM BO");
+ return false;
+ }
+
+ bool swizzled = false;
+
+ /* set_tiling overwrites the input on the error path, so we have to open
+ * code anv_ioctl.
+ */
+ struct drm_i915_gem_set_tiling set_tiling;
+ do {
+ VG_CLEAR(set_tiling);
+ set_tiling.handle = gem_create.handle;
+ set_tiling.tiling_mode = tiling;
+ set_tiling.stride = tiling == I915_TILING_X ? 512 : 128;
+
+ ret = ioctl(fd, DRM_IOCTL_I915_GEM_SET_TILING, &set_tiling);
+ } while (ret == -1 && (errno == EINTR || errno == EAGAIN));
+
+ if (ret != 0) {
+ assert(!"Failed to set BO tiling");
+ goto close_and_return;
+ }
+
+ struct drm_i915_gem_get_tiling get_tiling;
+ VG_CLEAR(get_tiling);
+ get_tiling.handle = gem_create.handle;
+
+ if (anv_ioctl(fd, DRM_IOCTL_I915_GEM_GET_TILING, &get_tiling)) {
+ assert(!"Failed to get BO tiling");
+ goto close_and_return;
+ }
+
+ swizzled = get_tiling.swizzle_mode != I915_BIT_6_SWIZZLE_NONE;
+
+close_and_return:
+
+ VG_CLEAR(close);
+ close.handle = gem_create.handle;
+ anv_ioctl(fd, DRM_IOCTL_GEM_CLOSE, &close);
+
+ return swizzled;
+}
+
+int
+anv_gem_create_context(struct anv_device *device)
+{
+ struct drm_i915_gem_context_create create;
+ int ret;
+
+ VG_CLEAR(create);
+
+ ret = anv_ioctl(device->fd, DRM_IOCTL_I915_GEM_CONTEXT_CREATE, &create);
+ if (ret == -1)
+ return -1;
+
+ return create.ctx_id;
+}
+
+int
+anv_gem_destroy_context(struct anv_device *device, int context)
+{
+ struct drm_i915_gem_context_destroy destroy;
+
+ VG_CLEAR(destroy);
+ destroy.ctx_id = context;
+
+ return anv_ioctl(device->fd, DRM_IOCTL_I915_GEM_CONTEXT_DESTROY, &destroy);
+}
+
+int
+anv_gem_get_aperture(int fd, uint64_t *size)
+{
+ struct drm_i915_gem_get_aperture aperture;
+ int ret;
+
+ VG_CLEAR(aperture);
+ ret = anv_ioctl(fd, DRM_IOCTL_I915_GEM_GET_APERTURE, &aperture);
+ if (ret == -1)
+ return -1;
+
+ *size = aperture.aper_available_size;
+
+ return 0;
+}
+
+int
+anv_gem_handle_to_fd(struct anv_device *device, uint32_t gem_handle)
+{
+ struct drm_prime_handle args;
+ int ret;
+
+ VG_CLEAR(args);
+ args.handle = gem_handle;
+ args.flags = DRM_CLOEXEC;
+
+ ret = anv_ioctl(device->fd, DRM_IOCTL_PRIME_HANDLE_TO_FD, &args);
+ if (ret == -1)
+ return -1;
+
+ return args.fd;
+}
+
+uint32_t
+anv_gem_fd_to_handle(struct anv_device *device, int fd)
+{
+ struct drm_prime_handle args;
+ int ret;
+
+ VG_CLEAR(args);
+ args.fd = fd;
+
+ ret = anv_ioctl(device->fd, DRM_IOCTL_PRIME_FD_TO_HANDLE, &args);
+ if (ret == -1)
+ return 0;
+
+ return args.handle;
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#define _DEFAULT_SOURCE
+
+#include <linux/memfd.h>
+#include <sys/mman.h>
+#include <sys/syscall.h>
+
+#include "anv_private.h"
+
+static inline int
+memfd_create(const char *name, unsigned int flags)
+{
+ return syscall(SYS_memfd_create, name, flags);
+}
+
+uint32_t
+anv_gem_create(struct anv_device *device, size_t size)
+{
+ int fd = memfd_create("fake bo", MFD_CLOEXEC);
+ if (fd == -1)
+ return 0;
+
+ assert(fd != 0);
+
+ if (ftruncate(fd, size) == -1)
+ return 0;
+
+ return fd;
+}
+
+void
+anv_gem_close(struct anv_device *device, uint32_t gem_handle)
+{
+ close(gem_handle);
+}
+
+void*
+anv_gem_mmap(struct anv_device *device, uint32_t gem_handle,
+ uint64_t offset, uint64_t size, uint32_t flags)
+{
+ /* Ignore flags, as they're specific to I915_GEM_MMAP. */
+ (void) flags;
+
+ return mmap(NULL, size, PROT_READ | PROT_WRITE, MAP_SHARED,
+ gem_handle, offset);
+}
+
+/* This is just a wrapper around munmap, but it also notifies valgrind that
+ * this map is no longer valid. Pair this with anv_gem_mmap().
+ */
+void
+anv_gem_munmap(void *p, uint64_t size)
+{
+ munmap(p, size);
+}
+
+uint32_t
+anv_gem_userptr(struct anv_device *device, void *mem, size_t size)
+{
+ return -1;
+}
+
+int
+anv_gem_wait(struct anv_device *device, uint32_t gem_handle, int64_t *timeout_ns)
+{
+ return 0;
+}
+
+int
+anv_gem_execbuffer(struct anv_device *device,
+ struct drm_i915_gem_execbuffer2 *execbuf)
+{
+ return 0;
+}
+
+int
+anv_gem_set_tiling(struct anv_device *device,
+ uint32_t gem_handle, uint32_t stride, uint32_t tiling)
+{
+ return 0;
+}
+
+int
+anv_gem_set_caching(struct anv_device *device, uint32_t gem_handle,
+ uint32_t caching)
+{
+ return 0;
+}
+
+int
+anv_gem_set_domain(struct anv_device *device, uint32_t gem_handle,
+ uint32_t read_domains, uint32_t write_domain)
+{
+ return 0;
+}
+
+int
+anv_gem_get_param(int fd, uint32_t param)
+{
+ unreachable("Unused");
+}
+
+bool
+anv_gem_get_bit6_swizzle(int fd, uint32_t tiling)
+{
+ unreachable("Unused");
+}
+
+int
+anv_gem_create_context(struct anv_device *device)
+{
+ unreachable("Unused");
+}
+
+int
+anv_gem_destroy_context(struct anv_device *device, int context)
+{
+ unreachable("Unused");
+}
+
+int
+anv_gem_get_aperture(int fd, uint64_t *size)
+{
+ unreachable("Unused");
+}
+
+int
+anv_gem_handle_to_fd(struct anv_device *device, uint32_t gem_handle)
+{
+ unreachable("Unused");
+}
+
+uint32_t
+anv_gem_fd_to_handle(struct anv_device *device, int fd)
+{
+ unreachable("Unused");
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#pragma once
+
+/* Macros for handling per-gen compilation.
+ *
+ * The prefixing macros GENX() and genX() automatically prefix whatever you
+ * give them by GENX_ or genX_ where X is the gen number.
+ *
+ * You can declare a function to be used on some range of gens like this:
+ *
+ * GENX_FUNC(GEN7, GEN75) void
+ * genX(my_function_name)(args...)
+ * {
+ * // Do stuff
+ * }
+ *
+ * If the file is compiled for any set of gens containing gen7 and gen75,
+ * the function will effectively only get compiled twice as
+ * gen7_my_function_nmae and gen75_my_function_name. The function has to
+ * be compilable on all gens, but it will become a static inline that gets
+ * discarded by the compiler on all gens not in range.
+ *
+ * You can do pseudo-runtime checks in your function such as
+ *
+ * if (ANV_GEN > 8 || ANV_IS_HASWELL) {
+ * // Do something
+ * }
+ *
+ * The contents of the if statement must be valid regardless of gen, but
+ * the if will get compiled away on everything except haswell.
+ *
+ * For places where you really do have a compile-time conflict, you can
+ * use preprocessor logic:
+ *
+ * #if (ANV_GEN > 8 || ANV_IS_HASWELL)
+ * // Do something
+ * #endif
+ *
+ * However, it is strongly recommended that the former be used whenever
+ * possible.
+ */
+
+/* Base macro defined on the command line. If we don't have this, we can't
+ * do anything.
+ */
+#ifdef ANV_GENx10
+
+/* Gen checking macros */
+#define ANV_GEN ((ANV_GENx10) / 10)
+#define ANV_IS_HASWELL ((ANV_GENx10) == 75)
+
+/* Prefixing macros */
+#if (ANV_GENx10 == 70)
+# define GENX(X) GEN7_##X
+# define genX(x) gen7_##x
+#elif (ANV_GENx10 == 75)
+# define GENX(X) GEN75_##X
+# define genX(x) gen75_##x
+#elif (ANV_GENx10 == 80)
+# define GENX(X) GEN8_##X
+# define genX(x) gen8_##x
+#elif (ANV_GENx10 == 90)
+# define GENX(X) GEN9_##X
+# define genX(x) gen9_##x
+#else
+# error "Need to add prefixing macros for your gen"
+#endif
+
+/* Macros for comparing gens */
+#if (ANV_GENx10 >= 70)
+#define __ANV_GEN_GE_GEN7(T, F) T
+#else
+#define __ANV_GEN_GE_GEN7(T, F) F
+#endif
+
+#if (ANV_GENx10 <= 70)
+#define __ANV_GEN_LE_GEN7(T, F) T
+#else
+#define __ANV_GEN_LE_GEN7(T, F) F
+#endif
+
+#if (ANV_GENx10 >= 75)
+#define __ANV_GEN_GE_GEN75(T, F) T
+#else
+#define __ANV_GEN_GE_GEN75(T, F) F
+#endif
+
+#if (ANV_GENx10 <= 75)
+#define __ANV_GEN_LE_GEN75(T, F) T
+#else
+#define __ANV_GEN_LE_GEN75(T, F) F
+#endif
+
+#if (ANV_GENx10 >= 80)
+#define __ANV_GEN_GE_GEN8(T, F) T
+#else
+#define __ANV_GEN_GE_GEN8(T, F) F
+#endif
+
+#if (ANV_GENx10 <= 80)
+#define __ANV_GEN_LE_GEN8(T, F) T
+#else
+#define __ANV_GEN_LE_GEN8(T, F) F
+#endif
+
+#if (ANV_GENx10 >= 90)
+#define __ANV_GEN_GE_GEN9(T, F) T
+#else
+#define __ANV_GEN_GE_GEN9(T, F) F
+#endif
+
+#if (ANV_GENx10 <= 90)
+#define __ANV_GEN_LE_GEN9(T, F) T
+#else
+#define __ANV_GEN_LE_GEN9(T, F) F
+#endif
+
+#define __ANV_GEN_IN_RANGE(start, end, T, F) \
+ __ANV_GEN_GE_##start(__ANV_GEN_LE_##end(T, F), F)
+
+/* Declares a function as static inlind if it's not in range */
+#define GENX_FUNC(start, end) __ANV_GEN_IN_RANGE(start, end, , static inline)
+
+#endif /* ANV_GENx10 */
--- /dev/null
+{
+ "file_format_version": "1.0.0",
+ "ICD": {
+ "library_path": "@abs_top_builddir@/lib/libvulkan.so.0.0.0",
+ "abi_versions": "0.210.1"
+ }
+}
+
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <assert.h>
+#include <stdbool.h>
+#include <string.h>
+#include <unistd.h>
+#include <fcntl.h>
+
+#include "anv_private.h"
+
+/**
+ * The \a format argument is required and overrides any format found in struct
+ * anv_image_create_info. Exactly one bit must be set in \a aspect.
+ */
+static isl_surf_usage_flags_t
+choose_isl_surf_usage(const struct anv_image_create_info *info,
+ VkImageAspectFlags aspect)
+{
+ const VkImageCreateInfo *vk_info = info->vk_info;
+ isl_surf_usage_flags_t isl_flags = 0;
+
+ /* FINISHME: Support aux surfaces */
+ isl_flags |= ISL_SURF_USAGE_DISABLE_AUX_BIT;
+
+ if (vk_info->usage & VK_IMAGE_USAGE_SAMPLED_BIT)
+ isl_flags |= ISL_SURF_USAGE_TEXTURE_BIT;
+
+ if (vk_info->usage & VK_IMAGE_USAGE_INPUT_ATTACHMENT_BIT)
+ isl_flags |= ISL_SURF_USAGE_TEXTURE_BIT;
+
+ if (vk_info->usage & VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT)
+ isl_flags |= ISL_SURF_USAGE_RENDER_TARGET_BIT;
+
+ if (vk_info->flags & VK_IMAGE_CREATE_CUBE_COMPATIBLE_BIT)
+ isl_flags |= ISL_SURF_USAGE_CUBE_BIT;
+
+ if (vk_info->usage & VK_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
+ switch (aspect) {
+ default:
+ unreachable("bad VkImageAspect");
+ case VK_IMAGE_ASPECT_DEPTH_BIT:
+ isl_flags |= ISL_SURF_USAGE_DEPTH_BIT;
+ break;
+ case VK_IMAGE_ASPECT_STENCIL_BIT:
+ isl_flags |= ISL_SURF_USAGE_STENCIL_BIT;
+ break;
+ }
+ }
+
+ if (vk_info->usage & VK_IMAGE_USAGE_TRANSFER_SRC_BIT) {
+ /* Meta implements transfers by sampling from the source image. */
+ isl_flags |= ISL_SURF_USAGE_TEXTURE_BIT;
+ }
+
+ if (vk_info->usage & VK_IMAGE_USAGE_TRANSFER_DST_BIT) {
+ /* Meta implements transfers by rendering into the destination image. */
+ isl_flags |= ISL_SURF_USAGE_RENDER_TARGET_BIT;
+ }
+
+ return isl_flags;
+}
+
+/**
+ * Exactly one bit must be set in \a aspect.
+ */
+static struct anv_surface *
+get_surface(struct anv_image *image, VkImageAspectFlags aspect)
+{
+ switch (aspect) {
+ default:
+ unreachable("bad VkImageAspect");
+ case VK_IMAGE_ASPECT_COLOR_BIT:
+ return &image->color_surface;
+ case VK_IMAGE_ASPECT_DEPTH_BIT:
+ return &image->depth_surface;
+ case VK_IMAGE_ASPECT_STENCIL_BIT:
+ return &image->stencil_surface;
+ }
+}
+
+/**
+ * Initialize the anv_image::*_surface selected by \a aspect. Then update the
+ * image's memory requirements (that is, the image's size and alignment).
+ *
+ * Exactly one bit must be set in \a aspect.
+ */
+static VkResult
+make_surface(const struct anv_device *dev,
+ struct anv_image *image,
+ const struct anv_image_create_info *anv_info,
+ VkImageAspectFlags aspect)
+{
+ const VkImageCreateInfo *vk_info = anv_info->vk_info;
+ bool ok UNUSED;
+
+ static const enum isl_surf_dim vk_to_isl_surf_dim[] = {
+ [VK_IMAGE_TYPE_1D] = ISL_SURF_DIM_1D,
+ [VK_IMAGE_TYPE_2D] = ISL_SURF_DIM_2D,
+ [VK_IMAGE_TYPE_3D] = ISL_SURF_DIM_3D,
+ };
+
+ isl_tiling_flags_t tiling_flags = anv_info->isl_tiling_flags;
+ if (vk_info->tiling == VK_IMAGE_TILING_LINEAR)
+ tiling_flags &= ISL_TILING_LINEAR_BIT;
+
+ struct anv_surface *anv_surf = get_surface(image, aspect);
+
+ ok = isl_surf_init(&dev->isl_dev, &anv_surf->isl,
+ .dim = vk_to_isl_surf_dim[vk_info->imageType],
+ .format = anv_get_isl_format(vk_info->format, aspect, vk_info->tiling),
+ .width = vk_info->extent.width,
+ .height = vk_info->extent.height,
+ .depth = vk_info->extent.depth,
+ .levels = vk_info->mipLevels,
+ .array_len = vk_info->arrayLayers,
+ .samples = vk_info->samples,
+ .min_alignment = 0,
+ .min_pitch = 0,
+ .usage = choose_isl_surf_usage(anv_info, aspect),
+ .tiling_flags = tiling_flags);
+
+ /* isl_surf_init() will fail only if provided invalid input. Invalid input
+ * is illegal in Vulkan.
+ */
+ assert(ok);
+
+ anv_surf->offset = align_u32(image->size, anv_surf->isl.alignment);
+ image->size = anv_surf->offset + anv_surf->isl.size;
+ image->alignment = MAX(image->alignment, anv_surf->isl.alignment);
+
+ return VK_SUCCESS;
+}
+
+static VkImageUsageFlags
+anv_image_get_full_usage(const VkImageCreateInfo *info)
+{
+ VkImageUsageFlags usage = info->usage;
+
+ if (usage & VK_IMAGE_USAGE_TRANSFER_SRC_BIT) {
+ /* Meta will transfer from the image by binding it as a texture. */
+ usage |= VK_IMAGE_USAGE_SAMPLED_BIT;
+ }
+
+ if (usage & VK_IMAGE_USAGE_TRANSFER_DST_BIT) {
+ /* Meta will transfer to the image by binding it as a color attachment,
+ * even if the image format is not a color format.
+ */
+ usage |= VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT;
+ }
+
+ return usage;
+}
+
+VkResult
+anv_image_create(VkDevice _device,
+ const struct anv_image_create_info *create_info,
+ const VkAllocationCallbacks* alloc,
+ VkImage *pImage)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ const VkImageCreateInfo *pCreateInfo = create_info->vk_info;
+ struct anv_image *image = NULL;
+ VkResult r;
+
+ assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO);
+
+ anv_assert(pCreateInfo->mipLevels > 0);
+ anv_assert(pCreateInfo->arrayLayers > 0);
+ anv_assert(pCreateInfo->samples == VK_SAMPLE_COUNT_1_BIT);
+ anv_assert(pCreateInfo->extent.width > 0);
+ anv_assert(pCreateInfo->extent.height > 0);
+ anv_assert(pCreateInfo->extent.depth > 0);
+
+ image = anv_alloc2(&device->alloc, alloc, sizeof(*image), 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (!image)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ memset(image, 0, sizeof(*image));
+ image->type = pCreateInfo->imageType;
+ image->extent = pCreateInfo->extent;
+ image->vk_format = pCreateInfo->format;
+ image->format = anv_format_for_vk_format(pCreateInfo->format);
+ image->levels = pCreateInfo->mipLevels;
+ image->array_size = pCreateInfo->arrayLayers;
+ image->usage = anv_image_get_full_usage(pCreateInfo);
+ image->tiling = pCreateInfo->tiling;
+
+ if (image->usage & VK_IMAGE_USAGE_SAMPLED_BIT) {
+ image->needs_nonrt_surface_state = true;
+ }
+
+ if (image->usage & VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT) {
+ image->needs_color_rt_surface_state = true;
+ }
+
+ if (image->usage & VK_IMAGE_USAGE_STORAGE_BIT) {
+ image->needs_storage_surface_state = true;
+ }
+
+ if (likely(anv_format_is_color(image->format))) {
+ r = make_surface(device, image, create_info,
+ VK_IMAGE_ASPECT_COLOR_BIT);
+ if (r != VK_SUCCESS)
+ goto fail;
+ } else {
+ if (image->format->depth_format) {
+ r = make_surface(device, image, create_info,
+ VK_IMAGE_ASPECT_DEPTH_BIT);
+ if (r != VK_SUCCESS)
+ goto fail;
+ }
+
+ if (image->format->has_stencil) {
+ r = make_surface(device, image, create_info,
+ VK_IMAGE_ASPECT_STENCIL_BIT);
+ if (r != VK_SUCCESS)
+ goto fail;
+ }
+ }
+
+ *pImage = anv_image_to_handle(image);
+
+ return VK_SUCCESS;
+
+fail:
+ if (image)
+ anv_free2(&device->alloc, alloc, image);
+
+ return r;
+}
+
+VkResult
+anv_CreateImage(VkDevice device,
+ const VkImageCreateInfo *pCreateInfo,
+ const VkAllocationCallbacks *pAllocator,
+ VkImage *pImage)
+{
+ return anv_image_create(device,
+ &(struct anv_image_create_info) {
+ .vk_info = pCreateInfo,
+ .isl_tiling_flags = ISL_TILING_ANY_MASK,
+ },
+ pAllocator,
+ pImage);
+}
+
+void
+anv_DestroyImage(VkDevice _device, VkImage _image,
+ const VkAllocationCallbacks *pAllocator)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+
+ anv_free2(&device->alloc, pAllocator, anv_image_from_handle(_image));
+}
+
+static void
+anv_surface_get_subresource_layout(struct anv_image *image,
+ struct anv_surface *surface,
+ const VkImageSubresource *subresource,
+ VkSubresourceLayout *layout)
+{
+ /* If we are on a non-zero mip level or array slice, we need to
+ * calculate a real offset.
+ */
+ anv_assert(subresource->mipLevel == 0);
+ anv_assert(subresource->arrayLayer == 0);
+
+ layout->offset = surface->offset;
+ layout->rowPitch = surface->isl.row_pitch;
+ layout->depthPitch = isl_surf_get_array_pitch(&surface->isl);
+ layout->arrayPitch = isl_surf_get_array_pitch(&surface->isl);
+ layout->size = surface->isl.size;
+}
+
+void anv_GetImageSubresourceLayout(
+ VkDevice device,
+ VkImage _image,
+ const VkImageSubresource* pSubresource,
+ VkSubresourceLayout* pLayout)
+{
+ ANV_FROM_HANDLE(anv_image, image, _image);
+
+ assert(__builtin_popcount(pSubresource->aspectMask) == 1);
+
+ switch (pSubresource->aspectMask) {
+ case VK_IMAGE_ASPECT_COLOR_BIT:
+ anv_surface_get_subresource_layout(image, &image->color_surface,
+ pSubresource, pLayout);
+ break;
+ case VK_IMAGE_ASPECT_DEPTH_BIT:
+ anv_surface_get_subresource_layout(image, &image->depth_surface,
+ pSubresource, pLayout);
+ break;
+ case VK_IMAGE_ASPECT_STENCIL_BIT:
+ anv_surface_get_subresource_layout(image, &image->stencil_surface,
+ pSubresource, pLayout);
+ break;
+ default:
+ assert(!"Invalid image aspect");
+ }
+}
+
+VkResult
+anv_validate_CreateImageView(VkDevice _device,
+ const VkImageViewCreateInfo *pCreateInfo,
+ const VkAllocationCallbacks *pAllocator,
+ VkImageView *pView)
+{
+ ANV_FROM_HANDLE(anv_image, image, pCreateInfo->image);
+ const VkImageSubresourceRange *subresource;
+ const struct anv_format *view_format_info;
+
+ /* Validate structure type before dereferencing it. */
+ assert(pCreateInfo);
+ assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO);
+ subresource = &pCreateInfo->subresourceRange;
+
+ /* Validate viewType is in range before using it. */
+ assert(pCreateInfo->viewType >= VK_IMAGE_VIEW_TYPE_BEGIN_RANGE);
+ assert(pCreateInfo->viewType <= VK_IMAGE_VIEW_TYPE_END_RANGE);
+
+ /* Validate format is in range before using it. */
+ assert(pCreateInfo->format >= VK_FORMAT_BEGIN_RANGE);
+ assert(pCreateInfo->format <= VK_FORMAT_END_RANGE);
+ view_format_info = anv_format_for_vk_format(pCreateInfo->format);
+
+ /* Validate channel swizzles. */
+ assert(pCreateInfo->components.r >= VK_COMPONENT_SWIZZLE_BEGIN_RANGE);
+ assert(pCreateInfo->components.r <= VK_COMPONENT_SWIZZLE_END_RANGE);
+ assert(pCreateInfo->components.g >= VK_COMPONENT_SWIZZLE_BEGIN_RANGE);
+ assert(pCreateInfo->components.g <= VK_COMPONENT_SWIZZLE_END_RANGE);
+ assert(pCreateInfo->components.b >= VK_COMPONENT_SWIZZLE_BEGIN_RANGE);
+ assert(pCreateInfo->components.b <= VK_COMPONENT_SWIZZLE_END_RANGE);
+ assert(pCreateInfo->components.a >= VK_COMPONENT_SWIZZLE_BEGIN_RANGE);
+ assert(pCreateInfo->components.a <= VK_COMPONENT_SWIZZLE_END_RANGE);
+
+ /* Validate subresource. */
+ assert(subresource->aspectMask != 0);
+ assert(subresource->levelCount > 0);
+ assert(subresource->layerCount > 0);
+ assert(subresource->baseMipLevel < image->levels);
+ assert(subresource->baseMipLevel + subresource->levelCount <= image->levels);
+ assert(subresource->baseArrayLayer < image->array_size);
+ assert(subresource->baseArrayLayer + subresource->layerCount <= image->array_size);
+ assert(pView);
+
+ const VkImageAspectFlags ds_flags = VK_IMAGE_ASPECT_DEPTH_BIT
+ | VK_IMAGE_ASPECT_STENCIL_BIT;
+
+ /* Validate format. */
+ if (subresource->aspectMask & VK_IMAGE_ASPECT_COLOR_BIT) {
+ assert(subresource->aspectMask == VK_IMAGE_ASPECT_COLOR_BIT);
+ assert(!image->format->depth_format);
+ assert(!image->format->has_stencil);
+ assert(!view_format_info->depth_format);
+ assert(!view_format_info->has_stencil);
+ assert(view_format_info->isl_layout->bs ==
+ image->format->isl_layout->bs);
+ } else if (subresource->aspectMask & ds_flags) {
+ assert((subresource->aspectMask & ~ds_flags) == 0);
+
+ if (subresource->aspectMask & VK_IMAGE_ASPECT_STENCIL_BIT) {
+ assert(image->format->depth_format);
+ assert(view_format_info->depth_format);
+ assert(view_format_info->isl_layout->bs ==
+ image->format->isl_layout->bs);
+ }
+
+ if (subresource->aspectMask & VK_IMAGE_ASPECT_STENCIL_BIT) {
+ /* FINISHME: Is it legal to have an R8 view of S8? */
+ assert(image->format->has_stencil);
+ assert(view_format_info->has_stencil);
+ }
+ } else {
+ assert(!"bad VkImageSubresourceRange::aspectFlags");
+ }
+
+ return anv_CreateImageView(_device, pCreateInfo, pAllocator, pView);
+}
+
+void
+anv_image_view_init(struct anv_image_view *iview,
+ struct anv_device *device,
+ const VkImageViewCreateInfo* pCreateInfo,
+ struct anv_cmd_buffer *cmd_buffer)
+{
+ ANV_FROM_HANDLE(anv_image, image, pCreateInfo->image);
+ const VkImageSubresourceRange *range = &pCreateInfo->subresourceRange;
+
+ assert(range->layerCount > 0);
+ assert(range->baseMipLevel < image->levels);
+ assert(image->usage & (VK_IMAGE_USAGE_SAMPLED_BIT |
+ VK_IMAGE_USAGE_STORAGE_BIT |
+ VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT |
+ VK_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT));
+
+ switch (image->type) {
+ default:
+ unreachable("bad VkImageType");
+ case VK_IMAGE_TYPE_1D:
+ case VK_IMAGE_TYPE_2D:
+ assert(range->baseArrayLayer + range->layerCount - 1 <= image->array_size);
+ break;
+ case VK_IMAGE_TYPE_3D:
+ assert(range->baseArrayLayer + range->layerCount - 1
+ <= anv_minify(image->extent.depth, range->baseMipLevel));
+ break;
+ }
+
+ struct anv_surface *surface =
+ anv_image_get_surface_for_aspect_mask(image, range->aspectMask);
+
+ iview->image = image;
+ iview->bo = image->bo;
+ iview->offset = image->offset + surface->offset;
+
+ iview->aspect_mask = pCreateInfo->subresourceRange.aspectMask;
+ iview->vk_format = pCreateInfo->format;
+ iview->format = anv_get_isl_format(pCreateInfo->format, iview->aspect_mask,
+ image->tiling);
+
+ iview->extent = (VkExtent3D) {
+ .width = anv_minify(image->extent.width, range->baseMipLevel),
+ .height = anv_minify(image->extent.height, range->baseMipLevel),
+ .depth = anv_minify(image->extent.depth, range->baseMipLevel),
+ };
+
+ switch (device->info.gen) {
+ case 7:
+ if (device->info.is_haswell)
+ gen75_image_view_init(iview, device, pCreateInfo, cmd_buffer);
+ else
+ gen7_image_view_init(iview, device, pCreateInfo, cmd_buffer);
+ break;
+ case 8:
+ gen8_image_view_init(iview, device, pCreateInfo, cmd_buffer);
+ break;
+ case 9:
+ gen9_image_view_init(iview, device, pCreateInfo, cmd_buffer);
+ break;
+ default:
+ unreachable("unsupported gen\n");
+ }
+}
+
+VkResult
+anv_CreateImageView(VkDevice _device,
+ const VkImageViewCreateInfo *pCreateInfo,
+ const VkAllocationCallbacks *pAllocator,
+ VkImageView *pView)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ struct anv_image_view *view;
+
+ view = anv_alloc2(&device->alloc, pAllocator, sizeof(*view), 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (view == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ anv_image_view_init(view, device, pCreateInfo, NULL);
+
+ *pView = anv_image_view_to_handle(view);
+
+ return VK_SUCCESS;
+}
+
+void
+anv_DestroyImageView(VkDevice _device, VkImageView _iview,
+ const VkAllocationCallbacks *pAllocator)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ ANV_FROM_HANDLE(anv_image_view, iview, _iview);
+
+ if (iview->image->needs_color_rt_surface_state) {
+ anv_state_pool_free(&device->surface_state_pool,
+ iview->color_rt_surface_state);
+ }
+
+ if (iview->image->needs_nonrt_surface_state) {
+ anv_state_pool_free(&device->surface_state_pool,
+ iview->nonrt_surface_state);
+ }
+
+ if (iview->image->needs_storage_surface_state) {
+ anv_state_pool_free(&device->surface_state_pool,
+ iview->storage_surface_state);
+ }
+
+ anv_free2(&device->alloc, pAllocator, iview);
+}
+
+VkResult
+anv_CreateBufferView(VkDevice _device,
+ const VkBufferViewCreateInfo *pCreateInfo,
+ const VkAllocationCallbacks *pAllocator,
+ VkBufferView *pView)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ ANV_FROM_HANDLE(anv_buffer, buffer, pCreateInfo->buffer);
+ struct anv_buffer_view *view;
+
+ view = anv_alloc2(&device->alloc, pAllocator, sizeof(*view), 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (!view)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ const struct anv_format *format =
+ anv_format_for_vk_format(pCreateInfo->format);
+
+ view->format = format->surface_format;
+ view->bo = buffer->bo;
+ view->offset = buffer->offset + pCreateInfo->offset;
+ view->range = pCreateInfo->range == VK_WHOLE_SIZE ?
+ buffer->size - view->offset : pCreateInfo->range;
+
+ if (buffer->usage & VK_BUFFER_USAGE_UNIFORM_TEXEL_BUFFER_BIT) {
+ view->surface_state =
+ anv_state_pool_alloc(&device->surface_state_pool, 64, 64);
+
+ anv_fill_buffer_surface_state(device, view->surface_state.map,
+ view->format,
+ view->offset, view->range,
+ format->isl_layout->bs);
+ } else {
+ view->surface_state = (struct anv_state){ 0 };
+ }
+
+ if (buffer->usage & VK_BUFFER_USAGE_STORAGE_TEXEL_BUFFER_BIT) {
+ view->storage_surface_state =
+ anv_state_pool_alloc(&device->surface_state_pool, 64, 64);
+
+ enum isl_format storage_format =
+ isl_lower_storage_image_format(&device->isl_dev, view->format);
+
+ anv_fill_buffer_surface_state(device, view->storage_surface_state.map,
+ storage_format,
+ view->offset, view->range,
+ format->isl_layout->bs);
+ } else {
+ view->storage_surface_state = (struct anv_state){ 0 };
+ }
+
+ *pView = anv_buffer_view_to_handle(view);
+
+ return VK_SUCCESS;
+}
+
+void
+anv_DestroyBufferView(VkDevice _device, VkBufferView bufferView,
+ const VkAllocationCallbacks *pAllocator)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ ANV_FROM_HANDLE(anv_buffer_view, view, bufferView);
+
+ if (view->surface_state.alloc_size > 0)
+ anv_state_pool_free(&device->surface_state_pool,
+ view->surface_state);
+
+ if (view->storage_surface_state.alloc_size > 0)
+ anv_state_pool_free(&device->surface_state_pool,
+ view->storage_surface_state);
+
+ anv_free2(&device->alloc, pAllocator, view);
+}
+
+struct anv_surface *
+anv_image_get_surface_for_aspect_mask(struct anv_image *image, VkImageAspectFlags aspect_mask)
+{
+ switch (aspect_mask) {
+ case VK_IMAGE_ASPECT_COLOR_BIT:
+ /* Dragons will eat you.
+ *
+ * Meta attaches all destination surfaces as color render targets. Guess
+ * what surface the Meta Dragons really want.
+ */
+ if (image->format->depth_format && image->format->has_stencil) {
+ anv_finishme("combined depth stencil formats");
+ return &image->depth_surface;
+ } else if (image->format->depth_format) {
+ return &image->depth_surface;
+ } else if (image->format->has_stencil) {
+ return &image->stencil_surface;
+ } else {
+ return &image->color_surface;
+ }
+ break;
+ case VK_IMAGE_ASPECT_DEPTH_BIT:
+ assert(image->format->depth_format);
+ return &image->depth_surface;
+ case VK_IMAGE_ASPECT_STENCIL_BIT:
+ assert(image->format->has_stencil);
+ return &image->stencil_surface;
+ case VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT:
+ if (image->format->depth_format && image->format->has_stencil) {
+ /* FINISHME: The Vulkan spec (git a511ba2) requires support for combined
+ * depth stencil formats. Specifically, it states:
+ *
+ * At least one of ename:VK_FORMAT_D24_UNORM_S8_UINT or
+ * ename:VK_FORMAT_D32_SFLOAT_S8_UINT must be supported.
+ */
+ anv_finishme("combined depthstencil aspect");
+ return &image->depth_surface;
+ } else if (image->format->depth_format) {
+ return &image->depth_surface;
+ } else if (image->format->has_stencil) {
+ return &image->stencil_surface;
+ }
+ /* fallthrough */
+ default:
+ unreachable("image does not have aspect");
+ return NULL;
+ }
+}
+
+void
+anv_image_view_fill_image_param(struct anv_device *device,
+ struct anv_image_view *view,
+ struct brw_image_param *param)
+{
+ memset(param, 0, sizeof *param);
+ anv_finishme("Actually fill out brw_image_param");
+}
+
+void
+anv_buffer_view_fill_image_param(struct anv_device *device,
+ struct anv_buffer_view *view,
+ struct brw_image_param *param)
+{
+ /* Set the swizzling shifts to all-ones to effectively disable swizzling --
+ * See emit_address_calculation() in brw_fs_surface_builder.cpp for a more
+ * detailed explanation of these parameters.
+ */
+ param->swizzling[0] = 0xff;
+ param->swizzling[1] = 0xff;
+
+ param->stride[0] = isl_format_layouts[view->format].bs;
+ param->size[0] = view->range / param->stride[0];
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <assert.h>
+#include <stdbool.h>
+#include <string.h>
+#include <unistd.h>
+#include <fcntl.h>
+
+#include "anv_private.h"
+
+VkResult anv_CreateDmaBufImageINTEL(
+ VkDevice _device,
+ const VkDmaBufImageCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkDeviceMemory* pMem,
+ VkImage* pImage)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ struct anv_device_memory *mem;
+ struct anv_image *image;
+ VkResult result;
+ VkImage image_h;
+
+ assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_DMA_BUF_IMAGE_CREATE_INFO_INTEL);
+
+ mem = anv_alloc2(&device->alloc, pAllocator, sizeof(*mem), 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (mem == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ mem->bo.gem_handle = anv_gem_fd_to_handle(device, pCreateInfo->fd);
+ if (!mem->bo.gem_handle) {
+ result = vk_error(VK_ERROR_OUT_OF_DEVICE_MEMORY);
+ goto fail;
+ }
+
+ mem->bo.map = NULL;
+ mem->bo.index = 0;
+ mem->bo.offset = 0;
+ mem->bo.size = pCreateInfo->strideInBytes * pCreateInfo->extent.height;
+
+ anv_image_create(_device,
+ &(struct anv_image_create_info) {
+ .isl_tiling_flags = ISL_TILING_X_BIT,
+ .stride = pCreateInfo->strideInBytes,
+ .vk_info =
+ &(VkImageCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO,
+ .imageType = VK_IMAGE_TYPE_2D,
+ .format = pCreateInfo->format,
+ .extent = pCreateInfo->extent,
+ .mipLevels = 1,
+ .arrayLayers = 1,
+ .samples = 1,
+ /* FIXME: Need a way to use X tiling to allow scanout */
+ .tiling = VK_IMAGE_TILING_OPTIMAL,
+ .usage = VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT,
+ .flags = 0,
+ }},
+ pAllocator, &image_h);
+
+ image = anv_image_from_handle(image_h);
+ image->bo = &mem->bo;
+ image->offset = 0;
+
+ assert(image->extent.width > 0);
+ assert(image->extent.height > 0);
+ assert(image->extent.depth == 1);
+
+ *pMem = anv_device_memory_to_handle(mem);
+ *pImage = anv_image_to_handle(image);
+
+ return VK_SUCCESS;
+
+ fail:
+ anv_free2(&device->alloc, pAllocator, mem);
+
+ return result;
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <assert.h>
+#include <stdbool.h>
+#include <string.h>
+#include <unistd.h>
+#include <fcntl.h>
+
+#include "anv_meta.h"
+#include "anv_meta_clear.h"
+#include "anv_private.h"
+#include "glsl/nir/nir_builder.h"
+
+struct anv_render_pass anv_meta_dummy_renderpass = {0};
+
+static nir_shader *
+build_nir_vertex_shader(bool attr_flat)
+{
+ nir_builder b;
+
+ const struct glsl_type *vertex_type = glsl_vec4_type();
+
+ nir_builder_init_simple_shader(&b, NULL, MESA_SHADER_VERTEX, NULL);
+ b.shader->info.name = ralloc_strdup(b.shader, "meta_blit_vs");
+
+ nir_variable *pos_in = nir_variable_create(b.shader, nir_var_shader_in,
+ vertex_type, "a_pos");
+ pos_in->data.location = VERT_ATTRIB_GENERIC0;
+ nir_variable *pos_out = nir_variable_create(b.shader, nir_var_shader_out,
+ vertex_type, "gl_Position");
+ pos_out->data.location = VARYING_SLOT_POS;
+ nir_copy_var(&b, pos_out, pos_in);
+
+ /* Add one more pass-through attribute. For clear shaders, this is used
+ * to store the color and for blit shaders it's the texture coordinate.
+ */
+ const struct glsl_type *attr_type = glsl_vec4_type();
+ nir_variable *attr_in = nir_variable_create(b.shader, nir_var_shader_in,
+ attr_type, "a_attr");
+ attr_in->data.location = VERT_ATTRIB_GENERIC1;
+ nir_variable *attr_out = nir_variable_create(b.shader, nir_var_shader_out,
+ attr_type, "v_attr");
+ attr_out->data.location = VARYING_SLOT_VAR0;
+ attr_out->data.interpolation = attr_flat ? INTERP_QUALIFIER_FLAT :
+ INTERP_QUALIFIER_SMOOTH;
+ nir_copy_var(&b, attr_out, attr_in);
+
+ return b.shader;
+}
+
+static nir_shader *
+build_nir_copy_fragment_shader(enum glsl_sampler_dim tex_dim)
+{
+ nir_builder b;
+
+ nir_builder_init_simple_shader(&b, NULL, MESA_SHADER_FRAGMENT, NULL);
+ b.shader->info.name = ralloc_strdup(b.shader, "meta_blit_fs");
+
+ const struct glsl_type *color_type = glsl_vec4_type();
+
+ nir_variable *tex_pos_in = nir_variable_create(b.shader, nir_var_shader_in,
+ glsl_vec4_type(), "v_attr");
+ tex_pos_in->data.location = VARYING_SLOT_VAR0;
+
+ const struct glsl_type *sampler_type =
+ glsl_sampler_type(tex_dim, false, false, glsl_get_base_type(color_type));
+ nir_variable *sampler = nir_variable_create(b.shader, nir_var_uniform,
+ sampler_type, "s_tex");
+ sampler->data.descriptor_set = 0;
+ sampler->data.binding = 0;
+
+ nir_tex_instr *tex = nir_tex_instr_create(b.shader, 1);
+ tex->sampler_dim = tex_dim;
+ tex->op = nir_texop_tex;
+ tex->src[0].src_type = nir_tex_src_coord;
+ tex->src[0].src = nir_src_for_ssa(nir_load_var(&b, tex_pos_in));
+ tex->dest_type = nir_type_float; /* TODO */
+
+ if (tex_dim != GLSL_SAMPLER_DIM_3D)
+ tex->is_array = true;
+
+ tex->coord_components = 3;
+
+ tex->sampler = nir_deref_var_create(tex, sampler);
+
+ nir_ssa_dest_init(&tex->instr, &tex->dest, 4, "tex");
+ nir_builder_instr_insert(&b, &tex->instr);
+
+ nir_variable *color_out = nir_variable_create(b.shader, nir_var_shader_out,
+ color_type, "f_color");
+ color_out->data.location = FRAG_RESULT_DATA0;
+ nir_store_var(&b, color_out, &tex->dest.ssa, 4);
+
+ return b.shader;
+}
+
+void
+anv_meta_save(struct anv_meta_saved_state *state,
+ const struct anv_cmd_buffer *cmd_buffer,
+ uint32_t dynamic_mask)
+{
+ state->old_pipeline = cmd_buffer->state.pipeline;
+ state->old_descriptor_set0 = cmd_buffer->state.descriptors[0];
+ memcpy(state->old_vertex_bindings, cmd_buffer->state.vertex_bindings,
+ sizeof(state->old_vertex_bindings));
+
+ state->dynamic_mask = dynamic_mask;
+ anv_dynamic_state_copy(&state->dynamic, &cmd_buffer->state.dynamic,
+ dynamic_mask);
+}
+
+void
+anv_meta_restore(const struct anv_meta_saved_state *state,
+ struct anv_cmd_buffer *cmd_buffer)
+{
+ cmd_buffer->state.pipeline = state->old_pipeline;
+ cmd_buffer->state.descriptors[0] = state->old_descriptor_set0;
+ memcpy(cmd_buffer->state.vertex_bindings, state->old_vertex_bindings,
+ sizeof(state->old_vertex_bindings));
+
+ cmd_buffer->state.vb_dirty |= (1 << ANV_META_VERTEX_BINDING_COUNT) - 1;
+ cmd_buffer->state.dirty |= ANV_CMD_DIRTY_PIPELINE;
+ cmd_buffer->state.descriptors_dirty |= VK_SHADER_STAGE_VERTEX_BIT;
+
+ anv_dynamic_state_copy(&cmd_buffer->state.dynamic, &state->dynamic,
+ state->dynamic_mask);
+ cmd_buffer->state.dirty |= state->dynamic_mask;
+
+ /* Since we've used the pipeline with the VS disabled, set
+ * need_query_wa. See CmdBeginQuery.
+ */
+ cmd_buffer->state.need_query_wa = true;
+}
+
+VkImageViewType
+anv_meta_get_view_type(const struct anv_image *image)
+{
+ switch (image->type) {
+ case VK_IMAGE_TYPE_1D: return VK_IMAGE_VIEW_TYPE_1D;
+ case VK_IMAGE_TYPE_2D: return VK_IMAGE_VIEW_TYPE_2D;
+ case VK_IMAGE_TYPE_3D: return VK_IMAGE_VIEW_TYPE_3D;
+ default:
+ unreachable("bad VkImageViewType");
+ }
+}
+
+static uint32_t
+meta_blit_get_dest_view_base_array_slice(const struct anv_image *dest_image,
+ const VkImageSubresourceLayers *dest_subresource,
+ const VkOffset3D *dest_offset)
+{
+ switch (dest_image->type) {
+ case VK_IMAGE_TYPE_1D:
+ case VK_IMAGE_TYPE_2D:
+ return dest_subresource->baseArrayLayer;
+ case VK_IMAGE_TYPE_3D:
+ /* HACK: Vulkan does not allow attaching a 3D image to a framebuffer,
+ * but meta does it anyway. When doing so, we translate the
+ * destination's z offset into an array offset.
+ */
+ return dest_offset->z;
+ default:
+ assert(!"bad VkImageType");
+ return 0;
+ }
+}
+
+static VkResult
+anv_device_init_meta_blit_state(struct anv_device *device)
+{
+ VkResult result;
+
+ result = anv_CreateRenderPass(anv_device_to_handle(device),
+ &(VkRenderPassCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO,
+ .attachmentCount = 1,
+ .pAttachments = &(VkAttachmentDescription) {
+ .format = VK_FORMAT_UNDEFINED, /* Our shaders don't care */
+ .loadOp = VK_ATTACHMENT_LOAD_OP_LOAD,
+ .storeOp = VK_ATTACHMENT_STORE_OP_STORE,
+ .initialLayout = VK_IMAGE_LAYOUT_GENERAL,
+ .finalLayout = VK_IMAGE_LAYOUT_GENERAL,
+ },
+ .subpassCount = 1,
+ .pSubpasses = &(VkSubpassDescription) {
+ .pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS,
+ .inputAttachmentCount = 0,
+ .colorAttachmentCount = 1,
+ .pColorAttachments = &(VkAttachmentReference) {
+ .attachment = 0,
+ .layout = VK_IMAGE_LAYOUT_GENERAL,
+ },
+ .pResolveAttachments = NULL,
+ .pDepthStencilAttachment = &(VkAttachmentReference) {
+ .attachment = VK_ATTACHMENT_UNUSED,
+ .layout = VK_IMAGE_LAYOUT_GENERAL,
+ },
+ .preserveAttachmentCount = 1,
+ .pPreserveAttachments = (uint32_t[]) { 0 },
+ },
+ .dependencyCount = 0,
+ }, &device->meta_state.alloc, &device->meta_state.blit.render_pass);
+ if (result != VK_SUCCESS)
+ goto fail;
+
+ /* We don't use a vertex shader for clearing, but instead build and pass
+ * the VUEs directly to the rasterization backend. However, we do need
+ * to provide GLSL source for the vertex shader so that the compiler
+ * does not dead-code our inputs.
+ */
+ struct anv_shader_module vs = {
+ .nir = build_nir_vertex_shader(false),
+ };
+
+ struct anv_shader_module fs_1d = {
+ .nir = build_nir_copy_fragment_shader(GLSL_SAMPLER_DIM_1D),
+ };
+
+ struct anv_shader_module fs_2d = {
+ .nir = build_nir_copy_fragment_shader(GLSL_SAMPLER_DIM_2D),
+ };
+
+ struct anv_shader_module fs_3d = {
+ .nir = build_nir_copy_fragment_shader(GLSL_SAMPLER_DIM_3D),
+ };
+
+ VkPipelineVertexInputStateCreateInfo vi_create_info = {
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO,
+ .vertexBindingDescriptionCount = 2,
+ .pVertexBindingDescriptions = (VkVertexInputBindingDescription[]) {
+ {
+ .binding = 0,
+ .stride = 0,
+ .inputRate = VK_VERTEX_INPUT_RATE_VERTEX
+ },
+ {
+ .binding = 1,
+ .stride = 5 * sizeof(float),
+ .inputRate = VK_VERTEX_INPUT_RATE_VERTEX
+ },
+ },
+ .vertexAttributeDescriptionCount = 3,
+ .pVertexAttributeDescriptions = (VkVertexInputAttributeDescription[]) {
+ {
+ /* VUE Header */
+ .location = 0,
+ .binding = 0,
+ .format = VK_FORMAT_R32G32B32A32_UINT,
+ .offset = 0
+ },
+ {
+ /* Position */
+ .location = 1,
+ .binding = 1,
+ .format = VK_FORMAT_R32G32_SFLOAT,
+ .offset = 0
+ },
+ {
+ /* Texture Coordinate */
+ .location = 2,
+ .binding = 1,
+ .format = VK_FORMAT_R32G32B32_SFLOAT,
+ .offset = 8
+ }
+ }
+ };
+
+ VkDescriptorSetLayoutCreateInfo ds_layout_info = {
+ .sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO,
+ .bindingCount = 1,
+ .pBindings = (VkDescriptorSetLayoutBinding[]) {
+ {
+ .binding = 0,
+ .descriptorType = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER,
+ .descriptorCount = 1,
+ .stageFlags = VK_SHADER_STAGE_FRAGMENT_BIT,
+ .pImmutableSamplers = NULL
+ },
+ }
+ };
+ result = anv_CreateDescriptorSetLayout(anv_device_to_handle(device),
+ &ds_layout_info,
+ &device->meta_state.alloc,
+ &device->meta_state.blit.ds_layout);
+ if (result != VK_SUCCESS)
+ goto fail_render_pass;
+
+ result = anv_CreatePipelineLayout(anv_device_to_handle(device),
+ &(VkPipelineLayoutCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO,
+ .setLayoutCount = 1,
+ .pSetLayouts = &device->meta_state.blit.ds_layout,
+ },
+ &device->meta_state.alloc, &device->meta_state.blit.pipeline_layout);
+ if (result != VK_SUCCESS)
+ goto fail_descriptor_set_layout;
+
+ VkPipelineShaderStageCreateInfo pipeline_shader_stages[] = {
+ {
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO,
+ .stage = VK_SHADER_STAGE_VERTEX_BIT,
+ .module = anv_shader_module_to_handle(&vs),
+ .pName = "main",
+ .pSpecializationInfo = NULL
+ }, {
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO,
+ .stage = VK_SHADER_STAGE_FRAGMENT_BIT,
+ .module = VK_NULL_HANDLE, /* TEMPLATE VALUE! FILL ME IN! */
+ .pName = "main",
+ .pSpecializationInfo = NULL
+ },
+ };
+
+ const VkGraphicsPipelineCreateInfo vk_pipeline_info = {
+ .sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO,
+ .stageCount = ARRAY_SIZE(pipeline_shader_stages),
+ .pStages = pipeline_shader_stages,
+ .pVertexInputState = &vi_create_info,
+ .pInputAssemblyState = &(VkPipelineInputAssemblyStateCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO,
+ .topology = VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP,
+ .primitiveRestartEnable = false,
+ },
+ .pViewportState = &(VkPipelineViewportStateCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO,
+ .viewportCount = 1,
+ .scissorCount = 1,
+ },
+ .pRasterizationState = &(VkPipelineRasterizationStateCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_STATE_CREATE_INFO,
+ .rasterizerDiscardEnable = false,
+ .polygonMode = VK_POLYGON_MODE_FILL,
+ .cullMode = VK_CULL_MODE_NONE,
+ .frontFace = VK_FRONT_FACE_COUNTER_CLOCKWISE
+ },
+ .pMultisampleState = &(VkPipelineMultisampleStateCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO,
+ .rasterizationSamples = 1,
+ .sampleShadingEnable = false,
+ .pSampleMask = (VkSampleMask[]) { UINT32_MAX },
+ },
+ .pColorBlendState = &(VkPipelineColorBlendStateCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO,
+ .attachmentCount = 1,
+ .pAttachments = (VkPipelineColorBlendAttachmentState []) {
+ { .colorWriteMask =
+ VK_COLOR_COMPONENT_A_BIT |
+ VK_COLOR_COMPONENT_R_BIT |
+ VK_COLOR_COMPONENT_G_BIT |
+ VK_COLOR_COMPONENT_B_BIT },
+ }
+ },
+ .pDynamicState = &(VkPipelineDynamicStateCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_DYNAMIC_STATE_CREATE_INFO,
+ .dynamicStateCount = 9,
+ .pDynamicStates = (VkDynamicState[]) {
+ VK_DYNAMIC_STATE_VIEWPORT,
+ VK_DYNAMIC_STATE_SCISSOR,
+ VK_DYNAMIC_STATE_LINE_WIDTH,
+ VK_DYNAMIC_STATE_DEPTH_BIAS,
+ VK_DYNAMIC_STATE_BLEND_CONSTANTS,
+ VK_DYNAMIC_STATE_DEPTH_BOUNDS,
+ VK_DYNAMIC_STATE_STENCIL_COMPARE_MASK,
+ VK_DYNAMIC_STATE_STENCIL_WRITE_MASK,
+ VK_DYNAMIC_STATE_STENCIL_REFERENCE,
+ },
+ },
+ .flags = 0,
+ .layout = device->meta_state.blit.pipeline_layout,
+ .renderPass = device->meta_state.blit.render_pass,
+ .subpass = 0,
+ };
+
+ const struct anv_graphics_pipeline_create_info anv_pipeline_info = {
+ .color_attachment_count = -1,
+ .use_repclear = false,
+ .disable_viewport = true,
+ .disable_scissor = true,
+ .disable_vs = true,
+ .use_rectlist = true
+ };
+
+ pipeline_shader_stages[1].module = anv_shader_module_to_handle(&fs_1d);
+ result = anv_graphics_pipeline_create(anv_device_to_handle(device),
+ VK_NULL_HANDLE,
+ &vk_pipeline_info, &anv_pipeline_info,
+ &device->meta_state.alloc, &device->meta_state.blit.pipeline_1d_src);
+ if (result != VK_SUCCESS)
+ goto fail_pipeline_layout;
+
+ pipeline_shader_stages[1].module = anv_shader_module_to_handle(&fs_2d);
+ result = anv_graphics_pipeline_create(anv_device_to_handle(device),
+ VK_NULL_HANDLE,
+ &vk_pipeline_info, &anv_pipeline_info,
+ &device->meta_state.alloc, &device->meta_state.blit.pipeline_2d_src);
+ if (result != VK_SUCCESS)
+ goto fail_pipeline_1d;
+
+ pipeline_shader_stages[1].module = anv_shader_module_to_handle(&fs_3d);
+ result = anv_graphics_pipeline_create(anv_device_to_handle(device),
+ VK_NULL_HANDLE,
+ &vk_pipeline_info, &anv_pipeline_info,
+ &device->meta_state.alloc, &device->meta_state.blit.pipeline_3d_src);
+ if (result != VK_SUCCESS)
+ goto fail_pipeline_2d;
+
+ ralloc_free(vs.nir);
+ ralloc_free(fs_1d.nir);
+ ralloc_free(fs_2d.nir);
+ ralloc_free(fs_3d.nir);
+
+ return VK_SUCCESS;
+
+ fail_pipeline_2d:
+ anv_DestroyPipeline(anv_device_to_handle(device),
+ device->meta_state.blit.pipeline_2d_src,
+ &device->meta_state.alloc);
+
+ fail_pipeline_1d:
+ anv_DestroyPipeline(anv_device_to_handle(device),
+ device->meta_state.blit.pipeline_1d_src,
+ &device->meta_state.alloc);
+
+ fail_pipeline_layout:
+ anv_DestroyPipelineLayout(anv_device_to_handle(device),
+ device->meta_state.blit.pipeline_layout,
+ &device->meta_state.alloc);
+ fail_descriptor_set_layout:
+ anv_DestroyDescriptorSetLayout(anv_device_to_handle(device),
+ device->meta_state.blit.ds_layout,
+ &device->meta_state.alloc);
+ fail_render_pass:
+ anv_DestroyRenderPass(anv_device_to_handle(device),
+ device->meta_state.blit.render_pass,
+ &device->meta_state.alloc);
+
+ ralloc_free(vs.nir);
+ ralloc_free(fs_1d.nir);
+ ralloc_free(fs_2d.nir);
+ ralloc_free(fs_3d.nir);
+ fail:
+ return result;
+}
+
+static void
+meta_prepare_blit(struct anv_cmd_buffer *cmd_buffer,
+ struct anv_meta_saved_state *saved_state)
+{
+ anv_meta_save(saved_state, cmd_buffer,
+ (1 << VK_DYNAMIC_STATE_VIEWPORT));
+}
+
+struct blit_region {
+ VkOffset3D src_offset;
+ VkExtent3D src_extent;
+ VkOffset3D dest_offset;
+ VkExtent3D dest_extent;
+};
+
+static void
+meta_emit_blit(struct anv_cmd_buffer *cmd_buffer,
+ struct anv_image *src_image,
+ struct anv_image_view *src_iview,
+ VkOffset3D src_offset,
+ VkExtent3D src_extent,
+ struct anv_image *dest_image,
+ struct anv_image_view *dest_iview,
+ VkOffset3D dest_offset,
+ VkExtent3D dest_extent,
+ VkFilter blit_filter)
+{
+ struct anv_device *device = cmd_buffer->device;
+ VkDescriptorPool dummy_desc_pool = (VkDescriptorPool)1;
+
+ struct blit_vb_data {
+ float pos[2];
+ float tex_coord[3];
+ } *vb_data;
+
+ unsigned vb_size = sizeof(struct anv_vue_header) + 3 * sizeof(*vb_data);
+
+ struct anv_state vb_state =
+ anv_cmd_buffer_alloc_dynamic_state(cmd_buffer, vb_size, 16);
+ memset(vb_state.map, 0, sizeof(struct anv_vue_header));
+ vb_data = vb_state.map + sizeof(struct anv_vue_header);
+
+ vb_data[0] = (struct blit_vb_data) {
+ .pos = {
+ dest_offset.x + dest_extent.width,
+ dest_offset.y + dest_extent.height,
+ },
+ .tex_coord = {
+ (float)(src_offset.x + src_extent.width) / (float)src_iview->extent.width,
+ (float)(src_offset.y + src_extent.height) / (float)src_iview->extent.height,
+ (float)src_offset.z / (float)src_iview->extent.depth,
+ },
+ };
+
+ vb_data[1] = (struct blit_vb_data) {
+ .pos = {
+ dest_offset.x,
+ dest_offset.y + dest_extent.height,
+ },
+ .tex_coord = {
+ (float)src_offset.x / (float)src_iview->extent.width,
+ (float)(src_offset.y + src_extent.height) / (float)src_iview->extent.height,
+ (float)src_offset.z / (float)src_iview->extent.depth,
+ },
+ };
+
+ vb_data[2] = (struct blit_vb_data) {
+ .pos = {
+ dest_offset.x,
+ dest_offset.y,
+ },
+ .tex_coord = {
+ (float)src_offset.x / (float)src_iview->extent.width,
+ (float)src_offset.y / (float)src_iview->extent.height,
+ (float)src_offset.z / (float)src_iview->extent.depth,
+ },
+ };
+
+ anv_state_clflush(vb_state);
+
+ struct anv_buffer vertex_buffer = {
+ .device = device,
+ .size = vb_size,
+ .bo = &device->dynamic_state_block_pool.bo,
+ .offset = vb_state.offset,
+ };
+
+ anv_CmdBindVertexBuffers(anv_cmd_buffer_to_handle(cmd_buffer), 0, 2,
+ (VkBuffer[]) {
+ anv_buffer_to_handle(&vertex_buffer),
+ anv_buffer_to_handle(&vertex_buffer)
+ },
+ (VkDeviceSize[]) {
+ 0,
+ sizeof(struct anv_vue_header),
+ });
+
+ VkSampler sampler;
+ ANV_CALL(CreateSampler)(anv_device_to_handle(device),
+ &(VkSamplerCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO,
+ .magFilter = blit_filter,
+ .minFilter = blit_filter,
+ }, &cmd_buffer->pool->alloc, &sampler);
+
+ VkDescriptorSet set;
+ anv_AllocateDescriptorSets(anv_device_to_handle(device),
+ &(VkDescriptorSetAllocateInfo) {
+ .sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_ALLOCATE_INFO,
+ .descriptorPool = dummy_desc_pool,
+ .descriptorSetCount = 1,
+ .pSetLayouts = &device->meta_state.blit.ds_layout
+ }, &set);
+ anv_UpdateDescriptorSets(anv_device_to_handle(device),
+ 1, /* writeCount */
+ (VkWriteDescriptorSet[]) {
+ {
+ .sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET,
+ .dstSet = set,
+ .dstBinding = 0,
+ .dstArrayElement = 0,
+ .descriptorCount = 1,
+ .descriptorType = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER,
+ .pImageInfo = (VkDescriptorImageInfo[]) {
+ {
+ .sampler = sampler,
+ .imageView = anv_image_view_to_handle(src_iview),
+ .imageLayout = VK_IMAGE_LAYOUT_GENERAL,
+ },
+ }
+ }
+ }, 0, NULL);
+
+ VkFramebuffer fb;
+ anv_CreateFramebuffer(anv_device_to_handle(device),
+ &(VkFramebufferCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO,
+ .attachmentCount = 1,
+ .pAttachments = (VkImageView[]) {
+ anv_image_view_to_handle(dest_iview),
+ },
+ .width = dest_iview->extent.width,
+ .height = dest_iview->extent.height,
+ .layers = 1
+ }, &cmd_buffer->pool->alloc, &fb);
+
+ ANV_CALL(CmdBeginRenderPass)(anv_cmd_buffer_to_handle(cmd_buffer),
+ &(VkRenderPassBeginInfo) {
+ .sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO,
+ .renderPass = device->meta_state.blit.render_pass,
+ .framebuffer = fb,
+ .renderArea = {
+ .offset = { dest_offset.x, dest_offset.y },
+ .extent = { dest_extent.width, dest_extent.height },
+ },
+ .clearValueCount = 0,
+ .pClearValues = NULL,
+ }, VK_SUBPASS_CONTENTS_INLINE);
+
+ VkPipeline pipeline;
+
+ switch (src_image->type) {
+ case VK_IMAGE_TYPE_1D:
+ pipeline = device->meta_state.blit.pipeline_1d_src;
+ break;
+ case VK_IMAGE_TYPE_2D:
+ pipeline = device->meta_state.blit.pipeline_2d_src;
+ break;
+ case VK_IMAGE_TYPE_3D:
+ pipeline = device->meta_state.blit.pipeline_3d_src;
+ break;
+ default:
+ unreachable(!"bad VkImageType");
+ }
+
+ if (cmd_buffer->state.pipeline != anv_pipeline_from_handle(pipeline)) {
+ anv_CmdBindPipeline(anv_cmd_buffer_to_handle(cmd_buffer),
+ VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
+ }
+
+ anv_CmdSetViewport(anv_cmd_buffer_to_handle(cmd_buffer), 0, 1,
+ &(VkViewport) {
+ .x = 0.0f,
+ .y = 0.0f,
+ .width = dest_iview->extent.width,
+ .height = dest_iview->extent.height,
+ .minDepth = 0.0f,
+ .maxDepth = 1.0f,
+ });
+
+ anv_CmdBindDescriptorSets(anv_cmd_buffer_to_handle(cmd_buffer),
+ VK_PIPELINE_BIND_POINT_GRAPHICS,
+ device->meta_state.blit.pipeline_layout, 0, 1,
+ &set, 0, NULL);
+
+ ANV_CALL(CmdDraw)(anv_cmd_buffer_to_handle(cmd_buffer), 3, 1, 0, 0);
+
+ ANV_CALL(CmdEndRenderPass)(anv_cmd_buffer_to_handle(cmd_buffer));
+
+ /* At the point where we emit the draw call, all data from the
+ * descriptor sets, etc. has been used. We are free to delete it.
+ */
+ anv_descriptor_set_destroy(device, anv_descriptor_set_from_handle(set));
+ anv_DestroySampler(anv_device_to_handle(device), sampler,
+ &cmd_buffer->pool->alloc);
+ anv_DestroyFramebuffer(anv_device_to_handle(device), fb,
+ &cmd_buffer->pool->alloc);
+}
+
+static void
+meta_finish_blit(struct anv_cmd_buffer *cmd_buffer,
+ const struct anv_meta_saved_state *saved_state)
+{
+ anv_meta_restore(saved_state, cmd_buffer);
+}
+
+static VkFormat
+vk_format_for_size(int bs)
+{
+ /* Note: We intentionally use the 4-channel formats whenever we can.
+ * This is so that, when we do a RGB <-> RGBX copy, the two formats will
+ * line up even though one of them is 3/4 the size of the other.
+ */
+ switch (bs) {
+ case 1: return VK_FORMAT_R8_UINT;
+ case 2: return VK_FORMAT_R8G8_UINT;
+ case 3: return VK_FORMAT_R8G8B8_UINT;
+ case 4: return VK_FORMAT_R8G8B8A8_UINT;
+ case 6: return VK_FORMAT_R16G16B16_UINT;
+ case 8: return VK_FORMAT_R16G16B16A16_UINT;
+ case 12: return VK_FORMAT_R32G32B32_UINT;
+ case 16: return VK_FORMAT_R32G32B32A32_UINT;
+ default:
+ unreachable("Invalid format block size");
+ }
+}
+
+static void
+do_buffer_copy(struct anv_cmd_buffer *cmd_buffer,
+ struct anv_bo *src, uint64_t src_offset,
+ struct anv_bo *dest, uint64_t dest_offset,
+ int width, int height, VkFormat copy_format)
+{
+ VkDevice vk_device = anv_device_to_handle(cmd_buffer->device);
+
+ VkImageCreateInfo image_info = {
+ .sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO,
+ .imageType = VK_IMAGE_TYPE_2D,
+ .format = copy_format,
+ .extent = {
+ .width = width,
+ .height = height,
+ .depth = 1,
+ },
+ .mipLevels = 1,
+ .arrayLayers = 1,
+ .samples = 1,
+ .tiling = VK_IMAGE_TILING_LINEAR,
+ .usage = 0,
+ .flags = 0,
+ };
+
+ VkImage src_image;
+ image_info.usage = VK_IMAGE_USAGE_SAMPLED_BIT;
+ anv_CreateImage(vk_device, &image_info,
+ &cmd_buffer->pool->alloc, &src_image);
+
+ VkImage dest_image;
+ image_info.usage = VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT;
+ anv_CreateImage(vk_device, &image_info,
+ &cmd_buffer->pool->alloc, &dest_image);
+
+ /* We could use a vk call to bind memory, but that would require
+ * creating a dummy memory object etc. so there's really no point.
+ */
+ anv_image_from_handle(src_image)->bo = src;
+ anv_image_from_handle(src_image)->offset = src_offset;
+ anv_image_from_handle(dest_image)->bo = dest;
+ anv_image_from_handle(dest_image)->offset = dest_offset;
+
+ struct anv_image_view src_iview;
+ anv_image_view_init(&src_iview, cmd_buffer->device,
+ &(VkImageViewCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
+ .image = src_image,
+ .viewType = VK_IMAGE_VIEW_TYPE_2D,
+ .format = copy_format,
+ .subresourceRange = {
+ .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
+ .baseMipLevel = 0,
+ .levelCount = 1,
+ .baseArrayLayer = 0,
+ .layerCount = 1
+ },
+ },
+ cmd_buffer);
+
+ struct anv_image_view dest_iview;
+ anv_image_view_init(&dest_iview, cmd_buffer->device,
+ &(VkImageViewCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
+ .image = dest_image,
+ .viewType = VK_IMAGE_VIEW_TYPE_2D,
+ .format = copy_format,
+ .subresourceRange = {
+ .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
+ .baseMipLevel = 0,
+ .levelCount = 1,
+ .baseArrayLayer = 0,
+ .layerCount = 1,
+ },
+ },
+ cmd_buffer);
+
+ meta_emit_blit(cmd_buffer,
+ anv_image_from_handle(src_image),
+ &src_iview,
+ (VkOffset3D) { 0, 0, 0 },
+ (VkExtent3D) { width, height, 1 },
+ anv_image_from_handle(dest_image),
+ &dest_iview,
+ (VkOffset3D) { 0, 0, 0 },
+ (VkExtent3D) { width, height, 1 },
+ VK_FILTER_NEAREST);
+
+ anv_DestroyImage(vk_device, src_image, &cmd_buffer->pool->alloc);
+ anv_DestroyImage(vk_device, dest_image, &cmd_buffer->pool->alloc);
+}
+
+void anv_CmdCopyBuffer(
+ VkCommandBuffer commandBuffer,
+ VkBuffer srcBuffer,
+ VkBuffer destBuffer,
+ uint32_t regionCount,
+ const VkBufferCopy* pRegions)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_buffer, src_buffer, srcBuffer);
+ ANV_FROM_HANDLE(anv_buffer, dest_buffer, destBuffer);
+
+ struct anv_meta_saved_state saved_state;
+
+ meta_prepare_blit(cmd_buffer, &saved_state);
+
+ for (unsigned r = 0; r < regionCount; r++) {
+ uint64_t src_offset = src_buffer->offset + pRegions[r].srcOffset;
+ uint64_t dest_offset = dest_buffer->offset + pRegions[r].dstOffset;
+ uint64_t copy_size = pRegions[r].size;
+
+ /* First, we compute the biggest format that can be used with the
+ * given offsets and size.
+ */
+ int bs = 16;
+
+ int fs = ffs(src_offset) - 1;
+ if (fs != -1)
+ bs = MIN2(bs, 1 << fs);
+ assert(src_offset % bs == 0);
+
+ fs = ffs(dest_offset) - 1;
+ if (fs != -1)
+ bs = MIN2(bs, 1 << fs);
+ assert(dest_offset % bs == 0);
+
+ fs = ffs(pRegions[r].size) - 1;
+ if (fs != -1)
+ bs = MIN2(bs, 1 << fs);
+ assert(pRegions[r].size % bs == 0);
+
+ VkFormat copy_format = vk_format_for_size(bs);
+
+ /* This is maximum possible width/height our HW can handle */
+ uint64_t max_surface_dim = 1 << 14;
+
+ /* First, we make a bunch of max-sized copies */
+ uint64_t max_copy_size = max_surface_dim * max_surface_dim * bs;
+ while (copy_size > max_copy_size) {
+ do_buffer_copy(cmd_buffer, src_buffer->bo, src_offset,
+ dest_buffer->bo, dest_offset,
+ max_surface_dim, max_surface_dim, copy_format);
+ copy_size -= max_copy_size;
+ src_offset += max_copy_size;
+ dest_offset += max_copy_size;
+ }
+
+ uint64_t height = copy_size / (max_surface_dim * bs);
+ assert(height < max_surface_dim);
+ if (height != 0) {
+ uint64_t rect_copy_size = height * max_surface_dim * bs;
+ do_buffer_copy(cmd_buffer, src_buffer->bo, src_offset,
+ dest_buffer->bo, dest_offset,
+ max_surface_dim, height, copy_format);
+ copy_size -= rect_copy_size;
+ src_offset += rect_copy_size;
+ dest_offset += rect_copy_size;
+ }
+
+ if (copy_size != 0) {
+ do_buffer_copy(cmd_buffer, src_buffer->bo, src_offset,
+ dest_buffer->bo, dest_offset,
+ copy_size / bs, 1, copy_format);
+ }
+ }
+
+ meta_finish_blit(cmd_buffer, &saved_state);
+}
+
+void anv_CmdUpdateBuffer(
+ VkCommandBuffer commandBuffer,
+ VkBuffer dstBuffer,
+ VkDeviceSize dstOffset,
+ VkDeviceSize dataSize,
+ const uint32_t* pData)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_buffer, dst_buffer, dstBuffer);
+ struct anv_meta_saved_state saved_state;
+
+ meta_prepare_blit(cmd_buffer, &saved_state);
+
+ /* We can't quite grab a full block because the state stream needs a
+ * little data at the top to build its linked list.
+ */
+ const uint32_t max_update_size =
+ cmd_buffer->device->dynamic_state_block_pool.block_size - 64;
+
+ assert(max_update_size < (1 << 14) * 4);
+
+ while (dataSize) {
+ const uint32_t copy_size = MIN2(dataSize, max_update_size);
+
+ struct anv_state tmp_data =
+ anv_cmd_buffer_alloc_dynamic_state(cmd_buffer, copy_size, 64);
+
+ memcpy(tmp_data.map, pData, copy_size);
+
+ VkFormat format;
+ int bs;
+ if ((copy_size & 15) == 0 && (dstOffset & 15) == 0) {
+ format = VK_FORMAT_R32G32B32A32_UINT;
+ bs = 16;
+ } else if ((copy_size & 7) == 0 && (dstOffset & 7) == 0) {
+ format = VK_FORMAT_R32G32_UINT;
+ bs = 8;
+ } else {
+ assert((copy_size & 3) == 0 && (dstOffset & 3) == 0);
+ format = VK_FORMAT_R32_UINT;
+ bs = 4;
+ }
+
+ do_buffer_copy(cmd_buffer,
+ &cmd_buffer->device->dynamic_state_block_pool.bo,
+ tmp_data.offset,
+ dst_buffer->bo, dst_buffer->offset + dstOffset,
+ copy_size / bs, 1, format);
+
+ dataSize -= copy_size;
+ dstOffset += copy_size;
+ pData = (void *)pData + copy_size;
+ }
+}
+
+static VkFormat
+choose_iview_format(struct anv_image *image, VkImageAspectFlagBits aspect)
+{
+ assert(__builtin_popcount(aspect) == 1);
+
+ struct isl_surf *surf =
+ &anv_image_get_surface_for_aspect_mask(image, aspect)->isl;
+
+ /* vkCmdCopyImage behaves like memcpy. Therefore we choose identical UINT
+ * formats for the source and destination image views.
+ *
+ * From the Vulkan spec (2015-12-30):
+ *
+ * vkCmdCopyImage performs image copies in a similar manner to a host
+ * memcpy. It does not perform general-purpose conversions such as
+ * scaling, resizing, blending, color-space conversion, or format
+ * conversions. Rather, it simply copies raw image data. vkCmdCopyImage
+ * can copy between images with different formats, provided the formats
+ * are compatible as defined below.
+ *
+ * [The spec later defines compatibility as having the same number of
+ * bytes per block].
+ */
+ return vk_format_for_size(isl_format_layouts[surf->format].bs);
+}
+
+static VkFormat
+choose_buffer_format(struct anv_image *image, VkImageAspectFlagBits aspect)
+{
+ assert(__builtin_popcount(aspect) == 1);
+
+ /* vkCmdCopy* commands behave like memcpy. Therefore we choose
+ * compatable UINT formats for the source and destination image views.
+ *
+ * For the buffer, we go back to the original image format and get a
+ * the format as if it were linear. This way, for RGB formats, we get
+ * an RGB format here even if the tiled image is RGBA. XXX: This doesn't
+ * work if the buffer is the destination.
+ */
+ enum isl_format linear_format = anv_get_isl_format(image->vk_format, aspect,
+ VK_IMAGE_TILING_LINEAR);
+
+ return vk_format_for_size(isl_format_layouts[linear_format].bs);
+}
+
+void anv_CmdCopyImage(
+ VkCommandBuffer commandBuffer,
+ VkImage srcImage,
+ VkImageLayout srcImageLayout,
+ VkImage destImage,
+ VkImageLayout destImageLayout,
+ uint32_t regionCount,
+ const VkImageCopy* pRegions)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_image, src_image, srcImage);
+ ANV_FROM_HANDLE(anv_image, dest_image, destImage);
+
+ struct anv_meta_saved_state saved_state;
+
+ meta_prepare_blit(cmd_buffer, &saved_state);
+
+ for (unsigned r = 0; r < regionCount; r++) {
+ assert(pRegions[r].srcSubresource.aspectMask ==
+ pRegions[r].dstSubresource.aspectMask);
+
+ VkImageAspectFlags aspect = pRegions[r].srcSubresource.aspectMask;
+
+ VkFormat src_format = choose_iview_format(src_image, aspect);
+ VkFormat dst_format = choose_iview_format(dest_image, aspect);
+
+ struct anv_image_view src_iview;
+ anv_image_view_init(&src_iview, cmd_buffer->device,
+ &(VkImageViewCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
+ .image = srcImage,
+ .viewType = anv_meta_get_view_type(src_image),
+ .format = src_format,
+ .subresourceRange = {
+ .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
+ .baseMipLevel = pRegions[r].srcSubresource.mipLevel,
+ .levelCount = 1,
+ .baseArrayLayer = pRegions[r].srcSubresource.baseArrayLayer,
+ .layerCount = pRegions[r].dstSubresource.layerCount,
+ },
+ },
+ cmd_buffer);
+
+ const VkOffset3D dest_offset = {
+ .x = pRegions[r].dstOffset.x,
+ .y = pRegions[r].dstOffset.y,
+ .z = 0,
+ };
+
+ unsigned num_slices;
+ if (src_image->type == VK_IMAGE_TYPE_3D) {
+ assert(pRegions[r].srcSubresource.layerCount == 1 &&
+ pRegions[r].dstSubresource.layerCount == 1);
+ num_slices = pRegions[r].extent.depth;
+ } else {
+ assert(pRegions[r].srcSubresource.layerCount ==
+ pRegions[r].dstSubresource.layerCount);
+ assert(pRegions[r].extent.depth == 1);
+ num_slices = pRegions[r].dstSubresource.layerCount;
+ }
+
+ const uint32_t dest_base_array_slice =
+ meta_blit_get_dest_view_base_array_slice(dest_image,
+ &pRegions[r].dstSubresource,
+ &pRegions[r].dstOffset);
+
+ for (unsigned slice = 0; slice < num_slices; slice++) {
+ VkOffset3D src_offset = pRegions[r].srcOffset;
+ src_offset.z += slice;
+
+ struct anv_image_view dest_iview;
+ anv_image_view_init(&dest_iview, cmd_buffer->device,
+ &(VkImageViewCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
+ .image = destImage,
+ .viewType = anv_meta_get_view_type(dest_image),
+ .format = dst_format,
+ .subresourceRange = {
+ .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
+ .baseMipLevel = pRegions[r].dstSubresource.mipLevel,
+ .levelCount = 1,
+ .baseArrayLayer = dest_base_array_slice + slice,
+ .layerCount = 1
+ },
+ },
+ cmd_buffer);
+
+ meta_emit_blit(cmd_buffer,
+ src_image, &src_iview,
+ src_offset,
+ pRegions[r].extent,
+ dest_image, &dest_iview,
+ dest_offset,
+ pRegions[r].extent,
+ VK_FILTER_NEAREST);
+ }
+ }
+
+ meta_finish_blit(cmd_buffer, &saved_state);
+}
+
+void anv_CmdBlitImage(
+ VkCommandBuffer commandBuffer,
+ VkImage srcImage,
+ VkImageLayout srcImageLayout,
+ VkImage destImage,
+ VkImageLayout destImageLayout,
+ uint32_t regionCount,
+ const VkImageBlit* pRegions,
+ VkFilter filter)
+
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_image, src_image, srcImage);
+ ANV_FROM_HANDLE(anv_image, dest_image, destImage);
+
+ struct anv_meta_saved_state saved_state;
+
+ anv_finishme("respect VkFilter");
+
+ meta_prepare_blit(cmd_buffer, &saved_state);
+
+ for (unsigned r = 0; r < regionCount; r++) {
+ struct anv_image_view src_iview;
+ anv_image_view_init(&src_iview, cmd_buffer->device,
+ &(VkImageViewCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
+ .image = srcImage,
+ .viewType = anv_meta_get_view_type(src_image),
+ .format = src_image->vk_format,
+ .subresourceRange = {
+ .aspectMask = pRegions[r].srcSubresource.aspectMask,
+ .baseMipLevel = pRegions[r].srcSubresource.mipLevel,
+ .levelCount = 1,
+ .baseArrayLayer = pRegions[r].srcSubresource.baseArrayLayer,
+ .layerCount = 1
+ },
+ },
+ cmd_buffer);
+
+ const VkOffset3D dest_offset = {
+ .x = pRegions[r].dstOffsets[0].x,
+ .y = pRegions[r].dstOffsets[0].y,
+ .z = 0,
+ };
+
+ if (pRegions[r].dstOffsets[1].x < pRegions[r].dstOffsets[0].x ||
+ pRegions[r].dstOffsets[1].y < pRegions[r].dstOffsets[0].y ||
+ pRegions[r].srcOffsets[1].x < pRegions[r].srcOffsets[0].x ||
+ pRegions[r].srcOffsets[1].y < pRegions[r].srcOffsets[0].y)
+ anv_finishme("FINISHME: Allow flipping in blits");
+
+ const VkExtent3D dest_extent = {
+ .width = pRegions[r].dstOffsets[1].x - pRegions[r].dstOffsets[0].x,
+ .height = pRegions[r].dstOffsets[1].y - pRegions[r].dstOffsets[0].y,
+ };
+
+ const VkExtent3D src_extent = {
+ .width = pRegions[r].srcOffsets[1].x - pRegions[r].srcOffsets[0].x,
+ .height = pRegions[r].srcOffsets[1].y - pRegions[r].srcOffsets[0].y,
+ };
+
+ const uint32_t dest_array_slice =
+ meta_blit_get_dest_view_base_array_slice(dest_image,
+ &pRegions[r].dstSubresource,
+ &pRegions[r].dstOffsets[0]);
+
+ if (pRegions[r].srcSubresource.layerCount > 1)
+ anv_finishme("FINISHME: copy multiple array layers");
+
+ if (pRegions[r].srcOffsets[0].z + 1 != pRegions[r].srcOffsets[1].z ||
+ pRegions[r].dstOffsets[0].z + 1 != pRegions[r].dstOffsets[1].z)
+ anv_finishme("FINISHME: copy multiple depth layers");
+
+ struct anv_image_view dest_iview;
+ anv_image_view_init(&dest_iview, cmd_buffer->device,
+ &(VkImageViewCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
+ .image = destImage,
+ .viewType = anv_meta_get_view_type(dest_image),
+ .format = dest_image->vk_format,
+ .subresourceRange = {
+ .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
+ .baseMipLevel = pRegions[r].dstSubresource.mipLevel,
+ .levelCount = 1,
+ .baseArrayLayer = dest_array_slice,
+ .layerCount = 1
+ },
+ },
+ cmd_buffer);
+
+ meta_emit_blit(cmd_buffer,
+ src_image, &src_iview,
+ pRegions[r].srcOffsets[0], src_extent,
+ dest_image, &dest_iview,
+ dest_offset, dest_extent,
+ filter);
+ }
+
+ meta_finish_blit(cmd_buffer, &saved_state);
+}
+
+static struct anv_image *
+make_image_for_buffer(VkDevice vk_device, VkBuffer vk_buffer, VkFormat format,
+ VkImageUsageFlags usage,
+ VkImageType image_type,
+ const VkAllocationCallbacks *alloc,
+ const VkBufferImageCopy *copy)
+{
+ ANV_FROM_HANDLE(anv_buffer, buffer, vk_buffer);
+
+ VkExtent3D extent = copy->imageExtent;
+ if (copy->bufferRowLength)
+ extent.width = copy->bufferRowLength;
+ if (copy->bufferImageHeight)
+ extent.height = copy->bufferImageHeight;
+ extent.depth = 1;
+
+ VkImage vk_image;
+ VkResult result = anv_CreateImage(vk_device,
+ &(VkImageCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO,
+ .imageType = VK_IMAGE_TYPE_2D,
+ .format = format,
+ .extent = extent,
+ .mipLevels = 1,
+ .arrayLayers = 1,
+ .samples = 1,
+ .tiling = VK_IMAGE_TILING_LINEAR,
+ .usage = usage,
+ .flags = 0,
+ }, alloc, &vk_image);
+ assert(result == VK_SUCCESS);
+
+ ANV_FROM_HANDLE(anv_image, image, vk_image);
+
+ /* We could use a vk call to bind memory, but that would require
+ * creating a dummy memory object etc. so there's really no point.
+ */
+ image->bo = buffer->bo;
+ image->offset = buffer->offset + copy->bufferOffset;
+
+ return image;
+}
+
+void anv_CmdCopyBufferToImage(
+ VkCommandBuffer commandBuffer,
+ VkBuffer srcBuffer,
+ VkImage destImage,
+ VkImageLayout destImageLayout,
+ uint32_t regionCount,
+ const VkBufferImageCopy* pRegions)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_image, dest_image, destImage);
+ VkDevice vk_device = anv_device_to_handle(cmd_buffer->device);
+ struct anv_meta_saved_state saved_state;
+
+ meta_prepare_blit(cmd_buffer, &saved_state);
+
+ for (unsigned r = 0; r < regionCount; r++) {
+ VkImageAspectFlags aspect = pRegions[r].imageSubresource.aspectMask;
+
+ VkFormat image_format = choose_iview_format(dest_image, aspect);
+ VkFormat buffer_format = choose_buffer_format(dest_image, aspect);
+
+ struct anv_image *src_image =
+ make_image_for_buffer(vk_device, srcBuffer, buffer_format,
+ VK_IMAGE_USAGE_SAMPLED_BIT,
+ dest_image->type, &cmd_buffer->pool->alloc,
+ &pRegions[r]);
+
+ const uint32_t dest_base_array_slice =
+ meta_blit_get_dest_view_base_array_slice(dest_image,
+ &pRegions[r].imageSubresource,
+ &pRegions[r].imageOffset);
+
+ unsigned num_slices;
+ if (dest_image->type == VK_IMAGE_TYPE_3D) {
+ assert(pRegions[r].imageSubresource.layerCount == 1);
+ num_slices = pRegions[r].imageExtent.depth;
+ } else {
+ assert(pRegions[r].imageExtent.depth == 1);
+ num_slices = pRegions[r].imageSubresource.layerCount;
+ }
+
+ for (unsigned slice = 0; slice < num_slices; slice++) {
+ struct anv_image_view src_iview;
+ anv_image_view_init(&src_iview, cmd_buffer->device,
+ &(VkImageViewCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
+ .image = anv_image_to_handle(src_image),
+ .viewType = VK_IMAGE_VIEW_TYPE_2D,
+ .format = buffer_format,
+ .subresourceRange = {
+ .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
+ .baseMipLevel = 0,
+ .levelCount = 1,
+ .baseArrayLayer = 0,
+ .layerCount = 1,
+ },
+ },
+ cmd_buffer);
+
+ struct anv_image_view dest_iview;
+ anv_image_view_init(&dest_iview, cmd_buffer->device,
+ &(VkImageViewCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
+ .image = anv_image_to_handle(dest_image),
+ .viewType = anv_meta_get_view_type(dest_image),
+ .format = image_format,
+ .subresourceRange = {
+ .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
+ .baseMipLevel = pRegions[r].imageSubresource.mipLevel,
+ .levelCount = 1,
+ .baseArrayLayer = dest_base_array_slice + slice,
+ .layerCount = 1
+ },
+ },
+ cmd_buffer);
+
+ VkOffset3D src_offset = { 0, 0, slice };
+
+ const VkOffset3D dest_offset = {
+ .x = pRegions[r].imageOffset.x,
+ .y = pRegions[r].imageOffset.y,
+ .z = 0,
+ };
+
+ meta_emit_blit(cmd_buffer,
+ src_image,
+ &src_iview,
+ src_offset,
+ pRegions[r].imageExtent,
+ dest_image,
+ &dest_iview,
+ dest_offset,
+ pRegions[r].imageExtent,
+ VK_FILTER_NEAREST);
+
+ /* Once we've done the blit, all of the actual information about
+ * the image is embedded in the command buffer so we can just
+ * increment the offset directly in the image effectively
+ * re-binding it to different backing memory.
+ */
+ src_image->offset += src_image->extent.width *
+ src_image->extent.height *
+ src_image->format->isl_layout->bs;
+ }
+
+ anv_DestroyImage(vk_device, anv_image_to_handle(src_image),
+ &cmd_buffer->pool->alloc);
+ }
+
+ meta_finish_blit(cmd_buffer, &saved_state);
+}
+
+void anv_CmdCopyImageToBuffer(
+ VkCommandBuffer commandBuffer,
+ VkImage srcImage,
+ VkImageLayout srcImageLayout,
+ VkBuffer destBuffer,
+ uint32_t regionCount,
+ const VkBufferImageCopy* pRegions)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_image, src_image, srcImage);
+ VkDevice vk_device = anv_device_to_handle(cmd_buffer->device);
+ struct anv_meta_saved_state saved_state;
+
+ meta_prepare_blit(cmd_buffer, &saved_state);
+
+ for (unsigned r = 0; r < regionCount; r++) {
+ VkImageAspectFlags aspect = pRegions[r].imageSubresource.aspectMask;
+
+ VkFormat image_format = choose_iview_format(src_image, aspect);
+ VkFormat buffer_format = choose_buffer_format(src_image, aspect);
+
+ struct anv_image_view src_iview;
+ anv_image_view_init(&src_iview, cmd_buffer->device,
+ &(VkImageViewCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
+ .image = srcImage,
+ .viewType = anv_meta_get_view_type(src_image),
+ .format = image_format,
+ .subresourceRange = {
+ .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
+ .baseMipLevel = pRegions[r].imageSubresource.mipLevel,
+ .levelCount = 1,
+ .baseArrayLayer = pRegions[r].imageSubresource.baseArrayLayer,
+ .layerCount = pRegions[r].imageSubresource.layerCount,
+ },
+ },
+ cmd_buffer);
+
+ struct anv_image *dest_image =
+ make_image_for_buffer(vk_device, destBuffer, buffer_format,
+ VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT,
+ src_image->type, &cmd_buffer->pool->alloc,
+ &pRegions[r]);
+
+ unsigned num_slices;
+ if (src_image->type == VK_IMAGE_TYPE_3D) {
+ assert(pRegions[r].imageSubresource.layerCount == 1);
+ num_slices = pRegions[r].imageExtent.depth;
+ } else {
+ assert(pRegions[r].imageExtent.depth == 1);
+ num_slices = pRegions[r].imageSubresource.layerCount;
+ }
+
+ for (unsigned slice = 0; slice < num_slices; slice++) {
+ VkOffset3D src_offset = pRegions[r].imageOffset;
+ src_offset.z += slice;
+
+ struct anv_image_view dest_iview;
+ anv_image_view_init(&dest_iview, cmd_buffer->device,
+ &(VkImageViewCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
+ .image = anv_image_to_handle(dest_image),
+ .viewType = VK_IMAGE_VIEW_TYPE_2D,
+ .format = buffer_format,
+ .subresourceRange = {
+ .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
+ .baseMipLevel = 0,
+ .levelCount = 1,
+ .baseArrayLayer = 0,
+ .layerCount = 1
+ },
+ },
+ cmd_buffer);
+
+ meta_emit_blit(cmd_buffer,
+ anv_image_from_handle(srcImage),
+ &src_iview,
+ src_offset,
+ pRegions[r].imageExtent,
+ dest_image,
+ &dest_iview,
+ (VkOffset3D) { 0, 0, 0 },
+ pRegions[r].imageExtent,
+ VK_FILTER_NEAREST);
+
+ /* Once we've done the blit, all of the actual information about
+ * the image is embedded in the command buffer so we can just
+ * increment the offset directly in the image effectively
+ * re-binding it to different backing memory.
+ */
+ dest_image->offset += dest_image->extent.width *
+ dest_image->extent.height *
+ src_image->format->isl_layout->bs;
+ }
+
+ anv_DestroyImage(vk_device, anv_image_to_handle(dest_image),
+ &cmd_buffer->pool->alloc);
+ }
+
+ meta_finish_blit(cmd_buffer, &saved_state);
+}
+
+void anv_CmdResolveImage(
+ VkCommandBuffer commandBuffer,
+ VkImage srcImage,
+ VkImageLayout srcImageLayout,
+ VkImage destImage,
+ VkImageLayout destImageLayout,
+ uint32_t regionCount,
+ const VkImageResolve* pRegions)
+{
+ stub();
+}
+
+static void *
+meta_alloc(void* _device, size_t size, size_t alignment,
+ VkSystemAllocationScope allocationScope)
+{
+ struct anv_device *device = _device;
+ return device->alloc.pfnAllocation(device->alloc.pUserData, size, alignment,
+ VK_SYSTEM_ALLOCATION_SCOPE_DEVICE);
+}
+
+static void *
+meta_realloc(void* _device, void *original, size_t size, size_t alignment,
+ VkSystemAllocationScope allocationScope)
+{
+ struct anv_device *device = _device;
+ return device->alloc.pfnReallocation(device->alloc.pUserData, original,
+ size, alignment,
+ VK_SYSTEM_ALLOCATION_SCOPE_DEVICE);
+}
+
+static void
+meta_free(void* _device, void *data)
+{
+ struct anv_device *device = _device;
+ return device->alloc.pfnFree(device->alloc.pUserData, data);
+}
+
+VkResult
+anv_device_init_meta(struct anv_device *device)
+{
+ device->meta_state.alloc = (VkAllocationCallbacks) {
+ .pUserData = device,
+ .pfnAllocation = meta_alloc,
+ .pfnReallocation = meta_realloc,
+ .pfnFree = meta_free,
+ };
+
+ VkResult result;
+ result = anv_device_init_meta_clear_state(device);
+ if (result != VK_SUCCESS)
+ return result;
+
+ result = anv_device_init_meta_blit_state(device);
+ if (result != VK_SUCCESS) {
+ anv_device_finish_meta_clear_state(device);
+ return result;
+ }
+
+ return VK_SUCCESS;
+}
+
+void
+anv_device_finish_meta(struct anv_device *device)
+{
+ anv_device_finish_meta_clear_state(device);
+
+ /* Blit */
+ anv_DestroyRenderPass(anv_device_to_handle(device),
+ device->meta_state.blit.render_pass,
+ &device->meta_state.alloc);
+ anv_DestroyPipeline(anv_device_to_handle(device),
+ device->meta_state.blit.pipeline_1d_src,
+ &device->meta_state.alloc);
+ anv_DestroyPipeline(anv_device_to_handle(device),
+ device->meta_state.blit.pipeline_2d_src,
+ &device->meta_state.alloc);
+ anv_DestroyPipeline(anv_device_to_handle(device),
+ device->meta_state.blit.pipeline_3d_src,
+ &device->meta_state.alloc);
+ anv_DestroyPipelineLayout(anv_device_to_handle(device),
+ device->meta_state.blit.pipeline_layout,
+ &device->meta_state.alloc);
+ anv_DestroyDescriptorSetLayout(anv_device_to_handle(device),
+ device->meta_state.blit.ds_layout,
+ &device->meta_state.alloc);
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#pragma once
+
+#include "anv_private.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#define ANV_META_VERTEX_BINDING_COUNT 2
+
+struct anv_meta_saved_state {
+ struct anv_vertex_binding old_vertex_bindings[ANV_META_VERTEX_BINDING_COUNT];
+ struct anv_descriptor_set *old_descriptor_set0;
+ struct anv_pipeline *old_pipeline;
+
+ /**
+ * Bitmask of (1 << VK_DYNAMIC_STATE_*). Defines the set of saved dynamic
+ * state.
+ */
+ uint32_t dynamic_mask;
+ struct anv_dynamic_state dynamic;
+};
+
+void
+anv_meta_save(struct anv_meta_saved_state *state,
+ const struct anv_cmd_buffer *cmd_buffer,
+ uint32_t dynamic_mask);
+
+void
+anv_meta_restore(const struct anv_meta_saved_state *state,
+ struct anv_cmd_buffer *cmd_buffer);
+
+VkImageViewType
+anv_meta_get_view_type(const struct anv_image *image);
+
+#ifdef __cplusplus
+}
+#endif
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include "anv_meta.h"
+#include "anv_meta_clear.h"
+#include "anv_private.h"
+#include "glsl/nir/nir_builder.h"
+
+/** Vertex attributes for color clears. */
+struct color_clear_vattrs {
+ struct anv_vue_header vue_header;
+ float position[2]; /**< 3DPRIM_RECTLIST */
+ VkClearColorValue color;
+};
+
+/** Vertex attributes for depthstencil clears. */
+struct depthstencil_clear_vattrs {
+ struct anv_vue_header vue_header;
+ float position[2]; /*<< 3DPRIM_RECTLIST */
+};
+
+static void
+meta_clear_begin(struct anv_meta_saved_state *saved_state,
+ struct anv_cmd_buffer *cmd_buffer)
+{
+ anv_meta_save(saved_state, cmd_buffer,
+ (1 << VK_DYNAMIC_STATE_VIEWPORT) |
+ (1 << VK_DYNAMIC_STATE_SCISSOR) |
+ (1 << VK_DYNAMIC_STATE_STENCIL_REFERENCE));
+
+ cmd_buffer->state.dynamic.viewport.count = 0;
+ cmd_buffer->state.dynamic.scissor.count = 0;
+}
+
+static void
+meta_clear_end(struct anv_meta_saved_state *saved_state,
+ struct anv_cmd_buffer *cmd_buffer)
+{
+ anv_meta_restore(saved_state, cmd_buffer);
+}
+
+static void
+build_color_shaders(struct nir_shader **out_vs,
+ struct nir_shader **out_fs,
+ uint32_t frag_output)
+{
+ nir_builder vs_b;
+ nir_builder fs_b;
+
+ nir_builder_init_simple_shader(&vs_b, NULL, MESA_SHADER_VERTEX, NULL);
+ nir_builder_init_simple_shader(&fs_b, NULL, MESA_SHADER_FRAGMENT, NULL);
+
+ vs_b.shader->info.name = ralloc_strdup(vs_b.shader, "meta_clear_color_vs");
+ fs_b.shader->info.name = ralloc_strdup(fs_b.shader, "meta_clear_color_fs");
+
+ const struct glsl_type *position_type = glsl_vec4_type();
+ const struct glsl_type *color_type = glsl_vec4_type();
+
+ nir_variable *vs_in_pos =
+ nir_variable_create(vs_b.shader, nir_var_shader_in, position_type,
+ "a_position");
+ vs_in_pos->data.location = VERT_ATTRIB_GENERIC0;
+
+ nir_variable *vs_out_pos =
+ nir_variable_create(vs_b.shader, nir_var_shader_out, position_type,
+ "gl_Position");
+ vs_out_pos->data.location = VARYING_SLOT_POS;
+
+ nir_variable *vs_in_color =
+ nir_variable_create(vs_b.shader, nir_var_shader_in, color_type,
+ "a_color");
+ vs_in_color->data.location = VERT_ATTRIB_GENERIC1;
+
+ nir_variable *vs_out_color =
+ nir_variable_create(vs_b.shader, nir_var_shader_out, color_type,
+ "v_color");
+ vs_out_color->data.location = VARYING_SLOT_VAR0;
+ vs_out_color->data.interpolation = INTERP_QUALIFIER_FLAT;
+
+ nir_variable *fs_in_color =
+ nir_variable_create(fs_b.shader, nir_var_shader_in, color_type,
+ "v_color");
+ fs_in_color->data.location = vs_out_color->data.location;
+ fs_in_color->data.interpolation = vs_out_color->data.interpolation;
+
+ nir_variable *fs_out_color =
+ nir_variable_create(fs_b.shader, nir_var_shader_out, color_type,
+ "f_color");
+ fs_out_color->data.location = FRAG_RESULT_DATA0 + frag_output;
+
+ nir_copy_var(&vs_b, vs_out_pos, vs_in_pos);
+ nir_copy_var(&vs_b, vs_out_color, vs_in_color);
+ nir_copy_var(&fs_b, fs_out_color, fs_in_color);
+
+ *out_vs = vs_b.shader;
+ *out_fs = fs_b.shader;
+}
+
+static VkResult
+create_pipeline(struct anv_device *device,
+ struct nir_shader *vs_nir,
+ struct nir_shader *fs_nir,
+ const VkPipelineVertexInputStateCreateInfo *vi_state,
+ const VkPipelineDepthStencilStateCreateInfo *ds_state,
+ const VkPipelineColorBlendStateCreateInfo *cb_state,
+ const VkAllocationCallbacks *alloc,
+ bool use_repclear,
+ struct anv_pipeline **pipeline)
+{
+ VkDevice device_h = anv_device_to_handle(device);
+ VkResult result;
+
+ struct anv_shader_module vs_m = { .nir = vs_nir };
+ struct anv_shader_module fs_m = { .nir = fs_nir };
+
+ VkPipeline pipeline_h = VK_NULL_HANDLE;
+ result = anv_graphics_pipeline_create(device_h,
+ VK_NULL_HANDLE,
+ &(VkGraphicsPipelineCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO,
+ .stageCount = 2,
+ .pStages = (VkPipelineShaderStageCreateInfo[]) {
+ {
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO,
+ .stage = VK_SHADER_STAGE_VERTEX_BIT,
+ .module = anv_shader_module_to_handle(&vs_m),
+ .pName = "main",
+ },
+ {
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO,
+ .stage = VK_SHADER_STAGE_FRAGMENT_BIT,
+ .module = anv_shader_module_to_handle(&fs_m),
+ .pName = "main",
+ },
+ },
+ .pVertexInputState = vi_state,
+ .pInputAssemblyState = &(VkPipelineInputAssemblyStateCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO,
+ .topology = VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP,
+ .primitiveRestartEnable = false,
+ },
+ .pViewportState = &(VkPipelineViewportStateCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO,
+ .viewportCount = 1,
+ .pViewports = NULL, /* dynamic */
+ .scissorCount = 1,
+ .pScissors = NULL, /* dynamic */
+ },
+ .pRasterizationState = &(VkPipelineRasterizationStateCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_STATE_CREATE_INFO,
+ .rasterizerDiscardEnable = false,
+ .polygonMode = VK_POLYGON_MODE_FILL,
+ .cullMode = VK_CULL_MODE_NONE,
+ .frontFace = VK_FRONT_FACE_COUNTER_CLOCKWISE,
+ .depthBiasEnable = false,
+ },
+ .pMultisampleState = &(VkPipelineMultisampleStateCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO,
+ .rasterizationSamples = 1, /* FINISHME: Multisampling */
+ .sampleShadingEnable = false,
+ .pSampleMask = (VkSampleMask[]) { UINT32_MAX },
+ .alphaToCoverageEnable = false,
+ .alphaToOneEnable = false,
+ },
+ .pDepthStencilState = ds_state,
+ .pColorBlendState = cb_state,
+ .pDynamicState = &(VkPipelineDynamicStateCreateInfo) {
+ /* The meta clear pipeline declares all state as dynamic.
+ * As a consequence, vkCmdBindPipeline writes no dynamic state
+ * to the cmd buffer. Therefore, at the end of the meta clear,
+ * we need only restore dynamic state was vkCmdSet.
+ */
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_DYNAMIC_STATE_CREATE_INFO,
+ .dynamicStateCount = 9,
+ .pDynamicStates = (VkDynamicState[]) {
+ VK_DYNAMIC_STATE_VIEWPORT,
+ VK_DYNAMIC_STATE_SCISSOR,
+ VK_DYNAMIC_STATE_LINE_WIDTH,
+ VK_DYNAMIC_STATE_DEPTH_BIAS,
+ VK_DYNAMIC_STATE_BLEND_CONSTANTS,
+ VK_DYNAMIC_STATE_DEPTH_BOUNDS,
+ VK_DYNAMIC_STATE_STENCIL_COMPARE_MASK,
+ VK_DYNAMIC_STATE_STENCIL_WRITE_MASK,
+ VK_DYNAMIC_STATE_STENCIL_REFERENCE,
+ },
+ },
+ .flags = 0,
+ .renderPass = anv_render_pass_to_handle(&anv_meta_dummy_renderpass),
+ .subpass = 0,
+ },
+ &(struct anv_graphics_pipeline_create_info) {
+ .color_attachment_count = MAX_RTS,
+ .use_repclear = use_repclear,
+ .disable_viewport = true,
+ .disable_vs = true,
+ .use_rectlist = true
+ },
+ alloc,
+ &pipeline_h);
+
+ ralloc_free(vs_nir);
+ ralloc_free(fs_nir);
+
+ *pipeline = anv_pipeline_from_handle(pipeline_h);
+
+ return result;
+}
+
+static VkResult
+create_color_pipeline(struct anv_device *device, uint32_t frag_output,
+ struct anv_pipeline **pipeline)
+{
+ struct nir_shader *vs_nir;
+ struct nir_shader *fs_nir;
+ build_color_shaders(&vs_nir, &fs_nir, frag_output);
+
+ const VkPipelineVertexInputStateCreateInfo vi_state = {
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO,
+ .vertexBindingDescriptionCount = 1,
+ .pVertexBindingDescriptions = (VkVertexInputBindingDescription[]) {
+ {
+ .binding = 0,
+ .stride = sizeof(struct color_clear_vattrs),
+ .inputRate = VK_VERTEX_INPUT_RATE_VERTEX
+ },
+ },
+ .vertexAttributeDescriptionCount = 3,
+ .pVertexAttributeDescriptions = (VkVertexInputAttributeDescription[]) {
+ {
+ /* VUE Header */
+ .location = 0,
+ .binding = 0,
+ .format = VK_FORMAT_R32G32B32A32_UINT,
+ .offset = offsetof(struct color_clear_vattrs, vue_header),
+ },
+ {
+ /* Position */
+ .location = 1,
+ .binding = 0,
+ .format = VK_FORMAT_R32G32_SFLOAT,
+ .offset = offsetof(struct color_clear_vattrs, position),
+ },
+ {
+ /* Color */
+ .location = 2,
+ .binding = 0,
+ .format = VK_FORMAT_R32G32B32A32_SFLOAT,
+ .offset = offsetof(struct color_clear_vattrs, color),
+ },
+ },
+ };
+
+ const VkPipelineDepthStencilStateCreateInfo ds_state = {
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO,
+ .depthTestEnable = false,
+ .depthWriteEnable = false,
+ .depthBoundsTestEnable = false,
+ .stencilTestEnable = false,
+ };
+
+ const VkPipelineColorBlendStateCreateInfo cb_state = {
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO,
+ .logicOpEnable = false,
+ .attachmentCount = 1,
+ .pAttachments = (VkPipelineColorBlendAttachmentState []) {
+ {
+ .blendEnable = false,
+ .colorWriteMask = VK_COLOR_COMPONENT_A_BIT |
+ VK_COLOR_COMPONENT_R_BIT |
+ VK_COLOR_COMPONENT_G_BIT |
+ VK_COLOR_COMPONENT_B_BIT,
+ },
+ },
+ };
+
+ /* Disable repclear because we do not want the compiler to replace the
+ * shader. We need the shader to write to the specified color attachment,
+ * but the repclear shader writes to all color attachments.
+ */
+ return
+ create_pipeline(device, vs_nir, fs_nir, &vi_state, &ds_state,
+ &cb_state, &device->meta_state.alloc,
+ /*use_repclear*/ false, pipeline);
+}
+
+static void
+free_color_pipelines(struct anv_device *device)
+{
+ for (uint32_t i = 0;
+ i < ARRAY_SIZE(device->meta_state.clear.color_pipelines); ++i) {
+ if (device->meta_state.clear.color_pipelines[i] == NULL)
+ continue;
+
+ ANV_CALL(DestroyPipeline)(
+ anv_device_to_handle(device),
+ anv_pipeline_to_handle(device->meta_state.clear.color_pipelines[i]),
+ &device->meta_state.alloc);
+ }
+}
+
+static VkResult
+init_color_pipelines(struct anv_device *device)
+{
+ VkResult result;
+ struct anv_pipeline **pipelines = device->meta_state.clear.color_pipelines;
+ uint32_t n = ARRAY_SIZE(device->meta_state.clear.color_pipelines);
+
+ zero(device->meta_state.clear.color_pipelines);
+
+ for (uint32_t i = 0; i < n; ++i) {
+ result = create_color_pipeline(device, i, &pipelines[i]);
+ if (result < 0)
+ goto fail;
+ }
+
+ return VK_SUCCESS;
+
+fail:
+ free_color_pipelines(device);
+
+ return result;
+}
+
+static void
+emit_color_clear(struct anv_cmd_buffer *cmd_buffer,
+ const VkClearAttachment *clear_att,
+ const VkClearRect *clear_rect)
+{
+ struct anv_device *device = cmd_buffer->device;
+ const struct anv_subpass *subpass = cmd_buffer->state.subpass;
+ const struct anv_framebuffer *fb = cmd_buffer->state.framebuffer;
+ VkClearColorValue clear_value = clear_att->clearValue.color;
+ struct anv_pipeline *pipeline =
+ device->meta_state.clear.color_pipelines[clear_att->colorAttachment];
+
+ VkCommandBuffer cmd_buffer_h = anv_cmd_buffer_to_handle(cmd_buffer);
+ VkPipeline pipeline_h = anv_pipeline_to_handle(pipeline);
+
+ assert(clear_att->aspectMask == VK_IMAGE_ASPECT_COLOR_BIT);
+ assert(clear_att->colorAttachment < subpass->color_count);
+
+ const struct color_clear_vattrs vertex_data[3] = {
+ {
+ .vue_header = { 0 },
+ .position = {
+ clear_rect->rect.offset.x,
+ clear_rect->rect.offset.y,
+ },
+ .color = clear_value,
+ },
+ {
+ .vue_header = { 0 },
+ .position = {
+ clear_rect->rect.offset.x + clear_rect->rect.extent.width,
+ clear_rect->rect.offset.y,
+ },
+ .color = clear_value,
+ },
+ {
+ .vue_header = { 0 },
+ .position = {
+ clear_rect->rect.offset.x + clear_rect->rect.extent.width,
+ clear_rect->rect.offset.y + clear_rect->rect.extent.height,
+ },
+ .color = clear_value,
+ },
+ };
+
+ struct anv_state state =
+ anv_cmd_buffer_emit_dynamic(cmd_buffer, vertex_data, sizeof(vertex_data), 16);
+
+ struct anv_buffer vertex_buffer = {
+ .device = device,
+ .size = sizeof(vertex_data),
+ .bo = &device->dynamic_state_block_pool.bo,
+ .offset = state.offset,
+ };
+
+ ANV_CALL(CmdSetViewport)(cmd_buffer_h, 0, 1,
+ (VkViewport[]) {
+ {
+ .x = 0,
+ .y = 0,
+ .width = fb->width,
+ .height = fb->height,
+ .minDepth = 0.0,
+ .maxDepth = 1.0,
+ },
+ });
+
+ ANV_CALL(CmdSetScissor)(cmd_buffer_h, 0, 1,
+ (VkRect2D[]) {
+ {
+ .offset = { 0, 0 },
+ .extent = { fb->width, fb->height },
+ }
+ });
+
+ ANV_CALL(CmdBindVertexBuffers)(cmd_buffer_h, 0, 1,
+ (VkBuffer[]) { anv_buffer_to_handle(&vertex_buffer) },
+ (VkDeviceSize[]) { 0 });
+
+ if (cmd_buffer->state.pipeline != pipeline) {
+ ANV_CALL(CmdBindPipeline)(cmd_buffer_h, VK_PIPELINE_BIND_POINT_GRAPHICS,
+ pipeline_h);
+ }
+
+ ANV_CALL(CmdDraw)(cmd_buffer_h, 3, 1, 0, 0);
+}
+
+
+static void
+build_depthstencil_shaders(struct nir_shader **out_vs,
+ struct nir_shader **out_fs)
+{
+ nir_builder vs_b;
+ nir_builder fs_b;
+
+ nir_builder_init_simple_shader(&vs_b, NULL, MESA_SHADER_VERTEX, NULL);
+ nir_builder_init_simple_shader(&fs_b, NULL, MESA_SHADER_FRAGMENT, NULL);
+
+ vs_b.shader->info.name = ralloc_strdup(vs_b.shader, "meta_clear_depthstencil_vs");
+ fs_b.shader->info.name = ralloc_strdup(fs_b.shader, "meta_clear_depthstencil_fs");
+
+ const struct glsl_type *position_type = glsl_vec4_type();
+
+ nir_variable *vs_in_pos =
+ nir_variable_create(vs_b.shader, nir_var_shader_in, position_type,
+ "a_position");
+ vs_in_pos->data.location = VERT_ATTRIB_GENERIC0;
+
+ nir_variable *vs_out_pos =
+ nir_variable_create(vs_b.shader, nir_var_shader_out, position_type,
+ "gl_Position");
+ vs_out_pos->data.location = VARYING_SLOT_POS;
+
+ nir_copy_var(&vs_b, vs_out_pos, vs_in_pos);
+
+ *out_vs = vs_b.shader;
+ *out_fs = fs_b.shader;
+}
+
+static VkResult
+create_depthstencil_pipeline(struct anv_device *device,
+ VkImageAspectFlags aspects,
+ struct anv_pipeline **pipeline)
+{
+ struct nir_shader *vs_nir;
+ struct nir_shader *fs_nir;
+
+ build_depthstencil_shaders(&vs_nir, &fs_nir);
+
+ const VkPipelineVertexInputStateCreateInfo vi_state = {
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO,
+ .vertexBindingDescriptionCount = 1,
+ .pVertexBindingDescriptions = (VkVertexInputBindingDescription[]) {
+ {
+ .binding = 0,
+ .stride = sizeof(struct depthstencil_clear_vattrs),
+ .inputRate = VK_VERTEX_INPUT_RATE_VERTEX
+ },
+ },
+ .vertexAttributeDescriptionCount = 2,
+ .pVertexAttributeDescriptions = (VkVertexInputAttributeDescription[]) {
+ {
+ /* VUE Header */
+ .location = 0,
+ .binding = 0,
+ .format = VK_FORMAT_R32G32B32A32_UINT,
+ .offset = offsetof(struct depthstencil_clear_vattrs, vue_header),
+ },
+ {
+ /* Position */
+ .location = 1,
+ .binding = 0,
+ .format = VK_FORMAT_R32G32_SFLOAT,
+ .offset = offsetof(struct depthstencil_clear_vattrs, position),
+ },
+ },
+ };
+
+ const VkPipelineDepthStencilStateCreateInfo ds_state = {
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO,
+ .depthTestEnable = (aspects & VK_IMAGE_ASPECT_DEPTH_BIT),
+ .depthCompareOp = VK_COMPARE_OP_ALWAYS,
+ .depthWriteEnable = (aspects & VK_IMAGE_ASPECT_DEPTH_BIT),
+ .depthBoundsTestEnable = false,
+ .stencilTestEnable = (aspects & VK_IMAGE_ASPECT_STENCIL_BIT),
+ .front = {
+ .passOp = VK_STENCIL_OP_REPLACE,
+ .compareOp = VK_COMPARE_OP_ALWAYS,
+ .writeMask = UINT32_MAX,
+ .reference = 0, /* dynamic */
+ },
+ .back = { 0 /* dont care */ },
+ };
+
+ const VkPipelineColorBlendStateCreateInfo cb_state = {
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO,
+ .logicOpEnable = false,
+ .attachmentCount = 0,
+ .pAttachments = NULL,
+ };
+
+ return create_pipeline(device, vs_nir, fs_nir, &vi_state, &ds_state,
+ &cb_state, &device->meta_state.alloc,
+ /*use_repclear*/ true, pipeline);
+}
+
+static void
+emit_depthstencil_clear(struct anv_cmd_buffer *cmd_buffer,
+ const VkClearAttachment *clear_att,
+ const VkClearRect *clear_rect)
+{
+ struct anv_device *device = cmd_buffer->device;
+ const struct anv_subpass *subpass = cmd_buffer->state.subpass;
+ const struct anv_framebuffer *fb = cmd_buffer->state.framebuffer;
+ uint32_t attachment = subpass->depth_stencil_attachment;
+ VkClearDepthStencilValue clear_value = clear_att->clearValue.depthStencil;
+ VkImageAspectFlags aspects = clear_att->aspectMask;
+
+ VkCommandBuffer cmd_buffer_h = anv_cmd_buffer_to_handle(cmd_buffer);
+
+ assert(aspects == VK_IMAGE_ASPECT_DEPTH_BIT ||
+ aspects == VK_IMAGE_ASPECT_STENCIL_BIT ||
+ aspects == (VK_IMAGE_ASPECT_DEPTH_BIT |
+ VK_IMAGE_ASPECT_STENCIL_BIT));
+ assert(attachment != VK_ATTACHMENT_UNUSED);
+
+ const struct depthstencil_clear_vattrs vertex_data[3] = {
+ {
+ .vue_header = { 0 },
+ .position = {
+ clear_rect->rect.offset.x,
+ clear_rect->rect.offset.y,
+ },
+ },
+ {
+ .vue_header = { 0 },
+ .position = {
+ clear_rect->rect.offset.x + clear_rect->rect.extent.width,
+ clear_rect->rect.offset.y,
+ },
+ },
+ {
+ .vue_header = { 0 },
+ .position = {
+ clear_rect->rect.offset.x + clear_rect->rect.extent.width,
+ clear_rect->rect.offset.y + clear_rect->rect.extent.height,
+ },
+ },
+ };
+
+ struct anv_state state =
+ anv_cmd_buffer_emit_dynamic(cmd_buffer, vertex_data, sizeof(vertex_data), 16);
+
+ struct anv_buffer vertex_buffer = {
+ .device = device,
+ .size = sizeof(vertex_data),
+ .bo = &device->dynamic_state_block_pool.bo,
+ .offset = state.offset,
+ };
+
+ ANV_CALL(CmdSetViewport)(cmd_buffer_h, 0, 1,
+ (VkViewport[]) {
+ {
+ .x = 0,
+ .y = 0,
+ .width = fb->width,
+ .height = fb->height,
+
+ /* Ignored when clearing only stencil. */
+ .minDepth = clear_value.depth,
+ .maxDepth = clear_value.depth,
+ },
+ });
+
+ ANV_CALL(CmdSetScissor)(cmd_buffer_h, 0, 1,
+ (VkRect2D[]) {
+ {
+ .offset = { 0, 0 },
+ .extent = { fb->width, fb->height },
+ }
+ });
+
+ if (aspects & VK_IMAGE_ASPECT_STENCIL_BIT) {
+ ANV_CALL(CmdSetStencilReference)(cmd_buffer_h, VK_STENCIL_FACE_FRONT_BIT,
+ clear_value.stencil);
+ }
+
+ ANV_CALL(CmdBindVertexBuffers)(cmd_buffer_h, 0, 1,
+ (VkBuffer[]) { anv_buffer_to_handle(&vertex_buffer) },
+ (VkDeviceSize[]) { 0 });
+
+ struct anv_pipeline *pipeline;
+ switch (aspects) {
+ case VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT:
+ pipeline = device->meta_state.clear.depthstencil_pipeline;
+ break;
+ case VK_IMAGE_ASPECT_DEPTH_BIT:
+ pipeline = device->meta_state.clear.depth_only_pipeline;
+ break;
+ case VK_IMAGE_ASPECT_STENCIL_BIT:
+ pipeline = device->meta_state.clear.stencil_only_pipeline;
+ break;
+ default:
+ unreachable("expected depth or stencil aspect");
+ }
+
+ if (cmd_buffer->state.pipeline != pipeline) {
+ ANV_CALL(CmdBindPipeline)(cmd_buffer_h, VK_PIPELINE_BIND_POINT_GRAPHICS,
+ anv_pipeline_to_handle(pipeline));
+ }
+
+ ANV_CALL(CmdDraw)(cmd_buffer_h, 3, 1, 0, 0);
+}
+
+static VkResult
+init_depthstencil_pipelines(struct anv_device *device)
+{
+ VkResult result;
+ struct anv_meta_state *state = &device->meta_state;
+
+ result =
+ create_depthstencil_pipeline(device, VK_IMAGE_ASPECT_DEPTH_BIT,
+ &state->clear.depth_only_pipeline);
+ if (result != VK_SUCCESS)
+ goto fail;
+
+ result =
+ create_depthstencil_pipeline(device, VK_IMAGE_ASPECT_STENCIL_BIT,
+ &state->clear.stencil_only_pipeline);
+ if (result != VK_SUCCESS)
+ goto fail_depth_only;
+
+ result =
+ create_depthstencil_pipeline(device,
+ VK_IMAGE_ASPECT_DEPTH_BIT |
+ VK_IMAGE_ASPECT_STENCIL_BIT,
+ &state->clear.depthstencil_pipeline);
+ if (result != VK_SUCCESS)
+ goto fail_stencil_only;
+
+ return result;
+
+ fail_stencil_only:
+ anv_DestroyPipeline(anv_device_to_handle(device),
+ anv_pipeline_to_handle(state->clear.stencil_only_pipeline),
+ NULL);
+ fail_depth_only:
+ anv_DestroyPipeline(anv_device_to_handle(device),
+ anv_pipeline_to_handle(state->clear.depth_only_pipeline),
+ NULL);
+ fail:
+ return result;
+}
+
+VkResult
+anv_device_init_meta_clear_state(struct anv_device *device)
+{
+ VkResult result;
+
+ result = init_color_pipelines(device);
+ if (result != VK_SUCCESS)
+ return result;
+
+ result = init_depthstencil_pipelines(device);
+ if (result != VK_SUCCESS) {
+ free_color_pipelines(device);
+ return result;
+ }
+
+ return VK_SUCCESS;
+}
+
+void
+anv_device_finish_meta_clear_state(struct anv_device *device)
+{
+ VkDevice device_h = anv_device_to_handle(device);
+
+ free_color_pipelines(device);
+
+ ANV_CALL(DestroyPipeline)(device_h,
+ anv_pipeline_to_handle(device->meta_state.clear.depth_only_pipeline),
+ NULL);
+ ANV_CALL(DestroyPipeline)(device_h,
+ anv_pipeline_to_handle(device->meta_state.clear.stencil_only_pipeline),
+ NULL);
+ ANV_CALL(DestroyPipeline)(device_h,
+ anv_pipeline_to_handle(device->meta_state.clear.depthstencil_pipeline),
+ NULL);
+}
+
+/**
+ * The parameters mean that same as those in vkCmdClearAttachments.
+ */
+static void
+emit_clear(struct anv_cmd_buffer *cmd_buffer,
+ const VkClearAttachment *clear_att,
+ const VkClearRect *clear_rect)
+{
+ if (clear_att->aspectMask & VK_IMAGE_ASPECT_COLOR_BIT) {
+ emit_color_clear(cmd_buffer, clear_att, clear_rect);
+ } else {
+ assert(clear_att->aspectMask & (VK_IMAGE_ASPECT_DEPTH_BIT |
+ VK_IMAGE_ASPECT_STENCIL_BIT));
+ emit_depthstencil_clear(cmd_buffer, clear_att, clear_rect);
+ }
+}
+
+static bool
+subpass_needs_clear(const struct anv_cmd_buffer *cmd_buffer)
+{
+ const struct anv_cmd_state *cmd_state = &cmd_buffer->state;
+ uint32_t ds = cmd_state->subpass->depth_stencil_attachment;
+
+ for (uint32_t i = 0; i < cmd_state->subpass->color_count; ++i) {
+ uint32_t a = cmd_state->subpass->color_attachments[i];
+ if (cmd_state->attachments[a].pending_clear_aspects) {
+ return true;
+ }
+ }
+
+ if (ds != VK_ATTACHMENT_UNUSED &&
+ cmd_state->attachments[ds].pending_clear_aspects) {
+ return true;
+ }
+
+ return false;
+}
+
+/**
+ * Emit any pending attachment clears for the current subpass.
+ *
+ * @see anv_attachment_state::pending_clear_aspects
+ */
+void
+anv_cmd_buffer_clear_subpass(struct anv_cmd_buffer *cmd_buffer)
+{
+ struct anv_cmd_state *cmd_state = &cmd_buffer->state;
+ struct anv_framebuffer *fb = cmd_buffer->state.framebuffer;
+ struct anv_meta_saved_state saved_state;
+
+ if (!subpass_needs_clear(cmd_buffer))
+ return;
+
+ meta_clear_begin(&saved_state, cmd_buffer);
+
+ if (cmd_state->framebuffer->layers > 1)
+ anv_finishme("clearing multi-layer framebuffer");
+
+ VkClearRect clear_rect = {
+ .rect = {
+ .offset = { 0, 0 },
+ .extent = { fb->width, fb->height },
+ },
+ .baseArrayLayer = 0,
+ .layerCount = 1, /* FINISHME: clear multi-layer framebuffer */
+ };
+
+ for (uint32_t i = 0; i < cmd_state->subpass->color_count; ++i) {
+ uint32_t a = cmd_state->subpass->color_attachments[i];
+
+ if (!cmd_state->attachments[a].pending_clear_aspects)
+ continue;
+
+ assert(cmd_state->attachments[a].pending_clear_aspects ==
+ VK_IMAGE_ASPECT_COLOR_BIT);
+
+ VkClearAttachment clear_att = {
+ .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
+ .colorAttachment = i, /* Use attachment index relative to subpass */
+ .clearValue = cmd_state->attachments[a].clear_value,
+ };
+
+ emit_clear(cmd_buffer, &clear_att, &clear_rect);
+ cmd_state->attachments[a].pending_clear_aspects = 0;
+ }
+
+ uint32_t ds = cmd_state->subpass->depth_stencil_attachment;
+
+ if (ds != VK_ATTACHMENT_UNUSED &&
+ cmd_state->attachments[ds].pending_clear_aspects) {
+
+ VkClearAttachment clear_att = {
+ .aspectMask = cmd_state->attachments[ds].pending_clear_aspects,
+ .clearValue = cmd_state->attachments[ds].clear_value,
+ };
+
+ emit_clear(cmd_buffer, &clear_att, &clear_rect);
+ cmd_state->attachments[ds].pending_clear_aspects = 0;
+ }
+
+ meta_clear_end(&saved_state, cmd_buffer);
+}
+
+static void
+anv_cmd_clear_image(struct anv_cmd_buffer *cmd_buffer,
+ struct anv_image *image,
+ VkImageLayout image_layout,
+ const VkClearValue *clear_value,
+ uint32_t range_count,
+ const VkImageSubresourceRange *ranges)
+{
+ VkDevice device_h = anv_device_to_handle(cmd_buffer->device);
+
+ for (uint32_t r = 0; r < range_count; r++) {
+ const VkImageSubresourceRange *range = &ranges[r];
+
+ for (uint32_t l = 0; l < range->levelCount; ++l) {
+ for (uint32_t s = 0; s < range->layerCount; ++s) {
+ struct anv_image_view iview;
+ anv_image_view_init(&iview, cmd_buffer->device,
+ &(VkImageViewCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
+ .image = anv_image_to_handle(image),
+ .viewType = anv_meta_get_view_type(image),
+ .format = image->vk_format,
+ .subresourceRange = {
+ .aspectMask = range->aspectMask,
+ .baseMipLevel = range->baseMipLevel + l,
+ .levelCount = 1,
+ .baseArrayLayer = range->baseArrayLayer + s,
+ .layerCount = 1
+ },
+ },
+ cmd_buffer);
+
+ VkFramebuffer fb;
+ anv_CreateFramebuffer(device_h,
+ &(VkFramebufferCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO,
+ .attachmentCount = 1,
+ .pAttachments = (VkImageView[]) {
+ anv_image_view_to_handle(&iview),
+ },
+ .width = iview.extent.width,
+ .height = iview.extent.height,
+ .layers = 1
+ },
+ &cmd_buffer->pool->alloc,
+ &fb);
+
+ VkAttachmentDescription att_desc = {
+ .format = iview.vk_format,
+ .loadOp = VK_ATTACHMENT_LOAD_OP_LOAD,
+ .storeOp = VK_ATTACHMENT_STORE_OP_STORE,
+ .stencilLoadOp = VK_ATTACHMENT_LOAD_OP_LOAD,
+ .stencilStoreOp = VK_ATTACHMENT_STORE_OP_STORE,
+ .initialLayout = image_layout,
+ .finalLayout = image_layout,
+ };
+
+ VkSubpassDescription subpass_desc = {
+ .pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS,
+ .inputAttachmentCount = 0,
+ .colorAttachmentCount = 0,
+ .pColorAttachments = NULL,
+ .pResolveAttachments = NULL,
+ .pDepthStencilAttachment = NULL,
+ .preserveAttachmentCount = 0,
+ .pPreserveAttachments = NULL,
+ };
+
+ const VkAttachmentReference att_ref = {
+ .attachment = 0,
+ .layout = image_layout,
+ };
+
+ if (range->aspectMask & VK_IMAGE_ASPECT_COLOR_BIT) {
+ subpass_desc.colorAttachmentCount = 1;
+ subpass_desc.pColorAttachments = &att_ref;
+ } else {
+ subpass_desc.pDepthStencilAttachment = &att_ref;
+ }
+
+ VkRenderPass pass;
+ anv_CreateRenderPass(device_h,
+ &(VkRenderPassCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO,
+ .attachmentCount = 1,
+ .pAttachments = &att_desc,
+ .subpassCount = 1,
+ .pSubpasses = &subpass_desc,
+ },
+ &cmd_buffer->pool->alloc,
+ &pass);
+
+ ANV_CALL(CmdBeginRenderPass)(anv_cmd_buffer_to_handle(cmd_buffer),
+ &(VkRenderPassBeginInfo) {
+ .sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO,
+ .renderArea = {
+ .offset = { 0, 0, },
+ .extent = {
+ .width = iview.extent.width,
+ .height = iview.extent.height,
+ },
+ },
+ .renderPass = pass,
+ .framebuffer = fb,
+ .clearValueCount = 0,
+ .pClearValues = NULL,
+ },
+ VK_SUBPASS_CONTENTS_INLINE);
+
+ VkClearAttachment clear_att = {
+ .aspectMask = range->aspectMask,
+ .colorAttachment = 0,
+ .clearValue = *clear_value,
+ };
+
+ VkClearRect clear_rect = {
+ .rect = {
+ .offset = { 0, 0 },
+ .extent = { iview.extent.width, iview.extent.height },
+ },
+ .baseArrayLayer = range->baseArrayLayer,
+ .layerCount = 1, /* FINISHME: clear multi-layer framebuffer */
+ };
+
+ emit_clear(cmd_buffer, &clear_att, &clear_rect);
+
+ ANV_CALL(CmdEndRenderPass)(anv_cmd_buffer_to_handle(cmd_buffer));
+ ANV_CALL(DestroyRenderPass)(device_h, pass,
+ &cmd_buffer->pool->alloc);
+ ANV_CALL(DestroyFramebuffer)(device_h, fb,
+ &cmd_buffer->pool->alloc);
+ }
+ }
+ }
+}
+
+void anv_CmdClearColorImage(
+ VkCommandBuffer commandBuffer,
+ VkImage image_h,
+ VkImageLayout imageLayout,
+ const VkClearColorValue* pColor,
+ uint32_t rangeCount,
+ const VkImageSubresourceRange* pRanges)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_image, image, image_h);
+ struct anv_meta_saved_state saved_state;
+
+ meta_clear_begin(&saved_state, cmd_buffer);
+
+ anv_cmd_clear_image(cmd_buffer, image, imageLayout,
+ (const VkClearValue *) pColor,
+ rangeCount, pRanges);
+
+ meta_clear_end(&saved_state, cmd_buffer);
+}
+
+void anv_CmdClearDepthStencilImage(
+ VkCommandBuffer commandBuffer,
+ VkImage image_h,
+ VkImageLayout imageLayout,
+ const VkClearDepthStencilValue* pDepthStencil,
+ uint32_t rangeCount,
+ const VkImageSubresourceRange* pRanges)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_image, image, image_h);
+ struct anv_meta_saved_state saved_state;
+
+ meta_clear_begin(&saved_state, cmd_buffer);
+
+ anv_cmd_clear_image(cmd_buffer, image, imageLayout,
+ (const VkClearValue *) pDepthStencil,
+ rangeCount, pRanges);
+
+ meta_clear_end(&saved_state, cmd_buffer);
+}
+
+void anv_CmdClearAttachments(
+ VkCommandBuffer commandBuffer,
+ uint32_t attachmentCount,
+ const VkClearAttachment* pAttachments,
+ uint32_t rectCount,
+ const VkClearRect* pRects)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ struct anv_meta_saved_state saved_state;
+
+ meta_clear_begin(&saved_state, cmd_buffer);
+
+ /* FINISHME: We can do better than this dumb loop. It thrashes too much
+ * state.
+ */
+ for (uint32_t a = 0; a < attachmentCount; ++a) {
+ for (uint32_t r = 0; r < rectCount; ++r) {
+ emit_clear(cmd_buffer, &pAttachments[a], &pRects[r]);
+ }
+ }
+
+ meta_clear_end(&saved_state, cmd_buffer);
+}
+
+static void
+do_buffer_fill(struct anv_cmd_buffer *cmd_buffer,
+ struct anv_bo *dest, uint64_t dest_offset,
+ int width, int height, VkFormat fill_format, uint32_t data)
+{
+ VkDevice vk_device = anv_device_to_handle(cmd_buffer->device);
+
+ VkImageCreateInfo image_info = {
+ .sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO,
+ .imageType = VK_IMAGE_TYPE_2D,
+ .format = fill_format,
+ .extent = {
+ .width = width,
+ .height = height,
+ .depth = 1,
+ },
+ .mipLevels = 1,
+ .arrayLayers = 1,
+ .samples = 1,
+ .tiling = VK_IMAGE_TILING_LINEAR,
+ .usage = VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT,
+ .flags = 0,
+ };
+
+ VkImage dest_image;
+ image_info.usage = VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT;
+ anv_CreateImage(vk_device, &image_info,
+ &cmd_buffer->pool->alloc, &dest_image);
+
+ /* We could use a vk call to bind memory, but that would require
+ * creating a dummy memory object etc. so there's really no point.
+ */
+ anv_image_from_handle(dest_image)->bo = dest;
+ anv_image_from_handle(dest_image)->offset = dest_offset;
+
+ const VkClearValue clear_value = {
+ .color = {
+ .uint32 = { data, data, data, data }
+ }
+ };
+
+ const VkImageSubresourceRange range = {
+ .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
+ .baseMipLevel = 0,
+ .levelCount = 1,
+ .baseArrayLayer = 0,
+ .layerCount = 1,
+ };
+
+ anv_cmd_clear_image(cmd_buffer, anv_image_from_handle(dest_image),
+ VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL,
+ &clear_value, 1, &range);
+}
+
+void anv_CmdFillBuffer(
+ VkCommandBuffer commandBuffer,
+ VkBuffer dstBuffer,
+ VkDeviceSize dstOffset,
+ VkDeviceSize fillSize,
+ uint32_t data)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_buffer, dst_buffer, dstBuffer);
+ struct anv_meta_saved_state saved_state;
+
+ meta_clear_begin(&saved_state, cmd_buffer);
+
+ VkFormat format;
+ int bs;
+ if ((fillSize & 15) == 0 && (dstOffset & 15) == 0) {
+ format = VK_FORMAT_R32G32B32A32_UINT;
+ bs = 16;
+ } else if ((fillSize & 7) == 0 && (dstOffset & 15) == 0) {
+ format = VK_FORMAT_R32G32_UINT;
+ bs = 8;
+ } else {
+ assert((fillSize & 3) == 0 && (dstOffset & 3) == 0);
+ format = VK_FORMAT_R32_UINT;
+ bs = 4;
+ }
+
+ /* This is maximum possible width/height our HW can handle */
+ const uint64_t max_surface_dim = 1 << 14;
+
+ /* First, we make a bunch of max-sized copies */
+ const uint64_t max_fill_size = max_surface_dim * max_surface_dim * bs;
+ while (fillSize > max_fill_size) {
+ do_buffer_fill(cmd_buffer, dst_buffer->bo,
+ dst_buffer->offset + dstOffset,
+ max_surface_dim, max_surface_dim, format, data);
+ fillSize -= max_fill_size;
+ dstOffset += max_fill_size;
+ }
+
+ uint64_t height = fillSize / (max_surface_dim * bs);
+ assert(height < max_surface_dim);
+ if (height != 0) {
+ const uint64_t rect_fill_size = height * max_surface_dim * bs;
+ do_buffer_fill(cmd_buffer, dst_buffer->bo,
+ dst_buffer->offset + dstOffset,
+ max_surface_dim, height, format, data);
+ fillSize -= rect_fill_size;
+ dstOffset += rect_fill_size;
+ }
+
+ if (fillSize != 0) {
+ do_buffer_fill(cmd_buffer, dst_buffer->bo,
+ dst_buffer->offset + dstOffset,
+ fillSize / bs, 1, format, data);
+ }
+
+ meta_clear_end(&saved_state, cmd_buffer);
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#pragma once
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+struct anv_device;
+
+VkResult anv_device_init_meta_clear_state(struct anv_device *device);
+void anv_device_finish_meta_clear_state(struct anv_device *device);
+
+#ifdef __cplusplus
+}
+#endif
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#pragma once
+
+#include "glsl/nir/nir.h"
+#include "anv_private.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+void anv_nir_lower_push_constants(nir_shader *shader, bool is_scalar);
+
+void anv_nir_apply_dynamic_offsets(struct anv_pipeline *pipeline,
+ nir_shader *shader,
+ struct brw_stage_prog_data *prog_data);
+bool anv_nir_apply_pipeline_layout(nir_shader *shader,
+ struct brw_stage_prog_data *prog_data,
+ const struct anv_pipeline_layout *layout);
+
+#ifdef __cplusplus
+}
+#endif
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include "anv_nir.h"
+#include "glsl/nir/nir_builder.h"
+
+struct apply_dynamic_offsets_state {
+ nir_shader *shader;
+ nir_builder builder;
+
+ struct anv_pipeline_layout *layout;
+
+ uint32_t indices_start;
+};
+
+static bool
+apply_dynamic_offsets_block(nir_block *block, void *void_state)
+{
+ struct apply_dynamic_offsets_state *state = void_state;
+ struct anv_descriptor_set_layout *set_layout;
+
+ nir_builder *b = &state->builder;
+
+ nir_foreach_instr_safe(block, instr) {
+ if (instr->type != nir_instr_type_intrinsic)
+ continue;
+
+ nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);
+
+ unsigned block_idx_src;
+ switch (intrin->intrinsic) {
+ case nir_intrinsic_load_ubo:
+ case nir_intrinsic_load_ssbo:
+ block_idx_src = 0;
+ break;
+ case nir_intrinsic_store_ssbo:
+ block_idx_src = 1;
+ break;
+ default:
+ continue; /* the loop */
+ }
+
+ nir_instr *res_instr = intrin->src[block_idx_src].ssa->parent_instr;
+ assert(res_instr->type == nir_instr_type_intrinsic);
+ nir_intrinsic_instr *res_intrin = nir_instr_as_intrinsic(res_instr);
+ assert(res_intrin->intrinsic == nir_intrinsic_vulkan_resource_index);
+
+ unsigned set = res_intrin->const_index[0];
+ unsigned binding = res_intrin->const_index[1];
+
+ set_layout = state->layout->set[set].layout;
+ if (set_layout->binding[binding].dynamic_offset_index < 0)
+ continue;
+
+ b->cursor = nir_before_instr(&intrin->instr);
+
+ /* First, we need to generate the uniform load for the buffer offset */
+ uint32_t index = state->layout->set[set].dynamic_offset_start +
+ set_layout->binding[binding].dynamic_offset_index;
+
+ nir_intrinsic_instr *offset_load =
+ nir_intrinsic_instr_create(state->shader, nir_intrinsic_load_uniform);
+ offset_load->num_components = 2;
+ offset_load->const_index[0] = state->indices_start + index * 8;
+ offset_load->src[0] = nir_src_for_ssa(nir_imul(b, res_intrin->src[0].ssa,
+ nir_imm_int(b, 8)));
+
+ nir_ssa_dest_init(&offset_load->instr, &offset_load->dest, 2, NULL);
+ nir_builder_instr_insert(b, &offset_load->instr);
+
+ nir_src *offset_src = nir_get_io_offset_src(intrin);
+ nir_ssa_def *new_offset = nir_iadd(b, offset_src->ssa,
+ &offset_load->dest.ssa);
+
+ /* In order to avoid out-of-bounds access, we predicate */
+ nir_ssa_def *pred = nir_uge(b, nir_channel(b, &offset_load->dest.ssa, 1),
+ offset_src->ssa);
+ nir_if *if_stmt = nir_if_create(b->shader);
+ if_stmt->condition = nir_src_for_ssa(pred);
+ nir_cf_node_insert(b->cursor, &if_stmt->cf_node);
+
+ nir_instr_remove(&intrin->instr);
+ *offset_src = nir_src_for_ssa(new_offset);
+ nir_instr_insert_after_cf_list(&if_stmt->then_list, &intrin->instr);
+
+ if (intrin->intrinsic != nir_intrinsic_store_ssbo) {
+ /* It's a load, we need a phi node */
+ nir_phi_instr *phi = nir_phi_instr_create(b->shader);
+ nir_ssa_dest_init(&phi->instr, &phi->dest,
+ intrin->num_components, NULL);
+
+ nir_phi_src *src1 = ralloc(phi, nir_phi_src);
+ struct exec_node *tnode = exec_list_get_tail(&if_stmt->then_list);
+ src1->pred = exec_node_data(nir_block, tnode, cf_node.node);
+ src1->src = nir_src_for_ssa(&intrin->dest.ssa);
+ exec_list_push_tail(&phi->srcs, &src1->node);
+
+ b->cursor = nir_after_cf_list(&if_stmt->else_list);
+ nir_ssa_def *zero = nir_build_imm(b, intrin->num_components,
+ (nir_const_value) { .u = { 0, 0, 0, 0 } });
+
+ nir_phi_src *src2 = ralloc(phi, nir_phi_src);
+ struct exec_node *enode = exec_list_get_tail(&if_stmt->else_list);
+ src2->pred = exec_node_data(nir_block, enode, cf_node.node);
+ src2->src = nir_src_for_ssa(zero);
+ exec_list_push_tail(&phi->srcs, &src2->node);
+
+ assert(intrin->dest.is_ssa);
+ nir_ssa_def_rewrite_uses(&intrin->dest.ssa,
+ nir_src_for_ssa(&phi->dest.ssa));
+
+ nir_instr_insert_after_cf(&if_stmt->cf_node, &phi->instr);
+ }
+ }
+
+ return true;
+}
+
+void
+anv_nir_apply_dynamic_offsets(struct anv_pipeline *pipeline,
+ nir_shader *shader,
+ struct brw_stage_prog_data *prog_data)
+{
+ struct apply_dynamic_offsets_state state = {
+ .shader = shader,
+ .layout = pipeline->layout,
+ .indices_start = shader->num_uniforms,
+ };
+
+ if (!state.layout || !state.layout->stage[shader->stage].has_dynamic_offsets)
+ return;
+
+ nir_foreach_function(shader, function) {
+ if (function->impl) {
+ nir_builder_init(&state.builder, function->impl);
+ nir_foreach_block(function->impl, apply_dynamic_offsets_block, &state);
+ nir_metadata_preserve(function->impl, nir_metadata_block_index |
+ nir_metadata_dominance);
+ }
+ }
+
+ struct anv_push_constants *null_data = NULL;
+ for (unsigned i = 0; i < MAX_DYNAMIC_BUFFERS; i++) {
+ prog_data->param[i * 2 + shader->num_uniforms] =
+ (const union gl_constant_value *)&null_data->dynamic[i].offset;
+ prog_data->param[i * 2 + 1 + shader->num_uniforms] =
+ (const union gl_constant_value *)&null_data->dynamic[i].range;
+ }
+
+ shader->num_uniforms += MAX_DYNAMIC_BUFFERS * 8;
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include "anv_nir.h"
+#include "program/prog_parameter.h"
+#include "glsl/nir/nir_builder.h"
+
+struct apply_pipeline_layout_state {
+ nir_shader *shader;
+ nir_builder builder;
+
+ const struct anv_pipeline_layout *layout;
+
+ bool progress;
+};
+
+static uint32_t
+get_surface_index(unsigned set, unsigned binding,
+ struct apply_pipeline_layout_state *state)
+{
+ assert(set < state->layout->num_sets);
+ struct anv_descriptor_set_layout *set_layout =
+ state->layout->set[set].layout;
+
+ gl_shader_stage stage = state->shader->stage;
+
+ assert(binding < set_layout->binding_count);
+
+ assert(set_layout->binding[binding].stage[stage].surface_index >= 0);
+
+ uint32_t surface_index =
+ state->layout->set[set].stage[stage].surface_start +
+ set_layout->binding[binding].stage[stage].surface_index;
+
+ assert(surface_index < state->layout->stage[stage].surface_count);
+
+ return surface_index;
+}
+
+static uint32_t
+get_sampler_index(unsigned set, unsigned binding, nir_texop tex_op,
+ struct apply_pipeline_layout_state *state)
+{
+ assert(set < state->layout->num_sets);
+ struct anv_descriptor_set_layout *set_layout =
+ state->layout->set[set].layout;
+
+ assert(binding < set_layout->binding_count);
+
+ gl_shader_stage stage = state->shader->stage;
+
+ if (set_layout->binding[binding].stage[stage].sampler_index < 0) {
+ assert(tex_op == nir_texop_txf);
+ return 0;
+ }
+
+ uint32_t sampler_index =
+ state->layout->set[set].stage[stage].sampler_start +
+ set_layout->binding[binding].stage[stage].sampler_index;
+
+ assert(sampler_index < state->layout->stage[stage].sampler_count);
+
+ return sampler_index;
+}
+
+static uint32_t
+get_image_index(unsigned set, unsigned binding,
+ struct apply_pipeline_layout_state *state)
+{
+ assert(set < state->layout->num_sets);
+ struct anv_descriptor_set_layout *set_layout =
+ state->layout->set[set].layout;
+
+ assert(binding < set_layout->binding_count);
+
+ gl_shader_stage stage = state->shader->stage;
+
+ assert(set_layout->binding[binding].stage[stage].image_index >= 0);
+
+ uint32_t image_index =
+ state->layout->set[set].stage[stage].image_start +
+ set_layout->binding[binding].stage[stage].image_index;
+
+ assert(image_index < state->layout->stage[stage].image_count);
+
+ return image_index;
+}
+
+static void
+lower_res_index_intrinsic(nir_intrinsic_instr *intrin,
+ struct apply_pipeline_layout_state *state)
+{
+ nir_builder *b = &state->builder;
+
+ b->cursor = nir_before_instr(&intrin->instr);
+
+ uint32_t set = intrin->const_index[0];
+ uint32_t binding = intrin->const_index[1];
+
+ uint32_t surface_index = get_surface_index(set, binding, state);
+
+ nir_const_value *const_block_idx =
+ nir_src_as_const_value(intrin->src[0]);
+
+ nir_ssa_def *block_index;
+ if (const_block_idx) {
+ block_index = nir_imm_int(b, surface_index + const_block_idx->u[0]);
+ } else {
+ block_index = nir_iadd(b, nir_imm_int(b, surface_index),
+ nir_ssa_for_src(b, intrin->src[0], 1));
+ }
+
+ assert(intrin->dest.is_ssa);
+ nir_ssa_def_rewrite_uses(&intrin->dest.ssa, nir_src_for_ssa(block_index));
+ nir_instr_remove(&intrin->instr);
+}
+
+static void
+lower_tex_deref(nir_tex_instr *tex, nir_deref_var *deref,
+ unsigned *const_index, nir_tex_src_type src_type,
+ struct apply_pipeline_layout_state *state)
+{
+ if (deref->deref.child) {
+ assert(deref->deref.child->deref_type == nir_deref_type_array);
+ nir_deref_array *deref_array = nir_deref_as_array(deref->deref.child);
+
+ *const_index += deref_array->base_offset;
+
+ if (deref_array->deref_array_type == nir_deref_array_type_indirect) {
+ nir_tex_src *new_srcs = rzalloc_array(tex, nir_tex_src,
+ tex->num_srcs + 1);
+
+ for (unsigned i = 0; i < tex->num_srcs; i++) {
+ new_srcs[i].src_type = tex->src[i].src_type;
+ nir_instr_move_src(&tex->instr, &new_srcs[i].src, &tex->src[i].src);
+ }
+
+ ralloc_free(tex->src);
+ tex->src = new_srcs;
+
+ /* Now we can go ahead and move the source over to being a
+ * first-class texture source.
+ */
+ tex->src[tex->num_srcs].src_type = src_type;
+ tex->num_srcs++;
+ assert(deref_array->indirect.is_ssa);
+ nir_instr_rewrite_src(&tex->instr, &tex->src[tex->num_srcs - 1].src,
+ deref_array->indirect);
+ }
+ }
+}
+
+static void
+cleanup_tex_deref(nir_tex_instr *tex, nir_deref_var *deref)
+{
+ if (deref->deref.child == NULL)
+ return;
+
+ nir_deref_array *deref_array = nir_deref_as_array(deref->deref.child);
+
+ if (deref_array->deref_array_type != nir_deref_array_type_indirect)
+ return;
+
+ nir_instr_rewrite_src(&tex->instr, &deref_array->indirect, NIR_SRC_INIT);
+}
+
+static void
+lower_tex(nir_tex_instr *tex, struct apply_pipeline_layout_state *state)
+{
+ /* No one should have come by and lowered it already */
+ assert(tex->sampler);
+
+ nir_deref_var *tex_deref = tex->texture ? tex->texture : tex->sampler;
+ tex->texture_index =
+ get_surface_index(tex_deref->var->data.descriptor_set,
+ tex_deref->var->data.binding, state);
+ lower_tex_deref(tex, tex_deref, &tex->texture_index,
+ nir_tex_src_texture_offset, state);
+
+ tex->sampler_index =
+ get_sampler_index(tex->sampler->var->data.descriptor_set,
+ tex->sampler->var->data.binding, tex->op, state);
+ lower_tex_deref(tex, tex->sampler, &tex->sampler_index,
+ nir_tex_src_sampler_offset, state);
+
+ /* The backend only ever uses this to mark used surfaces. We don't care
+ * about that little optimization so it just needs to be non-zero.
+ */
+ tex->texture_array_size = 1;
+
+ if (tex->texture)
+ cleanup_tex_deref(tex, tex->texture);
+ cleanup_tex_deref(tex, tex->sampler);
+ tex->texture = NULL;
+ tex->sampler = NULL;
+}
+
+static bool
+apply_pipeline_layout_block(nir_block *block, void *void_state)
+{
+ struct apply_pipeline_layout_state *state = void_state;
+
+ nir_foreach_instr_safe(block, instr) {
+ switch (instr->type) {
+ case nir_instr_type_intrinsic: {
+ nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);
+ if (intrin->intrinsic == nir_intrinsic_vulkan_resource_index) {
+ lower_res_index_intrinsic(intrin, state);
+ state->progress = true;
+ }
+ break;
+ }
+ case nir_instr_type_tex:
+ lower_tex(nir_instr_as_tex(instr), state);
+ /* All texture instructions need lowering */
+ state->progress = true;
+ break;
+ default:
+ continue;
+ }
+ }
+
+ return true;
+}
+
+static void
+setup_vec4_uniform_value(const union gl_constant_value **params,
+ const union gl_constant_value *values,
+ unsigned n)
+{
+ static const gl_constant_value zero = { 0 };
+
+ for (unsigned i = 0; i < n; ++i)
+ params[i] = &values[i];
+
+ for (unsigned i = n; i < 4; ++i)
+ params[i] = &zero;
+}
+
+bool
+anv_nir_apply_pipeline_layout(nir_shader *shader,
+ struct brw_stage_prog_data *prog_data,
+ const struct anv_pipeline_layout *layout)
+{
+ struct apply_pipeline_layout_state state = {
+ .shader = shader,
+ .layout = layout,
+ };
+
+ nir_foreach_function(shader, function) {
+ if (function->impl) {
+ nir_builder_init(&state.builder, function->impl);
+ nir_foreach_block(function->impl, apply_pipeline_layout_block, &state);
+ nir_metadata_preserve(function->impl, nir_metadata_block_index |
+ nir_metadata_dominance);
+ }
+ }
+
+ if (layout->stage[shader->stage].image_count > 0) {
+ nir_foreach_variable(var, &shader->uniforms) {
+ if (glsl_type_is_image(var->type) ||
+ (glsl_type_is_array(var->type) &&
+ glsl_type_is_image(glsl_get_array_element(var->type)))) {
+ /* Images are represented as uniform push constants and the actual
+ * information required for reading/writing to/from the image is
+ * storred in the uniform.
+ */
+ unsigned image_index = get_image_index(var->data.descriptor_set,
+ var->data.binding, &state);
+
+ var->data.driver_location = shader->num_uniforms +
+ image_index * BRW_IMAGE_PARAM_SIZE * 4;
+ }
+ }
+
+ struct anv_push_constants *null_data = NULL;
+ const gl_constant_value **param = prog_data->param + shader->num_uniforms;
+ const struct brw_image_param *image_param = null_data->images;
+ for (uint32_t i = 0; i < layout->stage[shader->stage].image_count; i++) {
+ setup_vec4_uniform_value(param + BRW_IMAGE_PARAM_SURFACE_IDX_OFFSET,
+ (const union gl_constant_value *)&image_param->surface_idx, 1);
+ setup_vec4_uniform_value(param + BRW_IMAGE_PARAM_OFFSET_OFFSET,
+ (const union gl_constant_value *)image_param->offset, 2);
+ setup_vec4_uniform_value(param + BRW_IMAGE_PARAM_SIZE_OFFSET,
+ (const union gl_constant_value *)image_param->size, 3);
+ setup_vec4_uniform_value(param + BRW_IMAGE_PARAM_STRIDE_OFFSET,
+ (const union gl_constant_value *)image_param->stride, 4);
+ setup_vec4_uniform_value(param + BRW_IMAGE_PARAM_TILING_OFFSET,
+ (const union gl_constant_value *)image_param->tiling, 3);
+ setup_vec4_uniform_value(param + BRW_IMAGE_PARAM_SWIZZLING_OFFSET,
+ (const union gl_constant_value *)image_param->swizzling, 2);
+
+ param += BRW_IMAGE_PARAM_SIZE;
+ image_param ++;
+ }
+
+ shader->num_uniforms += layout->stage[shader->stage].image_count *
+ BRW_IMAGE_PARAM_SIZE * 4;
+ }
+
+ return state.progress;
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include "anv_nir.h"
+
+struct lower_push_constants_state {
+ nir_shader *shader;
+ bool is_scalar;
+};
+
+static bool
+lower_push_constants_block(nir_block *block, void *void_state)
+{
+ struct lower_push_constants_state *state = void_state;
+
+ nir_foreach_instr(block, instr) {
+ if (instr->type != nir_instr_type_intrinsic)
+ continue;
+
+ nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);
+
+ /* TODO: Handle indirect push constants */
+ if (intrin->intrinsic != nir_intrinsic_load_push_constant)
+ continue;
+
+ assert(intrin->const_index[0] % 4 == 0);
+ unsigned dword_offset = intrin->const_index[0] / 4;
+
+ /* We just turn them into uniform loads with the appropreate offset */
+ intrin->intrinsic = nir_intrinsic_load_uniform;
+ intrin->const_index[0] = 0;
+ if (state->is_scalar) {
+ intrin->const_index[1] = dword_offset;
+ } else {
+ unsigned shift = dword_offset % 4;
+ /* Can't cross the vec4 boundary */
+ assert(shift + intrin->num_components <= 4);
+
+ /* vec4 shifts are in units of vec4's */
+ intrin->const_index[1] = dword_offset / 4;
+
+ if (shift) {
+ /* If there's a non-zero shift then we need to load a whole vec4
+ * and use a move to swizzle it into place.
+ */
+ assert(intrin->dest.is_ssa);
+ nir_alu_instr *mov = nir_alu_instr_create(state->shader,
+ nir_op_imov);
+ mov->src[0].src = nir_src_for_ssa(&intrin->dest.ssa);
+ for (unsigned i = 0; i < intrin->num_components; i++)
+ mov->src[0].swizzle[i] = i + shift;
+ mov->dest.write_mask = (1 << intrin->num_components) - 1;
+ nir_ssa_dest_init(&mov->instr, &mov->dest.dest,
+ intrin->num_components, NULL);
+
+ nir_ssa_def_rewrite_uses(&intrin->dest.ssa,
+ nir_src_for_ssa(&mov->dest.dest.ssa));
+ nir_instr_insert_after(&intrin->instr, &mov->instr);
+
+ /* Stomp the number of components to 4 */
+ intrin->num_components = 4;
+ intrin->dest.ssa.num_components = 4;
+ }
+ }
+ }
+
+ return true;
+}
+
+void
+anv_nir_lower_push_constants(nir_shader *shader, bool is_scalar)
+{
+ struct lower_push_constants_state state = {
+ .shader = shader,
+ .is_scalar = is_scalar,
+ };
+
+ nir_foreach_function(shader, function) {
+ if (function->impl)
+ nir_foreach_block(function->impl, lower_push_constants_block, &state);
+ }
+
+ assert(shader->num_uniforms % 4 == 0);
+ if (is_scalar)
+ shader->num_uniforms /= 4;
+ else
+ shader->num_uniforms = DIV_ROUND_UP(shader->num_uniforms, 16);
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include "anv_private.h"
+
+VkResult anv_CreateRenderPass(
+ VkDevice _device,
+ const VkRenderPassCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkRenderPass* pRenderPass)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ struct anv_render_pass *pass;
+ size_t size;
+ size_t attachments_offset;
+
+ assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO);
+
+ size = sizeof(*pass);
+ size += pCreateInfo->subpassCount * sizeof(pass->subpasses[0]);
+ attachments_offset = size;
+ size += pCreateInfo->attachmentCount * sizeof(pass->attachments[0]);
+
+ pass = anv_alloc2(&device->alloc, pAllocator, size, 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (pass == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ /* Clear the subpasses along with the parent pass. This required because
+ * each array member of anv_subpass must be a valid pointer if not NULL.
+ */
+ memset(pass, 0, size);
+ pass->attachment_count = pCreateInfo->attachmentCount;
+ pass->subpass_count = pCreateInfo->subpassCount;
+ pass->attachments = (void *) pass + attachments_offset;
+
+ for (uint32_t i = 0; i < pCreateInfo->attachmentCount; i++) {
+ struct anv_render_pass_attachment *att = &pass->attachments[i];
+
+ att->format = anv_format_for_vk_format(pCreateInfo->pAttachments[i].format);
+ att->samples = pCreateInfo->pAttachments[i].samples;
+ att->load_op = pCreateInfo->pAttachments[i].loadOp;
+ att->stencil_load_op = pCreateInfo->pAttachments[i].stencilLoadOp;
+ // att->store_op = pCreateInfo->pAttachments[i].storeOp;
+ // att->stencil_store_op = pCreateInfo->pAttachments[i].stencilStoreOp;
+ }
+
+ uint32_t subpass_attachment_count = 0, *p;
+ for (uint32_t i = 0; i < pCreateInfo->subpassCount; i++) {
+ const VkSubpassDescription *desc = &pCreateInfo->pSubpasses[i];
+
+ subpass_attachment_count +=
+ desc->inputAttachmentCount +
+ desc->colorAttachmentCount +
+ /* Count colorAttachmentCount again for resolve_attachments */
+ desc->colorAttachmentCount;
+ }
+
+ pass->subpass_attachments =
+ anv_alloc2(&device->alloc, pAllocator,
+ subpass_attachment_count * sizeof(uint32_t), 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (pass->subpass_attachments == NULL) {
+ anv_free2(&device->alloc, pAllocator, pass);
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+ }
+
+ p = pass->subpass_attachments;
+ for (uint32_t i = 0; i < pCreateInfo->subpassCount; i++) {
+ const VkSubpassDescription *desc = &pCreateInfo->pSubpasses[i];
+ struct anv_subpass *subpass = &pass->subpasses[i];
+
+ subpass->input_count = desc->inputAttachmentCount;
+ subpass->color_count = desc->colorAttachmentCount;
+
+ if (desc->inputAttachmentCount > 0) {
+ subpass->input_attachments = p;
+ p += desc->inputAttachmentCount;
+
+ for (uint32_t j = 0; j < desc->inputAttachmentCount; j++) {
+ subpass->input_attachments[j]
+ = desc->pInputAttachments[j].attachment;
+ }
+ }
+
+ if (desc->colorAttachmentCount > 0) {
+ subpass->color_attachments = p;
+ p += desc->colorAttachmentCount;
+
+ for (uint32_t j = 0; j < desc->colorAttachmentCount; j++) {
+ subpass->color_attachments[j]
+ = desc->pColorAttachments[j].attachment;
+ }
+ }
+
+ if (desc->pResolveAttachments) {
+ subpass->resolve_attachments = p;
+ p += desc->colorAttachmentCount;
+
+ for (uint32_t j = 0; j < desc->colorAttachmentCount; j++) {
+ subpass->resolve_attachments[j]
+ = desc->pResolveAttachments[j].attachment;
+ }
+ }
+
+ if (desc->pDepthStencilAttachment) {
+ subpass->depth_stencil_attachment =
+ desc->pDepthStencilAttachment->attachment;
+ } else {
+ subpass->depth_stencil_attachment = VK_ATTACHMENT_UNUSED;
+ }
+ }
+
+ *pRenderPass = anv_render_pass_to_handle(pass);
+
+ return VK_SUCCESS;
+}
+
+void anv_DestroyRenderPass(
+ VkDevice _device,
+ VkRenderPass _pass,
+ const VkAllocationCallbacks* pAllocator)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ ANV_FROM_HANDLE(anv_render_pass, pass, _pass);
+
+ anv_free2(&device->alloc, pAllocator, pass->subpass_attachments);
+ anv_free2(&device->alloc, pAllocator, pass);
+}
+
+void anv_GetRenderAreaGranularity(
+ VkDevice device,
+ VkRenderPass renderPass,
+ VkExtent2D* pGranularity)
+{
+ *pGranularity = (VkExtent2D) { 1, 1 };
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <assert.h>
+#include <stdbool.h>
+#include <string.h>
+#include <unistd.h>
+#include <fcntl.h>
+
+#include "anv_private.h"
+#include "brw_nir.h"
+#include "anv_nir.h"
+#include "glsl/nir/spirv/nir_spirv.h"
+
+/* Needed for SWIZZLE macros */
+#include "program/prog_instruction.h"
+
+// Shader functions
+
+VkResult anv_CreateShaderModule(
+ VkDevice _device,
+ const VkShaderModuleCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkShaderModule* pShaderModule)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ struct anv_shader_module *module;
+
+ assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO);
+ assert(pCreateInfo->flags == 0);
+
+ module = anv_alloc2(&device->alloc, pAllocator,
+ sizeof(*module) + pCreateInfo->codeSize, 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (module == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ module->nir = NULL;
+ module->size = pCreateInfo->codeSize;
+ memcpy(module->data, pCreateInfo->pCode, module->size);
+
+ *pShaderModule = anv_shader_module_to_handle(module);
+
+ return VK_SUCCESS;
+}
+
+void anv_DestroyShaderModule(
+ VkDevice _device,
+ VkShaderModule _module,
+ const VkAllocationCallbacks* pAllocator)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ ANV_FROM_HANDLE(anv_shader_module, module, _module);
+
+ anv_free2(&device->alloc, pAllocator, module);
+}
+
+#define SPIR_V_MAGIC_NUMBER 0x07230203
+
+/* Eventually, this will become part of anv_CreateShader. Unfortunately,
+ * we can't do that yet because we don't have the ability to copy nir.
+ */
+static nir_shader *
+anv_shader_compile_to_nir(struct anv_device *device,
+ struct anv_shader_module *module,
+ const char *entrypoint_name,
+ gl_shader_stage stage,
+ const VkSpecializationInfo *spec_info)
+{
+ if (strcmp(entrypoint_name, "main") != 0) {
+ anv_finishme("Multiple shaders per module not really supported");
+ }
+
+ const struct brw_compiler *compiler =
+ device->instance->physicalDevice.compiler;
+ const nir_shader_compiler_options *nir_options =
+ compiler->glsl_compiler_options[stage].NirOptions;
+
+ nir_shader *nir;
+ nir_function *entry_point;
+ if (module->nir) {
+ /* Some things such as our meta clear/blit code will give us a NIR
+ * shader directly. In that case, we just ignore the SPIR-V entirely
+ * and just use the NIR shader */
+ nir = module->nir;
+ nir->options = nir_options;
+ nir_validate_shader(nir);
+
+ assert(exec_list_length(&nir->functions) == 1);
+ struct exec_node *node = exec_list_get_head(&nir->functions);
+ entry_point = exec_node_data(nir_function, node, node);
+ } else {
+ uint32_t *spirv = (uint32_t *) module->data;
+ assert(spirv[0] == SPIR_V_MAGIC_NUMBER);
+ assert(module->size % 4 == 0);
+
+ uint32_t num_spec_entries = 0;
+ struct nir_spirv_specialization *spec_entries = NULL;
+ if (spec_info && spec_info->mapEntryCount > 0) {
+ num_spec_entries = spec_info->mapEntryCount;
+ spec_entries = malloc(num_spec_entries * sizeof(*spec_entries));
+ for (uint32_t i = 0; i < num_spec_entries; i++) {
+ const uint32_t *data =
+ spec_info->pData + spec_info->pMapEntries[i].offset;
+ assert((const void *)(data + 1) <=
+ spec_info->pData + spec_info->dataSize);
+
+ spec_entries[i].id = spec_info->pMapEntries[i].constantID;
+ spec_entries[i].data = *data;
+ }
+ }
+
+ entry_point = spirv_to_nir(spirv, module->size / 4,
+ spec_entries, num_spec_entries,
+ stage, entrypoint_name, nir_options);
+ nir = entry_point->shader;
+ assert(nir->stage == stage);
+ nir_validate_shader(nir);
+
+ free(spec_entries);
+
+ nir_lower_returns(nir);
+ nir_validate_shader(nir);
+
+ nir_inline_functions(nir);
+ nir_validate_shader(nir);
+
+ /* Pick off the single entrypoint that we want */
+ foreach_list_typed_safe(nir_function, func, node, &nir->functions) {
+ if (func != entry_point)
+ exec_node_remove(&func->node);
+ }
+ assert(exec_list_length(&nir->functions) == 1);
+ entry_point->name = ralloc_strdup(entry_point, "main");
+
+ nir_remove_dead_variables(nir, nir_var_shader_in);
+ nir_remove_dead_variables(nir, nir_var_shader_out);
+ nir_remove_dead_variables(nir, nir_var_system_value);
+ nir_validate_shader(nir);
+
+ nir_lower_outputs_to_temporaries(entry_point->shader, entry_point);
+
+ nir_lower_system_values(nir);
+ nir_validate_shader(nir);
+ }
+
+ /* Vulkan uses the separate-shader linking model */
+ nir->info.separate_shader = true;
+
+ nir = brw_preprocess_nir(nir, compiler->scalar_stage[stage]);
+
+ nir_shader_gather_info(nir, entry_point->impl);
+
+ uint32_t indirect_mask = (1 << nir_var_shader_in);
+ if (compiler->glsl_compiler_options[stage].EmitNoIndirectTemp)
+ indirect_mask |= 1 << nir_var_local;
+
+ nir_lower_indirect_derefs(nir, indirect_mask);
+
+ return nir;
+}
+
+void
+anv_pipeline_cache_init(struct anv_pipeline_cache *cache,
+ struct anv_device *device)
+{
+ cache->device = device;
+ anv_state_stream_init(&cache->program_stream,
+ &device->instruction_block_pool);
+ pthread_mutex_init(&cache->mutex, NULL);
+}
+
+void
+anv_pipeline_cache_finish(struct anv_pipeline_cache *cache)
+{
+ anv_state_stream_finish(&cache->program_stream);
+ pthread_mutex_destroy(&cache->mutex);
+}
+
+static uint32_t
+anv_pipeline_cache_upload_kernel(struct anv_pipeline_cache *cache,
+ const void *data, size_t size)
+{
+ pthread_mutex_lock(&cache->mutex);
+
+ struct anv_state state =
+ anv_state_stream_alloc(&cache->program_stream, size, 64);
+
+ pthread_mutex_unlock(&cache->mutex);
+
+ assert(size < cache->program_stream.block_pool->block_size);
+
+ memcpy(state.map, data, size);
+
+ if (!cache->device->info.has_llc)
+ anv_state_clflush(state);
+
+ return state.offset;
+}
+
+VkResult anv_CreatePipelineCache(
+ VkDevice _device,
+ const VkPipelineCacheCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkPipelineCache* pPipelineCache)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ struct anv_pipeline_cache *cache;
+
+ assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_PIPELINE_CACHE_CREATE_INFO);
+ assert(pCreateInfo->flags == 0);
+
+ cache = anv_alloc2(&device->alloc, pAllocator,
+ sizeof(*cache), 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (cache == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ anv_pipeline_cache_init(cache, device);
+
+ *pPipelineCache = anv_pipeline_cache_to_handle(cache);
+
+ return VK_SUCCESS;
+}
+
+void anv_DestroyPipelineCache(
+ VkDevice _device,
+ VkPipelineCache _cache,
+ const VkAllocationCallbacks* pAllocator)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ ANV_FROM_HANDLE(anv_pipeline_cache, cache, _cache);
+
+ anv_pipeline_cache_finish(cache);
+
+ anv_free2(&device->alloc, pAllocator, cache);
+}
+
+VkResult anv_GetPipelineCacheData(
+ VkDevice device,
+ VkPipelineCache pipelineCache,
+ size_t* pDataSize,
+ void* pData)
+{
+ *pDataSize = 0;
+
+ return VK_SUCCESS;
+}
+
+VkResult anv_MergePipelineCaches(
+ VkDevice device,
+ VkPipelineCache destCache,
+ uint32_t srcCacheCount,
+ const VkPipelineCache* pSrcCaches)
+{
+ stub_return(VK_SUCCESS);
+}
+
+void anv_DestroyPipeline(
+ VkDevice _device,
+ VkPipeline _pipeline,
+ const VkAllocationCallbacks* pAllocator)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ ANV_FROM_HANDLE(anv_pipeline, pipeline, _pipeline);
+
+ anv_reloc_list_finish(&pipeline->batch_relocs,
+ pAllocator ? pAllocator : &device->alloc);
+ if (pipeline->blend_state.map)
+ anv_state_pool_free(&device->dynamic_state_pool, pipeline->blend_state);
+ anv_free2(&device->alloc, pAllocator, pipeline);
+}
+
+static const uint32_t vk_to_gen_primitive_type[] = {
+ [VK_PRIMITIVE_TOPOLOGY_POINT_LIST] = _3DPRIM_POINTLIST,
+ [VK_PRIMITIVE_TOPOLOGY_LINE_LIST] = _3DPRIM_LINELIST,
+ [VK_PRIMITIVE_TOPOLOGY_LINE_STRIP] = _3DPRIM_LINESTRIP,
+ [VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST] = _3DPRIM_TRILIST,
+ [VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP] = _3DPRIM_TRISTRIP,
+ [VK_PRIMITIVE_TOPOLOGY_TRIANGLE_FAN] = _3DPRIM_TRIFAN,
+ [VK_PRIMITIVE_TOPOLOGY_LINE_LIST_WITH_ADJACENCY] = _3DPRIM_LINELIST_ADJ,
+ [VK_PRIMITIVE_TOPOLOGY_LINE_STRIP_WITH_ADJACENCY] = _3DPRIM_LINESTRIP_ADJ,
+ [VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST_WITH_ADJACENCY] = _3DPRIM_TRILIST_ADJ,
+ [VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP_WITH_ADJACENCY] = _3DPRIM_TRISTRIP_ADJ,
+/* [VK_PRIMITIVE_TOPOLOGY_PATCH_LIST] = _3DPRIM_PATCHLIST_1 */
+};
+
+static void
+populate_sampler_prog_key(const struct brw_device_info *devinfo,
+ struct brw_sampler_prog_key_data *key)
+{
+ /* XXX: Handle texture swizzle on HSW- */
+ for (int i = 0; i < MAX_SAMPLERS; i++) {
+ /* Assume color sampler, no swizzling. (Works for BDW+) */
+ key->swizzles[i] = SWIZZLE_XYZW;
+ }
+}
+
+static void
+populate_vs_prog_key(const struct brw_device_info *devinfo,
+ struct brw_vs_prog_key *key)
+{
+ memset(key, 0, sizeof(*key));
+
+ populate_sampler_prog_key(devinfo, &key->tex);
+
+ /* XXX: Handle vertex input work-arounds */
+
+ /* XXX: Handle sampler_prog_key */
+}
+
+static void
+populate_gs_prog_key(const struct brw_device_info *devinfo,
+ struct brw_gs_prog_key *key)
+{
+ memset(key, 0, sizeof(*key));
+
+ populate_sampler_prog_key(devinfo, &key->tex);
+}
+
+static void
+populate_wm_prog_key(const struct brw_device_info *devinfo,
+ const VkGraphicsPipelineCreateInfo *info,
+ const struct anv_graphics_pipeline_create_info *extra,
+ struct brw_wm_prog_key *key)
+{
+ ANV_FROM_HANDLE(anv_render_pass, render_pass, info->renderPass);
+
+ memset(key, 0, sizeof(*key));
+
+ populate_sampler_prog_key(devinfo, &key->tex);
+
+ /* TODO: Fill out key->input_slots_valid */
+
+ /* Vulkan doesn't specify a default */
+ key->high_quality_derivatives = false;
+
+ /* XXX Vulkan doesn't appear to specify */
+ key->clamp_fragment_color = false;
+
+ /* Vulkan always specifies upper-left coordinates */
+ key->drawable_height = 0;
+ key->render_to_fbo = false;
+
+ if (extra && extra->color_attachment_count >= 0) {
+ key->nr_color_regions = extra->color_attachment_count;
+ } else {
+ key->nr_color_regions =
+ render_pass->subpasses[info->subpass].color_count;
+ }
+
+ key->replicate_alpha = key->nr_color_regions > 1 &&
+ info->pMultisampleState &&
+ info->pMultisampleState->alphaToCoverageEnable;
+
+ if (info->pMultisampleState && info->pMultisampleState->rasterizationSamples > 1) {
+ /* We should probably pull this out of the shader, but it's fairly
+ * harmless to compute it and then let dead-code take care of it.
+ */
+ key->persample_shading = info->pMultisampleState->sampleShadingEnable;
+ if (key->persample_shading)
+ key->persample_2x = info->pMultisampleState->rasterizationSamples == 2;
+
+ key->compute_pos_offset = info->pMultisampleState->sampleShadingEnable;
+ key->compute_sample_id = info->pMultisampleState->sampleShadingEnable;
+ }
+}
+
+static void
+populate_cs_prog_key(const struct brw_device_info *devinfo,
+ struct brw_cs_prog_key *key)
+{
+ memset(key, 0, sizeof(*key));
+
+ populate_sampler_prog_key(devinfo, &key->tex);
+}
+
+static nir_shader *
+anv_pipeline_compile(struct anv_pipeline *pipeline,
+ struct anv_shader_module *module,
+ const char *entrypoint,
+ gl_shader_stage stage,
+ const VkSpecializationInfo *spec_info,
+ struct brw_stage_prog_data *prog_data)
+{
+ const struct brw_compiler *compiler =
+ pipeline->device->instance->physicalDevice.compiler;
+
+ nir_shader *nir = anv_shader_compile_to_nir(pipeline->device,
+ module, entrypoint, stage,
+ spec_info);
+ if (nir == NULL)
+ return NULL;
+
+ anv_nir_lower_push_constants(nir, compiler->scalar_stage[stage]);
+
+ /* Figure out the number of parameters */
+ prog_data->nr_params = 0;
+
+ if (nir->num_uniforms > 0) {
+ /* If the shader uses any push constants at all, we'll just give
+ * them the maximum possible number
+ */
+ prog_data->nr_params += MAX_PUSH_CONSTANTS_SIZE / sizeof(float);
+ }
+
+ if (pipeline->layout && pipeline->layout->stage[stage].has_dynamic_offsets)
+ prog_data->nr_params += MAX_DYNAMIC_BUFFERS * 2;
+
+ if (pipeline->layout && pipeline->layout->stage[stage].image_count > 0)
+ prog_data->nr_params += pipeline->layout->stage[stage].image_count *
+ BRW_IMAGE_PARAM_SIZE;
+
+ if (prog_data->nr_params > 0) {
+ /* XXX: I think we're leaking this */
+ prog_data->param = (const union gl_constant_value **)
+ malloc(prog_data->nr_params * sizeof(union gl_constant_value *));
+
+ /* We now set the param values to be offsets into a
+ * anv_push_constant_data structure. Since the compiler doesn't
+ * actually dereference any of the gl_constant_value pointers in the
+ * params array, it doesn't really matter what we put here.
+ */
+ struct anv_push_constants *null_data = NULL;
+ if (nir->num_uniforms > 0) {
+ /* Fill out the push constants section of the param array */
+ for (unsigned i = 0; i < MAX_PUSH_CONSTANTS_SIZE / sizeof(float); i++)
+ prog_data->param[i] = (const union gl_constant_value *)
+ &null_data->client_data[i * sizeof(float)];
+ }
+ }
+
+ /* Set up dynamic offsets */
+ anv_nir_apply_dynamic_offsets(pipeline, nir, prog_data);
+
+ /* Apply the actual pipeline layout to UBOs, SSBOs, and textures */
+ if (pipeline->layout)
+ anv_nir_apply_pipeline_layout(nir, prog_data, pipeline->layout);
+
+ /* All binding table offsets provided by apply_pipeline_layout() are
+ * relative to the start of the bindint table (plus MAX_RTS for VS).
+ */
+ unsigned bias;
+ switch (stage) {
+ case MESA_SHADER_FRAGMENT:
+ bias = MAX_RTS;
+ break;
+ case MESA_SHADER_COMPUTE:
+ bias = 1;
+ break;
+ default:
+ bias = 0;
+ break;
+ }
+ prog_data->binding_table.size_bytes = 0;
+ prog_data->binding_table.texture_start = bias;
+ prog_data->binding_table.ubo_start = bias;
+ prog_data->binding_table.ssbo_start = bias;
+ prog_data->binding_table.image_start = bias;
+
+ /* Finish the optimization and compilation process */
+ nir = brw_nir_lower_io(nir, &pipeline->device->info,
+ compiler->scalar_stage[stage]);
+
+ /* nir_lower_io will only handle the push constants; we need to set this
+ * to the full number of possible uniforms.
+ */
+ nir->num_uniforms = prog_data->nr_params * 4;
+
+ return nir;
+}
+
+static void
+anv_pipeline_add_compiled_stage(struct anv_pipeline *pipeline,
+ gl_shader_stage stage,
+ struct brw_stage_prog_data *prog_data)
+{
+ struct brw_device_info *devinfo = &pipeline->device->info;
+ uint32_t max_threads[] = {
+ [MESA_SHADER_VERTEX] = devinfo->max_vs_threads,
+ [MESA_SHADER_TESS_CTRL] = 0,
+ [MESA_SHADER_TESS_EVAL] = 0,
+ [MESA_SHADER_GEOMETRY] = devinfo->max_gs_threads,
+ [MESA_SHADER_FRAGMENT] = devinfo->max_wm_threads,
+ [MESA_SHADER_COMPUTE] = devinfo->max_cs_threads,
+ };
+
+ pipeline->prog_data[stage] = prog_data;
+ pipeline->active_stages |= mesa_to_vk_shader_stage(stage);
+ pipeline->scratch_start[stage] = pipeline->total_scratch;
+ pipeline->total_scratch =
+ align_u32(pipeline->total_scratch, 1024) +
+ prog_data->total_scratch * max_threads[stage];
+}
+
+static VkResult
+anv_pipeline_compile_vs(struct anv_pipeline *pipeline,
+ struct anv_pipeline_cache *cache,
+ const VkGraphicsPipelineCreateInfo *info,
+ struct anv_shader_module *module,
+ const char *entrypoint,
+ const VkSpecializationInfo *spec_info)
+{
+ const struct brw_compiler *compiler =
+ pipeline->device->instance->physicalDevice.compiler;
+ struct brw_vs_prog_data *prog_data = &pipeline->vs_prog_data;
+ struct brw_vs_prog_key key;
+
+ populate_vs_prog_key(&pipeline->device->info, &key);
+
+ /* TODO: Look up shader in cache */
+
+ memset(prog_data, 0, sizeof(*prog_data));
+
+ nir_shader *nir = anv_pipeline_compile(pipeline, module, entrypoint,
+ MESA_SHADER_VERTEX, spec_info,
+ &prog_data->base.base);
+ if (nir == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ void *mem_ctx = ralloc_context(NULL);
+
+ if (module->nir == NULL)
+ ralloc_steal(mem_ctx, nir);
+
+ prog_data->inputs_read = nir->info.inputs_read;
+ if (nir->info.outputs_written & (1ull << VARYING_SLOT_PSIZ))
+ pipeline->writes_point_size = true;
+
+ brw_compute_vue_map(&pipeline->device->info,
+ &prog_data->base.vue_map,
+ nir->info.outputs_written,
+ nir->info.separate_shader);
+
+ unsigned code_size;
+ const unsigned *shader_code =
+ brw_compile_vs(compiler, NULL, mem_ctx, &key, prog_data, nir,
+ NULL, false, -1, &code_size, NULL);
+ if (shader_code == NULL) {
+ ralloc_free(mem_ctx);
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+ }
+
+ const uint32_t offset =
+ anv_pipeline_cache_upload_kernel(cache, shader_code, code_size);
+ if (prog_data->base.dispatch_mode == DISPATCH_MODE_SIMD8) {
+ pipeline->vs_simd8 = offset;
+ pipeline->vs_vec4 = NO_KERNEL;
+ } else {
+ pipeline->vs_simd8 = NO_KERNEL;
+ pipeline->vs_vec4 = offset;
+ }
+
+ ralloc_free(mem_ctx);
+
+ anv_pipeline_add_compiled_stage(pipeline, MESA_SHADER_VERTEX,
+ &prog_data->base.base);
+
+ return VK_SUCCESS;
+}
+
+static VkResult
+anv_pipeline_compile_gs(struct anv_pipeline *pipeline,
+ struct anv_pipeline_cache *cache,
+ const VkGraphicsPipelineCreateInfo *info,
+ struct anv_shader_module *module,
+ const char *entrypoint,
+ const VkSpecializationInfo *spec_info)
+{
+ const struct brw_compiler *compiler =
+ pipeline->device->instance->physicalDevice.compiler;
+ struct brw_gs_prog_data *prog_data = &pipeline->gs_prog_data;
+ struct brw_gs_prog_key key;
+
+ populate_gs_prog_key(&pipeline->device->info, &key);
+
+ /* TODO: Look up shader in cache */
+
+ memset(prog_data, 0, sizeof(*prog_data));
+
+ nir_shader *nir = anv_pipeline_compile(pipeline, module, entrypoint,
+ MESA_SHADER_GEOMETRY, spec_info,
+ &prog_data->base.base);
+ if (nir == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ void *mem_ctx = ralloc_context(NULL);
+
+ if (module->nir == NULL)
+ ralloc_steal(mem_ctx, nir);
+
+ if (nir->info.outputs_written & (1ull << VARYING_SLOT_PSIZ))
+ pipeline->writes_point_size = true;
+
+ brw_compute_vue_map(&pipeline->device->info,
+ &prog_data->base.vue_map,
+ nir->info.outputs_written,
+ nir->info.separate_shader);
+
+ unsigned code_size;
+ const unsigned *shader_code =
+ brw_compile_gs(compiler, NULL, mem_ctx, &key, prog_data, nir,
+ NULL, -1, &code_size, NULL);
+ if (shader_code == NULL) {
+ ralloc_free(mem_ctx);
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+ }
+
+ /* TODO: SIMD8 GS */
+ pipeline->gs_kernel =
+ anv_pipeline_cache_upload_kernel(cache, shader_code, code_size);
+ pipeline->gs_vertex_count = nir->info.gs.vertices_in;
+
+ ralloc_free(mem_ctx);
+
+ anv_pipeline_add_compiled_stage(pipeline, MESA_SHADER_GEOMETRY,
+ &prog_data->base.base);
+
+ return VK_SUCCESS;
+}
+
+static VkResult
+anv_pipeline_compile_fs(struct anv_pipeline *pipeline,
+ struct anv_pipeline_cache *cache,
+ const VkGraphicsPipelineCreateInfo *info,
+ const struct anv_graphics_pipeline_create_info *extra,
+ struct anv_shader_module *module,
+ const char *entrypoint,
+ const VkSpecializationInfo *spec_info)
+{
+ const struct brw_compiler *compiler =
+ pipeline->device->instance->physicalDevice.compiler;
+ struct brw_wm_prog_data *prog_data = &pipeline->wm_prog_data;
+ struct brw_wm_prog_key key;
+
+ populate_wm_prog_key(&pipeline->device->info, info, extra, &key);
+
+ if (pipeline->use_repclear)
+ key.nr_color_regions = 1;
+
+ /* TODO: Look up shader in cache */
+
+ memset(prog_data, 0, sizeof(*prog_data));
+
+ prog_data->binding_table.render_target_start = 0;
+
+ nir_shader *nir = anv_pipeline_compile(pipeline, module, entrypoint,
+ MESA_SHADER_FRAGMENT, spec_info,
+ &prog_data->base);
+ if (nir == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ void *mem_ctx = ralloc_context(NULL);
+
+ if (module->nir == NULL)
+ ralloc_steal(mem_ctx, nir);
+
+ unsigned code_size;
+ const unsigned *shader_code =
+ brw_compile_fs(compiler, NULL, mem_ctx, &key, prog_data, nir,
+ NULL, -1, -1, pipeline->use_repclear, &code_size, NULL);
+ if (shader_code == NULL) {
+ ralloc_free(mem_ctx);
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+ }
+
+ uint32_t offset =
+ anv_pipeline_cache_upload_kernel(cache, shader_code, code_size);
+ if (prog_data->no_8)
+ pipeline->ps_simd8 = NO_KERNEL;
+ else
+ pipeline->ps_simd8 = offset;
+
+ if (prog_data->no_8 || prog_data->prog_offset_16) {
+ pipeline->ps_simd16 = offset + prog_data->prog_offset_16;
+ } else {
+ pipeline->ps_simd16 = NO_KERNEL;
+ }
+
+ pipeline->ps_ksp2 = 0;
+ pipeline->ps_grf_start2 = 0;
+ if (pipeline->ps_simd8 != NO_KERNEL) {
+ pipeline->ps_ksp0 = pipeline->ps_simd8;
+ pipeline->ps_grf_start0 = prog_data->base.dispatch_grf_start_reg;
+ if (pipeline->ps_simd16 != NO_KERNEL) {
+ pipeline->ps_ksp2 = pipeline->ps_simd16;
+ pipeline->ps_grf_start2 = prog_data->dispatch_grf_start_reg_16;
+ }
+ } else if (pipeline->ps_simd16 != NO_KERNEL) {
+ pipeline->ps_ksp0 = pipeline->ps_simd16;
+ pipeline->ps_grf_start0 = prog_data->dispatch_grf_start_reg_16;
+ }
+
+ ralloc_free(mem_ctx);
+
+ anv_pipeline_add_compiled_stage(pipeline, MESA_SHADER_FRAGMENT,
+ &prog_data->base);
+
+ return VK_SUCCESS;
+}
+
+VkResult
+anv_pipeline_compile_cs(struct anv_pipeline *pipeline,
+ struct anv_pipeline_cache *cache,
+ const VkComputePipelineCreateInfo *info,
+ struct anv_shader_module *module,
+ const char *entrypoint,
+ const VkSpecializationInfo *spec_info)
+{
+ const struct brw_compiler *compiler =
+ pipeline->device->instance->physicalDevice.compiler;
+ struct brw_cs_prog_data *prog_data = &pipeline->cs_prog_data;
+ struct brw_cs_prog_key key;
+
+ populate_cs_prog_key(&pipeline->device->info, &key);
+
+ /* TODO: Look up shader in cache */
+
+ memset(prog_data, 0, sizeof(*prog_data));
+
+ prog_data->binding_table.work_groups_start = 0;
+
+ nir_shader *nir = anv_pipeline_compile(pipeline, module, entrypoint,
+ MESA_SHADER_COMPUTE, spec_info,
+ &prog_data->base);
+ if (nir == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ prog_data->base.total_shared = nir->num_shared;
+
+ void *mem_ctx = ralloc_context(NULL);
+
+ if (module->nir == NULL)
+ ralloc_steal(mem_ctx, nir);
+
+ unsigned code_size;
+ const unsigned *shader_code =
+ brw_compile_cs(compiler, NULL, mem_ctx, &key, prog_data, nir,
+ -1, &code_size, NULL);
+ if (shader_code == NULL) {
+ ralloc_free(mem_ctx);
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+ }
+
+ pipeline->cs_simd =
+ anv_pipeline_cache_upload_kernel(cache, shader_code, code_size);
+ ralloc_free(mem_ctx);
+
+ anv_pipeline_add_compiled_stage(pipeline, MESA_SHADER_COMPUTE,
+ &prog_data->base);
+
+ return VK_SUCCESS;
+}
+
+static const int gen8_push_size = 32 * 1024;
+
+static void
+gen7_compute_urb_partition(struct anv_pipeline *pipeline)
+{
+ const struct brw_device_info *devinfo = &pipeline->device->info;
+ bool vs_present = pipeline->active_stages & VK_SHADER_STAGE_VERTEX_BIT;
+ unsigned vs_size = vs_present ? pipeline->vs_prog_data.base.urb_entry_size : 1;
+ unsigned vs_entry_size_bytes = vs_size * 64;
+ bool gs_present = pipeline->active_stages & VK_SHADER_STAGE_GEOMETRY_BIT;
+ unsigned gs_size = gs_present ? pipeline->gs_prog_data.base.urb_entry_size : 1;
+ unsigned gs_entry_size_bytes = gs_size * 64;
+
+ /* From p35 of the Ivy Bridge PRM (section 1.7.1: 3DSTATE_URB_GS):
+ *
+ * VS Number of URB Entries must be divisible by 8 if the VS URB Entry
+ * Allocation Size is less than 9 512-bit URB entries.
+ *
+ * Similar text exists for GS.
+ */
+ unsigned vs_granularity = (vs_size < 9) ? 8 : 1;
+ unsigned gs_granularity = (gs_size < 9) ? 8 : 1;
+
+ /* URB allocations must be done in 8k chunks. */
+ unsigned chunk_size_bytes = 8192;
+
+ /* Determine the size of the URB in chunks. */
+ unsigned urb_chunks = devinfo->urb.size * 1024 / chunk_size_bytes;
+
+ /* Reserve space for push constants */
+ unsigned push_constant_bytes = gen8_push_size;
+ unsigned push_constant_chunks =
+ push_constant_bytes / chunk_size_bytes;
+
+ /* Initially, assign each stage the minimum amount of URB space it needs,
+ * and make a note of how much additional space it "wants" (the amount of
+ * additional space it could actually make use of).
+ */
+
+ /* VS has a lower limit on the number of URB entries */
+ unsigned vs_chunks =
+ ALIGN(devinfo->urb.min_vs_entries * vs_entry_size_bytes,
+ chunk_size_bytes) / chunk_size_bytes;
+ unsigned vs_wants =
+ ALIGN(devinfo->urb.max_vs_entries * vs_entry_size_bytes,
+ chunk_size_bytes) / chunk_size_bytes - vs_chunks;
+
+ unsigned gs_chunks = 0;
+ unsigned gs_wants = 0;
+ if (gs_present) {
+ /* There are two constraints on the minimum amount of URB space we can
+ * allocate:
+ *
+ * (1) We need room for at least 2 URB entries, since we always operate
+ * the GS in DUAL_OBJECT mode.
+ *
+ * (2) We can't allocate less than nr_gs_entries_granularity.
+ */
+ gs_chunks = ALIGN(MAX2(gs_granularity, 2) * gs_entry_size_bytes,
+ chunk_size_bytes) / chunk_size_bytes;
+ gs_wants =
+ ALIGN(devinfo->urb.max_gs_entries * gs_entry_size_bytes,
+ chunk_size_bytes) / chunk_size_bytes - gs_chunks;
+ }
+
+ /* There should always be enough URB space to satisfy the minimum
+ * requirements of each stage.
+ */
+ unsigned total_needs = push_constant_chunks + vs_chunks + gs_chunks;
+ assert(total_needs <= urb_chunks);
+
+ /* Mete out remaining space (if any) in proportion to "wants". */
+ unsigned total_wants = vs_wants + gs_wants;
+ unsigned remaining_space = urb_chunks - total_needs;
+ if (remaining_space > total_wants)
+ remaining_space = total_wants;
+ if (remaining_space > 0) {
+ unsigned vs_additional = (unsigned)
+ round(vs_wants * (((double) remaining_space) / total_wants));
+ vs_chunks += vs_additional;
+ remaining_space -= vs_additional;
+ gs_chunks += remaining_space;
+ }
+
+ /* Sanity check that we haven't over-allocated. */
+ assert(push_constant_chunks + vs_chunks + gs_chunks <= urb_chunks);
+
+ /* Finally, compute the number of entries that can fit in the space
+ * allocated to each stage.
+ */
+ unsigned nr_vs_entries = vs_chunks * chunk_size_bytes / vs_entry_size_bytes;
+ unsigned nr_gs_entries = gs_chunks * chunk_size_bytes / gs_entry_size_bytes;
+
+ /* Since we rounded up when computing *_wants, this may be slightly more
+ * than the maximum allowed amount, so correct for that.
+ */
+ nr_vs_entries = MIN2(nr_vs_entries, devinfo->urb.max_vs_entries);
+ nr_gs_entries = MIN2(nr_gs_entries, devinfo->urb.max_gs_entries);
+
+ /* Ensure that we program a multiple of the granularity. */
+ nr_vs_entries = ROUND_DOWN_TO(nr_vs_entries, vs_granularity);
+ nr_gs_entries = ROUND_DOWN_TO(nr_gs_entries, gs_granularity);
+
+ /* Finally, sanity check to make sure we have at least the minimum number
+ * of entries needed for each stage.
+ */
+ assert(nr_vs_entries >= devinfo->urb.min_vs_entries);
+ if (gs_present)
+ assert(nr_gs_entries >= 2);
+
+ /* Lay out the URB in the following order:
+ * - push constants
+ * - VS
+ * - GS
+ */
+ pipeline->urb.vs_start = push_constant_chunks;
+ pipeline->urb.vs_size = vs_size;
+ pipeline->urb.nr_vs_entries = nr_vs_entries;
+
+ pipeline->urb.gs_start = push_constant_chunks + vs_chunks;
+ pipeline->urb.gs_size = gs_size;
+ pipeline->urb.nr_gs_entries = nr_gs_entries;
+}
+
+static void
+anv_pipeline_init_dynamic_state(struct anv_pipeline *pipeline,
+ const VkGraphicsPipelineCreateInfo *pCreateInfo)
+{
+ anv_cmd_dirty_mask_t states = ANV_CMD_DIRTY_DYNAMIC_ALL;
+ ANV_FROM_HANDLE(anv_render_pass, pass, pCreateInfo->renderPass);
+ struct anv_subpass *subpass = &pass->subpasses[pCreateInfo->subpass];
+
+ pipeline->dynamic_state = default_dynamic_state;
+
+ if (pCreateInfo->pDynamicState) {
+ /* Remove all of the states that are marked as dynamic */
+ uint32_t count = pCreateInfo->pDynamicState->dynamicStateCount;
+ for (uint32_t s = 0; s < count; s++)
+ states &= ~(1 << pCreateInfo->pDynamicState->pDynamicStates[s]);
+ }
+
+ struct anv_dynamic_state *dynamic = &pipeline->dynamic_state;
+
+ dynamic->viewport.count = pCreateInfo->pViewportState->viewportCount;
+ if (states & (1 << VK_DYNAMIC_STATE_VIEWPORT)) {
+ typed_memcpy(dynamic->viewport.viewports,
+ pCreateInfo->pViewportState->pViewports,
+ pCreateInfo->pViewportState->viewportCount);
+ }
+
+ dynamic->scissor.count = pCreateInfo->pViewportState->scissorCount;
+ if (states & (1 << VK_DYNAMIC_STATE_SCISSOR)) {
+ typed_memcpy(dynamic->scissor.scissors,
+ pCreateInfo->pViewportState->pScissors,
+ pCreateInfo->pViewportState->scissorCount);
+ }
+
+ if (states & (1 << VK_DYNAMIC_STATE_LINE_WIDTH)) {
+ assert(pCreateInfo->pRasterizationState);
+ dynamic->line_width = pCreateInfo->pRasterizationState->lineWidth;
+ }
+
+ if (states & (1 << VK_DYNAMIC_STATE_DEPTH_BIAS)) {
+ assert(pCreateInfo->pRasterizationState);
+ dynamic->depth_bias.bias =
+ pCreateInfo->pRasterizationState->depthBiasConstantFactor;
+ dynamic->depth_bias.clamp =
+ pCreateInfo->pRasterizationState->depthBiasClamp;
+ dynamic->depth_bias.slope =
+ pCreateInfo->pRasterizationState->depthBiasSlopeFactor;
+ }
+
+ if (states & (1 << VK_DYNAMIC_STATE_BLEND_CONSTANTS)) {
+ assert(pCreateInfo->pColorBlendState);
+ typed_memcpy(dynamic->blend_constants,
+ pCreateInfo->pColorBlendState->blendConstants, 4);
+ }
+
+ /* If there is no depthstencil attachment, then don't read
+ * pDepthStencilState. The Vulkan spec states that pDepthStencilState may
+ * be NULL in this case. Even if pDepthStencilState is non-NULL, there is
+ * no need to override the depthstencil defaults in
+ * anv_pipeline::dynamic_state when there is no depthstencil attachment.
+ *
+ * From the Vulkan spec (20 Oct 2015, git-aa308cb):
+ *
+ * pDepthStencilState [...] may only be NULL if renderPass and subpass
+ * specify a subpass that has no depth/stencil attachment.
+ */
+ if (subpass->depth_stencil_attachment != VK_ATTACHMENT_UNUSED) {
+ if (states & (1 << VK_DYNAMIC_STATE_DEPTH_BOUNDS)) {
+ assert(pCreateInfo->pDepthStencilState);
+ dynamic->depth_bounds.min =
+ pCreateInfo->pDepthStencilState->minDepthBounds;
+ dynamic->depth_bounds.max =
+ pCreateInfo->pDepthStencilState->maxDepthBounds;
+ }
+
+ if (states & (1 << VK_DYNAMIC_STATE_STENCIL_COMPARE_MASK)) {
+ assert(pCreateInfo->pDepthStencilState);
+ dynamic->stencil_compare_mask.front =
+ pCreateInfo->pDepthStencilState->front.compareMask;
+ dynamic->stencil_compare_mask.back =
+ pCreateInfo->pDepthStencilState->back.compareMask;
+ }
+
+ if (states & (1 << VK_DYNAMIC_STATE_STENCIL_WRITE_MASK)) {
+ assert(pCreateInfo->pDepthStencilState);
+ dynamic->stencil_write_mask.front =
+ pCreateInfo->pDepthStencilState->front.writeMask;
+ dynamic->stencil_write_mask.back =
+ pCreateInfo->pDepthStencilState->back.writeMask;
+ }
+
+ if (states & (1 << VK_DYNAMIC_STATE_STENCIL_REFERENCE)) {
+ assert(pCreateInfo->pDepthStencilState);
+ dynamic->stencil_reference.front =
+ pCreateInfo->pDepthStencilState->front.reference;
+ dynamic->stencil_reference.back =
+ pCreateInfo->pDepthStencilState->back.reference;
+ }
+ }
+
+ pipeline->dynamic_state_mask = states;
+}
+
+static void
+anv_pipeline_validate_create_info(const VkGraphicsPipelineCreateInfo *info)
+{
+ struct anv_render_pass *renderpass = NULL;
+ struct anv_subpass *subpass = NULL;
+
+ /* Assert that all required members of VkGraphicsPipelineCreateInfo are
+ * present, as explained by the Vulkan (20 Oct 2015, git-aa308cb), Section
+ * 4.2 Graphics Pipeline.
+ */
+ assert(info->sType == VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO);
+
+ renderpass = anv_render_pass_from_handle(info->renderPass);
+ assert(renderpass);
+
+ if (renderpass != &anv_meta_dummy_renderpass) {
+ assert(info->subpass < renderpass->subpass_count);
+ subpass = &renderpass->subpasses[info->subpass];
+ }
+
+ assert(info->stageCount >= 1);
+ assert(info->pVertexInputState);
+ assert(info->pInputAssemblyState);
+ assert(info->pViewportState);
+ assert(info->pRasterizationState);
+
+ if (subpass && subpass->depth_stencil_attachment != VK_ATTACHMENT_UNUSED)
+ assert(info->pDepthStencilState);
+
+ if (subpass && subpass->color_count > 0)
+ assert(info->pColorBlendState);
+
+ for (uint32_t i = 0; i < info->stageCount; ++i) {
+ switch (info->pStages[i].stage) {
+ case VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT:
+ case VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT:
+ assert(info->pTessellationState);
+ break;
+ default:
+ break;
+ }
+ }
+}
+
+VkResult
+anv_pipeline_init(struct anv_pipeline *pipeline,
+ struct anv_device *device,
+ struct anv_pipeline_cache *cache,
+ const VkGraphicsPipelineCreateInfo *pCreateInfo,
+ const struct anv_graphics_pipeline_create_info *extra,
+ const VkAllocationCallbacks *alloc)
+{
+ VkResult result;
+
+ anv_validate {
+ anv_pipeline_validate_create_info(pCreateInfo);
+ }
+
+ if (alloc == NULL)
+ alloc = &device->alloc;
+
+ pipeline->device = device;
+ pipeline->layout = anv_pipeline_layout_from_handle(pCreateInfo->layout);
+
+ result = anv_reloc_list_init(&pipeline->batch_relocs, alloc);
+ if (result != VK_SUCCESS)
+ return result;
+
+ pipeline->batch.alloc = alloc;
+ pipeline->batch.next = pipeline->batch.start = pipeline->batch_data;
+ pipeline->batch.end = pipeline->batch.start + sizeof(pipeline->batch_data);
+ pipeline->batch.relocs = &pipeline->batch_relocs;
+
+ anv_pipeline_init_dynamic_state(pipeline, pCreateInfo);
+
+ if (pCreateInfo->pTessellationState)
+ anv_finishme("VK_STRUCTURE_TYPE_PIPELINE_TESSELLATION_STATE_CREATE_INFO");
+ if (pCreateInfo->pMultisampleState &&
+ pCreateInfo->pMultisampleState->rasterizationSamples > 1)
+ anv_finishme("VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO");
+
+ pipeline->use_repclear = extra && extra->use_repclear;
+ pipeline->writes_point_size = false;
+
+ /* When we free the pipeline, we detect stages based on the NULL status
+ * of various prog_data pointers. Make them NULL by default.
+ */
+ memset(pipeline->prog_data, 0, sizeof(pipeline->prog_data));
+ memset(pipeline->scratch_start, 0, sizeof(pipeline->scratch_start));
+
+ pipeline->vs_simd8 = NO_KERNEL;
+ pipeline->vs_vec4 = NO_KERNEL;
+ pipeline->gs_kernel = NO_KERNEL;
+
+ pipeline->active_stages = 0;
+ pipeline->total_scratch = 0;
+
+ for (uint32_t i = 0; i < pCreateInfo->stageCount; i++) {
+ ANV_FROM_HANDLE(anv_shader_module, module,
+ pCreateInfo->pStages[i].module);
+
+ switch (pCreateInfo->pStages[i].stage) {
+ case VK_SHADER_STAGE_VERTEX_BIT:
+ anv_pipeline_compile_vs(pipeline, cache, pCreateInfo, module,
+ pCreateInfo->pStages[i].pName,
+ pCreateInfo->pStages[i].pSpecializationInfo);
+ break;
+ case VK_SHADER_STAGE_GEOMETRY_BIT:
+ anv_pipeline_compile_gs(pipeline, cache, pCreateInfo, module,
+ pCreateInfo->pStages[i].pName,
+ pCreateInfo->pStages[i].pSpecializationInfo);
+ break;
+ case VK_SHADER_STAGE_FRAGMENT_BIT:
+ anv_pipeline_compile_fs(pipeline, cache, pCreateInfo, extra, module,
+ pCreateInfo->pStages[i].pName,
+ pCreateInfo->pStages[i].pSpecializationInfo);
+ break;
+ default:
+ anv_finishme("Unsupported shader stage");
+ }
+ }
+
+ if (!(pipeline->active_stages & VK_SHADER_STAGE_VERTEX_BIT)) {
+ /* Vertex is only optional if disable_vs is set */
+ assert(extra->disable_vs);
+ memset(&pipeline->vs_prog_data, 0, sizeof(pipeline->vs_prog_data));
+ }
+
+ gen7_compute_urb_partition(pipeline);
+
+ const VkPipelineVertexInputStateCreateInfo *vi_info =
+ pCreateInfo->pVertexInputState;
+
+ uint64_t inputs_read;
+ if (extra && extra->disable_vs) {
+ /* If the VS is disabled, just assume the user knows what they're
+ * doing and apply the layout blindly. This can only come from
+ * meta, so this *should* be safe.
+ */
+ inputs_read = ~0ull;
+ } else {
+ inputs_read = pipeline->vs_prog_data.inputs_read;
+ }
+
+ pipeline->vb_used = 0;
+ for (uint32_t i = 0; i < vi_info->vertexAttributeDescriptionCount; i++) {
+ const VkVertexInputAttributeDescription *desc =
+ &vi_info->pVertexAttributeDescriptions[i];
+
+ if (inputs_read & (1 << (VERT_ATTRIB_GENERIC0 + desc->location)))
+ pipeline->vb_used |= 1 << desc->binding;
+ }
+
+ for (uint32_t i = 0; i < vi_info->vertexBindingDescriptionCount; i++) {
+ const VkVertexInputBindingDescription *desc =
+ &vi_info->pVertexBindingDescriptions[i];
+
+ pipeline->binding_stride[desc->binding] = desc->stride;
+
+ /* Step rate is programmed per vertex element (attribute), not
+ * binding. Set up a map of which bindings step per instance, for
+ * reference by vertex element setup. */
+ switch (desc->inputRate) {
+ default:
+ case VK_VERTEX_INPUT_RATE_VERTEX:
+ pipeline->instancing_enable[desc->binding] = false;
+ break;
+ case VK_VERTEX_INPUT_RATE_INSTANCE:
+ pipeline->instancing_enable[desc->binding] = true;
+ break;
+ }
+ }
+
+ const VkPipelineInputAssemblyStateCreateInfo *ia_info =
+ pCreateInfo->pInputAssemblyState;
+ pipeline->primitive_restart = ia_info->primitiveRestartEnable;
+ pipeline->topology = vk_to_gen_primitive_type[ia_info->topology];
+
+ if (extra && extra->use_rectlist)
+ pipeline->topology = _3DPRIM_RECTLIST;
+
+ return VK_SUCCESS;
+}
+
+VkResult
+anv_graphics_pipeline_create(
+ VkDevice _device,
+ VkPipelineCache _cache,
+ const VkGraphicsPipelineCreateInfo *pCreateInfo,
+ const struct anv_graphics_pipeline_create_info *extra,
+ const VkAllocationCallbacks *pAllocator,
+ VkPipeline *pPipeline)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ ANV_FROM_HANDLE(anv_pipeline_cache, cache, _cache);
+
+ if (cache == NULL)
+ cache = &device->default_pipeline_cache;
+
+ switch (device->info.gen) {
+ case 7:
+ if (device->info.is_haswell)
+ return gen75_graphics_pipeline_create(_device, cache, pCreateInfo, extra, pAllocator, pPipeline);
+ else
+ return gen7_graphics_pipeline_create(_device, cache, pCreateInfo, extra, pAllocator, pPipeline);
+ case 8:
+ return gen8_graphics_pipeline_create(_device, cache, pCreateInfo, extra, pAllocator, pPipeline);
+ case 9:
+ return gen9_graphics_pipeline_create(_device, cache, pCreateInfo, extra, pAllocator, pPipeline);
+ default:
+ unreachable("unsupported gen\n");
+ }
+}
+
+VkResult anv_CreateGraphicsPipelines(
+ VkDevice _device,
+ VkPipelineCache pipelineCache,
+ uint32_t count,
+ const VkGraphicsPipelineCreateInfo* pCreateInfos,
+ const VkAllocationCallbacks* pAllocator,
+ VkPipeline* pPipelines)
+{
+ VkResult result = VK_SUCCESS;
+
+ unsigned i = 0;
+ for (; i < count; i++) {
+ result = anv_graphics_pipeline_create(_device,
+ pipelineCache,
+ &pCreateInfos[i],
+ NULL, pAllocator, &pPipelines[i]);
+ if (result != VK_SUCCESS) {
+ for (unsigned j = 0; j < i; j++) {
+ anv_DestroyPipeline(_device, pPipelines[j], pAllocator);
+ }
+
+ return result;
+ }
+ }
+
+ return VK_SUCCESS;
+}
+
+static VkResult anv_compute_pipeline_create(
+ VkDevice _device,
+ VkPipelineCache _cache,
+ const VkComputePipelineCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkPipeline* pPipeline)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ ANV_FROM_HANDLE(anv_pipeline_cache, cache, _cache);
+
+ if (cache == NULL)
+ cache = &device->default_pipeline_cache;
+
+ switch (device->info.gen) {
+ case 7:
+ if (device->info.is_haswell)
+ return gen75_compute_pipeline_create(_device, cache, pCreateInfo, pAllocator, pPipeline);
+ else
+ return gen7_compute_pipeline_create(_device, cache, pCreateInfo, pAllocator, pPipeline);
+ case 8:
+ return gen8_compute_pipeline_create(_device, cache, pCreateInfo, pAllocator, pPipeline);
+ case 9:
+ return gen9_compute_pipeline_create(_device, cache, pCreateInfo, pAllocator, pPipeline);
+ default:
+ unreachable("unsupported gen\n");
+ }
+}
+
+VkResult anv_CreateComputePipelines(
+ VkDevice _device,
+ VkPipelineCache pipelineCache,
+ uint32_t count,
+ const VkComputePipelineCreateInfo* pCreateInfos,
+ const VkAllocationCallbacks* pAllocator,
+ VkPipeline* pPipelines)
+{
+ VkResult result = VK_SUCCESS;
+
+ unsigned i = 0;
+ for (; i < count; i++) {
+ result = anv_compute_pipeline_create(_device, pipelineCache,
+ &pCreateInfos[i],
+ pAllocator, &pPipelines[i]);
+ if (result != VK_SUCCESS) {
+ for (unsigned j = 0; j < i; j++) {
+ anv_DestroyPipeline(_device, pPipelines[j], pAllocator);
+ }
+
+ return result;
+ }
+ }
+
+ return VK_SUCCESS;
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#pragma once
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <stdbool.h>
+#include <pthread.h>
+#include <assert.h>
+#include <stdint.h>
+#include <i915_drm.h>
+
+#ifdef HAVE_VALGRIND
+#include <valgrind.h>
+#include <memcheck.h>
+#define VG(x) x
+#define __gen_validate_value(x) VALGRIND_CHECK_MEM_IS_DEFINED(&(x), sizeof(x))
+#else
+#define VG(x)
+#endif
+
+#include "brw_device_info.h"
+#include "util/macros.h"
+#include "util/list.h"
+
+/* Pre-declarations needed for WSI entrypoints */
+struct wl_surface;
+struct wl_display;
+typedef struct xcb_connection_t xcb_connection_t;
+typedef uint32_t xcb_visualid_t;
+typedef uint32_t xcb_window_t;
+
+#define VK_USE_PLATFORM_XCB_KHR
+#define VK_USE_PLATFORM_WAYLAND_KHR
+
+#define VK_PROTOTYPES
+#include <vulkan/vulkan.h>
+#include <vulkan/vulkan_intel.h>
+
+#include "anv_entrypoints.h"
+#include "anv_gen_macros.h"
+#include "brw_context.h"
+#include "isl.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#define MAX_VBS 32
+#define MAX_SETS 8
+#define MAX_RTS 8
+#define MAX_VIEWPORTS 16
+#define MAX_SCISSORS 16
+#define MAX_PUSH_CONSTANTS_SIZE 128
+#define MAX_DYNAMIC_BUFFERS 16
+#define MAX_IMAGES 8
+
+#define ICD_LOADER_MAGIC 0x01CDC0DE
+
+typedef union _VK_LOADER_DATA {
+ uintptr_t loaderMagic;
+ void *loaderData;
+} VK_LOADER_DATA;
+
+#define anv_noreturn __attribute__((__noreturn__))
+#define anv_printflike(a, b) __attribute__((__format__(__printf__, a, b)))
+
+#define MIN(a, b) ((a) < (b) ? (a) : (b))
+#define MAX(a, b) ((a) > (b) ? (a) : (b))
+
+static inline uint32_t
+align_u32(uint32_t v, uint32_t a)
+{
+ assert(a != 0 && a == (a & -a));
+ return (v + a - 1) & ~(a - 1);
+}
+
+static inline uint64_t
+align_u64(uint64_t v, uint64_t a)
+{
+ assert(a != 0 && a == (a & -a));
+ return (v + a - 1) & ~(a - 1);
+}
+
+static inline int32_t
+align_i32(int32_t v, int32_t a)
+{
+ assert(a != 0 && a == (a & -a));
+ return (v + a - 1) & ~(a - 1);
+}
+
+/** Alignment must be a power of 2. */
+static inline bool
+anv_is_aligned(uintmax_t n, uintmax_t a)
+{
+ assert(a == (a & -a));
+ return (n & (a - 1)) == 0;
+}
+
+static inline uint32_t
+anv_minify(uint32_t n, uint32_t levels)
+{
+ if (unlikely(n == 0))
+ return 0;
+ else
+ return MAX(n >> levels, 1);
+}
+
+static inline float
+anv_clamp_f(float f, float min, float max)
+{
+ assert(min < max);
+
+ if (f > max)
+ return max;
+ else if (f < min)
+ return min;
+ else
+ return f;
+}
+
+static inline bool
+anv_clear_mask(uint32_t *inout_mask, uint32_t clear_mask)
+{
+ if (*inout_mask & clear_mask) {
+ *inout_mask &= ~clear_mask;
+ return true;
+ } else {
+ return false;
+ }
+}
+
+#define for_each_bit(b, dword) \
+ for (uint32_t __dword = (dword); \
+ (b) = __builtin_ffs(__dword) - 1, __dword; \
+ __dword &= ~(1 << (b)))
+
+#define typed_memcpy(dest, src, count) ({ \
+ static_assert(sizeof(*src) == sizeof(*dest), ""); \
+ memcpy((dest), (src), (count) * sizeof(*(src))); \
+})
+
+#define zero(x) (memset(&(x), 0, sizeof(x)))
+
+/* Define no kernel as 1, since that's an illegal offset for a kernel */
+#define NO_KERNEL 1
+
+struct anv_common {
+ VkStructureType sType;
+ const void* pNext;
+};
+
+/* Whenever we generate an error, pass it through this function. Useful for
+ * debugging, where we can break on it. Only call at error site, not when
+ * propagating errors. Might be useful to plug in a stack trace here.
+ */
+
+VkResult __vk_errorf(VkResult error, const char *file, int line, const char *format, ...);
+
+#ifdef DEBUG
+#define vk_error(error) __vk_errorf(error, __FILE__, __LINE__, NULL);
+#define vk_errorf(error, format, ...) __vk_errorf(error, __FILE__, __LINE__, format, ## __VA_ARGS__);
+#else
+#define vk_error(error) error
+#define vk_errorf(error, format, ...) error
+#endif
+
+void __anv_finishme(const char *file, int line, const char *format, ...)
+ anv_printflike(3, 4);
+void anv_loge(const char *format, ...) anv_printflike(1, 2);
+void anv_loge_v(const char *format, va_list va);
+
+/**
+ * Print a FINISHME message, including its source location.
+ */
+#define anv_finishme(format, ...) \
+ __anv_finishme(__FILE__, __LINE__, format, ##__VA_ARGS__);
+
+/* A non-fatal assert. Useful for debugging. */
+#ifdef DEBUG
+#define anv_assert(x) ({ \
+ if (unlikely(!(x))) \
+ fprintf(stderr, "%s:%d ASSERT: %s\n", __FILE__, __LINE__, #x); \
+})
+#else
+#define anv_assert(x)
+#endif
+
+/**
+ * If a block of code is annotated with anv_validate, then the block runs only
+ * in debug builds.
+ */
+#ifdef DEBUG
+#define anv_validate if (1)
+#else
+#define anv_validate if (0)
+#endif
+
+void anv_abortf(const char *format, ...) anv_noreturn anv_printflike(1, 2);
+void anv_abortfv(const char *format, va_list va) anv_noreturn;
+
+#define stub_return(v) \
+ do { \
+ anv_finishme("stub %s", __func__); \
+ return (v); \
+ } while (0)
+
+#define stub() \
+ do { \
+ anv_finishme("stub %s", __func__); \
+ return; \
+ } while (0)
+
+/**
+ * A dynamically growable, circular buffer. Elements are added at head and
+ * removed from tail. head and tail are free-running uint32_t indices and we
+ * only compute the modulo with size when accessing the array. This way,
+ * number of bytes in the queue is always head - tail, even in case of
+ * wraparound.
+ */
+
+struct anv_vector {
+ uint32_t head;
+ uint32_t tail;
+ uint32_t element_size;
+ uint32_t size;
+ void *data;
+};
+
+int anv_vector_init(struct anv_vector *queue, uint32_t element_size, uint32_t size);
+void *anv_vector_add(struct anv_vector *queue);
+void *anv_vector_remove(struct anv_vector *queue);
+
+static inline int
+anv_vector_length(struct anv_vector *queue)
+{
+ return (queue->head - queue->tail) / queue->element_size;
+}
+
+static inline void *
+anv_vector_head(struct anv_vector *vector)
+{
+ assert(vector->tail < vector->head);
+ return (void *)((char *)vector->data +
+ ((vector->head - vector->element_size) &
+ (vector->size - 1)));
+}
+
+static inline void *
+anv_vector_tail(struct anv_vector *vector)
+{
+ return (void *)((char *)vector->data + (vector->tail & (vector->size - 1)));
+}
+
+static inline void
+anv_vector_finish(struct anv_vector *queue)
+{
+ free(queue->data);
+}
+
+#define anv_vector_foreach(elem, queue) \
+ static_assert(__builtin_types_compatible_p(__typeof__(queue), struct anv_vector *), ""); \
+ for (uint32_t __anv_vector_offset = (queue)->tail; \
+ elem = (queue)->data + (__anv_vector_offset & ((queue)->size - 1)), __anv_vector_offset < (queue)->head; \
+ __anv_vector_offset += (queue)->element_size)
+
+struct anv_bo {
+ uint32_t gem_handle;
+
+ /* Index into the current validation list. This is used by the
+ * validation list building alrogithm to track which buffers are already
+ * in the validation list so that we can ensure uniqueness.
+ */
+ uint32_t index;
+
+ /* Last known offset. This value is provided by the kernel when we
+ * execbuf and is used as the presumed offset for the next bunch of
+ * relocations.
+ */
+ uint64_t offset;
+
+ uint64_t size;
+ void *map;
+};
+
+/* Represents a lock-free linked list of "free" things. This is used by
+ * both the block pool and the state pools. Unfortunately, in order to
+ * solve the ABA problem, we can't use a single uint32_t head.
+ */
+union anv_free_list {
+ struct {
+ int32_t offset;
+
+ /* A simple count that is incremented every time the head changes. */
+ uint32_t count;
+ };
+ uint64_t u64;
+};
+
+#define ANV_FREE_LIST_EMPTY ((union anv_free_list) { { 1, 0 } })
+
+struct anv_block_state {
+ union {
+ struct {
+ uint32_t next;
+ uint32_t end;
+ };
+ uint64_t u64;
+ };
+};
+
+struct anv_block_pool {
+ struct anv_device *device;
+
+ struct anv_bo bo;
+
+ /* The offset from the start of the bo to the "center" of the block
+ * pool. Pointers to allocated blocks are given by
+ * bo.map + center_bo_offset + offsets.
+ */
+ uint32_t center_bo_offset;
+
+ /* Current memory map of the block pool. This pointer may or may not
+ * point to the actual beginning of the block pool memory. If
+ * anv_block_pool_alloc_back has ever been called, then this pointer
+ * will point to the "center" position of the buffer and all offsets
+ * (negative or positive) given out by the block pool alloc functions
+ * will be valid relative to this pointer.
+ *
+ * In particular, map == bo.map + center_offset
+ */
+ void *map;
+ int fd;
+
+ /**
+ * Array of mmaps and gem handles owned by the block pool, reclaimed when
+ * the block pool is destroyed.
+ */
+ struct anv_vector mmap_cleanups;
+
+ uint32_t block_size;
+
+ union anv_free_list free_list;
+ struct anv_block_state state;
+
+ union anv_free_list back_free_list;
+ struct anv_block_state back_state;
+};
+
+/* Block pools are backed by a fixed-size 2GB memfd */
+#define BLOCK_POOL_MEMFD_SIZE (1ull << 32)
+
+/* The center of the block pool is also the middle of the memfd. This may
+ * change in the future if we decide differently for some reason.
+ */
+#define BLOCK_POOL_MEMFD_CENTER (BLOCK_POOL_MEMFD_SIZE / 2)
+
+static inline uint32_t
+anv_block_pool_size(struct anv_block_pool *pool)
+{
+ return pool->state.end + pool->back_state.end;
+}
+
+struct anv_state {
+ int32_t offset;
+ uint32_t alloc_size;
+ void *map;
+};
+
+struct anv_fixed_size_state_pool {
+ size_t state_size;
+ union anv_free_list free_list;
+ struct anv_block_state block;
+};
+
+#define ANV_MIN_STATE_SIZE_LOG2 6
+#define ANV_MAX_STATE_SIZE_LOG2 10
+
+#define ANV_STATE_BUCKETS (ANV_MAX_STATE_SIZE_LOG2 - ANV_MIN_STATE_SIZE_LOG2)
+
+struct anv_state_pool {
+ struct anv_block_pool *block_pool;
+ struct anv_fixed_size_state_pool buckets[ANV_STATE_BUCKETS];
+};
+
+struct anv_state_stream_block;
+
+struct anv_state_stream {
+ struct anv_block_pool *block_pool;
+
+ /* The current working block */
+ struct anv_state_stream_block *block;
+
+ /* Offset at which the current block starts */
+ uint32_t start;
+ /* Offset at which to allocate the next state */
+ uint32_t next;
+ /* Offset at which the current block ends */
+ uint32_t end;
+};
+
+#define CACHELINE_SIZE 64
+#define CACHELINE_MASK 63
+
+static void inline
+anv_state_clflush(struct anv_state state)
+{
+ /* state.map may not be cacheline aligned, so round down the start pointer
+ * to a cacheline boundary so we flush all pages that contain the state.
+ */
+ void *end = state.map + state.alloc_size;
+ void *p = (void *) (((uintptr_t) state.map) & ~CACHELINE_MASK);
+
+ __builtin_ia32_sfence();
+ while (p < end) {
+ __builtin_ia32_clflush(p);
+ p += CACHELINE_SIZE;
+ }
+}
+
+void anv_block_pool_init(struct anv_block_pool *pool,
+ struct anv_device *device, uint32_t block_size);
+void anv_block_pool_finish(struct anv_block_pool *pool);
+int32_t anv_block_pool_alloc(struct anv_block_pool *pool);
+int32_t anv_block_pool_alloc_back(struct anv_block_pool *pool);
+void anv_block_pool_free(struct anv_block_pool *pool, int32_t offset);
+void anv_state_pool_init(struct anv_state_pool *pool,
+ struct anv_block_pool *block_pool);
+void anv_state_pool_finish(struct anv_state_pool *pool);
+struct anv_state anv_state_pool_alloc(struct anv_state_pool *pool,
+ size_t state_size, size_t alignment);
+void anv_state_pool_free(struct anv_state_pool *pool, struct anv_state state);
+void anv_state_stream_init(struct anv_state_stream *stream,
+ struct anv_block_pool *block_pool);
+void anv_state_stream_finish(struct anv_state_stream *stream);
+struct anv_state anv_state_stream_alloc(struct anv_state_stream *stream,
+ uint32_t size, uint32_t alignment);
+
+/**
+ * Implements a pool of re-usable BOs. The interface is identical to that
+ * of block_pool except that each block is its own BO.
+ */
+struct anv_bo_pool {
+ struct anv_device *device;
+
+ uint32_t bo_size;
+
+ void *free_list;
+};
+
+void anv_bo_pool_init(struct anv_bo_pool *pool,
+ struct anv_device *device, uint32_t block_size);
+void anv_bo_pool_finish(struct anv_bo_pool *pool);
+VkResult anv_bo_pool_alloc(struct anv_bo_pool *pool, struct anv_bo *bo);
+void anv_bo_pool_free(struct anv_bo_pool *pool, const struct anv_bo *bo);
+
+
+void *anv_resolve_entrypoint(uint32_t index);
+
+extern struct anv_dispatch_table dtable;
+
+#define ANV_CALL(func) ({ \
+ if (dtable.func == NULL) { \
+ size_t idx = offsetof(struct anv_dispatch_table, func) / sizeof(void *); \
+ dtable.entrypoints[idx] = anv_resolve_entrypoint(idx); \
+ } \
+ dtable.func; \
+})
+
+static inline void *
+anv_alloc(const VkAllocationCallbacks *alloc,
+ size_t size, size_t align,
+ VkSystemAllocationScope scope)
+{
+ return alloc->pfnAllocation(alloc->pUserData, size, align, scope);
+}
+
+static inline void *
+anv_realloc(const VkAllocationCallbacks *alloc,
+ void *ptr, size_t size, size_t align,
+ VkSystemAllocationScope scope)
+{
+ return alloc->pfnReallocation(alloc->pUserData, ptr, size, align, scope);
+}
+
+static inline void
+anv_free(const VkAllocationCallbacks *alloc, void *data)
+{
+ alloc->pfnFree(alloc->pUserData, data);
+}
+
+static inline void *
+anv_alloc2(const VkAllocationCallbacks *parent_alloc,
+ const VkAllocationCallbacks *alloc,
+ size_t size, size_t align,
+ VkSystemAllocationScope scope)
+{
+ if (alloc)
+ return anv_alloc(alloc, size, align, scope);
+ else
+ return anv_alloc(parent_alloc, size, align, scope);
+}
+
+static inline void
+anv_free2(const VkAllocationCallbacks *parent_alloc,
+ const VkAllocationCallbacks *alloc,
+ void *data)
+{
+ if (alloc)
+ anv_free(alloc, data);
+ else
+ anv_free(parent_alloc, data);
+}
+
+struct anv_physical_device {
+ VK_LOADER_DATA _loader_data;
+
+ struct anv_instance * instance;
+ uint32_t chipset_id;
+ const char * path;
+ const char * name;
+ const struct brw_device_info * info;
+ uint64_t aperture_size;
+ struct brw_compiler * compiler;
+ struct isl_device isl_dev;
+};
+
+struct anv_instance {
+ VK_LOADER_DATA _loader_data;
+
+ VkAllocationCallbacks alloc;
+
+ uint32_t apiVersion;
+ int physicalDeviceCount;
+ struct anv_physical_device physicalDevice;
+
+ void * wayland_wsi;
+};
+
+VkResult anv_init_wsi(struct anv_instance *instance);
+void anv_finish_wsi(struct anv_instance *instance);
+
+struct anv_meta_state {
+ VkAllocationCallbacks alloc;
+
+ struct {
+ /**
+ * Pipeline N is used to clear color attachment N of the current
+ * subpass.
+ *
+ * HACK: We use one pipeline per color attachment to work around the
+ * compiler's inability to dynamically set the render target index of
+ * the render target write message.
+ */
+ struct anv_pipeline *color_pipelines[MAX_RTS];
+
+ struct anv_pipeline *depth_only_pipeline;
+ struct anv_pipeline *stencil_only_pipeline;
+ struct anv_pipeline *depthstencil_pipeline;
+ } clear;
+
+ struct {
+ VkRenderPass render_pass;
+
+ /** Pipeline that blits from a 1D image. */
+ VkPipeline pipeline_1d_src;
+
+ /** Pipeline that blits from a 2D image. */
+ VkPipeline pipeline_2d_src;
+
+ /** Pipeline that blits from a 3D image. */
+ VkPipeline pipeline_3d_src;
+
+ VkPipelineLayout pipeline_layout;
+ VkDescriptorSetLayout ds_layout;
+ } blit;
+};
+
+struct anv_queue {
+ VK_LOADER_DATA _loader_data;
+
+ struct anv_device * device;
+
+ struct anv_state_pool * pool;
+};
+
+struct anv_pipeline_cache {
+ struct anv_device * device;
+ struct anv_state_stream program_stream;
+ pthread_mutex_t mutex;
+};
+
+void anv_pipeline_cache_init(struct anv_pipeline_cache *cache,
+ struct anv_device *device);
+void anv_pipeline_cache_finish(struct anv_pipeline_cache *cache);
+
+struct anv_device {
+ VK_LOADER_DATA _loader_data;
+
+ VkAllocationCallbacks alloc;
+
+ struct anv_instance * instance;
+ uint32_t chipset_id;
+ struct brw_device_info info;
+ struct isl_device isl_dev;
+ int context_id;
+ int fd;
+
+ struct anv_bo_pool batch_bo_pool;
+
+ struct anv_block_pool dynamic_state_block_pool;
+ struct anv_state_pool dynamic_state_pool;
+
+ struct anv_block_pool instruction_block_pool;
+ struct anv_pipeline_cache default_pipeline_cache;
+
+ struct anv_block_pool surface_state_block_pool;
+ struct anv_state_pool surface_state_pool;
+
+ struct anv_bo workaround_bo;
+
+ struct anv_meta_state meta_state;
+
+ struct anv_state border_colors;
+
+ struct anv_queue queue;
+
+ struct anv_block_pool scratch_block_pool;
+
+ pthread_mutex_t mutex;
+};
+
+void* anv_gem_mmap(struct anv_device *device,
+ uint32_t gem_handle, uint64_t offset, uint64_t size, uint32_t flags);
+void anv_gem_munmap(void *p, uint64_t size);
+uint32_t anv_gem_create(struct anv_device *device, size_t size);
+void anv_gem_close(struct anv_device *device, uint32_t gem_handle);
+uint32_t anv_gem_userptr(struct anv_device *device, void *mem, size_t size);
+int anv_gem_wait(struct anv_device *device, uint32_t gem_handle, int64_t *timeout_ns);
+int anv_gem_execbuffer(struct anv_device *device,
+ struct drm_i915_gem_execbuffer2 *execbuf);
+int anv_gem_set_tiling(struct anv_device *device, uint32_t gem_handle,
+ uint32_t stride, uint32_t tiling);
+int anv_gem_create_context(struct anv_device *device);
+int anv_gem_destroy_context(struct anv_device *device, int context);
+int anv_gem_get_param(int fd, uint32_t param);
+bool anv_gem_get_bit6_swizzle(int fd, uint32_t tiling);
+int anv_gem_get_aperture(int fd, uint64_t *size);
+int anv_gem_handle_to_fd(struct anv_device *device, uint32_t gem_handle);
+uint32_t anv_gem_fd_to_handle(struct anv_device *device, int fd);
+int anv_gem_set_caching(struct anv_device *device, uint32_t gem_handle, uint32_t caching);
+int anv_gem_set_domain(struct anv_device *device, uint32_t gem_handle,
+ uint32_t read_domains, uint32_t write_domain);
+
+VkResult anv_bo_init_new(struct anv_bo *bo, struct anv_device *device, uint64_t size);
+
+struct anv_reloc_list {
+ size_t num_relocs;
+ size_t array_length;
+ struct drm_i915_gem_relocation_entry * relocs;
+ struct anv_bo ** reloc_bos;
+};
+
+VkResult anv_reloc_list_init(struct anv_reloc_list *list,
+ const VkAllocationCallbacks *alloc);
+void anv_reloc_list_finish(struct anv_reloc_list *list,
+ const VkAllocationCallbacks *alloc);
+
+uint64_t anv_reloc_list_add(struct anv_reloc_list *list,
+ const VkAllocationCallbacks *alloc,
+ uint32_t offset, struct anv_bo *target_bo,
+ uint32_t delta);
+
+struct anv_batch_bo {
+ /* Link in the anv_cmd_buffer.owned_batch_bos list */
+ struct list_head link;
+
+ struct anv_bo bo;
+
+ /* Bytes actually consumed in this batch BO */
+ size_t length;
+
+ /* Last seen surface state block pool bo offset */
+ uint32_t last_ss_pool_bo_offset;
+
+ struct anv_reloc_list relocs;
+};
+
+struct anv_batch {
+ const VkAllocationCallbacks * alloc;
+
+ void * start;
+ void * end;
+ void * next;
+
+ struct anv_reloc_list * relocs;
+
+ /* This callback is called (with the associated user data) in the event
+ * that the batch runs out of space.
+ */
+ VkResult (*extend_cb)(struct anv_batch *, void *);
+ void * user_data;
+};
+
+void *anv_batch_emit_dwords(struct anv_batch *batch, int num_dwords);
+void anv_batch_emit_batch(struct anv_batch *batch, struct anv_batch *other);
+uint64_t anv_batch_emit_reloc(struct anv_batch *batch,
+ void *location, struct anv_bo *bo, uint32_t offset);
+
+struct anv_address {
+ struct anv_bo *bo;
+ uint32_t offset;
+};
+
+#define __gen_address_type struct anv_address
+#define __gen_user_data struct anv_batch
+
+static inline uint64_t
+__gen_combine_address(struct anv_batch *batch, void *location,
+ const struct anv_address address, uint32_t delta)
+{
+ if (address.bo == NULL) {
+ return address.offset + delta;
+ } else {
+ assert(batch->start <= location && location < batch->end);
+
+ return anv_batch_emit_reloc(batch, location, address.bo, address.offset + delta);
+ }
+}
+
+/* Wrapper macros needed to work around preprocessor argument issues. In
+ * particular, arguments don't get pre-evaluated if they are concatenated.
+ * This means that, if you pass GENX(3DSTATE_PS) into the emit macro, the
+ * GENX macro won't get evaluated if the emit macro contains "cmd ## foo".
+ * We can work around this easily enough with these helpers.
+ */
+#define __anv_cmd_length(cmd) cmd ## _length
+#define __anv_cmd_length_bias(cmd) cmd ## _length_bias
+#define __anv_cmd_header(cmd) cmd ## _header
+#define __anv_cmd_pack(cmd) cmd ## _pack
+
+#define anv_batch_emit(batch, cmd, ...) do { \
+ void *__dst = anv_batch_emit_dwords(batch, __anv_cmd_length(cmd)); \
+ struct cmd __template = { \
+ __anv_cmd_header(cmd), \
+ __VA_ARGS__ \
+ }; \
+ __anv_cmd_pack(cmd)(batch, __dst, &__template); \
+ VG(VALGRIND_CHECK_MEM_IS_DEFINED(__dst, __anv_cmd_length(cmd) * 4)); \
+ } while (0)
+
+#define anv_batch_emitn(batch, n, cmd, ...) ({ \
+ void *__dst = anv_batch_emit_dwords(batch, n); \
+ struct cmd __template = { \
+ __anv_cmd_header(cmd), \
+ .DwordLength = n - __anv_cmd_length_bias(cmd), \
+ __VA_ARGS__ \
+ }; \
+ __anv_cmd_pack(cmd)(batch, __dst, &__template); \
+ __dst; \
+ })
+
+#define anv_batch_emit_merge(batch, dwords0, dwords1) \
+ do { \
+ uint32_t *dw; \
+ \
+ static_assert(ARRAY_SIZE(dwords0) == ARRAY_SIZE(dwords1), "mismatch merge"); \
+ dw = anv_batch_emit_dwords((batch), ARRAY_SIZE(dwords0)); \
+ for (uint32_t i = 0; i < ARRAY_SIZE(dwords0); i++) \
+ dw[i] = (dwords0)[i] | (dwords1)[i]; \
+ VG(VALGRIND_CHECK_MEM_IS_DEFINED(dw, ARRAY_SIZE(dwords0) * 4));\
+ } while (0)
+
+#define anv_state_pool_emit(pool, cmd, align, ...) ({ \
+ const uint32_t __size = __anv_cmd_length(cmd) * 4; \
+ struct anv_state __state = \
+ anv_state_pool_alloc((pool), __size, align); \
+ struct cmd __template = { \
+ __VA_ARGS__ \
+ }; \
+ __anv_cmd_pack(cmd)(NULL, __state.map, &__template); \
+ VG(VALGRIND_CHECK_MEM_IS_DEFINED(__state.map, __anv_cmd_length(cmd) * 4)); \
+ if (!(pool)->block_pool->device->info.has_llc) \
+ anv_state_clflush(__state); \
+ __state; \
+ })
+
+#define GEN7_MOCS (struct GEN7_MEMORY_OBJECT_CONTROL_STATE) { \
+ .GraphicsDataTypeGFDT = 0, \
+ .LLCCacheabilityControlLLCCC = 0, \
+ .L3CacheabilityControlL3CC = 1, \
+}
+
+#define GEN75_MOCS (struct GEN75_MEMORY_OBJECT_CONTROL_STATE) { \
+ .LLCeLLCCacheabilityControlLLCCC = 0, \
+ .L3CacheabilityControlL3CC = 1, \
+}
+
+#define GEN8_MOCS { \
+ .MemoryTypeLLCeLLCCacheabilityControl = WB, \
+ .TargetCache = L3DefertoPATforLLCeLLCselection, \
+ .AgeforQUADLRU = 0 \
+ }
+
+/* Skylake: MOCS is now an index into an array of 62 different caching
+ * configurations programmed by the kernel.
+ */
+
+#define GEN9_MOCS { \
+ /* TC=LLC/eLLC, LeCC=WB, LRUM=3, L3CC=WB */ \
+ .IndextoMOCSTables = 2 \
+ }
+
+#define GEN9_MOCS_PTE { \
+ /* TC=LLC/eLLC, LeCC=WB, LRUM=3, L3CC=WB */ \
+ .IndextoMOCSTables = 1 \
+ }
+
+struct anv_device_memory {
+ struct anv_bo bo;
+ uint32_t type_index;
+ VkDeviceSize map_size;
+ void * map;
+};
+
+/**
+ * Header for Vertex URB Entry (VUE)
+ */
+struct anv_vue_header {
+ uint32_t Reserved;
+ uint32_t RTAIndex; /* RenderTargetArrayIndex */
+ uint32_t ViewportIndex;
+ float PointWidth;
+};
+
+struct anv_descriptor_set_binding_layout {
+ /* Number of array elements in this binding */
+ uint16_t array_size;
+
+ /* Index into the flattend descriptor set */
+ uint16_t descriptor_index;
+
+ /* Index into the dynamic state array for a dynamic buffer */
+ int16_t dynamic_offset_index;
+
+ /* Index into the descriptor set buffer views */
+ int16_t buffer_index;
+
+ struct {
+ /* Index into the binding table for the associated surface */
+ int16_t surface_index;
+
+ /* Index into the sampler table for the associated sampler */
+ int16_t sampler_index;
+
+ /* Index into the image table for the associated image */
+ int16_t image_index;
+ } stage[MESA_SHADER_STAGES];
+
+ /* Immutable samplers (or NULL if no immutable samplers) */
+ struct anv_sampler **immutable_samplers;
+};
+
+struct anv_descriptor_set_layout {
+ /* Number of bindings in this descriptor set */
+ uint16_t binding_count;
+
+ /* Total size of the descriptor set with room for all array entries */
+ uint16_t size;
+
+ /* Shader stages affected by this descriptor set */
+ uint16_t shader_stages;
+
+ /* Number of buffers in this descriptor set */
+ uint16_t buffer_count;
+
+ /* Number of dynamic offsets used by this descriptor set */
+ uint16_t dynamic_offset_count;
+
+ /* Bindings in this descriptor set */
+ struct anv_descriptor_set_binding_layout binding[0];
+};
+
+struct anv_descriptor {
+ VkDescriptorType type;
+
+ union {
+ struct {
+ union {
+ struct anv_image_view *image_view;
+ };
+ struct anv_sampler *sampler;
+ };
+
+ struct anv_buffer_view *buffer_view;
+ };
+};
+
+struct anv_descriptor_set {
+ const struct anv_descriptor_set_layout *layout;
+ struct anv_buffer_view *buffer_views;
+ struct anv_descriptor descriptors[0];
+};
+
+VkResult
+anv_descriptor_set_create(struct anv_device *device,
+ const struct anv_descriptor_set_layout *layout,
+ struct anv_descriptor_set **out_set);
+
+void
+anv_descriptor_set_destroy(struct anv_device *device,
+ struct anv_descriptor_set *set);
+
+struct anv_pipeline_binding {
+ /* The descriptor set this surface corresponds to */
+ uint16_t set;
+
+ /* Offset into the descriptor set */
+ uint16_t offset;
+};
+
+struct anv_pipeline_layout {
+ struct {
+ struct anv_descriptor_set_layout *layout;
+ uint32_t dynamic_offset_start;
+ struct {
+ uint32_t surface_start;
+ uint32_t sampler_start;
+ uint32_t image_start;
+ } stage[MESA_SHADER_STAGES];
+ } set[MAX_SETS];
+
+ uint32_t num_sets;
+
+ struct {
+ bool has_dynamic_offsets;
+ uint32_t surface_count;
+ struct anv_pipeline_binding *surface_to_descriptor;
+ uint32_t sampler_count;
+ struct anv_pipeline_binding *sampler_to_descriptor;
+ uint32_t image_count;
+ } stage[MESA_SHADER_STAGES];
+
+ struct anv_pipeline_binding entries[0];
+};
+
+struct anv_buffer {
+ struct anv_device * device;
+ VkDeviceSize size;
+
+ VkBufferUsageFlags usage;
+
+ /* Set when bound */
+ struct anv_bo * bo;
+ VkDeviceSize offset;
+};
+
+enum anv_cmd_dirty_bits {
+ ANV_CMD_DIRTY_DYNAMIC_VIEWPORT = 1 << 0, /* VK_DYNAMIC_STATE_VIEWPORT */
+ ANV_CMD_DIRTY_DYNAMIC_SCISSOR = 1 << 1, /* VK_DYNAMIC_STATE_SCISSOR */
+ ANV_CMD_DIRTY_DYNAMIC_LINE_WIDTH = 1 << 2, /* VK_DYNAMIC_STATE_LINE_WIDTH */
+ ANV_CMD_DIRTY_DYNAMIC_DEPTH_BIAS = 1 << 3, /* VK_DYNAMIC_STATE_DEPTH_BIAS */
+ ANV_CMD_DIRTY_DYNAMIC_BLEND_CONSTANTS = 1 << 4, /* VK_DYNAMIC_STATE_BLEND_CONSTANTS */
+ ANV_CMD_DIRTY_DYNAMIC_DEPTH_BOUNDS = 1 << 5, /* VK_DYNAMIC_STATE_DEPTH_BOUNDS */
+ ANV_CMD_DIRTY_DYNAMIC_STENCIL_COMPARE_MASK = 1 << 6, /* VK_DYNAMIC_STATE_STENCIL_COMPARE_MASK */
+ ANV_CMD_DIRTY_DYNAMIC_STENCIL_WRITE_MASK = 1 << 7, /* VK_DYNAMIC_STATE_STENCIL_WRITE_MASK */
+ ANV_CMD_DIRTY_DYNAMIC_STENCIL_REFERENCE = 1 << 8, /* VK_DYNAMIC_STATE_STENCIL_REFERENCE */
+ ANV_CMD_DIRTY_DYNAMIC_ALL = (1 << 9) - 1,
+ ANV_CMD_DIRTY_PIPELINE = 1 << 9,
+ ANV_CMD_DIRTY_INDEX_BUFFER = 1 << 10,
+ ANV_CMD_DIRTY_RENDER_TARGETS = 1 << 11,
+};
+typedef uint32_t anv_cmd_dirty_mask_t;
+
+struct anv_vertex_binding {
+ struct anv_buffer * buffer;
+ VkDeviceSize offset;
+};
+
+struct anv_push_constants {
+ /* Current allocated size of this push constants data structure.
+ * Because a decent chunk of it may not be used (images on SKL, for
+ * instance), we won't actually allocate the entire structure up-front.
+ */
+ uint32_t size;
+
+ /* Push constant data provided by the client through vkPushConstants */
+ uint8_t client_data[MAX_PUSH_CONSTANTS_SIZE];
+
+ /* Our hardware only provides zero-based vertex and instance id so, in
+ * order to satisfy the vulkan requirements, we may have to push one or
+ * both of these into the shader.
+ */
+ uint32_t base_vertex;
+ uint32_t base_instance;
+
+ /* Offsets and ranges for dynamically bound buffers */
+ struct {
+ uint32_t offset;
+ uint32_t range;
+ } dynamic[MAX_DYNAMIC_BUFFERS];
+
+ /* Image data for image_load_store on pre-SKL */
+ struct brw_image_param images[MAX_IMAGES];
+};
+
+struct anv_dynamic_state {
+ struct {
+ uint32_t count;
+ VkViewport viewports[MAX_VIEWPORTS];
+ } viewport;
+
+ struct {
+ uint32_t count;
+ VkRect2D scissors[MAX_SCISSORS];
+ } scissor;
+
+ float line_width;
+
+ struct {
+ float bias;
+ float clamp;
+ float slope;
+ } depth_bias;
+
+ float blend_constants[4];
+
+ struct {
+ float min;
+ float max;
+ } depth_bounds;
+
+ struct {
+ uint32_t front;
+ uint32_t back;
+ } stencil_compare_mask;
+
+ struct {
+ uint32_t front;
+ uint32_t back;
+ } stencil_write_mask;
+
+ struct {
+ uint32_t front;
+ uint32_t back;
+ } stencil_reference;
+};
+
+extern const struct anv_dynamic_state default_dynamic_state;
+
+void anv_dynamic_state_copy(struct anv_dynamic_state *dest,
+ const struct anv_dynamic_state *src,
+ uint32_t copy_mask);
+
+/**
+ * Attachment state when recording a renderpass instance.
+ *
+ * The clear value is valid only if there exists a pending clear.
+ */
+struct anv_attachment_state {
+ VkImageAspectFlags pending_clear_aspects;
+ VkClearValue clear_value;
+};
+
+/** State required while building cmd buffer */
+struct anv_cmd_state {
+ /* PIPELINE_SELECT.PipelineSelection */
+ uint32_t current_pipeline;
+ uint32_t current_l3_config;
+ uint32_t vb_dirty;
+ anv_cmd_dirty_mask_t dirty;
+ anv_cmd_dirty_mask_t compute_dirty;
+ uint32_t num_workgroups_offset;
+ struct anv_bo *num_workgroups_bo;
+ VkShaderStageFlags descriptors_dirty;
+ VkShaderStageFlags push_constants_dirty;
+ uint32_t scratch_size;
+ struct anv_pipeline * pipeline;
+ struct anv_pipeline * compute_pipeline;
+ struct anv_framebuffer * framebuffer;
+ struct anv_render_pass * pass;
+ struct anv_subpass * subpass;
+ uint32_t restart_index;
+ struct anv_vertex_binding vertex_bindings[MAX_VBS];
+ struct anv_descriptor_set * descriptors[MAX_SETS];
+ struct anv_push_constants * push_constants[MESA_SHADER_STAGES];
+ struct anv_state binding_tables[MESA_SHADER_STAGES];
+ struct anv_state samplers[MESA_SHADER_STAGES];
+ struct anv_dynamic_state dynamic;
+ bool need_query_wa;
+
+ /**
+ * Array length is anv_cmd_state::pass::attachment_count. Array content is
+ * valid only when recording a render pass instance.
+ */
+ struct anv_attachment_state * attachments;
+
+ struct {
+ struct anv_buffer * index_buffer;
+ uint32_t index_type; /**< 3DSTATE_INDEX_BUFFER.IndexFormat */
+ uint32_t index_offset;
+ } gen7;
+};
+
+struct anv_cmd_pool {
+ VkAllocationCallbacks alloc;
+ struct list_head cmd_buffers;
+};
+
+#define ANV_CMD_BUFFER_BATCH_SIZE 8192
+
+enum anv_cmd_buffer_exec_mode {
+ ANV_CMD_BUFFER_EXEC_MODE_PRIMARY,
+ ANV_CMD_BUFFER_EXEC_MODE_EMIT,
+ ANV_CMD_BUFFER_EXEC_MODE_CHAIN,
+ ANV_CMD_BUFFER_EXEC_MODE_COPY_AND_CHAIN,
+};
+
+struct anv_cmd_buffer {
+ VK_LOADER_DATA _loader_data;
+
+ struct anv_device * device;
+
+ struct anv_cmd_pool * pool;
+ struct list_head pool_link;
+
+ struct anv_batch batch;
+
+ /* Fields required for the actual chain of anv_batch_bo's.
+ *
+ * These fields are initialized by anv_cmd_buffer_init_batch_bo_chain().
+ */
+ struct list_head batch_bos;
+ enum anv_cmd_buffer_exec_mode exec_mode;
+
+ /* A vector of anv_batch_bo pointers for every batch or surface buffer
+ * referenced by this command buffer
+ *
+ * initialized by anv_cmd_buffer_init_batch_bo_chain()
+ */
+ struct anv_vector seen_bbos;
+
+ /* A vector of int32_t's for every block of binding tables.
+ *
+ * initialized by anv_cmd_buffer_init_batch_bo_chain()
+ */
+ struct anv_vector bt_blocks;
+ uint32_t bt_next;
+ struct anv_reloc_list surface_relocs;
+
+ /* Information needed for execbuf
+ *
+ * These fields are generated by anv_cmd_buffer_prepare_execbuf().
+ */
+ struct {
+ struct drm_i915_gem_execbuffer2 execbuf;
+
+ struct drm_i915_gem_exec_object2 * objects;
+ uint32_t bo_count;
+ struct anv_bo ** bos;
+
+ /* Allocated length of the 'objects' and 'bos' arrays */
+ uint32_t array_length;
+
+ bool need_reloc;
+ } execbuf2;
+
+ /* Serial for tracking buffer completion */
+ uint32_t serial;
+
+ /* Stream objects for storing temporary data */
+ struct anv_state_stream surface_state_stream;
+ struct anv_state_stream dynamic_state_stream;
+
+ VkCommandBufferUsageFlags usage_flags;
+ VkCommandBufferLevel level;
+
+ struct anv_cmd_state state;
+};
+
+VkResult anv_cmd_buffer_init_batch_bo_chain(struct anv_cmd_buffer *cmd_buffer);
+void anv_cmd_buffer_fini_batch_bo_chain(struct anv_cmd_buffer *cmd_buffer);
+void anv_cmd_buffer_reset_batch_bo_chain(struct anv_cmd_buffer *cmd_buffer);
+void anv_cmd_buffer_end_batch_buffer(struct anv_cmd_buffer *cmd_buffer);
+void anv_cmd_buffer_add_secondary(struct anv_cmd_buffer *primary,
+ struct anv_cmd_buffer *secondary);
+void anv_cmd_buffer_prepare_execbuf(struct anv_cmd_buffer *cmd_buffer);
+
+VkResult anv_cmd_buffer_emit_binding_table(struct anv_cmd_buffer *cmd_buffer,
+ unsigned stage, struct anv_state *bt_state);
+VkResult anv_cmd_buffer_emit_samplers(struct anv_cmd_buffer *cmd_buffer,
+ unsigned stage, struct anv_state *state);
+uint32_t gen7_cmd_buffer_flush_descriptor_sets(struct anv_cmd_buffer *cmd_buffer);
+void gen7_cmd_buffer_emit_descriptor_pointers(struct anv_cmd_buffer *cmd_buffer,
+ uint32_t stages);
+
+struct anv_state anv_cmd_buffer_emit_dynamic(struct anv_cmd_buffer *cmd_buffer,
+ const void *data, uint32_t size, uint32_t alignment);
+struct anv_state anv_cmd_buffer_merge_dynamic(struct anv_cmd_buffer *cmd_buffer,
+ uint32_t *a, uint32_t *b,
+ uint32_t dwords, uint32_t alignment);
+
+struct anv_address
+anv_cmd_buffer_surface_base_address(struct anv_cmd_buffer *cmd_buffer);
+struct anv_state
+anv_cmd_buffer_alloc_binding_table(struct anv_cmd_buffer *cmd_buffer,
+ uint32_t entries, uint32_t *state_offset);
+struct anv_state
+anv_cmd_buffer_alloc_surface_state(struct anv_cmd_buffer *cmd_buffer);
+struct anv_state
+anv_cmd_buffer_alloc_dynamic_state(struct anv_cmd_buffer *cmd_buffer,
+ uint32_t size, uint32_t alignment);
+
+VkResult
+anv_cmd_buffer_new_binding_table_block(struct anv_cmd_buffer *cmd_buffer);
+
+void gen8_cmd_buffer_emit_viewport(struct anv_cmd_buffer *cmd_buffer);
+void gen7_cmd_buffer_emit_scissor(struct anv_cmd_buffer *cmd_buffer);
+
+void gen7_cmd_buffer_emit_state_base_address(struct anv_cmd_buffer *cmd_buffer);
+void gen75_cmd_buffer_emit_state_base_address(struct anv_cmd_buffer *cmd_buffer);
+void gen8_cmd_buffer_emit_state_base_address(struct anv_cmd_buffer *cmd_buffer);
+void gen9_cmd_buffer_emit_state_base_address(struct anv_cmd_buffer *cmd_buffer);
+
+void anv_cmd_buffer_emit_state_base_address(struct anv_cmd_buffer *cmd_buffer);
+
+void anv_cmd_state_setup_attachments(struct anv_cmd_buffer *cmd_buffer,
+ const VkRenderPassBeginInfo *info);
+
+void gen7_cmd_buffer_set_subpass(struct anv_cmd_buffer *cmd_buffer,
+ struct anv_subpass *subpass);
+void gen8_cmd_buffer_set_subpass(struct anv_cmd_buffer *cmd_buffer,
+ struct anv_subpass *subpass);
+void gen9_cmd_buffer_set_subpass(struct anv_cmd_buffer *cmd_buffer,
+ struct anv_subpass *subpass);
+void anv_cmd_buffer_set_subpass(struct anv_cmd_buffer *cmd_buffer,
+ struct anv_subpass *subpass);
+
+struct anv_state
+anv_cmd_buffer_push_constants(struct anv_cmd_buffer *cmd_buffer,
+ gl_shader_stage stage);
+struct anv_state
+anv_cmd_buffer_cs_push_constants(struct anv_cmd_buffer *cmd_buffer);
+
+void anv_cmd_buffer_clear_subpass(struct anv_cmd_buffer *cmd_buffer);
+
+const struct anv_image_view *
+anv_cmd_buffer_get_depth_stencil_view(const struct anv_cmd_buffer *cmd_buffer);
+
+void anv_cmd_buffer_dump(struct anv_cmd_buffer *cmd_buffer);
+
+struct anv_fence {
+ struct anv_bo bo;
+ struct drm_i915_gem_execbuffer2 execbuf;
+ struct drm_i915_gem_exec_object2 exec2_objects[1];
+ bool ready;
+};
+
+struct anv_event {
+ uint32_t semaphore;
+ struct anv_state state;
+};
+
+struct nir_shader;
+
+struct anv_shader_module {
+ struct nir_shader * nir;
+
+ uint32_t size;
+ char data[0];
+};
+
+static inline gl_shader_stage
+vk_to_mesa_shader_stage(VkShaderStageFlagBits vk_stage)
+{
+ assert(__builtin_popcount(vk_stage) == 1);
+ return ffs(vk_stage) - 1;
+}
+
+static inline VkShaderStageFlagBits
+mesa_to_vk_shader_stage(gl_shader_stage mesa_stage)
+{
+ return (1 << mesa_stage);
+}
+
+#define ANV_STAGE_MASK ((1 << MESA_SHADER_STAGES) - 1)
+
+#define anv_foreach_stage(stage, stage_bits) \
+ for (gl_shader_stage stage, \
+ __tmp = (gl_shader_stage)((stage_bits) & ANV_STAGE_MASK); \
+ stage = __builtin_ffs(__tmp) - 1, __tmp; \
+ __tmp &= ~(1 << (stage)))
+
+struct anv_pipeline {
+ struct anv_device * device;
+ struct anv_batch batch;
+ uint32_t batch_data[512];
+ struct anv_reloc_list batch_relocs;
+ uint32_t dynamic_state_mask;
+ struct anv_dynamic_state dynamic_state;
+
+ struct anv_pipeline_layout * layout;
+ bool use_repclear;
+
+ struct brw_vs_prog_data vs_prog_data;
+ struct brw_wm_prog_data wm_prog_data;
+ struct brw_gs_prog_data gs_prog_data;
+ struct brw_cs_prog_data cs_prog_data;
+ bool writes_point_size;
+ struct brw_stage_prog_data * prog_data[MESA_SHADER_STAGES];
+ uint32_t scratch_start[MESA_SHADER_STAGES];
+ uint32_t total_scratch;
+ struct {
+ uint32_t vs_start;
+ uint32_t vs_size;
+ uint32_t nr_vs_entries;
+ uint32_t gs_start;
+ uint32_t gs_size;
+ uint32_t nr_gs_entries;
+ } urb;
+
+ VkShaderStageFlags active_stages;
+ struct anv_state blend_state;
+ uint32_t vs_simd8;
+ uint32_t vs_vec4;
+ uint32_t ps_simd8;
+ uint32_t ps_simd16;
+ uint32_t ps_ksp0;
+ uint32_t ps_ksp2;
+ uint32_t ps_grf_start0;
+ uint32_t ps_grf_start2;
+ uint32_t gs_kernel;
+ uint32_t gs_vertex_count;
+ uint32_t cs_simd;
+
+ uint32_t vb_used;
+ uint32_t binding_stride[MAX_VBS];
+ bool instancing_enable[MAX_VBS];
+ bool primitive_restart;
+ uint32_t topology;
+
+ uint32_t cs_thread_width_max;
+ uint32_t cs_right_mask;
+
+ struct {
+ uint32_t sf[7];
+ uint32_t depth_stencil_state[3];
+ } gen7;
+
+ struct {
+ uint32_t sf[4];
+ uint32_t raster[5];
+ uint32_t wm_depth_stencil[3];
+ } gen8;
+
+ struct {
+ uint32_t wm_depth_stencil[4];
+ } gen9;
+};
+
+struct anv_graphics_pipeline_create_info {
+ /**
+ * If non-negative, overrides the color attachment count of the pipeline's
+ * subpass.
+ */
+ int8_t color_attachment_count;
+
+ bool use_repclear;
+ bool disable_viewport;
+ bool disable_scissor;
+ bool disable_vs;
+ bool use_rectlist;
+};
+
+VkResult
+anv_pipeline_init(struct anv_pipeline *pipeline, struct anv_device *device,
+ struct anv_pipeline_cache *cache,
+ const VkGraphicsPipelineCreateInfo *pCreateInfo,
+ const struct anv_graphics_pipeline_create_info *extra,
+ const VkAllocationCallbacks *alloc);
+
+VkResult
+anv_pipeline_compile_cs(struct anv_pipeline *pipeline,
+ struct anv_pipeline_cache *cache,
+ const VkComputePipelineCreateInfo *info,
+ struct anv_shader_module *module,
+ const char *entrypoint,
+ const VkSpecializationInfo *spec_info);
+
+VkResult
+anv_graphics_pipeline_create(VkDevice device,
+ VkPipelineCache cache,
+ const VkGraphicsPipelineCreateInfo *pCreateInfo,
+ const struct anv_graphics_pipeline_create_info *extra,
+ const VkAllocationCallbacks *alloc,
+ VkPipeline *pPipeline);
+
+VkResult
+gen7_graphics_pipeline_create(VkDevice _device,
+ struct anv_pipeline_cache *cache,
+ const VkGraphicsPipelineCreateInfo *pCreateInfo,
+ const struct anv_graphics_pipeline_create_info *extra,
+ const VkAllocationCallbacks *alloc,
+ VkPipeline *pPipeline);
+
+VkResult
+gen75_graphics_pipeline_create(VkDevice _device,
+ struct anv_pipeline_cache *cache,
+ const VkGraphicsPipelineCreateInfo *pCreateInfo,
+ const struct anv_graphics_pipeline_create_info *extra,
+ const VkAllocationCallbacks *alloc,
+ VkPipeline *pPipeline);
+
+VkResult
+gen8_graphics_pipeline_create(VkDevice _device,
+ struct anv_pipeline_cache *cache,
+ const VkGraphicsPipelineCreateInfo *pCreateInfo,
+ const struct anv_graphics_pipeline_create_info *extra,
+ const VkAllocationCallbacks *alloc,
+ VkPipeline *pPipeline);
+VkResult
+gen9_graphics_pipeline_create(VkDevice _device,
+ struct anv_pipeline_cache *cache,
+ const VkGraphicsPipelineCreateInfo *pCreateInfo,
+ const struct anv_graphics_pipeline_create_info *extra,
+ const VkAllocationCallbacks *alloc,
+ VkPipeline *pPipeline);
+VkResult
+gen7_compute_pipeline_create(VkDevice _device,
+ struct anv_pipeline_cache *cache,
+ const VkComputePipelineCreateInfo *pCreateInfo,
+ const VkAllocationCallbacks *alloc,
+ VkPipeline *pPipeline);
+VkResult
+gen75_compute_pipeline_create(VkDevice _device,
+ struct anv_pipeline_cache *cache,
+ const VkComputePipelineCreateInfo *pCreateInfo,
+ const VkAllocationCallbacks *alloc,
+ VkPipeline *pPipeline);
+
+VkResult
+gen8_compute_pipeline_create(VkDevice _device,
+ struct anv_pipeline_cache *cache,
+ const VkComputePipelineCreateInfo *pCreateInfo,
+ const VkAllocationCallbacks *alloc,
+ VkPipeline *pPipeline);
+VkResult
+gen9_compute_pipeline_create(VkDevice _device,
+ struct anv_pipeline_cache *cache,
+ const VkComputePipelineCreateInfo *pCreateInfo,
+ const VkAllocationCallbacks *alloc,
+ VkPipeline *pPipeline);
+
+struct anv_format {
+ const VkFormat vk_format;
+ const char *name;
+ enum isl_format surface_format; /**< RENDER_SURFACE_STATE.SurfaceFormat */
+ const struct isl_format_layout *isl_layout;
+ uint16_t depth_format; /**< 3DSTATE_DEPTH_BUFFER.SurfaceFormat */
+ bool has_stencil;
+};
+
+const struct anv_format *
+anv_format_for_vk_format(VkFormat format);
+
+enum isl_format
+anv_get_isl_format(VkFormat format, VkImageAspectFlags aspect,
+ VkImageTiling tiling);
+
+static inline bool
+anv_format_is_color(const struct anv_format *format)
+{
+ return !format->depth_format && !format->has_stencil;
+}
+
+static inline bool
+anv_format_is_depth_or_stencil(const struct anv_format *format)
+{
+ return format->depth_format || format->has_stencil;
+}
+
+/**
+ * Subsurface of an anv_image.
+ */
+struct anv_surface {
+ struct isl_surf isl;
+
+ /**
+ * Offset from VkImage's base address, as bound by vkBindImageMemory().
+ */
+ uint32_t offset;
+};
+
+struct anv_image {
+ VkImageType type;
+ /* The original VkFormat provided by the client. This may not match any
+ * of the actual surface formats.
+ */
+ VkFormat vk_format;
+ const struct anv_format *format;
+ VkExtent3D extent;
+ uint32_t levels;
+ uint32_t array_size;
+ VkImageUsageFlags usage; /**< Superset of VkImageCreateInfo::usage. */
+ VkImageTiling tiling; /** VkImageCreateInfo::tiling */
+
+ VkDeviceSize size;
+ uint32_t alignment;
+
+ /* Set when bound */
+ struct anv_bo *bo;
+ VkDeviceSize offset;
+
+ bool needs_nonrt_surface_state:1;
+ bool needs_color_rt_surface_state:1;
+ bool needs_storage_surface_state:1;
+
+ /**
+ * Image subsurfaces
+ *
+ * For each foo, anv_image::foo_surface is valid if and only if
+ * anv_image::format has a foo aspect.
+ *
+ * The hardware requires that the depth buffer and stencil buffer be
+ * separate surfaces. From Vulkan's perspective, though, depth and stencil
+ * reside in the same VkImage. To satisfy both the hardware and Vulkan, we
+ * allocate the depth and stencil buffers as separate surfaces in the same
+ * bo.
+ */
+ union {
+ struct anv_surface color_surface;
+
+ struct {
+ struct anv_surface depth_surface;
+ struct anv_surface stencil_surface;
+ };
+ };
+};
+
+struct anv_image_view {
+ const struct anv_image *image; /**< VkImageViewCreateInfo::image */
+ struct anv_bo *bo;
+ uint32_t offset; /**< Offset into bo. */
+
+ VkImageAspectFlags aspect_mask;
+ VkFormat vk_format;
+ enum isl_format format;
+ VkExtent3D extent; /**< Extent of VkImageViewCreateInfo::baseMipLevel. */
+
+ /** RENDER_SURFACE_STATE when using image as a color render target. */
+ struct anv_state color_rt_surface_state;
+
+ /** RENDER_SURFACE_STATE when using image as a non render target. */
+ struct anv_state nonrt_surface_state;
+
+ /** RENDER_SURFACE_STATE when using image as a storage image. */
+ struct anv_state storage_surface_state;
+};
+
+struct anv_image_create_info {
+ const VkImageCreateInfo *vk_info;
+ isl_tiling_flags_t isl_tiling_flags;
+ uint32_t stride;
+};
+
+VkResult anv_image_create(VkDevice _device,
+ const struct anv_image_create_info *info,
+ const VkAllocationCallbacks* alloc,
+ VkImage *pImage);
+
+struct anv_surface *
+anv_image_get_surface_for_aspect_mask(struct anv_image *image,
+ VkImageAspectFlags aspect_mask);
+
+void anv_image_view_init(struct anv_image_view *view,
+ struct anv_device *device,
+ const VkImageViewCreateInfo* pCreateInfo,
+ struct anv_cmd_buffer *cmd_buffer);
+
+void
+gen7_image_view_init(struct anv_image_view *iview,
+ struct anv_device *device,
+ const VkImageViewCreateInfo* pCreateInfo,
+ struct anv_cmd_buffer *cmd_buffer);
+
+void
+gen75_image_view_init(struct anv_image_view *iview,
+ struct anv_device *device,
+ const VkImageViewCreateInfo* pCreateInfo,
+ struct anv_cmd_buffer *cmd_buffer);
+
+void
+gen8_image_view_init(struct anv_image_view *iview,
+ struct anv_device *device,
+ const VkImageViewCreateInfo* pCreateInfo,
+ struct anv_cmd_buffer *cmd_buffer);
+
+void
+gen9_image_view_init(struct anv_image_view *iview,
+ struct anv_device *device,
+ const VkImageViewCreateInfo* pCreateInfo,
+ struct anv_cmd_buffer *cmd_buffer);
+
+struct anv_buffer_view {
+ enum isl_format format; /**< VkBufferViewCreateInfo::format */
+ struct anv_bo *bo;
+ uint32_t offset; /**< Offset into bo. */
+ uint64_t range; /**< VkBufferViewCreateInfo::range */
+
+ struct anv_state surface_state;
+ struct anv_state storage_surface_state;
+};
+
+const struct anv_format *
+anv_format_for_descriptor_type(VkDescriptorType type);
+
+void anv_fill_buffer_surface_state(struct anv_device *device, void *state,
+ enum isl_format format,
+ uint32_t offset, uint32_t range,
+ uint32_t stride);
+
+void gen7_fill_buffer_surface_state(void *state, enum isl_format format,
+ uint32_t offset, uint32_t range,
+ uint32_t stride);
+void gen75_fill_buffer_surface_state(void *state, enum isl_format format,
+ uint32_t offset, uint32_t range,
+ uint32_t stride);
+void gen8_fill_buffer_surface_state(void *state, enum isl_format format,
+ uint32_t offset, uint32_t range,
+ uint32_t stride);
+void gen9_fill_buffer_surface_state(void *state, enum isl_format format,
+ uint32_t offset, uint32_t range,
+ uint32_t stride);
+
+void anv_image_view_fill_image_param(struct anv_device *device,
+ struct anv_image_view *view,
+ struct brw_image_param *param);
+void anv_buffer_view_fill_image_param(struct anv_device *device,
+ struct anv_buffer_view *view,
+ struct brw_image_param *param);
+
+struct anv_sampler {
+ uint32_t state[4];
+};
+
+struct anv_framebuffer {
+ uint32_t width;
+ uint32_t height;
+ uint32_t layers;
+
+ uint32_t attachment_count;
+ const struct anv_image_view * attachments[0];
+};
+
+struct anv_subpass {
+ uint32_t input_count;
+ uint32_t * input_attachments;
+ uint32_t color_count;
+ uint32_t * color_attachments;
+ uint32_t * resolve_attachments;
+ uint32_t depth_stencil_attachment;
+};
+
+struct anv_render_pass_attachment {
+ const struct anv_format *format;
+ uint32_t samples;
+ VkAttachmentLoadOp load_op;
+ VkAttachmentLoadOp stencil_load_op;
+};
+
+struct anv_render_pass {
+ uint32_t attachment_count;
+ uint32_t subpass_count;
+ uint32_t * subpass_attachments;
+ struct anv_render_pass_attachment * attachments;
+ struct anv_subpass subpasses[0];
+};
+
+extern struct anv_render_pass anv_meta_dummy_renderpass;
+
+struct anv_query_pool_slot {
+ uint64_t begin;
+ uint64_t end;
+ uint64_t available;
+};
+
+struct anv_query_pool {
+ VkQueryType type;
+ uint32_t slots;
+ struct anv_bo bo;
+};
+
+VkResult anv_device_init_meta(struct anv_device *device);
+void anv_device_finish_meta(struct anv_device *device);
+
+void *anv_lookup_entrypoint(const char *name);
+
+void anv_dump_image_to_ppm(struct anv_device *device,
+ struct anv_image *image, unsigned miplevel,
+ unsigned array_layer, const char *filename);
+
+#define ANV_DEFINE_HANDLE_CASTS(__anv_type, __VkType) \
+ \
+ static inline struct __anv_type * \
+ __anv_type ## _from_handle(__VkType _handle) \
+ { \
+ return (struct __anv_type *) _handle; \
+ } \
+ \
+ static inline __VkType \
+ __anv_type ## _to_handle(struct __anv_type *_obj) \
+ { \
+ return (__VkType) _obj; \
+ }
+
+#define ANV_DEFINE_NONDISP_HANDLE_CASTS(__anv_type, __VkType) \
+ \
+ static inline struct __anv_type * \
+ __anv_type ## _from_handle(__VkType _handle) \
+ { \
+ return (struct __anv_type *)(uintptr_t) _handle; \
+ } \
+ \
+ static inline __VkType \
+ __anv_type ## _to_handle(struct __anv_type *_obj) \
+ { \
+ return (__VkType)(uintptr_t) _obj; \
+ }
+
+#define ANV_FROM_HANDLE(__anv_type, __name, __handle) \
+ struct __anv_type *__name = __anv_type ## _from_handle(__handle)
+
+ANV_DEFINE_HANDLE_CASTS(anv_cmd_buffer, VkCommandBuffer)
+ANV_DEFINE_HANDLE_CASTS(anv_device, VkDevice)
+ANV_DEFINE_HANDLE_CASTS(anv_instance, VkInstance)
+ANV_DEFINE_HANDLE_CASTS(anv_physical_device, VkPhysicalDevice)
+ANV_DEFINE_HANDLE_CASTS(anv_queue, VkQueue)
+
+ANV_DEFINE_NONDISP_HANDLE_CASTS(anv_cmd_pool, VkCommandPool)
+ANV_DEFINE_NONDISP_HANDLE_CASTS(anv_buffer, VkBuffer)
+ANV_DEFINE_NONDISP_HANDLE_CASTS(anv_buffer_view, VkBufferView)
+ANV_DEFINE_NONDISP_HANDLE_CASTS(anv_descriptor_set, VkDescriptorSet)
+ANV_DEFINE_NONDISP_HANDLE_CASTS(anv_descriptor_set_layout, VkDescriptorSetLayout)
+ANV_DEFINE_NONDISP_HANDLE_CASTS(anv_device_memory, VkDeviceMemory)
+ANV_DEFINE_NONDISP_HANDLE_CASTS(anv_fence, VkFence)
+ANV_DEFINE_NONDISP_HANDLE_CASTS(anv_event, VkEvent)
+ANV_DEFINE_NONDISP_HANDLE_CASTS(anv_framebuffer, VkFramebuffer)
+ANV_DEFINE_NONDISP_HANDLE_CASTS(anv_image, VkImage)
+ANV_DEFINE_NONDISP_HANDLE_CASTS(anv_image_view, VkImageView);
+ANV_DEFINE_NONDISP_HANDLE_CASTS(anv_pipeline_cache, VkPipelineCache)
+ANV_DEFINE_NONDISP_HANDLE_CASTS(anv_pipeline, VkPipeline)
+ANV_DEFINE_NONDISP_HANDLE_CASTS(anv_pipeline_layout, VkPipelineLayout)
+ANV_DEFINE_NONDISP_HANDLE_CASTS(anv_query_pool, VkQueryPool)
+ANV_DEFINE_NONDISP_HANDLE_CASTS(anv_render_pass, VkRenderPass)
+ANV_DEFINE_NONDISP_HANDLE_CASTS(anv_sampler, VkSampler)
+ANV_DEFINE_NONDISP_HANDLE_CASTS(anv_shader_module, VkShaderModule)
+
+#define ANV_DEFINE_STRUCT_CASTS(__anv_type, __VkType) \
+ \
+ static inline const __VkType * \
+ __anv_type ## _to_ ## __VkType(const struct __anv_type *__anv_obj) \
+ { \
+ return (const __VkType *) __anv_obj; \
+ }
+
+#define ANV_COMMON_TO_STRUCT(__VkType, __vk_name, __common_name) \
+ const __VkType *__vk_name = anv_common_to_ ## __VkType(__common_name)
+
+ANV_DEFINE_STRUCT_CASTS(anv_common, VkMemoryBarrier)
+ANV_DEFINE_STRUCT_CASTS(anv_common, VkBufferMemoryBarrier)
+ANV_DEFINE_STRUCT_CASTS(anv_common, VkImageMemoryBarrier)
+
+#ifdef __cplusplus
+}
+#endif
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <assert.h>
+#include <stdbool.h>
+#include <string.h>
+#include <unistd.h>
+#include <fcntl.h>
+
+#include "anv_private.h"
+
+VkResult anv_CreateQueryPool(
+ VkDevice _device,
+ const VkQueryPoolCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkQueryPool* pQueryPool)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ struct anv_query_pool *pool;
+ VkResult result;
+ uint32_t slot_size;
+ uint64_t size;
+
+ assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_QUERY_POOL_CREATE_INFO);
+
+ switch (pCreateInfo->queryType) {
+ case VK_QUERY_TYPE_OCCLUSION:
+ case VK_QUERY_TYPE_TIMESTAMP:
+ break;
+ case VK_QUERY_TYPE_PIPELINE_STATISTICS:
+ return VK_ERROR_INCOMPATIBLE_DRIVER;
+ default:
+ assert(!"Invalid query type");
+ }
+
+ slot_size = sizeof(struct anv_query_pool_slot);
+ pool = anv_alloc2(&device->alloc, pAllocator, sizeof(*pool), 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (pool == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ pool->type = pCreateInfo->queryType;
+ pool->slots = pCreateInfo->queryCount;
+
+ size = pCreateInfo->queryCount * slot_size;
+ result = anv_bo_init_new(&pool->bo, device, size);
+ if (result != VK_SUCCESS)
+ goto fail;
+
+ pool->bo.map = anv_gem_mmap(device, pool->bo.gem_handle, 0, size, 0);
+
+ *pQueryPool = anv_query_pool_to_handle(pool);
+
+ return VK_SUCCESS;
+
+ fail:
+ anv_free2(&device->alloc, pAllocator, pool);
+
+ return result;
+}
+
+void anv_DestroyQueryPool(
+ VkDevice _device,
+ VkQueryPool _pool,
+ const VkAllocationCallbacks* pAllocator)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ ANV_FROM_HANDLE(anv_query_pool, pool, _pool);
+
+ anv_gem_munmap(pool->bo.map, pool->bo.size);
+ anv_gem_close(device, pool->bo.gem_handle);
+ anv_free2(&device->alloc, pAllocator, pool);
+}
+
+VkResult anv_GetQueryPoolResults(
+ VkDevice _device,
+ VkQueryPool queryPool,
+ uint32_t firstQuery,
+ uint32_t queryCount,
+ size_t dataSize,
+ void* pData,
+ VkDeviceSize stride,
+ VkQueryResultFlags flags)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ ANV_FROM_HANDLE(anv_query_pool, pool, queryPool);
+ int64_t timeout = INT64_MAX;
+ uint64_t result;
+ int ret;
+
+ assert(pool->type == VK_QUERY_TYPE_OCCLUSION ||
+ pool->type == VK_QUERY_TYPE_TIMESTAMP);
+
+ if (pData == NULL)
+ return VK_SUCCESS;
+
+ if (flags & VK_QUERY_RESULT_WAIT_BIT) {
+ ret = anv_gem_wait(device, pool->bo.gem_handle, &timeout);
+ if (ret == -1) {
+ /* We don't know the real error. */
+ return vk_errorf(VK_ERROR_OUT_OF_DEVICE_MEMORY,
+ "gem_wait failed %m");
+ }
+ }
+
+ void *data_end = pData + dataSize;
+ struct anv_query_pool_slot *slot = pool->bo.map;
+
+ for (uint32_t i = 0; i < queryCount; i++) {
+ switch (pool->type) {
+ case VK_QUERY_TYPE_OCCLUSION: {
+ result = slot[firstQuery + i].end - slot[firstQuery + i].begin;
+ break;
+ }
+ case VK_QUERY_TYPE_PIPELINE_STATISTICS:
+ /* Not yet implemented */
+ break;
+ case VK_QUERY_TYPE_TIMESTAMP: {
+ result = slot[firstQuery + i].begin;
+ break;
+ }
+ default:
+ assert(!"Invalid query type");
+ }
+
+ if (flags & VK_QUERY_RESULT_64_BIT) {
+ uint64_t *dst = pData;
+ dst[0] = result;
+ if (flags & VK_QUERY_RESULT_WITH_AVAILABILITY_BIT)
+ dst[1] = slot[firstQuery + i].available;
+ } else {
+ uint32_t *dst = pData;
+ if (result > UINT32_MAX)
+ result = UINT32_MAX;
+ dst[0] = result;
+ if (flags & VK_QUERY_RESULT_WITH_AVAILABILITY_BIT)
+ dst[1] = slot[firstQuery + i].available;
+ }
+
+ pData += stride;
+ if (pData >= data_end)
+ break;
+ }
+
+ return VK_SUCCESS;
+}
+
+void anv_CmdResetQueryPool(
+ VkCommandBuffer commandBuffer,
+ VkQueryPool queryPool,
+ uint32_t firstQuery,
+ uint32_t queryCount)
+{
+ ANV_FROM_HANDLE(anv_query_pool, pool, queryPool);
+
+ for (uint32_t i = 0; i < queryCount; i++) {
+ switch (pool->type) {
+ case VK_QUERY_TYPE_OCCLUSION:
+ case VK_QUERY_TYPE_TIMESTAMP: {
+ struct anv_query_pool_slot *slot = pool->bo.map;
+ slot[firstQuery + i].available = 0;
+ break;
+ }
+ default:
+ assert(!"Invalid query type");
+ }
+ }
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <stdarg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <errno.h>
+#include <assert.h>
+
+#include "anv_private.h"
+
+/** Log an error message. */
+void anv_printflike(1, 2)
+anv_loge(const char *format, ...)
+{
+ va_list va;
+
+ va_start(va, format);
+ anv_loge_v(format, va);
+ va_end(va);
+}
+
+/** \see anv_loge() */
+void
+anv_loge_v(const char *format, va_list va)
+{
+ fprintf(stderr, "vk: error: ");
+ vfprintf(stderr, format, va);
+ fprintf(stderr, "\n");
+}
+
+void anv_printflike(3, 4)
+__anv_finishme(const char *file, int line, const char *format, ...)
+{
+ va_list ap;
+ char buffer[256];
+
+ va_start(ap, format);
+ vsnprintf(buffer, sizeof(buffer), format, ap);
+ va_end(ap);
+
+ fprintf(stderr, "%s:%d: FINISHME: %s\n", file, line, buffer);
+}
+
+void anv_noreturn anv_printflike(1, 2)
+anv_abortf(const char *format, ...)
+{
+ va_list va;
+
+ va_start(va, format);
+ anv_abortfv(format, va);
+ va_end(va);
+}
+
+void anv_noreturn
+anv_abortfv(const char *format, va_list va)
+{
+ fprintf(stderr, "vk: error: ");
+ vfprintf(stderr, format, va);
+ fprintf(stderr, "\n");
+ abort();
+}
+
+VkResult
+__vk_errorf(VkResult error, const char *file, int line, const char *format, ...)
+{
+ va_list ap;
+ char buffer[256];
+
+#define ERROR_CASE(error) case error: error_str = #error; break;
+
+ const char *error_str;
+ switch ((int32_t)error) {
+
+ /* Core errors */
+ ERROR_CASE(VK_ERROR_OUT_OF_HOST_MEMORY)
+ ERROR_CASE(VK_ERROR_OUT_OF_DEVICE_MEMORY)
+ ERROR_CASE(VK_ERROR_INITIALIZATION_FAILED)
+ ERROR_CASE(VK_ERROR_DEVICE_LOST)
+ ERROR_CASE(VK_ERROR_MEMORY_MAP_FAILED)
+ ERROR_CASE(VK_ERROR_LAYER_NOT_PRESENT)
+ ERROR_CASE(VK_ERROR_EXTENSION_NOT_PRESENT)
+ ERROR_CASE(VK_ERROR_INCOMPATIBLE_DRIVER)
+
+ /* Extension errors */
+ ERROR_CASE(VK_ERROR_OUT_OF_DATE_KHR)
+
+ default:
+ assert(!"Unknown error");
+ error_str = "unknown error";
+ }
+
+#undef ERROR_CASE
+
+ if (format) {
+ va_start(ap, format);
+ vsnprintf(buffer, sizeof(buffer), format, ap);
+ va_end(ap);
+
+ fprintf(stderr, "%s:%d: %s (%s)\n", file, line, buffer, error_str);
+ } else {
+ fprintf(stderr, "%s:%d: %s\n", file, line, error_str);
+ }
+
+ return error;
+}
+
+int
+anv_vector_init(struct anv_vector *vector, uint32_t element_size, uint32_t size)
+{
+ assert(util_is_power_of_two(size));
+ assert(element_size < size && util_is_power_of_two(element_size));
+
+ vector->head = 0;
+ vector->tail = 0;
+ vector->element_size = element_size;
+ vector->size = size;
+ vector->data = malloc(size);
+
+ return vector->data != NULL;
+}
+
+void *
+anv_vector_add(struct anv_vector *vector)
+{
+ uint32_t offset, size, split, tail;
+ void *data;
+
+ if (vector->head - vector->tail == vector->size) {
+ size = vector->size * 2;
+ data = malloc(size);
+ if (data == NULL)
+ return NULL;
+ split = align_u32(vector->tail, vector->size);
+ tail = vector->tail & (vector->size - 1);
+ if (vector->head - split < vector->size) {
+ memcpy(data + tail,
+ vector->data + tail,
+ split - vector->tail);
+ memcpy(data + vector->size,
+ vector->data, vector->head - split);
+ } else {
+ memcpy(data + tail,
+ vector->data + tail,
+ vector->head - vector->tail);
+ }
+ free(vector->data);
+ vector->data = data;
+ vector->size = size;
+ }
+
+ assert(vector->head - vector->tail < vector->size);
+
+ offset = vector->head & (vector->size - 1);
+ vector->head += vector->element_size;
+
+ return vector->data + offset;
+}
+
+void *
+anv_vector_remove(struct anv_vector *vector)
+{
+ uint32_t offset;
+
+ if (vector->head == vector->tail)
+ return NULL;
+
+ assert(vector->head - vector->tail <= vector->size);
+
+ offset = vector->tail & (vector->size - 1);
+ vector->tail += vector->element_size;
+
+ return vector->data + offset;
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include "anv_wsi.h"
+
+VkResult
+anv_init_wsi(struct anv_instance *instance)
+{
+ VkResult result;
+
+ result = anv_x11_init_wsi(instance);
+ if (result != VK_SUCCESS)
+ return result;
+
+#ifdef HAVE_WAYLAND_PLATFORM
+ result = anv_wl_init_wsi(instance);
+ if (result != VK_SUCCESS) {
+ anv_x11_finish_wsi(instance);
+ return result;
+ }
+#endif
+
+ return VK_SUCCESS;
+}
+
+void
+anv_finish_wsi(struct anv_instance *instance)
+{
+#ifdef HAVE_WAYLAND_PLATFORM
+ anv_wl_finish_wsi(instance);
+#endif
+ anv_x11_finish_wsi(instance);
+}
+
+void anv_DestroySurfaceKHR(
+ VkInstance instance,
+ VkSurfaceKHR _surface,
+ const VkAllocationCallbacks* pAllocator)
+{
+ ANV_FROM_HANDLE(anv_wsi_surface, surface, _surface);
+
+ surface->destroy(surface, pAllocator);
+}
+
+VkResult anv_GetPhysicalDeviceSurfaceSupportKHR(
+ VkPhysicalDevice physicalDevice,
+ uint32_t queueFamilyIndex,
+ VkSurfaceKHR _surface,
+ VkBool32* pSupported)
+{
+ ANV_FROM_HANDLE(anv_physical_device, device, physicalDevice);
+ ANV_FROM_HANDLE(anv_wsi_surface, surface, _surface);
+
+ return surface->get_support(surface, device, queueFamilyIndex, pSupported);
+}
+
+VkResult anv_GetPhysicalDeviceSurfaceCapabilitiesKHR(
+ VkPhysicalDevice physicalDevice,
+ VkSurfaceKHR _surface,
+ VkSurfaceCapabilitiesKHR* pSurfaceCapabilities)
+{
+ ANV_FROM_HANDLE(anv_physical_device, device, physicalDevice);
+ ANV_FROM_HANDLE(anv_wsi_surface, surface, _surface);
+
+ return surface->get_capabilities(surface, device, pSurfaceCapabilities);
+}
+
+VkResult anv_GetPhysicalDeviceSurfaceFormatsKHR(
+ VkPhysicalDevice physicalDevice,
+ VkSurfaceKHR _surface,
+ uint32_t* pSurfaceFormatCount,
+ VkSurfaceFormatKHR* pSurfaceFormats)
+{
+ ANV_FROM_HANDLE(anv_physical_device, device, physicalDevice);
+ ANV_FROM_HANDLE(anv_wsi_surface, surface, _surface);
+
+ return surface->get_formats(surface, device, pSurfaceFormatCount,
+ pSurfaceFormats);
+}
+
+VkResult anv_GetPhysicalDeviceSurfacePresentModesKHR(
+ VkPhysicalDevice physicalDevice,
+ VkSurfaceKHR _surface,
+ uint32_t* pPresentModeCount,
+ VkPresentModeKHR* pPresentModes)
+{
+ ANV_FROM_HANDLE(anv_physical_device, device, physicalDevice);
+ ANV_FROM_HANDLE(anv_wsi_surface, surface, _surface);
+
+ return surface->get_present_modes(surface, device, pPresentModeCount,
+ pPresentModes);
+}
+
+VkResult anv_CreateSwapchainKHR(
+ VkDevice _device,
+ const VkSwapchainCreateInfoKHR* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkSwapchainKHR* pSwapchain)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ ANV_FROM_HANDLE(anv_wsi_surface, surface, pCreateInfo->surface);
+ struct anv_swapchain *swapchain;
+
+ VkResult result = surface->create_swapchain(surface, device, pCreateInfo,
+ pAllocator, &swapchain);
+ if (result != VK_SUCCESS)
+ return result;
+
+ *pSwapchain = anv_swapchain_to_handle(swapchain);
+
+ return VK_SUCCESS;
+}
+
+void anv_DestroySwapchainKHR(
+ VkDevice device,
+ VkSwapchainKHR _swapchain,
+ const VkAllocationCallbacks* pAllocator)
+{
+ ANV_FROM_HANDLE(anv_swapchain, swapchain, _swapchain);
+
+ swapchain->destroy(swapchain, pAllocator);
+}
+
+VkResult anv_GetSwapchainImagesKHR(
+ VkDevice device,
+ VkSwapchainKHR _swapchain,
+ uint32_t* pSwapchainImageCount,
+ VkImage* pSwapchainImages)
+{
+ ANV_FROM_HANDLE(anv_swapchain, swapchain, _swapchain);
+
+ return swapchain->get_images(swapchain, pSwapchainImageCount,
+ pSwapchainImages);
+}
+
+VkResult anv_AcquireNextImageKHR(
+ VkDevice device,
+ VkSwapchainKHR _swapchain,
+ uint64_t timeout,
+ VkSemaphore semaphore,
+ VkFence fence,
+ uint32_t* pImageIndex)
+{
+ ANV_FROM_HANDLE(anv_swapchain, swapchain, _swapchain);
+
+ return swapchain->acquire_next_image(swapchain, timeout, semaphore,
+ pImageIndex);
+}
+
+VkResult anv_QueuePresentKHR(
+ VkQueue _queue,
+ const VkPresentInfoKHR* pPresentInfo)
+{
+ ANV_FROM_HANDLE(anv_queue, queue, _queue);
+ VkResult result;
+
+ for (uint32_t i = 0; i < pPresentInfo->swapchainCount; i++) {
+ ANV_FROM_HANDLE(anv_swapchain, swapchain, pPresentInfo->pSwapchains[i]);
+
+ assert(swapchain->device == queue->device);
+
+ result = swapchain->queue_present(swapchain, queue,
+ pPresentInfo->pImageIndices[i]);
+ /* TODO: What if one of them returns OUT_OF_DATE? */
+ if (result != VK_SUCCESS)
+ return result;
+ }
+
+ return VK_SUCCESS;
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#pragma once
+
+#include "anv_private.h"
+
+struct anv_swapchain;
+
+struct anv_wsi_surface {
+ struct anv_instance *instance;
+
+ void (*destroy)(struct anv_wsi_surface *surface,
+ const VkAllocationCallbacks *pAllocator);
+ VkResult (*get_support)(struct anv_wsi_surface *surface,
+ struct anv_physical_device *device,
+ uint32_t queueFamilyIndex,
+ VkBool32* pSupported);
+ VkResult (*get_capabilities)(struct anv_wsi_surface *surface,
+ struct anv_physical_device *device,
+ VkSurfaceCapabilitiesKHR* pSurfaceCapabilities);
+ VkResult (*get_formats)(struct anv_wsi_surface *surface,
+ struct anv_physical_device *device,
+ uint32_t* pSurfaceFormatCount,
+ VkSurfaceFormatKHR* pSurfaceFormats);
+ VkResult (*get_present_modes)(struct anv_wsi_surface *surface,
+ struct anv_physical_device *device,
+ uint32_t* pPresentModeCount,
+ VkPresentModeKHR* pPresentModes);
+ VkResult (*create_swapchain)(struct anv_wsi_surface *surface,
+ struct anv_device *device,
+ const VkSwapchainCreateInfoKHR* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ struct anv_swapchain **swapchain);
+};
+
+struct anv_swapchain {
+ struct anv_device *device;
+
+ VkResult (*destroy)(struct anv_swapchain *swapchain,
+ const VkAllocationCallbacks *pAllocator);
+ VkResult (*get_images)(struct anv_swapchain *swapchain,
+ uint32_t *pCount, VkImage *pSwapchainImages);
+ VkResult (*acquire_next_image)(struct anv_swapchain *swap_chain,
+ uint64_t timeout, VkSemaphore semaphore,
+ uint32_t *image_index);
+ VkResult (*queue_present)(struct anv_swapchain *swap_chain,
+ struct anv_queue *queue,
+ uint32_t image_index);
+};
+
+ANV_DEFINE_NONDISP_HANDLE_CASTS(anv_wsi_surface, VkSurfaceKHR)
+ANV_DEFINE_NONDISP_HANDLE_CASTS(anv_swapchain, VkSwapchainKHR)
+
+VkResult anv_x11_init_wsi(struct anv_instance *instance);
+void anv_x11_finish_wsi(struct anv_instance *instance);
+VkResult anv_wl_init_wsi(struct anv_instance *instance);
+void anv_wl_finish_wsi(struct anv_instance *instance);
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <wayland-client.h>
+#include <wayland-drm-client-protocol.h>
+
+#include "anv_wsi.h"
+
+#include <util/hash_table.h>
+
+#define MIN_NUM_IMAGES 2
+
+struct wsi_wl_surface {
+ struct anv_wsi_surface base;
+
+ struct wl_display *display;
+ struct wl_surface *surface;
+};
+
+struct wsi_wl_display {
+ struct wl_display * display;
+ struct wl_drm * drm;
+
+ /* Vector of VkFormats supported */
+ struct anv_vector formats;
+
+ uint32_t capabilities;
+};
+
+struct wsi_wayland {
+ struct anv_instance * instance;
+
+ pthread_mutex_t mutex;
+ /* Hash table of wl_display -> wsi_wl_display mappings */
+ struct hash_table * displays;
+};
+
+static void
+wsi_wl_display_add_vk_format(struct wsi_wl_display *display, VkFormat format)
+{
+ /* Don't add a format that's already in the list */
+ VkFormat *f;
+ anv_vector_foreach(f, &display->formats)
+ if (*f == format)
+ return;
+
+ /* Don't add formats which aren't supported by the driver */
+ if (anv_format_for_vk_format(format)->surface_format ==
+ ISL_FORMAT_UNSUPPORTED) {
+ return;
+ }
+
+ f = anv_vector_add(&display->formats);
+ if (f)
+ *f = format;
+}
+
+static void
+drm_handle_device(void *data, struct wl_drm *drm, const char *name)
+{
+ fprintf(stderr, "wl_drm.device(%s)\n", name);
+}
+
+static uint32_t
+wl_drm_format_for_vk_format(VkFormat vk_format, bool alpha)
+{
+ switch (vk_format) {
+ /* TODO: Figure out what all the formats mean and make this table
+ * correct.
+ */
+#if 0
+ case VK_FORMAT_R4G4B4A4_UNORM:
+ return alpha ? WL_DRM_FORMAT_ABGR4444 : WL_DRM_FORMAT_XBGR4444;
+ case VK_FORMAT_R5G6B5_UNORM:
+ return WL_DRM_FORMAT_BGR565;
+ case VK_FORMAT_R5G5B5A1_UNORM:
+ return alpha ? WL_DRM_FORMAT_ABGR1555 : WL_DRM_FORMAT_XBGR1555;
+ case VK_FORMAT_R8G8B8_UNORM:
+ return WL_DRM_FORMAT_XBGR8888;
+ case VK_FORMAT_R8G8B8A8_UNORM:
+ return alpha ? WL_DRM_FORMAT_ABGR8888 : WL_DRM_FORMAT_XBGR8888;
+ case VK_FORMAT_R10G10B10A2_UNORM:
+ return alpha ? WL_DRM_FORMAT_ABGR2101010 : WL_DRM_FORMAT_XBGR2101010;
+ case VK_FORMAT_B4G4R4A4_UNORM:
+ return alpha ? WL_DRM_FORMAT_ARGB4444 : WL_DRM_FORMAT_XRGB4444;
+ case VK_FORMAT_B5G6R5_UNORM:
+ return WL_DRM_FORMAT_RGB565;
+ case VK_FORMAT_B5G5R5A1_UNORM:
+ return alpha ? WL_DRM_FORMAT_XRGB1555 : WL_DRM_FORMAT_XRGB1555;
+#endif
+ case VK_FORMAT_B8G8R8_UNORM:
+ return WL_DRM_FORMAT_BGRX8888;
+ case VK_FORMAT_B8G8R8A8_UNORM:
+ return alpha ? WL_DRM_FORMAT_ARGB8888 : WL_DRM_FORMAT_XRGB8888;
+#if 0
+ case VK_FORMAT_B10G10R10A2_UNORM:
+ return alpha ? WL_DRM_FORMAT_ARGB2101010 : WL_DRM_FORMAT_XRGB2101010;
+#endif
+
+ default:
+ assert("!Unsupported Vulkan format");
+ return 0;
+ }
+}
+
+static void
+drm_handle_format(void *data, struct wl_drm *drm, uint32_t wl_format)
+{
+ struct wsi_wl_display *display = data;
+
+ switch (wl_format) {
+#if 0
+ case WL_DRM_FORMAT_ABGR4444:
+ case WL_DRM_FORMAT_XBGR4444:
+ wsi_wl_display_add_vk_format(display, VK_FORMAT_R4G4B4A4_UNORM);
+ break;
+ case WL_DRM_FORMAT_BGR565:
+ wsi_wl_display_add_vk_format(display, VK_FORMAT_R5G6B5_UNORM);
+ break;
+ case WL_DRM_FORMAT_ABGR1555:
+ case WL_DRM_FORMAT_XBGR1555:
+ wsi_wl_display_add_vk_format(display, VK_FORMAT_R5G5B5A1_UNORM);
+ break;
+ case WL_DRM_FORMAT_XBGR8888:
+ wsi_wl_display_add_vk_format(display, VK_FORMAT_R8G8B8_UNORM);
+ /* fallthrough */
+ case WL_DRM_FORMAT_ABGR8888:
+ wsi_wl_display_add_vk_format(display, VK_FORMAT_R8G8B8A8_UNORM);
+ break;
+ case WL_DRM_FORMAT_ABGR2101010:
+ case WL_DRM_FORMAT_XBGR2101010:
+ wsi_wl_display_add_vk_format(display, VK_FORMAT_R10G10B10A2_UNORM);
+ break;
+ case WL_DRM_FORMAT_ARGB4444:
+ case WL_DRM_FORMAT_XRGB4444:
+ wsi_wl_display_add_vk_format(display, VK_FORMAT_B4G4R4A4_UNORM);
+ break;
+ case WL_DRM_FORMAT_RGB565:
+ wsi_wl_display_add_vk_format(display, VK_FORMAT_B5G6R5_UNORM);
+ break;
+ case WL_DRM_FORMAT_ARGB1555:
+ case WL_DRM_FORMAT_XRGB1555:
+ wsi_wl_display_add_vk_format(display, VK_FORMAT_B5G5R5A1_UNORM);
+ break;
+#endif
+ case WL_DRM_FORMAT_XRGB8888:
+ wsi_wl_display_add_vk_format(display, VK_FORMAT_B8G8R8_UNORM);
+ /* fallthrough */
+ case WL_DRM_FORMAT_ARGB8888:
+ wsi_wl_display_add_vk_format(display, VK_FORMAT_B8G8R8A8_UNORM);
+ break;
+#if 0
+ case WL_DRM_FORMAT_ARGB2101010:
+ case WL_DRM_FORMAT_XRGB2101010:
+ wsi_wl_display_add_vk_format(display, VK_FORMAT_B10G10R10A2_UNORM);
+ break;
+#endif
+ }
+}
+
+static void
+drm_handle_authenticated(void *data, struct wl_drm *drm)
+{
+}
+
+static void
+drm_handle_capabilities(void *data, struct wl_drm *drm, uint32_t capabilities)
+{
+ struct wsi_wl_display *display = data;
+
+ display->capabilities = capabilities;
+}
+
+static const struct wl_drm_listener drm_listener = {
+ drm_handle_device,
+ drm_handle_format,
+ drm_handle_authenticated,
+ drm_handle_capabilities,
+};
+
+static void
+registry_handle_global(void *data, struct wl_registry *registry,
+ uint32_t name, const char *interface, uint32_t version)
+{
+ struct wsi_wl_display *display = data;
+
+ if (strcmp(interface, "wl_drm") == 0) {
+ assert(display->drm == NULL);
+
+ assert(version >= 2);
+ display->drm = wl_registry_bind(registry, name, &wl_drm_interface, 2);
+
+ if (display->drm)
+ wl_drm_add_listener(display->drm, &drm_listener, display);
+ }
+}
+
+static void
+registry_handle_global_remove(void *data, struct wl_registry *registry,
+ uint32_t name)
+{ /* No-op */ }
+
+static const struct wl_registry_listener registry_listener = {
+ registry_handle_global,
+ registry_handle_global_remove
+};
+
+static void
+wsi_wl_display_destroy(struct wsi_wayland *wsi, struct wsi_wl_display *display)
+{
+ anv_vector_finish(&display->formats);
+ if (display->drm)
+ wl_drm_destroy(display->drm);
+ anv_free(&wsi->instance->alloc, display);
+}
+
+static struct wsi_wl_display *
+wsi_wl_display_create(struct wsi_wayland *wsi, struct wl_display *wl_display)
+{
+ struct wsi_wl_display *display =
+ anv_alloc(&wsi->instance->alloc, sizeof(*display), 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_INSTANCE);
+ if (!display)
+ return NULL;
+
+ memset(display, 0, sizeof(*display));
+
+ display->display = wl_display;
+
+ if (!anv_vector_init(&display->formats, sizeof(VkFormat), 8))
+ goto fail;
+
+ struct wl_registry *registry = wl_display_get_registry(wl_display);
+ if (!registry)
+ return NULL;
+
+ wl_registry_add_listener(registry, ®istry_listener, display);
+
+ /* Round-rip to get the wl_drm global */
+ wl_display_roundtrip(wl_display);
+
+ if (!display->drm)
+ goto fail;
+
+ /* Round-rip to get wl_drm formats and capabilities */
+ wl_display_roundtrip(wl_display);
+
+ /* We need prime support */
+ if (!(display->capabilities & WL_DRM_CAPABILITY_PRIME))
+ goto fail;
+
+ /* We don't need this anymore */
+ wl_registry_destroy(registry);
+
+ return display;
+
+fail:
+ if (registry)
+ wl_registry_destroy(registry);
+
+ wsi_wl_display_destroy(wsi, display);
+ return NULL;
+}
+
+static struct wsi_wl_display *
+wsi_wl_get_display(struct anv_instance *instance, struct wl_display *wl_display)
+{
+ struct wsi_wayland *wsi = instance->wayland_wsi;
+
+ pthread_mutex_lock(&wsi->mutex);
+
+ struct hash_entry *entry = _mesa_hash_table_search(wsi->displays,
+ wl_display);
+ if (!entry) {
+ /* We're about to make a bunch of blocking calls. Let's drop the
+ * mutex for now so we don't block up too badly.
+ */
+ pthread_mutex_unlock(&wsi->mutex);
+
+ struct wsi_wl_display *display = wsi_wl_display_create(wsi, wl_display);
+
+ pthread_mutex_lock(&wsi->mutex);
+
+ entry = _mesa_hash_table_search(wsi->displays, wl_display);
+ if (entry) {
+ /* Oops, someone raced us to it */
+ wsi_wl_display_destroy(wsi, display);
+ } else {
+ entry = _mesa_hash_table_insert(wsi->displays, wl_display, display);
+ }
+ }
+
+ pthread_mutex_unlock(&wsi->mutex);
+
+ return entry->data;
+}
+
+VkBool32 anv_GetPhysicalDeviceWaylandPresentationSupportKHR(
+ VkPhysicalDevice physicalDevice,
+ uint32_t queueFamilyIndex,
+ struct wl_display* display)
+{
+ ANV_FROM_HANDLE(anv_physical_device, physical_device, physicalDevice);
+
+ return wsi_wl_get_display(physical_device->instance, display) != NULL;
+}
+
+static VkResult
+wsi_wl_surface_get_support(struct anv_wsi_surface *surface,
+ struct anv_physical_device *device,
+ uint32_t queueFamilyIndex,
+ VkBool32* pSupported)
+{
+ *pSupported = true;
+
+ return VK_SUCCESS;
+}
+
+static const VkPresentModeKHR present_modes[] = {
+ VK_PRESENT_MODE_MAILBOX_KHR,
+ VK_PRESENT_MODE_FIFO_KHR,
+};
+
+static VkResult
+wsi_wl_surface_get_capabilities(struct anv_wsi_surface *surface,
+ struct anv_physical_device *device,
+ VkSurfaceCapabilitiesKHR* caps)
+{
+ caps->minImageCount = MIN_NUM_IMAGES;
+ caps->maxImageCount = 4;
+ caps->currentExtent = (VkExtent2D) { -1, -1 };
+ caps->minImageExtent = (VkExtent2D) { 1, 1 };
+ caps->maxImageExtent = (VkExtent2D) { INT16_MAX, INT16_MAX };
+ caps->supportedTransforms = VK_SURFACE_TRANSFORM_IDENTITY_BIT_KHR;
+ caps->currentTransform = VK_SURFACE_TRANSFORM_IDENTITY_BIT_KHR;
+ caps->maxImageArrayLayers = 1;
+
+ caps->supportedCompositeAlpha =
+ VK_COMPOSITE_ALPHA_OPAQUE_BIT_KHR |
+ VK_COMPOSITE_ALPHA_PRE_MULTIPLIED_BIT_KHR;
+
+ caps->supportedUsageFlags =
+ VK_IMAGE_USAGE_TRANSFER_DST_BIT |
+ VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT;
+
+ return VK_SUCCESS;
+}
+
+static VkResult
+wsi_wl_surface_get_formats(struct anv_wsi_surface *wsi_surface,
+ struct anv_physical_device *device,
+ uint32_t* pSurfaceFormatCount,
+ VkSurfaceFormatKHR* pSurfaceFormats)
+{
+ struct wsi_wl_surface *surface = (struct wsi_wl_surface *)wsi_surface;
+ struct wsi_wl_display *display =
+ wsi_wl_get_display(device->instance, surface->display);
+
+ uint32_t count = anv_vector_length(&display->formats);
+
+ if (pSurfaceFormats == NULL) {
+ *pSurfaceFormatCount = count;
+ return VK_SUCCESS;
+ }
+
+ assert(*pSurfaceFormatCount >= count);
+ *pSurfaceFormatCount = count;
+
+ VkFormat *f;
+ anv_vector_foreach(f, &display->formats) {
+ *(pSurfaceFormats++) = (VkSurfaceFormatKHR) {
+ .format = *f,
+ /* TODO: We should get this from the compositor somehow */
+ .colorSpace = VK_COLORSPACE_SRGB_NONLINEAR_KHR,
+ };
+ }
+
+ return VK_SUCCESS;
+}
+
+static VkResult
+wsi_wl_surface_get_present_modes(struct anv_wsi_surface *surface,
+ struct anv_physical_device *device,
+ uint32_t* pPresentModeCount,
+ VkPresentModeKHR* pPresentModes)
+{
+ if (pPresentModes == NULL) {
+ *pPresentModeCount = ARRAY_SIZE(present_modes);
+ return VK_SUCCESS;
+ }
+
+ assert(*pPresentModeCount >= ARRAY_SIZE(present_modes));
+ typed_memcpy(pPresentModes, present_modes, *pPresentModeCount);
+ *pPresentModeCount = ARRAY_SIZE(present_modes);
+
+ return VK_SUCCESS;
+}
+
+static void
+wsi_wl_surface_destroy(struct anv_wsi_surface *surface,
+ const VkAllocationCallbacks *pAllocator)
+{
+ anv_free2(&surface->instance->alloc, pAllocator, surface);
+}
+
+static VkResult
+wsi_wl_surface_create_swapchain(struct anv_wsi_surface *surface,
+ struct anv_device *device,
+ const VkSwapchainCreateInfoKHR* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ struct anv_swapchain **swapchain);
+
+VkResult anv_CreateWaylandSurfaceKHR(
+ VkInstance _instance,
+ const VkWaylandSurfaceCreateInfoKHR* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkSurfaceKHR* pSurface)
+{
+ assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_WAYLAND_SURFACE_CREATE_INFO_KHR);
+
+ ANV_FROM_HANDLE(anv_instance, instance, _instance);
+ struct wsi_wl_surface *surface;
+
+ surface = anv_alloc2(&instance->alloc, pAllocator, sizeof *surface, 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (surface == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ surface->display = pCreateInfo->display;
+ surface->surface = pCreateInfo->surface;
+
+ surface->base.instance = instance;
+ surface->base.destroy = wsi_wl_surface_destroy;
+ surface->base.get_support = wsi_wl_surface_get_support;
+ surface->base.get_capabilities = wsi_wl_surface_get_capabilities;
+ surface->base.get_formats = wsi_wl_surface_get_formats;
+ surface->base.get_present_modes = wsi_wl_surface_get_present_modes;
+ surface->base.create_swapchain = wsi_wl_surface_create_swapchain;
+
+ *pSurface = anv_wsi_surface_to_handle(&surface->base);
+
+ return VK_SUCCESS;
+}
+
+struct wsi_wl_image {
+ struct anv_image * image;
+ struct anv_device_memory * memory;
+ struct wl_buffer * buffer;
+ bool busy;
+};
+
+struct wsi_wl_swapchain {
+ struct anv_swapchain base;
+
+ struct wsi_wl_display * display;
+ struct wl_event_queue * queue;
+ struct wl_surface * surface;
+
+ VkExtent2D extent;
+ VkFormat vk_format;
+ uint32_t drm_format;
+
+ VkPresentModeKHR present_mode;
+ bool fifo_ready;
+
+ uint32_t image_count;
+ struct wsi_wl_image images[0];
+};
+
+static VkResult
+wsi_wl_swapchain_get_images(struct anv_swapchain *anv_chain,
+ uint32_t *pCount, VkImage *pSwapchainImages)
+{
+ struct wsi_wl_swapchain *chain = (struct wsi_wl_swapchain *)anv_chain;
+
+ if (pSwapchainImages == NULL) {
+ *pCount = chain->image_count;
+ return VK_SUCCESS;
+ }
+
+ assert(chain->image_count <= *pCount);
+ for (uint32_t i = 0; i < chain->image_count; i++)
+ pSwapchainImages[i] = anv_image_to_handle(chain->images[i].image);
+
+ *pCount = chain->image_count;
+
+ return VK_SUCCESS;
+}
+
+static VkResult
+wsi_wl_swapchain_acquire_next_image(struct anv_swapchain *anv_chain,
+ uint64_t timeout,
+ VkSemaphore semaphore,
+ uint32_t *image_index)
+{
+ struct wsi_wl_swapchain *chain = (struct wsi_wl_swapchain *)anv_chain;
+
+ int ret = wl_display_dispatch_queue_pending(chain->display->display,
+ chain->queue);
+ /* XXX: I'm not sure if out-of-date is the right error here. If
+ * wl_display_dispatch_queue_pending fails it most likely means we got
+ * kicked by the server so this seems more-or-less correct.
+ */
+ if (ret < 0)
+ return vk_error(VK_ERROR_OUT_OF_DATE_KHR);
+
+ while (1) {
+ for (uint32_t i = 0; i < chain->image_count; i++) {
+ if (!chain->images[i].busy) {
+ /* We found a non-busy image */
+ *image_index = i;
+ return VK_SUCCESS;
+ }
+ }
+
+ /* This time we do a blocking dispatch because we can't go
+ * anywhere until we get an event.
+ */
+ int ret = wl_display_dispatch_queue(chain->display->display,
+ chain->queue);
+ if (ret < 0)
+ return vk_error(VK_ERROR_OUT_OF_DATE_KHR);
+ }
+}
+
+static void
+frame_handle_done(void *data, struct wl_callback *callback, uint32_t serial)
+{
+ struct wsi_wl_swapchain *chain = data;
+
+ chain->fifo_ready = true;
+
+ wl_callback_destroy(callback);
+}
+
+static const struct wl_callback_listener frame_listener = {
+ frame_handle_done,
+};
+
+static VkResult
+wsi_wl_swapchain_queue_present(struct anv_swapchain *anv_chain,
+ struct anv_queue *queue,
+ uint32_t image_index)
+{
+ struct wsi_wl_swapchain *chain = (struct wsi_wl_swapchain *)anv_chain;
+
+ if (chain->present_mode == VK_PRESENT_MODE_FIFO_KHR) {
+ while (!chain->fifo_ready) {
+ int ret = wl_display_dispatch_queue(chain->display->display,
+ chain->queue);
+ if (ret < 0)
+ return vk_error(VK_ERROR_OUT_OF_DATE_KHR);
+ }
+ }
+
+ assert(image_index < chain->image_count);
+ wl_surface_attach(chain->surface, chain->images[image_index].buffer, 0, 0);
+ wl_surface_damage(chain->surface, 0, 0, INT32_MAX, INT32_MAX);
+
+ if (chain->present_mode == VK_PRESENT_MODE_FIFO_KHR) {
+ struct wl_callback *frame = wl_surface_frame(chain->surface);
+ wl_proxy_set_queue((struct wl_proxy *)frame, chain->queue);
+ wl_callback_add_listener(frame, &frame_listener, chain);
+ chain->fifo_ready = false;
+ }
+
+ wl_surface_commit(chain->surface);
+ wl_display_flush(chain->display->display);
+
+ return VK_SUCCESS;
+}
+
+static void
+wsi_wl_image_finish(struct wsi_wl_swapchain *chain, struct wsi_wl_image *image,
+ const VkAllocationCallbacks* pAllocator)
+{
+ VkDevice vk_device = anv_device_to_handle(chain->base.device);
+ anv_FreeMemory(vk_device, anv_device_memory_to_handle(image->memory),
+ pAllocator);
+ anv_DestroyImage(vk_device, anv_image_to_handle(image->image),
+ pAllocator);
+}
+
+static void
+buffer_handle_release(void *data, struct wl_buffer *buffer)
+{
+ struct wsi_wl_image *image = data;
+
+ assert(image->buffer == buffer);
+
+ image->busy = false;
+}
+
+static const struct wl_buffer_listener buffer_listener = {
+ buffer_handle_release,
+};
+
+static VkResult
+wsi_wl_image_init(struct wsi_wl_swapchain *chain, struct wsi_wl_image *image,
+ const VkAllocationCallbacks* pAllocator)
+{
+ VkDevice vk_device = anv_device_to_handle(chain->base.device);
+ VkResult result;
+
+ VkImage vk_image;
+ result = anv_image_create(vk_device,
+ &(struct anv_image_create_info) {
+ .isl_tiling_flags = ISL_TILING_X_BIT,
+ .stride = 0,
+ .vk_info =
+ &(VkImageCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO,
+ .imageType = VK_IMAGE_TYPE_2D,
+ .format = chain->vk_format,
+ .extent = {
+ .width = chain->extent.width,
+ .height = chain->extent.height,
+ .depth = 1
+ },
+ .mipLevels = 1,
+ .arrayLayers = 1,
+ .samples = 1,
+ /* FIXME: Need a way to use X tiling to allow scanout */
+ .tiling = VK_IMAGE_TILING_OPTIMAL,
+ .usage = VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT,
+ .flags = 0,
+ }},
+ pAllocator,
+ &vk_image);
+
+ if (result != VK_SUCCESS)
+ return result;
+
+ image->image = anv_image_from_handle(vk_image);
+ assert(anv_format_is_color(image->image->format));
+
+ struct anv_surface *surface = &image->image->color_surface;
+
+ VkDeviceMemory vk_memory;
+ result = anv_AllocateMemory(vk_device,
+ &(VkMemoryAllocateInfo) {
+ .sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO,
+ .allocationSize = image->image->size,
+ .memoryTypeIndex = 0,
+ },
+ pAllocator,
+ &vk_memory);
+
+ if (result != VK_SUCCESS)
+ goto fail_image;
+
+ image->memory = anv_device_memory_from_handle(vk_memory);
+
+ result = anv_BindImageMemory(vk_device, vk_image, vk_memory, 0);
+
+ if (result != VK_SUCCESS)
+ goto fail_mem;
+
+ int ret = anv_gem_set_tiling(chain->base.device,
+ image->memory->bo.gem_handle,
+ surface->isl.row_pitch, I915_TILING_X);
+ if (ret) {
+ /* FINISHME: Choose a better error. */
+ result = vk_error(VK_ERROR_OUT_OF_DEVICE_MEMORY);
+ goto fail_mem;
+ }
+
+ int fd = anv_gem_handle_to_fd(chain->base.device,
+ image->memory->bo.gem_handle);
+ if (fd == -1) {
+ /* FINISHME: Choose a better error. */
+ result = vk_error(VK_ERROR_OUT_OF_DEVICE_MEMORY);
+ goto fail_mem;
+ }
+
+ image->buffer = wl_drm_create_prime_buffer(chain->display->drm,
+ fd, /* name */
+ chain->extent.width,
+ chain->extent.height,
+ chain->drm_format,
+ surface->offset,
+ surface->isl.row_pitch,
+ 0, 0, 0, 0 /* unused */);
+ wl_display_roundtrip(chain->display->display);
+ close(fd);
+
+ wl_proxy_set_queue((struct wl_proxy *)image->buffer, chain->queue);
+ wl_buffer_add_listener(image->buffer, &buffer_listener, image);
+
+ return VK_SUCCESS;
+
+fail_mem:
+ anv_FreeMemory(vk_device, vk_memory, pAllocator);
+fail_image:
+ anv_DestroyImage(vk_device, vk_image, pAllocator);
+
+ return result;
+}
+
+static VkResult
+wsi_wl_swapchain_destroy(struct anv_swapchain *anv_chain,
+ const VkAllocationCallbacks *pAllocator)
+{
+ struct wsi_wl_swapchain *chain = (struct wsi_wl_swapchain *)anv_chain;
+
+ for (uint32_t i = 0; i < chain->image_count; i++) {
+ if (chain->images[i].buffer)
+ wsi_wl_image_finish(chain, &chain->images[i], pAllocator);
+ }
+
+ anv_free2(&chain->base.device->alloc, pAllocator, chain);
+
+ return VK_SUCCESS;
+}
+
+static VkResult
+wsi_wl_surface_create_swapchain(struct anv_wsi_surface *wsi_surface,
+ struct anv_device *device,
+ const VkSwapchainCreateInfoKHR* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ struct anv_swapchain **swapchain_out)
+{
+ struct wsi_wl_surface *surface = (struct wsi_wl_surface *)wsi_surface;
+ struct wsi_wl_swapchain *chain;
+ VkResult result;
+
+ assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_SWAPCHAIN_CREATE_INFO_KHR);
+
+ int num_images = pCreateInfo->minImageCount;
+
+ assert(num_images >= MIN_NUM_IMAGES);
+
+ /* For true mailbox mode, we need at least 4 images:
+ * 1) One to scan out from
+ * 2) One to have queued for scan-out
+ * 3) One to be currently held by the Wayland compositor
+ * 4) One to render to
+ */
+ if (pCreateInfo->presentMode == VK_PRESENT_MODE_MAILBOX_KHR)
+ num_images = MAX2(num_images, 4);
+
+ size_t size = sizeof(*chain) + num_images * sizeof(chain->images[0]);
+ chain = anv_alloc2(&device->alloc, pAllocator, size, 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (chain == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ chain->base.device = device;
+ chain->base.destroy = wsi_wl_swapchain_destroy;
+ chain->base.get_images = wsi_wl_swapchain_get_images;
+ chain->base.acquire_next_image = wsi_wl_swapchain_acquire_next_image;
+ chain->base.queue_present = wsi_wl_swapchain_queue_present;
+
+ chain->surface = surface->surface;
+ chain->extent = pCreateInfo->imageExtent;
+ chain->vk_format = pCreateInfo->imageFormat;
+ chain->drm_format = wl_drm_format_for_vk_format(chain->vk_format, false);
+
+ chain->present_mode = pCreateInfo->presentMode;
+ chain->fifo_ready = true;
+
+ chain->image_count = num_images;
+
+ /* Mark a bunch of stuff as NULL. This way we can just call
+ * destroy_swapchain for cleanup.
+ */
+ for (uint32_t i = 0; i < chain->image_count; i++)
+ chain->images[i].buffer = NULL;
+ chain->queue = NULL;
+
+ chain->display = wsi_wl_get_display(device->instance, surface->display);
+ if (!chain->display)
+ goto fail;
+
+ chain->queue = wl_display_create_queue(chain->display->display);
+ if (!chain->queue)
+ goto fail;
+
+ for (uint32_t i = 0; i < chain->image_count; i++) {
+ result = wsi_wl_image_init(chain, &chain->images[i], pAllocator);
+ if (result != VK_SUCCESS)
+ goto fail;
+ chain->images[i].busy = false;
+ }
+
+ *swapchain_out = &chain->base;
+
+ return VK_SUCCESS;
+
+fail:
+ wsi_wl_swapchain_destroy(&chain->base, pAllocator);
+
+ return result;
+}
+
+VkResult
+anv_wl_init_wsi(struct anv_instance *instance)
+{
+ struct wsi_wayland *wsi;
+ VkResult result;
+
+ wsi = anv_alloc(&instance->alloc, sizeof(*wsi), 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_INSTANCE);
+ if (!wsi) {
+ result = vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+ goto fail;
+ }
+
+ wsi->instance = instance;
+
+ int ret = pthread_mutex_init(&wsi->mutex, NULL);
+ if (ret != 0) {
+ if (ret == ENOMEM) {
+ result = vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+ } else {
+ /* FINISHME: Choose a better error. */
+ result = vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+ }
+
+ goto fail_alloc;
+ }
+
+ wsi->displays = _mesa_hash_table_create(NULL, _mesa_hash_pointer,
+ _mesa_key_pointer_equal);
+ if (!wsi->displays) {
+ result = vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+ goto fail_mutex;
+ }
+
+ instance->wayland_wsi = wsi;
+
+ return VK_SUCCESS;
+
+fail_mutex:
+ pthread_mutex_destroy(&wsi->mutex);
+
+fail_alloc:
+ anv_free(&instance->alloc, wsi);
+fail:
+ instance->wayland_wsi = NULL;
+
+ return result;
+}
+
+void
+anv_wl_finish_wsi(struct anv_instance *instance)
+{
+ struct wsi_wayland *wsi = instance->wayland_wsi;
+
+ if (wsi) {
+ _mesa_hash_table_destroy(wsi->displays, NULL);
+
+ pthread_mutex_destroy(&wsi->mutex);
+
+ anv_free(&instance->alloc, wsi);
+ }
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <xcb/xcb.h>
+#include <xcb/dri3.h>
+#include <xcb/present.h>
+
+#include "anv_wsi.h"
+
+struct x11_surface {
+ struct anv_wsi_surface base;
+
+ xcb_connection_t *connection;
+ xcb_window_t window;
+};
+
+static const VkSurfaceFormatKHR formats[] = {
+ { .format = VK_FORMAT_B8G8R8A8_UNORM, },
+};
+
+static const VkPresentModeKHR present_modes[] = {
+ VK_PRESENT_MODE_MAILBOX_KHR,
+};
+
+VkBool32 anv_GetPhysicalDeviceXcbPresentationSupportKHR(
+ VkPhysicalDevice physicalDevice,
+ uint32_t queueFamilyIndex,
+ xcb_connection_t* connection,
+ xcb_visualid_t visual_id)
+{
+ anv_finishme("Check that we actually have DRI3");
+ stub_return(true);
+}
+
+static VkResult
+x11_surface_get_capabilities(struct anv_wsi_surface *wsi_surface,
+ struct anv_physical_device *device,
+ VkSurfaceCapabilitiesKHR *caps)
+{
+ struct x11_surface *surface = (struct x11_surface *)wsi_surface;
+
+ xcb_get_geometry_cookie_t cookie = xcb_get_geometry(surface->connection,
+ surface->window);
+ xcb_generic_error_t *err;
+ xcb_get_geometry_reply_t *geom = xcb_get_geometry_reply(surface->connection,
+ cookie, &err);
+ if (geom) {
+ VkExtent2D extent = { geom->width, geom->height };
+ caps->currentExtent = extent;
+ caps->minImageExtent = extent;
+ caps->maxImageExtent = extent;
+ } else {
+ /* This can happen if the client didn't wait for the configure event
+ * to come back from the compositor. In that case, we don't know the
+ * size of the window so we just return valid "I don't know" stuff.
+ */
+ caps->currentExtent = (VkExtent2D) { -1, -1 };
+ caps->minImageExtent = (VkExtent2D) { 1, 1 };
+ caps->maxImageExtent = (VkExtent2D) { INT16_MAX, INT16_MAX };
+ }
+ free(err);
+ free(geom);
+
+ caps->minImageCount = 2;
+ caps->maxImageCount = 4;
+ caps->supportedTransforms = VK_SURFACE_TRANSFORM_IDENTITY_BIT_KHR;
+ caps->currentTransform = VK_SURFACE_TRANSFORM_IDENTITY_BIT_KHR;
+ caps->maxImageArrayLayers = 1;
+ caps->supportedCompositeAlpha = VK_COMPOSITE_ALPHA_INHERIT_BIT_KHR;
+ caps->supportedUsageFlags =
+ VK_IMAGE_USAGE_TRANSFER_DST_BIT |
+ VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT;
+
+ return VK_SUCCESS;
+}
+
+static VkResult
+x11_surface_get_formats(struct anv_wsi_surface *surface,
+ struct anv_physical_device *device,
+ uint32_t *pSurfaceFormatCount,
+ VkSurfaceFormatKHR *pSurfaceFormats)
+{
+ if (pSurfaceFormats == NULL) {
+ *pSurfaceFormatCount = ARRAY_SIZE(formats);
+ return VK_SUCCESS;
+ }
+
+ assert(*pSurfaceFormatCount >= ARRAY_SIZE(formats));
+ typed_memcpy(pSurfaceFormats, formats, *pSurfaceFormatCount);
+ *pSurfaceFormatCount = ARRAY_SIZE(formats);
+
+ return VK_SUCCESS;
+}
+
+static VkResult
+x11_surface_get_present_modes(struct anv_wsi_surface *surface,
+ struct anv_physical_device *device,
+ uint32_t *pPresentModeCount,
+ VkPresentModeKHR *pPresentModes)
+{
+ if (pPresentModes == NULL) {
+ *pPresentModeCount = ARRAY_SIZE(present_modes);
+ return VK_SUCCESS;
+ }
+
+ assert(*pPresentModeCount >= ARRAY_SIZE(present_modes));
+ typed_memcpy(pPresentModes, present_modes, *pPresentModeCount);
+ *pPresentModeCount = ARRAY_SIZE(present_modes);
+
+ return VK_SUCCESS;
+}
+
+static void
+x11_surface_destroy(struct anv_wsi_surface *surface,
+ const VkAllocationCallbacks *pAllocator)
+{
+ anv_free2(&surface->instance->alloc, pAllocator, surface);
+}
+
+static VkResult
+x11_surface_create_swapchain(struct anv_wsi_surface *surface,
+ struct anv_device *device,
+ const VkSwapchainCreateInfoKHR* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ struct anv_swapchain **swapchain);
+
+VkResult anv_CreateXcbSurfaceKHR(
+ VkInstance _instance,
+ const VkXcbSurfaceCreateInfoKHR* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkSurfaceKHR* pSurface)
+{
+ assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_XCB_SURFACE_CREATE_INFO_KHR);
+
+ ANV_FROM_HANDLE(anv_instance, instance, _instance);
+ struct x11_surface *surface;
+
+ surface = anv_alloc2(&instance->alloc, pAllocator, sizeof *surface, 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (surface == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ surface->connection = pCreateInfo->connection;
+ surface->window = pCreateInfo->window;
+
+ surface->base.instance = instance;
+ surface->base.destroy = x11_surface_destroy;
+ surface->base.get_capabilities = x11_surface_get_capabilities;
+ surface->base.get_formats = x11_surface_get_formats;
+ surface->base.get_present_modes = x11_surface_get_present_modes;
+ surface->base.create_swapchain = x11_surface_create_swapchain;
+
+ *pSurface = anv_wsi_surface_to_handle(&surface->base);
+
+ return VK_SUCCESS;
+}
+
+struct x11_image {
+ struct anv_image * image;
+ struct anv_device_memory * memory;
+ xcb_pixmap_t pixmap;
+ xcb_get_geometry_cookie_t geom_cookie;
+ bool busy;
+};
+
+struct x11_swapchain {
+ struct anv_swapchain base;
+
+ xcb_connection_t * conn;
+ xcb_window_t window;
+ xcb_gc_t gc;
+ VkExtent2D extent;
+ uint32_t image_count;
+ uint32_t next_image;
+ struct x11_image images[0];
+};
+
+static VkResult
+x11_get_images(struct anv_swapchain *anv_chain,
+ uint32_t* pCount, VkImage *pSwapchainImages)
+{
+ struct x11_swapchain *chain = (struct x11_swapchain *)anv_chain;
+
+ if (pSwapchainImages == NULL) {
+ *pCount = chain->image_count;
+ return VK_SUCCESS;
+ }
+
+ assert(chain->image_count <= *pCount);
+ for (uint32_t i = 0; i < chain->image_count; i++)
+ pSwapchainImages[i] = anv_image_to_handle(chain->images[i].image);
+
+ *pCount = chain->image_count;
+
+ return VK_SUCCESS;
+}
+
+static VkResult
+x11_acquire_next_image(struct anv_swapchain *anv_chain,
+ uint64_t timeout,
+ VkSemaphore semaphore,
+ uint32_t *image_index)
+{
+ struct x11_swapchain *chain = (struct x11_swapchain *)anv_chain;
+ struct x11_image *image = &chain->images[chain->next_image];
+
+ if (image->busy) {
+ xcb_generic_error_t *err;
+ xcb_get_geometry_reply_t *geom =
+ xcb_get_geometry_reply(chain->conn, image->geom_cookie, &err);
+ if (!geom) {
+ free(err);
+ return vk_error(VK_ERROR_OUT_OF_DATE_KHR);
+ }
+
+ if (geom->width != chain->extent.width ||
+ geom->height != chain->extent.height) {
+ free(geom);
+ return vk_error(VK_ERROR_OUT_OF_DATE_KHR);
+ }
+ free(geom);
+
+ image->busy = false;
+ }
+
+ *image_index = chain->next_image;
+ chain->next_image = (chain->next_image + 1) % chain->image_count;
+ return VK_SUCCESS;
+}
+
+static VkResult
+x11_queue_present(struct anv_swapchain *anv_chain,
+ struct anv_queue *queue,
+ uint32_t image_index)
+{
+ struct x11_swapchain *chain = (struct x11_swapchain *)anv_chain;
+ struct x11_image *image = &chain->images[image_index];
+
+ assert(image_index < chain->image_count);
+
+ xcb_void_cookie_t cookie;
+
+ cookie = xcb_copy_area(chain->conn,
+ image->pixmap,
+ chain->window,
+ chain->gc,
+ 0, 0,
+ 0, 0,
+ chain->extent.width,
+ chain->extent.height);
+ xcb_discard_reply(chain->conn, cookie.sequence);
+
+ image->geom_cookie = xcb_get_geometry(chain->conn, chain->window);
+ image->busy = true;
+
+ xcb_flush(chain->conn);
+
+ return VK_SUCCESS;
+}
+
+static VkResult
+x11_swapchain_destroy(struct anv_swapchain *anv_chain,
+ const VkAllocationCallbacks *pAllocator)
+{
+ struct x11_swapchain *chain = (struct x11_swapchain *)anv_chain;
+ xcb_void_cookie_t cookie;
+
+ for (uint32_t i = 0; i < chain->image_count; i++) {
+ struct x11_image *image = &chain->images[i];
+
+ if (image->busy)
+ xcb_discard_reply(chain->conn, image->geom_cookie.sequence);
+
+ cookie = xcb_free_pixmap(chain->conn, image->pixmap);
+ xcb_discard_reply(chain->conn, cookie.sequence);
+
+ /* TODO: Delete images and free memory */
+ }
+
+ anv_free(NULL /* XXX: pAllocator */, chain);
+
+ return VK_SUCCESS;
+}
+
+static VkResult
+x11_surface_create_swapchain(struct anv_wsi_surface *wsi_surface,
+ struct anv_device *device,
+ const VkSwapchainCreateInfoKHR *pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ struct anv_swapchain **swapchain_out)
+{
+ struct x11_surface *surface = (struct x11_surface *)wsi_surface;
+ struct x11_swapchain *chain;
+ xcb_void_cookie_t cookie;
+ VkResult result;
+
+ int num_images = pCreateInfo->minImageCount;
+
+ assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_SWAPCHAIN_CREATE_INFO_KHR);
+
+ size_t size = sizeof(*chain) + num_images * sizeof(chain->images[0]);
+ chain = anv_alloc2(&device->alloc, pAllocator, size, 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (chain == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ chain->base.device = device;
+ chain->base.destroy = x11_swapchain_destroy;
+ chain->base.get_images = x11_get_images;
+ chain->base.acquire_next_image = x11_acquire_next_image;
+ chain->base.queue_present = x11_queue_present;
+
+ chain->conn = surface->connection;
+ chain->window = surface->window;
+ chain->extent = pCreateInfo->imageExtent;
+ chain->image_count = num_images;
+ chain->next_image = 0;
+
+ for (uint32_t i = 0; i < chain->image_count; i++) {
+ VkDeviceMemory memory_h;
+ VkImage image_h;
+ struct anv_image *image;
+ struct anv_surface *surface;
+ struct anv_device_memory *memory;
+
+ anv_image_create(anv_device_to_handle(device),
+ &(struct anv_image_create_info) {
+ .isl_tiling_flags = ISL_TILING_X_BIT,
+ .stride = 0,
+ .vk_info =
+ &(VkImageCreateInfo) {
+ .sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO,
+ .imageType = VK_IMAGE_TYPE_2D,
+ .format = pCreateInfo->imageFormat,
+ .extent = {
+ .width = pCreateInfo->imageExtent.width,
+ .height = pCreateInfo->imageExtent.height,
+ .depth = 1
+ },
+ .mipLevels = 1,
+ .arrayLayers = 1,
+ .samples = 1,
+ /* FIXME: Need a way to use X tiling to allow scanout */
+ .tiling = VK_IMAGE_TILING_OPTIMAL,
+ .usage = VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT,
+ .flags = 0,
+ }},
+ NULL,
+ &image_h);
+
+ image = anv_image_from_handle(image_h);
+ assert(anv_format_is_color(image->format));
+
+ surface = &image->color_surface;
+
+ anv_AllocateMemory(anv_device_to_handle(device),
+ &(VkMemoryAllocateInfo) {
+ .sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO,
+ .allocationSize = image->size,
+ .memoryTypeIndex = 0,
+ },
+ NULL /* XXX: pAllocator */,
+ &memory_h);
+
+ memory = anv_device_memory_from_handle(memory_h);
+
+ anv_BindImageMemory(VK_NULL_HANDLE, anv_image_to_handle(image),
+ memory_h, 0);
+
+ int ret = anv_gem_set_tiling(device, memory->bo.gem_handle,
+ surface->isl.row_pitch, I915_TILING_X);
+ if (ret) {
+ /* FINISHME: Choose a better error. */
+ result = vk_errorf(VK_ERROR_OUT_OF_DEVICE_MEMORY,
+ "set_tiling failed: %m");
+ goto fail;
+ }
+
+ int fd = anv_gem_handle_to_fd(device, memory->bo.gem_handle);
+ if (fd == -1) {
+ /* FINISHME: Choose a better error. */
+ result = vk_errorf(VK_ERROR_OUT_OF_DEVICE_MEMORY,
+ "handle_to_fd failed: %m");
+ goto fail;
+ }
+
+ uint32_t bpp = 32;
+ uint32_t depth = 24;
+ xcb_pixmap_t pixmap = xcb_generate_id(chain->conn);
+
+ cookie =
+ xcb_dri3_pixmap_from_buffer_checked(chain->conn,
+ pixmap,
+ chain->window,
+ image->size,
+ pCreateInfo->imageExtent.width,
+ pCreateInfo->imageExtent.height,
+ surface->isl.row_pitch,
+ depth, bpp, fd);
+
+ chain->images[i].image = image;
+ chain->images[i].memory = memory;
+ chain->images[i].pixmap = pixmap;
+ chain->images[i].busy = false;
+
+ xcb_discard_reply(chain->conn, cookie.sequence);
+ }
+
+ chain->gc = xcb_generate_id(chain->conn);
+ if (!chain->gc) {
+ /* FINISHME: Choose a better error. */
+ result = vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+ goto fail;
+ }
+
+ cookie = xcb_create_gc(chain->conn,
+ chain->gc,
+ chain->window,
+ XCB_GC_GRAPHICS_EXPOSURES,
+ (uint32_t []) { 0 });
+ xcb_discard_reply(chain->conn, cookie.sequence);
+
+ *swapchain_out = &chain->base;
+
+ return VK_SUCCESS;
+
+ fail:
+ return result;
+}
+
+VkResult
+anv_x11_init_wsi(struct anv_instance *instance)
+{
+ return VK_SUCCESS;
+}
+
+void
+anv_x11_finish_wsi(struct anv_instance *instance)
+{ }
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+
+/* Instructions, enums and structures for HSW.
+ *
+ * This file has been generated, do not hand edit.
+ */
+
+#pragma once
+
+#include <stdio.h>
+#include <assert.h>
+
+#ifndef __gen_validate_value
+#define __gen_validate_value(x)
+#endif
+
+#ifndef __gen_field_functions
+#define __gen_field_functions
+
+union __gen_value {
+ float f;
+ uint32_t dw;
+};
+
+static inline uint64_t
+__gen_mbo(uint32_t start, uint32_t end)
+{
+ return (~0ul >> (64 - (end - start + 1))) << start;
+}
+
+static inline uint64_t
+__gen_field(uint64_t v, uint32_t start, uint32_t end)
+{
+ __gen_validate_value(v);
+#if DEBUG
+ if (end - start + 1 < 64)
+ assert(v < 1ul << (end - start + 1));
+#endif
+
+ return v << start;
+}
+
+static inline uint64_t
+__gen_fixed(float v, uint32_t start, uint32_t end,
+ bool is_signed, uint32_t fract_bits)
+{
+ __gen_validate_value(v);
+
+ const float factor = (1 << fract_bits);
+
+ float max, min;
+ if (is_signed) {
+ max = ((1 << (end - start)) - 1) / factor;
+ min = -(1 << (end - start)) / factor;
+ } else {
+ max = ((1 << (end - start + 1)) - 1) / factor;
+ min = 0.0f;
+ }
+
+ if (v > max)
+ v = max;
+ else if (v < min)
+ v = min;
+
+ int32_t int_val = roundf(v * factor);
+
+ if (is_signed)
+ int_val &= (1 << (end - start + 1)) - 1;
+
+ return int_val << start;
+}
+
+static inline uint64_t
+__gen_offset(uint64_t v, uint32_t start, uint32_t end)
+{
+ __gen_validate_value(v);
+#if DEBUG
+ uint64_t mask = (~0ul >> (64 - (end - start + 1))) << start;
+
+ assert((v & ~mask) == 0);
+#endif
+
+ return v;
+}
+
+static inline uint32_t
+__gen_float(float v)
+{
+ __gen_validate_value(v);
+ return ((union __gen_value) { .f = (v) }).dw;
+}
+
+#ifndef __gen_address_type
+#error #define __gen_address_type before including this file
+#endif
+
+#ifndef __gen_user_data
+#error #define __gen_combine_address before including this file
+#endif
+
+#endif
+
+#define GEN75_3DSTATE_URB_VS_length_bias 0x00000002
+#define GEN75_3DSTATE_URB_VS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 48, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_URB_VS_length 0x00000002
+
+struct GEN75_3DSTATE_URB_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t VSURBStartingAddress;
+ uint32_t VSURBEntryAllocationSize;
+ uint32_t VSNumberofURBEntries;
+};
+
+static inline void
+GEN75_3DSTATE_URB_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_URB_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->VSURBStartingAddress, 25, 30) |
+ __gen_field(values->VSURBEntryAllocationSize, 16, 24) |
+ __gen_field(values->VSNumberofURBEntries, 0, 15) |
+ 0;
+
+}
+
+#define GEN75_GPGPU_CSR_BASE_ADDRESS_length_bias 0x00000002
+#define GEN75_GPGPU_CSR_BASE_ADDRESS_header \
+ .CommandType = 3, \
+ .CommandSubType = 0, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 4, \
+ .DwordLength = 0
+
+#define GEN75_GPGPU_CSR_BASE_ADDRESS_length 0x00000002
+
+struct GEN75_GPGPU_CSR_BASE_ADDRESS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ __gen_address_type GPGPUCSRBaseAddress;
+};
+
+static inline void
+GEN75_GPGPU_CSR_BASE_ADDRESS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_GPGPU_CSR_BASE_ADDRESS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw1 =
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->GPGPUCSRBaseAddress, dw1);
+
+}
+
+#define GEN75_MI_STORE_REGISTER_MEM_length_bias 0x00000002
+#define GEN75_MI_STORE_REGISTER_MEM_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 36, \
+ .DwordLength = 1
+
+#define GEN75_MI_STORE_REGISTER_MEM_length 0x00000003
+
+struct GEN75_MI_STORE_REGISTER_MEM {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool UseGlobalGTT;
+ uint32_t PredicateEnable;
+ uint32_t DwordLength;
+ uint32_t RegisterAddress;
+ __gen_address_type MemoryAddress;
+};
+
+static inline void
+GEN75_MI_STORE_REGISTER_MEM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_STORE_REGISTER_MEM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->UseGlobalGTT, 22, 22) |
+ __gen_field(values->PredicateEnable, 21, 21) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->RegisterAddress, 2, 22) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->MemoryAddress, dw2);
+
+}
+
+#define GEN75_PIPELINE_SELECT_length_bias 0x00000001
+#define GEN75_PIPELINE_SELECT_header \
+ .CommandType = 3, \
+ .CommandSubType = 1, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 4
+
+#define GEN75_PIPELINE_SELECT_length 0x00000001
+
+struct GEN75_PIPELINE_SELECT {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+#define _3D 0
+#define Media 1
+#define GPGPU 2
+ uint32_t PipelineSelection;
+};
+
+static inline void
+GEN75_PIPELINE_SELECT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_PIPELINE_SELECT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->PipelineSelection, 0, 1) |
+ 0;
+
+}
+
+#define GEN75_STATE_BASE_ADDRESS_length_bias 0x00000002
+#define GEN75_STATE_BASE_ADDRESS_header \
+ .CommandType = 3, \
+ .CommandSubType = 0, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 1, \
+ .DwordLength = 8
+
+#define GEN75_STATE_BASE_ADDRESS_length 0x0000000a
+
+#define GEN75_MEMORY_OBJECT_CONTROL_STATE_length 0x00000001
+
+struct GEN75_MEMORY_OBJECT_CONTROL_STATE {
+ uint32_t LLCeLLCCacheabilityControlLLCCC;
+ uint32_t L3CacheabilityControlL3CC;
+};
+
+static inline void
+GEN75_MEMORY_OBJECT_CONTROL_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MEMORY_OBJECT_CONTROL_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->LLCeLLCCacheabilityControlLLCCC, 1, 2) |
+ __gen_field(values->L3CacheabilityControlL3CC, 0, 0) |
+ 0;
+
+}
+
+struct GEN75_STATE_BASE_ADDRESS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ __gen_address_type GeneralStateBaseAddress;
+ struct GEN75_MEMORY_OBJECT_CONTROL_STATE GeneralStateMemoryObjectControlState;
+ struct GEN75_MEMORY_OBJECT_CONTROL_STATE StatelessDataPortAccessMemoryObjectControlState;
+ bool GeneralStateBaseAddressModifyEnable;
+ __gen_address_type SurfaceStateBaseAddress;
+ struct GEN75_MEMORY_OBJECT_CONTROL_STATE SurfaceStateMemoryObjectControlState;
+ bool SurfaceStateBaseAddressModifyEnable;
+ __gen_address_type DynamicStateBaseAddress;
+ struct GEN75_MEMORY_OBJECT_CONTROL_STATE DynamicStateMemoryObjectControlState;
+ bool DynamicStateBaseAddressModifyEnable;
+ __gen_address_type IndirectObjectBaseAddress;
+ struct GEN75_MEMORY_OBJECT_CONTROL_STATE IndirectObjectMemoryObjectControlState;
+ bool IndirectObjectBaseAddressModifyEnable;
+ __gen_address_type InstructionBaseAddress;
+ struct GEN75_MEMORY_OBJECT_CONTROL_STATE InstructionMemoryObjectControlState;
+ bool InstructionBaseAddressModifyEnable;
+ __gen_address_type GeneralStateAccessUpperBound;
+ bool GeneralStateAccessUpperBoundModifyEnable;
+ __gen_address_type DynamicStateAccessUpperBound;
+ bool DynamicStateAccessUpperBoundModifyEnable;
+ __gen_address_type IndirectObjectAccessUpperBound;
+ bool IndirectObjectAccessUpperBoundModifyEnable;
+ __gen_address_type InstructionAccessUpperBound;
+ bool InstructionAccessUpperBoundModifyEnable;
+};
+
+static inline void
+GEN75_STATE_BASE_ADDRESS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_STATE_BASE_ADDRESS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw_GeneralStateMemoryObjectControlState;
+ GEN75_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_GeneralStateMemoryObjectControlState, &values->GeneralStateMemoryObjectControlState);
+ uint32_t dw_StatelessDataPortAccessMemoryObjectControlState;
+ GEN75_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_StatelessDataPortAccessMemoryObjectControlState, &values->StatelessDataPortAccessMemoryObjectControlState);
+ uint32_t dw1 =
+ __gen_field(dw_GeneralStateMemoryObjectControlState, 8, 11) |
+ __gen_field(dw_StatelessDataPortAccessMemoryObjectControlState, 4, 7) |
+ __gen_field(values->GeneralStateBaseAddressModifyEnable, 0, 0) |
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->GeneralStateBaseAddress, dw1);
+
+ uint32_t dw_SurfaceStateMemoryObjectControlState;
+ GEN75_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_SurfaceStateMemoryObjectControlState, &values->SurfaceStateMemoryObjectControlState);
+ uint32_t dw2 =
+ __gen_field(dw_SurfaceStateMemoryObjectControlState, 8, 11) |
+ __gen_field(values->SurfaceStateBaseAddressModifyEnable, 0, 0) |
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->SurfaceStateBaseAddress, dw2);
+
+ uint32_t dw_DynamicStateMemoryObjectControlState;
+ GEN75_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_DynamicStateMemoryObjectControlState, &values->DynamicStateMemoryObjectControlState);
+ uint32_t dw3 =
+ __gen_field(dw_DynamicStateMemoryObjectControlState, 8, 11) |
+ __gen_field(values->DynamicStateBaseAddressModifyEnable, 0, 0) |
+ 0;
+
+ dw[3] =
+ __gen_combine_address(data, &dw[3], values->DynamicStateBaseAddress, dw3);
+
+ uint32_t dw_IndirectObjectMemoryObjectControlState;
+ GEN75_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_IndirectObjectMemoryObjectControlState, &values->IndirectObjectMemoryObjectControlState);
+ uint32_t dw4 =
+ __gen_field(dw_IndirectObjectMemoryObjectControlState, 8, 11) |
+ __gen_field(values->IndirectObjectBaseAddressModifyEnable, 0, 0) |
+ 0;
+
+ dw[4] =
+ __gen_combine_address(data, &dw[4], values->IndirectObjectBaseAddress, dw4);
+
+ uint32_t dw_InstructionMemoryObjectControlState;
+ GEN75_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_InstructionMemoryObjectControlState, &values->InstructionMemoryObjectControlState);
+ uint32_t dw5 =
+ __gen_field(dw_InstructionMemoryObjectControlState, 8, 11) |
+ __gen_field(values->InstructionBaseAddressModifyEnable, 0, 0) |
+ 0;
+
+ dw[5] =
+ __gen_combine_address(data, &dw[5], values->InstructionBaseAddress, dw5);
+
+ uint32_t dw6 =
+ __gen_field(values->GeneralStateAccessUpperBoundModifyEnable, 0, 0) |
+ 0;
+
+ dw[6] =
+ __gen_combine_address(data, &dw[6], values->GeneralStateAccessUpperBound, dw6);
+
+ uint32_t dw7 =
+ __gen_field(values->DynamicStateAccessUpperBoundModifyEnable, 0, 0) |
+ 0;
+
+ dw[7] =
+ __gen_combine_address(data, &dw[7], values->DynamicStateAccessUpperBound, dw7);
+
+ uint32_t dw8 =
+ __gen_field(values->IndirectObjectAccessUpperBoundModifyEnable, 0, 0) |
+ 0;
+
+ dw[8] =
+ __gen_combine_address(data, &dw[8], values->IndirectObjectAccessUpperBound, dw8);
+
+ uint32_t dw9 =
+ __gen_field(values->InstructionAccessUpperBoundModifyEnable, 0, 0) |
+ 0;
+
+ dw[9] =
+ __gen_combine_address(data, &dw[9], values->InstructionAccessUpperBound, dw9);
+
+}
+
+#define GEN75_STATE_PREFETCH_length_bias 0x00000002
+#define GEN75_STATE_PREFETCH_header \
+ .CommandType = 3, \
+ .CommandSubType = 0, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 3, \
+ .DwordLength = 0
+
+#define GEN75_STATE_PREFETCH_length 0x00000002
+
+struct GEN75_STATE_PREFETCH {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ __gen_address_type PrefetchPointer;
+ uint32_t PrefetchCount;
+};
+
+static inline void
+GEN75_STATE_PREFETCH_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_STATE_PREFETCH * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw1 =
+ __gen_field(values->PrefetchCount, 0, 2) |
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->PrefetchPointer, dw1);
+
+}
+
+#define GEN75_STATE_SIP_length_bias 0x00000002
+#define GEN75_STATE_SIP_header \
+ .CommandType = 3, \
+ .CommandSubType = 0, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 2, \
+ .DwordLength = 0
+
+#define GEN75_STATE_SIP_length 0x00000002
+
+struct GEN75_STATE_SIP {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t SystemInstructionPointer;
+};
+
+static inline void
+GEN75_STATE_SIP_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_STATE_SIP * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->SystemInstructionPointer, 4, 31) |
+ 0;
+
+}
+
+#define GEN75_SWTESS_BASE_ADDRESS_length_bias 0x00000002
+#define GEN75_SWTESS_BASE_ADDRESS_header \
+ .CommandType = 3, \
+ .CommandSubType = 0, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 3, \
+ .DwordLength = 0
+
+#define GEN75_SWTESS_BASE_ADDRESS_length 0x00000002
+
+struct GEN75_SWTESS_BASE_ADDRESS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ __gen_address_type SWTessellationBaseAddress;
+ struct GEN75_MEMORY_OBJECT_CONTROL_STATE SWTessellationMemoryObjectControlState;
+};
+
+static inline void
+GEN75_SWTESS_BASE_ADDRESS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_SWTESS_BASE_ADDRESS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw_SWTessellationMemoryObjectControlState;
+ GEN75_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_SWTessellationMemoryObjectControlState, &values->SWTessellationMemoryObjectControlState);
+ uint32_t dw1 =
+ __gen_field(dw_SWTessellationMemoryObjectControlState, 8, 11) |
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->SWTessellationBaseAddress, dw1);
+
+}
+
+#define GEN75_3DPRIMITIVE_length_bias 0x00000002
+#define GEN75_3DPRIMITIVE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 3, \
+ ._3DCommandSubOpcode = 0, \
+ .DwordLength = 5
+
+#define GEN75_3DPRIMITIVE_length 0x00000007
+
+struct GEN75_3DPRIMITIVE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ bool IndirectParameterEnable;
+ uint32_t UAVCoherencyRequired;
+ bool PredicateEnable;
+ uint32_t DwordLength;
+ bool EndOffsetEnable;
+#define SEQUENTIAL 0
+#define RANDOM 1
+ uint32_t VertexAccessType;
+ uint32_t PrimitiveTopologyType;
+ uint32_t VertexCountPerInstance;
+ uint32_t StartVertexLocation;
+ uint32_t InstanceCount;
+ uint32_t StartInstanceLocation;
+ uint32_t BaseVertexLocation;
+};
+
+static inline void
+GEN75_3DPRIMITIVE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DPRIMITIVE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->IndirectParameterEnable, 10, 10) |
+ __gen_field(values->UAVCoherencyRequired, 9, 9) |
+ __gen_field(values->PredicateEnable, 8, 8) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->EndOffsetEnable, 9, 9) |
+ __gen_field(values->VertexAccessType, 8, 8) |
+ __gen_field(values->PrimitiveTopologyType, 0, 5) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->VertexCountPerInstance, 0, 31) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->StartVertexLocation, 0, 31) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->InstanceCount, 0, 31) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->StartInstanceLocation, 0, 31) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->BaseVertexLocation, 0, 31) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_AA_LINE_PARAMETERS_length_bias 0x00000002
+#define GEN75_3DSTATE_AA_LINE_PARAMETERS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 10, \
+ .DwordLength = 1
+
+#define GEN75_3DSTATE_AA_LINE_PARAMETERS_length 0x00000003
+
+struct GEN75_3DSTATE_AA_LINE_PARAMETERS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ float AACoverageBias;
+ float AACoverageSlope;
+ float AACoverageEndCapBias;
+ float AACoverageEndCapSlope;
+};
+
+static inline void
+GEN75_3DSTATE_AA_LINE_PARAMETERS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_AA_LINE_PARAMETERS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->AACoverageBias * (1 << 8), 16, 23) |
+ __gen_field(values->AACoverageSlope * (1 << 8), 0, 7) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->AACoverageEndCapBias * (1 << 8), 16, 23) |
+ __gen_field(values->AACoverageEndCapSlope * (1 << 8), 0, 7) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_BINDING_TABLE_EDIT_DS_length_bias 0x00000002
+#define GEN75_3DSTATE_BINDING_TABLE_EDIT_DS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 70
+
+#define GEN75_3DSTATE_BINDING_TABLE_EDIT_DS_length 0x00000000
+
+#define GEN75_BINDING_TABLE_EDIT_ENTRY_length 0x00000001
+
+struct GEN75_BINDING_TABLE_EDIT_ENTRY {
+ uint32_t BindingTableIndex;
+ uint32_t SurfaceStatePointer;
+};
+
+static inline void
+GEN75_BINDING_TABLE_EDIT_ENTRY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_BINDING_TABLE_EDIT_ENTRY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->BindingTableIndex, 16, 23) |
+ __gen_offset(values->SurfaceStatePointer, 0, 15) |
+ 0;
+
+}
+
+struct GEN75_3DSTATE_BINDING_TABLE_EDIT_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t BindingTableBlockClear;
+#define AllCores 3
+#define Core1 2
+#define Core0 1
+ uint32_t BindingTableEditTarget;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN75_3DSTATE_BINDING_TABLE_EDIT_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_BINDING_TABLE_EDIT_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 8) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->BindingTableBlockClear, 16, 31) |
+ __gen_field(values->BindingTableEditTarget, 0, 1) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN75_3DSTATE_BINDING_TABLE_EDIT_GS_length_bias 0x00000002
+#define GEN75_3DSTATE_BINDING_TABLE_EDIT_GS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 68
+
+#define GEN75_3DSTATE_BINDING_TABLE_EDIT_GS_length 0x00000000
+
+struct GEN75_3DSTATE_BINDING_TABLE_EDIT_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t BindingTableBlockClear;
+#define AllCores 3
+#define Core1 2
+#define Core0 1
+ uint32_t BindingTableEditTarget;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN75_3DSTATE_BINDING_TABLE_EDIT_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_BINDING_TABLE_EDIT_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 8) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->BindingTableBlockClear, 16, 31) |
+ __gen_field(values->BindingTableEditTarget, 0, 1) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN75_3DSTATE_BINDING_TABLE_EDIT_HS_length_bias 0x00000002
+#define GEN75_3DSTATE_BINDING_TABLE_EDIT_HS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 69
+
+#define GEN75_3DSTATE_BINDING_TABLE_EDIT_HS_length 0x00000000
+
+struct GEN75_3DSTATE_BINDING_TABLE_EDIT_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t BindingTableBlockClear;
+#define AllCores 3
+#define Core1 2
+#define Core0 1
+ uint32_t BindingTableEditTarget;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN75_3DSTATE_BINDING_TABLE_EDIT_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_BINDING_TABLE_EDIT_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 8) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->BindingTableBlockClear, 16, 31) |
+ __gen_field(values->BindingTableEditTarget, 0, 1) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN75_3DSTATE_BINDING_TABLE_EDIT_PS_length_bias 0x00000002
+#define GEN75_3DSTATE_BINDING_TABLE_EDIT_PS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 71
+
+#define GEN75_3DSTATE_BINDING_TABLE_EDIT_PS_length 0x00000000
+
+struct GEN75_3DSTATE_BINDING_TABLE_EDIT_PS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t BindingTableBlockClear;
+#define AllCores 3
+#define Core1 2
+#define Core0 1
+ uint32_t BindingTableEditTarget;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN75_3DSTATE_BINDING_TABLE_EDIT_PS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_BINDING_TABLE_EDIT_PS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 8) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->BindingTableBlockClear, 16, 31) |
+ __gen_field(values->BindingTableEditTarget, 0, 1) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN75_3DSTATE_BINDING_TABLE_EDIT_VS_length_bias 0x00000002
+#define GEN75_3DSTATE_BINDING_TABLE_EDIT_VS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 67
+
+#define GEN75_3DSTATE_BINDING_TABLE_EDIT_VS_length 0x00000000
+
+struct GEN75_3DSTATE_BINDING_TABLE_EDIT_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t BindingTableBlockClear;
+#define AllCores 3
+#define Core1 2
+#define Core0 1
+ uint32_t BindingTableEditTarget;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN75_3DSTATE_BINDING_TABLE_EDIT_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_BINDING_TABLE_EDIT_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 8) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->BindingTableBlockClear, 16, 31) |
+ __gen_field(values->BindingTableEditTarget, 0, 1) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN75_3DSTATE_BINDING_TABLE_POINTERS_DS_length_bias 0x00000002
+#define GEN75_3DSTATE_BINDING_TABLE_POINTERS_DS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 40, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_BINDING_TABLE_POINTERS_DS_length 0x00000002
+
+struct GEN75_3DSTATE_BINDING_TABLE_POINTERS_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoDSBindingTable;
+};
+
+static inline void
+GEN75_3DSTATE_BINDING_TABLE_POINTERS_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_BINDING_TABLE_POINTERS_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoDSBindingTable, 5, 15) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_BINDING_TABLE_POINTERS_GS_length_bias 0x00000002
+#define GEN75_3DSTATE_BINDING_TABLE_POINTERS_GS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 41, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_BINDING_TABLE_POINTERS_GS_length 0x00000002
+
+struct GEN75_3DSTATE_BINDING_TABLE_POINTERS_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoGSBindingTable;
+};
+
+static inline void
+GEN75_3DSTATE_BINDING_TABLE_POINTERS_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_BINDING_TABLE_POINTERS_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoGSBindingTable, 5, 15) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_BINDING_TABLE_POINTERS_HS_length_bias 0x00000002
+#define GEN75_3DSTATE_BINDING_TABLE_POINTERS_HS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 39, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_BINDING_TABLE_POINTERS_HS_length 0x00000002
+
+struct GEN75_3DSTATE_BINDING_TABLE_POINTERS_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoHSBindingTable;
+};
+
+static inline void
+GEN75_3DSTATE_BINDING_TABLE_POINTERS_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_BINDING_TABLE_POINTERS_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoHSBindingTable, 5, 15) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_BINDING_TABLE_POINTERS_PS_length_bias 0x00000002
+#define GEN75_3DSTATE_BINDING_TABLE_POINTERS_PS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 42, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_BINDING_TABLE_POINTERS_PS_length 0x00000002
+
+struct GEN75_3DSTATE_BINDING_TABLE_POINTERS_PS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoPSBindingTable;
+};
+
+static inline void
+GEN75_3DSTATE_BINDING_TABLE_POINTERS_PS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_BINDING_TABLE_POINTERS_PS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoPSBindingTable, 5, 15) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_BINDING_TABLE_POINTERS_VS_length_bias 0x00000002
+#define GEN75_3DSTATE_BINDING_TABLE_POINTERS_VS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 38, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_BINDING_TABLE_POINTERS_VS_length 0x00000002
+
+struct GEN75_3DSTATE_BINDING_TABLE_POINTERS_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoVSBindingTable;
+};
+
+static inline void
+GEN75_3DSTATE_BINDING_TABLE_POINTERS_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_BINDING_TABLE_POINTERS_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoVSBindingTable, 5, 15) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_BINDING_TABLE_POOL_ALLOC_length_bias 0x00000002
+#define GEN75_3DSTATE_BINDING_TABLE_POOL_ALLOC_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 25, \
+ .DwordLength = 1
+
+#define GEN75_3DSTATE_BINDING_TABLE_POOL_ALLOC_length 0x00000003
+
+struct GEN75_3DSTATE_BINDING_TABLE_POOL_ALLOC {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ __gen_address_type BindingTablePoolBaseAddress;
+ uint32_t BindingTablePoolEnable;
+ struct GEN75_MEMORY_OBJECT_CONTROL_STATE SurfaceObjectControlState;
+ __gen_address_type BindingTablePoolUpperBound;
+};
+
+static inline void
+GEN75_3DSTATE_BINDING_TABLE_POOL_ALLOC_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_BINDING_TABLE_POOL_ALLOC * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw_SurfaceObjectControlState;
+ GEN75_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_SurfaceObjectControlState, &values->SurfaceObjectControlState);
+ uint32_t dw1 =
+ __gen_field(values->BindingTablePoolEnable, 11, 11) |
+ __gen_field(dw_SurfaceObjectControlState, 7, 10) |
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->BindingTablePoolBaseAddress, dw1);
+
+ uint32_t dw2 =
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->BindingTablePoolUpperBound, dw2);
+
+}
+
+#define GEN75_3DSTATE_BLEND_STATE_POINTERS_length_bias 0x00000002
+#define GEN75_3DSTATE_BLEND_STATE_POINTERS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 36, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_BLEND_STATE_POINTERS_length 0x00000002
+
+struct GEN75_3DSTATE_BLEND_STATE_POINTERS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t BlendStatePointer;
+};
+
+static inline void
+GEN75_3DSTATE_BLEND_STATE_POINTERS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_BLEND_STATE_POINTERS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->BlendStatePointer, 6, 31) |
+ __gen_mbo(0, 0) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_CC_STATE_POINTERS_length_bias 0x00000002
+#define GEN75_3DSTATE_CC_STATE_POINTERS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 14, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_CC_STATE_POINTERS_length 0x00000002
+
+struct GEN75_3DSTATE_CC_STATE_POINTERS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ColorCalcStatePointer;
+};
+
+static inline void
+GEN75_3DSTATE_CC_STATE_POINTERS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_CC_STATE_POINTERS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->ColorCalcStatePointer, 6, 31) |
+ __gen_mbo(0, 0) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_CHROMA_KEY_length_bias 0x00000002
+#define GEN75_3DSTATE_CHROMA_KEY_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 4, \
+ .DwordLength = 2
+
+#define GEN75_3DSTATE_CHROMA_KEY_length 0x00000004
+
+struct GEN75_3DSTATE_CHROMA_KEY {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ChromaKeyTableIndex;
+ uint32_t ChromaKeyLowValue;
+ uint32_t ChromaKeyHighValue;
+};
+
+static inline void
+GEN75_3DSTATE_CHROMA_KEY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_CHROMA_KEY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ChromaKeyTableIndex, 30, 31) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ChromaKeyLowValue, 0, 31) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->ChromaKeyHighValue, 0, 31) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_CLEAR_PARAMS_length_bias 0x00000002
+#define GEN75_3DSTATE_CLEAR_PARAMS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 4, \
+ .DwordLength = 1
+
+#define GEN75_3DSTATE_CLEAR_PARAMS_length 0x00000003
+
+struct GEN75_3DSTATE_CLEAR_PARAMS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t DepthClearValue;
+ bool DepthClearValueValid;
+};
+
+static inline void
+GEN75_3DSTATE_CLEAR_PARAMS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_CLEAR_PARAMS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->DepthClearValue, 0, 31) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->DepthClearValueValid, 0, 0) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_CLIP_length_bias 0x00000002
+#define GEN75_3DSTATE_CLIP_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 18, \
+ .DwordLength = 2
+
+#define GEN75_3DSTATE_CLIP_length 0x00000004
+
+struct GEN75_3DSTATE_CLIP {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t FrontWinding;
+ uint32_t VertexSubPixelPrecisionSelect;
+ bool EarlyCullEnable;
+#define CULLMODE_BOTH 0
+#define CULLMODE_NONE 1
+#define CULLMODE_FRONT 2
+#define CULLMODE_BACK 3
+ uint32_t CullMode;
+ bool ClipperStatisticsEnable;
+ uint32_t UserClipDistanceCullTestEnableBitmask;
+ bool ClipEnable;
+#define APIMODE_OGL 0
+ uint32_t APIMode;
+ bool ViewportXYClipTestEnable;
+ bool ViewportZClipTestEnable;
+ bool GuardbandClipTestEnable;
+ uint32_t UserClipDistanceClipTestEnableBitmask;
+#define CLIPMODE_NORMAL 0
+#define CLIPMODE_REJECT_ALL 3
+#define CLIPMODE_ACCEPT_ALL 4
+ uint32_t ClipMode;
+ bool PerspectiveDivideDisable;
+ bool NonPerspectiveBarycentricEnable;
+#define Vertex0 0
+#define Vertex1 1
+#define Vertex2 2
+ uint32_t TriangleStripListProvokingVertexSelect;
+#define Vertex0 0
+#define Vertex1 1
+ uint32_t LineStripListProvokingVertexSelect;
+#define Vertex0 0
+#define Vertex1 1
+#define Vertex2 2
+ uint32_t TriangleFanProvokingVertexSelect;
+ float MinimumPointWidth;
+ float MaximumPointWidth;
+ bool ForceZeroRTAIndexEnable;
+ uint32_t MaximumVPIndex;
+};
+
+static inline void
+GEN75_3DSTATE_CLIP_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_CLIP * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->FrontWinding, 20, 20) |
+ __gen_field(values->VertexSubPixelPrecisionSelect, 19, 19) |
+ __gen_field(values->EarlyCullEnable, 18, 18) |
+ __gen_field(values->CullMode, 16, 17) |
+ __gen_field(values->ClipperStatisticsEnable, 10, 10) |
+ __gen_field(values->UserClipDistanceCullTestEnableBitmask, 0, 7) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ClipEnable, 31, 31) |
+ __gen_field(values->APIMode, 30, 30) |
+ __gen_field(values->ViewportXYClipTestEnable, 28, 28) |
+ __gen_field(values->ViewportZClipTestEnable, 27, 27) |
+ __gen_field(values->GuardbandClipTestEnable, 26, 26) |
+ __gen_field(values->UserClipDistanceClipTestEnableBitmask, 16, 23) |
+ __gen_field(values->ClipMode, 13, 15) |
+ __gen_field(values->PerspectiveDivideDisable, 9, 9) |
+ __gen_field(values->NonPerspectiveBarycentricEnable, 8, 8) |
+ __gen_field(values->TriangleStripListProvokingVertexSelect, 4, 5) |
+ __gen_field(values->LineStripListProvokingVertexSelect, 2, 3) |
+ __gen_field(values->TriangleFanProvokingVertexSelect, 0, 1) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->MinimumPointWidth * (1 << 3), 17, 27) |
+ __gen_field(values->MaximumPointWidth * (1 << 3), 6, 16) |
+ __gen_field(values->ForceZeroRTAIndexEnable, 5, 5) |
+ __gen_field(values->MaximumVPIndex, 0, 3) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_CONSTANT_DS_length_bias 0x00000002
+#define GEN75_3DSTATE_CONSTANT_DS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 26, \
+ .DwordLength = 5
+
+#define GEN75_3DSTATE_CONSTANT_DS_length 0x00000007
+
+#define GEN75_3DSTATE_CONSTANT_BODY_length 0x00000006
+
+struct GEN75_3DSTATE_CONSTANT_BODY {
+ uint32_t ConstantBuffer1ReadLength;
+ uint32_t ConstantBuffer0ReadLength;
+ uint32_t ConstantBuffer3ReadLength;
+ uint32_t ConstantBuffer2ReadLength;
+ __gen_address_type PointerToConstantBuffer0;
+ struct GEN75_MEMORY_OBJECT_CONTROL_STATE ConstantBufferObjectControlState;
+ __gen_address_type PointerToConstantBuffer1;
+ __gen_address_type PointerToConstantBuffer2;
+ __gen_address_type PointerToConstantBuffer3;
+};
+
+static inline void
+GEN75_3DSTATE_CONSTANT_BODY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_CONSTANT_BODY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->ConstantBuffer1ReadLength, 16, 31) |
+ __gen_field(values->ConstantBuffer0ReadLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBuffer3ReadLength, 16, 31) |
+ __gen_field(values->ConstantBuffer2ReadLength, 0, 15) |
+ 0;
+
+ uint32_t dw_ConstantBufferObjectControlState;
+ GEN75_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_ConstantBufferObjectControlState, &values->ConstantBufferObjectControlState);
+ uint32_t dw2 =
+ __gen_field(dw_ConstantBufferObjectControlState, 0, 4) |
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->PointerToConstantBuffer0, dw2);
+
+ uint32_t dw3 =
+ 0;
+
+ dw[3] =
+ __gen_combine_address(data, &dw[3], values->PointerToConstantBuffer1, dw3);
+
+ uint32_t dw4 =
+ 0;
+
+ dw[4] =
+ __gen_combine_address(data, &dw[4], values->PointerToConstantBuffer2, dw4);
+
+ uint32_t dw5 =
+ 0;
+
+ dw[5] =
+ __gen_combine_address(data, &dw[5], values->PointerToConstantBuffer3, dw5);
+
+}
+
+struct GEN75_3DSTATE_CONSTANT_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ struct GEN75_3DSTATE_CONSTANT_BODY ConstantBody;
+};
+
+static inline void
+GEN75_3DSTATE_CONSTANT_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_CONSTANT_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ GEN75_3DSTATE_CONSTANT_BODY_pack(data, &dw[1], &values->ConstantBody);
+}
+
+#define GEN75_3DSTATE_CONSTANT_GS_length_bias 0x00000002
+#define GEN75_3DSTATE_CONSTANT_GS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 22, \
+ .DwordLength = 5
+
+#define GEN75_3DSTATE_CONSTANT_GS_length 0x00000007
+
+struct GEN75_3DSTATE_CONSTANT_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ struct GEN75_3DSTATE_CONSTANT_BODY ConstantBody;
+};
+
+static inline void
+GEN75_3DSTATE_CONSTANT_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_CONSTANT_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ GEN75_3DSTATE_CONSTANT_BODY_pack(data, &dw[1], &values->ConstantBody);
+}
+
+#define GEN75_3DSTATE_CONSTANT_HS_length_bias 0x00000002
+#define GEN75_3DSTATE_CONSTANT_HS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 25, \
+ .DwordLength = 5
+
+#define GEN75_3DSTATE_CONSTANT_HS_length 0x00000007
+
+struct GEN75_3DSTATE_CONSTANT_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ struct GEN75_3DSTATE_CONSTANT_BODY ConstantBody;
+};
+
+static inline void
+GEN75_3DSTATE_CONSTANT_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_CONSTANT_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ GEN75_3DSTATE_CONSTANT_BODY_pack(data, &dw[1], &values->ConstantBody);
+}
+
+#define GEN75_3DSTATE_CONSTANT_PS_length_bias 0x00000002
+#define GEN75_3DSTATE_CONSTANT_PS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 23, \
+ .DwordLength = 5
+
+#define GEN75_3DSTATE_CONSTANT_PS_length 0x00000007
+
+struct GEN75_3DSTATE_CONSTANT_PS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ struct GEN75_3DSTATE_CONSTANT_BODY ConstantBody;
+};
+
+static inline void
+GEN75_3DSTATE_CONSTANT_PS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_CONSTANT_PS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ GEN75_3DSTATE_CONSTANT_BODY_pack(data, &dw[1], &values->ConstantBody);
+}
+
+#define GEN75_3DSTATE_CONSTANT_VS_length_bias 0x00000002
+#define GEN75_3DSTATE_CONSTANT_VS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 21, \
+ .DwordLength = 5
+
+#define GEN75_3DSTATE_CONSTANT_VS_length 0x00000007
+
+struct GEN75_3DSTATE_CONSTANT_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ struct GEN75_3DSTATE_CONSTANT_BODY ConstantBody;
+};
+
+static inline void
+GEN75_3DSTATE_CONSTANT_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_CONSTANT_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ GEN75_3DSTATE_CONSTANT_BODY_pack(data, &dw[1], &values->ConstantBody);
+}
+
+#define GEN75_3DSTATE_DEPTH_BUFFER_length_bias 0x00000002
+#define GEN75_3DSTATE_DEPTH_BUFFER_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 5, \
+ .DwordLength = 5
+
+#define GEN75_3DSTATE_DEPTH_BUFFER_length 0x00000007
+
+struct GEN75_3DSTATE_DEPTH_BUFFER {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define SURFTYPE_1D 0
+#define SURFTYPE_2D 1
+#define SURFTYPE_3D 2
+#define SURFTYPE_CUBE 3
+#define SURFTYPE_NULL 7
+ uint32_t SurfaceType;
+ bool DepthWriteEnable;
+ bool StencilWriteEnable;
+ bool HierarchicalDepthBufferEnable;
+#define D32_FLOAT 1
+#define D24_UNORM_X8_UINT 3
+#define D16_UNORM 5
+ uint32_t SurfaceFormat;
+ uint32_t SurfacePitch;
+ __gen_address_type SurfaceBaseAddress;
+ uint32_t Height;
+ uint32_t Width;
+ uint32_t LOD;
+#define SURFTYPE_CUBEmustbezero 0
+ uint32_t Depth;
+ uint32_t MinimumArrayElement;
+ struct GEN75_MEMORY_OBJECT_CONTROL_STATE DepthBufferObjectControlState;
+ uint32_t DepthCoordinateOffsetY;
+ uint32_t DepthCoordinateOffsetX;
+ uint32_t RenderTargetViewExtent;
+};
+
+static inline void
+GEN75_3DSTATE_DEPTH_BUFFER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_DEPTH_BUFFER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->SurfaceType, 29, 31) |
+ __gen_field(values->DepthWriteEnable, 28, 28) |
+ __gen_field(values->StencilWriteEnable, 27, 27) |
+ __gen_field(values->HierarchicalDepthBufferEnable, 22, 22) |
+ __gen_field(values->SurfaceFormat, 18, 20) |
+ __gen_field(values->SurfacePitch, 0, 17) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->SurfaceBaseAddress, dw2);
+
+ dw[3] =
+ __gen_field(values->Height, 18, 31) |
+ __gen_field(values->Width, 4, 17) |
+ __gen_field(values->LOD, 0, 3) |
+ 0;
+
+ uint32_t dw_DepthBufferObjectControlState;
+ GEN75_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_DepthBufferObjectControlState, &values->DepthBufferObjectControlState);
+ dw[4] =
+ __gen_field(values->Depth, 21, 31) |
+ __gen_field(values->MinimumArrayElement, 10, 20) |
+ __gen_field(dw_DepthBufferObjectControlState, 0, 3) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->DepthCoordinateOffsetY, 16, 31) |
+ __gen_field(values->DepthCoordinateOffsetX, 0, 15) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->RenderTargetViewExtent, 21, 31) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_DEPTH_STENCIL_STATE_POINTERS_length_bias 0x00000002
+#define GEN75_3DSTATE_DEPTH_STENCIL_STATE_POINTERS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 37, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_DEPTH_STENCIL_STATE_POINTERS_length 0x00000002
+
+struct GEN75_3DSTATE_DEPTH_STENCIL_STATE_POINTERS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoDEPTH_STENCIL_STATE;
+};
+
+static inline void
+GEN75_3DSTATE_DEPTH_STENCIL_STATE_POINTERS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_DEPTH_STENCIL_STATE_POINTERS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoDEPTH_STENCIL_STATE, 6, 31) |
+ __gen_mbo(0, 0) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_DRAWING_RECTANGLE_length_bias 0x00000002
+#define GEN75_3DSTATE_DRAWING_RECTANGLE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 0, \
+ .DwordLength = 2
+
+#define GEN75_3DSTATE_DRAWING_RECTANGLE_length 0x00000004
+
+struct GEN75_3DSTATE_DRAWING_RECTANGLE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+#define Legacy 0
+#define Core0Enabled 1
+#define Core1Enabled 2
+ uint32_t CoreModeSelect;
+ uint32_t DwordLength;
+ uint32_t ClippedDrawingRectangleYMin;
+ uint32_t ClippedDrawingRectangleXMin;
+ uint32_t ClippedDrawingRectangleYMax;
+ uint32_t ClippedDrawingRectangleXMax;
+ uint32_t DrawingRectangleOriginY;
+ uint32_t DrawingRectangleOriginX;
+};
+
+static inline void
+GEN75_3DSTATE_DRAWING_RECTANGLE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_DRAWING_RECTANGLE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->CoreModeSelect, 14, 15) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ClippedDrawingRectangleYMin, 16, 31) |
+ __gen_field(values->ClippedDrawingRectangleXMin, 0, 15) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ClippedDrawingRectangleYMax, 16, 31) |
+ __gen_field(values->ClippedDrawingRectangleXMax, 0, 15) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->DrawingRectangleOriginY, 16, 31) |
+ __gen_field(values->DrawingRectangleOriginX, 0, 15) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_DS_length_bias 0x00000002
+#define GEN75_3DSTATE_DS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 29, \
+ .DwordLength = 4
+
+#define GEN75_3DSTATE_DS_length 0x00000006
+
+struct GEN75_3DSTATE_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t KernelStartPointer;
+#define Multiple 0
+#define Single 1
+ uint32_t SingleDomainPointDispatch;
+#define Dmask 0
+#define Vmask 1
+ uint32_t VectorMaskEnable;
+#define NoSamplers 0
+#define _14Samplers 1
+#define _58Samplers 2
+#define _912Samplers 3
+#define _1316Samplers 4
+ uint32_t SamplerCount;
+ uint32_t BindingTableEntryCount;
+#define Normal 0
+#define High 1
+ uint32_t ThreadDispatchPriority;
+#define IEEE754 0
+#define Alternate 1
+ uint32_t FloatingPointMode;
+ bool AccessesUAV;
+ bool IllegalOpcodeExceptionEnable;
+ bool SoftwareExceptionEnable;
+ uint32_t ScratchSpaceBasePointer;
+ uint32_t PerThreadScratchSpace;
+ uint32_t DispatchGRFStartRegisterForURBData;
+ uint32_t PatchURBEntryReadLength;
+ uint32_t PatchURBEntryReadOffset;
+ uint32_t MaximumNumberofThreads;
+ bool StatisticsEnable;
+ bool ComputeWCoordinateEnable;
+ bool DSCacheDisable;
+ bool DSFunctionEnable;
+};
+
+static inline void
+GEN75_3DSTATE_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->KernelStartPointer, 6, 31) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->SingleDomainPointDispatch, 31, 31) |
+ __gen_field(values->VectorMaskEnable, 30, 30) |
+ __gen_field(values->SamplerCount, 27, 29) |
+ __gen_field(values->BindingTableEntryCount, 18, 25) |
+ __gen_field(values->ThreadDispatchPriority, 17, 17) |
+ __gen_field(values->FloatingPointMode, 16, 16) |
+ __gen_field(values->AccessesUAV, 14, 14) |
+ __gen_field(values->IllegalOpcodeExceptionEnable, 13, 13) |
+ __gen_field(values->SoftwareExceptionEnable, 7, 7) |
+ 0;
+
+ dw[3] =
+ __gen_offset(values->ScratchSpaceBasePointer, 10, 31) |
+ __gen_field(values->PerThreadScratchSpace, 0, 3) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->DispatchGRFStartRegisterForURBData, 20, 24) |
+ __gen_field(values->PatchURBEntryReadLength, 11, 17) |
+ __gen_field(values->PatchURBEntryReadOffset, 4, 9) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->MaximumNumberofThreads, 21, 29) |
+ __gen_field(values->StatisticsEnable, 10, 10) |
+ __gen_field(values->ComputeWCoordinateEnable, 2, 2) |
+ __gen_field(values->DSCacheDisable, 1, 1) |
+ __gen_field(values->DSFunctionEnable, 0, 0) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_GATHER_CONSTANT_DS_length_bias 0x00000002
+#define GEN75_3DSTATE_GATHER_CONSTANT_DS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 55
+
+#define GEN75_3DSTATE_GATHER_CONSTANT_DS_length 0x00000000
+
+#define GEN75_GATHER_CONSTANT_ENTRY_length 0x00000001
+
+struct GEN75_GATHER_CONSTANT_ENTRY {
+ uint32_t ConstantBufferOffset;
+ uint32_t ChannelMask;
+ uint32_t BindingTableIndexOffset;
+};
+
+static inline void
+GEN75_GATHER_CONSTANT_ENTRY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_GATHER_CONSTANT_ENTRY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_offset(values->ConstantBufferOffset, 8, 15) |
+ __gen_field(values->ChannelMask, 4, 7) |
+ __gen_field(values->BindingTableIndexOffset, 0, 3) |
+ 0;
+
+}
+
+struct GEN75_3DSTATE_GATHER_CONSTANT_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferValid;
+ uint32_t ConstantBufferBindingTableBlock;
+ uint32_t GatherBufferOffset;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN75_3DSTATE_GATHER_CONSTANT_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_GATHER_CONSTANT_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferValid, 16, 31) |
+ __gen_field(values->ConstantBufferBindingTableBlock, 12, 15) |
+ 0;
+
+ dw[2] =
+ __gen_offset(values->GatherBufferOffset, 6, 22) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN75_3DSTATE_GATHER_CONSTANT_GS_length_bias 0x00000002
+#define GEN75_3DSTATE_GATHER_CONSTANT_GS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 53
+
+#define GEN75_3DSTATE_GATHER_CONSTANT_GS_length 0x00000000
+
+struct GEN75_3DSTATE_GATHER_CONSTANT_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferValid;
+ uint32_t ConstantBufferBindingTableBlock;
+ uint32_t GatherBufferOffset;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN75_3DSTATE_GATHER_CONSTANT_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_GATHER_CONSTANT_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferValid, 16, 31) |
+ __gen_field(values->ConstantBufferBindingTableBlock, 12, 15) |
+ 0;
+
+ dw[2] =
+ __gen_offset(values->GatherBufferOffset, 6, 22) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN75_3DSTATE_GATHER_CONSTANT_HS_length_bias 0x00000002
+#define GEN75_3DSTATE_GATHER_CONSTANT_HS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 54
+
+#define GEN75_3DSTATE_GATHER_CONSTANT_HS_length 0x00000000
+
+struct GEN75_3DSTATE_GATHER_CONSTANT_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferValid;
+ uint32_t ConstantBufferBindingTableBlock;
+ uint32_t GatherBufferOffset;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN75_3DSTATE_GATHER_CONSTANT_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_GATHER_CONSTANT_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferValid, 16, 31) |
+ __gen_field(values->ConstantBufferBindingTableBlock, 12, 15) |
+ 0;
+
+ dw[2] =
+ __gen_offset(values->GatherBufferOffset, 6, 22) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN75_3DSTATE_GATHER_CONSTANT_PS_length_bias 0x00000002
+#define GEN75_3DSTATE_GATHER_CONSTANT_PS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 56
+
+#define GEN75_3DSTATE_GATHER_CONSTANT_PS_length 0x00000000
+
+struct GEN75_3DSTATE_GATHER_CONSTANT_PS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferValid;
+ uint32_t ConstantBufferBindingTableBlock;
+ uint32_t GatherBufferOffset;
+ bool ConstantBufferDx9Enable;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN75_3DSTATE_GATHER_CONSTANT_PS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_GATHER_CONSTANT_PS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferValid, 16, 31) |
+ __gen_field(values->ConstantBufferBindingTableBlock, 12, 15) |
+ 0;
+
+ dw[2] =
+ __gen_offset(values->GatherBufferOffset, 6, 22) |
+ __gen_field(values->ConstantBufferDx9Enable, 4, 4) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN75_3DSTATE_GATHER_CONSTANT_VS_length_bias 0x00000002
+#define GEN75_3DSTATE_GATHER_CONSTANT_VS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 52
+
+#define GEN75_3DSTATE_GATHER_CONSTANT_VS_length 0x00000000
+
+struct GEN75_3DSTATE_GATHER_CONSTANT_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferValid;
+ uint32_t ConstantBufferBindingTableBlock;
+ uint32_t GatherBufferOffset;
+ bool ConstantBufferDx9Enable;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN75_3DSTATE_GATHER_CONSTANT_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_GATHER_CONSTANT_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferValid, 16, 31) |
+ __gen_field(values->ConstantBufferBindingTableBlock, 12, 15) |
+ 0;
+
+ dw[2] =
+ __gen_offset(values->GatherBufferOffset, 6, 22) |
+ __gen_field(values->ConstantBufferDx9Enable, 4, 4) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN75_3DSTATE_GATHER_POOL_ALLOC_length_bias 0x00000002
+#define GEN75_3DSTATE_GATHER_POOL_ALLOC_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 26, \
+ .DwordLength = 1
+
+#define GEN75_3DSTATE_GATHER_POOL_ALLOC_length 0x00000003
+
+struct GEN75_3DSTATE_GATHER_POOL_ALLOC {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ __gen_address_type GatherPoolBaseAddress;
+ bool GatherPoolEnable;
+ struct GEN75_MEMORY_OBJECT_CONTROL_STATE MemoryObjectControlState;
+ __gen_address_type GatherPoolUpperBound;
+};
+
+static inline void
+GEN75_3DSTATE_GATHER_POOL_ALLOC_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_GATHER_POOL_ALLOC * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw_MemoryObjectControlState;
+ GEN75_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_MemoryObjectControlState, &values->MemoryObjectControlState);
+ uint32_t dw1 =
+ __gen_field(values->GatherPoolEnable, 11, 11) |
+ __gen_mbo(4, 5) |
+ __gen_field(dw_MemoryObjectControlState, 0, 3) |
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->GatherPoolBaseAddress, dw1);
+
+ uint32_t dw2 =
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->GatherPoolUpperBound, dw2);
+
+}
+
+#define GEN75_3DSTATE_GS_length_bias 0x00000002
+#define GEN75_3DSTATE_GS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 17, \
+ .DwordLength = 5
+
+#define GEN75_3DSTATE_GS_length 0x00000007
+
+struct GEN75_3DSTATE_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t KernelStartPointer;
+ uint32_t SingleProgramFlowSPF;
+#define Dmask 0
+#define Vmask 1
+ uint32_t VectorMaskEnableVME;
+#define NoSamplers 0
+#define _14Samplers 1
+#define _58Samplers 2
+#define _912Samplers 3
+#define _1316Samplers 4
+ uint32_t SamplerCount;
+ uint32_t BindingTableEntryCount;
+#define NormalPriority 0
+#define HighPriority 1
+ uint32_t ThreadPriority;
+#define IEEE754 0
+#define alternate 1
+ uint32_t FloatingPointMode;
+ bool IllegalOpcodeExceptionEnable;
+ uint32_t GSaccessesUAV;
+ bool MaskStackExceptionEnable;
+ bool SoftwareExceptionEnable;
+ uint32_t ScratchSpaceBasePointer;
+ uint32_t PerThreadScratchSpace;
+ uint32_t OutputVertexSize;
+ uint32_t OutputTopology;
+ uint32_t VertexURBEntryReadLength;
+ bool IncludeVertexHandles;
+ uint32_t VertexURBEntryReadOffset;
+ uint32_t DispatchGRFStartRegisterforURBData;
+ uint32_t MaximumNumberofThreads;
+ uint32_t ControlDataHeaderSize;
+ uint32_t InstanceControl;
+ uint32_t DefaultStreamID;
+#define SINGLE 0
+#define DUAL_INSTANCE 1
+#define DUAL_OBJECT 2
+ uint32_t DispatchMode;
+ uint32_t GSStatisticsEnable;
+ uint32_t GSInvocationsIncrementValue;
+ bool IncludePrimitiveID;
+ uint32_t Hint;
+#define REORDER_LEADING 0
+#define REORDER_TRAILING 1
+ uint32_t ReorderMode;
+ bool DiscardAdjacency;
+ bool GSEnable;
+#define GSCTL_CUT 0
+#define GSCTL_SID 1
+ uint32_t ControlDataFormat;
+ uint32_t SemaphoreHandle;
+};
+
+static inline void
+GEN75_3DSTATE_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->KernelStartPointer, 6, 31) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->SingleProgramFlowSPF, 31, 31) |
+ __gen_field(values->VectorMaskEnableVME, 30, 30) |
+ __gen_field(values->SamplerCount, 27, 29) |
+ __gen_field(values->BindingTableEntryCount, 18, 25) |
+ __gen_field(values->ThreadPriority, 17, 17) |
+ __gen_field(values->FloatingPointMode, 16, 16) |
+ __gen_field(values->IllegalOpcodeExceptionEnable, 13, 13) |
+ __gen_field(values->GSaccessesUAV, 12, 12) |
+ __gen_field(values->MaskStackExceptionEnable, 11, 11) |
+ __gen_field(values->SoftwareExceptionEnable, 7, 7) |
+ 0;
+
+ dw[3] =
+ __gen_offset(values->ScratchSpaceBasePointer, 10, 31) |
+ __gen_field(values->PerThreadScratchSpace, 0, 3) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->OutputVertexSize, 23, 28) |
+ __gen_field(values->OutputTopology, 17, 22) |
+ __gen_field(values->VertexURBEntryReadLength, 11, 16) |
+ __gen_field(values->IncludeVertexHandles, 10, 10) |
+ __gen_field(values->VertexURBEntryReadOffset, 4, 9) |
+ __gen_field(values->DispatchGRFStartRegisterforURBData, 0, 3) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->MaximumNumberofThreads, 24, 31) |
+ __gen_field(values->ControlDataHeaderSize, 20, 23) |
+ __gen_field(values->InstanceControl, 15, 19) |
+ __gen_field(values->DefaultStreamID, 13, 14) |
+ __gen_field(values->DispatchMode, 11, 12) |
+ __gen_field(values->GSStatisticsEnable, 10, 10) |
+ __gen_field(values->GSInvocationsIncrementValue, 5, 9) |
+ __gen_field(values->IncludePrimitiveID, 4, 4) |
+ __gen_field(values->Hint, 3, 3) |
+ __gen_field(values->ReorderMode, 2, 2) |
+ __gen_field(values->DiscardAdjacency, 1, 1) |
+ __gen_field(values->GSEnable, 0, 0) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->ControlDataFormat, 31, 31) |
+ __gen_offset(values->SemaphoreHandle, 0, 12) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_HIER_DEPTH_BUFFER_length_bias 0x00000002
+#define GEN75_3DSTATE_HIER_DEPTH_BUFFER_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 7, \
+ .DwordLength = 1
+
+#define GEN75_3DSTATE_HIER_DEPTH_BUFFER_length 0x00000003
+
+struct GEN75_3DSTATE_HIER_DEPTH_BUFFER {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ struct GEN75_MEMORY_OBJECT_CONTROL_STATE HierarchicalDepthBufferObjectControlState;
+ uint32_t SurfacePitch;
+ __gen_address_type SurfaceBaseAddress;
+};
+
+static inline void
+GEN75_3DSTATE_HIER_DEPTH_BUFFER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_HIER_DEPTH_BUFFER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw_HierarchicalDepthBufferObjectControlState;
+ GEN75_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_HierarchicalDepthBufferObjectControlState, &values->HierarchicalDepthBufferObjectControlState);
+ dw[1] =
+ __gen_field(dw_HierarchicalDepthBufferObjectControlState, 25, 28) |
+ __gen_field(values->SurfacePitch, 0, 16) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->SurfaceBaseAddress, dw2);
+
+}
+
+#define GEN75_3DSTATE_HS_length_bias 0x00000002
+#define GEN75_3DSTATE_HS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 27, \
+ .DwordLength = 5
+
+#define GEN75_3DSTATE_HS_length 0x00000007
+
+struct GEN75_3DSTATE_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define NoSamplers 0
+#define _14Samplers 1
+#define _58Samplers 2
+#define _912Samplers 3
+#define _1316Samplers 4
+ uint32_t SamplerCount;
+ uint32_t BindingTableEntryCount;
+#define Normal 0
+#define High 1
+ uint32_t ThreadDispatchPriority;
+#define IEEE754 0
+#define alternate 1
+ uint32_t FloatingPointMode;
+ bool IllegalOpcodeExceptionEnable;
+ bool SoftwareExceptionEnable;
+ uint32_t MaximumNumberofThreads;
+ bool Enable;
+ bool StatisticsEnable;
+ uint32_t InstanceCount;
+ uint32_t KernelStartPointer;
+ uint32_t ScratchSpaceBasePointer;
+ uint32_t PerThreadScratchSpace;
+ uint32_t SingleProgramFlow;
+#define Dmask 0
+#define Vmask 1
+ uint32_t VectorMaskEnable;
+ bool HSaccessesUAV;
+ bool IncludeVertexHandles;
+ uint32_t DispatchGRFStartRegisterForURBData;
+ uint32_t VertexURBEntryReadLength;
+ uint32_t VertexURBEntryReadOffset;
+ uint32_t SemaphoreHandle;
+};
+
+static inline void
+GEN75_3DSTATE_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->SamplerCount, 27, 29) |
+ __gen_field(values->BindingTableEntryCount, 18, 25) |
+ __gen_field(values->ThreadDispatchPriority, 17, 17) |
+ __gen_field(values->FloatingPointMode, 16, 16) |
+ __gen_field(values->IllegalOpcodeExceptionEnable, 13, 13) |
+ __gen_field(values->SoftwareExceptionEnable, 12, 12) |
+ __gen_field(values->MaximumNumberofThreads, 0, 7) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->Enable, 31, 31) |
+ __gen_field(values->StatisticsEnable, 29, 29) |
+ __gen_field(values->InstanceCount, 0, 3) |
+ 0;
+
+ dw[3] =
+ __gen_offset(values->KernelStartPointer, 6, 31) |
+ 0;
+
+ dw[4] =
+ __gen_offset(values->ScratchSpaceBasePointer, 10, 31) |
+ __gen_field(values->PerThreadScratchSpace, 0, 3) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->SingleProgramFlow, 27, 27) |
+ __gen_field(values->VectorMaskEnable, 26, 26) |
+ __gen_field(values->HSaccessesUAV, 25, 25) |
+ __gen_field(values->IncludeVertexHandles, 24, 24) |
+ __gen_field(values->DispatchGRFStartRegisterForURBData, 19, 23) |
+ __gen_field(values->VertexURBEntryReadLength, 11, 16) |
+ __gen_field(values->VertexURBEntryReadOffset, 4, 9) |
+ 0;
+
+ dw[6] =
+ __gen_offset(values->SemaphoreHandle, 0, 12) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_INDEX_BUFFER_length_bias 0x00000002
+#define GEN75_3DSTATE_INDEX_BUFFER_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 10, \
+ .DwordLength = 1
+
+#define GEN75_3DSTATE_INDEX_BUFFER_length 0x00000003
+
+struct GEN75_3DSTATE_INDEX_BUFFER {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ struct GEN75_MEMORY_OBJECT_CONTROL_STATE MemoryObjectControlState;
+#define INDEX_BYTE 0
+#define INDEX_WORD 1
+#define INDEX_DWORD 2
+ uint32_t IndexFormat;
+ uint32_t DwordLength;
+ __gen_address_type BufferStartingAddress;
+ __gen_address_type BufferEndingAddress;
+};
+
+static inline void
+GEN75_3DSTATE_INDEX_BUFFER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_INDEX_BUFFER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ uint32_t dw_MemoryObjectControlState;
+ GEN75_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_MemoryObjectControlState, &values->MemoryObjectControlState);
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(dw_MemoryObjectControlState, 12, 15) |
+ __gen_field(values->IndexFormat, 8, 9) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw1 =
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->BufferStartingAddress, dw1);
+
+ uint32_t dw2 =
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->BufferEndingAddress, dw2);
+
+}
+
+#define GEN75_3DSTATE_LINE_STIPPLE_length_bias 0x00000002
+#define GEN75_3DSTATE_LINE_STIPPLE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 8, \
+ .DwordLength = 1
+
+#define GEN75_3DSTATE_LINE_STIPPLE_length 0x00000003
+
+struct GEN75_3DSTATE_LINE_STIPPLE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool ModifyEnableCurrentRepeatCounterCurrentStippleIndex;
+ uint32_t CurrentRepeatCounter;
+ uint32_t CurrentStippleIndex;
+ uint32_t LineStipplePattern;
+ float LineStippleInverseRepeatCount;
+ uint32_t LineStippleRepeatCount;
+};
+
+static inline void
+GEN75_3DSTATE_LINE_STIPPLE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_LINE_STIPPLE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ModifyEnableCurrentRepeatCounterCurrentStippleIndex, 31, 31) |
+ __gen_field(values->CurrentRepeatCounter, 21, 29) |
+ __gen_field(values->CurrentStippleIndex, 16, 19) |
+ __gen_field(values->LineStipplePattern, 0, 15) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->LineStippleInverseRepeatCount * (1 << 16), 15, 31) |
+ __gen_field(values->LineStippleRepeatCount, 0, 8) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_MONOFILTER_SIZE_length_bias 0x00000002
+#define GEN75_3DSTATE_MONOFILTER_SIZE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 17, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_MONOFILTER_SIZE_length 0x00000002
+
+struct GEN75_3DSTATE_MONOFILTER_SIZE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t MonochromeFilterWidth;
+ uint32_t MonochromeFilterHeight;
+};
+
+static inline void
+GEN75_3DSTATE_MONOFILTER_SIZE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_MONOFILTER_SIZE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->MonochromeFilterWidth, 3, 5) |
+ __gen_field(values->MonochromeFilterHeight, 0, 2) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_MULTISAMPLE_length_bias 0x00000002
+#define GEN75_3DSTATE_MULTISAMPLE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 13, \
+ .DwordLength = 2
+
+#define GEN75_3DSTATE_MULTISAMPLE_length 0x00000004
+
+struct GEN75_3DSTATE_MULTISAMPLE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool MultiSampleEnable;
+#define PIXLOC_CENTER 0
+#define PIXLOC_UL_CORNER 1
+ uint32_t PixelLocation;
+#define NUMSAMPLES_1 0
+#define NUMSAMPLES_4 2
+#define NUMSAMPLES_8 3
+ uint32_t NumberofMultisamples;
+ float Sample3XOffset;
+ float Sample3YOffset;
+ float Sample2XOffset;
+ float Sample2YOffset;
+ float Sample1XOffset;
+ float Sample1YOffset;
+ float Sample0XOffset;
+ float Sample0YOffset;
+ float Sample7XOffset;
+ float Sample7YOffset;
+ float Sample6XOffset;
+ float Sample6YOffset;
+ float Sample5XOffset;
+ float Sample5YOffset;
+ float Sample4XOffset;
+ float Sample4YOffset;
+};
+
+static inline void
+GEN75_3DSTATE_MULTISAMPLE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_MULTISAMPLE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->MultiSampleEnable, 5, 5) |
+ __gen_field(values->PixelLocation, 4, 4) |
+ __gen_field(values->NumberofMultisamples, 1, 3) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->Sample3XOffset * (1 << 4), 28, 31) |
+ __gen_field(values->Sample3YOffset * (1 << 4), 24, 27) |
+ __gen_field(values->Sample2XOffset * (1 << 4), 20, 23) |
+ __gen_field(values->Sample2YOffset * (1 << 4), 16, 19) |
+ __gen_field(values->Sample1XOffset * (1 << 4), 12, 15) |
+ __gen_field(values->Sample1YOffset * (1 << 4), 8, 11) |
+ __gen_field(values->Sample0XOffset * (1 << 4), 4, 7) |
+ __gen_field(values->Sample0YOffset * (1 << 4), 0, 3) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->Sample7XOffset * (1 << 4), 28, 31) |
+ __gen_field(values->Sample7YOffset * (1 << 4), 24, 27) |
+ __gen_field(values->Sample6XOffset * (1 << 4), 20, 23) |
+ __gen_field(values->Sample6YOffset * (1 << 4), 16, 19) |
+ __gen_field(values->Sample5XOffset * (1 << 4), 12, 15) |
+ __gen_field(values->Sample5YOffset * (1 << 4), 8, 11) |
+ __gen_field(values->Sample4XOffset * (1 << 4), 4, 7) |
+ __gen_field(values->Sample4YOffset * (1 << 4), 0, 3) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_POLY_STIPPLE_OFFSET_length_bias 0x00000002
+#define GEN75_3DSTATE_POLY_STIPPLE_OFFSET_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 6, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_POLY_STIPPLE_OFFSET_length 0x00000002
+
+struct GEN75_3DSTATE_POLY_STIPPLE_OFFSET {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PolygonStippleXOffset;
+ uint32_t PolygonStippleYOffset;
+};
+
+static inline void
+GEN75_3DSTATE_POLY_STIPPLE_OFFSET_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_POLY_STIPPLE_OFFSET * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->PolygonStippleXOffset, 8, 12) |
+ __gen_field(values->PolygonStippleYOffset, 0, 4) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_POLY_STIPPLE_PATTERN_length_bias 0x00000002
+#define GEN75_3DSTATE_POLY_STIPPLE_PATTERN_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 7, \
+ .DwordLength = 31
+
+#define GEN75_3DSTATE_POLY_STIPPLE_PATTERN_length 0x00000021
+
+struct GEN75_3DSTATE_POLY_STIPPLE_PATTERN {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PatternRow[32];
+};
+
+static inline void
+GEN75_3DSTATE_POLY_STIPPLE_PATTERN_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_POLY_STIPPLE_PATTERN * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ for (uint32_t i = 0, j = 1; i < 32; i += 1, j++) {
+ dw[j] =
+ __gen_field(values->PatternRow[i + 0], 0, 31) |
+ 0;
+ }
+
+}
+
+#define GEN75_3DSTATE_PS_length_bias 0x00000002
+#define GEN75_3DSTATE_PS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 32, \
+ .DwordLength = 6
+
+#define GEN75_3DSTATE_PS_length 0x00000008
+
+struct GEN75_3DSTATE_PS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t KernelStartPointer0;
+#define Multiple 0
+#define Single 1
+ uint32_t SingleProgramFlowSPF;
+#define Dmask 0
+#define Vmask 1
+ uint32_t VectorMaskEnableVME;
+ uint32_t SamplerCount;
+#define FTZ 0
+#define RET 1
+ uint32_t DenormalMode;
+ uint32_t BindingTableEntryCount;
+#define Normal 0
+#define High 1
+ uint32_t ThreadPriority;
+#define IEEE745 0
+#define Alt 1
+ uint32_t FloatingPointMode;
+#define RTNE 0
+#define RU 1
+#define RD 2
+#define RTZ 3
+ uint32_t RoundingMode;
+ bool IllegalOpcodeExceptionEnable;
+ bool MaskStackExceptionEnable;
+ bool SoftwareExceptionEnable;
+ uint32_t ScratchSpaceBasePointer;
+ uint32_t PerThreadScratchSpace;
+ uint32_t MaximumNumberofThreads;
+ uint32_t SampleMask;
+ bool PushConstantEnable;
+ bool AttributeEnable;
+ bool oMaskPresenttoRenderTarget;
+ bool RenderTargetFastClearEnable;
+ bool DualSourceBlendEnable;
+ bool RenderTargetResolveEnable;
+ bool PSAccessesUAV;
+#define POSOFFSET_NONE 0
+#define POSOFFSET_CENTROID 2
+#define POSOFFSET_SAMPLE 3
+ uint32_t PositionXYOffsetSelect;
+ bool _32PixelDispatchEnable;
+ bool _16PixelDispatchEnable;
+ bool _8PixelDispatchEnable;
+ uint32_t DispatchGRFStartRegisterforConstantSetupData0;
+ uint32_t DispatchGRFStartRegisterforConstantSetupData1;
+ uint32_t DispatchGRFStartRegisterforConstantSetupData2;
+ uint32_t KernelStartPointer1;
+ uint32_t KernelStartPointer2;
+};
+
+static inline void
+GEN75_3DSTATE_PS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_PS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->KernelStartPointer0, 6, 31) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->SingleProgramFlowSPF, 31, 31) |
+ __gen_field(values->VectorMaskEnableVME, 30, 30) |
+ __gen_field(values->SamplerCount, 27, 29) |
+ __gen_field(values->DenormalMode, 26, 26) |
+ __gen_field(values->BindingTableEntryCount, 18, 25) |
+ __gen_field(values->ThreadPriority, 17, 17) |
+ __gen_field(values->FloatingPointMode, 16, 16) |
+ __gen_field(values->RoundingMode, 14, 15) |
+ __gen_field(values->IllegalOpcodeExceptionEnable, 13, 13) |
+ __gen_field(values->MaskStackExceptionEnable, 11, 11) |
+ __gen_field(values->SoftwareExceptionEnable, 7, 7) |
+ 0;
+
+ dw[3] =
+ __gen_offset(values->ScratchSpaceBasePointer, 10, 31) |
+ __gen_field(values->PerThreadScratchSpace, 0, 3) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->MaximumNumberofThreads, 23, 31) |
+ __gen_field(values->SampleMask, 12, 19) |
+ __gen_field(values->PushConstantEnable, 11, 11) |
+ __gen_field(values->AttributeEnable, 10, 10) |
+ __gen_field(values->oMaskPresenttoRenderTarget, 9, 9) |
+ __gen_field(values->RenderTargetFastClearEnable, 8, 8) |
+ __gen_field(values->DualSourceBlendEnable, 7, 7) |
+ __gen_field(values->RenderTargetResolveEnable, 6, 6) |
+ __gen_field(values->PSAccessesUAV, 5, 5) |
+ __gen_field(values->PositionXYOffsetSelect, 3, 4) |
+ __gen_field(values->_32PixelDispatchEnable, 2, 2) |
+ __gen_field(values->_16PixelDispatchEnable, 1, 1) |
+ __gen_field(values->_8PixelDispatchEnable, 0, 0) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->DispatchGRFStartRegisterforConstantSetupData0, 16, 22) |
+ __gen_field(values->DispatchGRFStartRegisterforConstantSetupData1, 8, 14) |
+ __gen_field(values->DispatchGRFStartRegisterforConstantSetupData2, 0, 6) |
+ 0;
+
+ dw[6] =
+ __gen_offset(values->KernelStartPointer1, 6, 31) |
+ 0;
+
+ dw[7] =
+ __gen_offset(values->KernelStartPointer2, 6, 31) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_DS_length_bias 0x00000002
+#define GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_DS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 20, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_DS_length 0x00000002
+
+struct GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferOffset;
+ uint32_t ConstantBufferSize;
+};
+
+static inline void
+GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferOffset, 16, 20) |
+ __gen_field(values->ConstantBufferSize, 0, 5) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_GS_length_bias 0x00000002
+#define GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_GS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 21, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_GS_length 0x00000002
+
+struct GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferOffset;
+ uint32_t ConstantBufferSize;
+};
+
+static inline void
+GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferOffset, 16, 20) |
+ __gen_field(values->ConstantBufferSize, 0, 5) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_HS_length_bias 0x00000002
+#define GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_HS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 19, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_HS_length 0x00000002
+
+struct GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferOffset;
+ uint32_t ConstantBufferSize;
+};
+
+static inline void
+GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferOffset, 16, 20) |
+ __gen_field(values->ConstantBufferSize, 0, 5) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_PS_length_bias 0x00000002
+#define GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_PS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 22, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_PS_length 0x00000002
+
+struct GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_PS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferOffset;
+ uint32_t ConstantBufferSize;
+};
+
+static inline void
+GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_PS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_PS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferOffset, 16, 20) |
+ __gen_field(values->ConstantBufferSize, 0, 5) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_VS_length_bias 0x00000002
+#define GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_VS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 18, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_VS_length 0x00000002
+
+struct GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferOffset;
+ uint32_t ConstantBufferSize;
+};
+
+static inline void
+GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_PUSH_CONSTANT_ALLOC_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferOffset, 16, 20) |
+ __gen_field(values->ConstantBufferSize, 0, 5) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_RAST_MULTISAMPLE_length_bias 0x00000002
+#define GEN75_3DSTATE_RAST_MULTISAMPLE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 14, \
+ .DwordLength = 4
+
+#define GEN75_3DSTATE_RAST_MULTISAMPLE_length 0x00000006
+
+struct GEN75_3DSTATE_RAST_MULTISAMPLE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define NRM_NUMRASTSAMPLES_1 0
+#define NRM_NUMRASTSAMPLES_2 1
+#define NRM_NUMRASTSAMPLES_4 2
+#define NRM_NUMRASTSAMPLES_8 3
+#define NRM_NUMRASTSAMPLES_16 4
+ uint32_t NumberofRasterizationMultisamples;
+ float Sample3XOffset;
+ float Sample3YOffset;
+ float Sample2XOffset;
+ float Sample2YOffset;
+ float Sample1XOffset;
+ float Sample1YOffset;
+ float Sample0XOffset;
+ float Sample0YOffset;
+ float Sample7XOffset;
+ float Sample7YOffset;
+ float Sample6XOffset;
+ float Sample6YOffset;
+ float Sample5XOffset;
+ float Sample5YOffset;
+ float Sample4XOffset;
+ float Sample4YOffset;
+ float Sample11XOffset;
+ float Sample11YOffset;
+ float Sample10XOffset;
+ float Sample10YOffset;
+ float Sample9XOffset;
+ float Sample9YOffset;
+ float Sample8XOffset;
+ float Sample8YOffset;
+ float Sample15XOffset;
+ float Sample15YOffset;
+ float Sample14XOffset;
+ float Sample14YOffset;
+ float Sample13XOffset;
+ float Sample13YOffset;
+ float Sample12XOffset;
+ float Sample12YOffset;
+};
+
+static inline void
+GEN75_3DSTATE_RAST_MULTISAMPLE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_RAST_MULTISAMPLE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->NumberofRasterizationMultisamples, 1, 3) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->Sample3XOffset * (1 << 4), 28, 31) |
+ __gen_field(values->Sample3YOffset * (1 << 4), 24, 27) |
+ __gen_field(values->Sample2XOffset * (1 << 4), 20, 23) |
+ __gen_field(values->Sample2YOffset * (1 << 4), 16, 19) |
+ __gen_field(values->Sample1XOffset * (1 << 4), 12, 15) |
+ __gen_field(values->Sample1YOffset * (1 << 4), 8, 11) |
+ __gen_field(values->Sample0XOffset * (1 << 4), 4, 7) |
+ __gen_field(values->Sample0YOffset * (1 << 4), 0, 3) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->Sample7XOffset * (1 << 4), 28, 31) |
+ __gen_field(values->Sample7YOffset * (1 << 4), 24, 27) |
+ __gen_field(values->Sample6XOffset * (1 << 4), 20, 23) |
+ __gen_field(values->Sample6YOffset * (1 << 4), 16, 19) |
+ __gen_field(values->Sample5XOffset * (1 << 4), 12, 15) |
+ __gen_field(values->Sample5YOffset * (1 << 4), 8, 11) |
+ __gen_field(values->Sample4XOffset * (1 << 4), 4, 7) |
+ __gen_field(values->Sample4YOffset * (1 << 4), 0, 3) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->Sample11XOffset * (1 << 4), 28, 31) |
+ __gen_field(values->Sample11YOffset * (1 << 4), 24, 27) |
+ __gen_field(values->Sample10XOffset * (1 << 4), 20, 23) |
+ __gen_field(values->Sample10YOffset * (1 << 4), 16, 19) |
+ __gen_field(values->Sample9XOffset * (1 << 4), 12, 15) |
+ __gen_field(values->Sample9YOffset * (1 << 4), 8, 11) |
+ __gen_field(values->Sample8XOffset * (1 << 4), 4, 7) |
+ __gen_field(values->Sample8YOffset * (1 << 4), 0, 3) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->Sample15XOffset * (1 << 4), 28, 31) |
+ __gen_field(values->Sample15YOffset * (1 << 4), 24, 27) |
+ __gen_field(values->Sample14XOffset * (1 << 4), 20, 23) |
+ __gen_field(values->Sample14YOffset * (1 << 4), 16, 19) |
+ __gen_field(values->Sample13XOffset * (1 << 4), 12, 15) |
+ __gen_field(values->Sample13YOffset * (1 << 4), 8, 11) |
+ __gen_field(values->Sample12XOffset * (1 << 4), 4, 7) |
+ __gen_field(values->Sample12YOffset * (1 << 4), 0, 3) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_SAMPLER_PALETTE_LOAD0_length_bias 0x00000002
+#define GEN75_3DSTATE_SAMPLER_PALETTE_LOAD0_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 2
+
+#define GEN75_3DSTATE_SAMPLER_PALETTE_LOAD0_length 0x00000000
+
+#define GEN75_PALETTE_ENTRY_length 0x00000001
+
+struct GEN75_PALETTE_ENTRY {
+ uint32_t Alpha;
+ uint32_t Red;
+ uint32_t Green;
+ uint32_t Blue;
+};
+
+static inline void
+GEN75_PALETTE_ENTRY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_PALETTE_ENTRY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->Alpha, 24, 31) |
+ __gen_field(values->Red, 16, 23) |
+ __gen_field(values->Green, 8, 15) |
+ __gen_field(values->Blue, 0, 7) |
+ 0;
+
+}
+
+struct GEN75_3DSTATE_SAMPLER_PALETTE_LOAD0 {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN75_3DSTATE_SAMPLER_PALETTE_LOAD0_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_SAMPLER_PALETTE_LOAD0 * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN75_3DSTATE_SAMPLER_PALETTE_LOAD1_length_bias 0x00000002
+#define GEN75_3DSTATE_SAMPLER_PALETTE_LOAD1_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 12
+
+#define GEN75_3DSTATE_SAMPLER_PALETTE_LOAD1_length 0x00000000
+
+struct GEN75_3DSTATE_SAMPLER_PALETTE_LOAD1 {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN75_3DSTATE_SAMPLER_PALETTE_LOAD1_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_SAMPLER_PALETTE_LOAD1 * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN75_3DSTATE_SAMPLER_STATE_POINTERS_DS_length_bias 0x00000002
+#define GEN75_3DSTATE_SAMPLER_STATE_POINTERS_DS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 45, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_SAMPLER_STATE_POINTERS_DS_length 0x00000002
+
+struct GEN75_3DSTATE_SAMPLER_STATE_POINTERS_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoDSSamplerState;
+};
+
+static inline void
+GEN75_3DSTATE_SAMPLER_STATE_POINTERS_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_SAMPLER_STATE_POINTERS_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoDSSamplerState, 5, 31) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_SAMPLER_STATE_POINTERS_GS_length_bias 0x00000002
+#define GEN75_3DSTATE_SAMPLER_STATE_POINTERS_GS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 46, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_SAMPLER_STATE_POINTERS_GS_length 0x00000002
+
+struct GEN75_3DSTATE_SAMPLER_STATE_POINTERS_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoGSSamplerState;
+};
+
+static inline void
+GEN75_3DSTATE_SAMPLER_STATE_POINTERS_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_SAMPLER_STATE_POINTERS_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoGSSamplerState, 5, 31) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_SAMPLER_STATE_POINTERS_HS_length_bias 0x00000002
+#define GEN75_3DSTATE_SAMPLER_STATE_POINTERS_HS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 44, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_SAMPLER_STATE_POINTERS_HS_length 0x00000002
+
+struct GEN75_3DSTATE_SAMPLER_STATE_POINTERS_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoHSSamplerState;
+};
+
+static inline void
+GEN75_3DSTATE_SAMPLER_STATE_POINTERS_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_SAMPLER_STATE_POINTERS_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoHSSamplerState, 5, 31) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_SAMPLER_STATE_POINTERS_PS_length_bias 0x00000002
+#define GEN75_3DSTATE_SAMPLER_STATE_POINTERS_PS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 47, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_SAMPLER_STATE_POINTERS_PS_length 0x00000002
+
+struct GEN75_3DSTATE_SAMPLER_STATE_POINTERS_PS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoPSSamplerState;
+};
+
+static inline void
+GEN75_3DSTATE_SAMPLER_STATE_POINTERS_PS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_SAMPLER_STATE_POINTERS_PS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoPSSamplerState, 5, 31) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_SAMPLER_STATE_POINTERS_VS_length_bias 0x00000002
+#define GEN75_3DSTATE_SAMPLER_STATE_POINTERS_VS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 43, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_SAMPLER_STATE_POINTERS_VS_length 0x00000002
+
+struct GEN75_3DSTATE_SAMPLER_STATE_POINTERS_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoVSSamplerState;
+};
+
+static inline void
+GEN75_3DSTATE_SAMPLER_STATE_POINTERS_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_SAMPLER_STATE_POINTERS_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoVSSamplerState, 5, 31) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_SAMPLE_MASK_length_bias 0x00000002
+#define GEN75_3DSTATE_SAMPLE_MASK_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 24, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_SAMPLE_MASK_length 0x00000002
+
+struct GEN75_3DSTATE_SAMPLE_MASK {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t SampleMask;
+};
+
+static inline void
+GEN75_3DSTATE_SAMPLE_MASK_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_SAMPLE_MASK * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->SampleMask, 0, 7) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_SBE_length_bias 0x00000002
+#define GEN75_3DSTATE_SBE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 31, \
+ .DwordLength = 12
+
+#define GEN75_3DSTATE_SBE_length 0x0000000e
+
+struct GEN75_3DSTATE_SBE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t AttributeSwizzleControlMode;
+ uint32_t NumberofSFOutputAttributes;
+ bool AttributeSwizzleEnable;
+#define UPPERLEFT 0
+#define LOWERLEFT 1
+ uint32_t PointSpriteTextureCoordinateOrigin;
+ uint32_t VertexURBEntryReadLength;
+ uint32_t VertexURBEntryReadOffset;
+ bool Attribute2n1ComponentOverrideW;
+ bool Attribute2n1ComponentOverrideZ;
+ bool Attribute2n1ComponentOverrideY;
+ bool Attribute2n1ComponentOverrideX;
+#define CONST_0000 0
+#define CONST_0001_FLOAT 1
+#define CONST_1111_FLOAT 2
+#define PRIM_ID 3
+ uint32_t Attribute2n1ConstantSource;
+#define INPUTATTR 0
+#define INPUTATTR_FACING 1
+#define INPUTATTR_W 2
+#define INPUTATTR_FACING_W 3
+ uint32_t Attribute2n1SwizzleSelect;
+ uint32_t Attribute2n1SourceAttribute;
+ bool Attribute2nComponentOverrideW;
+ bool Attribute2nComponentOverrideZ;
+ bool Attribute2nComponentOverrideY;
+ bool Attribute2nComponentOverrideX;
+#define CONST_0000 0
+#define CONST_0001_FLOAT 1
+#define CONST_1111_FLOAT 2
+#define PRIM_ID 3
+ uint32_t Attribute2nConstantSource;
+#define INPUTATTR 0
+#define INPUTATTR_FACING 1
+#define INPUTATTR_W 2
+#define INPUTATTR_FACING_W 3
+ uint32_t Attribute2nSwizzleSelect;
+ uint32_t Attribute2nSourceAttribute;
+ uint32_t PointSpriteTextureCoordinateEnable;
+ uint32_t ConstantInterpolationEnable310;
+ uint32_t Attribute7WrapShortestEnables;
+ uint32_t Attribute6WrapShortestEnables;
+ uint32_t Attribute5WrapShortestEnables;
+ uint32_t Attribute4WrapShortestEnables;
+ uint32_t Attribute3WrapShortestEnables;
+ uint32_t Attribute2WrapShortestEnables;
+ uint32_t Attribute1WrapShortestEnables;
+ uint32_t Attribute0WrapShortestEnables;
+ uint32_t Attribute15WrapShortestEnables;
+ uint32_t Attribute14WrapShortestEnables;
+ uint32_t Attribute13WrapShortestEnables;
+ uint32_t Attribute12WrapShortestEnables;
+ uint32_t Attribute11WrapShortestEnables;
+ uint32_t Attribute10WrapShortestEnables;
+ uint32_t Attribute9WrapShortestEnables;
+ uint32_t Attribute8WrapShortestEnables;
+};
+
+static inline void
+GEN75_3DSTATE_SBE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_SBE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->AttributeSwizzleControlMode, 28, 28) |
+ __gen_field(values->NumberofSFOutputAttributes, 22, 27) |
+ __gen_field(values->AttributeSwizzleEnable, 21, 21) |
+ __gen_field(values->PointSpriteTextureCoordinateOrigin, 20, 20) |
+ __gen_field(values->VertexURBEntryReadLength, 11, 15) |
+ __gen_field(values->VertexURBEntryReadOffset, 4, 9) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->Attribute2n1ComponentOverrideW, 31, 31) |
+ __gen_field(values->Attribute2n1ComponentOverrideZ, 30, 30) |
+ __gen_field(values->Attribute2n1ComponentOverrideY, 29, 29) |
+ __gen_field(values->Attribute2n1ComponentOverrideX, 28, 28) |
+ __gen_field(values->Attribute2n1ConstantSource, 25, 26) |
+ __gen_field(values->Attribute2n1SwizzleSelect, 22, 23) |
+ __gen_field(values->Attribute2n1SourceAttribute, 16, 20) |
+ __gen_field(values->Attribute2nComponentOverrideW, 15, 15) |
+ __gen_field(values->Attribute2nComponentOverrideZ, 14, 14) |
+ __gen_field(values->Attribute2nComponentOverrideY, 13, 13) |
+ __gen_field(values->Attribute2nComponentOverrideX, 12, 12) |
+ __gen_field(values->Attribute2nConstantSource, 9, 10) |
+ __gen_field(values->Attribute2nSwizzleSelect, 6, 7) |
+ __gen_field(values->Attribute2nSourceAttribute, 0, 4) |
+ 0;
+
+ dw[10] =
+ __gen_field(values->PointSpriteTextureCoordinateEnable, 0, 31) |
+ 0;
+
+ dw[11] =
+ __gen_field(values->ConstantInterpolationEnable310, 0, 31) |
+ 0;
+
+ dw[12] =
+ __gen_field(values->Attribute7WrapShortestEnables, 28, 31) |
+ __gen_field(values->Attribute6WrapShortestEnables, 24, 27) |
+ __gen_field(values->Attribute5WrapShortestEnables, 20, 23) |
+ __gen_field(values->Attribute4WrapShortestEnables, 16, 19) |
+ __gen_field(values->Attribute3WrapShortestEnables, 12, 15) |
+ __gen_field(values->Attribute2WrapShortestEnables, 8, 11) |
+ __gen_field(values->Attribute1WrapShortestEnables, 4, 7) |
+ __gen_field(values->Attribute0WrapShortestEnables, 0, 3) |
+ 0;
+
+ dw[13] =
+ __gen_field(values->Attribute15WrapShortestEnables, 28, 31) |
+ __gen_field(values->Attribute14WrapShortestEnables, 24, 27) |
+ __gen_field(values->Attribute13WrapShortestEnables, 20, 23) |
+ __gen_field(values->Attribute12WrapShortestEnables, 16, 19) |
+ __gen_field(values->Attribute11WrapShortestEnables, 12, 15) |
+ __gen_field(values->Attribute10WrapShortestEnables, 8, 11) |
+ __gen_field(values->Attribute9WrapShortestEnables, 4, 7) |
+ __gen_field(values->Attribute8WrapShortestEnables, 0, 3) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_SCISSOR_STATE_POINTERS_length_bias 0x00000002
+#define GEN75_3DSTATE_SCISSOR_STATE_POINTERS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 15, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_SCISSOR_STATE_POINTERS_length 0x00000002
+
+struct GEN75_3DSTATE_SCISSOR_STATE_POINTERS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ScissorRectPointer;
+};
+
+static inline void
+GEN75_3DSTATE_SCISSOR_STATE_POINTERS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_SCISSOR_STATE_POINTERS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->ScissorRectPointer, 5, 31) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_SF_length_bias 0x00000002
+#define GEN75_3DSTATE_SF_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 19, \
+ .DwordLength = 5
+
+#define GEN75_3DSTATE_SF_length 0x00000007
+
+struct GEN75_3DSTATE_SF {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define D32_FLOAT_S8X24_UINT 0
+#define D32_FLOAT 1
+#define D24_UNORM_S8_UINT 2
+#define D24_UNORM_X8_UINT 3
+#define D16_UNORM 5
+ uint32_t DepthBufferSurfaceFormat;
+ bool LegacyGlobalDepthBiasEnable;
+ bool StatisticsEnable;
+ bool GlobalDepthOffsetEnableSolid;
+ bool GlobalDepthOffsetEnableWireframe;
+ bool GlobalDepthOffsetEnablePoint;
+#define RASTER_SOLID 0
+#define RASTER_WIREFRAME 1
+#define RASTER_POINT 2
+ uint32_t FrontFaceFillMode;
+#define RASTER_SOLID 0
+#define RASTER_WIREFRAME 1
+#define RASTER_POINT 2
+ uint32_t BackFaceFillMode;
+ bool ViewTransformEnable;
+ uint32_t FrontWinding;
+ bool AntiAliasingEnable;
+#define CULLMODE_BOTH 0
+#define CULLMODE_NONE 1
+#define CULLMODE_FRONT 2
+#define CULLMODE_BACK 3
+ uint32_t CullMode;
+ float LineWidth;
+ uint32_t LineEndCapAntialiasingRegionWidth;
+ bool LineStippleEnable;
+ bool ScissorRectangleEnable;
+ bool RTIndependentRasterizationEnable;
+ uint32_t MultisampleRasterizationMode;
+ bool LastPixelEnable;
+#define Vertex0 0
+#define Vertex1 1
+#define Vertex2 2
+ uint32_t TriangleStripListProvokingVertexSelect;
+ uint32_t LineStripListProvokingVertexSelect;
+#define Vertex0 0
+#define Vertex1 1
+#define Vertex2 2
+ uint32_t TriangleFanProvokingVertexSelect;
+#define AALINEDISTANCE_TRUE 1
+ uint32_t AALineDistanceMode;
+ uint32_t VertexSubPixelPrecisionSelect;
+ uint32_t UsePointWidthState;
+ float PointWidth;
+ float GlobalDepthOffsetConstant;
+ float GlobalDepthOffsetScale;
+ float GlobalDepthOffsetClamp;
+};
+
+static inline void
+GEN75_3DSTATE_SF_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_SF * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->DepthBufferSurfaceFormat, 12, 14) |
+ __gen_field(values->LegacyGlobalDepthBiasEnable, 11, 11) |
+ __gen_field(values->StatisticsEnable, 10, 10) |
+ __gen_field(values->GlobalDepthOffsetEnableSolid, 9, 9) |
+ __gen_field(values->GlobalDepthOffsetEnableWireframe, 8, 8) |
+ __gen_field(values->GlobalDepthOffsetEnablePoint, 7, 7) |
+ __gen_field(values->FrontFaceFillMode, 5, 6) |
+ __gen_field(values->BackFaceFillMode, 3, 4) |
+ __gen_field(values->ViewTransformEnable, 1, 1) |
+ __gen_field(values->FrontWinding, 0, 0) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->AntiAliasingEnable, 31, 31) |
+ __gen_field(values->CullMode, 29, 30) |
+ __gen_field(values->LineWidth * (1 << 7), 18, 27) |
+ __gen_field(values->LineEndCapAntialiasingRegionWidth, 16, 17) |
+ __gen_field(values->LineStippleEnable, 14, 14) |
+ __gen_field(values->ScissorRectangleEnable, 11, 11) |
+ __gen_field(values->RTIndependentRasterizationEnable, 10, 10) |
+ __gen_field(values->MultisampleRasterizationMode, 8, 9) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->LastPixelEnable, 31, 31) |
+ __gen_field(values->TriangleStripListProvokingVertexSelect, 29, 30) |
+ __gen_field(values->LineStripListProvokingVertexSelect, 27, 28) |
+ __gen_field(values->TriangleFanProvokingVertexSelect, 25, 26) |
+ __gen_field(values->AALineDistanceMode, 14, 14) |
+ __gen_field(values->VertexSubPixelPrecisionSelect, 12, 12) |
+ __gen_field(values->UsePointWidthState, 11, 11) |
+ __gen_field(values->PointWidth * (1 << 3), 0, 10) |
+ 0;
+
+ dw[4] =
+ __gen_float(values->GlobalDepthOffsetConstant) |
+ 0;
+
+ dw[5] =
+ __gen_float(values->GlobalDepthOffsetScale) |
+ 0;
+
+ dw[6] =
+ __gen_float(values->GlobalDepthOffsetClamp) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_SO_BUFFER_length_bias 0x00000002
+#define GEN75_3DSTATE_SO_BUFFER_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 24, \
+ .DwordLength = 2
+
+#define GEN75_3DSTATE_SO_BUFFER_length 0x00000004
+
+struct GEN75_3DSTATE_SO_BUFFER {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t SOBufferIndex;
+ struct GEN75_MEMORY_OBJECT_CONTROL_STATE SOBufferObjectControlState;
+ uint32_t SurfacePitch;
+ __gen_address_type SurfaceBaseAddress;
+ __gen_address_type SurfaceEndAddress;
+};
+
+static inline void
+GEN75_3DSTATE_SO_BUFFER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_SO_BUFFER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw_SOBufferObjectControlState;
+ GEN75_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_SOBufferObjectControlState, &values->SOBufferObjectControlState);
+ dw[1] =
+ __gen_field(values->SOBufferIndex, 29, 30) |
+ __gen_field(dw_SOBufferObjectControlState, 25, 28) |
+ __gen_field(values->SurfacePitch, 0, 11) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->SurfaceBaseAddress, dw2);
+
+ uint32_t dw3 =
+ 0;
+
+ dw[3] =
+ __gen_combine_address(data, &dw[3], values->SurfaceEndAddress, dw3);
+
+}
+
+#define GEN75_3DSTATE_SO_DECL_LIST_length_bias 0x00000002
+#define GEN75_3DSTATE_SO_DECL_LIST_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 23
+
+#define GEN75_3DSTATE_SO_DECL_LIST_length 0x00000000
+
+#define GEN75_SO_DECL_ENTRY_length 0x00000002
+
+#define GEN75_SO_DECL_length 0x00000001
+
+struct GEN75_SO_DECL {
+ uint32_t OutputBufferSlot;
+ uint32_t HoleFlag;
+ uint32_t RegisterIndex;
+ uint32_t ComponentMask;
+};
+
+static inline void
+GEN75_SO_DECL_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_SO_DECL * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->OutputBufferSlot, 12, 13) |
+ __gen_field(values->HoleFlag, 11, 11) |
+ __gen_field(values->RegisterIndex, 4, 9) |
+ __gen_field(values->ComponentMask, 0, 3) |
+ 0;
+
+}
+
+struct GEN75_SO_DECL_ENTRY {
+ struct GEN75_SO_DECL Stream3Decl;
+ struct GEN75_SO_DECL Stream2Decl;
+ struct GEN75_SO_DECL Stream1Decl;
+ struct GEN75_SO_DECL Stream0Decl;
+};
+
+static inline void
+GEN75_SO_DECL_ENTRY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_SO_DECL_ENTRY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ uint32_t dw_Stream3Decl;
+ GEN75_SO_DECL_pack(data, &dw_Stream3Decl, &values->Stream3Decl);
+ uint32_t dw_Stream2Decl;
+ GEN75_SO_DECL_pack(data, &dw_Stream2Decl, &values->Stream2Decl);
+ uint32_t dw_Stream1Decl;
+ GEN75_SO_DECL_pack(data, &dw_Stream1Decl, &values->Stream1Decl);
+ uint32_t dw_Stream0Decl;
+ GEN75_SO_DECL_pack(data, &dw_Stream0Decl, &values->Stream0Decl);
+ uint64_t qw0 =
+ __gen_field(dw_Stream3Decl, 48, 63) |
+ __gen_field(dw_Stream2Decl, 32, 47) |
+ __gen_field(dw_Stream1Decl, 16, 31) |
+ __gen_field(dw_Stream0Decl, 0, 15) |
+ 0;
+
+ dw[0] = qw0;
+ dw[1] = qw0 >> 32;
+
+}
+
+struct GEN75_3DSTATE_SO_DECL_LIST {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t StreamtoBufferSelects3;
+ uint32_t StreamtoBufferSelects2;
+ uint32_t StreamtoBufferSelects1;
+ uint32_t StreamtoBufferSelects0;
+ uint32_t NumEntries3;
+ uint32_t NumEntries2;
+ uint32_t NumEntries1;
+ uint32_t NumEntries0;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN75_3DSTATE_SO_DECL_LIST_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_SO_DECL_LIST * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 8) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->StreamtoBufferSelects3, 12, 15) |
+ __gen_field(values->StreamtoBufferSelects2, 8, 11) |
+ __gen_field(values->StreamtoBufferSelects1, 4, 7) |
+ __gen_field(values->StreamtoBufferSelects0, 0, 3) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->NumEntries3, 24, 31) |
+ __gen_field(values->NumEntries2, 16, 23) |
+ __gen_field(values->NumEntries1, 8, 15) |
+ __gen_field(values->NumEntries0, 0, 7) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN75_3DSTATE_STENCIL_BUFFER_length_bias 0x00000002
+#define GEN75_3DSTATE_STENCIL_BUFFER_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 6, \
+ .DwordLength = 1
+
+#define GEN75_3DSTATE_STENCIL_BUFFER_length 0x00000003
+
+struct GEN75_3DSTATE_STENCIL_BUFFER {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t StencilBufferEnable;
+ struct GEN75_MEMORY_OBJECT_CONTROL_STATE StencilBufferObjectControlState;
+ uint32_t SurfacePitch;
+ __gen_address_type SurfaceBaseAddress;
+};
+
+static inline void
+GEN75_3DSTATE_STENCIL_BUFFER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_STENCIL_BUFFER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw_StencilBufferObjectControlState;
+ GEN75_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_StencilBufferObjectControlState, &values->StencilBufferObjectControlState);
+ dw[1] =
+ __gen_field(values->StencilBufferEnable, 31, 31) |
+ __gen_field(dw_StencilBufferObjectControlState, 25, 28) |
+ __gen_field(values->SurfacePitch, 0, 16) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->SurfaceBaseAddress, dw2);
+
+}
+
+#define GEN75_3DSTATE_STREAMOUT_length_bias 0x00000002
+#define GEN75_3DSTATE_STREAMOUT_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 30, \
+ .DwordLength = 1
+
+#define GEN75_3DSTATE_STREAMOUT_length 0x00000003
+
+struct GEN75_3DSTATE_STREAMOUT {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t SOFunctionEnable;
+ uint32_t RenderingDisable;
+ uint32_t RenderStreamSelect;
+#define LEADING 0
+#define TRAILING 1
+ uint32_t ReorderMode;
+ bool SOStatisticsEnable;
+ uint32_t SOBufferEnable3;
+ uint32_t SOBufferEnable2;
+ uint32_t SOBufferEnable1;
+ uint32_t SOBufferEnable0;
+ uint32_t Stream3VertexReadOffset;
+ uint32_t Stream3VertexReadLength;
+ uint32_t Stream2VertexReadOffset;
+ uint32_t Stream2VertexReadLength;
+ uint32_t Stream1VertexReadOffset;
+ uint32_t Stream1VertexReadLength;
+ uint32_t Stream0VertexReadOffset;
+ uint32_t Stream0VertexReadLength;
+};
+
+static inline void
+GEN75_3DSTATE_STREAMOUT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_STREAMOUT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->SOFunctionEnable, 31, 31) |
+ __gen_field(values->RenderingDisable, 30, 30) |
+ __gen_field(values->RenderStreamSelect, 27, 28) |
+ __gen_field(values->ReorderMode, 26, 26) |
+ __gen_field(values->SOStatisticsEnable, 25, 25) |
+ __gen_field(values->SOBufferEnable3, 11, 11) |
+ __gen_field(values->SOBufferEnable2, 10, 10) |
+ __gen_field(values->SOBufferEnable1, 9, 9) |
+ __gen_field(values->SOBufferEnable0, 8, 8) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->Stream3VertexReadOffset, 29, 29) |
+ __gen_field(values->Stream3VertexReadLength, 24, 28) |
+ __gen_field(values->Stream2VertexReadOffset, 21, 21) |
+ __gen_field(values->Stream2VertexReadLength, 16, 20) |
+ __gen_field(values->Stream1VertexReadOffset, 13, 13) |
+ __gen_field(values->Stream1VertexReadLength, 8, 12) |
+ __gen_field(values->Stream0VertexReadOffset, 5, 5) |
+ __gen_field(values->Stream0VertexReadLength, 0, 4) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_TE_length_bias 0x00000002
+#define GEN75_3DSTATE_TE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 28, \
+ .DwordLength = 2
+
+#define GEN75_3DSTATE_TE_length 0x00000004
+
+struct GEN75_3DSTATE_TE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define INTEGER 0
+#define ODD_FRACTIONAL 1
+#define EVEN_FRACTIONAL 2
+ uint32_t Partitioning;
+#define POINT 0
+#define OUTPUT_LINE 1
+#define OUTPUT_TRI_CW 2
+#define OUTPUT_TRI_CCW 3
+ uint32_t OutputTopology;
+#define QUAD 0
+#define TRI 1
+#define ISOLINE 2
+ uint32_t TEDomain;
+#define HW_TESS 0
+#define SW_TESS 1
+ uint32_t TEMode;
+ bool TEEnable;
+ float MaximumTessellationFactorOdd;
+ float MaximumTessellationFactorNotOdd;
+};
+
+static inline void
+GEN75_3DSTATE_TE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_TE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->Partitioning, 12, 13) |
+ __gen_field(values->OutputTopology, 8, 9) |
+ __gen_field(values->TEDomain, 4, 5) |
+ __gen_field(values->TEMode, 1, 2) |
+ __gen_field(values->TEEnable, 0, 0) |
+ 0;
+
+ dw[2] =
+ __gen_float(values->MaximumTessellationFactorOdd) |
+ 0;
+
+ dw[3] =
+ __gen_float(values->MaximumTessellationFactorNotOdd) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_URB_DS_length_bias 0x00000002
+#define GEN75_3DSTATE_URB_DS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 50, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_URB_DS_length 0x00000002
+
+struct GEN75_3DSTATE_URB_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t DSURBStartingAddress;
+ uint32_t DSURBEntryAllocationSize;
+ uint32_t DSNumberofURBEntries;
+};
+
+static inline void
+GEN75_3DSTATE_URB_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_URB_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->DSURBStartingAddress, 25, 30) |
+ __gen_field(values->DSURBEntryAllocationSize, 16, 24) |
+ __gen_field(values->DSNumberofURBEntries, 0, 15) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_URB_GS_length_bias 0x00000002
+#define GEN75_3DSTATE_URB_GS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 51, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_URB_GS_length 0x00000002
+
+struct GEN75_3DSTATE_URB_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t GSURBStartingAddress;
+ uint32_t GSURBEntryAllocationSize;
+ uint32_t GSNumberofURBEntries;
+};
+
+static inline void
+GEN75_3DSTATE_URB_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_URB_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->GSURBStartingAddress, 25, 30) |
+ __gen_field(values->GSURBEntryAllocationSize, 16, 24) |
+ __gen_field(values->GSNumberofURBEntries, 0, 15) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_URB_HS_length_bias 0x00000002
+#define GEN75_3DSTATE_URB_HS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 49, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_URB_HS_length 0x00000002
+
+struct GEN75_3DSTATE_URB_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t HSURBStartingAddress;
+ uint32_t HSURBEntryAllocationSize;
+ uint32_t HSNumberofURBEntries;
+};
+
+static inline void
+GEN75_3DSTATE_URB_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_URB_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->HSURBStartingAddress, 25, 30) |
+ __gen_field(values->HSURBEntryAllocationSize, 16, 24) |
+ __gen_field(values->HSNumberofURBEntries, 0, 15) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_VERTEX_BUFFERS_length_bias 0x00000002
+#define GEN75_3DSTATE_VERTEX_BUFFERS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 8
+
+#define GEN75_3DSTATE_VERTEX_BUFFERS_length 0x00000000
+
+#define GEN75_VERTEX_BUFFER_STATE_length 0x00000004
+
+struct GEN75_VERTEX_BUFFER_STATE {
+ uint32_t VertexBufferIndex;
+#define VERTEXDATA 0
+#define INSTANCEDATA 1
+ uint32_t BufferAccessType;
+ struct GEN75_MEMORY_OBJECT_CONTROL_STATE VertexBufferMemoryObjectControlState;
+ uint32_t AddressModifyEnable;
+ bool NullVertexBuffer;
+ uint32_t VertexFetchInvalidate;
+ uint32_t BufferPitch;
+ __gen_address_type BufferStartingAddress;
+ __gen_address_type EndAddress;
+ uint32_t InstanceDataStepRate;
+};
+
+static inline void
+GEN75_VERTEX_BUFFER_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_VERTEX_BUFFER_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ uint32_t dw_VertexBufferMemoryObjectControlState;
+ GEN75_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_VertexBufferMemoryObjectControlState, &values->VertexBufferMemoryObjectControlState);
+ dw[0] =
+ __gen_field(values->VertexBufferIndex, 26, 31) |
+ __gen_field(values->BufferAccessType, 20, 20) |
+ __gen_field(dw_VertexBufferMemoryObjectControlState, 16, 19) |
+ __gen_field(values->AddressModifyEnable, 14, 14) |
+ __gen_field(values->NullVertexBuffer, 13, 13) |
+ __gen_field(values->VertexFetchInvalidate, 12, 12) |
+ __gen_field(values->BufferPitch, 0, 11) |
+ 0;
+
+ uint32_t dw1 =
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->BufferStartingAddress, dw1);
+
+ uint32_t dw2 =
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->EndAddress, dw2);
+
+ dw[3] =
+ __gen_field(values->InstanceDataStepRate, 0, 31) |
+ 0;
+
+}
+
+struct GEN75_3DSTATE_VERTEX_BUFFERS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN75_3DSTATE_VERTEX_BUFFERS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_VERTEX_BUFFERS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN75_3DSTATE_VERTEX_ELEMENTS_length_bias 0x00000002
+#define GEN75_3DSTATE_VERTEX_ELEMENTS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 9
+
+#define GEN75_3DSTATE_VERTEX_ELEMENTS_length 0x00000000
+
+#define GEN75_VERTEX_ELEMENT_STATE_length 0x00000002
+
+struct GEN75_VERTEX_ELEMENT_STATE {
+ uint32_t VertexBufferIndex;
+ bool Valid;
+ uint32_t SourceElementFormat;
+ bool EdgeFlagEnable;
+ uint32_t SourceElementOffset;
+ uint32_t Component0Control;
+ uint32_t Component1Control;
+ uint32_t Component2Control;
+ uint32_t Component3Control;
+};
+
+static inline void
+GEN75_VERTEX_ELEMENT_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_VERTEX_ELEMENT_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->VertexBufferIndex, 26, 31) |
+ __gen_field(values->Valid, 25, 25) |
+ __gen_field(values->SourceElementFormat, 16, 24) |
+ __gen_field(values->EdgeFlagEnable, 15, 15) |
+ __gen_field(values->SourceElementOffset, 0, 11) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->Component0Control, 28, 30) |
+ __gen_field(values->Component1Control, 24, 26) |
+ __gen_field(values->Component2Control, 20, 22) |
+ __gen_field(values->Component3Control, 16, 18) |
+ 0;
+
+}
+
+struct GEN75_3DSTATE_VERTEX_ELEMENTS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN75_3DSTATE_VERTEX_ELEMENTS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_VERTEX_ELEMENTS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN75_3DSTATE_VF_length_bias 0x00000002
+#define GEN75_3DSTATE_VF_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 12, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_VF_length 0x00000002
+
+struct GEN75_3DSTATE_VF {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ bool IndexedDrawCutIndexEnable;
+ uint32_t DwordLength;
+ uint32_t CutIndex;
+};
+
+static inline void
+GEN75_3DSTATE_VF_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_VF * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->IndexedDrawCutIndexEnable, 8, 8) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->CutIndex, 0, 31) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_VF_STATISTICS_length_bias 0x00000001
+#define GEN75_3DSTATE_VF_STATISTICS_header \
+ .CommandType = 3, \
+ .CommandSubType = 1, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 11
+
+#define GEN75_3DSTATE_VF_STATISTICS_length 0x00000001
+
+struct GEN75_3DSTATE_VF_STATISTICS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ bool StatisticsEnable;
+};
+
+static inline void
+GEN75_3DSTATE_VF_STATISTICS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_VF_STATISTICS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->StatisticsEnable, 0, 0) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_VIEWPORT_STATE_POINTERS_CC_length_bias 0x00000002
+#define GEN75_3DSTATE_VIEWPORT_STATE_POINTERS_CC_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 35, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_VIEWPORT_STATE_POINTERS_CC_length 0x00000002
+
+struct GEN75_3DSTATE_VIEWPORT_STATE_POINTERS_CC {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t CCViewportPointer;
+};
+
+static inline void
+GEN75_3DSTATE_VIEWPORT_STATE_POINTERS_CC_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_VIEWPORT_STATE_POINTERS_CC * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->CCViewportPointer, 5, 31) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP_length_bias 0x00000002
+#define GEN75_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 33, \
+ .DwordLength = 0
+
+#define GEN75_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP_length 0x00000002
+
+struct GEN75_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t SFClipViewportPointer;
+};
+
+static inline void
+GEN75_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->SFClipViewportPointer, 6, 31) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_VS_length_bias 0x00000002
+#define GEN75_3DSTATE_VS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 16, \
+ .DwordLength = 4
+
+#define GEN75_3DSTATE_VS_length 0x00000006
+
+struct GEN75_3DSTATE_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t KernelStartPointer;
+#define Multiple 0
+#define Single 1
+ uint32_t SingleVertexDispatch;
+#define Dmask 0
+#define Vmask 1
+ uint32_t VectorMaskEnableVME;
+#define NoSamplers 0
+#define _14Samplers 1
+#define _58Samplers 2
+#define _912Samplers 3
+#define _1316Samplers 4
+ uint32_t SamplerCount;
+ uint32_t BindingTableEntryCount;
+#define NormalPriority 0
+#define HighPriority 1
+ uint32_t ThreadPriority;
+#define IEEE754 0
+#define Alternate 1
+ uint32_t FloatingPointMode;
+ bool IllegalOpcodeExceptionEnable;
+ bool VSaccessesUAV;
+ bool SoftwareExceptionEnable;
+ uint32_t ScratchSpaceBaseOffset;
+ uint32_t PerThreadScratchSpace;
+ uint32_t DispatchGRFStartRegisterforURBData;
+ uint32_t VertexURBEntryReadLength;
+ uint32_t VertexURBEntryReadOffset;
+ uint32_t MaximumNumberofThreads;
+ bool StatisticsEnable;
+ bool VertexCacheDisable;
+ bool VSFunctionEnable;
+};
+
+static inline void
+GEN75_3DSTATE_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->KernelStartPointer, 6, 31) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->SingleVertexDispatch, 31, 31) |
+ __gen_field(values->VectorMaskEnableVME, 30, 30) |
+ __gen_field(values->SamplerCount, 27, 29) |
+ __gen_field(values->BindingTableEntryCount, 18, 25) |
+ __gen_field(values->ThreadPriority, 17, 17) |
+ __gen_field(values->FloatingPointMode, 16, 16) |
+ __gen_field(values->IllegalOpcodeExceptionEnable, 13, 13) |
+ __gen_field(values->VSaccessesUAV, 12, 12) |
+ __gen_field(values->SoftwareExceptionEnable, 7, 7) |
+ 0;
+
+ dw[3] =
+ __gen_offset(values->ScratchSpaceBaseOffset, 10, 31) |
+ __gen_field(values->PerThreadScratchSpace, 0, 3) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->DispatchGRFStartRegisterforURBData, 20, 24) |
+ __gen_field(values->VertexURBEntryReadLength, 11, 16) |
+ __gen_field(values->VertexURBEntryReadOffset, 4, 9) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->MaximumNumberofThreads, 23, 31) |
+ __gen_field(values->StatisticsEnable, 10, 10) |
+ __gen_field(values->VertexCacheDisable, 1, 1) |
+ __gen_field(values->VSFunctionEnable, 0, 0) |
+ 0;
+
+}
+
+#define GEN75_3DSTATE_WM_length_bias 0x00000002
+#define GEN75_3DSTATE_WM_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 20, \
+ .DwordLength = 1
+
+#define GEN75_3DSTATE_WM_length 0x00000003
+
+struct GEN75_3DSTATE_WM {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool StatisticsEnable;
+ bool DepthBufferClear;
+ bool ThreadDispatchEnable;
+ bool DepthBufferResolveEnable;
+ bool HierarchicalDepthBufferResolveEnable;
+ bool LegacyDiamondLineRasterization;
+ bool PixelShaderKillPixel;
+#define PSCDEPTH_OFF 0
+#define PSCDEPTH_ON 1
+#define PSCDEPTH_ON_GE 2
+#define PSCDEPTH_ON_LE 3
+ uint32_t PixelShaderComputedDepthMode;
+#define EDSC_NORMAL 0
+#define EDSC_PSEXEC 1
+#define EDSC_PREPS 2
+ uint32_t EarlyDepthStencilControl;
+ bool PixelShaderUsesSourceDepth;
+ bool PixelShaderUsesSourceW;
+#define INTERP_PIXEL 0
+#define INTERP_CENTROID 2
+#define INTERP_SAMPLE 3
+ uint32_t PositionZWInterpolationMode;
+ uint32_t BarycentricInterpolationMode;
+ bool PixelShaderUsesInputCoverageMask;
+ uint32_t LineEndCapAntialiasingRegionWidth;
+ uint32_t LineAntialiasingRegionWidth;
+ bool RTIndependentRasterizationEnable;
+ bool PolygonStippleEnable;
+ bool LineStippleEnable;
+#define RASTRULE_UPPER_LEFT 0
+#define RASTRULE_UPPER_RIGHT 1
+ uint32_t PointRasterizationRule;
+#define MSRASTMODE_OFF_PIXEL 0
+#define MSRASTMODE_OFF_PATTERN 1
+#define MSRASTMODE_ON_PIXEL 2
+#define MSRASTMODE_ON_PATTERN 3
+ uint32_t MultisampleRasterizationMode;
+#define MSDISPMODE_PERSAMPLE 0
+#define MSDISPMODE_PERPIXEL 1
+ uint32_t MultisampleDispatchMode;
+#define OFF 0
+#define ON 1
+ uint32_t PSUAVonly;
+};
+
+static inline void
+GEN75_3DSTATE_WM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_3DSTATE_WM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->StatisticsEnable, 31, 31) |
+ __gen_field(values->DepthBufferClear, 30, 30) |
+ __gen_field(values->ThreadDispatchEnable, 29, 29) |
+ __gen_field(values->DepthBufferResolveEnable, 28, 28) |
+ __gen_field(values->HierarchicalDepthBufferResolveEnable, 27, 27) |
+ __gen_field(values->LegacyDiamondLineRasterization, 26, 26) |
+ __gen_field(values->PixelShaderKillPixel, 25, 25) |
+ __gen_field(values->PixelShaderComputedDepthMode, 23, 24) |
+ __gen_field(values->EarlyDepthStencilControl, 21, 22) |
+ __gen_field(values->PixelShaderUsesSourceDepth, 20, 20) |
+ __gen_field(values->PixelShaderUsesSourceW, 19, 19) |
+ __gen_field(values->PositionZWInterpolationMode, 17, 18) |
+ __gen_field(values->BarycentricInterpolationMode, 11, 16) |
+ __gen_field(values->PixelShaderUsesInputCoverageMask, 10, 10) |
+ __gen_field(values->LineEndCapAntialiasingRegionWidth, 8, 9) |
+ __gen_field(values->LineAntialiasingRegionWidth, 6, 7) |
+ __gen_field(values->RTIndependentRasterizationEnable, 5, 5) |
+ __gen_field(values->PolygonStippleEnable, 4, 4) |
+ __gen_field(values->LineStippleEnable, 3, 3) |
+ __gen_field(values->PointRasterizationRule, 2, 2) |
+ __gen_field(values->MultisampleRasterizationMode, 0, 1) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->MultisampleDispatchMode, 31, 31) |
+ __gen_field(values->PSUAVonly, 30, 30) |
+ 0;
+
+}
+
+#define GEN75_GPGPU_OBJECT_length_bias 0x00000002
+#define GEN75_GPGPU_OBJECT_header \
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 1, \
+ .SubOpcode = 4, \
+ .DwordLength = 6
+
+#define GEN75_GPGPU_OBJECT_length 0x00000008
+
+struct GEN75_GPGPU_OBJECT {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ bool PredicateEnable;
+ uint32_t DwordLength;
+ uint32_t SharedLocalMemoryFixedOffset;
+ uint32_t InterfaceDescriptorOffset;
+ uint32_t SharedLocalMemoryOffset;
+ uint32_t EndofThreadGroup;
+#define Slice0 0
+#define Slice1 1
+ uint32_t SliceDestinationSelect;
+#define HalfSlice1 2
+#define HalfSlice0 1
+#define EitherHalfSlice 0
+ uint32_t HalfSliceDestinationSelect;
+ uint32_t IndirectDataLength;
+ uint32_t IndirectDataStartAddress;
+ uint32_t ThreadGroupIDX;
+ uint32_t ThreadGroupIDY;
+ uint32_t ThreadGroupIDZ;
+ uint32_t ExecutionMask;
+};
+
+static inline void
+GEN75_GPGPU_OBJECT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_GPGPU_OBJECT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->PredicateEnable, 8, 8) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->SharedLocalMemoryFixedOffset, 7, 7) |
+ __gen_field(values->InterfaceDescriptorOffset, 0, 5) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->SharedLocalMemoryOffset, 28, 31) |
+ __gen_field(values->EndofThreadGroup, 24, 24) |
+ __gen_field(values->SliceDestinationSelect, 19, 19) |
+ __gen_field(values->HalfSliceDestinationSelect, 17, 18) |
+ __gen_field(values->IndirectDataLength, 0, 16) |
+ 0;
+
+ dw[3] =
+ __gen_offset(values->IndirectDataStartAddress, 0, 31) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->ThreadGroupIDX, 0, 31) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->ThreadGroupIDY, 0, 31) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->ThreadGroupIDZ, 0, 31) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->ExecutionMask, 0, 31) |
+ 0;
+
+}
+
+#define GEN75_GPGPU_WALKER_length_bias 0x00000002
+#define GEN75_GPGPU_WALKER_header \
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 1, \
+ .SubOpcodeA = 5, \
+ .DwordLength = 9
+
+#define GEN75_GPGPU_WALKER_length 0x0000000b
+
+struct GEN75_GPGPU_WALKER {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcodeA;
+ bool IndirectParameterEnable;
+ bool PredicateEnable;
+ uint32_t DwordLength;
+ uint32_t InterfaceDescriptorOffset;
+#define SIMD8 0
+#define SIMD16 1
+#define SIMD32 2
+ uint32_t SIMDSize;
+ uint32_t ThreadDepthCounterMaximum;
+ uint32_t ThreadHeightCounterMaximum;
+ uint32_t ThreadWidthCounterMaximum;
+ uint32_t ThreadGroupIDStartingX;
+ uint32_t ThreadGroupIDXDimension;
+ uint32_t ThreadGroupIDStartingY;
+ uint32_t ThreadGroupIDYDimension;
+ uint32_t ThreadGroupIDStartingZ;
+ uint32_t ThreadGroupIDZDimension;
+ uint32_t RightExecutionMask;
+ uint32_t BottomExecutionMask;
+};
+
+static inline void
+GEN75_GPGPU_WALKER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_GPGPU_WALKER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcodeA, 16, 23) |
+ __gen_field(values->IndirectParameterEnable, 10, 10) |
+ __gen_field(values->PredicateEnable, 8, 8) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->InterfaceDescriptorOffset, 0, 5) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->SIMDSize, 30, 31) |
+ __gen_field(values->ThreadDepthCounterMaximum, 16, 21) |
+ __gen_field(values->ThreadHeightCounterMaximum, 8, 13) |
+ __gen_field(values->ThreadWidthCounterMaximum, 0, 5) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->ThreadGroupIDStartingX, 0, 31) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->ThreadGroupIDXDimension, 0, 31) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->ThreadGroupIDStartingY, 0, 31) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->ThreadGroupIDYDimension, 0, 31) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->ThreadGroupIDStartingZ, 0, 31) |
+ 0;
+
+ dw[8] =
+ __gen_field(values->ThreadGroupIDZDimension, 0, 31) |
+ 0;
+
+ dw[9] =
+ __gen_field(values->RightExecutionMask, 0, 31) |
+ 0;
+
+ dw[10] =
+ __gen_field(values->BottomExecutionMask, 0, 31) |
+ 0;
+
+}
+
+#define GEN75_MEDIA_CURBE_LOAD_length_bias 0x00000002
+#define GEN75_MEDIA_CURBE_LOAD_header \
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 0, \
+ .SubOpcode = 1, \
+ .DwordLength = 2
+
+#define GEN75_MEDIA_CURBE_LOAD_length 0x00000004
+
+struct GEN75_MEDIA_CURBE_LOAD {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ uint32_t DwordLength;
+ uint32_t CURBETotalDataLength;
+ uint32_t CURBEDataStartAddress;
+};
+
+static inline void
+GEN75_MEDIA_CURBE_LOAD_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MEDIA_CURBE_LOAD * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ 0;
+
+ dw[2] =
+ __gen_field(values->CURBETotalDataLength, 0, 16) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->CURBEDataStartAddress, 0, 31) |
+ 0;
+
+}
+
+#define GEN75_MEDIA_INTERFACE_DESCRIPTOR_LOAD_length_bias 0x00000002
+#define GEN75_MEDIA_INTERFACE_DESCRIPTOR_LOAD_header\
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 0, \
+ .SubOpcode = 2, \
+ .DwordLength = 2
+
+#define GEN75_MEDIA_INTERFACE_DESCRIPTOR_LOAD_length 0x00000004
+
+struct GEN75_MEDIA_INTERFACE_DESCRIPTOR_LOAD {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ uint32_t DwordLength;
+ uint32_t InterfaceDescriptorTotalLength;
+ uint32_t InterfaceDescriptorDataStartAddress;
+};
+
+static inline void
+GEN75_MEDIA_INTERFACE_DESCRIPTOR_LOAD_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MEDIA_INTERFACE_DESCRIPTOR_LOAD * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ 0;
+
+ dw[2] =
+ __gen_field(values->InterfaceDescriptorTotalLength, 0, 16) |
+ 0;
+
+ dw[3] =
+ __gen_offset(values->InterfaceDescriptorDataStartAddress, 0, 31) |
+ 0;
+
+}
+
+#define GEN75_MEDIA_OBJECT_length_bias 0x00000002
+#define GEN75_MEDIA_OBJECT_header \
+ .CommandType = 3, \
+ .MediaCommandPipeline = 2, \
+ .MediaCommandOpcode = 1, \
+ .MediaCommandSubOpcode = 0
+
+#define GEN75_MEDIA_OBJECT_length 0x00000000
+
+struct GEN75_MEDIA_OBJECT {
+ uint32_t CommandType;
+ uint32_t MediaCommandPipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t MediaCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t InterfaceDescriptorOffset;
+ bool ChildrenPresent;
+#define Nothreadsynchronization 0
+#define Threaddispatchissynchronizedbythespawnrootthreadmessage 1
+ uint32_t ThreadSynchronization;
+#define Notusingscoreboard 0
+#define Usingscoreboard 1
+ uint32_t UseScoreboard;
+#define Slice0 0
+#define Slice1 1
+#define EitherSlice 0
+ uint32_t SliceDestinationSelect;
+#define HalfSlice1 2
+#define HalfSlice0 1
+#define Eitherhalfslice 0
+ uint32_t HalfSliceDestinationSelect;
+ uint32_t IndirectDataLength;
+ __gen_address_type IndirectDataStartAddress;
+ uint32_t ScoredboardY;
+ uint32_t ScoreboardX;
+ uint32_t ScoreboardColor;
+ bool ScoreboardMask;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN75_MEDIA_OBJECT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MEDIA_OBJECT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MediaCommandPipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->MediaCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->InterfaceDescriptorOffset, 0, 5) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ChildrenPresent, 31, 31) |
+ __gen_field(values->ThreadSynchronization, 24, 24) |
+ __gen_field(values->UseScoreboard, 21, 21) |
+ __gen_field(values->SliceDestinationSelect, 19, 19) |
+ __gen_field(values->HalfSliceDestinationSelect, 17, 18) |
+ __gen_field(values->IndirectDataLength, 0, 16) |
+ 0;
+
+ uint32_t dw3 =
+ 0;
+
+ dw[3] =
+ __gen_combine_address(data, &dw[3], values->IndirectDataStartAddress, dw3);
+
+ dw[4] =
+ __gen_field(values->ScoredboardY, 16, 24) |
+ __gen_field(values->ScoreboardX, 0, 8) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->ScoreboardColor, 16, 19) |
+ __gen_field(values->ScoreboardMask, 0, 7) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN75_MEDIA_OBJECT_PRT_length_bias 0x00000002
+#define GEN75_MEDIA_OBJECT_PRT_header \
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 1, \
+ .SubOpcode = 2, \
+ .DwordLength = 14
+
+#define GEN75_MEDIA_OBJECT_PRT_length 0x00000010
+
+struct GEN75_MEDIA_OBJECT_PRT {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ uint32_t DwordLength;
+ uint32_t InterfaceDescriptorOffset;
+ bool ChildrenPresent;
+ bool PRT_FenceNeeded;
+#define Rootthreadqueue 0
+#define VFEstateflush 1
+ uint32_t PRT_FenceType;
+ uint32_t InlineData[12];
+};
+
+static inline void
+GEN75_MEDIA_OBJECT_PRT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MEDIA_OBJECT_PRT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->InterfaceDescriptorOffset, 0, 5) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ChildrenPresent, 31, 31) |
+ __gen_field(values->PRT_FenceNeeded, 23, 23) |
+ __gen_field(values->PRT_FenceType, 22, 22) |
+ 0;
+
+ dw[3] =
+ 0;
+
+ for (uint32_t i = 0, j = 4; i < 12; i += 1, j++) {
+ dw[j] =
+ __gen_field(values->InlineData[i + 0], 0, 31) |
+ 0;
+ }
+
+}
+
+#define GEN75_MEDIA_OBJECT_WALKER_length_bias 0x00000002
+#define GEN75_MEDIA_OBJECT_WALKER_header \
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 1, \
+ .SubOpcode = 3
+
+#define GEN75_MEDIA_OBJECT_WALKER_length 0x00000000
+
+struct GEN75_MEDIA_OBJECT_WALKER {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ uint32_t DwordLength;
+ uint32_t InterfaceDescriptorOffset;
+ bool ChildrenPresent;
+#define Nothreadsynchronization 0
+#define Threaddispatchissynchronizedbythespawnrootthreadmessage 1
+ uint32_t ThreadSynchronization;
+#define Notusingscoreboard 0
+#define Usingscoreboard 1
+ uint32_t UseScoreboard;
+ uint32_t IndirectDataLength;
+ uint32_t IndirectDataStartAddress;
+ bool ScoreboardMask;
+ bool DualMode;
+ bool Repel;
+ bool QuadMode;
+ uint32_t ColorCountMinusOne;
+ uint32_t MiddleLoopExtraSteps;
+ uint32_t LocalMidLoopUnitY;
+ uint32_t MidLoopUnitX;
+ uint32_t GlobalLoopExecCount;
+ uint32_t LocalLoopExecCount;
+ uint32_t BlockResolutionY;
+ uint32_t BlockResolutionX;
+ uint32_t LocalStartY;
+ uint32_t LocalStartX;
+ uint32_t LocalOuterLoopStrideY;
+ uint32_t LocalOuterLoopStrideX;
+ uint32_t LocalInnerLoopUnitY;
+ uint32_t LocalInnerLoopUnitX;
+ uint32_t GlobalResolutionY;
+ uint32_t GlobalResolutionX;
+ uint32_t GlobalStartY;
+ uint32_t GlobalStartX;
+ uint32_t GlobalOuterLoopStrideY;
+ uint32_t GlobalOuterLoopStrideX;
+ uint32_t GlobalInnerLoopUnitY;
+ uint32_t GlobalInnerLoopUnitX;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN75_MEDIA_OBJECT_WALKER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MEDIA_OBJECT_WALKER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->InterfaceDescriptorOffset, 0, 5) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ChildrenPresent, 31, 31) |
+ __gen_field(values->ThreadSynchronization, 24, 24) |
+ __gen_field(values->UseScoreboard, 21, 21) |
+ __gen_field(values->IndirectDataLength, 0, 16) |
+ 0;
+
+ dw[3] =
+ __gen_offset(values->IndirectDataStartAddress, 0, 31) |
+ 0;
+
+ dw[4] =
+ 0;
+
+ dw[5] =
+ __gen_field(values->ScoreboardMask, 0, 7) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->DualMode, 31, 31) |
+ __gen_field(values->Repel, 30, 30) |
+ __gen_field(values->QuadMode, 29, 29) |
+ __gen_field(values->ColorCountMinusOne, 24, 27) |
+ __gen_field(values->MiddleLoopExtraSteps, 16, 20) |
+ __gen_field(values->LocalMidLoopUnitY, 12, 13) |
+ __gen_field(values->MidLoopUnitX, 8, 9) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->GlobalLoopExecCount, 16, 25) |
+ __gen_field(values->LocalLoopExecCount, 0, 9) |
+ 0;
+
+ dw[8] =
+ __gen_field(values->BlockResolutionY, 16, 24) |
+ __gen_field(values->BlockResolutionX, 0, 8) |
+ 0;
+
+ dw[9] =
+ __gen_field(values->LocalStartY, 16, 24) |
+ __gen_field(values->LocalStartX, 0, 8) |
+ 0;
+
+ dw[10] =
+ 0;
+
+ dw[11] =
+ __gen_field(values->LocalOuterLoopStrideY, 16, 25) |
+ __gen_field(values->LocalOuterLoopStrideX, 0, 9) |
+ 0;
+
+ dw[12] =
+ __gen_field(values->LocalInnerLoopUnitY, 16, 25) |
+ __gen_field(values->LocalInnerLoopUnitX, 0, 9) |
+ 0;
+
+ dw[13] =
+ __gen_field(values->GlobalResolutionY, 16, 24) |
+ __gen_field(values->GlobalResolutionX, 0, 8) |
+ 0;
+
+ dw[14] =
+ __gen_field(values->GlobalStartY, 16, 25) |
+ __gen_field(values->GlobalStartX, 0, 9) |
+ 0;
+
+ dw[15] =
+ __gen_field(values->GlobalOuterLoopStrideY, 16, 25) |
+ __gen_field(values->GlobalOuterLoopStrideX, 0, 9) |
+ 0;
+
+ dw[16] =
+ __gen_field(values->GlobalInnerLoopUnitY, 16, 25) |
+ __gen_field(values->GlobalInnerLoopUnitX, 0, 9) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN75_MEDIA_STATE_FLUSH_length_bias 0x00000002
+#define GEN75_MEDIA_STATE_FLUSH_header \
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 0, \
+ .SubOpcode = 4, \
+ .DwordLength = 0
+
+#define GEN75_MEDIA_STATE_FLUSH_length 0x00000002
+
+struct GEN75_MEDIA_STATE_FLUSH {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ uint32_t DwordLength;
+ bool DisablePreemption;
+ bool FlushtoGO;
+ uint32_t WatermarkRequired;
+ uint32_t InterfaceDescriptorOffset;
+};
+
+static inline void
+GEN75_MEDIA_STATE_FLUSH_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MEDIA_STATE_FLUSH * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->DisablePreemption, 8, 8) |
+ __gen_field(values->FlushtoGO, 7, 7) |
+ __gen_field(values->WatermarkRequired, 6, 6) |
+ __gen_field(values->InterfaceDescriptorOffset, 0, 5) |
+ 0;
+
+}
+
+#define GEN75_MEDIA_VFE_STATE_length_bias 0x00000002
+#define GEN75_MEDIA_VFE_STATE_header \
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 0, \
+ .SubOpcode = 0, \
+ .DwordLength = 6
+
+#define GEN75_MEDIA_VFE_STATE_length 0x00000008
+
+struct GEN75_MEDIA_VFE_STATE {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ uint32_t DwordLength;
+ uint32_t ScratchSpaceBasePointer;
+ uint32_t StackSize;
+ uint32_t PerThreadScratchSpace;
+ uint32_t MaximumNumberofThreads;
+ uint32_t NumberofURBEntries;
+#define Maintainingtheexistingtimestampstate 0
+#define Resettingrelativetimerandlatchingtheglobaltimestamp 1
+ uint32_t ResetGatewayTimer;
+#define MaintainingOpenGatewayForwardMsgCloseGatewayprotocollegacymode 0
+#define BypassingOpenGatewayCloseGatewayprotocol 1
+ uint32_t BypassGatewayControl;
+ uint32_t GPGPUMode;
+ uint32_t HalfSliceDisable;
+ uint32_t URBEntryAllocationSize;
+ uint32_t CURBEAllocationSize;
+#define Scoreboarddisabled 0
+#define Scoreboardenabled 1
+ uint32_t ScoreboardEnable;
+#define StallingScoreboard 0
+#define NonStallingScoreboard 1
+ uint32_t ScoreboardType;
+ uint32_t ScoreboardMask;
+ uint32_t Scoreboard3DeltaY;
+ uint32_t Scoreboard3DeltaX;
+ uint32_t Scoreboard2DeltaY;
+ uint32_t Scoreboard2DeltaX;
+ uint32_t Scoreboard1DeltaY;
+ uint32_t Scoreboard1DeltaX;
+ uint32_t Scoreboard0DeltaY;
+ uint32_t Scoreboard0DeltaX;
+ uint32_t Scoreboard7DeltaY;
+ uint32_t Scoreboard7DeltaX;
+ uint32_t Scoreboard6DeltaY;
+ uint32_t Scoreboard6DeltaX;
+ uint32_t Scoreboard5DeltaY;
+ uint32_t Scoreboard5DeltaX;
+ uint32_t Scoreboard4DeltaY;
+ uint32_t Scoreboard4DeltaX;
+};
+
+static inline void
+GEN75_MEDIA_VFE_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MEDIA_VFE_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->ScratchSpaceBasePointer, 10, 31) |
+ __gen_field(values->StackSize, 4, 7) |
+ __gen_field(values->PerThreadScratchSpace, 0, 3) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->MaximumNumberofThreads, 16, 31) |
+ __gen_field(values->NumberofURBEntries, 8, 15) |
+ __gen_field(values->ResetGatewayTimer, 7, 7) |
+ __gen_field(values->BypassGatewayControl, 6, 6) |
+ __gen_field(values->GPGPUMode, 2, 2) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->HalfSliceDisable, 0, 1) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->URBEntryAllocationSize, 16, 31) |
+ __gen_field(values->CURBEAllocationSize, 0, 15) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->ScoreboardEnable, 31, 31) |
+ __gen_field(values->ScoreboardType, 30, 30) |
+ __gen_field(values->ScoreboardMask, 0, 7) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->Scoreboard3DeltaY, 28, 31) |
+ __gen_field(values->Scoreboard3DeltaX, 24, 27) |
+ __gen_field(values->Scoreboard2DeltaY, 20, 23) |
+ __gen_field(values->Scoreboard2DeltaX, 16, 19) |
+ __gen_field(values->Scoreboard1DeltaY, 12, 15) |
+ __gen_field(values->Scoreboard1DeltaX, 8, 11) |
+ __gen_field(values->Scoreboard0DeltaY, 4, 7) |
+ __gen_field(values->Scoreboard0DeltaX, 0, 3) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->Scoreboard7DeltaY, 28, 31) |
+ __gen_field(values->Scoreboard7DeltaX, 24, 27) |
+ __gen_field(values->Scoreboard6DeltaY, 20, 23) |
+ __gen_field(values->Scoreboard6DeltaX, 16, 19) |
+ __gen_field(values->Scoreboard5DeltaY, 12, 15) |
+ __gen_field(values->Scoreboard5DeltaX, 8, 11) |
+ __gen_field(values->Scoreboard4DeltaY, 4, 7) |
+ __gen_field(values->Scoreboard4DeltaX, 0, 3) |
+ 0;
+
+}
+
+#define GEN75_MI_ARB_CHECK_length_bias 0x00000001
+#define GEN75_MI_ARB_CHECK_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 5
+
+#define GEN75_MI_ARB_CHECK_length 0x00000001
+
+struct GEN75_MI_ARB_CHECK {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+};
+
+static inline void
+GEN75_MI_ARB_CHECK_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_ARB_CHECK * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ 0;
+
+}
+
+#define GEN75_MI_ARB_ON_OFF_length_bias 0x00000001
+#define GEN75_MI_ARB_ON_OFF_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 8
+
+#define GEN75_MI_ARB_ON_OFF_length 0x00000001
+
+struct GEN75_MI_ARB_ON_OFF {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool ArbitrationEnable;
+};
+
+static inline void
+GEN75_MI_ARB_ON_OFF_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_ARB_ON_OFF * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->ArbitrationEnable, 0, 0) |
+ 0;
+
+}
+
+#define GEN75_MI_BATCH_BUFFER_END_length_bias 0x00000001
+#define GEN75_MI_BATCH_BUFFER_END_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 10
+
+#define GEN75_MI_BATCH_BUFFER_END_length 0x00000001
+
+struct GEN75_MI_BATCH_BUFFER_END {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+};
+
+static inline void
+GEN75_MI_BATCH_BUFFER_END_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_BATCH_BUFFER_END * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ 0;
+
+}
+
+#define GEN75_MI_BATCH_BUFFER_START_length_bias 0x00000002
+#define GEN75_MI_BATCH_BUFFER_START_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 49, \
+ .DwordLength = 0
+
+#define GEN75_MI_BATCH_BUFFER_START_length 0x00000002
+
+struct GEN75_MI_BATCH_BUFFER_START {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define _1stlevelbatch 0
+#define _2ndlevelbatch 1
+ uint32_t _2ndLevelBatchBuffer;
+ bool AddOffsetEnable;
+ bool PredicationEnable;
+ uint32_t NonPrivileged;
+ bool ClearCommandBufferEnable;
+ bool ResourceStreamerEnable;
+#define ASI_GGTT 0
+#define ASI_PPGTT 1
+ uint32_t AddressSpaceIndicator;
+ uint32_t DwordLength;
+ __gen_address_type BatchBufferStartAddress;
+};
+
+static inline void
+GEN75_MI_BATCH_BUFFER_START_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_BATCH_BUFFER_START * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->_2ndLevelBatchBuffer, 22, 22) |
+ __gen_field(values->AddOffsetEnable, 16, 16) |
+ __gen_field(values->PredicationEnable, 15, 15) |
+ __gen_field(values->NonPrivileged, 13, 13) |
+ __gen_field(values->ClearCommandBufferEnable, 11, 11) |
+ __gen_field(values->ResourceStreamerEnable, 10, 10) |
+ __gen_field(values->AddressSpaceIndicator, 8, 8) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw1 =
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->BatchBufferStartAddress, dw1);
+
+}
+
+#define GEN75_MI_CLFLUSH_length_bias 0x00000002
+#define GEN75_MI_CLFLUSH_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 39
+
+#define GEN75_MI_CLFLUSH_length 0x00000000
+
+struct GEN75_MI_CLFLUSH {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define PerProcessGraphicsAddress 0
+#define GlobalGraphicsAddress 1
+ uint32_t UseGlobalGTT;
+ uint32_t DwordLength;
+ __gen_address_type PageBaseAddress;
+ uint32_t StartingCachelineOffset;
+ __gen_address_type PageBaseAddressHigh;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN75_MI_CLFLUSH_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_CLFLUSH * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->UseGlobalGTT, 22, 22) |
+ __gen_field(values->DwordLength, 0, 9) |
+ 0;
+
+ uint32_t dw1 =
+ __gen_field(values->StartingCachelineOffset, 6, 11) |
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->PageBaseAddress, dw1);
+
+ uint32_t dw2 =
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->PageBaseAddressHigh, dw2);
+
+ /* variable length fields follow */
+}
+
+#define GEN75_MI_CONDITIONAL_BATCH_BUFFER_END_length_bias 0x00000002
+#define GEN75_MI_CONDITIONAL_BATCH_BUFFER_END_header\
+ .CommandType = 0, \
+ .MICommandOpcode = 54, \
+ .UseGlobalGTT = 0, \
+ .CompareSemaphore = 0, \
+ .DwordLength = 0
+
+#define GEN75_MI_CONDITIONAL_BATCH_BUFFER_END_length 0x00000002
+
+struct GEN75_MI_CONDITIONAL_BATCH_BUFFER_END {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t UseGlobalGTT;
+ uint32_t CompareSemaphore;
+ uint32_t DwordLength;
+ uint32_t CompareDataDword;
+ __gen_address_type CompareAddress;
+};
+
+static inline void
+GEN75_MI_CONDITIONAL_BATCH_BUFFER_END_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_CONDITIONAL_BATCH_BUFFER_END * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->UseGlobalGTT, 22, 22) |
+ __gen_field(values->CompareSemaphore, 21, 21) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->CompareDataDword, 0, 31) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->CompareAddress, dw2);
+
+}
+
+#define GEN75_MI_FLUSH_length_bias 0x00000001
+#define GEN75_MI_FLUSH_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 4
+
+#define GEN75_MI_FLUSH_length 0x00000001
+
+struct GEN75_MI_FLUSH {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool IndirectStatePointersDisable;
+ bool GenericMediaStateClear;
+#define DontReset 0
+#define Reset 1
+ bool GlobalSnapshotCountReset;
+#define Flush 0
+#define DontFlush 1
+ bool RenderCacheFlushInhibit;
+#define DontInvalidate 0
+#define Invalidate 1
+ bool StateInstructionCacheInvalidate;
+};
+
+static inline void
+GEN75_MI_FLUSH_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_FLUSH * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->IndirectStatePointersDisable, 5, 5) |
+ __gen_field(values->GenericMediaStateClear, 4, 4) |
+ __gen_field(values->GlobalSnapshotCountReset, 3, 3) |
+ __gen_field(values->RenderCacheFlushInhibit, 2, 2) |
+ __gen_field(values->StateInstructionCacheInvalidate, 1, 1) |
+ 0;
+
+}
+
+#define GEN75_MI_LOAD_REGISTER_IMM_length_bias 0x00000002
+#define GEN75_MI_LOAD_REGISTER_IMM_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 34, \
+ .DwordLength = 1
+
+#define GEN75_MI_LOAD_REGISTER_IMM_length 0x00000003
+
+struct GEN75_MI_LOAD_REGISTER_IMM {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t ByteWriteDisables;
+ uint32_t DwordLength;
+ uint32_t RegisterOffset;
+ uint32_t DataDWord;
+};
+
+static inline void
+GEN75_MI_LOAD_REGISTER_IMM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_LOAD_REGISTER_IMM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->ByteWriteDisables, 8, 11) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->RegisterOffset, 2, 22) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->DataDWord, 0, 31) |
+ 0;
+
+}
+
+#define GEN75_MI_LOAD_REGISTER_MEM_length_bias 0x00000002
+#define GEN75_MI_LOAD_REGISTER_MEM_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 41, \
+ .DwordLength = 1
+
+#define GEN75_MI_LOAD_REGISTER_MEM_length 0x00000003
+
+struct GEN75_MI_LOAD_REGISTER_MEM {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool UseGlobalGTT;
+ uint32_t AsyncModeEnable;
+ uint32_t DwordLength;
+ uint32_t RegisterAddress;
+ __gen_address_type MemoryAddress;
+};
+
+static inline void
+GEN75_MI_LOAD_REGISTER_MEM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_LOAD_REGISTER_MEM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->UseGlobalGTT, 22, 22) |
+ __gen_field(values->AsyncModeEnable, 21, 21) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->RegisterAddress, 2, 22) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->MemoryAddress, dw2);
+
+}
+
+#define GEN75_MI_LOAD_REGISTER_REG_length_bias 0x00000002
+#define GEN75_MI_LOAD_REGISTER_REG_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 42, \
+ .DwordLength = 1
+
+#define GEN75_MI_LOAD_REGISTER_REG_length 0x00000003
+
+struct GEN75_MI_LOAD_REGISTER_REG {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t DwordLength;
+ uint32_t SourceRegisterAddress;
+ uint32_t DestinationRegisterAddress;
+};
+
+static inline void
+GEN75_MI_LOAD_REGISTER_REG_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_LOAD_REGISTER_REG * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->SourceRegisterAddress, 2, 22) |
+ 0;
+
+ dw[2] =
+ __gen_offset(values->DestinationRegisterAddress, 2, 22) |
+ 0;
+
+}
+
+#define GEN75_MI_LOAD_SCAN_LINES_EXCL_length_bias 0x00000002
+#define GEN75_MI_LOAD_SCAN_LINES_EXCL_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 19, \
+ .DwordLength = 0
+
+#define GEN75_MI_LOAD_SCAN_LINES_EXCL_length 0x00000002
+
+struct GEN75_MI_LOAD_SCAN_LINES_EXCL {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define DisplayPlaneA 0
+#define DisplayPlaneB 1
+#define DisplayPlaneC 4
+ uint32_t DisplayPlaneSelect;
+ uint32_t DwordLength;
+ uint32_t StartScanLineNumber;
+ uint32_t EndScanLineNumber;
+};
+
+static inline void
+GEN75_MI_LOAD_SCAN_LINES_EXCL_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_LOAD_SCAN_LINES_EXCL * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DisplayPlaneSelect, 19, 21) |
+ __gen_field(values->DwordLength, 0, 5) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->StartScanLineNumber, 16, 28) |
+ __gen_field(values->EndScanLineNumber, 0, 12) |
+ 0;
+
+}
+
+#define GEN75_MI_LOAD_SCAN_LINES_INCL_length_bias 0x00000002
+#define GEN75_MI_LOAD_SCAN_LINES_INCL_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 18, \
+ .DwordLength = 0
+
+#define GEN75_MI_LOAD_SCAN_LINES_INCL_length 0x00000002
+
+struct GEN75_MI_LOAD_SCAN_LINES_INCL {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define DisplayPlaneA 0
+#define DisplayPlaneB 1
+#define DisplayPlaneC 4
+ uint32_t DisplayPlaneSelect;
+ uint32_t DwordLength;
+ uint32_t StartScanLineNumber;
+ uint32_t EndScanLineNumber;
+};
+
+static inline void
+GEN75_MI_LOAD_SCAN_LINES_INCL_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_LOAD_SCAN_LINES_INCL * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DisplayPlaneSelect, 19, 21) |
+ __gen_field(values->DwordLength, 0, 5) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->StartScanLineNumber, 16, 28) |
+ __gen_field(values->EndScanLineNumber, 0, 12) |
+ 0;
+
+}
+
+#define GEN75_MI_LOAD_URB_MEM_length_bias 0x00000002
+#define GEN75_MI_LOAD_URB_MEM_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 44, \
+ .DwordLength = 1
+
+#define GEN75_MI_LOAD_URB_MEM_length 0x00000003
+
+struct GEN75_MI_LOAD_URB_MEM {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t DwordLength;
+ uint32_t URBAddress;
+ __gen_address_type MemoryAddress;
+};
+
+static inline void
+GEN75_MI_LOAD_URB_MEM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_LOAD_URB_MEM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->URBAddress, 2, 14) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->MemoryAddress, dw2);
+
+}
+
+#define GEN75_MI_MATH_length_bias 0x00000002
+#define GEN75_MI_MATH_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 26
+
+#define GEN75_MI_MATH_length 0x00000000
+
+struct GEN75_MI_MATH {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t DwordLength;
+ uint32_t ALUINSTRUCTION1;
+ uint32_t ALUINSTRUCTION2;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN75_MI_MATH_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_MATH * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DwordLength, 0, 5) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ALUINSTRUCTION1, 0, 31) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ALUINSTRUCTION2, 0, 31) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN75_MI_NOOP_length_bias 0x00000001
+#define GEN75_MI_NOOP_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 0
+
+#define GEN75_MI_NOOP_length 0x00000001
+
+struct GEN75_MI_NOOP {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool IdentificationNumberRegisterWriteEnable;
+ uint32_t IdentificationNumber;
+};
+
+static inline void
+GEN75_MI_NOOP_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_NOOP * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->IdentificationNumberRegisterWriteEnable, 22, 22) |
+ __gen_field(values->IdentificationNumber, 0, 21) |
+ 0;
+
+}
+
+#define GEN75_MI_PREDICATE_length_bias 0x00000001
+#define GEN75_MI_PREDICATE_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 12
+
+#define GEN75_MI_PREDICATE_length 0x00000001
+
+struct GEN75_MI_PREDICATE {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define LOAD_KEEP 0
+#define LOAD_LOAD 2
+#define LOAD_LOADINV 3
+ uint32_t LoadOperation;
+#define COMBINE_SET 0
+#define COMBINE_AND 1
+#define COMBINE_OR 2
+#define COMBINE_XOR 3
+ uint32_t CombineOperation;
+#define COMPARE_SRCS_EQUAL 2
+#define COMPARE_DELTAS_EQUAL 3
+ uint32_t CompareOperation;
+};
+
+static inline void
+GEN75_MI_PREDICATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_PREDICATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->LoadOperation, 6, 7) |
+ __gen_field(values->CombineOperation, 3, 4) |
+ __gen_field(values->CompareOperation, 0, 1) |
+ 0;
+
+}
+
+#define GEN75_MI_REPORT_HEAD_length_bias 0x00000001
+#define GEN75_MI_REPORT_HEAD_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 7
+
+#define GEN75_MI_REPORT_HEAD_length 0x00000001
+
+struct GEN75_MI_REPORT_HEAD {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+};
+
+static inline void
+GEN75_MI_REPORT_HEAD_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_REPORT_HEAD * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ 0;
+
+}
+
+#define GEN75_MI_RS_CONTEXT_length_bias 0x00000001
+#define GEN75_MI_RS_CONTEXT_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 15
+
+#define GEN75_MI_RS_CONTEXT_length 0x00000001
+
+struct GEN75_MI_RS_CONTEXT {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define RS_RESTORE 0
+#define RS_SAVE 1
+ uint32_t ResourceStreamerSave;
+};
+
+static inline void
+GEN75_MI_RS_CONTEXT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_RS_CONTEXT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->ResourceStreamerSave, 0, 0) |
+ 0;
+
+}
+
+#define GEN75_MI_RS_CONTROL_length_bias 0x00000001
+#define GEN75_MI_RS_CONTROL_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 6
+
+#define GEN75_MI_RS_CONTROL_length 0x00000001
+
+struct GEN75_MI_RS_CONTROL {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define RS_STOP 0
+#define RS_START 1
+ uint32_t ResourceStreamerControl;
+};
+
+static inline void
+GEN75_MI_RS_CONTROL_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_RS_CONTROL * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->ResourceStreamerControl, 0, 0) |
+ 0;
+
+}
+
+#define GEN75_MI_RS_STORE_DATA_IMM_length_bias 0x00000002
+#define GEN75_MI_RS_STORE_DATA_IMM_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 43, \
+ .DwordLength = 2
+
+#define GEN75_MI_RS_STORE_DATA_IMM_length 0x00000004
+
+struct GEN75_MI_RS_STORE_DATA_IMM {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t DwordLength;
+ __gen_address_type DestinationAddress;
+ uint32_t CoreModeEnable;
+ uint32_t DataDWord0;
+};
+
+static inline void
+GEN75_MI_RS_STORE_DATA_IMM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_RS_STORE_DATA_IMM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ 0;
+
+ uint32_t dw2 =
+ __gen_field(values->CoreModeEnable, 0, 0) |
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->DestinationAddress, dw2);
+
+ dw[3] =
+ __gen_field(values->DataDWord0, 0, 31) |
+ 0;
+
+}
+
+#define GEN75_MI_SEMAPHORE_MBOX_length_bias 0x00000002
+#define GEN75_MI_SEMAPHORE_MBOX_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 22, \
+ .DwordLength = 1
+
+#define GEN75_MI_SEMAPHORE_MBOX_length 0x00000003
+
+struct GEN75_MI_SEMAPHORE_MBOX {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define RVSYNC 0
+#define RVESYNC 1
+#define RBSYNC 2
+#define UseGeneralRegisterSelect 3
+ uint32_t RegisterSelect;
+ uint32_t GeneralRegisterSelect;
+ uint32_t DwordLength;
+ uint32_t SemaphoreDataDword;
+};
+
+static inline void
+GEN75_MI_SEMAPHORE_MBOX_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_SEMAPHORE_MBOX * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->RegisterSelect, 16, 17) |
+ __gen_field(values->GeneralRegisterSelect, 8, 13) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->SemaphoreDataDword, 0, 31) |
+ 0;
+
+ dw[2] =
+ 0;
+
+}
+
+#define GEN75_MI_SET_CONTEXT_length_bias 0x00000002
+#define GEN75_MI_SET_CONTEXT_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 24, \
+ .DwordLength = 0
+
+#define GEN75_MI_SET_CONTEXT_length 0x00000002
+
+struct GEN75_MI_SET_CONTEXT {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t DwordLength;
+ __gen_address_type LogicalContextAddress;
+ uint32_t ReservedMustbe1;
+ bool CoreModeEnable;
+ bool ResourceStreamerStateSaveEnable;
+ bool ResourceStreamerStateRestoreEnable;
+ uint32_t ForceRestore;
+ uint32_t RestoreInhibit;
+};
+
+static inline void
+GEN75_MI_SET_CONTEXT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_SET_CONTEXT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw1 =
+ __gen_field(values->ReservedMustbe1, 8, 8) |
+ __gen_field(values->CoreModeEnable, 4, 4) |
+ __gen_field(values->ResourceStreamerStateSaveEnable, 3, 3) |
+ __gen_field(values->ResourceStreamerStateRestoreEnable, 2, 2) |
+ __gen_field(values->ForceRestore, 1, 1) |
+ __gen_field(values->RestoreInhibit, 0, 0) |
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->LogicalContextAddress, dw1);
+
+}
+
+#define GEN75_MI_SET_PREDICATE_length_bias 0x00000001
+#define GEN75_MI_SET_PREDICATE_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 1, \
+ .PREDICATEENABLE = 6
+
+#define GEN75_MI_SET_PREDICATE_length 0x00000001
+
+struct GEN75_MI_SET_PREDICATE {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define PredicateAlways 0
+#define PredicateonClear 1
+#define PredicateonSet 2
+#define PredicateDisable 3
+ bool PREDICATEENABLE;
+};
+
+static inline void
+GEN75_MI_SET_PREDICATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_SET_PREDICATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->PREDICATEENABLE, 0, 1) |
+ 0;
+
+}
+
+#define GEN75_MI_STORE_DATA_IMM_length_bias 0x00000002
+#define GEN75_MI_STORE_DATA_IMM_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 32, \
+ .DwordLength = 2
+
+#define GEN75_MI_STORE_DATA_IMM_length 0x00000004
+
+struct GEN75_MI_STORE_DATA_IMM {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool UseGlobalGTT;
+ uint32_t DwordLength;
+ uint32_t Address;
+ uint32_t CoreModeEnable;
+ uint32_t DataDWord0;
+ uint32_t DataDWord1;
+};
+
+static inline void
+GEN75_MI_STORE_DATA_IMM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_STORE_DATA_IMM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->UseGlobalGTT, 22, 22) |
+ __gen_field(values->DwordLength, 0, 5) |
+ 0;
+
+ dw[1] =
+ 0;
+
+ dw[2] =
+ __gen_field(values->Address, 2, 31) |
+ __gen_field(values->CoreModeEnable, 0, 0) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->DataDWord0, 0, 31) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->DataDWord1, 0, 31) |
+ 0;
+
+}
+
+#define GEN75_MI_STORE_DATA_INDEX_length_bias 0x00000002
+#define GEN75_MI_STORE_DATA_INDEX_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 33, \
+ .DwordLength = 1
+
+#define GEN75_MI_STORE_DATA_INDEX_length 0x00000003
+
+struct GEN75_MI_STORE_DATA_INDEX {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t DwordLength;
+ uint32_t Offset;
+ uint32_t DataDWord0;
+ uint32_t DataDWord1;
+};
+
+static inline void
+GEN75_MI_STORE_DATA_INDEX_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_STORE_DATA_INDEX * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->Offset, 2, 11) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->DataDWord0, 0, 31) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->DataDWord1, 0, 31) |
+ 0;
+
+}
+
+#define GEN75_MI_STORE_URB_MEM_length_bias 0x00000002
+#define GEN75_MI_STORE_URB_MEM_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 45, \
+ .DwordLength = 1
+
+#define GEN75_MI_STORE_URB_MEM_length 0x00000003
+
+struct GEN75_MI_STORE_URB_MEM {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t DwordLength;
+ uint32_t URBAddress;
+ __gen_address_type MemoryAddress;
+};
+
+static inline void
+GEN75_MI_STORE_URB_MEM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_STORE_URB_MEM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->URBAddress, 2, 14) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->MemoryAddress, dw2);
+
+}
+
+#define GEN75_MI_SUSPEND_FLUSH_length_bias 0x00000001
+#define GEN75_MI_SUSPEND_FLUSH_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 11
+
+#define GEN75_MI_SUSPEND_FLUSH_length 0x00000001
+
+struct GEN75_MI_SUSPEND_FLUSH {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool SuspendFlush;
+};
+
+static inline void
+GEN75_MI_SUSPEND_FLUSH_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_SUSPEND_FLUSH * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->SuspendFlush, 0, 0) |
+ 0;
+
+}
+
+#define GEN75_MI_TOPOLOGY_FILTER_length_bias 0x00000001
+#define GEN75_MI_TOPOLOGY_FILTER_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 13
+
+#define GEN75_MI_TOPOLOGY_FILTER_length 0x00000001
+
+struct GEN75_MI_TOPOLOGY_FILTER {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t TopologyFilterValue;
+};
+
+static inline void
+GEN75_MI_TOPOLOGY_FILTER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_TOPOLOGY_FILTER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->TopologyFilterValue, 0, 5) |
+ 0;
+
+}
+
+#define GEN75_MI_UPDATE_GTT_length_bias 0x00000002
+#define GEN75_MI_UPDATE_GTT_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 35
+
+#define GEN75_MI_UPDATE_GTT_length 0x00000000
+
+struct GEN75_MI_UPDATE_GTT {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define PerProcessGraphicsAddress 0
+#define GlobalGraphicsAddress 1
+ uint32_t UseGlobalGTT;
+ uint32_t DwordLength;
+ __gen_address_type EntryAddress;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN75_MI_UPDATE_GTT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_UPDATE_GTT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->UseGlobalGTT, 22, 22) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw1 =
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->EntryAddress, dw1);
+
+ /* variable length fields follow */
+}
+
+#define GEN75_MI_URB_ATOMIC_ALLOC_length_bias 0x00000001
+#define GEN75_MI_URB_ATOMIC_ALLOC_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 9
+
+#define GEN75_MI_URB_ATOMIC_ALLOC_length 0x00000001
+
+struct GEN75_MI_URB_ATOMIC_ALLOC {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t URBAtomicStorageOffset;
+ uint32_t URBAtomicStorageSize;
+};
+
+static inline void
+GEN75_MI_URB_ATOMIC_ALLOC_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_URB_ATOMIC_ALLOC * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->URBAtomicStorageOffset, 12, 19) |
+ __gen_field(values->URBAtomicStorageSize, 0, 8) |
+ 0;
+
+}
+
+#define GEN75_MI_URB_CLEAR_length_bias 0x00000002
+#define GEN75_MI_URB_CLEAR_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 25, \
+ .DwordLength = 0
+
+#define GEN75_MI_URB_CLEAR_length 0x00000002
+
+struct GEN75_MI_URB_CLEAR {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t DwordLength;
+ uint32_t URBClearLength;
+ uint32_t URBAddress;
+};
+
+static inline void
+GEN75_MI_URB_CLEAR_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_URB_CLEAR * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->URBClearLength, 16, 29) |
+ __gen_offset(values->URBAddress, 0, 14) |
+ 0;
+
+}
+
+#define GEN75_MI_USER_INTERRUPT_length_bias 0x00000001
+#define GEN75_MI_USER_INTERRUPT_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 2
+
+#define GEN75_MI_USER_INTERRUPT_length 0x00000001
+
+struct GEN75_MI_USER_INTERRUPT {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+};
+
+static inline void
+GEN75_MI_USER_INTERRUPT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_USER_INTERRUPT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ 0;
+
+}
+
+#define GEN75_MI_WAIT_FOR_EVENT_length_bias 0x00000001
+#define GEN75_MI_WAIT_FOR_EVENT_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 3
+
+#define GEN75_MI_WAIT_FOR_EVENT_length 0x00000001
+
+struct GEN75_MI_WAIT_FOR_EVENT {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool DisplayPipeCHorizontalBlankWaitEnable;
+ bool DisplayPipeCVerticalBlankWaitEnable;
+ bool DisplaySpriteCFlipPendingWaitEnable;
+#define Notenabled 0
+ uint32_t ConditionCodeWaitSelect;
+ bool DisplayPlaneCFlipPendingWaitEnable;
+ bool DisplayPipeCScanLineWaitEnable;
+ bool DisplayPipeBHorizontalBlankWaitEnable;
+ bool DisplayPipeBVerticalBlankWaitEnable;
+ bool DisplaySpriteBFlipPendingWaitEnable;
+ bool DisplayPlaneBFlipPendingWaitEnable;
+ bool DisplayPipeBScanLineWaitEnable;
+ bool DisplayPipeAHorizontalBlankWaitEnable;
+ bool DisplayPipeAVerticalBlankWaitEnable;
+ bool DisplaySpriteAFlipPendingWaitEnable;
+ bool DisplayPlaneAFlipPendingWaitEnable;
+ bool DisplayPipeAScanLineWaitEnable;
+};
+
+static inline void
+GEN75_MI_WAIT_FOR_EVENT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_MI_WAIT_FOR_EVENT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DisplayPipeCHorizontalBlankWaitEnable, 22, 22) |
+ __gen_field(values->DisplayPipeCVerticalBlankWaitEnable, 21, 21) |
+ __gen_field(values->DisplaySpriteCFlipPendingWaitEnable, 20, 20) |
+ __gen_field(values->ConditionCodeWaitSelect, 16, 19) |
+ __gen_field(values->DisplayPlaneCFlipPendingWaitEnable, 15, 15) |
+ __gen_field(values->DisplayPipeCScanLineWaitEnable, 14, 14) |
+ __gen_field(values->DisplayPipeBHorizontalBlankWaitEnable, 13, 13) |
+ __gen_field(values->DisplayPipeBVerticalBlankWaitEnable, 11, 11) |
+ __gen_field(values->DisplaySpriteBFlipPendingWaitEnable, 10, 10) |
+ __gen_field(values->DisplayPlaneBFlipPendingWaitEnable, 9, 9) |
+ __gen_field(values->DisplayPipeBScanLineWaitEnable, 8, 8) |
+ __gen_field(values->DisplayPipeAHorizontalBlankWaitEnable, 5, 5) |
+ __gen_field(values->DisplayPipeAVerticalBlankWaitEnable, 3, 3) |
+ __gen_field(values->DisplaySpriteAFlipPendingWaitEnable, 2, 2) |
+ __gen_field(values->DisplayPlaneAFlipPendingWaitEnable, 1, 1) |
+ __gen_field(values->DisplayPipeAScanLineWaitEnable, 0, 0) |
+ 0;
+
+}
+
+#define GEN75_PIPE_CONTROL_length_bias 0x00000002
+#define GEN75_PIPE_CONTROL_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 2, \
+ ._3DCommandSubOpcode = 0, \
+ .DwordLength = 3
+
+#define GEN75_PIPE_CONTROL_length 0x00000005
+
+struct GEN75_PIPE_CONTROL {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define DAT_PPGTT 0
+#define DAT_GGTT 1
+ uint32_t DestinationAddressType;
+#define NoLRIOperation 0
+#define MMIOWriteImmediateData 1
+ uint32_t LRIPostSyncOperation;
+ uint32_t StoreDataIndex;
+ uint32_t CommandStreamerStallEnable;
+#define DontReset 0
+#define Reset 1
+ uint32_t GlobalSnapshotCountReset;
+ uint32_t TLBInvalidate;
+ bool GenericMediaStateClear;
+#define NoWrite 0
+#define WriteImmediateData 1
+#define WritePSDepthCount 2
+#define WriteTimestamp 3
+ uint32_t PostSyncOperation;
+ bool DepthStallEnable;
+#define DisableFlush 0
+#define EnableFlush 1
+ bool RenderTargetCacheFlushEnable;
+ bool InstructionCacheInvalidateEnable;
+ bool TextureCacheInvalidationEnable;
+ bool IndirectStatePointersDisable;
+ bool NotifyEnable;
+ bool PipeControlFlushEnable;
+ bool DCFlushEnable;
+ bool VFCacheInvalidationEnable;
+ bool ConstantCacheInvalidationEnable;
+ bool StateCacheInvalidationEnable;
+ bool StallAtPixelScoreboard;
+#define FlushDisabled 0
+#define FlushEnabled 1
+ bool DepthCacheFlushEnable;
+ __gen_address_type Address;
+ uint32_t ImmediateData;
+ uint32_t ImmediateData0;
+};
+
+static inline void
+GEN75_PIPE_CONTROL_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_PIPE_CONTROL * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->DestinationAddressType, 24, 24) |
+ __gen_field(values->LRIPostSyncOperation, 23, 23) |
+ __gen_field(values->StoreDataIndex, 21, 21) |
+ __gen_field(values->CommandStreamerStallEnable, 20, 20) |
+ __gen_field(values->GlobalSnapshotCountReset, 19, 19) |
+ __gen_field(values->TLBInvalidate, 18, 18) |
+ __gen_field(values->GenericMediaStateClear, 16, 16) |
+ __gen_field(values->PostSyncOperation, 14, 15) |
+ __gen_field(values->DepthStallEnable, 13, 13) |
+ __gen_field(values->RenderTargetCacheFlushEnable, 12, 12) |
+ __gen_field(values->InstructionCacheInvalidateEnable, 11, 11) |
+ __gen_field(values->TextureCacheInvalidationEnable, 10, 10) |
+ __gen_field(values->IndirectStatePointersDisable, 9, 9) |
+ __gen_field(values->NotifyEnable, 8, 8) |
+ __gen_field(values->PipeControlFlushEnable, 7, 7) |
+ __gen_field(values->DCFlushEnable, 5, 5) |
+ __gen_field(values->VFCacheInvalidationEnable, 4, 4) |
+ __gen_field(values->ConstantCacheInvalidationEnable, 3, 3) |
+ __gen_field(values->StateCacheInvalidationEnable, 2, 2) |
+ __gen_field(values->StallAtPixelScoreboard, 1, 1) |
+ __gen_field(values->DepthCacheFlushEnable, 0, 0) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->Address, dw2);
+
+ dw[3] =
+ __gen_field(values->ImmediateData, 0, 31) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->ImmediateData, 0, 31) |
+ 0;
+
+}
+
+#define GEN75_SCISSOR_RECT_length 0x00000002
+
+struct GEN75_SCISSOR_RECT {
+ uint32_t ScissorRectangleYMin;
+ uint32_t ScissorRectangleXMin;
+ uint32_t ScissorRectangleYMax;
+ uint32_t ScissorRectangleXMax;
+};
+
+static inline void
+GEN75_SCISSOR_RECT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_SCISSOR_RECT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->ScissorRectangleYMin, 16, 31) |
+ __gen_field(values->ScissorRectangleXMin, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ScissorRectangleYMax, 16, 31) |
+ __gen_field(values->ScissorRectangleXMax, 0, 15) |
+ 0;
+
+}
+
+#define GEN75_SF_CLIP_VIEWPORT_length 0x00000010
+
+struct GEN75_SF_CLIP_VIEWPORT {
+ float ViewportMatrixElementm00;
+ float ViewportMatrixElementm11;
+ float ViewportMatrixElementm22;
+ float ViewportMatrixElementm30;
+ float ViewportMatrixElementm31;
+ float ViewportMatrixElementm32;
+ float XMinClipGuardband;
+ float XMaxClipGuardband;
+ float YMinClipGuardband;
+ float YMaxClipGuardband;
+};
+
+static inline void
+GEN75_SF_CLIP_VIEWPORT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_SF_CLIP_VIEWPORT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_float(values->ViewportMatrixElementm00) |
+ 0;
+
+ dw[1] =
+ __gen_float(values->ViewportMatrixElementm11) |
+ 0;
+
+ dw[2] =
+ __gen_float(values->ViewportMatrixElementm22) |
+ 0;
+
+ dw[3] =
+ __gen_float(values->ViewportMatrixElementm30) |
+ 0;
+
+ dw[4] =
+ __gen_float(values->ViewportMatrixElementm31) |
+ 0;
+
+ dw[5] =
+ __gen_float(values->ViewportMatrixElementm32) |
+ 0;
+
+ dw[6] =
+ 0;
+
+ dw[7] =
+ 0;
+
+ dw[8] =
+ __gen_float(values->XMinClipGuardband) |
+ 0;
+
+ dw[9] =
+ __gen_float(values->XMaxClipGuardband) |
+ 0;
+
+ dw[10] =
+ __gen_float(values->YMinClipGuardband) |
+ 0;
+
+ dw[11] =
+ __gen_float(values->YMaxClipGuardband) |
+ 0;
+
+ for (uint32_t i = 0, j = 12; i < 4; i += 1, j++) {
+ dw[j] =
+ 0;
+ }
+
+}
+
+#define GEN75_BLEND_STATE_length 0x00000002
+
+struct GEN75_BLEND_STATE {
+ bool ColorBufferBlendEnable;
+ bool IndependentAlphaBlendEnable;
+#define BLENDFUNCTION_ADD 0
+#define BLENDFUNCTION_SUBTRACT 1
+#define BLENDFUNCTION_REVERSE_SUBTRACT 2
+#define BLENDFUNCTION_MIN 3
+#define BLENDFUNCTION_MAX 4
+ uint32_t AlphaBlendFunction;
+#define BLENDFACTOR_ONE 1
+#define BLENDFACTOR_SRC_COLOR 2
+#define BLENDFACTOR_SRC_ALPHA 3
+#define BLENDFACTOR_DST_ALPHA 4
+#define BLENDFACTOR_DST_COLOR 5
+#define BLENDFACTOR_SRC_ALPHA_SATURATE 6
+#define BLENDFACTOR_CONST_COLOR 7
+#define BLENDFACTOR_CONST_ALPHA 8
+#define BLENDFACTOR_SRC1_COLOR 9
+#define BLENDFACTOR_SRC1_ALPHA 10
+#define BLENDFACTOR_ZERO 17
+#define BLENDFACTOR_INV_SRC_COLOR 18
+#define BLENDFACTOR_INV_SRC_ALPHA 19
+#define BLENDFACTOR_INV_DST_ALPHA 20
+#define BLENDFACTOR_INV_DST_COLOR 21
+#define BLENDFACTOR_INV_CONST_COLOR 23
+#define BLENDFACTOR_INV_CONST_ALPHA 24
+#define BLENDFACTOR_INV_SRC1_COLOR 25
+#define BLENDFACTOR_INV_SRC1_ALPHA 26
+ uint32_t SourceAlphaBlendFactor;
+ uint32_t DestinationAlphaBlendFactor;
+#define BLENDFUNCTION_ADD 0
+#define BLENDFUNCTION_SUBTRACT 1
+#define BLENDFUNCTION_REVERSE_SUBTRACT 2
+#define BLENDFUNCTION_MIN 3
+#define BLENDFUNCTION_MAX 4
+ uint32_t ColorBlendFunction;
+ uint32_t SourceBlendFactor;
+ uint32_t DestinationBlendFactor;
+ bool AlphaToCoverageEnable;
+ bool AlphaToOneEnable;
+ bool AlphaToCoverageDitherEnable;
+ bool WriteDisableAlpha;
+ bool WriteDisableRed;
+ bool WriteDisableGreen;
+ bool WriteDisableBlue;
+ bool LogicOpEnable;
+#define LOGICOP_CLEAR 0
+#define LOGICOP_NOR 1
+#define LOGICOP_AND_INVERTED 2
+#define LOGICOP_COPY_INVERTED 3
+#define LOGICOP_AND_REVERSE 4
+#define LOGICOP_INVERT 5
+#define LOGICOP_XOR 6
+#define LOGICOP_NAND 7
+#define LOGICOP_AND 8
+#define LOGICOP_EQUIV 9
+#define LOGICOP_NOOP 10
+#define LOGICOP_OR_INVERTED 11
+#define LOGICOP_COPY 12
+#define LOGICOP_OR_REVERSE 13
+#define LOGICOP_OR 14
+#define LOGICOP_SET 15
+ uint32_t LogicOpFunction;
+ bool AlphaTestEnable;
+#define COMPAREFUNCTION_ALWAYS 0
+#define COMPAREFUNCTION_NEVER 1
+#define COMPAREFUNCTION_LESS 2
+#define COMPAREFUNCTION_EQUAL 3
+#define COMPAREFUNCTION_LEQUAL 4
+#define COMPAREFUNCTION_GREATER 5
+#define COMPAREFUNCTION_NOTEQUAL 6
+#define COMPAREFUNCTION_GEQUAL 7
+ uint32_t AlphaTestFunction;
+ bool ColorDitherEnable;
+ uint32_t XDitherOffset;
+ uint32_t YDitherOffset;
+#define COLORCLAMP_UNORM 0
+#define COLORCLAMP_SNORM 1
+#define COLORCLAMP_RTFORMAT 2
+ uint32_t ColorClampRange;
+ bool PreBlendColorClampEnable;
+ bool PostBlendColorClampEnable;
+};
+
+static inline void
+GEN75_BLEND_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_BLEND_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->ColorBufferBlendEnable, 31, 31) |
+ __gen_field(values->IndependentAlphaBlendEnable, 30, 30) |
+ __gen_field(values->AlphaBlendFunction, 26, 28) |
+ __gen_field(values->SourceAlphaBlendFactor, 20, 24) |
+ __gen_field(values->DestinationAlphaBlendFactor, 15, 19) |
+ __gen_field(values->ColorBlendFunction, 11, 13) |
+ __gen_field(values->SourceBlendFactor, 5, 9) |
+ __gen_field(values->DestinationBlendFactor, 0, 4) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->AlphaToCoverageEnable, 31, 31) |
+ __gen_field(values->AlphaToOneEnable, 30, 30) |
+ __gen_field(values->AlphaToCoverageDitherEnable, 29, 29) |
+ __gen_field(values->WriteDisableAlpha, 27, 27) |
+ __gen_field(values->WriteDisableRed, 26, 26) |
+ __gen_field(values->WriteDisableGreen, 25, 25) |
+ __gen_field(values->WriteDisableBlue, 24, 24) |
+ __gen_field(values->LogicOpEnable, 22, 22) |
+ __gen_field(values->LogicOpFunction, 18, 21) |
+ __gen_field(values->AlphaTestEnable, 16, 16) |
+ __gen_field(values->AlphaTestFunction, 13, 15) |
+ __gen_field(values->ColorDitherEnable, 12, 12) |
+ __gen_field(values->XDitherOffset, 10, 11) |
+ __gen_field(values->YDitherOffset, 8, 9) |
+ __gen_field(values->ColorClampRange, 2, 3) |
+ __gen_field(values->PreBlendColorClampEnable, 1, 1) |
+ __gen_field(values->PostBlendColorClampEnable, 0, 0) |
+ 0;
+
+}
+
+#define GEN75_CC_VIEWPORT_length 0x00000002
+
+struct GEN75_CC_VIEWPORT {
+ float MinimumDepth;
+ float MaximumDepth;
+};
+
+static inline void
+GEN75_CC_VIEWPORT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_CC_VIEWPORT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_float(values->MinimumDepth) |
+ 0;
+
+ dw[1] =
+ __gen_float(values->MaximumDepth) |
+ 0;
+
+}
+
+#define GEN75_COLOR_CALC_STATE_length 0x00000006
+
+struct GEN75_COLOR_CALC_STATE {
+ uint32_t StencilReferenceValue;
+ uint32_t BackFaceStencilReferenceValue;
+#define Cancelled 0
+#define NotCancelled 1
+ uint32_t RoundDisableFunctionDisable;
+#define ALPHATEST_UNORM8 0
+#define ALPHATEST_FLOAT32 1
+ uint32_t AlphaTestFormat;
+ uint32_t AlphaReferenceValueAsUNORM8;
+ float AlphaReferenceValueAsFLOAT32;
+ float BlendConstantColorRed;
+ float BlendConstantColorGreen;
+ float BlendConstantColorBlue;
+ float BlendConstantColorAlpha;
+};
+
+static inline void
+GEN75_COLOR_CALC_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_COLOR_CALC_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->StencilReferenceValue, 24, 31) |
+ __gen_field(values->BackFaceStencilReferenceValue, 16, 23) |
+ __gen_field(values->RoundDisableFunctionDisable, 15, 15) |
+ __gen_field(values->AlphaTestFormat, 0, 0) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->AlphaReferenceValueAsUNORM8, 0, 31) |
+ __gen_float(values->AlphaReferenceValueAsFLOAT32) |
+ 0;
+
+ dw[2] =
+ __gen_float(values->BlendConstantColorRed) |
+ 0;
+
+ dw[3] =
+ __gen_float(values->BlendConstantColorGreen) |
+ 0;
+
+ dw[4] =
+ __gen_float(values->BlendConstantColorBlue) |
+ 0;
+
+ dw[5] =
+ __gen_float(values->BlendConstantColorAlpha) |
+ 0;
+
+}
+
+#define GEN75_DEPTH_STENCIL_STATE_length 0x00000003
+
+struct GEN75_DEPTH_STENCIL_STATE {
+ bool StencilTestEnable;
+#define COMPAREFUNCTION_ALWAYS 0
+#define COMPAREFUNCTION_NEVER 1
+#define COMPAREFUNCTION_LESS 2
+#define COMPAREFUNCTION_EQUAL 3
+#define COMPAREFUNCTION_LEQUAL 4
+#define COMPAREFUNCTION_GREATER 5
+#define COMPAREFUNCTION_NOTEQUAL 6
+#define COMPAREFUNCTION_GEQUAL 7
+ uint32_t StencilTestFunction;
+#define STENCILOP_KEEP 0
+#define STENCILOP_ZERO 1
+#define STENCILOP_REPLACE 2
+#define STENCILOP_INCRSAT 3
+#define STENCILOP_DECRSAT 4
+#define STENCILOP_INCR 5
+#define STENCILOP_DECR 6
+#define STENCILOP_INVERT 7
+ uint32_t StencilFailOp;
+ uint32_t StencilPassDepthFailOp;
+ uint32_t StencilPassDepthPassOp;
+ bool StencilBufferWriteEnable;
+ bool DoubleSidedStencilEnable;
+#define COMPAREFUNCTION_ALWAYS 0
+#define COMPAREFUNCTION_NEVER 1
+#define COMPAREFUNCTION_LESS 2
+#define COMPAREFUNCTION_EQUAL 3
+#define COMPAREFUNCTION_LEQUAL 4
+#define COMPAREFUNCTION_GREATER 5
+#define COMPAREFUNCTION_NOTEQUAL 6
+#define COMPAREFUNCTION_GEQUAL 7
+ uint32_t BackFaceStencilTestFunction;
+#define STENCILOP_KEEP 0
+#define STENCILOP_ZERO 1
+#define STENCILOP_REPLACE 2
+#define STENCILOP_INCRSAT 3
+#define STENCILOP_DECRSAT 4
+#define STENCILOP_INCR 5
+#define STENCILOP_DECR 6
+#define STENCILOP_INVERT 7
+ uint32_t BackfaceStencilFailOp;
+ uint32_t BackfaceStencilPassDepthFailOp;
+ uint32_t BackfaceStencilPassDepthPassOp;
+ uint32_t StencilTestMask;
+ uint32_t StencilWriteMask;
+ uint32_t BackfaceStencilTestMask;
+ uint32_t BackfaceStencilWriteMask;
+ bool DepthTestEnable;
+#define COMPAREFUNCTION_ALWAYS 0
+#define COMPAREFUNCTION_NEVER 1
+#define COMPAREFUNCTION_LESS 2
+#define COMPAREFUNCTION_EQUAL 3
+#define COMPAREFUNCTION_LEQUAL 4
+#define COMPAREFUNCTION_GREATER 5
+#define COMPAREFUNCTION_NOTEQUAL 6
+#define COMPAREFUNCTION_GEQUAL 7
+ uint32_t DepthTestFunction;
+ bool DepthBufferWriteEnable;
+};
+
+static inline void
+GEN75_DEPTH_STENCIL_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_DEPTH_STENCIL_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->StencilTestEnable, 31, 31) |
+ __gen_field(values->StencilTestFunction, 28, 30) |
+ __gen_field(values->StencilFailOp, 25, 27) |
+ __gen_field(values->StencilPassDepthFailOp, 22, 24) |
+ __gen_field(values->StencilPassDepthPassOp, 19, 21) |
+ __gen_field(values->StencilBufferWriteEnable, 18, 18) |
+ __gen_field(values->DoubleSidedStencilEnable, 15, 15) |
+ __gen_field(values->BackFaceStencilTestFunction, 12, 14) |
+ __gen_field(values->BackfaceStencilFailOp, 9, 11) |
+ __gen_field(values->BackfaceStencilPassDepthFailOp, 6, 8) |
+ __gen_field(values->BackfaceStencilPassDepthPassOp, 3, 5) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->StencilTestMask, 24, 31) |
+ __gen_field(values->StencilWriteMask, 16, 23) |
+ __gen_field(values->BackfaceStencilTestMask, 8, 15) |
+ __gen_field(values->BackfaceStencilWriteMask, 0, 7) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->DepthTestEnable, 31, 31) |
+ __gen_field(values->DepthTestFunction, 27, 29) |
+ __gen_field(values->DepthBufferWriteEnable, 26, 26) |
+ 0;
+
+}
+
+#define GEN75_INTERFACE_DESCRIPTOR_DATA_length 0x00000008
+
+struct GEN75_INTERFACE_DESCRIPTOR_DATA {
+ uint32_t KernelStartPointer;
+#define Multiple 0
+#define Single 1
+ uint32_t SingleProgramFlow;
+#define NormalPriority 0
+#define HighPriority 1
+ uint32_t ThreadPriority;
+#define IEEE754 0
+#define Alternate 1
+ uint32_t FloatingPointMode;
+ bool IllegalOpcodeExceptionEnable;
+ bool MaskStackExceptionEnable;
+ bool SoftwareExceptionEnable;
+ uint32_t SamplerStatePointer;
+#define Nosamplersused 0
+#define Between1and4samplersused 1
+#define Between5and8samplersused 2
+#define Between9and12samplersused 3
+#define Between13and16samplersused 4
+ uint32_t SamplerCount;
+ uint32_t BindingTablePointer;
+ uint32_t BindingTableEntryCount;
+ uint32_t ConstantURBEntryReadLength;
+#define RTNE 0
+#define RU 1
+#define RD 2
+#define RTZ 3
+ uint32_t RoundingMode;
+ bool BarrierEnable;
+ uint32_t SharedLocalMemorySize;
+ uint32_t NumberofThreadsinGPGPUThreadGroup;
+ uint32_t CrossThreadConstantDataReadLength;
+};
+
+static inline void
+GEN75_INTERFACE_DESCRIPTOR_DATA_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_INTERFACE_DESCRIPTOR_DATA * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_offset(values->KernelStartPointer, 6, 31) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->SingleProgramFlow, 18, 18) |
+ __gen_field(values->ThreadPriority, 17, 17) |
+ __gen_field(values->FloatingPointMode, 16, 16) |
+ __gen_field(values->IllegalOpcodeExceptionEnable, 13, 13) |
+ __gen_field(values->MaskStackExceptionEnable, 11, 11) |
+ __gen_field(values->SoftwareExceptionEnable, 7, 7) |
+ 0;
+
+ dw[2] =
+ __gen_offset(values->SamplerStatePointer, 5, 31) |
+ __gen_field(values->SamplerCount, 2, 4) |
+ 0;
+
+ dw[3] =
+ __gen_offset(values->BindingTablePointer, 5, 15) |
+ __gen_field(values->BindingTableEntryCount, 0, 4) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->ConstantURBEntryReadLength, 16, 31) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->RoundingMode, 22, 23) |
+ __gen_field(values->BarrierEnable, 21, 21) |
+ __gen_field(values->SharedLocalMemorySize, 16, 20) |
+ __gen_field(values->NumberofThreadsinGPGPUThreadGroup, 0, 7) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->CrossThreadConstantDataReadLength, 0, 7) |
+ 0;
+
+ dw[7] =
+ 0;
+
+}
+
+#define GEN75_BINDING_TABLE_STATE_length 0x00000001
+
+struct GEN75_BINDING_TABLE_STATE {
+ uint32_t SurfaceStatePointer;
+};
+
+static inline void
+GEN75_BINDING_TABLE_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_BINDING_TABLE_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_offset(values->SurfaceStatePointer, 5, 31) |
+ 0;
+
+}
+
+#define GEN75_RENDER_SURFACE_STATE_length 0x00000008
+
+struct GEN75_RENDER_SURFACE_STATE {
+#define SURFTYPE_1D 0
+#define SURFTYPE_2D 1
+#define SURFTYPE_3D 2
+#define SURFTYPE_CUBE 3
+#define SURFTYPE_BUFFER 4
+#define SURFTYPE_STRBUF 5
+#define SURFTYPE_NULL 7
+ uint32_t SurfaceType;
+ bool SurfaceArray;
+ uint32_t SurfaceFormat;
+#define VALIGN_2 0
+#define VALIGN_4 1
+ uint32_t SurfaceVerticalAlignment;
+#define HALIGN_4 0
+#define HALIGN_8 1
+ uint32_t SurfaceHorizontalAlignment;
+ uint32_t TiledSurface;
+#define TILEWALK_XMAJOR 0
+#define TILEWALK_YMAJOR 1
+ uint32_t TileWalk;
+ uint32_t VerticalLineStride;
+ uint32_t VerticalLineStrideOffset;
+#define ARYSPC_FULL 0
+#define ARYSPC_LOD0 1
+ uint32_t SurfaceArraySpacing;
+ uint32_t RenderCacheReadWriteMode;
+#define NORMAL_MODE 0
+#define PROGRESSIVE_FRAME 2
+#define INTERLACED_FRAME 3
+ uint32_t MediaBoundaryPixelMode;
+ uint32_t CubeFaceEnables;
+ __gen_address_type SurfaceBaseAddress;
+ uint32_t Height;
+ uint32_t Width;
+ uint32_t Depth;
+ uint32_t IntegerSurfaceFormat;
+ uint32_t SurfacePitch;
+#define RTROTATE_0DEG 0
+#define RTROTATE_90DEG 1
+#define RTROTATE_270DEG 3
+ uint32_t RenderTargetRotation;
+ uint32_t MinimumArrayElement;
+ uint32_t RenderTargetViewExtent;
+#define MSFMT_MSS 0
+#define MSFMT_DEPTH_STENCIL 1
+ uint32_t MultisampledSurfaceStorageFormat;
+#define MULTISAMPLECOUNT_1 0
+#define MULTISAMPLECOUNT_4 2
+#define MULTISAMPLECOUNT_8 3
+ uint32_t NumberofMultisamples;
+ uint32_t MultisamplePositionPaletteIndex;
+ uint32_t MinimumArrayElement0;
+ uint32_t XOffset;
+ uint32_t YOffset;
+ struct GEN75_MEMORY_OBJECT_CONTROL_STATE SurfaceObjectControlState;
+ uint32_t SurfaceMinLOD;
+ uint32_t MIPCountLOD;
+ __gen_address_type MCSBaseAddress;
+ uint32_t MCSSurfacePitch;
+ __gen_address_type AppendCounterAddress;
+ bool AppendCounterEnable;
+ bool MCSEnable;
+ uint32_t XOffsetforUVPlane;
+ uint32_t YOffsetforUVPlane;
+#define SCS_ZERO 0
+#define SCS_ONE 1
+#define SCS_RED 4
+#define SCS_GREEN 5
+#define SCS_BLUE 6
+#define SCS_ALPHA 7
+ uint32_t ShaderChannelSelectR;
+ uint32_t ShaderChannelSelectG;
+ uint32_t ShaderChannelSelectB;
+ uint32_t ShaderChannelSelectA;
+ float ResourceMinLOD;
+};
+
+static inline void
+GEN75_RENDER_SURFACE_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_RENDER_SURFACE_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->SurfaceType, 29, 31) |
+ __gen_field(values->SurfaceArray, 28, 28) |
+ __gen_field(values->SurfaceFormat, 18, 26) |
+ __gen_field(values->SurfaceVerticalAlignment, 16, 17) |
+ __gen_field(values->SurfaceHorizontalAlignment, 15, 15) |
+ __gen_field(values->TiledSurface, 14, 14) |
+ __gen_field(values->TileWalk, 13, 13) |
+ __gen_field(values->VerticalLineStride, 12, 12) |
+ __gen_field(values->VerticalLineStrideOffset, 11, 11) |
+ __gen_field(values->SurfaceArraySpacing, 10, 10) |
+ __gen_field(values->RenderCacheReadWriteMode, 8, 8) |
+ __gen_field(values->MediaBoundaryPixelMode, 6, 7) |
+ __gen_field(values->CubeFaceEnables, 0, 5) |
+ 0;
+
+ uint32_t dw1 =
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->SurfaceBaseAddress, dw1);
+
+ dw[2] =
+ __gen_field(values->Height, 16, 29) |
+ __gen_field(values->Width, 0, 13) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->Depth, 21, 31) |
+ __gen_field(values->IntegerSurfaceFormat, 18, 20) |
+ __gen_field(values->SurfacePitch, 0, 17) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->RenderTargetRotation, 29, 30) |
+ __gen_field(values->MinimumArrayElement, 18, 28) |
+ __gen_field(values->RenderTargetViewExtent, 7, 17) |
+ __gen_field(values->MultisampledSurfaceStorageFormat, 6, 6) |
+ __gen_field(values->NumberofMultisamples, 3, 5) |
+ __gen_field(values->MultisamplePositionPaletteIndex, 0, 2) |
+ __gen_field(values->MinimumArrayElement, 0, 26) |
+ 0;
+
+ uint32_t dw_SurfaceObjectControlState;
+ GEN75_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_SurfaceObjectControlState, &values->SurfaceObjectControlState);
+ dw[5] =
+ __gen_offset(values->XOffset, 25, 31) |
+ __gen_offset(values->YOffset, 20, 23) |
+ __gen_field(dw_SurfaceObjectControlState, 16, 19) |
+ __gen_field(values->SurfaceMinLOD, 4, 7) |
+ __gen_field(values->MIPCountLOD, 0, 3) |
+ 0;
+
+ uint32_t dw6 =
+ __gen_field(values->MCSSurfacePitch, 3, 11) |
+ __gen_field(values->AppendCounterEnable, 1, 1) |
+ __gen_field(values->MCSEnable, 0, 0) |
+ __gen_field(values->XOffsetforUVPlane, 16, 29) |
+ __gen_field(values->YOffsetforUVPlane, 0, 13) |
+ 0;
+
+ dw[6] =
+ __gen_combine_address(data, &dw[6], values->AppendCounterAddress, dw6);
+
+ dw[7] =
+ __gen_field(values->ShaderChannelSelectR, 25, 27) |
+ __gen_field(values->ShaderChannelSelectG, 22, 24) |
+ __gen_field(values->ShaderChannelSelectB, 19, 21) |
+ __gen_field(values->ShaderChannelSelectA, 16, 18) |
+ __gen_field(values->ResourceMinLOD * (1 << 8), 0, 11) |
+ 0;
+
+}
+
+#define GEN75_SAMPLER_BORDER_COLOR_STATE_length 0x00000014
+
+#define GEN75_BORDER_COLOR_UINT32_SINT32_length 0x00000004
+
+struct GEN75_BORDER_COLOR_UINT32_SINT32 {
+ uint32_t BorderColorRedui32integerunclamp;
+ uint32_t BorderColorRedsi32integerunclamp;
+ uint32_t BorderColorGreenui32integerunclamp;
+ uint32_t BorderColorGreensi32integerunclamp;
+ uint32_t BorderColorBlueui32integerunclamp;
+ uint32_t BorderColorBluesi32integerunclamp;
+ uint32_t BorderColorGreenui32integerunclamp0;
+ uint32_t BorderColorGreensi32integerunclamp0;
+ uint32_t BorderColorAlphaui32integerunclamp;
+ uint32_t BorderColorAlphasi32integerunclamp;
+};
+
+static inline void
+GEN75_BORDER_COLOR_UINT32_SINT32_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_BORDER_COLOR_UINT32_SINT32 * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->BorderColorRedui32integerunclamp, 0, 31) |
+ __gen_field(values->BorderColorRedsi32integerunclamp, 0, 31) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->BorderColorGreenui32integerunclamp, 0, 31) |
+ __gen_field(values->BorderColorGreensi32integerunclamp, 0, 31) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->BorderColorBlueui32integerunclamp, 0, 31) |
+ __gen_field(values->BorderColorBluesi32integerunclamp, 0, 31) |
+ __gen_field(values->BorderColorGreenui32integerunclamp, 0, 31) |
+ __gen_field(values->BorderColorGreensi32integerunclamp, 0, 31) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->BorderColorAlphaui32integerunclamp, 0, 31) |
+ __gen_field(values->BorderColorAlphasi32integerunclamp, 0, 31) |
+ 0;
+
+}
+
+#define GEN75_BORDER_COLOR_UINT16_SINT16_length 0x00000004
+
+struct GEN75_BORDER_COLOR_UINT16_SINT16 {
+ uint32_t BorderColorGreenclamptouint16;
+ uint32_t BorderColorGreenclamptosint16;
+ uint32_t BorderColorRedclamptouint16;
+ uint32_t BorderColorRedclamptosint16;
+ uint32_t BorderColorAlphaclamptouint16;
+ uint32_t BorderColorAlphaclamptosint16;
+ uint32_t BorderColorBlueclamptouint16;
+ uint32_t BorderColorBlueclamptosint16;
+};
+
+static inline void
+GEN75_BORDER_COLOR_UINT16_SINT16_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_BORDER_COLOR_UINT16_SINT16 * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->BorderColorGreenclamptouint16, 16, 31) |
+ __gen_field(values->BorderColorGreenclamptosint16, 16, 31) |
+ __gen_field(values->BorderColorRedclamptouint16, 0, 15) |
+ __gen_field(values->BorderColorRedclamptosint16, 0, 15) |
+ 0;
+
+ dw[1] =
+ 0;
+
+ dw[2] =
+ __gen_field(values->BorderColorAlphaclamptouint16, 16, 31) |
+ __gen_field(values->BorderColorAlphaclamptosint16, 16, 31) |
+ __gen_field(values->BorderColorBlueclamptouint16, 0, 15) |
+ __gen_field(values->BorderColorBlueclamptosint16, 0, 15) |
+ 0;
+
+ dw[3] =
+ 0;
+
+}
+
+#define GEN75_BORDER_COLOR_UINT8_SINT8_length 0x00000004
+
+struct GEN75_BORDER_COLOR_UINT8_SINT8 {
+ uint32_t BorderColorAlphaclamptouint8;
+ uint32_t BorderColorAlphaclamptosint8;
+ uint32_t BorderColorBlueclamptouint8;
+ uint32_t BorderColorBlueclamptosint8;
+ uint32_t BorderColorGreenclamptouint8;
+ uint32_t BorderColorGreenclamptosint8;
+ uint32_t BorderRedAlphaclamptouint8;
+ uint32_t BorderRedAlphaclamptosint8;
+};
+
+static inline void
+GEN75_BORDER_COLOR_UINT8_SINT8_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_BORDER_COLOR_UINT8_SINT8 * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->BorderColorAlphaclamptouint8, 24, 31) |
+ __gen_field(values->BorderColorAlphaclamptosint8, 24, 31) |
+ __gen_field(values->BorderColorBlueclamptouint8, 16, 23) |
+ __gen_field(values->BorderColorBlueclamptosint8, 16, 23) |
+ __gen_field(values->BorderColorGreenclamptouint8, 8, 15) |
+ __gen_field(values->BorderColorGreenclamptosint8, 8, 15) |
+ __gen_field(values->BorderRedAlphaclamptouint8, 0, 7) |
+ __gen_field(values->BorderRedAlphaclamptosint8, 0, 7) |
+ 0;
+
+ dw[1] =
+ 0;
+
+ dw[2] =
+ 0;
+
+ dw[3] =
+ 0;
+
+}
+
+struct GEN75_SAMPLER_BORDER_COLOR_STATE {
+ float BorderColorRedDX100GL;
+ uint32_t BorderColorAlpha;
+ uint32_t BorderColorBlue;
+ uint32_t BorderColorGreen;
+ uint32_t BorderColorRedDX9;
+ float BorderColorGreen0;
+ float BorderColorBlue0;
+ float BorderColorAlpha0;
+ struct GEN75_BORDER_COLOR_UINT32_SINT32 BorderColor;
+ struct GEN75_BORDER_COLOR_UINT16_SINT16 BorderColor0;
+ struct GEN75_BORDER_COLOR_UINT8_SINT8 BorderColor1;
+};
+
+static inline void
+GEN75_SAMPLER_BORDER_COLOR_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_SAMPLER_BORDER_COLOR_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_float(values->BorderColorRedDX100GL) |
+ __gen_field(values->BorderColorAlpha, 24, 31) |
+ __gen_field(values->BorderColorBlue, 16, 23) |
+ __gen_field(values->BorderColorGreen, 8, 15) |
+ __gen_field(values->BorderColorRedDX9, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_float(values->BorderColorGreen) |
+ 0;
+
+ dw[2] =
+ __gen_float(values->BorderColorBlue) |
+ 0;
+
+ dw[3] =
+ __gen_float(values->BorderColorAlpha) |
+ 0;
+
+ for (uint32_t i = 0, j = 4; i < 12; i += 1, j++) {
+ dw[j] =
+ 0;
+ }
+
+ GEN75_BORDER_COLOR_UINT32_SINT32_pack(data, &dw[16], &values->BorderColor);
+}
+
+#define GEN75_SAMPLER_STATE_length 0x00000004
+
+struct GEN75_SAMPLER_STATE {
+ bool SamplerDisable;
+#define DX10OGL 0
+#define DX9 1
+ uint32_t TextureBorderColorMode;
+#define OGL 1
+ uint32_t LODPreClampEnable;
+ float BaseMipLevel;
+#define MIPFILTER_NONE 0
+#define MIPFILTER_NEAREST 1
+#define MIPFILTER_LINEAR 3
+ uint32_t MipModeFilter;
+#define MAPFILTER_NEAREST 0
+#define MAPFILTER_LINEAR 1
+#define MAPFILTER_ANISOTROPIC 2
+#define MAPFILTER_MONO 6
+ uint32_t MagModeFilter;
+#define MAPFILTER_NEAREST 0
+#define MAPFILTER_LINEAR 1
+#define MAPFILTER_ANISOTROPIC 2
+#define MAPFILTER_MONO 6
+ uint32_t MinModeFilter;
+ float TextureLODBias;
+#define LEGACY 0
+#define EWAApproximation 1
+ uint32_t AnisotropicAlgorithm;
+ float MinLOD;
+ float MaxLOD;
+#define PREFILTEROPALWAYS 0
+#define PREFILTEROPNEVER 1
+#define PREFILTEROPLESS 2
+#define PREFILTEROPEQUAL 3
+#define PREFILTEROPLEQUAL 4
+#define PREFILTEROPGREATER 5
+#define PREFILTEROPNOTEQUAL 6
+#define PREFILTEROPGEQUAL 7
+ uint32_t ShadowFunction;
+#define PROGRAMMED 0
+#define OVERRIDE 1
+ uint32_t CubeSurfaceControlMode;
+ uint32_t BorderColorPointer;
+ bool ChromaKeyEnable;
+ uint32_t ChromaKeyIndex;
+#define KEYFILTER_KILL_ON_ANY_MATCH 0
+#define KEYFILTER_REPLACE_BLACK 1
+ uint32_t ChromaKeyMode;
+#define RATIO21 0
+#define RATIO41 1
+#define RATIO61 2
+#define RATIO81 3
+#define RATIO101 4
+#define RATIO121 5
+#define RATIO141 6
+#define RATIO161 7
+ uint32_t MaximumAnisotropy;
+ bool RAddressMinFilterRoundingEnable;
+ bool RAddressMagFilterRoundingEnable;
+ bool VAddressMinFilterRoundingEnable;
+ bool VAddressMagFilterRoundingEnable;
+ bool UAddressMinFilterRoundingEnable;
+ bool UAddressMagFilterRoundingEnable;
+#define FULL 0
+#define TRIQUAL_HIGHMAG_CLAMP_MIPFILTER 1
+#define MED 2
+#define LOW 3
+ uint32_t TrilinearFilterQuality;
+ bool NonnormalizedCoordinateEnable;
+ uint32_t TCXAddressControlMode;
+ uint32_t TCYAddressControlMode;
+ uint32_t TCZAddressControlMode;
+};
+
+static inline void
+GEN75_SAMPLER_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN75_SAMPLER_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->SamplerDisable, 31, 31) |
+ __gen_field(values->TextureBorderColorMode, 29, 29) |
+ __gen_field(values->LODPreClampEnable, 28, 28) |
+ __gen_field(values->BaseMipLevel * (1 << 1), 22, 26) |
+ __gen_field(values->MipModeFilter, 20, 21) |
+ __gen_field(values->MagModeFilter, 17, 19) |
+ __gen_field(values->MinModeFilter, 14, 16) |
+ __gen_fixed(values->TextureLODBias, 1, 13, true, 8) |
+ __gen_field(values->AnisotropicAlgorithm, 0, 0) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->MinLOD * (1 << 8), 20, 31) |
+ __gen_field(values->MaxLOD * (1 << 8), 8, 19) |
+ __gen_field(values->ShadowFunction, 1, 3) |
+ __gen_field(values->CubeSurfaceControlMode, 0, 0) |
+ 0;
+
+ dw[2] =
+ __gen_offset(values->BorderColorPointer, 5, 31) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->ChromaKeyEnable, 25, 25) |
+ __gen_field(values->ChromaKeyIndex, 23, 24) |
+ __gen_field(values->ChromaKeyMode, 22, 22) |
+ __gen_field(values->MaximumAnisotropy, 19, 21) |
+ __gen_field(values->RAddressMinFilterRoundingEnable, 13, 13) |
+ __gen_field(values->RAddressMagFilterRoundingEnable, 14, 14) |
+ __gen_field(values->VAddressMinFilterRoundingEnable, 15, 15) |
+ __gen_field(values->VAddressMagFilterRoundingEnable, 16, 16) |
+ __gen_field(values->UAddressMinFilterRoundingEnable, 17, 17) |
+ __gen_field(values->UAddressMagFilterRoundingEnable, 18, 18) |
+ __gen_field(values->TrilinearFilterQuality, 11, 12) |
+ __gen_field(values->NonnormalizedCoordinateEnable, 10, 10) |
+ __gen_field(values->TCXAddressControlMode, 6, 8) |
+ __gen_field(values->TCYAddressControlMode, 3, 5) |
+ __gen_field(values->TCZAddressControlMode, 0, 2) |
+ 0;
+
+}
+
+/* Enum 3D_Prim_Topo_Type */
+#define _3DPRIM_POINTLIST 1
+#define _3DPRIM_LINELIST 2
+#define _3DPRIM_LINESTRIP 3
+#define _3DPRIM_TRILIST 4
+#define _3DPRIM_TRISTRIP 5
+#define _3DPRIM_TRIFAN 6
+#define _3DPRIM_QUADLIST 7
+#define _3DPRIM_QUADSTRIP 8
+#define _3DPRIM_LINELIST_ADJ 9
+#define _3DPRIM_LINESTRIP_ADJ 10
+#define _3DPRIM_TRILIST_ADJ 11
+#define _3DPRIM_TRISTRIP_ADJ 12
+#define _3DPRIM_TRISTRIP_REVERSE 13
+#define _3DPRIM_POLYGON 14
+#define _3DPRIM_RECTLIST 15
+#define _3DPRIM_LINELOOP 16
+#define _3DPRIM_POINTLIST_BF 17
+#define _3DPRIM_LINESTRIP_CONT 18
+#define _3DPRIM_LINESTRIP_BF 19
+#define _3DPRIM_LINESTRIP_CONT_BF 20
+#define _3DPRIM_TRIFAN_NOSTIPPLE 22
+#define _3DPRIM_PATCHLIST_1 32
+#define _3DPRIM_PATCHLIST_2 33
+#define _3DPRIM_PATCHLIST_3 34
+#define _3DPRIM_PATCHLIST_4 35
+#define _3DPRIM_PATCHLIST_5 36
+#define _3DPRIM_PATCHLIST_6 37
+#define _3DPRIM_PATCHLIST_7 38
+#define _3DPRIM_PATCHLIST_8 39
+#define _3DPRIM_PATCHLIST_9 40
+#define _3DPRIM_PATCHLIST_10 41
+#define _3DPRIM_PATCHLIST_11 42
+#define _3DPRIM_PATCHLIST_12 43
+#define _3DPRIM_PATCHLIST_13 44
+#define _3DPRIM_PATCHLIST_14 45
+#define _3DPRIM_PATCHLIST_15 46
+#define _3DPRIM_PATCHLIST_16 47
+#define _3DPRIM_PATCHLIST_17 48
+#define _3DPRIM_PATCHLIST_18 49
+#define _3DPRIM_PATCHLIST_19 50
+#define _3DPRIM_PATCHLIST_20 51
+#define _3DPRIM_PATCHLIST_21 52
+#define _3DPRIM_PATCHLIST_22 53
+#define _3DPRIM_PATCHLIST_23 54
+#define _3DPRIM_PATCHLIST_24 55
+#define _3DPRIM_PATCHLIST_25 56
+#define _3DPRIM_PATCHLIST_26 57
+#define _3DPRIM_PATCHLIST_27 58
+#define _3DPRIM_PATCHLIST_28 59
+#define _3DPRIM_PATCHLIST_29 60
+#define _3DPRIM_PATCHLIST_30 61
+#define _3DPRIM_PATCHLIST_31 62
+#define _3DPRIM_PATCHLIST_32 63
+
+/* Enum 3D_Vertex_Component_Control */
+#define VFCOMP_NOSTORE 0
+#define VFCOMP_STORE_SRC 1
+#define VFCOMP_STORE_0 2
+#define VFCOMP_STORE_1_FP 3
+#define VFCOMP_STORE_1_INT 4
+#define VFCOMP_STORE_VID 5
+#define VFCOMP_STORE_IID 6
+#define VFCOMP_STORE_PID 7
+
+/* Enum 3D_Compare_Function */
+#define COMPAREFUNCTION_ALWAYS 0
+#define COMPAREFUNCTION_NEVER 1
+#define COMPAREFUNCTION_LESS 2
+#define COMPAREFUNCTION_EQUAL 3
+#define COMPAREFUNCTION_LEQUAL 4
+#define COMPAREFUNCTION_GREATER 5
+#define COMPAREFUNCTION_NOTEQUAL 6
+#define COMPAREFUNCTION_GEQUAL 7
+
+/* Enum SURFACE_FORMAT */
+#define R32G32B32A32_FLOAT 0
+#define R32G32B32A32_SINT 1
+#define R32G32B32A32_UINT 2
+#define R32G32B32A32_UNORM 3
+#define R32G32B32A32_SNORM 4
+#define R64G64_FLOAT 5
+#define R32G32B32X32_FLOAT 6
+#define R32G32B32A32_SSCALED 7
+#define R32G32B32A32_USCALED 8
+#define R32G32B32A32_SFIXED 32
+#define R64G64_PASSTHRU 33
+#define R32G32B32_FLOAT 64
+#define R32G32B32_SINT 65
+#define R32G32B32_UINT 66
+#define R32G32B32_UNORM 67
+#define R32G32B32_SNORM 68
+#define R32G32B32_SSCALED 69
+#define R32G32B32_USCALED 70
+#define R32G32B32_SFIXED 80
+#define R16G16B16A16_UNORM 128
+#define R16G16B16A16_SNORM 129
+#define R16G16B16A16_SINT 130
+#define R16G16B16A16_UINT 131
+#define R16G16B16A16_FLOAT 132
+#define R32G32_FLOAT 133
+#define R32G32_SINT 134
+#define R32G32_UINT 135
+#define R32_FLOAT_X8X24_TYPELESS 136
+#define X32_TYPELESS_G8X24_UINT 137
+#define L32A32_FLOAT 138
+#define R32G32_UNORM 139
+#define R32G32_SNORM 140
+#define R64_FLOAT 141
+#define R16G16B16X16_UNORM 142
+#define R16G16B16X16_FLOAT 143
+#define A32X32_FLOAT 144
+#define L32X32_FLOAT 145
+#define I32X32_FLOAT 146
+#define R16G16B16A16_SSCALED 147
+#define R16G16B16A16_USCALED 148
+#define R32G32_SSCALED 149
+#define R32G32_USCALED 150
+#define R32G32_SFIXED 160
+#define R64_PASSTHRU 161
+#define B8G8R8A8_UNORM 192
+#define B8G8R8A8_UNORM_SRGB 193
+#define R10G10B10A2_UNORM 194
+#define R10G10B10A2_UNORM_SRGB 195
+#define R10G10B10A2_UINT 196
+#define R10G10B10_SNORM_A2_UNORM 197
+#define R8G8B8A8_UNORM 199
+#define R8G8B8A8_UNORM_SRGB 200
+#define R8G8B8A8_SNORM 201
+#define R8G8B8A8_SINT 202
+#define R8G8B8A8_UINT 203
+#define R16G16_UNORM 204
+#define R16G16_SNORM 205
+#define R16G16_SINT 206
+#define R16G16_UINT 207
+#define R16G16_FLOAT 208
+#define B10G10R10A2_UNORM 209
+#define B10G10R10A2_UNORM_SRGB 210
+#define R11G11B10_FLOAT 211
+#define R32_SINT 214
+#define R32_UINT 215
+#define R32_FLOAT 216
+#define R24_UNORM_X8_TYPELESS 217
+#define X24_TYPELESS_G8_UINT 218
+#define L32_UNORM 221
+#define A32_UNORM 222
+#define L16A16_UNORM 223
+#define I24X8_UNORM 224
+#define L24X8_UNORM 225
+#define A24X8_UNORM 226
+#define I32_FLOAT 227
+#define L32_FLOAT 228
+#define A32_FLOAT 229
+#define X8B8_UNORM_G8R8_SNORM 230
+#define A8X8_UNORM_G8R8_SNORM 231
+#define B8X8_UNORM_G8R8_SNORM 232
+#define B8G8R8X8_UNORM 233
+#define B8G8R8X8_UNORM_SRGB 234
+#define R8G8B8X8_UNORM 235
+#define R8G8B8X8_UNORM_SRGB 236
+#define R9G9B9E5_SHAREDEXP 237
+#define B10G10R10X2_UNORM 238
+#define L16A16_FLOAT 240
+#define R32_UNORM 241
+#define R32_SNORM 242
+#define R10G10B10X2_USCALED 243
+#define R8G8B8A8_SSCALED 244
+#define R8G8B8A8_USCALED 245
+#define R16G16_SSCALED 246
+#define R16G16_USCALED 247
+#define R32_SSCALED 248
+#define R32_USCALED 249
+#define B5G6R5_UNORM 256
+#define B5G6R5_UNORM_SRGB 257
+#define B5G5R5A1_UNORM 258
+#define B5G5R5A1_UNORM_SRGB 259
+#define B4G4R4A4_UNORM 260
+#define B4G4R4A4_UNORM_SRGB 261
+#define R8G8_UNORM 262
+#define R8G8_SNORM 263
+#define R8G8_SINT 264
+#define R8G8_UINT 265
+#define R16_UNORM 266
+#define R16_SNORM 267
+#define R16_SINT 268
+#define R16_UINT 269
+#define R16_FLOAT 270
+#define A8P8_UNORM_PALETTE0 271
+#define A8P8_UNORM_PALETTE1 272
+#define I16_UNORM 273
+#define L16_UNORM 274
+#define A16_UNORM 275
+#define L8A8_UNORM 276
+#define I16_FLOAT 277
+#define L16_FLOAT 278
+#define A16_FLOAT 279
+#define L8A8_UNORM_SRGB 280
+#define R5G5_SNORM_B6_UNORM 281
+#define B5G5R5X1_UNORM 282
+#define B5G5R5X1_UNORM_SRGB 283
+#define R8G8_SSCALED 284
+#define R8G8_USCALED 285
+#define R16_SSCALED 286
+#define R16_USCALED 287
+#define P8A8_UNORM_PALETTE0 290
+#define P8A8_UNORM_PALETTE1 291
+#define A1B5G5R5_UNORM 292
+#define A4B4G4R4_UNORM 293
+#define L8A8_UINT 294
+#define L8A8_SINT 295
+#define R8_UNORM 320
+#define R8_SNORM 321
+#define R8_SINT 322
+#define R8_UINT 323
+#define A8_UNORM 324
+#define I8_UNORM 325
+#define L8_UNORM 326
+#define P4A4_UNORM_PALETTE0 327
+#define A4P4_UNORM_PALETTE0 328
+#define R8_SSCALED 329
+#define R8_USCALED 330
+#define P8_UNORM_PALETTE0 331
+#define L8_UNORM_SRGB 332
+#define P8_UNORM_PALETTE1 333
+#define P4A4_UNORM_PALETTE1 334
+#define A4P4_UNORM_PALETTE1 335
+#define Y8_UNORM 336
+#define L8_UINT 338
+#define L8_SINT 339
+#define I8_UINT 340
+#define I8_SINT 341
+#define DXT1_RGB_SRGB 384
+#define R1_UNORM 385
+#define YCRCB_NORMAL 386
+#define YCRCB_SWAPUVY 387
+#define P2_UNORM_PALETTE0 388
+#define P2_UNORM_PALETTE1 389
+#define BC1_UNORM 390
+#define BC2_UNORM 391
+#define BC3_UNORM 392
+#define BC4_UNORM 393
+#define BC5_UNORM 394
+#define BC1_UNORM_SRGB 395
+#define BC2_UNORM_SRGB 396
+#define BC3_UNORM_SRGB 397
+#define MONO8 398
+#define YCRCB_SWAPUV 399
+#define YCRCB_SWAPY 400
+#define DXT1_RGB 401
+#define FXT1 402
+#define R8G8B8_UNORM 403
+#define R8G8B8_SNORM 404
+#define R8G8B8_SSCALED 405
+#define R8G8B8_USCALED 406
+#define R64G64B64A64_FLOAT 407
+#define R64G64B64_FLOAT 408
+#define BC4_SNORM 409
+#define BC5_SNORM 410
+#define R16G16B16_FLOAT 411
+#define R16G16B16_UNORM 412
+#define R16G16B16_SNORM 413
+#define R16G16B16_SSCALED 414
+#define R16G16B16_USCALED 415
+#define BC6H_SF16 417
+#define BC7_UNORM 418
+#define BC7_UNORM_SRGB 419
+#define BC6H_UF16 420
+#define PLANAR_420_8 421
+#define R8G8B8_UNORM_SRGB 424
+#define ETC1_RGB8 425
+#define ETC2_RGB8 426
+#define EAC_R11 427
+#define EAC_RG11 428
+#define EAC_SIGNED_R11 429
+#define EAC_SIGNED_RG11 430
+#define ETC2_SRGB8 431
+#define R16G16B16_UINT 432
+#define R16G16B16_SINT 433
+#define R32_SFIXED 434
+#define R10G10B10A2_SNORM 435
+#define R10G10B10A2_USCALED 436
+#define R10G10B10A2_SSCALED 437
+#define R10G10B10A2_SINT 438
+#define B10G10R10A2_SNORM 439
+#define B10G10R10A2_USCALED 440
+#define B10G10R10A2_SSCALED 441
+#define B10G10R10A2_UINT 442
+#define B10G10R10A2_SINT 443
+#define R64G64B64A64_PASSTHRU 444
+#define R64G64B64_PASSTHRU 445
+#define ETC2_RGB8_PTA 448
+#define ETC2_SRGB8_PTA 449
+#define ETC2_EAC_RGBA8 450
+#define ETC2_EAC_SRGB8_A8 451
+#define R8G8B8_UINT 456
+#define R8G8B8_SINT 457
+#define RAW 511
+
+/* Enum Texture Coordinate Mode */
+#define TCM_WRAP 0
+#define TCM_MIRROR 1
+#define TCM_CLAMP 2
+#define TCM_CUBE 3
+#define TCM_CLAMP_BORDER 4
+#define TCM_MIRROR_ONCE 5
+
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <assert.h>
+#include <stdbool.h>
+#include <string.h>
+#include <unistd.h>
+#include <fcntl.h>
+
+#include "anv_private.h"
+
+#include "gen7_pack.h"
+#include "gen75_pack.h"
+
+static uint32_t
+cmd_buffer_flush_push_constants(struct anv_cmd_buffer *cmd_buffer)
+{
+ static const uint32_t push_constant_opcodes[] = {
+ [MESA_SHADER_VERTEX] = 21,
+ [MESA_SHADER_TESS_CTRL] = 25, /* HS */
+ [MESA_SHADER_TESS_EVAL] = 26, /* DS */
+ [MESA_SHADER_GEOMETRY] = 22,
+ [MESA_SHADER_FRAGMENT] = 23,
+ [MESA_SHADER_COMPUTE] = 0,
+ };
+
+ VkShaderStageFlags flushed = 0;
+
+ anv_foreach_stage(stage, cmd_buffer->state.push_constants_dirty) {
+ struct anv_state state = anv_cmd_buffer_push_constants(cmd_buffer, stage);
+
+ if (state.offset == 0)
+ continue;
+
+ anv_batch_emit(&cmd_buffer->batch, GEN7_3DSTATE_CONSTANT_VS,
+ ._3DCommandSubOpcode = push_constant_opcodes[stage],
+ .ConstantBody = {
+ .PointerToConstantBuffer0 = { .offset = state.offset },
+ .ConstantBuffer0ReadLength = DIV_ROUND_UP(state.alloc_size, 32),
+ });
+
+ flushed |= mesa_to_vk_shader_stage(stage);
+ }
+
+ cmd_buffer->state.push_constants_dirty &= ~flushed;
+
+ return flushed;
+}
+
+GENX_FUNC(GEN7, GEN7) void
+genX(cmd_buffer_emit_descriptor_pointers)(struct anv_cmd_buffer *cmd_buffer,
+ uint32_t stages)
+{
+ static const uint32_t sampler_state_opcodes[] = {
+ [MESA_SHADER_VERTEX] = 43,
+ [MESA_SHADER_TESS_CTRL] = 44, /* HS */
+ [MESA_SHADER_TESS_EVAL] = 45, /* DS */
+ [MESA_SHADER_GEOMETRY] = 46,
+ [MESA_SHADER_FRAGMENT] = 47,
+ [MESA_SHADER_COMPUTE] = 0,
+ };
+
+ static const uint32_t binding_table_opcodes[] = {
+ [MESA_SHADER_VERTEX] = 38,
+ [MESA_SHADER_TESS_CTRL] = 39,
+ [MESA_SHADER_TESS_EVAL] = 40,
+ [MESA_SHADER_GEOMETRY] = 41,
+ [MESA_SHADER_FRAGMENT] = 42,
+ [MESA_SHADER_COMPUTE] = 0,
+ };
+
+ anv_foreach_stage(s, stages) {
+ if (cmd_buffer->state.samplers[s].alloc_size > 0) {
+ anv_batch_emit(&cmd_buffer->batch,
+ GEN7_3DSTATE_SAMPLER_STATE_POINTERS_VS,
+ ._3DCommandSubOpcode = sampler_state_opcodes[s],
+ .PointertoVSSamplerState = cmd_buffer->state.samplers[s].offset);
+ }
+
+ /* Always emit binding table pointers if we're asked to, since on SKL
+ * this is what flushes push constants. */
+ anv_batch_emit(&cmd_buffer->batch,
+ GEN7_3DSTATE_BINDING_TABLE_POINTERS_VS,
+ ._3DCommandSubOpcode = binding_table_opcodes[s],
+ .PointertoVSBindingTable = cmd_buffer->state.binding_tables[s].offset);
+ }
+}
+
+GENX_FUNC(GEN7, GEN7) uint32_t
+genX(cmd_buffer_flush_descriptor_sets)(struct anv_cmd_buffer *cmd_buffer)
+{
+ VkShaderStageFlags dirty = cmd_buffer->state.descriptors_dirty &
+ cmd_buffer->state.pipeline->active_stages;
+
+ VkResult result = VK_SUCCESS;
+ anv_foreach_stage(s, dirty) {
+ result = anv_cmd_buffer_emit_samplers(cmd_buffer, s,
+ &cmd_buffer->state.samplers[s]);
+ if (result != VK_SUCCESS)
+ break;
+ result = anv_cmd_buffer_emit_binding_table(cmd_buffer, s,
+ &cmd_buffer->state.binding_tables[s]);
+ if (result != VK_SUCCESS)
+ break;
+ }
+
+ if (result != VK_SUCCESS) {
+ assert(result == VK_ERROR_OUT_OF_DEVICE_MEMORY);
+
+ result = anv_cmd_buffer_new_binding_table_block(cmd_buffer);
+ assert(result == VK_SUCCESS);
+
+ /* Re-emit state base addresses so we get the new surface state base
+ * address before we start emitting binding tables etc.
+ */
+ anv_cmd_buffer_emit_state_base_address(cmd_buffer);
+
+ /* Re-emit all active binding tables */
+ dirty |= cmd_buffer->state.pipeline->active_stages;
+ anv_foreach_stage(s, dirty) {
+ result = anv_cmd_buffer_emit_samplers(cmd_buffer, s,
+ &cmd_buffer->state.samplers[s]);
+ if (result != VK_SUCCESS)
+ return result;
+ result = anv_cmd_buffer_emit_binding_table(cmd_buffer, s,
+ &cmd_buffer->state.binding_tables[s]);
+ if (result != VK_SUCCESS)
+ return result;
+ }
+ }
+
+ cmd_buffer->state.descriptors_dirty &= ~dirty;
+
+ return dirty;
+}
+
+static inline int64_t
+clamp_int64(int64_t x, int64_t min, int64_t max)
+{
+ if (x < min)
+ return min;
+ else if (x < max)
+ return x;
+ else
+ return max;
+}
+
+static void
+emit_scissor_state(struct anv_cmd_buffer *cmd_buffer,
+ uint32_t count, const VkRect2D *scissors)
+{
+ struct anv_state scissor_state =
+ anv_cmd_buffer_alloc_dynamic_state(cmd_buffer, count * 32, 32);
+
+ for (uint32_t i = 0; i < count; i++) {
+ const VkRect2D *s = &scissors[i];
+
+ /* Since xmax and ymax are inclusive, we have to have xmax < xmin or
+ * ymax < ymin for empty clips. In case clip x, y, width height are all
+ * 0, the clamps below produce 0 for xmin, ymin, xmax, ymax, which isn't
+ * what we want. Just special case empty clips and produce a canonical
+ * empty clip. */
+ static const struct GEN7_SCISSOR_RECT empty_scissor = {
+ .ScissorRectangleYMin = 1,
+ .ScissorRectangleXMin = 1,
+ .ScissorRectangleYMax = 0,
+ .ScissorRectangleXMax = 0
+ };
+
+ const int max = 0xffff;
+ struct GEN7_SCISSOR_RECT scissor = {
+ /* Do this math using int64_t so overflow gets clamped correctly. */
+ .ScissorRectangleYMin = clamp_int64(s->offset.y, 0, max),
+ .ScissorRectangleXMin = clamp_int64(s->offset.x, 0, max),
+ .ScissorRectangleYMax = clamp_int64((uint64_t) s->offset.y + s->extent.height - 1, 0, max),
+ .ScissorRectangleXMax = clamp_int64((uint64_t) s->offset.x + s->extent.width - 1, 0, max)
+ };
+
+ if (s->extent.width <= 0 || s->extent.height <= 0) {
+ GEN7_SCISSOR_RECT_pack(NULL, scissor_state.map + i * 32,
+ &empty_scissor);
+ } else {
+ GEN7_SCISSOR_RECT_pack(NULL, scissor_state.map + i * 32, &scissor);
+ }
+ }
+
+ anv_batch_emit(&cmd_buffer->batch, GEN7_3DSTATE_SCISSOR_STATE_POINTERS,
+ .ScissorRectPointer = scissor_state.offset);
+
+ if (!cmd_buffer->device->info.has_llc)
+ anv_state_clflush(scissor_state);
+}
+
+GENX_FUNC(GEN7, GEN7) void
+genX(cmd_buffer_emit_scissor)(struct anv_cmd_buffer *cmd_buffer)
+{
+ if (cmd_buffer->state.dynamic.scissor.count > 0) {
+ emit_scissor_state(cmd_buffer, cmd_buffer->state.dynamic.scissor.count,
+ cmd_buffer->state.dynamic.scissor.scissors);
+ } else {
+ /* Emit a default scissor based on the currently bound framebuffer */
+ emit_scissor_state(cmd_buffer, 1,
+ &(VkRect2D) {
+ .offset = { .x = 0, .y = 0, },
+ .extent = {
+ .width = cmd_buffer->state.framebuffer->width,
+ .height = cmd_buffer->state.framebuffer->height,
+ },
+ });
+ }
+}
+
+static const uint32_t vk_to_gen_index_type[] = {
+ [VK_INDEX_TYPE_UINT16] = INDEX_WORD,
+ [VK_INDEX_TYPE_UINT32] = INDEX_DWORD,
+};
+
+static const uint32_t restart_index_for_type[] = {
+ [VK_INDEX_TYPE_UINT16] = UINT16_MAX,
+ [VK_INDEX_TYPE_UINT32] = UINT32_MAX,
+};
+
+void genX(CmdBindIndexBuffer)(
+ VkCommandBuffer commandBuffer,
+ VkBuffer _buffer,
+ VkDeviceSize offset,
+ VkIndexType indexType)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_buffer, buffer, _buffer);
+
+ cmd_buffer->state.dirty |= ANV_CMD_DIRTY_INDEX_BUFFER;
+ if (ANV_IS_HASWELL)
+ cmd_buffer->state.restart_index = restart_index_for_type[indexType];
+ cmd_buffer->state.gen7.index_buffer = buffer;
+ cmd_buffer->state.gen7.index_type = vk_to_gen_index_type[indexType];
+ cmd_buffer->state.gen7.index_offset = offset;
+}
+
+static VkResult
+flush_compute_descriptor_set(struct anv_cmd_buffer *cmd_buffer)
+{
+ struct anv_device *device = cmd_buffer->device;
+ struct anv_pipeline *pipeline = cmd_buffer->state.compute_pipeline;
+ struct anv_state surfaces = { 0, }, samplers = { 0, };
+ VkResult result;
+
+ result = anv_cmd_buffer_emit_samplers(cmd_buffer,
+ MESA_SHADER_COMPUTE, &samplers);
+ if (result != VK_SUCCESS)
+ return result;
+ result = anv_cmd_buffer_emit_binding_table(cmd_buffer,
+ MESA_SHADER_COMPUTE, &surfaces);
+ if (result != VK_SUCCESS)
+ return result;
+
+ const struct brw_cs_prog_data *cs_prog_data = &pipeline->cs_prog_data;
+
+ struct anv_state state =
+ anv_state_pool_emit(&device->dynamic_state_pool,
+ GEN7_INTERFACE_DESCRIPTOR_DATA, 64,
+ .KernelStartPointer = pipeline->cs_simd,
+ .BindingTablePointer = surfaces.offset,
+ .SamplerStatePointer = samplers.offset,
+ .BarrierEnable = cs_prog_data->uses_barrier,
+ .NumberofThreadsinGPGPUThreadGroup =
+ pipeline->cs_thread_width_max);
+
+ const uint32_t size = GEN7_INTERFACE_DESCRIPTOR_DATA_length * sizeof(uint32_t);
+ anv_batch_emit(&cmd_buffer->batch, GEN7_MEDIA_INTERFACE_DESCRIPTOR_LOAD,
+ .InterfaceDescriptorTotalLength = size,
+ .InterfaceDescriptorDataStartAddress = state.offset);
+
+ return VK_SUCCESS;
+}
+
+static void
+cmd_buffer_flush_compute_state(struct anv_cmd_buffer *cmd_buffer)
+{
+ struct anv_pipeline *pipeline = cmd_buffer->state.compute_pipeline;
+ VkResult result;
+
+ assert(pipeline->active_stages == VK_SHADER_STAGE_COMPUTE_BIT);
+
+ if (cmd_buffer->state.current_pipeline != GPGPU) {
+ anv_batch_emit(&cmd_buffer->batch, GEN7_PIPELINE_SELECT,
+ .PipelineSelection = GPGPU);
+ cmd_buffer->state.current_pipeline = GPGPU;
+ }
+
+ if (cmd_buffer->state.compute_dirty & ANV_CMD_DIRTY_PIPELINE)
+ anv_batch_emit_batch(&cmd_buffer->batch, &pipeline->batch);
+
+ if ((cmd_buffer->state.descriptors_dirty & VK_SHADER_STAGE_COMPUTE_BIT) ||
+ (cmd_buffer->state.compute_dirty & ANV_CMD_DIRTY_PIPELINE)) {
+ /* FIXME: figure out descriptors for gen7 */
+ result = flush_compute_descriptor_set(cmd_buffer);
+ assert(result == VK_SUCCESS);
+ cmd_buffer->state.descriptors_dirty &= ~VK_SHADER_STAGE_COMPUTE_BIT;
+ }
+
+ cmd_buffer->state.compute_dirty = 0;
+}
+
+static void
+cmd_buffer_flush_state(struct anv_cmd_buffer *cmd_buffer)
+{
+ struct anv_pipeline *pipeline = cmd_buffer->state.pipeline;
+ uint32_t *p;
+
+ uint32_t vb_emit = cmd_buffer->state.vb_dirty & pipeline->vb_used;
+
+ assert((pipeline->active_stages & VK_SHADER_STAGE_COMPUTE_BIT) == 0);
+
+ if (cmd_buffer->state.current_pipeline != _3D) {
+ anv_batch_emit(&cmd_buffer->batch, GEN7_PIPELINE_SELECT,
+ .PipelineSelection = _3D);
+ cmd_buffer->state.current_pipeline = _3D;
+ }
+
+ if (vb_emit) {
+ const uint32_t num_buffers = __builtin_popcount(vb_emit);
+ const uint32_t num_dwords = 1 + num_buffers * 4;
+
+ p = anv_batch_emitn(&cmd_buffer->batch, num_dwords,
+ GEN7_3DSTATE_VERTEX_BUFFERS);
+ uint32_t vb, i = 0;
+ for_each_bit(vb, vb_emit) {
+ struct anv_buffer *buffer = cmd_buffer->state.vertex_bindings[vb].buffer;
+ uint32_t offset = cmd_buffer->state.vertex_bindings[vb].offset;
+
+ struct GEN7_VERTEX_BUFFER_STATE state = {
+ .VertexBufferIndex = vb,
+ .BufferAccessType = pipeline->instancing_enable[vb] ? INSTANCEDATA : VERTEXDATA,
+ .VertexBufferMemoryObjectControlState = GEN7_MOCS,
+ .AddressModifyEnable = true,
+ .BufferPitch = pipeline->binding_stride[vb],
+ .BufferStartingAddress = { buffer->bo, buffer->offset + offset },
+ .EndAddress = { buffer->bo, buffer->offset + buffer->size - 1},
+ .InstanceDataStepRate = 1
+ };
+
+ GEN7_VERTEX_BUFFER_STATE_pack(&cmd_buffer->batch, &p[1 + i * 4], &state);
+ i++;
+ }
+ }
+
+ if (cmd_buffer->state.dirty & ANV_CMD_DIRTY_PIPELINE) {
+ /* If somebody compiled a pipeline after starting a command buffer the
+ * scratch bo may have grown since we started this cmd buffer (and
+ * emitted STATE_BASE_ADDRESS). If we're binding that pipeline now,
+ * reemit STATE_BASE_ADDRESS so that we use the bigger scratch bo. */
+ if (cmd_buffer->state.scratch_size < pipeline->total_scratch)
+ gen7_cmd_buffer_emit_state_base_address(cmd_buffer);
+
+ anv_batch_emit_batch(&cmd_buffer->batch, &pipeline->batch);
+ }
+
+ if (cmd_buffer->state.descriptors_dirty & VK_SHADER_STAGE_VERTEX_BIT ||
+ cmd_buffer->state.push_constants_dirty & VK_SHADER_STAGE_VERTEX_BIT) {
+ /* From the IVB PRM Vol. 2, Part 1, Section 3.2.1:
+ *
+ * "A PIPE_CONTROL with Post-Sync Operation set to 1h and a depth
+ * stall needs to be sent just prior to any 3DSTATE_VS,
+ * 3DSTATE_URB_VS, 3DSTATE_CONSTANT_VS,
+ * 3DSTATE_BINDING_TABLE_POINTER_VS,
+ * 3DSTATE_SAMPLER_STATE_POINTER_VS command. Only one
+ * PIPE_CONTROL needs to be sent before any combination of VS
+ * associated 3DSTATE."
+ */
+ anv_batch_emit(&cmd_buffer->batch, GEN7_PIPE_CONTROL,
+ .DepthStallEnable = true,
+ .PostSyncOperation = WriteImmediateData,
+ .Address = { &cmd_buffer->device->workaround_bo, 0 });
+ }
+
+ uint32_t dirty = 0;
+ if (cmd_buffer->state.descriptors_dirty) {
+ dirty = gen7_cmd_buffer_flush_descriptor_sets(cmd_buffer);
+ gen7_cmd_buffer_emit_descriptor_pointers(cmd_buffer, dirty);
+ }
+
+ if (cmd_buffer->state.push_constants_dirty)
+ cmd_buffer_flush_push_constants(cmd_buffer);
+
+ /* We use the gen8 state here because it only contains the additional
+ * min/max fields and, since they occur at the end of the packet and
+ * don't change the stride, they work on gen7 too.
+ */
+ if (cmd_buffer->state.dirty & ANV_CMD_DIRTY_DYNAMIC_VIEWPORT)
+ gen8_cmd_buffer_emit_viewport(cmd_buffer);
+
+ if (cmd_buffer->state.dirty & ANV_CMD_DIRTY_DYNAMIC_SCISSOR)
+ gen7_cmd_buffer_emit_scissor(cmd_buffer);
+
+ if (cmd_buffer->state.dirty & (ANV_CMD_DIRTY_PIPELINE |
+ ANV_CMD_DIRTY_DYNAMIC_LINE_WIDTH |
+ ANV_CMD_DIRTY_DYNAMIC_DEPTH_BIAS)) {
+
+ bool enable_bias = cmd_buffer->state.dynamic.depth_bias.bias != 0.0f ||
+ cmd_buffer->state.dynamic.depth_bias.slope != 0.0f;
+
+ uint32_t sf_dw[GEN7_3DSTATE_SF_length];
+ struct GEN7_3DSTATE_SF sf = {
+ GEN7_3DSTATE_SF_header,
+ .LineWidth = cmd_buffer->state.dynamic.line_width,
+ .GlobalDepthOffsetEnableSolid = enable_bias,
+ .GlobalDepthOffsetEnableWireframe = enable_bias,
+ .GlobalDepthOffsetEnablePoint = enable_bias,
+ .GlobalDepthOffsetConstant = cmd_buffer->state.dynamic.depth_bias.bias,
+ .GlobalDepthOffsetScale = cmd_buffer->state.dynamic.depth_bias.slope,
+ .GlobalDepthOffsetClamp = cmd_buffer->state.dynamic.depth_bias.clamp
+ };
+ GEN7_3DSTATE_SF_pack(NULL, sf_dw, &sf);
+
+ anv_batch_emit_merge(&cmd_buffer->batch, sf_dw, pipeline->gen7.sf);
+ }
+
+ if (cmd_buffer->state.dirty & (ANV_CMD_DIRTY_DYNAMIC_BLEND_CONSTANTS |
+ ANV_CMD_DIRTY_DYNAMIC_STENCIL_REFERENCE)) {
+ struct anv_state cc_state =
+ anv_cmd_buffer_alloc_dynamic_state(cmd_buffer,
+ GEN7_COLOR_CALC_STATE_length * 4,
+ 64);
+ struct GEN7_COLOR_CALC_STATE cc = {
+ .BlendConstantColorRed = cmd_buffer->state.dynamic.blend_constants[0],
+ .BlendConstantColorGreen = cmd_buffer->state.dynamic.blend_constants[1],
+ .BlendConstantColorBlue = cmd_buffer->state.dynamic.blend_constants[2],
+ .BlendConstantColorAlpha = cmd_buffer->state.dynamic.blend_constants[3],
+ .StencilReferenceValue =
+ cmd_buffer->state.dynamic.stencil_reference.front,
+ .BackFaceStencilReferenceValue =
+ cmd_buffer->state.dynamic.stencil_reference.back,
+ };
+ GEN7_COLOR_CALC_STATE_pack(NULL, cc_state.map, &cc);
+ if (!cmd_buffer->device->info.has_llc)
+ anv_state_clflush(cc_state);
+
+ anv_batch_emit(&cmd_buffer->batch,
+ GEN7_3DSTATE_CC_STATE_POINTERS,
+ .ColorCalcStatePointer = cc_state.offset);
+ }
+
+ if (cmd_buffer->state.dirty & (ANV_CMD_DIRTY_PIPELINE |
+ ANV_CMD_DIRTY_RENDER_TARGETS |
+ ANV_CMD_DIRTY_DYNAMIC_STENCIL_COMPARE_MASK |
+ ANV_CMD_DIRTY_DYNAMIC_STENCIL_WRITE_MASK)) {
+ uint32_t depth_stencil_dw[GEN7_DEPTH_STENCIL_STATE_length];
+
+ const struct anv_image_view *iview =
+ anv_cmd_buffer_get_depth_stencil_view(cmd_buffer);
+
+ struct GEN7_DEPTH_STENCIL_STATE depth_stencil = {
+ .StencilBufferWriteEnable = iview && (iview->aspect_mask & VK_IMAGE_ASPECT_STENCIL_BIT),
+
+ .StencilTestMask =
+ cmd_buffer->state.dynamic.stencil_compare_mask.front & 0xff,
+ .StencilWriteMask =
+ cmd_buffer->state.dynamic.stencil_write_mask.front & 0xff,
+
+ .BackfaceStencilTestMask =
+ cmd_buffer->state.dynamic.stencil_compare_mask.back & 0xff,
+ .BackfaceStencilWriteMask =
+ cmd_buffer->state.dynamic.stencil_write_mask.back & 0xff,
+ };
+ GEN7_DEPTH_STENCIL_STATE_pack(NULL, depth_stencil_dw, &depth_stencil);
+
+ struct anv_state ds_state =
+ anv_cmd_buffer_merge_dynamic(cmd_buffer, depth_stencil_dw,
+ pipeline->gen7.depth_stencil_state,
+ GEN7_DEPTH_STENCIL_STATE_length, 64);
+
+ anv_batch_emit(&cmd_buffer->batch,
+ GEN7_3DSTATE_DEPTH_STENCIL_STATE_POINTERS,
+ .PointertoDEPTH_STENCIL_STATE = ds_state.offset);
+ }
+
+ if (cmd_buffer->state.gen7.index_buffer &&
+ cmd_buffer->state.dirty & (ANV_CMD_DIRTY_PIPELINE |
+ ANV_CMD_DIRTY_INDEX_BUFFER)) {
+ struct anv_buffer *buffer = cmd_buffer->state.gen7.index_buffer;
+ uint32_t offset = cmd_buffer->state.gen7.index_offset;
+
+ if (ANV_IS_HASWELL) {
+ anv_batch_emit(&cmd_buffer->batch, GEN75_3DSTATE_VF,
+ .IndexedDrawCutIndexEnable = pipeline->primitive_restart,
+ .CutIndex = cmd_buffer->state.restart_index);
+ }
+
+ anv_batch_emit(&cmd_buffer->batch, GEN7_3DSTATE_INDEX_BUFFER,
+ .CutIndexEnable = pipeline->primitive_restart,
+ .IndexFormat = cmd_buffer->state.gen7.index_type,
+ .MemoryObjectControlState = GEN7_MOCS,
+ .BufferStartingAddress = { buffer->bo, buffer->offset + offset },
+ .BufferEndingAddress = { buffer->bo, buffer->offset + buffer->size });
+ }
+
+ cmd_buffer->state.vb_dirty &= ~vb_emit;
+ cmd_buffer->state.dirty = 0;
+}
+
+void genX(CmdDraw)(
+ VkCommandBuffer commandBuffer,
+ uint32_t vertexCount,
+ uint32_t instanceCount,
+ uint32_t firstVertex,
+ uint32_t firstInstance)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ struct anv_pipeline *pipeline = cmd_buffer->state.pipeline;
+
+ cmd_buffer_flush_state(cmd_buffer);
+
+ anv_batch_emit(&cmd_buffer->batch, GEN7_3DPRIMITIVE,
+ .VertexAccessType = SEQUENTIAL,
+ .PrimitiveTopologyType = pipeline->topology,
+ .VertexCountPerInstance = vertexCount,
+ .StartVertexLocation = firstVertex,
+ .InstanceCount = instanceCount,
+ .StartInstanceLocation = firstInstance,
+ .BaseVertexLocation = 0);
+}
+
+void genX(CmdDrawIndexed)(
+ VkCommandBuffer commandBuffer,
+ uint32_t indexCount,
+ uint32_t instanceCount,
+ uint32_t firstIndex,
+ int32_t vertexOffset,
+ uint32_t firstInstance)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ struct anv_pipeline *pipeline = cmd_buffer->state.pipeline;
+
+ cmd_buffer_flush_state(cmd_buffer);
+
+ anv_batch_emit(&cmd_buffer->batch, GEN7_3DPRIMITIVE,
+ .VertexAccessType = RANDOM,
+ .PrimitiveTopologyType = pipeline->topology,
+ .VertexCountPerInstance = indexCount,
+ .StartVertexLocation = firstIndex,
+ .InstanceCount = instanceCount,
+ .StartInstanceLocation = firstInstance,
+ .BaseVertexLocation = vertexOffset);
+}
+
+static void
+gen7_batch_lrm(struct anv_batch *batch,
+ uint32_t reg, struct anv_bo *bo, uint32_t offset)
+{
+ anv_batch_emit(batch, GEN7_MI_LOAD_REGISTER_MEM,
+ .RegisterAddress = reg,
+ .MemoryAddress = { bo, offset });
+}
+
+static void
+gen7_batch_lri(struct anv_batch *batch, uint32_t reg, uint32_t imm)
+{
+ anv_batch_emit(batch, GEN7_MI_LOAD_REGISTER_IMM,
+ .RegisterOffset = reg,
+ .DataDWord = imm);
+}
+
+/* Auto-Draw / Indirect Registers */
+#define GEN7_3DPRIM_END_OFFSET 0x2420
+#define GEN7_3DPRIM_START_VERTEX 0x2430
+#define GEN7_3DPRIM_VERTEX_COUNT 0x2434
+#define GEN7_3DPRIM_INSTANCE_COUNT 0x2438
+#define GEN7_3DPRIM_START_INSTANCE 0x243C
+#define GEN7_3DPRIM_BASE_VERTEX 0x2440
+
+void genX(CmdDrawIndirect)(
+ VkCommandBuffer commandBuffer,
+ VkBuffer _buffer,
+ VkDeviceSize offset,
+ uint32_t drawCount,
+ uint32_t stride)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_buffer, buffer, _buffer);
+ struct anv_pipeline *pipeline = cmd_buffer->state.pipeline;
+ struct anv_bo *bo = buffer->bo;
+ uint32_t bo_offset = buffer->offset + offset;
+
+ cmd_buffer_flush_state(cmd_buffer);
+
+ gen7_batch_lrm(&cmd_buffer->batch, GEN7_3DPRIM_VERTEX_COUNT, bo, bo_offset);
+ gen7_batch_lrm(&cmd_buffer->batch, GEN7_3DPRIM_INSTANCE_COUNT, bo, bo_offset + 4);
+ gen7_batch_lrm(&cmd_buffer->batch, GEN7_3DPRIM_START_VERTEX, bo, bo_offset + 8);
+ gen7_batch_lrm(&cmd_buffer->batch, GEN7_3DPRIM_START_INSTANCE, bo, bo_offset + 12);
+ gen7_batch_lri(&cmd_buffer->batch, GEN7_3DPRIM_BASE_VERTEX, 0);
+
+ anv_batch_emit(&cmd_buffer->batch, GEN7_3DPRIMITIVE,
+ .IndirectParameterEnable = true,
+ .VertexAccessType = SEQUENTIAL,
+ .PrimitiveTopologyType = pipeline->topology);
+}
+
+void genX(CmdDrawIndexedIndirect)(
+ VkCommandBuffer commandBuffer,
+ VkBuffer _buffer,
+ VkDeviceSize offset,
+ uint32_t drawCount,
+ uint32_t stride)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_buffer, buffer, _buffer);
+ struct anv_pipeline *pipeline = cmd_buffer->state.pipeline;
+ struct anv_bo *bo = buffer->bo;
+ uint32_t bo_offset = buffer->offset + offset;
+
+ cmd_buffer_flush_state(cmd_buffer);
+
+ gen7_batch_lrm(&cmd_buffer->batch, GEN7_3DPRIM_VERTEX_COUNT, bo, bo_offset);
+ gen7_batch_lrm(&cmd_buffer->batch, GEN7_3DPRIM_INSTANCE_COUNT, bo, bo_offset + 4);
+ gen7_batch_lrm(&cmd_buffer->batch, GEN7_3DPRIM_START_VERTEX, bo, bo_offset + 8);
+ gen7_batch_lrm(&cmd_buffer->batch, GEN7_3DPRIM_BASE_VERTEX, bo, bo_offset + 12);
+ gen7_batch_lrm(&cmd_buffer->batch, GEN7_3DPRIM_START_INSTANCE, bo, bo_offset + 16);
+
+ anv_batch_emit(&cmd_buffer->batch, GEN7_3DPRIMITIVE,
+ .IndirectParameterEnable = true,
+ .VertexAccessType = RANDOM,
+ .PrimitiveTopologyType = pipeline->topology);
+}
+
+void genX(CmdDispatch)(
+ VkCommandBuffer commandBuffer,
+ uint32_t x,
+ uint32_t y,
+ uint32_t z)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ struct anv_pipeline *pipeline = cmd_buffer->state.compute_pipeline;
+ struct brw_cs_prog_data *prog_data = &pipeline->cs_prog_data;
+
+ cmd_buffer_flush_compute_state(cmd_buffer);
+
+ anv_batch_emit(&cmd_buffer->batch, GEN7_GPGPU_WALKER,
+ .SIMDSize = prog_data->simd_size / 16,
+ .ThreadDepthCounterMaximum = 0,
+ .ThreadHeightCounterMaximum = 0,
+ .ThreadWidthCounterMaximum = pipeline->cs_thread_width_max - 1,
+ .ThreadGroupIDXDimension = x,
+ .ThreadGroupIDYDimension = y,
+ .ThreadGroupIDZDimension = z,
+ .RightExecutionMask = pipeline->cs_right_mask,
+ .BottomExecutionMask = 0xffffffff);
+
+ anv_batch_emit(&cmd_buffer->batch, GEN7_MEDIA_STATE_FLUSH);
+}
+
+#define GPGPU_DISPATCHDIMX 0x2500
+#define GPGPU_DISPATCHDIMY 0x2504
+#define GPGPU_DISPATCHDIMZ 0x2508
+
+void genX(CmdDispatchIndirect)(
+ VkCommandBuffer commandBuffer,
+ VkBuffer _buffer,
+ VkDeviceSize offset)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_buffer, buffer, _buffer);
+ struct anv_pipeline *pipeline = cmd_buffer->state.compute_pipeline;
+ struct brw_cs_prog_data *prog_data = &pipeline->cs_prog_data;
+ struct anv_bo *bo = buffer->bo;
+ uint32_t bo_offset = buffer->offset + offset;
+
+ cmd_buffer_flush_compute_state(cmd_buffer);
+
+ gen7_batch_lrm(&cmd_buffer->batch, GPGPU_DISPATCHDIMX, bo, bo_offset);
+ gen7_batch_lrm(&cmd_buffer->batch, GPGPU_DISPATCHDIMY, bo, bo_offset + 4);
+ gen7_batch_lrm(&cmd_buffer->batch, GPGPU_DISPATCHDIMZ, bo, bo_offset + 8);
+
+ anv_batch_emit(&cmd_buffer->batch, GEN7_GPGPU_WALKER,
+ .IndirectParameterEnable = true,
+ .SIMDSize = prog_data->simd_size / 16,
+ .ThreadDepthCounterMaximum = 0,
+ .ThreadHeightCounterMaximum = 0,
+ .ThreadWidthCounterMaximum = pipeline->cs_thread_width_max - 1,
+ .RightExecutionMask = pipeline->cs_right_mask,
+ .BottomExecutionMask = 0xffffffff);
+
+ anv_batch_emit(&cmd_buffer->batch, GEN7_MEDIA_STATE_FLUSH);
+}
+
+static void
+cmd_buffer_emit_depth_stencil(struct anv_cmd_buffer *cmd_buffer)
+{
+ const struct anv_framebuffer *fb = cmd_buffer->state.framebuffer;
+ const struct anv_image_view *iview =
+ anv_cmd_buffer_get_depth_stencil_view(cmd_buffer);
+ const struct anv_image *image = iview ? iview->image : NULL;
+
+ /* XXX: isl needs to grow depth format support */
+ const struct anv_format *anv_format =
+ iview ? anv_format_for_vk_format(iview->vk_format) : NULL;
+
+ const bool has_depth = iview && anv_format->depth_format;
+ const bool has_stencil = iview && anv_format->has_stencil;
+
+ /* Emit 3DSTATE_DEPTH_BUFFER */
+ if (has_depth) {
+ anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_DEPTH_BUFFER),
+ .SurfaceType = SURFTYPE_2D,
+ .DepthWriteEnable = true,
+ .StencilWriteEnable = has_stencil,
+ .HierarchicalDepthBufferEnable = false,
+ .SurfaceFormat = anv_format->depth_format,
+ .SurfacePitch = image->depth_surface.isl.row_pitch - 1,
+ .SurfaceBaseAddress = {
+ .bo = image->bo,
+ .offset = image->depth_surface.offset,
+ },
+ .Height = fb->height - 1,
+ .Width = fb->width - 1,
+ .LOD = 0,
+ .Depth = 1 - 1,
+ .MinimumArrayElement = 0,
+ .DepthBufferObjectControlState = GENX(MOCS),
+ .RenderTargetViewExtent = 1 - 1);
+ } else {
+ /* Even when no depth buffer is present, the hardware requires that
+ * 3DSTATE_DEPTH_BUFFER be programmed correctly. The Broadwell PRM says:
+ *
+ * If a null depth buffer is bound, the driver must instead bind depth as:
+ * 3DSTATE_DEPTH.SurfaceType = SURFTYPE_2D
+ * 3DSTATE_DEPTH.Width = 1
+ * 3DSTATE_DEPTH.Height = 1
+ * 3DSTATE_DEPTH.SuraceFormat = D16_UNORM
+ * 3DSTATE_DEPTH.SurfaceBaseAddress = 0
+ * 3DSTATE_DEPTH.HierarchicalDepthBufferEnable = 0
+ * 3DSTATE_WM_DEPTH_STENCIL.DepthTestEnable = 0
+ * 3DSTATE_WM_DEPTH_STENCIL.DepthBufferWriteEnable = 0
+ *
+ * The PRM is wrong, though. The width and height must be programmed to
+ * actual framebuffer's width and height, even when neither depth buffer
+ * nor stencil buffer is present.
+ */
+ anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_DEPTH_BUFFER),
+ .SurfaceType = SURFTYPE_2D,
+ .SurfaceFormat = D16_UNORM,
+ .Width = fb->width - 1,
+ .Height = fb->height - 1,
+ .StencilWriteEnable = has_stencil);
+ }
+
+ /* Emit 3DSTATE_STENCIL_BUFFER */
+ if (has_stencil) {
+ anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_STENCIL_BUFFER),
+# if (ANV_IS_HASWELL)
+ .StencilBufferEnable = true,
+# endif
+ .StencilBufferObjectControlState = GENX(MOCS),
+
+ /* Stencil buffers have strange pitch. The PRM says:
+ *
+ * The pitch must be set to 2x the value computed based on width,
+ * as the stencil buffer is stored with two rows interleaved.
+ */
+ .SurfacePitch = 2 * image->stencil_surface.isl.row_pitch - 1,
+
+ .SurfaceBaseAddress = {
+ .bo = image->bo,
+ .offset = image->offset + image->stencil_surface.offset,
+ });
+ } else {
+ anv_batch_emit(&cmd_buffer->batch, GEN7_3DSTATE_STENCIL_BUFFER);
+ }
+
+ /* Disable hierarchial depth buffers. */
+ anv_batch_emit(&cmd_buffer->batch, GEN7_3DSTATE_HIER_DEPTH_BUFFER);
+
+ /* Clear the clear params. */
+ anv_batch_emit(&cmd_buffer->batch, GEN7_3DSTATE_CLEAR_PARAMS);
+}
+
+/**
+ * @see anv_cmd_buffer_set_subpass()
+ */
+GENX_FUNC(GEN7, GEN7) void
+genX(cmd_buffer_set_subpass)(struct anv_cmd_buffer *cmd_buffer,
+ struct anv_subpass *subpass)
+{
+ cmd_buffer->state.subpass = subpass;
+ cmd_buffer->state.descriptors_dirty |= VK_SHADER_STAGE_FRAGMENT_BIT;
+ cmd_buffer->state.dirty |= ANV_CMD_DIRTY_RENDER_TARGETS;
+
+ cmd_buffer_emit_depth_stencil(cmd_buffer);
+}
+
+void genX(CmdBeginRenderPass)(
+ VkCommandBuffer commandBuffer,
+ const VkRenderPassBeginInfo* pRenderPassBegin,
+ VkSubpassContents contents)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_render_pass, pass, pRenderPassBegin->renderPass);
+ ANV_FROM_HANDLE(anv_framebuffer, framebuffer, pRenderPassBegin->framebuffer);
+
+ cmd_buffer->state.framebuffer = framebuffer;
+ cmd_buffer->state.pass = pass;
+ anv_cmd_state_setup_attachments(cmd_buffer, pRenderPassBegin);
+
+ const VkRect2D *render_area = &pRenderPassBegin->renderArea;
+
+ anv_batch_emit(&cmd_buffer->batch, GEN7_3DSTATE_DRAWING_RECTANGLE,
+ .ClippedDrawingRectangleYMin = render_area->offset.y,
+ .ClippedDrawingRectangleXMin = render_area->offset.x,
+ .ClippedDrawingRectangleYMax =
+ render_area->offset.y + render_area->extent.height - 1,
+ .ClippedDrawingRectangleXMax =
+ render_area->offset.x + render_area->extent.width - 1,
+ .DrawingRectangleOriginY = 0,
+ .DrawingRectangleOriginX = 0);
+
+ gen7_cmd_buffer_set_subpass(cmd_buffer, pass->subpasses);
+ anv_cmd_buffer_clear_subpass(cmd_buffer);
+}
+
+void genX(CmdNextSubpass)(
+ VkCommandBuffer commandBuffer,
+ VkSubpassContents contents)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+
+ assert(cmd_buffer->level == VK_COMMAND_BUFFER_LEVEL_PRIMARY);
+
+ gen7_cmd_buffer_set_subpass(cmd_buffer, cmd_buffer->state.subpass + 1);
+ anv_cmd_buffer_clear_subpass(cmd_buffer);
+}
+
+void genX(CmdEndRenderPass)(
+ VkCommandBuffer commandBuffer)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+
+ /* Emit a flushing pipe control at the end of a pass. This is kind of a
+ * hack but it ensures that render targets always actually get written.
+ * Eventually, we should do flushing based on image format transitions
+ * or something of that nature.
+ */
+ anv_batch_emit(&cmd_buffer->batch, GEN7_PIPE_CONTROL,
+ .PostSyncOperation = NoWrite,
+ .RenderTargetCacheFlushEnable = true,
+ .InstructionCacheInvalidateEnable = true,
+ .DepthCacheFlushEnable = true,
+ .VFCacheInvalidationEnable = true,
+ .TextureCacheInvalidationEnable = true,
+ .CommandStreamerStallEnable = true);
+}
+
+void genX(CmdSetEvent)(
+ VkCommandBuffer commandBuffer,
+ VkEvent event,
+ VkPipelineStageFlags stageMask)
+{
+ stub();
+}
+
+void genX(CmdResetEvent)(
+ VkCommandBuffer commandBuffer,
+ VkEvent event,
+ VkPipelineStageFlags stageMask)
+{
+ stub();
+}
+
+void genX(CmdWaitEvents)(
+ VkCommandBuffer commandBuffer,
+ uint32_t eventCount,
+ const VkEvent* pEvents,
+ VkPipelineStageFlags srcStageMask,
+ VkPipelineStageFlags destStageMask,
+ uint32_t memoryBarrierCount,
+ const VkMemoryBarrier* pMemoryBarriers,
+ uint32_t bufferMemoryBarrierCount,
+ const VkBufferMemoryBarrier* pBufferMemoryBarriers,
+ uint32_t imageMemoryBarrierCount,
+ const VkImageMemoryBarrier* pImageMemoryBarriers)
+{
+ stub();
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+
+/* Instructions, enums and structures for IVB.
+ *
+ * This file has been generated, do not hand edit.
+ */
+
+#pragma once
+
+#include <stdio.h>
+#include <assert.h>
+
+#ifndef __gen_validate_value
+#define __gen_validate_value(x)
+#endif
+
+#ifndef __gen_field_functions
+#define __gen_field_functions
+
+union __gen_value {
+ float f;
+ uint32_t dw;
+};
+
+static inline uint64_t
+__gen_mbo(uint32_t start, uint32_t end)
+{
+ return (~0ul >> (64 - (end - start + 1))) << start;
+}
+
+static inline uint64_t
+__gen_field(uint64_t v, uint32_t start, uint32_t end)
+{
+ __gen_validate_value(v);
+#if DEBUG
+ if (end - start + 1 < 64)
+ assert(v < 1ul << (end - start + 1));
+#endif
+
+ return v << start;
+}
+
+static inline uint64_t
+__gen_fixed(float v, uint32_t start, uint32_t end,
+ bool is_signed, uint32_t fract_bits)
+{
+ __gen_validate_value(v);
+
+ const float factor = (1 << fract_bits);
+
+ float max, min;
+ if (is_signed) {
+ max = ((1 << (end - start)) - 1) / factor;
+ min = -(1 << (end - start)) / factor;
+ } else {
+ max = ((1 << (end - start + 1)) - 1) / factor;
+ min = 0.0f;
+ }
+
+ if (v > max)
+ v = max;
+ else if (v < min)
+ v = min;
+
+ int32_t int_val = roundf(v * factor);
+
+ if (is_signed)
+ int_val &= (1 << (end - start + 1)) - 1;
+
+ return int_val << start;
+}
+
+static inline uint64_t
+__gen_offset(uint64_t v, uint32_t start, uint32_t end)
+{
+ __gen_validate_value(v);
+#if DEBUG
+ uint64_t mask = (~0ul >> (64 - (end - start + 1))) << start;
+
+ assert((v & ~mask) == 0);
+#endif
+
+ return v;
+}
+
+static inline uint32_t
+__gen_float(float v)
+{
+ __gen_validate_value(v);
+ return ((union __gen_value) { .f = (v) }).dw;
+}
+
+#ifndef __gen_address_type
+#error #define __gen_address_type before including this file
+#endif
+
+#ifndef __gen_user_data
+#error #define __gen_combine_address before including this file
+#endif
+
+#endif
+
+#define GEN7_3DSTATE_URB_VS_length_bias 0x00000002
+#define GEN7_3DSTATE_URB_VS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 48, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_URB_VS_length 0x00000002
+
+struct GEN7_3DSTATE_URB_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t VSURBStartingAddress;
+ uint32_t VSURBEntryAllocationSize;
+ uint32_t VSNumberofURBEntries;
+};
+
+static inline void
+GEN7_3DSTATE_URB_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_URB_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->VSURBStartingAddress, 25, 29) |
+ __gen_field(values->VSURBEntryAllocationSize, 16, 24) |
+ __gen_field(values->VSNumberofURBEntries, 0, 15) |
+ 0;
+
+}
+
+#define GEN7_MI_STORE_REGISTER_MEM_length_bias 0x00000002
+#define GEN7_MI_STORE_REGISTER_MEM_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 36, \
+ .DwordLength = 1
+
+#define GEN7_MI_STORE_REGISTER_MEM_length 0x00000003
+
+struct GEN7_MI_STORE_REGISTER_MEM {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool UseGlobalGTT;
+ uint32_t DwordLength;
+ uint32_t RegisterAddress;
+ __gen_address_type MemoryAddress;
+};
+
+static inline void
+GEN7_MI_STORE_REGISTER_MEM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MI_STORE_REGISTER_MEM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->UseGlobalGTT, 22, 22) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->RegisterAddress, 2, 22) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->MemoryAddress, dw2);
+
+}
+
+#define GEN7_PIPELINE_SELECT_length_bias 0x00000001
+#define GEN7_PIPELINE_SELECT_header \
+ .CommandType = 3, \
+ .CommandSubType = 1, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 4
+
+#define GEN7_PIPELINE_SELECT_length 0x00000001
+
+struct GEN7_PIPELINE_SELECT {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+#define _3D 0
+#define Media 1
+#define GPGPU 2
+ uint32_t PipelineSelection;
+};
+
+static inline void
+GEN7_PIPELINE_SELECT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_PIPELINE_SELECT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->PipelineSelection, 0, 1) |
+ 0;
+
+}
+
+#define GEN7_STATE_BASE_ADDRESS_length_bias 0x00000002
+#define GEN7_STATE_BASE_ADDRESS_header \
+ .CommandType = 3, \
+ .CommandSubType = 0, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 1, \
+ .DwordLength = 8
+
+#define GEN7_STATE_BASE_ADDRESS_length 0x0000000a
+
+#define GEN7_MEMORY_OBJECT_CONTROL_STATE_length 0x00000001
+
+struct GEN7_MEMORY_OBJECT_CONTROL_STATE {
+ uint32_t GraphicsDataTypeGFDT;
+ uint32_t LLCCacheabilityControlLLCCC;
+ uint32_t L3CacheabilityControlL3CC;
+};
+
+static inline void
+GEN7_MEMORY_OBJECT_CONTROL_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MEMORY_OBJECT_CONTROL_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->GraphicsDataTypeGFDT, 2, 2) |
+ __gen_field(values->LLCCacheabilityControlLLCCC, 1, 1) |
+ __gen_field(values->L3CacheabilityControlL3CC, 0, 0) |
+ 0;
+
+}
+
+struct GEN7_STATE_BASE_ADDRESS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ __gen_address_type GeneralStateBaseAddress;
+ struct GEN7_MEMORY_OBJECT_CONTROL_STATE GeneralStateMemoryObjectControlState;
+ struct GEN7_MEMORY_OBJECT_CONTROL_STATE StatelessDataPortAccessMemoryObjectControlState;
+ uint32_t StatelessDataPortAccessForceWriteThru;
+ bool GeneralStateBaseAddressModifyEnable;
+ __gen_address_type SurfaceStateBaseAddress;
+ struct GEN7_MEMORY_OBJECT_CONTROL_STATE SurfaceStateMemoryObjectControlState;
+ bool SurfaceStateBaseAddressModifyEnable;
+ __gen_address_type DynamicStateBaseAddress;
+ struct GEN7_MEMORY_OBJECT_CONTROL_STATE DynamicStateMemoryObjectControlState;
+ bool DynamicStateBaseAddressModifyEnable;
+ __gen_address_type IndirectObjectBaseAddress;
+ struct GEN7_MEMORY_OBJECT_CONTROL_STATE IndirectObjectMemoryObjectControlState;
+ bool IndirectObjectBaseAddressModifyEnable;
+ __gen_address_type InstructionBaseAddress;
+ struct GEN7_MEMORY_OBJECT_CONTROL_STATE InstructionMemoryObjectControlState;
+ bool InstructionBaseAddressModifyEnable;
+ __gen_address_type GeneralStateAccessUpperBound;
+ bool GeneralStateAccessUpperBoundModifyEnable;
+ __gen_address_type DynamicStateAccessUpperBound;
+ bool DynamicStateAccessUpperBoundModifyEnable;
+ __gen_address_type IndirectObjectAccessUpperBound;
+ bool IndirectObjectAccessUpperBoundModifyEnable;
+ __gen_address_type InstructionAccessUpperBound;
+ bool InstructionAccessUpperBoundModifyEnable;
+};
+
+static inline void
+GEN7_STATE_BASE_ADDRESS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_STATE_BASE_ADDRESS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw_GeneralStateMemoryObjectControlState;
+ GEN7_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_GeneralStateMemoryObjectControlState, &values->GeneralStateMemoryObjectControlState);
+ uint32_t dw_StatelessDataPortAccessMemoryObjectControlState;
+ GEN7_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_StatelessDataPortAccessMemoryObjectControlState, &values->StatelessDataPortAccessMemoryObjectControlState);
+ uint32_t dw1 =
+ __gen_field(dw_GeneralStateMemoryObjectControlState, 8, 11) |
+ __gen_field(dw_StatelessDataPortAccessMemoryObjectControlState, 4, 7) |
+ __gen_field(values->StatelessDataPortAccessForceWriteThru, 3, 3) |
+ __gen_field(values->GeneralStateBaseAddressModifyEnable, 0, 0) |
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->GeneralStateBaseAddress, dw1);
+
+ uint32_t dw_SurfaceStateMemoryObjectControlState;
+ GEN7_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_SurfaceStateMemoryObjectControlState, &values->SurfaceStateMemoryObjectControlState);
+ uint32_t dw2 =
+ __gen_field(dw_SurfaceStateMemoryObjectControlState, 8, 11) |
+ __gen_field(values->SurfaceStateBaseAddressModifyEnable, 0, 0) |
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->SurfaceStateBaseAddress, dw2);
+
+ uint32_t dw_DynamicStateMemoryObjectControlState;
+ GEN7_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_DynamicStateMemoryObjectControlState, &values->DynamicStateMemoryObjectControlState);
+ uint32_t dw3 =
+ __gen_field(dw_DynamicStateMemoryObjectControlState, 8, 11) |
+ __gen_field(values->DynamicStateBaseAddressModifyEnable, 0, 0) |
+ 0;
+
+ dw[3] =
+ __gen_combine_address(data, &dw[3], values->DynamicStateBaseAddress, dw3);
+
+ uint32_t dw_IndirectObjectMemoryObjectControlState;
+ GEN7_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_IndirectObjectMemoryObjectControlState, &values->IndirectObjectMemoryObjectControlState);
+ uint32_t dw4 =
+ __gen_field(dw_IndirectObjectMemoryObjectControlState, 8, 11) |
+ __gen_field(values->IndirectObjectBaseAddressModifyEnable, 0, 0) |
+ 0;
+
+ dw[4] =
+ __gen_combine_address(data, &dw[4], values->IndirectObjectBaseAddress, dw4);
+
+ uint32_t dw_InstructionMemoryObjectControlState;
+ GEN7_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_InstructionMemoryObjectControlState, &values->InstructionMemoryObjectControlState);
+ uint32_t dw5 =
+ __gen_field(dw_InstructionMemoryObjectControlState, 8, 11) |
+ __gen_field(values->InstructionBaseAddressModifyEnable, 0, 0) |
+ 0;
+
+ dw[5] =
+ __gen_combine_address(data, &dw[5], values->InstructionBaseAddress, dw5);
+
+ uint32_t dw6 =
+ __gen_field(values->GeneralStateAccessUpperBoundModifyEnable, 0, 0) |
+ 0;
+
+ dw[6] =
+ __gen_combine_address(data, &dw[6], values->GeneralStateAccessUpperBound, dw6);
+
+ uint32_t dw7 =
+ __gen_field(values->DynamicStateAccessUpperBoundModifyEnable, 0, 0) |
+ 0;
+
+ dw[7] =
+ __gen_combine_address(data, &dw[7], values->DynamicStateAccessUpperBound, dw7);
+
+ uint32_t dw8 =
+ __gen_field(values->IndirectObjectAccessUpperBoundModifyEnable, 0, 0) |
+ 0;
+
+ dw[8] =
+ __gen_combine_address(data, &dw[8], values->IndirectObjectAccessUpperBound, dw8);
+
+ uint32_t dw9 =
+ __gen_field(values->InstructionAccessUpperBoundModifyEnable, 0, 0) |
+ 0;
+
+ dw[9] =
+ __gen_combine_address(data, &dw[9], values->InstructionAccessUpperBound, dw9);
+
+}
+
+#define GEN7_STATE_PREFETCH_length_bias 0x00000002
+#define GEN7_STATE_PREFETCH_header \
+ .CommandType = 3, \
+ .CommandSubType = 0, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 3, \
+ .DwordLength = 0
+
+#define GEN7_STATE_PREFETCH_length 0x00000002
+
+struct GEN7_STATE_PREFETCH {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ __gen_address_type PrefetchPointer;
+ uint32_t PrefetchCount;
+};
+
+static inline void
+GEN7_STATE_PREFETCH_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_STATE_PREFETCH * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw1 =
+ __gen_field(values->PrefetchCount, 0, 2) |
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->PrefetchPointer, dw1);
+
+}
+
+#define GEN7_STATE_SIP_length_bias 0x00000002
+#define GEN7_STATE_SIP_header \
+ .CommandType = 3, \
+ .CommandSubType = 0, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 2, \
+ .DwordLength = 0
+
+#define GEN7_STATE_SIP_length 0x00000002
+
+struct GEN7_STATE_SIP {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t SystemInstructionPointer;
+};
+
+static inline void
+GEN7_STATE_SIP_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_STATE_SIP * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->SystemInstructionPointer, 4, 31) |
+ 0;
+
+}
+
+#define GEN7_SWTESS_BASE_ADDRESS_length_bias 0x00000002
+#define GEN7_SWTESS_BASE_ADDRESS_header \
+ .CommandType = 3, \
+ .CommandSubType = 0, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 3, \
+ .DwordLength = 0
+
+#define GEN7_SWTESS_BASE_ADDRESS_length 0x00000002
+
+struct GEN7_SWTESS_BASE_ADDRESS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ __gen_address_type SWTessellationBaseAddress;
+ struct GEN7_MEMORY_OBJECT_CONTROL_STATE SWTessellationMemoryObjectControlState;
+};
+
+static inline void
+GEN7_SWTESS_BASE_ADDRESS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_SWTESS_BASE_ADDRESS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw_SWTessellationMemoryObjectControlState;
+ GEN7_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_SWTessellationMemoryObjectControlState, &values->SWTessellationMemoryObjectControlState);
+ uint32_t dw1 =
+ __gen_field(dw_SWTessellationMemoryObjectControlState, 8, 11) |
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->SWTessellationBaseAddress, dw1);
+
+}
+
+#define GEN7_3DPRIMITIVE_length_bias 0x00000002
+#define GEN7_3DPRIMITIVE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 3, \
+ ._3DCommandSubOpcode = 0, \
+ .DwordLength = 5
+
+#define GEN7_3DPRIMITIVE_length 0x00000007
+
+struct GEN7_3DPRIMITIVE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ bool IndirectParameterEnable;
+ bool PredicateEnable;
+ uint32_t DwordLength;
+ bool EndOffsetEnable;
+#define SEQUENTIAL 0
+#define RANDOM 1
+ uint32_t VertexAccessType;
+ uint32_t PrimitiveTopologyType;
+ uint32_t VertexCountPerInstance;
+ uint32_t StartVertexLocation;
+ uint32_t InstanceCount;
+ uint32_t StartInstanceLocation;
+ uint32_t BaseVertexLocation;
+};
+
+static inline void
+GEN7_3DPRIMITIVE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DPRIMITIVE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->IndirectParameterEnable, 10, 10) |
+ __gen_field(values->PredicateEnable, 8, 8) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->EndOffsetEnable, 9, 9) |
+ __gen_field(values->VertexAccessType, 8, 8) |
+ __gen_field(values->PrimitiveTopologyType, 0, 5) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->VertexCountPerInstance, 0, 31) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->StartVertexLocation, 0, 31) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->InstanceCount, 0, 31) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->StartInstanceLocation, 0, 31) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->BaseVertexLocation, 0, 31) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_AA_LINE_PARAMETERS_length_bias 0x00000002
+#define GEN7_3DSTATE_AA_LINE_PARAMETERS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 10, \
+ .DwordLength = 1
+
+#define GEN7_3DSTATE_AA_LINE_PARAMETERS_length 0x00000003
+
+struct GEN7_3DSTATE_AA_LINE_PARAMETERS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ float AACoverageBias;
+ float AACoverageSlope;
+ float AACoverageEndCapBias;
+ float AACoverageEndCapSlope;
+};
+
+static inline void
+GEN7_3DSTATE_AA_LINE_PARAMETERS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_AA_LINE_PARAMETERS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->AACoverageBias * (1 << 8), 16, 23) |
+ __gen_field(values->AACoverageSlope * (1 << 8), 0, 7) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->AACoverageEndCapBias * (1 << 8), 16, 23) |
+ __gen_field(values->AACoverageEndCapSlope * (1 << 8), 0, 7) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_BINDING_TABLE_POINTERS_DS_length_bias 0x00000002
+#define GEN7_3DSTATE_BINDING_TABLE_POINTERS_DS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 40, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_BINDING_TABLE_POINTERS_DS_length 0x00000002
+
+struct GEN7_3DSTATE_BINDING_TABLE_POINTERS_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoDSBindingTable;
+};
+
+static inline void
+GEN7_3DSTATE_BINDING_TABLE_POINTERS_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_BINDING_TABLE_POINTERS_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoDSBindingTable, 5, 15) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_BINDING_TABLE_POINTERS_GS_length_bias 0x00000002
+#define GEN7_3DSTATE_BINDING_TABLE_POINTERS_GS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 41, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_BINDING_TABLE_POINTERS_GS_length 0x00000002
+
+struct GEN7_3DSTATE_BINDING_TABLE_POINTERS_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoGSBindingTable;
+};
+
+static inline void
+GEN7_3DSTATE_BINDING_TABLE_POINTERS_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_BINDING_TABLE_POINTERS_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoGSBindingTable, 5, 15) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_BINDING_TABLE_POINTERS_HS_length_bias 0x00000002
+#define GEN7_3DSTATE_BINDING_TABLE_POINTERS_HS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 39, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_BINDING_TABLE_POINTERS_HS_length 0x00000002
+
+struct GEN7_3DSTATE_BINDING_TABLE_POINTERS_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoHSBindingTable;
+};
+
+static inline void
+GEN7_3DSTATE_BINDING_TABLE_POINTERS_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_BINDING_TABLE_POINTERS_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoHSBindingTable, 5, 15) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_BINDING_TABLE_POINTERS_PS_length_bias 0x00000002
+#define GEN7_3DSTATE_BINDING_TABLE_POINTERS_PS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 42, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_BINDING_TABLE_POINTERS_PS_length 0x00000002
+
+struct GEN7_3DSTATE_BINDING_TABLE_POINTERS_PS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoPSBindingTable;
+};
+
+static inline void
+GEN7_3DSTATE_BINDING_TABLE_POINTERS_PS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_BINDING_TABLE_POINTERS_PS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoPSBindingTable, 5, 15) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_BINDING_TABLE_POINTERS_VS_length_bias 0x00000002
+#define GEN7_3DSTATE_BINDING_TABLE_POINTERS_VS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 38, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_BINDING_TABLE_POINTERS_VS_length 0x00000002
+
+struct GEN7_3DSTATE_BINDING_TABLE_POINTERS_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoVSBindingTable;
+};
+
+static inline void
+GEN7_3DSTATE_BINDING_TABLE_POINTERS_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_BINDING_TABLE_POINTERS_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoVSBindingTable, 5, 15) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_BLEND_STATE_POINTERS_length_bias 0x00000002
+#define GEN7_3DSTATE_BLEND_STATE_POINTERS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 36, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_BLEND_STATE_POINTERS_length 0x00000002
+
+struct GEN7_3DSTATE_BLEND_STATE_POINTERS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t BlendStatePointer;
+};
+
+static inline void
+GEN7_3DSTATE_BLEND_STATE_POINTERS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_BLEND_STATE_POINTERS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->BlendStatePointer, 6, 31) |
+ __gen_mbo(0, 0) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_CC_STATE_POINTERS_length_bias 0x00000002
+#define GEN7_3DSTATE_CC_STATE_POINTERS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 14, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_CC_STATE_POINTERS_length 0x00000002
+
+struct GEN7_3DSTATE_CC_STATE_POINTERS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ColorCalcStatePointer;
+};
+
+static inline void
+GEN7_3DSTATE_CC_STATE_POINTERS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_CC_STATE_POINTERS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->ColorCalcStatePointer, 6, 31) |
+ __gen_mbo(0, 0) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_CHROMA_KEY_length_bias 0x00000002
+#define GEN7_3DSTATE_CHROMA_KEY_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 4, \
+ .DwordLength = 2
+
+#define GEN7_3DSTATE_CHROMA_KEY_length 0x00000004
+
+struct GEN7_3DSTATE_CHROMA_KEY {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ChromaKeyTableIndex;
+ uint32_t ChromaKeyLowValue;
+ uint32_t ChromaKeyHighValue;
+};
+
+static inline void
+GEN7_3DSTATE_CHROMA_KEY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_CHROMA_KEY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ChromaKeyTableIndex, 30, 31) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ChromaKeyLowValue, 0, 31) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->ChromaKeyHighValue, 0, 31) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_CLEAR_PARAMS_length_bias 0x00000002
+#define GEN7_3DSTATE_CLEAR_PARAMS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 4, \
+ .DwordLength = 1
+
+#define GEN7_3DSTATE_CLEAR_PARAMS_length 0x00000003
+
+struct GEN7_3DSTATE_CLEAR_PARAMS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t DepthClearValue;
+ bool DepthClearValueValid;
+};
+
+static inline void
+GEN7_3DSTATE_CLEAR_PARAMS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_CLEAR_PARAMS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->DepthClearValue, 0, 31) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->DepthClearValueValid, 0, 0) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_CLIP_length_bias 0x00000002
+#define GEN7_3DSTATE_CLIP_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 18, \
+ .DwordLength = 2
+
+#define GEN7_3DSTATE_CLIP_length 0x00000004
+
+struct GEN7_3DSTATE_CLIP {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t FrontWinding;
+ uint32_t VertexSubPixelPrecisionSelect;
+ bool EarlyCullEnable;
+#define CULLMODE_BOTH 0
+#define CULLMODE_NONE 1
+#define CULLMODE_FRONT 2
+#define CULLMODE_BACK 3
+ uint32_t CullMode;
+ bool ClipperStatisticsEnable;
+ uint32_t UserClipDistanceCullTestEnableBitmask;
+ bool ClipEnable;
+#define APIMODE_OGL 0
+ uint32_t APIMode;
+ bool ViewportXYClipTestEnable;
+ bool ViewportZClipTestEnable;
+ bool GuardbandClipTestEnable;
+ uint32_t UserClipDistanceClipTestEnableBitmask;
+#define CLIPMODE_NORMAL 0
+#define CLIPMODE_REJECT_ALL 3
+#define CLIPMODE_ACCEPT_ALL 4
+ uint32_t ClipMode;
+ bool PerspectiveDivideDisable;
+ bool NonPerspectiveBarycentricEnable;
+#define Vertex0 0
+#define Vertex1 1
+#define Vertex2 2
+ uint32_t TriangleStripListProvokingVertexSelect;
+#define Vertex0 0
+#define Vertex1 1
+ uint32_t LineStripListProvokingVertexSelect;
+#define Vertex0 0
+#define Vertex1 1
+#define Vertex2 2
+ uint32_t TriangleFanProvokingVertexSelect;
+ float MinimumPointWidth;
+ float MaximumPointWidth;
+ bool ForceZeroRTAIndexEnable;
+ uint32_t MaximumVPIndex;
+};
+
+static inline void
+GEN7_3DSTATE_CLIP_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_CLIP * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->FrontWinding, 20, 20) |
+ __gen_field(values->VertexSubPixelPrecisionSelect, 19, 19) |
+ __gen_field(values->EarlyCullEnable, 18, 18) |
+ __gen_field(values->CullMode, 16, 17) |
+ __gen_field(values->ClipperStatisticsEnable, 10, 10) |
+ __gen_field(values->UserClipDistanceCullTestEnableBitmask, 0, 7) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ClipEnable, 31, 31) |
+ __gen_field(values->APIMode, 30, 30) |
+ __gen_field(values->ViewportXYClipTestEnable, 28, 28) |
+ __gen_field(values->ViewportZClipTestEnable, 27, 27) |
+ __gen_field(values->GuardbandClipTestEnable, 26, 26) |
+ __gen_field(values->UserClipDistanceClipTestEnableBitmask, 16, 23) |
+ __gen_field(values->ClipMode, 13, 15) |
+ __gen_field(values->PerspectiveDivideDisable, 9, 9) |
+ __gen_field(values->NonPerspectiveBarycentricEnable, 8, 8) |
+ __gen_field(values->TriangleStripListProvokingVertexSelect, 4, 5) |
+ __gen_field(values->LineStripListProvokingVertexSelect, 2, 3) |
+ __gen_field(values->TriangleFanProvokingVertexSelect, 0, 1) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->MinimumPointWidth * (1 << 3), 17, 27) |
+ __gen_field(values->MaximumPointWidth * (1 << 3), 6, 16) |
+ __gen_field(values->ForceZeroRTAIndexEnable, 5, 5) |
+ __gen_field(values->MaximumVPIndex, 0, 3) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_CONSTANT_DS_length_bias 0x00000002
+#define GEN7_3DSTATE_CONSTANT_DS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 26, \
+ .DwordLength = 5
+
+#define GEN7_3DSTATE_CONSTANT_DS_length 0x00000007
+
+#define GEN7_3DSTATE_CONSTANT_BODY_length 0x00000006
+
+struct GEN7_3DSTATE_CONSTANT_BODY {
+ uint32_t ConstantBuffer1ReadLength;
+ uint32_t ConstantBuffer0ReadLength;
+ uint32_t ConstantBuffer3ReadLength;
+ uint32_t ConstantBuffer2ReadLength;
+ __gen_address_type PointerToConstantBuffer0;
+ struct GEN7_MEMORY_OBJECT_CONTROL_STATE ConstantBufferObjectControlState;
+ __gen_address_type PointerToConstantBuffer1;
+ __gen_address_type PointerToConstantBuffer2;
+ __gen_address_type PointerToConstantBuffer3;
+};
+
+static inline void
+GEN7_3DSTATE_CONSTANT_BODY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_CONSTANT_BODY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->ConstantBuffer1ReadLength, 16, 31) |
+ __gen_field(values->ConstantBuffer0ReadLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBuffer3ReadLength, 16, 31) |
+ __gen_field(values->ConstantBuffer2ReadLength, 0, 15) |
+ 0;
+
+ uint32_t dw_ConstantBufferObjectControlState;
+ GEN7_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_ConstantBufferObjectControlState, &values->ConstantBufferObjectControlState);
+ uint32_t dw2 =
+ __gen_field(dw_ConstantBufferObjectControlState, 0, 4) |
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->PointerToConstantBuffer0, dw2);
+
+ uint32_t dw3 =
+ 0;
+
+ dw[3] =
+ __gen_combine_address(data, &dw[3], values->PointerToConstantBuffer1, dw3);
+
+ uint32_t dw4 =
+ 0;
+
+ dw[4] =
+ __gen_combine_address(data, &dw[4], values->PointerToConstantBuffer2, dw4);
+
+ uint32_t dw5 =
+ 0;
+
+ dw[5] =
+ __gen_combine_address(data, &dw[5], values->PointerToConstantBuffer3, dw5);
+
+}
+
+struct GEN7_3DSTATE_CONSTANT_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ struct GEN7_3DSTATE_CONSTANT_BODY ConstantBody;
+};
+
+static inline void
+GEN7_3DSTATE_CONSTANT_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_CONSTANT_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ GEN7_3DSTATE_CONSTANT_BODY_pack(data, &dw[1], &values->ConstantBody);
+}
+
+#define GEN7_3DSTATE_CONSTANT_GS_length_bias 0x00000002
+#define GEN7_3DSTATE_CONSTANT_GS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 22, \
+ .DwordLength = 5
+
+#define GEN7_3DSTATE_CONSTANT_GS_length 0x00000007
+
+struct GEN7_3DSTATE_CONSTANT_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ struct GEN7_3DSTATE_CONSTANT_BODY ConstantBody;
+};
+
+static inline void
+GEN7_3DSTATE_CONSTANT_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_CONSTANT_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ GEN7_3DSTATE_CONSTANT_BODY_pack(data, &dw[1], &values->ConstantBody);
+}
+
+#define GEN7_3DSTATE_CONSTANT_HS_length_bias 0x00000002
+#define GEN7_3DSTATE_CONSTANT_HS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 25, \
+ .DwordLength = 5
+
+#define GEN7_3DSTATE_CONSTANT_HS_length 0x00000007
+
+struct GEN7_3DSTATE_CONSTANT_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ struct GEN7_3DSTATE_CONSTANT_BODY ConstantBody;
+};
+
+static inline void
+GEN7_3DSTATE_CONSTANT_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_CONSTANT_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ GEN7_3DSTATE_CONSTANT_BODY_pack(data, &dw[1], &values->ConstantBody);
+}
+
+#define GEN7_3DSTATE_CONSTANT_PS_length_bias 0x00000002
+#define GEN7_3DSTATE_CONSTANT_PS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 23, \
+ .DwordLength = 5
+
+#define GEN7_3DSTATE_CONSTANT_PS_length 0x00000007
+
+struct GEN7_3DSTATE_CONSTANT_PS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ struct GEN7_3DSTATE_CONSTANT_BODY ConstantBody;
+};
+
+static inline void
+GEN7_3DSTATE_CONSTANT_PS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_CONSTANT_PS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ GEN7_3DSTATE_CONSTANT_BODY_pack(data, &dw[1], &values->ConstantBody);
+}
+
+#define GEN7_3DSTATE_CONSTANT_VS_length_bias 0x00000002
+#define GEN7_3DSTATE_CONSTANT_VS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 21, \
+ .DwordLength = 5
+
+#define GEN7_3DSTATE_CONSTANT_VS_length 0x00000007
+
+struct GEN7_3DSTATE_CONSTANT_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ struct GEN7_3DSTATE_CONSTANT_BODY ConstantBody;
+};
+
+static inline void
+GEN7_3DSTATE_CONSTANT_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_CONSTANT_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ GEN7_3DSTATE_CONSTANT_BODY_pack(data, &dw[1], &values->ConstantBody);
+}
+
+#define GEN7_3DSTATE_DEPTH_BUFFER_length_bias 0x00000002
+#define GEN7_3DSTATE_DEPTH_BUFFER_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 5, \
+ .DwordLength = 5
+
+#define GEN7_3DSTATE_DEPTH_BUFFER_length 0x00000007
+
+struct GEN7_3DSTATE_DEPTH_BUFFER {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define SURFTYPE_1D 0
+#define SURFTYPE_2D 1
+#define SURFTYPE_3D 2
+#define SURFTYPE_CUBE 3
+#define SURFTYPE_NULL 7
+ uint32_t SurfaceType;
+ bool DepthWriteEnable;
+ bool StencilWriteEnable;
+ bool HierarchicalDepthBufferEnable;
+#define D32_FLOAT 1
+#define D24_UNORM_X8_UINT 3
+#define D16_UNORM 5
+ uint32_t SurfaceFormat;
+ uint32_t SurfacePitch;
+ __gen_address_type SurfaceBaseAddress;
+ uint32_t Height;
+ uint32_t Width;
+ uint32_t LOD;
+#define SURFTYPE_CUBEmustbezero 0
+ uint32_t Depth;
+ uint32_t MinimumArrayElement;
+ struct GEN7_MEMORY_OBJECT_CONTROL_STATE DepthBufferObjectControlState;
+ uint32_t DepthCoordinateOffsetY;
+ uint32_t DepthCoordinateOffsetX;
+ uint32_t RenderTargetViewExtent;
+};
+
+static inline void
+GEN7_3DSTATE_DEPTH_BUFFER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_DEPTH_BUFFER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->SurfaceType, 29, 31) |
+ __gen_field(values->DepthWriteEnable, 28, 28) |
+ __gen_field(values->StencilWriteEnable, 27, 27) |
+ __gen_field(values->HierarchicalDepthBufferEnable, 22, 22) |
+ __gen_field(values->SurfaceFormat, 18, 20) |
+ __gen_field(values->SurfacePitch, 0, 17) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->SurfaceBaseAddress, dw2);
+
+ dw[3] =
+ __gen_field(values->Height, 18, 31) |
+ __gen_field(values->Width, 4, 17) |
+ __gen_field(values->LOD, 0, 3) |
+ 0;
+
+ uint32_t dw_DepthBufferObjectControlState;
+ GEN7_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_DepthBufferObjectControlState, &values->DepthBufferObjectControlState);
+ dw[4] =
+ __gen_field(values->Depth, 21, 31) |
+ __gen_field(values->MinimumArrayElement, 10, 20) |
+ __gen_field(dw_DepthBufferObjectControlState, 0, 3) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->DepthCoordinateOffsetY, 16, 31) |
+ __gen_field(values->DepthCoordinateOffsetX, 0, 15) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->RenderTargetViewExtent, 21, 31) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_DEPTH_STENCIL_STATE_POINTERS_length_bias 0x00000002
+#define GEN7_3DSTATE_DEPTH_STENCIL_STATE_POINTERS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 37, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_DEPTH_STENCIL_STATE_POINTERS_length 0x00000002
+
+struct GEN7_3DSTATE_DEPTH_STENCIL_STATE_POINTERS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoDEPTH_STENCIL_STATE;
+};
+
+static inline void
+GEN7_3DSTATE_DEPTH_STENCIL_STATE_POINTERS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_DEPTH_STENCIL_STATE_POINTERS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoDEPTH_STENCIL_STATE, 6, 31) |
+ __gen_mbo(0, 0) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_DRAWING_RECTANGLE_length_bias 0x00000002
+#define GEN7_3DSTATE_DRAWING_RECTANGLE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 0, \
+ .DwordLength = 2
+
+#define GEN7_3DSTATE_DRAWING_RECTANGLE_length 0x00000004
+
+struct GEN7_3DSTATE_DRAWING_RECTANGLE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ClippedDrawingRectangleYMin;
+ uint32_t ClippedDrawingRectangleXMin;
+ uint32_t ClippedDrawingRectangleYMax;
+ uint32_t ClippedDrawingRectangleXMax;
+ uint32_t DrawingRectangleOriginY;
+ uint32_t DrawingRectangleOriginX;
+};
+
+static inline void
+GEN7_3DSTATE_DRAWING_RECTANGLE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_DRAWING_RECTANGLE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ClippedDrawingRectangleYMin, 16, 31) |
+ __gen_field(values->ClippedDrawingRectangleXMin, 0, 15) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ClippedDrawingRectangleYMax, 16, 31) |
+ __gen_field(values->ClippedDrawingRectangleXMax, 0, 15) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->DrawingRectangleOriginY, 16, 31) |
+ __gen_field(values->DrawingRectangleOriginX, 0, 15) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_DS_length_bias 0x00000002
+#define GEN7_3DSTATE_DS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 29, \
+ .DwordLength = 4
+
+#define GEN7_3DSTATE_DS_length 0x00000006
+
+struct GEN7_3DSTATE_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t KernelStartPointer;
+#define Multiple 0
+#define Single 1
+ uint32_t SingleDomainPointDispatch;
+#define Dmask 0
+#define Vmask 1
+ uint32_t VectorMaskEnable;
+#define NoSamplers 0
+#define _14Samplers 1
+#define _58Samplers 2
+#define _912Samplers 3
+#define _1316Samplers 4
+ uint32_t SamplerCount;
+ uint32_t BindingTableEntryCount;
+#define IEEE754 0
+#define Alternate 1
+ uint32_t FloatingPointMode;
+ bool IllegalOpcodeExceptionEnable;
+ bool SoftwareExceptionEnable;
+ uint32_t ScratchSpaceBasePointer;
+ uint32_t PerThreadScratchSpace;
+ uint32_t DispatchGRFStartRegisterForURBData;
+ uint32_t PatchURBEntryReadLength;
+ uint32_t PatchURBEntryReadOffset;
+ uint32_t MaximumNumberofThreads;
+ bool StatisticsEnable;
+ bool ComputeWCoordinateEnable;
+ bool DSCacheDisable;
+ bool DSFunctionEnable;
+};
+
+static inline void
+GEN7_3DSTATE_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->KernelStartPointer, 6, 31) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->SingleDomainPointDispatch, 31, 31) |
+ __gen_field(values->VectorMaskEnable, 30, 30) |
+ __gen_field(values->SamplerCount, 27, 29) |
+ __gen_field(values->BindingTableEntryCount, 18, 25) |
+ __gen_field(values->FloatingPointMode, 16, 16) |
+ __gen_field(values->IllegalOpcodeExceptionEnable, 13, 13) |
+ __gen_field(values->SoftwareExceptionEnable, 7, 7) |
+ 0;
+
+ dw[3] =
+ __gen_offset(values->ScratchSpaceBasePointer, 10, 31) |
+ __gen_field(values->PerThreadScratchSpace, 0, 3) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->DispatchGRFStartRegisterForURBData, 20, 24) |
+ __gen_field(values->PatchURBEntryReadLength, 11, 17) |
+ __gen_field(values->PatchURBEntryReadOffset, 4, 9) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->MaximumNumberofThreads, 25, 31) |
+ __gen_field(values->StatisticsEnable, 10, 10) |
+ __gen_field(values->ComputeWCoordinateEnable, 2, 2) |
+ __gen_field(values->DSCacheDisable, 1, 1) |
+ __gen_field(values->DSFunctionEnable, 0, 0) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_GS_length_bias 0x00000002
+#define GEN7_3DSTATE_GS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 17, \
+ .DwordLength = 5
+
+#define GEN7_3DSTATE_GS_length 0x00000007
+
+struct GEN7_3DSTATE_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t KernelStartPointer;
+ uint32_t SingleProgramFlowSPF;
+#define Dmask 0
+#define Vmask 1
+ uint32_t VectorMaskEnableVME;
+#define NoSamplers 0
+#define _14Samplers 1
+#define _58Samplers 2
+#define _912Samplers 3
+#define _1316Samplers 4
+ uint32_t SamplerCount;
+ uint32_t BindingTableEntryCount;
+#define NormalPriority 0
+#define HighPriority 1
+ uint32_t ThreadPriority;
+#define IEEE754 0
+#define alternate 1
+ uint32_t FloatingPointMode;
+ bool IllegalOpcodeExceptionEnable;
+ bool MaskStackExceptionEnable;
+ bool SoftwareExceptionEnable;
+ uint32_t ScratchSpaceBasePointer;
+ uint32_t PerThreadScratchSpace;
+ uint32_t OutputVertexSize;
+ uint32_t OutputTopology;
+ uint32_t VertexURBEntryReadLength;
+ bool IncludeVertexHandles;
+ uint32_t VertexURBEntryReadOffset;
+ uint32_t DispatchGRFStartRegisterforURBData;
+ uint32_t MaximumNumberofThreads;
+#define GSCTL_CUT 0
+#define GSCTL_SID 1
+ uint32_t ControlDataFormat;
+ uint32_t ControlDataHeaderSize;
+ uint32_t InstanceControl;
+ uint32_t DefaultStreamID;
+#define SINGLE 0
+#define DUAL_INSTANCE 1
+#define DUAL_OBJECT 2
+ uint32_t DispatchMode;
+ uint32_t GSStatisticsEnable;
+ uint32_t GSInvocationsIncrementValue;
+ bool IncludePrimitiveID;
+ uint32_t Hint;
+ bool ReorderEnable;
+ bool DiscardAdjacency;
+ bool GSEnable;
+ uint32_t SemaphoreHandle;
+};
+
+static inline void
+GEN7_3DSTATE_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->KernelStartPointer, 6, 31) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->SingleProgramFlowSPF, 31, 31) |
+ __gen_field(values->VectorMaskEnableVME, 30, 30) |
+ __gen_field(values->SamplerCount, 27, 29) |
+ __gen_field(values->BindingTableEntryCount, 18, 25) |
+ __gen_field(values->ThreadPriority, 17, 17) |
+ __gen_field(values->FloatingPointMode, 16, 16) |
+ __gen_field(values->IllegalOpcodeExceptionEnable, 13, 13) |
+ __gen_field(values->MaskStackExceptionEnable, 11, 11) |
+ __gen_field(values->SoftwareExceptionEnable, 7, 7) |
+ 0;
+
+ dw[3] =
+ __gen_offset(values->ScratchSpaceBasePointer, 10, 31) |
+ __gen_field(values->PerThreadScratchSpace, 0, 3) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->OutputVertexSize, 23, 28) |
+ __gen_field(values->OutputTopology, 17, 22) |
+ __gen_field(values->VertexURBEntryReadLength, 11, 16) |
+ __gen_field(values->IncludeVertexHandles, 10, 10) |
+ __gen_field(values->VertexURBEntryReadOffset, 4, 9) |
+ __gen_field(values->DispatchGRFStartRegisterforURBData, 0, 3) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->MaximumNumberofThreads, 25, 31) |
+ __gen_field(values->ControlDataFormat, 24, 24) |
+ __gen_field(values->ControlDataHeaderSize, 20, 23) |
+ __gen_field(values->InstanceControl, 15, 19) |
+ __gen_field(values->DefaultStreamID, 13, 14) |
+ __gen_field(values->DispatchMode, 11, 12) |
+ __gen_field(values->GSStatisticsEnable, 10, 10) |
+ __gen_field(values->GSInvocationsIncrementValue, 5, 9) |
+ __gen_field(values->IncludePrimitiveID, 4, 4) |
+ __gen_field(values->Hint, 3, 3) |
+ __gen_field(values->ReorderEnable, 2, 2) |
+ __gen_field(values->DiscardAdjacency, 1, 1) |
+ __gen_field(values->GSEnable, 0, 0) |
+ 0;
+
+ dw[6] =
+ __gen_offset(values->SemaphoreHandle, 0, 11) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_HIER_DEPTH_BUFFER_length_bias 0x00000002
+#define GEN7_3DSTATE_HIER_DEPTH_BUFFER_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 7, \
+ .DwordLength = 1
+
+#define GEN7_3DSTATE_HIER_DEPTH_BUFFER_length 0x00000003
+
+struct GEN7_3DSTATE_HIER_DEPTH_BUFFER {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ struct GEN7_MEMORY_OBJECT_CONTROL_STATE HierarchicalDepthBufferObjectControlState;
+ uint32_t SurfacePitch;
+ __gen_address_type SurfaceBaseAddress;
+};
+
+static inline void
+GEN7_3DSTATE_HIER_DEPTH_BUFFER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_HIER_DEPTH_BUFFER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw_HierarchicalDepthBufferObjectControlState;
+ GEN7_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_HierarchicalDepthBufferObjectControlState, &values->HierarchicalDepthBufferObjectControlState);
+ dw[1] =
+ __gen_field(dw_HierarchicalDepthBufferObjectControlState, 25, 28) |
+ __gen_field(values->SurfacePitch, 0, 16) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->SurfaceBaseAddress, dw2);
+
+}
+
+#define GEN7_3DSTATE_HS_length_bias 0x00000002
+#define GEN7_3DSTATE_HS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 27, \
+ .DwordLength = 5
+
+#define GEN7_3DSTATE_HS_length 0x00000007
+
+struct GEN7_3DSTATE_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define NoSamplers 0
+#define _14Samplers 1
+#define _58Samplers 2
+#define _912Samplers 3
+#define _1316Samplers 4
+ uint32_t SamplerCount;
+ uint32_t BindingTableEntryCount;
+#define IEEE754 0
+#define alternate 1
+ uint32_t FloatingPointMode;
+ bool IllegalOpcodeExceptionEnable;
+ bool SoftwareExceptionEnable;
+ uint32_t MaximumNumberofThreads;
+ bool Enable;
+ bool StatisticsEnable;
+ uint32_t InstanceCount;
+ uint32_t KernelStartPointer;
+ uint32_t ScratchSpaceBasePointer;
+ uint32_t PerThreadScratchSpace;
+ uint32_t SingleProgramFlow;
+#define Dmask 0
+#define Vmask 1
+ uint32_t VectorMaskEnable;
+ bool IncludeVertexHandles;
+ uint32_t DispatchGRFStartRegisterForURBData;
+ uint32_t VertexURBEntryReadLength;
+ uint32_t VertexURBEntryReadOffset;
+ uint32_t SemaphoreHandle;
+};
+
+static inline void
+GEN7_3DSTATE_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->SamplerCount, 27, 29) |
+ __gen_field(values->BindingTableEntryCount, 18, 25) |
+ __gen_field(values->FloatingPointMode, 16, 16) |
+ __gen_field(values->IllegalOpcodeExceptionEnable, 13, 13) |
+ __gen_field(values->SoftwareExceptionEnable, 7, 7) |
+ __gen_field(values->MaximumNumberofThreads, 0, 6) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->Enable, 31, 31) |
+ __gen_field(values->StatisticsEnable, 29, 29) |
+ __gen_field(values->InstanceCount, 0, 3) |
+ 0;
+
+ dw[3] =
+ __gen_offset(values->KernelStartPointer, 6, 31) |
+ 0;
+
+ dw[4] =
+ __gen_offset(values->ScratchSpaceBasePointer, 10, 31) |
+ __gen_field(values->PerThreadScratchSpace, 0, 3) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->SingleProgramFlow, 27, 27) |
+ __gen_field(values->VectorMaskEnable, 26, 26) |
+ __gen_field(values->IncludeVertexHandles, 24, 24) |
+ __gen_field(values->DispatchGRFStartRegisterForURBData, 19, 23) |
+ __gen_field(values->VertexURBEntryReadLength, 11, 16) |
+ __gen_field(values->VertexURBEntryReadOffset, 4, 9) |
+ 0;
+
+ dw[6] =
+ __gen_offset(values->SemaphoreHandle, 0, 11) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_INDEX_BUFFER_length_bias 0x00000002
+#define GEN7_3DSTATE_INDEX_BUFFER_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 10, \
+ .DwordLength = 1
+
+#define GEN7_3DSTATE_INDEX_BUFFER_length 0x00000003
+
+struct GEN7_3DSTATE_INDEX_BUFFER {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ struct GEN7_MEMORY_OBJECT_CONTROL_STATE MemoryObjectControlState;
+ bool CutIndexEnable;
+#define INDEX_BYTE 0
+#define INDEX_WORD 1
+#define INDEX_DWORD 2
+ uint32_t IndexFormat;
+ uint32_t DwordLength;
+ __gen_address_type BufferStartingAddress;
+ __gen_address_type BufferEndingAddress;
+};
+
+static inline void
+GEN7_3DSTATE_INDEX_BUFFER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_INDEX_BUFFER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ uint32_t dw_MemoryObjectControlState;
+ GEN7_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_MemoryObjectControlState, &values->MemoryObjectControlState);
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(dw_MemoryObjectControlState, 12, 15) |
+ __gen_field(values->CutIndexEnable, 10, 10) |
+ __gen_field(values->IndexFormat, 8, 9) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw1 =
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->BufferStartingAddress, dw1);
+
+ uint32_t dw2 =
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->BufferEndingAddress, dw2);
+
+}
+
+#define GEN7_3DSTATE_LINE_STIPPLE_length_bias 0x00000002
+#define GEN7_3DSTATE_LINE_STIPPLE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 8, \
+ .DwordLength = 1
+
+#define GEN7_3DSTATE_LINE_STIPPLE_length 0x00000003
+
+struct GEN7_3DSTATE_LINE_STIPPLE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool ModifyEnableCurrentRepeatCounterCurrentStippleIndex;
+ uint32_t CurrentRepeatCounter;
+ uint32_t CurrentStippleIndex;
+ uint32_t LineStipplePattern;
+ float LineStippleInverseRepeatCount;
+ uint32_t LineStippleRepeatCount;
+};
+
+static inline void
+GEN7_3DSTATE_LINE_STIPPLE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_LINE_STIPPLE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ModifyEnableCurrentRepeatCounterCurrentStippleIndex, 31, 31) |
+ __gen_field(values->CurrentRepeatCounter, 21, 29) |
+ __gen_field(values->CurrentStippleIndex, 16, 19) |
+ __gen_field(values->LineStipplePattern, 0, 15) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->LineStippleInverseRepeatCount * (1 << 16), 15, 31) |
+ __gen_field(values->LineStippleRepeatCount, 0, 8) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_MONOFILTER_SIZE_length_bias 0x00000002
+#define GEN7_3DSTATE_MONOFILTER_SIZE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 17, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_MONOFILTER_SIZE_length 0x00000002
+
+struct GEN7_3DSTATE_MONOFILTER_SIZE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t MonochromeFilterWidth;
+ uint32_t MonochromeFilterHeight;
+};
+
+static inline void
+GEN7_3DSTATE_MONOFILTER_SIZE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_MONOFILTER_SIZE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->MonochromeFilterWidth, 3, 5) |
+ __gen_field(values->MonochromeFilterHeight, 0, 2) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_MULTISAMPLE_length_bias 0x00000002
+#define GEN7_3DSTATE_MULTISAMPLE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 13, \
+ .DwordLength = 2
+
+#define GEN7_3DSTATE_MULTISAMPLE_length 0x00000004
+
+struct GEN7_3DSTATE_MULTISAMPLE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define PIXLOC_CENTER 0
+#define PIXLOC_UL_CORNER 1
+ uint32_t PixelLocation;
+#define NUMSAMPLES_1 0
+#define NUMSAMPLES_4 2
+#define NUMSAMPLES_8 3
+ uint32_t NumberofMultisamples;
+ float Sample3XOffset;
+ float Sample3YOffset;
+ float Sample2XOffset;
+ float Sample2YOffset;
+ float Sample1XOffset;
+ float Sample1YOffset;
+ float Sample0XOffset;
+ float Sample0YOffset;
+ float Sample7XOffset;
+ float Sample7YOffset;
+ float Sample6XOffset;
+ float Sample6YOffset;
+ float Sample5XOffset;
+ float Sample5YOffset;
+ float Sample4XOffset;
+ float Sample4YOffset;
+};
+
+static inline void
+GEN7_3DSTATE_MULTISAMPLE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_MULTISAMPLE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->PixelLocation, 4, 4) |
+ __gen_field(values->NumberofMultisamples, 1, 3) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->Sample3XOffset * (1 << 4), 28, 31) |
+ __gen_field(values->Sample3YOffset * (1 << 4), 24, 27) |
+ __gen_field(values->Sample2XOffset * (1 << 4), 20, 23) |
+ __gen_field(values->Sample2YOffset * (1 << 4), 16, 19) |
+ __gen_field(values->Sample1XOffset * (1 << 4), 12, 15) |
+ __gen_field(values->Sample1YOffset * (1 << 4), 8, 11) |
+ __gen_field(values->Sample0XOffset * (1 << 4), 4, 7) |
+ __gen_field(values->Sample0YOffset * (1 << 4), 0, 3) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->Sample7XOffset * (1 << 4), 28, 31) |
+ __gen_field(values->Sample7YOffset * (1 << 4), 24, 27) |
+ __gen_field(values->Sample6XOffset * (1 << 4), 20, 23) |
+ __gen_field(values->Sample6YOffset * (1 << 4), 16, 19) |
+ __gen_field(values->Sample5XOffset * (1 << 4), 12, 15) |
+ __gen_field(values->Sample5YOffset * (1 << 4), 8, 11) |
+ __gen_field(values->Sample4XOffset * (1 << 4), 4, 7) |
+ __gen_field(values->Sample4YOffset * (1 << 4), 0, 3) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_POLY_STIPPLE_OFFSET_length_bias 0x00000002
+#define GEN7_3DSTATE_POLY_STIPPLE_OFFSET_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 6, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_POLY_STIPPLE_OFFSET_length 0x00000002
+
+struct GEN7_3DSTATE_POLY_STIPPLE_OFFSET {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PolygonStippleXOffset;
+ uint32_t PolygonStippleYOffset;
+};
+
+static inline void
+GEN7_3DSTATE_POLY_STIPPLE_OFFSET_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_POLY_STIPPLE_OFFSET * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->PolygonStippleXOffset, 8, 12) |
+ __gen_field(values->PolygonStippleYOffset, 0, 4) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_POLY_STIPPLE_PATTERN_length_bias 0x00000002
+#define GEN7_3DSTATE_POLY_STIPPLE_PATTERN_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 7, \
+ .DwordLength = 31
+
+#define GEN7_3DSTATE_POLY_STIPPLE_PATTERN_length 0x00000021
+
+struct GEN7_3DSTATE_POLY_STIPPLE_PATTERN {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PatternRow[32];
+};
+
+static inline void
+GEN7_3DSTATE_POLY_STIPPLE_PATTERN_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_POLY_STIPPLE_PATTERN * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ for (uint32_t i = 0, j = 1; i < 32; i += 1, j++) {
+ dw[j] =
+ __gen_field(values->PatternRow[i + 0], 0, 31) |
+ 0;
+ }
+
+}
+
+#define GEN7_3DSTATE_PS_length_bias 0x00000002
+#define GEN7_3DSTATE_PS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 32, \
+ .DwordLength = 6
+
+#define GEN7_3DSTATE_PS_length 0x00000008
+
+struct GEN7_3DSTATE_PS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t KernelStartPointer0;
+#define Multiple 0
+#define Single 1
+ uint32_t SingleProgramFlowSPF;
+#define Dmask 0
+#define Vmask 1
+ uint32_t VectorMaskEnableVME;
+ uint32_t SamplerCount;
+#define FTZ 0
+#define RET 1
+ uint32_t DenormalMode;
+ uint32_t BindingTableEntryCount;
+#define IEEE745 0
+#define Alt 1
+ uint32_t FloatingPointMode;
+#define RTNE 0
+#define RU 1
+#define RD 2
+#define RTZ 3
+ uint32_t RoundingMode;
+ bool IllegalOpcodeExceptionEnable;
+ bool MaskStackExceptionEnable;
+ bool SoftwareExceptionEnable;
+ uint32_t ScratchSpaceBasePointer;
+ uint32_t PerThreadScratchSpace;
+ uint32_t MaximumNumberofThreads;
+ bool PushConstantEnable;
+ bool AttributeEnable;
+ bool oMaskPresenttoRenderTarget;
+ bool RenderTargetFastClearEnable;
+ bool DualSourceBlendEnable;
+ bool RenderTargetResolveEnable;
+#define POSOFFSET_NONE 0
+#define POSOFFSET_CENTROID 2
+#define POSOFFSET_SAMPLE 3
+ uint32_t PositionXYOffsetSelect;
+ bool _32PixelDispatchEnable;
+ bool _16PixelDispatchEnable;
+ bool _8PixelDispatchEnable;
+ uint32_t DispatchGRFStartRegisterforConstantSetupData0;
+ uint32_t DispatchGRFStartRegisterforConstantSetupData1;
+ uint32_t DispatchGRFStartRegisterforConstantSetupData2;
+ uint32_t KernelStartPointer1;
+ uint32_t KernelStartPointer2;
+};
+
+static inline void
+GEN7_3DSTATE_PS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_PS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->KernelStartPointer0, 6, 31) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->SingleProgramFlowSPF, 31, 31) |
+ __gen_field(values->VectorMaskEnableVME, 30, 30) |
+ __gen_field(values->SamplerCount, 27, 29) |
+ __gen_field(values->DenormalMode, 26, 26) |
+ __gen_field(values->BindingTableEntryCount, 18, 25) |
+ __gen_field(values->FloatingPointMode, 16, 16) |
+ __gen_field(values->RoundingMode, 14, 15) |
+ __gen_field(values->IllegalOpcodeExceptionEnable, 13, 13) |
+ __gen_field(values->MaskStackExceptionEnable, 11, 11) |
+ __gen_field(values->SoftwareExceptionEnable, 7, 7) |
+ 0;
+
+ dw[3] =
+ __gen_offset(values->ScratchSpaceBasePointer, 10, 31) |
+ __gen_field(values->PerThreadScratchSpace, 0, 3) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->MaximumNumberofThreads, 24, 31) |
+ __gen_field(values->PushConstantEnable, 11, 11) |
+ __gen_field(values->AttributeEnable, 10, 10) |
+ __gen_field(values->oMaskPresenttoRenderTarget, 9, 9) |
+ __gen_field(values->RenderTargetFastClearEnable, 8, 8) |
+ __gen_field(values->DualSourceBlendEnable, 7, 7) |
+ __gen_field(values->RenderTargetResolveEnable, 6, 6) |
+ __gen_field(values->PositionXYOffsetSelect, 3, 4) |
+ __gen_field(values->_32PixelDispatchEnable, 2, 2) |
+ __gen_field(values->_16PixelDispatchEnable, 1, 1) |
+ __gen_field(values->_8PixelDispatchEnable, 0, 0) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->DispatchGRFStartRegisterforConstantSetupData0, 16, 22) |
+ __gen_field(values->DispatchGRFStartRegisterforConstantSetupData1, 8, 14) |
+ __gen_field(values->DispatchGRFStartRegisterforConstantSetupData2, 0, 6) |
+ 0;
+
+ dw[6] =
+ __gen_offset(values->KernelStartPointer1, 6, 31) |
+ 0;
+
+ dw[7] =
+ __gen_offset(values->KernelStartPointer2, 6, 31) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_DS_length_bias 0x00000002
+#define GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_DS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 20, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_DS_length 0x00000002
+
+struct GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define _0KB 0
+ uint32_t ConstantBufferOffset;
+#define _0KB 0
+ uint32_t ConstantBufferSize;
+};
+
+static inline void
+GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferOffset, 16, 19) |
+ __gen_field(values->ConstantBufferSize, 0, 4) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_GS_length_bias 0x00000002
+#define GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_GS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 21, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_GS_length 0x00000002
+
+struct GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define _0KB 0
+ uint32_t ConstantBufferOffset;
+#define _0KB 0
+ uint32_t ConstantBufferSize;
+};
+
+static inline void
+GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferOffset, 16, 19) |
+ __gen_field(values->ConstantBufferSize, 0, 4) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_HS_length_bias 0x00000002
+#define GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_HS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 19, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_HS_length 0x00000002
+
+struct GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define _0KB 0
+ uint32_t ConstantBufferOffset;
+#define _0KB 0
+ uint32_t ConstantBufferSize;
+};
+
+static inline void
+GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferOffset, 16, 19) |
+ __gen_field(values->ConstantBufferSize, 0, 4) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_PS_length_bias 0x00000002
+#define GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_PS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 22, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_PS_length 0x00000002
+
+struct GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_PS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define _0KB 0
+ uint32_t ConstantBufferOffset;
+#define _0KB 0
+ uint32_t ConstantBufferSize;
+};
+
+static inline void
+GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_PS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_PS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferOffset, 16, 19) |
+ __gen_field(values->ConstantBufferSize, 0, 4) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_VS_length_bias 0x00000002
+#define GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_VS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 18, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_VS_length 0x00000002
+
+struct GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define _0KB 0
+ uint32_t ConstantBufferOffset;
+#define _0KB 0
+ uint32_t ConstantBufferSize;
+};
+
+static inline void
+GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferOffset, 16, 19) |
+ __gen_field(values->ConstantBufferSize, 0, 4) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_SAMPLER_PALETTE_LOAD0_length_bias 0x00000002
+#define GEN7_3DSTATE_SAMPLER_PALETTE_LOAD0_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 2
+
+#define GEN7_3DSTATE_SAMPLER_PALETTE_LOAD0_length 0x00000000
+
+#define GEN7_PALETTE_ENTRY_length 0x00000001
+
+struct GEN7_PALETTE_ENTRY {
+ uint32_t Alpha;
+ uint32_t Red;
+ uint32_t Green;
+ uint32_t Blue;
+};
+
+static inline void
+GEN7_PALETTE_ENTRY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_PALETTE_ENTRY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->Alpha, 24, 31) |
+ __gen_field(values->Red, 16, 23) |
+ __gen_field(values->Green, 8, 15) |
+ __gen_field(values->Blue, 0, 7) |
+ 0;
+
+}
+
+struct GEN7_3DSTATE_SAMPLER_PALETTE_LOAD0 {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN7_3DSTATE_SAMPLER_PALETTE_LOAD0_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_SAMPLER_PALETTE_LOAD0 * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN7_3DSTATE_SAMPLER_PALETTE_LOAD1_length_bias 0x00000002
+#define GEN7_3DSTATE_SAMPLER_PALETTE_LOAD1_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 12
+
+#define GEN7_3DSTATE_SAMPLER_PALETTE_LOAD1_length 0x00000000
+
+struct GEN7_3DSTATE_SAMPLER_PALETTE_LOAD1 {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN7_3DSTATE_SAMPLER_PALETTE_LOAD1_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_SAMPLER_PALETTE_LOAD1 * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN7_3DSTATE_SAMPLER_STATE_POINTERS_DS_length_bias 0x00000002
+#define GEN7_3DSTATE_SAMPLER_STATE_POINTERS_DS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 45, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_SAMPLER_STATE_POINTERS_DS_length 0x00000002
+
+struct GEN7_3DSTATE_SAMPLER_STATE_POINTERS_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoDSSamplerState;
+};
+
+static inline void
+GEN7_3DSTATE_SAMPLER_STATE_POINTERS_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_SAMPLER_STATE_POINTERS_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoDSSamplerState, 5, 31) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_SAMPLER_STATE_POINTERS_GS_length_bias 0x00000002
+#define GEN7_3DSTATE_SAMPLER_STATE_POINTERS_GS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 46, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_SAMPLER_STATE_POINTERS_GS_length 0x00000002
+
+struct GEN7_3DSTATE_SAMPLER_STATE_POINTERS_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoGSSamplerState;
+};
+
+static inline void
+GEN7_3DSTATE_SAMPLER_STATE_POINTERS_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_SAMPLER_STATE_POINTERS_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoGSSamplerState, 5, 31) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_SAMPLER_STATE_POINTERS_HS_length_bias 0x00000002
+#define GEN7_3DSTATE_SAMPLER_STATE_POINTERS_HS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 44, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_SAMPLER_STATE_POINTERS_HS_length 0x00000002
+
+struct GEN7_3DSTATE_SAMPLER_STATE_POINTERS_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoHSSamplerState;
+};
+
+static inline void
+GEN7_3DSTATE_SAMPLER_STATE_POINTERS_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_SAMPLER_STATE_POINTERS_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoHSSamplerState, 5, 31) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_SAMPLER_STATE_POINTERS_PS_length_bias 0x00000002
+#define GEN7_3DSTATE_SAMPLER_STATE_POINTERS_PS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 47, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_SAMPLER_STATE_POINTERS_PS_length 0x00000002
+
+struct GEN7_3DSTATE_SAMPLER_STATE_POINTERS_PS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoPSSamplerState;
+};
+
+static inline void
+GEN7_3DSTATE_SAMPLER_STATE_POINTERS_PS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_SAMPLER_STATE_POINTERS_PS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoPSSamplerState, 5, 31) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_SAMPLER_STATE_POINTERS_VS_length_bias 0x00000002
+#define GEN7_3DSTATE_SAMPLER_STATE_POINTERS_VS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 43, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_SAMPLER_STATE_POINTERS_VS_length 0x00000002
+
+struct GEN7_3DSTATE_SAMPLER_STATE_POINTERS_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoVSSamplerState;
+};
+
+static inline void
+GEN7_3DSTATE_SAMPLER_STATE_POINTERS_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_SAMPLER_STATE_POINTERS_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoVSSamplerState, 5, 31) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_SAMPLE_MASK_length_bias 0x00000002
+#define GEN7_3DSTATE_SAMPLE_MASK_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 24, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_SAMPLE_MASK_length 0x00000002
+
+struct GEN7_3DSTATE_SAMPLE_MASK {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t SampleMask;
+};
+
+static inline void
+GEN7_3DSTATE_SAMPLE_MASK_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_SAMPLE_MASK * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->SampleMask, 0, 7) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_SBE_length_bias 0x00000002
+#define GEN7_3DSTATE_SBE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 31, \
+ .DwordLength = 12
+
+#define GEN7_3DSTATE_SBE_length 0x0000000e
+
+struct GEN7_3DSTATE_SBE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define SWIZ_0_15 0
+#define SWIZ_16_31 1
+ uint32_t AttributeSwizzleControlMode;
+ uint32_t NumberofSFOutputAttributes;
+ bool AttributeSwizzleEnable;
+#define UPPERLEFT 0
+#define LOWERLEFT 1
+ uint32_t PointSpriteTextureCoordinateOrigin;
+ uint32_t VertexURBEntryReadLength;
+ uint32_t VertexURBEntryReadOffset;
+ bool Attribute2n1ComponentOverrideW;
+ bool Attribute2n1ComponentOverrideZ;
+ bool Attribute2n1ComponentOverrideY;
+ bool Attribute2n1ComponentOverrideX;
+#define CONST_0000 0
+#define CONST_0001_FLOAT 1
+#define CONST_1111_FLOAT 2
+#define PRIM_ID 3
+ uint32_t Attribute2n1ConstantSource;
+#define INPUTATTR 0
+#define INPUTATTR_FACING 1
+#define INPUTATTR_W 2
+#define INPUTATTR_FACING_W 3
+ uint32_t Attribute2n1SwizzleSelect;
+ uint32_t Attribute2n1SourceAttribute;
+ bool Attribute2nComponentOverrideW;
+ bool Attribute2nComponentOverrideZ;
+ bool Attribute2nComponentOverrideY;
+ bool Attribute2nComponentOverrideX;
+#define CONST_0000 0
+#define CONST_0001_FLOAT 1
+#define CONST_1111_FLOAT 2
+#define PRIM_ID 3
+ uint32_t Attribute2nConstantSource;
+#define INPUTATTR 0
+#define INPUTATTR_FACING 1
+#define INPUTATTR_W 2
+#define INPUTATTR_FACING_W 3
+ uint32_t Attribute2nSwizzleSelect;
+ uint32_t Attribute2nSourceAttribute;
+ uint32_t PointSpriteTextureCoordinateEnable;
+ uint32_t ConstantInterpolationEnable310;
+ uint32_t Attribute7WrapShortestEnables;
+ uint32_t Attribute6WrapShortestEnables;
+ uint32_t Attribute5WrapShortestEnables;
+ uint32_t Attribute4WrapShortestEnables;
+ uint32_t Attribute3WrapShortestEnables;
+ uint32_t Attribute2WrapShortestEnables;
+ uint32_t Attribute1WrapShortestEnables;
+ uint32_t Attribute0WrapShortestEnables;
+ uint32_t Attribute15WrapShortestEnables;
+ uint32_t Attribute14WrapShortestEnables;
+ uint32_t Attribute13WrapShortestEnables;
+ uint32_t Attribute12WrapShortestEnables;
+ uint32_t Attribute11WrapShortestEnables;
+ uint32_t Attribute10WrapShortestEnables;
+ uint32_t Attribute9WrapShortestEnables;
+ uint32_t Attribute8WrapShortestEnables;
+};
+
+static inline void
+GEN7_3DSTATE_SBE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_SBE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->AttributeSwizzleControlMode, 28, 28) |
+ __gen_field(values->NumberofSFOutputAttributes, 22, 27) |
+ __gen_field(values->AttributeSwizzleEnable, 21, 21) |
+ __gen_field(values->PointSpriteTextureCoordinateOrigin, 20, 20) |
+ __gen_field(values->VertexURBEntryReadLength, 11, 15) |
+ __gen_field(values->VertexURBEntryReadOffset, 4, 9) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->Attribute2n1ComponentOverrideW, 31, 31) |
+ __gen_field(values->Attribute2n1ComponentOverrideZ, 30, 30) |
+ __gen_field(values->Attribute2n1ComponentOverrideY, 29, 29) |
+ __gen_field(values->Attribute2n1ComponentOverrideX, 28, 28) |
+ __gen_field(values->Attribute2n1ConstantSource, 25, 26) |
+ __gen_field(values->Attribute2n1SwizzleSelect, 22, 23) |
+ __gen_field(values->Attribute2n1SourceAttribute, 16, 20) |
+ __gen_field(values->Attribute2nComponentOverrideW, 15, 15) |
+ __gen_field(values->Attribute2nComponentOverrideZ, 14, 14) |
+ __gen_field(values->Attribute2nComponentOverrideY, 13, 13) |
+ __gen_field(values->Attribute2nComponentOverrideX, 12, 12) |
+ __gen_field(values->Attribute2nConstantSource, 9, 10) |
+ __gen_field(values->Attribute2nSwizzleSelect, 6, 7) |
+ __gen_field(values->Attribute2nSourceAttribute, 0, 4) |
+ 0;
+
+ dw[10] =
+ __gen_field(values->PointSpriteTextureCoordinateEnable, 0, 31) |
+ 0;
+
+ dw[11] =
+ __gen_field(values->ConstantInterpolationEnable310, 0, 31) |
+ 0;
+
+ dw[12] =
+ __gen_field(values->Attribute7WrapShortestEnables, 28, 31) |
+ __gen_field(values->Attribute6WrapShortestEnables, 24, 27) |
+ __gen_field(values->Attribute5WrapShortestEnables, 20, 23) |
+ __gen_field(values->Attribute4WrapShortestEnables, 16, 19) |
+ __gen_field(values->Attribute3WrapShortestEnables, 12, 15) |
+ __gen_field(values->Attribute2WrapShortestEnables, 8, 11) |
+ __gen_field(values->Attribute1WrapShortestEnables, 4, 7) |
+ __gen_field(values->Attribute0WrapShortestEnables, 0, 3) |
+ 0;
+
+ dw[13] =
+ __gen_field(values->Attribute15WrapShortestEnables, 28, 31) |
+ __gen_field(values->Attribute14WrapShortestEnables, 24, 27) |
+ __gen_field(values->Attribute13WrapShortestEnables, 20, 23) |
+ __gen_field(values->Attribute12WrapShortestEnables, 16, 19) |
+ __gen_field(values->Attribute11WrapShortestEnables, 12, 15) |
+ __gen_field(values->Attribute10WrapShortestEnables, 8, 11) |
+ __gen_field(values->Attribute9WrapShortestEnables, 4, 7) |
+ __gen_field(values->Attribute8WrapShortestEnables, 0, 3) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_SCISSOR_STATE_POINTERS_length_bias 0x00000002
+#define GEN7_3DSTATE_SCISSOR_STATE_POINTERS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 15, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_SCISSOR_STATE_POINTERS_length 0x00000002
+
+struct GEN7_3DSTATE_SCISSOR_STATE_POINTERS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ScissorRectPointer;
+};
+
+static inline void
+GEN7_3DSTATE_SCISSOR_STATE_POINTERS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_SCISSOR_STATE_POINTERS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->ScissorRectPointer, 5, 31) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_SF_length_bias 0x00000002
+#define GEN7_3DSTATE_SF_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 19, \
+ .DwordLength = 5
+
+#define GEN7_3DSTATE_SF_length 0x00000007
+
+struct GEN7_3DSTATE_SF {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define D32_FLOAT_S8X24_UINT 0
+#define D32_FLOAT 1
+#define D24_UNORM_S8_UINT 2
+#define D24_UNORM_X8_UINT 3
+#define D16_UNORM 5
+ uint32_t DepthBufferSurfaceFormat;
+ bool LegacyGlobalDepthBiasEnable;
+ bool StatisticsEnable;
+ bool GlobalDepthOffsetEnableSolid;
+ bool GlobalDepthOffsetEnableWireframe;
+ bool GlobalDepthOffsetEnablePoint;
+#define RASTER_SOLID 0
+#define RASTER_WIREFRAME 1
+#define RASTER_POINT 2
+ uint32_t FrontFaceFillMode;
+#define RASTER_SOLID 0
+#define RASTER_WIREFRAME 1
+#define RASTER_POINT 2
+ uint32_t BackFaceFillMode;
+ bool ViewTransformEnable;
+ uint32_t FrontWinding;
+ bool AntiAliasingEnable;
+#define CULLMODE_BOTH 0
+#define CULLMODE_NONE 1
+#define CULLMODE_FRONT 2
+#define CULLMODE_BACK 3
+ uint32_t CullMode;
+ float LineWidth;
+ uint32_t LineEndCapAntialiasingRegionWidth;
+ bool ScissorRectangleEnable;
+ uint32_t MultisampleRasterizationMode;
+ bool LastPixelEnable;
+#define Vertex0 0
+#define Vertex1 1
+#define Vertex2 2
+ uint32_t TriangleStripListProvokingVertexSelect;
+ uint32_t LineStripListProvokingVertexSelect;
+#define Vertex0 0
+#define Vertex1 1
+#define Vertex2 2
+ uint32_t TriangleFanProvokingVertexSelect;
+#define AALINEDISTANCE_TRUE 1
+ uint32_t AALineDistanceMode;
+ uint32_t VertexSubPixelPrecisionSelect;
+ uint32_t UsePointWidthState;
+ float PointWidth;
+ float GlobalDepthOffsetConstant;
+ float GlobalDepthOffsetScale;
+ float GlobalDepthOffsetClamp;
+};
+
+static inline void
+GEN7_3DSTATE_SF_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_SF * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->DepthBufferSurfaceFormat, 12, 14) |
+ __gen_field(values->LegacyGlobalDepthBiasEnable, 11, 11) |
+ __gen_field(values->StatisticsEnable, 10, 10) |
+ __gen_field(values->GlobalDepthOffsetEnableSolid, 9, 9) |
+ __gen_field(values->GlobalDepthOffsetEnableWireframe, 8, 8) |
+ __gen_field(values->GlobalDepthOffsetEnablePoint, 7, 7) |
+ __gen_field(values->FrontFaceFillMode, 5, 6) |
+ __gen_field(values->BackFaceFillMode, 3, 4) |
+ __gen_field(values->ViewTransformEnable, 1, 1) |
+ __gen_field(values->FrontWinding, 0, 0) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->AntiAliasingEnable, 31, 31) |
+ __gen_field(values->CullMode, 29, 30) |
+ __gen_field(values->LineWidth * (1 << 7), 18, 27) |
+ __gen_field(values->LineEndCapAntialiasingRegionWidth, 16, 17) |
+ __gen_field(values->ScissorRectangleEnable, 11, 11) |
+ __gen_field(values->MultisampleRasterizationMode, 8, 9) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->LastPixelEnable, 31, 31) |
+ __gen_field(values->TriangleStripListProvokingVertexSelect, 29, 30) |
+ __gen_field(values->LineStripListProvokingVertexSelect, 27, 28) |
+ __gen_field(values->TriangleFanProvokingVertexSelect, 25, 26) |
+ __gen_field(values->AALineDistanceMode, 14, 14) |
+ __gen_field(values->VertexSubPixelPrecisionSelect, 12, 12) |
+ __gen_field(values->UsePointWidthState, 11, 11) |
+ __gen_field(values->PointWidth * (1 << 3), 0, 10) |
+ 0;
+
+ dw[4] =
+ __gen_float(values->GlobalDepthOffsetConstant) |
+ 0;
+
+ dw[5] =
+ __gen_float(values->GlobalDepthOffsetScale) |
+ 0;
+
+ dw[6] =
+ __gen_float(values->GlobalDepthOffsetClamp) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_SO_BUFFER_length_bias 0x00000002
+#define GEN7_3DSTATE_SO_BUFFER_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 24, \
+ .DwordLength = 2
+
+#define GEN7_3DSTATE_SO_BUFFER_length 0x00000004
+
+struct GEN7_3DSTATE_SO_BUFFER {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t SOBufferIndex;
+ struct GEN7_MEMORY_OBJECT_CONTROL_STATE SOBufferObjectControlState;
+ uint32_t SurfacePitch;
+ __gen_address_type SurfaceBaseAddress;
+ __gen_address_type SurfaceEndAddress;
+};
+
+static inline void
+GEN7_3DSTATE_SO_BUFFER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_SO_BUFFER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw_SOBufferObjectControlState;
+ GEN7_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_SOBufferObjectControlState, &values->SOBufferObjectControlState);
+ dw[1] =
+ __gen_field(values->SOBufferIndex, 29, 30) |
+ __gen_field(dw_SOBufferObjectControlState, 25, 28) |
+ __gen_field(values->SurfacePitch, 0, 11) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->SurfaceBaseAddress, dw2);
+
+ uint32_t dw3 =
+ 0;
+
+ dw[3] =
+ __gen_combine_address(data, &dw[3], values->SurfaceEndAddress, dw3);
+
+}
+
+#define GEN7_3DSTATE_SO_DECL_LIST_length_bias 0x00000002
+#define GEN7_3DSTATE_SO_DECL_LIST_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 23
+
+#define GEN7_3DSTATE_SO_DECL_LIST_length 0x00000000
+
+#define GEN7_SO_DECL_ENTRY_length 0x00000002
+
+#define GEN7_SO_DECL_length 0x00000001
+
+struct GEN7_SO_DECL {
+ uint32_t OutputBufferSlot;
+ uint32_t HoleFlag;
+ uint32_t RegisterIndex;
+ uint32_t ComponentMask;
+};
+
+static inline void
+GEN7_SO_DECL_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_SO_DECL * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->OutputBufferSlot, 12, 13) |
+ __gen_field(values->HoleFlag, 11, 11) |
+ __gen_field(values->RegisterIndex, 4, 9) |
+ __gen_field(values->ComponentMask, 0, 3) |
+ 0;
+
+}
+
+struct GEN7_SO_DECL_ENTRY {
+ struct GEN7_SO_DECL Stream3Decl;
+ struct GEN7_SO_DECL Stream2Decl;
+ struct GEN7_SO_DECL Stream1Decl;
+ struct GEN7_SO_DECL Stream0Decl;
+};
+
+static inline void
+GEN7_SO_DECL_ENTRY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_SO_DECL_ENTRY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ uint32_t dw_Stream3Decl;
+ GEN7_SO_DECL_pack(data, &dw_Stream3Decl, &values->Stream3Decl);
+ uint32_t dw_Stream2Decl;
+ GEN7_SO_DECL_pack(data, &dw_Stream2Decl, &values->Stream2Decl);
+ uint32_t dw_Stream1Decl;
+ GEN7_SO_DECL_pack(data, &dw_Stream1Decl, &values->Stream1Decl);
+ uint32_t dw_Stream0Decl;
+ GEN7_SO_DECL_pack(data, &dw_Stream0Decl, &values->Stream0Decl);
+ uint64_t qw0 =
+ __gen_field(dw_Stream3Decl, 48, 63) |
+ __gen_field(dw_Stream2Decl, 32, 47) |
+ __gen_field(dw_Stream1Decl, 16, 31) |
+ __gen_field(dw_Stream0Decl, 0, 15) |
+ 0;
+
+ dw[0] = qw0;
+ dw[1] = qw0 >> 32;
+
+}
+
+struct GEN7_3DSTATE_SO_DECL_LIST {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t StreamtoBufferSelects3;
+ uint32_t StreamtoBufferSelects2;
+ uint32_t StreamtoBufferSelects1;
+ uint32_t StreamtoBufferSelects0;
+ uint32_t NumEntries3;
+ uint32_t NumEntries2;
+ uint32_t NumEntries1;
+ uint32_t NumEntries0;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN7_3DSTATE_SO_DECL_LIST_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_SO_DECL_LIST * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 8) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->StreamtoBufferSelects3, 12, 15) |
+ __gen_field(values->StreamtoBufferSelects2, 8, 11) |
+ __gen_field(values->StreamtoBufferSelects1, 4, 7) |
+ __gen_field(values->StreamtoBufferSelects0, 0, 3) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->NumEntries3, 24, 31) |
+ __gen_field(values->NumEntries2, 16, 23) |
+ __gen_field(values->NumEntries1, 8, 15) |
+ __gen_field(values->NumEntries0, 0, 7) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN7_3DSTATE_STENCIL_BUFFER_length_bias 0x00000002
+#define GEN7_3DSTATE_STENCIL_BUFFER_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 6, \
+ .DwordLength = 1
+
+#define GEN7_3DSTATE_STENCIL_BUFFER_length 0x00000003
+
+struct GEN7_3DSTATE_STENCIL_BUFFER {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ struct GEN7_MEMORY_OBJECT_CONTROL_STATE StencilBufferObjectControlState;
+ uint32_t SurfacePitch;
+ __gen_address_type SurfaceBaseAddress;
+};
+
+static inline void
+GEN7_3DSTATE_STENCIL_BUFFER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_STENCIL_BUFFER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw_StencilBufferObjectControlState;
+ GEN7_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_StencilBufferObjectControlState, &values->StencilBufferObjectControlState);
+ dw[1] =
+ __gen_field(dw_StencilBufferObjectControlState, 25, 28) |
+ __gen_field(values->SurfacePitch, 0, 16) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->SurfaceBaseAddress, dw2);
+
+}
+
+#define GEN7_3DSTATE_STREAMOUT_length_bias 0x00000002
+#define GEN7_3DSTATE_STREAMOUT_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 30, \
+ .DwordLength = 1
+
+#define GEN7_3DSTATE_STREAMOUT_length 0x00000003
+
+struct GEN7_3DSTATE_STREAMOUT {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t SOFunctionEnable;
+ uint32_t RenderingDisable;
+ uint32_t RenderStreamSelect;
+#define LEADING 0
+#define TRAILING 1
+ uint32_t ReorderMode;
+ bool SOStatisticsEnable;
+ uint32_t SOBufferEnable3;
+ uint32_t SOBufferEnable2;
+ uint32_t SOBufferEnable1;
+ uint32_t SOBufferEnable0;
+ uint32_t Stream3VertexReadOffset;
+ uint32_t Stream3VertexReadLength;
+ uint32_t Stream2VertexReadOffset;
+ uint32_t Stream2VertexReadLength;
+ uint32_t Stream1VertexReadOffset;
+ uint32_t Stream1VertexReadLength;
+ uint32_t Stream0VertexReadOffset;
+ uint32_t Stream0VertexReadLength;
+};
+
+static inline void
+GEN7_3DSTATE_STREAMOUT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_STREAMOUT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->SOFunctionEnable, 31, 31) |
+ __gen_field(values->RenderingDisable, 30, 30) |
+ __gen_field(values->RenderStreamSelect, 27, 28) |
+ __gen_field(values->ReorderMode, 26, 26) |
+ __gen_field(values->SOStatisticsEnable, 25, 25) |
+ __gen_field(values->SOBufferEnable3, 11, 11) |
+ __gen_field(values->SOBufferEnable2, 10, 10) |
+ __gen_field(values->SOBufferEnable1, 9, 9) |
+ __gen_field(values->SOBufferEnable0, 8, 8) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->Stream3VertexReadOffset, 29, 29) |
+ __gen_field(values->Stream3VertexReadLength, 24, 28) |
+ __gen_field(values->Stream2VertexReadOffset, 21, 21) |
+ __gen_field(values->Stream2VertexReadLength, 16, 20) |
+ __gen_field(values->Stream1VertexReadOffset, 13, 13) |
+ __gen_field(values->Stream1VertexReadLength, 8, 12) |
+ __gen_field(values->Stream0VertexReadOffset, 5, 5) |
+ __gen_field(values->Stream0VertexReadLength, 0, 4) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_TE_length_bias 0x00000002
+#define GEN7_3DSTATE_TE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 28, \
+ .DwordLength = 2
+
+#define GEN7_3DSTATE_TE_length 0x00000004
+
+struct GEN7_3DSTATE_TE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define INTEGER 0
+#define ODD_FRACTIONAL 1
+#define EVEN_FRACTIONAL 2
+ uint32_t Partitioning;
+#define POINT 0
+#define OUTPUT_LINE 1
+#define OUTPUT_TRI_CW 2
+#define OUTPUT_TRI_CCW 3
+ uint32_t OutputTopology;
+#define QUAD 0
+#define TRI 1
+#define ISOLINE 2
+ uint32_t TEDomain;
+#define HW_TESS 0
+#define SW_TESS 1
+ uint32_t TEMode;
+ bool TEEnable;
+ float MaximumTessellationFactorOdd;
+ float MaximumTessellationFactorNotOdd;
+};
+
+static inline void
+GEN7_3DSTATE_TE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_TE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->Partitioning, 12, 13) |
+ __gen_field(values->OutputTopology, 8, 9) |
+ __gen_field(values->TEDomain, 4, 5) |
+ __gen_field(values->TEMode, 1, 2) |
+ __gen_field(values->TEEnable, 0, 0) |
+ 0;
+
+ dw[2] =
+ __gen_float(values->MaximumTessellationFactorOdd) |
+ 0;
+
+ dw[3] =
+ __gen_float(values->MaximumTessellationFactorNotOdd) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_URB_DS_length_bias 0x00000002
+#define GEN7_3DSTATE_URB_DS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 50, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_URB_DS_length 0x00000002
+
+struct GEN7_3DSTATE_URB_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t DSURBStartingAddress;
+ uint32_t DSURBEntryAllocationSize;
+ uint32_t DSNumberofURBEntries;
+};
+
+static inline void
+GEN7_3DSTATE_URB_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_URB_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->DSURBStartingAddress, 25, 29) |
+ __gen_field(values->DSURBEntryAllocationSize, 16, 24) |
+ __gen_field(values->DSNumberofURBEntries, 0, 15) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_URB_GS_length_bias 0x00000002
+#define GEN7_3DSTATE_URB_GS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 51, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_URB_GS_length 0x00000002
+
+struct GEN7_3DSTATE_URB_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t GSURBStartingAddress;
+ uint32_t GSURBEntryAllocationSize;
+ uint32_t GSNumberofURBEntries;
+};
+
+static inline void
+GEN7_3DSTATE_URB_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_URB_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->GSURBStartingAddress, 25, 29) |
+ __gen_field(values->GSURBEntryAllocationSize, 16, 24) |
+ __gen_field(values->GSNumberofURBEntries, 0, 15) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_URB_HS_length_bias 0x00000002
+#define GEN7_3DSTATE_URB_HS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 49, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_URB_HS_length 0x00000002
+
+struct GEN7_3DSTATE_URB_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t HSURBStartingAddress;
+ uint32_t HSURBEntryAllocationSize;
+ uint32_t HSNumberofURBEntries;
+};
+
+static inline void
+GEN7_3DSTATE_URB_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_URB_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->HSURBStartingAddress, 25, 29) |
+ __gen_field(values->HSURBEntryAllocationSize, 16, 24) |
+ __gen_field(values->HSNumberofURBEntries, 0, 15) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_VERTEX_BUFFERS_length_bias 0x00000002
+#define GEN7_3DSTATE_VERTEX_BUFFERS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 8
+
+#define GEN7_3DSTATE_VERTEX_BUFFERS_length 0x00000000
+
+#define GEN7_VERTEX_BUFFER_STATE_length 0x00000004
+
+struct GEN7_VERTEX_BUFFER_STATE {
+ uint32_t VertexBufferIndex;
+#define VERTEXDATA 0
+#define INSTANCEDATA 1
+ uint32_t BufferAccessType;
+ struct GEN7_MEMORY_OBJECT_CONTROL_STATE VertexBufferMemoryObjectControlState;
+ uint32_t AddressModifyEnable;
+ bool NullVertexBuffer;
+ uint32_t VertexFetchInvalidate;
+ uint32_t BufferPitch;
+ __gen_address_type BufferStartingAddress;
+ __gen_address_type EndAddress;
+ uint32_t InstanceDataStepRate;
+};
+
+static inline void
+GEN7_VERTEX_BUFFER_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_VERTEX_BUFFER_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ uint32_t dw_VertexBufferMemoryObjectControlState;
+ GEN7_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_VertexBufferMemoryObjectControlState, &values->VertexBufferMemoryObjectControlState);
+ dw[0] =
+ __gen_field(values->VertexBufferIndex, 26, 31) |
+ __gen_field(values->BufferAccessType, 20, 20) |
+ __gen_field(dw_VertexBufferMemoryObjectControlState, 16, 19) |
+ __gen_field(values->AddressModifyEnable, 14, 14) |
+ __gen_field(values->NullVertexBuffer, 13, 13) |
+ __gen_field(values->VertexFetchInvalidate, 12, 12) |
+ __gen_field(values->BufferPitch, 0, 11) |
+ 0;
+
+ uint32_t dw1 =
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->BufferStartingAddress, dw1);
+
+ uint32_t dw2 =
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->EndAddress, dw2);
+
+ dw[3] =
+ __gen_field(values->InstanceDataStepRate, 0, 31) |
+ 0;
+
+}
+
+struct GEN7_3DSTATE_VERTEX_BUFFERS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN7_3DSTATE_VERTEX_BUFFERS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_VERTEX_BUFFERS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN7_3DSTATE_VERTEX_ELEMENTS_length_bias 0x00000002
+#define GEN7_3DSTATE_VERTEX_ELEMENTS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 9
+
+#define GEN7_3DSTATE_VERTEX_ELEMENTS_length 0x00000000
+
+#define GEN7_VERTEX_ELEMENT_STATE_length 0x00000002
+
+struct GEN7_VERTEX_ELEMENT_STATE {
+ uint32_t VertexBufferIndex;
+ bool Valid;
+ uint32_t SourceElementFormat;
+ bool EdgeFlagEnable;
+ uint32_t SourceElementOffset;
+ uint32_t Component0Control;
+ uint32_t Component1Control;
+ uint32_t Component2Control;
+ uint32_t Component3Control;
+};
+
+static inline void
+GEN7_VERTEX_ELEMENT_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_VERTEX_ELEMENT_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->VertexBufferIndex, 26, 31) |
+ __gen_field(values->Valid, 25, 25) |
+ __gen_field(values->SourceElementFormat, 16, 24) |
+ __gen_field(values->EdgeFlagEnable, 15, 15) |
+ __gen_field(values->SourceElementOffset, 0, 11) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->Component0Control, 28, 30) |
+ __gen_field(values->Component1Control, 24, 26) |
+ __gen_field(values->Component2Control, 20, 22) |
+ __gen_field(values->Component3Control, 16, 18) |
+ 0;
+
+}
+
+struct GEN7_3DSTATE_VERTEX_ELEMENTS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN7_3DSTATE_VERTEX_ELEMENTS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_VERTEX_ELEMENTS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN7_3DSTATE_VF_STATISTICS_length_bias 0x00000001
+#define GEN7_3DSTATE_VF_STATISTICS_header \
+ .CommandType = 3, \
+ .CommandSubType = 1, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 11
+
+#define GEN7_3DSTATE_VF_STATISTICS_length 0x00000001
+
+struct GEN7_3DSTATE_VF_STATISTICS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ bool StatisticsEnable;
+};
+
+static inline void
+GEN7_3DSTATE_VF_STATISTICS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_VF_STATISTICS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->StatisticsEnable, 0, 0) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_VIEWPORT_STATE_POINTERS_CC_length_bias 0x00000002
+#define GEN7_3DSTATE_VIEWPORT_STATE_POINTERS_CC_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 35, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_VIEWPORT_STATE_POINTERS_CC_length 0x00000002
+
+struct GEN7_3DSTATE_VIEWPORT_STATE_POINTERS_CC {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t CCViewportPointer;
+};
+
+static inline void
+GEN7_3DSTATE_VIEWPORT_STATE_POINTERS_CC_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_VIEWPORT_STATE_POINTERS_CC * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->CCViewportPointer, 5, 31) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP_length_bias 0x00000002
+#define GEN7_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 33, \
+ .DwordLength = 0
+
+#define GEN7_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP_length 0x00000002
+
+struct GEN7_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t SFClipViewportPointer;
+};
+
+static inline void
+GEN7_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->SFClipViewportPointer, 6, 31) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_VS_length_bias 0x00000002
+#define GEN7_3DSTATE_VS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 16, \
+ .DwordLength = 4
+
+#define GEN7_3DSTATE_VS_length 0x00000006
+
+struct GEN7_3DSTATE_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t KernelStartPointer;
+#define Multiple 0
+#define Single 1
+ uint32_t SingleVertexDispatch;
+#define Dmask 0
+#define Vmask 1
+ uint32_t VectorMaskEnableVME;
+#define NoSamplers 0
+#define _14Samplers 1
+#define _58Samplers 2
+#define _912Samplers 3
+#define _1316Samplers 4
+ uint32_t SamplerCount;
+ uint32_t BindingTableEntryCount;
+#define IEEE754 0
+#define Alternate 1
+ uint32_t FloatingPointMode;
+ bool IllegalOpcodeExceptionEnable;
+ bool SoftwareExceptionEnable;
+ uint32_t ScratchSpaceBaseOffset;
+ uint32_t PerThreadScratchSpace;
+ uint32_t DispatchGRFStartRegisterforURBData;
+ uint32_t VertexURBEntryReadLength;
+ uint32_t VertexURBEntryReadOffset;
+ uint32_t MaximumNumberofThreads;
+ bool StatisticsEnable;
+ bool VertexCacheDisable;
+ bool VSFunctionEnable;
+};
+
+static inline void
+GEN7_3DSTATE_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->KernelStartPointer, 6, 31) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->SingleVertexDispatch, 31, 31) |
+ __gen_field(values->VectorMaskEnableVME, 30, 30) |
+ __gen_field(values->SamplerCount, 27, 29) |
+ __gen_field(values->BindingTableEntryCount, 18, 25) |
+ __gen_field(values->FloatingPointMode, 16, 16) |
+ __gen_field(values->IllegalOpcodeExceptionEnable, 13, 13) |
+ __gen_field(values->SoftwareExceptionEnable, 7, 7) |
+ 0;
+
+ dw[3] =
+ __gen_offset(values->ScratchSpaceBaseOffset, 10, 31) |
+ __gen_field(values->PerThreadScratchSpace, 0, 3) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->DispatchGRFStartRegisterforURBData, 20, 24) |
+ __gen_field(values->VertexURBEntryReadLength, 11, 16) |
+ __gen_field(values->VertexURBEntryReadOffset, 4, 9) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->MaximumNumberofThreads, 25, 31) |
+ __gen_field(values->StatisticsEnable, 10, 10) |
+ __gen_field(values->VertexCacheDisable, 1, 1) |
+ __gen_field(values->VSFunctionEnable, 0, 0) |
+ 0;
+
+}
+
+#define GEN7_3DSTATE_WM_length_bias 0x00000002
+#define GEN7_3DSTATE_WM_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 20, \
+ .DwordLength = 1
+
+#define GEN7_3DSTATE_WM_length 0x00000003
+
+struct GEN7_3DSTATE_WM {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool StatisticsEnable;
+ bool DepthBufferClear;
+ bool ThreadDispatchEnable;
+ bool DepthBufferResolveEnable;
+ bool HierarchicalDepthBufferResolveEnable;
+ bool LegacyDiamondLineRasterization;
+ bool PixelShaderKillPixel;
+#define PSCDEPTH_OFF 0
+#define PSCDEPTH_ON 1
+#define PSCDEPTH_ON_GE 2
+#define PSCDEPTH_ON_LE 3
+ uint32_t PixelShaderComputedDepthMode;
+#define EDSC_NORMAL 0
+#define EDSC_PSEXEC 1
+#define EDSC_PREPS 2
+ uint32_t EarlyDepthStencilControl;
+ bool PixelShaderUsesSourceDepth;
+ bool PixelShaderUsesSourceW;
+#define INTERP_PIXEL 0
+#define INTERP_CENTROID 2
+#define INTERP_SAMPLE 3
+ uint32_t PositionZWInterpolationMode;
+ uint32_t BarycentricInterpolationMode;
+ bool PixelShaderUsesInputCoverageMask;
+ uint32_t LineEndCapAntialiasingRegionWidth;
+ uint32_t LineAntialiasingRegionWidth;
+ bool PolygonStippleEnable;
+ bool LineStippleEnable;
+#define RASTRULE_UPPER_LEFT 0
+#define RASTRULE_UPPER_RIGHT 1
+ uint32_t PointRasterizationRule;
+#define MSRASTMODE_OFF_PIXEL 0
+#define MSRASTMODE_OFF_PATTERN 1
+#define MSRASTMODE_ON_PIXEL 2
+#define MSRASTMODE_ON_PATTERN 3
+ uint32_t MultisampleRasterizationMode;
+#define MSDISPMODE_PERSAMPLE 0
+#define MSDISPMODE_PERPIXEL 1
+ uint32_t MultisampleDispatchMode;
+};
+
+static inline void
+GEN7_3DSTATE_WM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_3DSTATE_WM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->StatisticsEnable, 31, 31) |
+ __gen_field(values->DepthBufferClear, 30, 30) |
+ __gen_field(values->ThreadDispatchEnable, 29, 29) |
+ __gen_field(values->DepthBufferResolveEnable, 28, 28) |
+ __gen_field(values->HierarchicalDepthBufferResolveEnable, 27, 27) |
+ __gen_field(values->LegacyDiamondLineRasterization, 26, 26) |
+ __gen_field(values->PixelShaderKillPixel, 25, 25) |
+ __gen_field(values->PixelShaderComputedDepthMode, 23, 24) |
+ __gen_field(values->EarlyDepthStencilControl, 21, 22) |
+ __gen_field(values->PixelShaderUsesSourceDepth, 20, 20) |
+ __gen_field(values->PixelShaderUsesSourceW, 19, 19) |
+ __gen_field(values->PositionZWInterpolationMode, 17, 18) |
+ __gen_field(values->BarycentricInterpolationMode, 11, 16) |
+ __gen_field(values->PixelShaderUsesInputCoverageMask, 10, 10) |
+ __gen_field(values->LineEndCapAntialiasingRegionWidth, 8, 9) |
+ __gen_field(values->LineAntialiasingRegionWidth, 6, 7) |
+ __gen_field(values->PolygonStippleEnable, 4, 4) |
+ __gen_field(values->LineStippleEnable, 3, 3) |
+ __gen_field(values->PointRasterizationRule, 2, 2) |
+ __gen_field(values->MultisampleRasterizationMode, 0, 1) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->MultisampleDispatchMode, 31, 31) |
+ 0;
+
+}
+
+#define GEN7_GPGPU_OBJECT_length_bias 0x00000002
+#define GEN7_GPGPU_OBJECT_header \
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 1, \
+ .SubOpcode = 4, \
+ .DwordLength = 6
+
+#define GEN7_GPGPU_OBJECT_length 0x00000008
+
+struct GEN7_GPGPU_OBJECT {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ bool PredicateEnable;
+ uint32_t DwordLength;
+ uint32_t SharedLocalMemoryFixedOffset;
+ uint32_t InterfaceDescriptorOffset;
+ uint32_t SharedLocalMemoryOffset;
+ uint32_t EndofThreadGroup;
+#define HalfSlice1 2
+#define HalfSlice0 1
+#define EitherHalfSlice 0
+ uint32_t HalfSliceDestinationSelect;
+ uint32_t IndirectDataLength;
+ uint32_t IndirectDataStartAddress;
+ uint32_t ThreadGroupIDX;
+ uint32_t ThreadGroupIDY;
+ uint32_t ThreadGroupIDZ;
+ uint32_t ExecutionMask;
+};
+
+static inline void
+GEN7_GPGPU_OBJECT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_GPGPU_OBJECT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->PredicateEnable, 8, 8) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->SharedLocalMemoryFixedOffset, 7, 7) |
+ __gen_field(values->InterfaceDescriptorOffset, 0, 4) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->SharedLocalMemoryOffset, 28, 31) |
+ __gen_field(values->EndofThreadGroup, 24, 24) |
+ __gen_field(values->HalfSliceDestinationSelect, 17, 18) |
+ __gen_field(values->IndirectDataLength, 0, 16) |
+ 0;
+
+ dw[3] =
+ __gen_offset(values->IndirectDataStartAddress, 0, 31) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->ThreadGroupIDX, 0, 31) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->ThreadGroupIDY, 0, 31) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->ThreadGroupIDZ, 0, 31) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->ExecutionMask, 0, 31) |
+ 0;
+
+}
+
+#define GEN7_GPGPU_WALKER_length_bias 0x00000002
+#define GEN7_GPGPU_WALKER_header \
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 1, \
+ .SubOpcodeA = 5, \
+ .DwordLength = 9
+
+#define GEN7_GPGPU_WALKER_length 0x0000000b
+
+struct GEN7_GPGPU_WALKER {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcodeA;
+ bool IndirectParameterEnable;
+ bool PredicateEnable;
+ uint32_t DwordLength;
+ uint32_t InterfaceDescriptorOffset;
+#define SIMD8 0
+#define SIMD16 1
+#define SIMD32 2
+ uint32_t SIMDSize;
+ uint32_t ThreadDepthCounterMaximum;
+ uint32_t ThreadHeightCounterMaximum;
+ uint32_t ThreadWidthCounterMaximum;
+ uint32_t ThreadGroupIDStartingX;
+ uint32_t ThreadGroupIDXDimension;
+ uint32_t ThreadGroupIDStartingY;
+ uint32_t ThreadGroupIDYDimension;
+ uint32_t ThreadGroupIDStartingZ;
+ uint32_t ThreadGroupIDZDimension;
+ uint32_t RightExecutionMask;
+ uint32_t BottomExecutionMask;
+};
+
+static inline void
+GEN7_GPGPU_WALKER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_GPGPU_WALKER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcodeA, 16, 23) |
+ __gen_field(values->IndirectParameterEnable, 10, 10) |
+ __gen_field(values->PredicateEnable, 8, 8) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->InterfaceDescriptorOffset, 0, 4) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->SIMDSize, 30, 31) |
+ __gen_field(values->ThreadDepthCounterMaximum, 16, 21) |
+ __gen_field(values->ThreadHeightCounterMaximum, 8, 13) |
+ __gen_field(values->ThreadWidthCounterMaximum, 0, 5) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->ThreadGroupIDStartingX, 0, 31) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->ThreadGroupIDXDimension, 0, 31) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->ThreadGroupIDStartingY, 0, 31) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->ThreadGroupIDYDimension, 0, 31) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->ThreadGroupIDStartingZ, 0, 31) |
+ 0;
+
+ dw[8] =
+ __gen_field(values->ThreadGroupIDZDimension, 0, 31) |
+ 0;
+
+ dw[9] =
+ __gen_field(values->RightExecutionMask, 0, 31) |
+ 0;
+
+ dw[10] =
+ __gen_field(values->BottomExecutionMask, 0, 31) |
+ 0;
+
+}
+
+#define GEN7_MEDIA_CURBE_LOAD_length_bias 0x00000002
+#define GEN7_MEDIA_CURBE_LOAD_header \
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 0, \
+ .SubOpcode = 1, \
+ .DwordLength = 2
+
+#define GEN7_MEDIA_CURBE_LOAD_length 0x00000004
+
+struct GEN7_MEDIA_CURBE_LOAD {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ uint32_t DwordLength;
+ uint32_t CURBETotalDataLength;
+ uint32_t CURBEDataStartAddress;
+};
+
+static inline void
+GEN7_MEDIA_CURBE_LOAD_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MEDIA_CURBE_LOAD * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ 0;
+
+ dw[2] =
+ __gen_field(values->CURBETotalDataLength, 0, 16) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->CURBEDataStartAddress, 0, 31) |
+ 0;
+
+}
+
+#define GEN7_MEDIA_INTERFACE_DESCRIPTOR_LOAD_length_bias 0x00000002
+#define GEN7_MEDIA_INTERFACE_DESCRIPTOR_LOAD_header\
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 0, \
+ .SubOpcode = 2, \
+ .DwordLength = 2
+
+#define GEN7_MEDIA_INTERFACE_DESCRIPTOR_LOAD_length 0x00000004
+
+struct GEN7_MEDIA_INTERFACE_DESCRIPTOR_LOAD {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ uint32_t DwordLength;
+ uint32_t InterfaceDescriptorTotalLength;
+ uint32_t InterfaceDescriptorDataStartAddress;
+};
+
+static inline void
+GEN7_MEDIA_INTERFACE_DESCRIPTOR_LOAD_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MEDIA_INTERFACE_DESCRIPTOR_LOAD * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ 0;
+
+ dw[2] =
+ __gen_field(values->InterfaceDescriptorTotalLength, 0, 16) |
+ 0;
+
+ dw[3] =
+ __gen_offset(values->InterfaceDescriptorDataStartAddress, 0, 31) |
+ 0;
+
+}
+
+#define GEN7_MEDIA_OBJECT_length_bias 0x00000002
+#define GEN7_MEDIA_OBJECT_header \
+ .CommandType = 3, \
+ .MediaCommandPipeline = 2, \
+ .MediaCommandOpcode = 1, \
+ .MediaCommandSubOpcode = 0
+
+#define GEN7_MEDIA_OBJECT_length 0x00000000
+
+struct GEN7_MEDIA_OBJECT {
+ uint32_t CommandType;
+ uint32_t MediaCommandPipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t MediaCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t InterfaceDescriptorOffset;
+ bool ChildrenPresent;
+#define Nothreadsynchronization 0
+#define Threaddispatchissynchronizedbythespawnrootthreadmessage 1
+ uint32_t ThreadSynchronization;
+#define Notusingscoreboard 0
+#define Usingscoreboard 1
+ uint32_t UseScoreboard;
+#define HalfSlice1 2
+#define HalfSlice0 1
+#define Eitherhalfslice 0
+ uint32_t HalfSliceDestinationSelect;
+ uint32_t IndirectDataLength;
+ __gen_address_type IndirectDataStartAddress;
+ uint32_t ScoredboardY;
+ uint32_t ScoreboardX;
+ uint32_t ScoreboardColor;
+ bool ScoreboardMask;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN7_MEDIA_OBJECT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MEDIA_OBJECT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MediaCommandPipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->MediaCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->InterfaceDescriptorOffset, 0, 4) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ChildrenPresent, 31, 31) |
+ __gen_field(values->ThreadSynchronization, 24, 24) |
+ __gen_field(values->UseScoreboard, 21, 21) |
+ __gen_field(values->HalfSliceDestinationSelect, 17, 18) |
+ __gen_field(values->IndirectDataLength, 0, 16) |
+ 0;
+
+ uint32_t dw3 =
+ 0;
+
+ dw[3] =
+ __gen_combine_address(data, &dw[3], values->IndirectDataStartAddress, dw3);
+
+ dw[4] =
+ __gen_field(values->ScoredboardY, 16, 24) |
+ __gen_field(values->ScoreboardX, 0, 8) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->ScoreboardColor, 16, 19) |
+ __gen_field(values->ScoreboardMask, 0, 7) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN7_MEDIA_OBJECT_PRT_length_bias 0x00000002
+#define GEN7_MEDIA_OBJECT_PRT_header \
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 1, \
+ .SubOpcode = 2, \
+ .DwordLength = 14
+
+#define GEN7_MEDIA_OBJECT_PRT_length 0x00000010
+
+struct GEN7_MEDIA_OBJECT_PRT {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ uint32_t DwordLength;
+ uint32_t InterfaceDescriptorOffset;
+ bool ChildrenPresent;
+ bool PRT_FenceNeeded;
+#define Rootthreadqueue 0
+#define VFEstateflush 1
+ uint32_t PRT_FenceType;
+ uint32_t InlineData[12];
+};
+
+static inline void
+GEN7_MEDIA_OBJECT_PRT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MEDIA_OBJECT_PRT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->InterfaceDescriptorOffset, 0, 4) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ChildrenPresent, 31, 31) |
+ __gen_field(values->PRT_FenceNeeded, 23, 23) |
+ __gen_field(values->PRT_FenceType, 22, 22) |
+ 0;
+
+ dw[3] =
+ 0;
+
+ for (uint32_t i = 0, j = 4; i < 12; i += 1, j++) {
+ dw[j] =
+ __gen_field(values->InlineData[i + 0], 0, 31) |
+ 0;
+ }
+
+}
+
+#define GEN7_MEDIA_OBJECT_WALKER_length_bias 0x00000002
+#define GEN7_MEDIA_OBJECT_WALKER_header \
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 1, \
+ .SubOpcode = 3
+
+#define GEN7_MEDIA_OBJECT_WALKER_length 0x00000000
+
+struct GEN7_MEDIA_OBJECT_WALKER {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ uint32_t DwordLength;
+ uint32_t InterfaceDescriptorOffset;
+ bool ChildrenPresent;
+#define Nothreadsynchronization 0
+#define Threaddispatchissynchronizedbythespawnrootthreadmessage 1
+ uint32_t ThreadSynchronization;
+#define Notusingscoreboard 0
+#define Usingscoreboard 1
+ uint32_t UseScoreboard;
+ uint32_t IndirectDataLength;
+ uint32_t IndirectDataStartAddress;
+ bool ScoreboardMask;
+ bool DualMode;
+ bool Repel;
+ uint32_t ColorCountMinusOne;
+ uint32_t MiddleLoopExtraSteps;
+ uint32_t LocalMidLoopUnitY;
+ uint32_t MidLoopUnitX;
+ uint32_t GlobalLoopExecCount;
+ uint32_t LocalLoopExecCount;
+ uint32_t BlockResolutionY;
+ uint32_t BlockResolutionX;
+ uint32_t LocalStartY;
+ uint32_t LocalStartX;
+ uint32_t LocalEndY;
+ uint32_t LocalEndX;
+ uint32_t LocalOuterLoopStrideY;
+ uint32_t LocalOuterLoopStrideX;
+ uint32_t LocalInnerLoopUnitY;
+ uint32_t LocalInnerLoopUnitX;
+ uint32_t GlobalResolutionY;
+ uint32_t GlobalResolutionX;
+ uint32_t GlobalStartY;
+ uint32_t GlobalStartX;
+ uint32_t GlobalOuterLoopStrideY;
+ uint32_t GlobalOuterLoopStrideX;
+ uint32_t GlobalInnerLoopUnitY;
+ uint32_t GlobalInnerLoopUnitX;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN7_MEDIA_OBJECT_WALKER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MEDIA_OBJECT_WALKER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->InterfaceDescriptorOffset, 0, 4) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ChildrenPresent, 31, 31) |
+ __gen_field(values->ThreadSynchronization, 24, 24) |
+ __gen_field(values->UseScoreboard, 21, 21) |
+ __gen_field(values->IndirectDataLength, 0, 16) |
+ 0;
+
+ dw[3] =
+ __gen_offset(values->IndirectDataStartAddress, 0, 31) |
+ 0;
+
+ dw[4] =
+ 0;
+
+ dw[5] =
+ __gen_field(values->ScoreboardMask, 0, 7) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->DualMode, 31, 31) |
+ __gen_field(values->Repel, 30, 30) |
+ __gen_field(values->ColorCountMinusOne, 24, 27) |
+ __gen_field(values->MiddleLoopExtraSteps, 16, 20) |
+ __gen_field(values->LocalMidLoopUnitY, 12, 13) |
+ __gen_field(values->MidLoopUnitX, 8, 9) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->GlobalLoopExecCount, 16, 25) |
+ __gen_field(values->LocalLoopExecCount, 0, 9) |
+ 0;
+
+ dw[8] =
+ __gen_field(values->BlockResolutionY, 16, 24) |
+ __gen_field(values->BlockResolutionX, 0, 8) |
+ 0;
+
+ dw[9] =
+ __gen_field(values->LocalStartY, 16, 24) |
+ __gen_field(values->LocalStartX, 0, 8) |
+ 0;
+
+ dw[10] =
+ __gen_field(values->LocalEndY, 16, 24) |
+ __gen_field(values->LocalEndX, 0, 8) |
+ 0;
+
+ dw[11] =
+ __gen_field(values->LocalOuterLoopStrideY, 16, 25) |
+ __gen_field(values->LocalOuterLoopStrideX, 0, 9) |
+ 0;
+
+ dw[12] =
+ __gen_field(values->LocalInnerLoopUnitY, 16, 25) |
+ __gen_field(values->LocalInnerLoopUnitX, 0, 9) |
+ 0;
+
+ dw[13] =
+ __gen_field(values->GlobalResolutionY, 16, 24) |
+ __gen_field(values->GlobalResolutionX, 0, 8) |
+ 0;
+
+ dw[14] =
+ __gen_field(values->GlobalStartY, 16, 25) |
+ __gen_field(values->GlobalStartX, 0, 9) |
+ 0;
+
+ dw[15] =
+ __gen_field(values->GlobalOuterLoopStrideY, 16, 25) |
+ __gen_field(values->GlobalOuterLoopStrideX, 0, 9) |
+ 0;
+
+ dw[16] =
+ __gen_field(values->GlobalInnerLoopUnitY, 16, 25) |
+ __gen_field(values->GlobalInnerLoopUnitX, 0, 9) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN7_MEDIA_STATE_FLUSH_length_bias 0x00000002
+#define GEN7_MEDIA_STATE_FLUSH_header \
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 0, \
+ .SubOpcode = 4, \
+ .DwordLength = 0
+
+#define GEN7_MEDIA_STATE_FLUSH_length 0x00000002
+
+struct GEN7_MEDIA_STATE_FLUSH {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ uint32_t DwordLength;
+ uint32_t WatermarkRequired;
+ uint32_t InterfaceDescriptorOffset;
+};
+
+static inline void
+GEN7_MEDIA_STATE_FLUSH_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MEDIA_STATE_FLUSH * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->WatermarkRequired, 6, 6) |
+ __gen_field(values->InterfaceDescriptorOffset, 0, 5) |
+ 0;
+
+}
+
+#define GEN7_MEDIA_VFE_STATE_length_bias 0x00000002
+#define GEN7_MEDIA_VFE_STATE_header \
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 0, \
+ .SubOpcode = 0, \
+ .DwordLength = 6
+
+#define GEN7_MEDIA_VFE_STATE_length 0x00000008
+
+struct GEN7_MEDIA_VFE_STATE {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ uint32_t DwordLength;
+ uint32_t ScratchSpaceBasePointer;
+ uint32_t PerThreadScratchSpace;
+ uint32_t MaximumNumberofThreads;
+ uint32_t NumberofURBEntries;
+#define Maintainingtheexistingtimestampstate 0
+#define Resettingrelativetimerandlatchingtheglobaltimestamp 1
+ uint32_t ResetGatewayTimer;
+#define MaintainingOpenGatewayForwardMsgCloseGatewayprotocollegacymode 0
+#define BypassingOpenGatewayCloseGatewayprotocol 1
+ uint32_t BypassGatewayControl;
+#define NoMMIOreadwriteallowed 0
+#define MMIOreadwritetoanyaddress 2
+ uint32_t GatewayMMIOAccessControl;
+ uint32_t GPGPUMode;
+ uint32_t URBEntryAllocationSize;
+ uint32_t CURBEAllocationSize;
+#define Scoreboarddisabled 0
+#define Scoreboardenabled 1
+ uint32_t ScoreboardEnable;
+#define StallingScoreboard 0
+#define NonStallingScoreboard 1
+ uint32_t ScoreboardType;
+ uint32_t ScoreboardMask;
+ uint32_t Scoreboard3DeltaY;
+ uint32_t Scoreboard3DeltaX;
+ uint32_t Scoreboard2DeltaY;
+ uint32_t Scoreboard2DeltaX;
+ uint32_t Scoreboard1DeltaY;
+ uint32_t Scoreboard1DeltaX;
+ uint32_t Scoreboard0DeltaY;
+ uint32_t Scoreboard0DeltaX;
+ uint32_t Scoreboard7DeltaY;
+ uint32_t Scoreboard7DeltaX;
+ uint32_t Scoreboard6DeltaY;
+ uint32_t Scoreboard6DeltaX;
+ uint32_t Scoreboard5DeltaY;
+ uint32_t Scoreboard5DeltaX;
+ uint32_t Scoreboard4DeltaY;
+ uint32_t Scoreboard4DeltaX;
+};
+
+static inline void
+GEN7_MEDIA_VFE_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MEDIA_VFE_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->ScratchSpaceBasePointer, 10, 31) |
+ __gen_field(values->PerThreadScratchSpace, 0, 3) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->MaximumNumberofThreads, 16, 31) |
+ __gen_field(values->NumberofURBEntries, 8, 15) |
+ __gen_field(values->ResetGatewayTimer, 7, 7) |
+ __gen_field(values->BypassGatewayControl, 6, 6) |
+ __gen_field(values->GatewayMMIOAccessControl, 3, 4) |
+ __gen_field(values->GPGPUMode, 2, 2) |
+ 0;
+
+ dw[3] =
+ 0;
+
+ dw[4] =
+ __gen_field(values->URBEntryAllocationSize, 16, 31) |
+ __gen_field(values->CURBEAllocationSize, 0, 15) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->ScoreboardEnable, 31, 31) |
+ __gen_field(values->ScoreboardType, 30, 30) |
+ __gen_field(values->ScoreboardMask, 0, 7) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->Scoreboard3DeltaY, 28, 31) |
+ __gen_field(values->Scoreboard3DeltaX, 24, 27) |
+ __gen_field(values->Scoreboard2DeltaY, 20, 23) |
+ __gen_field(values->Scoreboard2DeltaX, 16, 19) |
+ __gen_field(values->Scoreboard1DeltaY, 12, 15) |
+ __gen_field(values->Scoreboard1DeltaX, 8, 11) |
+ __gen_field(values->Scoreboard0DeltaY, 4, 7) |
+ __gen_field(values->Scoreboard0DeltaX, 0, 3) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->Scoreboard7DeltaY, 28, 31) |
+ __gen_field(values->Scoreboard7DeltaX, 24, 27) |
+ __gen_field(values->Scoreboard6DeltaY, 20, 23) |
+ __gen_field(values->Scoreboard6DeltaX, 16, 19) |
+ __gen_field(values->Scoreboard5DeltaY, 12, 15) |
+ __gen_field(values->Scoreboard5DeltaX, 8, 11) |
+ __gen_field(values->Scoreboard4DeltaY, 4, 7) |
+ __gen_field(values->Scoreboard4DeltaX, 0, 3) |
+ 0;
+
+}
+
+#define GEN7_MI_ARB_CHECK_length_bias 0x00000001
+#define GEN7_MI_ARB_CHECK_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 5
+
+#define GEN7_MI_ARB_CHECK_length 0x00000001
+
+struct GEN7_MI_ARB_CHECK {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+};
+
+static inline void
+GEN7_MI_ARB_CHECK_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MI_ARB_CHECK * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ 0;
+
+}
+
+#define GEN7_MI_ARB_ON_OFF_length_bias 0x00000001
+#define GEN7_MI_ARB_ON_OFF_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 8
+
+#define GEN7_MI_ARB_ON_OFF_length 0x00000001
+
+struct GEN7_MI_ARB_ON_OFF {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool ArbitrationEnable;
+};
+
+static inline void
+GEN7_MI_ARB_ON_OFF_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MI_ARB_ON_OFF * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->ArbitrationEnable, 0, 0) |
+ 0;
+
+}
+
+#define GEN7_MI_BATCH_BUFFER_END_length_bias 0x00000001
+#define GEN7_MI_BATCH_BUFFER_END_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 10
+
+#define GEN7_MI_BATCH_BUFFER_END_length 0x00000001
+
+struct GEN7_MI_BATCH_BUFFER_END {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+};
+
+static inline void
+GEN7_MI_BATCH_BUFFER_END_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MI_BATCH_BUFFER_END * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ 0;
+
+}
+
+#define GEN7_MI_BATCH_BUFFER_START_length_bias 0x00000002
+#define GEN7_MI_BATCH_BUFFER_START_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 49, \
+ .DwordLength = 0
+
+#define GEN7_MI_BATCH_BUFFER_START_length 0x00000002
+
+struct GEN7_MI_BATCH_BUFFER_START {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool ClearCommandBufferEnable;
+#define ASI_GGTT 0
+#define ASI_PPGTT 1
+ uint32_t AddressSpaceIndicator;
+ uint32_t DwordLength;
+ __gen_address_type BatchBufferStartAddress;
+};
+
+static inline void
+GEN7_MI_BATCH_BUFFER_START_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MI_BATCH_BUFFER_START * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->ClearCommandBufferEnable, 11, 11) |
+ __gen_field(values->AddressSpaceIndicator, 8, 8) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw1 =
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->BatchBufferStartAddress, dw1);
+
+}
+
+#define GEN7_MI_CLFLUSH_length_bias 0x00000002
+#define GEN7_MI_CLFLUSH_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 39
+
+#define GEN7_MI_CLFLUSH_length 0x00000000
+
+struct GEN7_MI_CLFLUSH {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define PerProcessGraphicsAddress 0
+#define GlobalGraphicsAddress 1
+ uint32_t UseGlobalGTT;
+ uint32_t DwordLength;
+ __gen_address_type PageBaseAddress;
+ uint32_t StartingCachelineOffset;
+ __gen_address_type PageBaseAddressHigh;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN7_MI_CLFLUSH_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MI_CLFLUSH * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->UseGlobalGTT, 22, 22) |
+ __gen_field(values->DwordLength, 0, 9) |
+ 0;
+
+ uint32_t dw1 =
+ __gen_field(values->StartingCachelineOffset, 6, 11) |
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->PageBaseAddress, dw1);
+
+ uint32_t dw2 =
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->PageBaseAddressHigh, dw2);
+
+ /* variable length fields follow */
+}
+
+#define GEN7_MI_CONDITIONAL_BATCH_BUFFER_END_length_bias 0x00000002
+#define GEN7_MI_CONDITIONAL_BATCH_BUFFER_END_header\
+ .CommandType = 0, \
+ .MICommandOpcode = 54, \
+ .UseGlobalGTT = 0, \
+ .CompareSemaphore = 0, \
+ .DwordLength = 0
+
+#define GEN7_MI_CONDITIONAL_BATCH_BUFFER_END_length 0x00000002
+
+struct GEN7_MI_CONDITIONAL_BATCH_BUFFER_END {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t UseGlobalGTT;
+ uint32_t CompareSemaphore;
+ uint32_t DwordLength;
+ uint32_t CompareDataDword;
+ __gen_address_type CompareAddress;
+};
+
+static inline void
+GEN7_MI_CONDITIONAL_BATCH_BUFFER_END_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MI_CONDITIONAL_BATCH_BUFFER_END * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->UseGlobalGTT, 22, 22) |
+ __gen_field(values->CompareSemaphore, 21, 21) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->CompareDataDword, 0, 31) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->CompareAddress, dw2);
+
+}
+
+#define GEN7_MI_FLUSH_length_bias 0x00000001
+#define GEN7_MI_FLUSH_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 4
+
+#define GEN7_MI_FLUSH_length 0x00000001
+
+struct GEN7_MI_FLUSH {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool IndirectStatePointersDisable;
+ bool GenericMediaStateClear;
+#define DontReset 0
+#define Reset 1
+ bool GlobalSnapshotCountReset;
+#define Flush 0
+#define DontFlush 1
+ bool RenderCacheFlushInhibit;
+#define DontInvalidate 0
+#define Invalidate 1
+ bool StateInstructionCacheInvalidate;
+};
+
+static inline void
+GEN7_MI_FLUSH_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MI_FLUSH * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->IndirectStatePointersDisable, 5, 5) |
+ __gen_field(values->GenericMediaStateClear, 4, 4) |
+ __gen_field(values->GlobalSnapshotCountReset, 3, 3) |
+ __gen_field(values->RenderCacheFlushInhibit, 2, 2) |
+ __gen_field(values->StateInstructionCacheInvalidate, 1, 1) |
+ 0;
+
+}
+
+#define GEN7_MI_LOAD_REGISTER_IMM_length_bias 0x00000002
+#define GEN7_MI_LOAD_REGISTER_IMM_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 34, \
+ .DwordLength = 1
+
+#define GEN7_MI_LOAD_REGISTER_IMM_length 0x00000003
+
+struct GEN7_MI_LOAD_REGISTER_IMM {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t ByteWriteDisables;
+ uint32_t DwordLength;
+ uint32_t RegisterOffset;
+ uint32_t DataDWord;
+};
+
+static inline void
+GEN7_MI_LOAD_REGISTER_IMM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MI_LOAD_REGISTER_IMM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->ByteWriteDisables, 8, 11) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->RegisterOffset, 2, 22) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->DataDWord, 0, 31) |
+ 0;
+
+}
+
+#define GEN7_MI_LOAD_REGISTER_MEM_length_bias 0x00000002
+#define GEN7_MI_LOAD_REGISTER_MEM_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 41, \
+ .DwordLength = 1
+
+#define GEN7_MI_LOAD_REGISTER_MEM_length 0x00000003
+
+struct GEN7_MI_LOAD_REGISTER_MEM {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool UseGlobalGTT;
+ uint32_t AsyncModeEnable;
+ uint32_t DwordLength;
+ uint32_t RegisterAddress;
+ __gen_address_type MemoryAddress;
+};
+
+static inline void
+GEN7_MI_LOAD_REGISTER_MEM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MI_LOAD_REGISTER_MEM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->UseGlobalGTT, 22, 22) |
+ __gen_field(values->AsyncModeEnable, 21, 21) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->RegisterAddress, 2, 22) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->MemoryAddress, dw2);
+
+}
+
+#define GEN7_MI_NOOP_length_bias 0x00000001
+#define GEN7_MI_NOOP_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 0
+
+#define GEN7_MI_NOOP_length 0x00000001
+
+struct GEN7_MI_NOOP {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool IdentificationNumberRegisterWriteEnable;
+ uint32_t IdentificationNumber;
+};
+
+static inline void
+GEN7_MI_NOOP_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MI_NOOP * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->IdentificationNumberRegisterWriteEnable, 22, 22) |
+ __gen_field(values->IdentificationNumber, 0, 21) |
+ 0;
+
+}
+
+#define GEN7_MI_PREDICATE_length_bias 0x00000001
+#define GEN7_MI_PREDICATE_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 12
+
+#define GEN7_MI_PREDICATE_length 0x00000001
+
+struct GEN7_MI_PREDICATE {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define LOAD_KEEP 0
+#define LOAD_LOAD 2
+#define LOAD_LOADINV 3
+ uint32_t LoadOperation;
+#define COMBINE_SET 0
+#define COMBINE_AND 1
+#define COMBINE_OR 2
+#define COMBINE_XOR 3
+ uint32_t CombineOperation;
+#define COMPARE_SRCS_EQUAL 2
+#define COMPARE_DELTAS_EQUAL 3
+ uint32_t CompareOperation;
+};
+
+static inline void
+GEN7_MI_PREDICATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MI_PREDICATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->LoadOperation, 6, 7) |
+ __gen_field(values->CombineOperation, 3, 4) |
+ __gen_field(values->CompareOperation, 0, 1) |
+ 0;
+
+}
+
+#define GEN7_MI_REPORT_HEAD_length_bias 0x00000001
+#define GEN7_MI_REPORT_HEAD_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 7
+
+#define GEN7_MI_REPORT_HEAD_length 0x00000001
+
+struct GEN7_MI_REPORT_HEAD {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+};
+
+static inline void
+GEN7_MI_REPORT_HEAD_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MI_REPORT_HEAD * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ 0;
+
+}
+
+#define GEN7_MI_SEMAPHORE_MBOX_length_bias 0x00000002
+#define GEN7_MI_SEMAPHORE_MBOX_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 22, \
+ .DwordLength = 1
+
+#define GEN7_MI_SEMAPHORE_MBOX_length 0x00000003
+
+struct GEN7_MI_SEMAPHORE_MBOX {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define RVSYNC 0
+#define RBSYNC 2
+#define UseGeneralRegisterSelect 3
+ uint32_t RegisterSelect;
+ uint32_t DwordLength;
+ uint32_t SemaphoreDataDword;
+};
+
+static inline void
+GEN7_MI_SEMAPHORE_MBOX_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MI_SEMAPHORE_MBOX * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->RegisterSelect, 16, 17) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->SemaphoreDataDword, 0, 31) |
+ 0;
+
+ dw[2] =
+ 0;
+
+}
+
+#define GEN7_MI_SET_CONTEXT_length_bias 0x00000002
+#define GEN7_MI_SET_CONTEXT_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 24, \
+ .DwordLength = 0
+
+#define GEN7_MI_SET_CONTEXT_length 0x00000002
+
+struct GEN7_MI_SET_CONTEXT {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t DwordLength;
+ __gen_address_type LogicalContextAddress;
+ uint32_t ReservedMustbe1;
+ bool ExtendedStateSaveEnable;
+ bool ExtendedStateRestoreEnable;
+ uint32_t ForceRestore;
+ uint32_t RestoreInhibit;
+};
+
+static inline void
+GEN7_MI_SET_CONTEXT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MI_SET_CONTEXT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw1 =
+ __gen_field(values->ReservedMustbe1, 8, 8) |
+ __gen_field(values->ExtendedStateSaveEnable, 3, 3) |
+ __gen_field(values->ExtendedStateRestoreEnable, 2, 2) |
+ __gen_field(values->ForceRestore, 1, 1) |
+ __gen_field(values->RestoreInhibit, 0, 0) |
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->LogicalContextAddress, dw1);
+
+}
+
+#define GEN7_MI_STORE_DATA_IMM_length_bias 0x00000002
+#define GEN7_MI_STORE_DATA_IMM_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 32, \
+ .DwordLength = 2
+
+#define GEN7_MI_STORE_DATA_IMM_length 0x00000004
+
+struct GEN7_MI_STORE_DATA_IMM {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool UseGlobalGTT;
+ uint32_t DwordLength;
+ uint32_t Address;
+ uint32_t CoreModeEnable;
+ uint32_t DataDWord0;
+ uint32_t DataDWord1;
+};
+
+static inline void
+GEN7_MI_STORE_DATA_IMM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MI_STORE_DATA_IMM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->UseGlobalGTT, 22, 22) |
+ __gen_field(values->DwordLength, 0, 5) |
+ 0;
+
+ dw[1] =
+ 0;
+
+ dw[2] =
+ __gen_field(values->Address, 2, 31) |
+ __gen_field(values->CoreModeEnable, 0, 0) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->DataDWord0, 0, 31) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->DataDWord1, 0, 31) |
+ 0;
+
+}
+
+#define GEN7_MI_STORE_DATA_INDEX_length_bias 0x00000002
+#define GEN7_MI_STORE_DATA_INDEX_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 33, \
+ .DwordLength = 1
+
+#define GEN7_MI_STORE_DATA_INDEX_length 0x00000003
+
+struct GEN7_MI_STORE_DATA_INDEX {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t DwordLength;
+ uint32_t Offset;
+ uint32_t DataDWord0;
+ uint32_t DataDWord1;
+};
+
+static inline void
+GEN7_MI_STORE_DATA_INDEX_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MI_STORE_DATA_INDEX * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->Offset, 2, 11) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->DataDWord0, 0, 31) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->DataDWord1, 0, 31) |
+ 0;
+
+}
+
+#define GEN7_MI_SUSPEND_FLUSH_length_bias 0x00000001
+#define GEN7_MI_SUSPEND_FLUSH_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 11
+
+#define GEN7_MI_SUSPEND_FLUSH_length 0x00000001
+
+struct GEN7_MI_SUSPEND_FLUSH {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool SuspendFlush;
+};
+
+static inline void
+GEN7_MI_SUSPEND_FLUSH_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MI_SUSPEND_FLUSH * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->SuspendFlush, 0, 0) |
+ 0;
+
+}
+
+#define GEN7_MI_TOPOLOGY_FILTER_length_bias 0x00000001
+#define GEN7_MI_TOPOLOGY_FILTER_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 13
+
+#define GEN7_MI_TOPOLOGY_FILTER_length 0x00000001
+
+struct GEN7_MI_TOPOLOGY_FILTER {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t TopologyFilterValue;
+};
+
+static inline void
+GEN7_MI_TOPOLOGY_FILTER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MI_TOPOLOGY_FILTER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->TopologyFilterValue, 0, 5) |
+ 0;
+
+}
+
+#define GEN7_MI_UPDATE_GTT_length_bias 0x00000002
+#define GEN7_MI_UPDATE_GTT_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 35
+
+#define GEN7_MI_UPDATE_GTT_length 0x00000000
+
+struct GEN7_MI_UPDATE_GTT {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define PerProcessGraphicsAddress 0
+#define GlobalGraphicsAddress 1
+ uint32_t UseGlobalGTT;
+ uint32_t DwordLength;
+ __gen_address_type EntryAddress;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN7_MI_UPDATE_GTT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MI_UPDATE_GTT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->UseGlobalGTT, 22, 22) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw1 =
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->EntryAddress, dw1);
+
+ /* variable length fields follow */
+}
+
+#define GEN7_MI_URB_CLEAR_length_bias 0x00000002
+#define GEN7_MI_URB_CLEAR_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 25, \
+ .DwordLength = 0
+
+#define GEN7_MI_URB_CLEAR_length 0x00000002
+
+struct GEN7_MI_URB_CLEAR {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t DwordLength;
+ uint32_t URBClearLength;
+ uint32_t URBAddress;
+};
+
+static inline void
+GEN7_MI_URB_CLEAR_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MI_URB_CLEAR * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->URBClearLength, 16, 28) |
+ __gen_offset(values->URBAddress, 0, 13) |
+ 0;
+
+}
+
+#define GEN7_MI_USER_INTERRUPT_length_bias 0x00000001
+#define GEN7_MI_USER_INTERRUPT_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 2
+
+#define GEN7_MI_USER_INTERRUPT_length 0x00000001
+
+struct GEN7_MI_USER_INTERRUPT {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+};
+
+static inline void
+GEN7_MI_USER_INTERRUPT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MI_USER_INTERRUPT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ 0;
+
+}
+
+#define GEN7_MI_WAIT_FOR_EVENT_length_bias 0x00000001
+#define GEN7_MI_WAIT_FOR_EVENT_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 3
+
+#define GEN7_MI_WAIT_FOR_EVENT_length 0x00000001
+
+struct GEN7_MI_WAIT_FOR_EVENT {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool DisplayPipeCHorizontalBlankWaitEnable;
+ bool DisplayPipeCVerticalBlankWaitEnable;
+ bool DisplaySpriteCFlipPendingWaitEnable;
+#define Notenabled 0
+ uint32_t ConditionCodeWaitSelect;
+ bool DisplayPlaneCFlipPendingWaitEnable;
+ bool DisplayPipeCScanLineWaitEnable;
+ bool DisplayPipeBHorizontalBlankWaitEnable;
+ bool DisplayPipeBVerticalBlankWaitEnable;
+ bool DisplaySpriteBFlipPendingWaitEnable;
+ bool DisplayPlaneBFlipPendingWaitEnable;
+ bool DisplayPipeBScanLineWaitEnable;
+ bool DisplayPipeAHorizontalBlankWaitEnable;
+ bool DisplayPipeAVerticalBlankWaitEnable;
+ bool DisplaySpriteAFlipPendingWaitEnable;
+ bool DisplayPlaneAFlipPendingWaitEnable;
+ bool DisplayPipeAScanLineWaitEnable;
+};
+
+static inline void
+GEN7_MI_WAIT_FOR_EVENT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_MI_WAIT_FOR_EVENT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DisplayPipeCHorizontalBlankWaitEnable, 22, 22) |
+ __gen_field(values->DisplayPipeCVerticalBlankWaitEnable, 21, 21) |
+ __gen_field(values->DisplaySpriteCFlipPendingWaitEnable, 20, 20) |
+ __gen_field(values->ConditionCodeWaitSelect, 16, 19) |
+ __gen_field(values->DisplayPlaneCFlipPendingWaitEnable, 15, 15) |
+ __gen_field(values->DisplayPipeCScanLineWaitEnable, 14, 14) |
+ __gen_field(values->DisplayPipeBHorizontalBlankWaitEnable, 13, 13) |
+ __gen_field(values->DisplayPipeBVerticalBlankWaitEnable, 11, 11) |
+ __gen_field(values->DisplaySpriteBFlipPendingWaitEnable, 10, 10) |
+ __gen_field(values->DisplayPlaneBFlipPendingWaitEnable, 9, 9) |
+ __gen_field(values->DisplayPipeBScanLineWaitEnable, 8, 8) |
+ __gen_field(values->DisplayPipeAHorizontalBlankWaitEnable, 5, 5) |
+ __gen_field(values->DisplayPipeAVerticalBlankWaitEnable, 3, 3) |
+ __gen_field(values->DisplaySpriteAFlipPendingWaitEnable, 2, 2) |
+ __gen_field(values->DisplayPlaneAFlipPendingWaitEnable, 1, 1) |
+ __gen_field(values->DisplayPipeAScanLineWaitEnable, 0, 0) |
+ 0;
+
+}
+
+#define GEN7_PIPE_CONTROL_length_bias 0x00000002
+#define GEN7_PIPE_CONTROL_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 2, \
+ ._3DCommandSubOpcode = 0, \
+ .DwordLength = 3
+
+#define GEN7_PIPE_CONTROL_length 0x00000005
+
+struct GEN7_PIPE_CONTROL {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define DAT_PPGTT 0
+#define DAT_GGTT 1
+ uint32_t DestinationAddressType;
+#define NoLRIOperation 0
+#define MMIOWriteImmediateData 1
+ uint32_t LRIPostSyncOperation;
+ uint32_t StoreDataIndex;
+ uint32_t CommandStreamerStallEnable;
+#define DontReset 0
+#define Reset 1
+ uint32_t GlobalSnapshotCountReset;
+ uint32_t TLBInvalidate;
+ bool GenericMediaStateClear;
+#define NoWrite 0
+#define WriteImmediateData 1
+#define WritePSDepthCount 2
+#define WriteTimestamp 3
+ uint32_t PostSyncOperation;
+ bool DepthStallEnable;
+#define DisableFlush 0
+#define EnableFlush 1
+ bool RenderTargetCacheFlushEnable;
+ bool InstructionCacheInvalidateEnable;
+ bool TextureCacheInvalidationEnable;
+ bool IndirectStatePointersDisable;
+ bool NotifyEnable;
+ bool PipeControlFlushEnable;
+ bool DCFlushEnable;
+ bool VFCacheInvalidationEnable;
+ bool ConstantCacheInvalidationEnable;
+ bool StateCacheInvalidationEnable;
+ bool StallAtPixelScoreboard;
+#define FlushDisabled 0
+#define FlushEnabled 1
+ bool DepthCacheFlushEnable;
+ __gen_address_type Address;
+ uint32_t ImmediateData;
+ uint32_t ImmediateData0;
+};
+
+static inline void
+GEN7_PIPE_CONTROL_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_PIPE_CONTROL * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->DestinationAddressType, 24, 24) |
+ __gen_field(values->LRIPostSyncOperation, 23, 23) |
+ __gen_field(values->StoreDataIndex, 21, 21) |
+ __gen_field(values->CommandStreamerStallEnable, 20, 20) |
+ __gen_field(values->GlobalSnapshotCountReset, 19, 19) |
+ __gen_field(values->TLBInvalidate, 18, 18) |
+ __gen_field(values->GenericMediaStateClear, 16, 16) |
+ __gen_field(values->PostSyncOperation, 14, 15) |
+ __gen_field(values->DepthStallEnable, 13, 13) |
+ __gen_field(values->RenderTargetCacheFlushEnable, 12, 12) |
+ __gen_field(values->InstructionCacheInvalidateEnable, 11, 11) |
+ __gen_field(values->TextureCacheInvalidationEnable, 10, 10) |
+ __gen_field(values->IndirectStatePointersDisable, 9, 9) |
+ __gen_field(values->NotifyEnable, 8, 8) |
+ __gen_field(values->PipeControlFlushEnable, 7, 7) |
+ __gen_field(values->DCFlushEnable, 5, 5) |
+ __gen_field(values->VFCacheInvalidationEnable, 4, 4) |
+ __gen_field(values->ConstantCacheInvalidationEnable, 3, 3) |
+ __gen_field(values->StateCacheInvalidationEnable, 2, 2) |
+ __gen_field(values->StallAtPixelScoreboard, 1, 1) |
+ __gen_field(values->DepthCacheFlushEnable, 0, 0) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->Address, dw2);
+
+ dw[3] =
+ __gen_field(values->ImmediateData, 0, 31) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->ImmediateData, 0, 31) |
+ 0;
+
+}
+
+#define GEN7_SCISSOR_RECT_length 0x00000002
+
+struct GEN7_SCISSOR_RECT {
+ uint32_t ScissorRectangleYMin;
+ uint32_t ScissorRectangleXMin;
+ uint32_t ScissorRectangleYMax;
+ uint32_t ScissorRectangleXMax;
+};
+
+static inline void
+GEN7_SCISSOR_RECT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_SCISSOR_RECT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->ScissorRectangleYMin, 16, 31) |
+ __gen_field(values->ScissorRectangleXMin, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ScissorRectangleYMax, 16, 31) |
+ __gen_field(values->ScissorRectangleXMax, 0, 15) |
+ 0;
+
+}
+
+#define GEN7_SF_CLIP_VIEWPORT_length 0x00000010
+
+struct GEN7_SF_CLIP_VIEWPORT {
+ float ViewportMatrixElementm00;
+ float ViewportMatrixElementm11;
+ float ViewportMatrixElementm22;
+ float ViewportMatrixElementm30;
+ float ViewportMatrixElementm31;
+ float ViewportMatrixElementm32;
+ float XMinClipGuardband;
+ float XMaxClipGuardband;
+ float YMinClipGuardband;
+ float YMaxClipGuardband;
+};
+
+static inline void
+GEN7_SF_CLIP_VIEWPORT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_SF_CLIP_VIEWPORT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_float(values->ViewportMatrixElementm00) |
+ 0;
+
+ dw[1] =
+ __gen_float(values->ViewportMatrixElementm11) |
+ 0;
+
+ dw[2] =
+ __gen_float(values->ViewportMatrixElementm22) |
+ 0;
+
+ dw[3] =
+ __gen_float(values->ViewportMatrixElementm30) |
+ 0;
+
+ dw[4] =
+ __gen_float(values->ViewportMatrixElementm31) |
+ 0;
+
+ dw[5] =
+ __gen_float(values->ViewportMatrixElementm32) |
+ 0;
+
+ dw[6] =
+ 0;
+
+ dw[7] =
+ 0;
+
+ dw[8] =
+ __gen_float(values->XMinClipGuardband) |
+ 0;
+
+ dw[9] =
+ __gen_float(values->XMaxClipGuardband) |
+ 0;
+
+ dw[10] =
+ __gen_float(values->YMinClipGuardband) |
+ 0;
+
+ dw[11] =
+ __gen_float(values->YMaxClipGuardband) |
+ 0;
+
+ for (uint32_t i = 0, j = 12; i < 4; i += 1, j++) {
+ dw[j] =
+ 0;
+ }
+
+}
+
+#define GEN7_BLEND_STATE_length 0x00000002
+
+struct GEN7_BLEND_STATE {
+ bool ColorBufferBlendEnable;
+ bool IndependentAlphaBlendEnable;
+#define BLENDFUNCTION_ADD 0
+#define BLENDFUNCTION_SUBTRACT 1
+#define BLENDFUNCTION_REVERSE_SUBTRACT 2
+#define BLENDFUNCTION_MIN 3
+#define BLENDFUNCTION_MAX 4
+ uint32_t AlphaBlendFunction;
+#define BLENDFACTOR_ONE 1
+#define BLENDFACTOR_SRC_COLOR 2
+#define BLENDFACTOR_SRC_ALPHA 3
+#define BLENDFACTOR_DST_ALPHA 4
+#define BLENDFACTOR_DST_COLOR 5
+#define BLENDFACTOR_SRC_ALPHA_SATURATE 6
+#define BLENDFACTOR_CONST_COLOR 7
+#define BLENDFACTOR_CONST_ALPHA 8
+#define BLENDFACTOR_SRC1_COLOR 9
+#define BLENDFACTOR_SRC1_ALPHA 10
+#define BLENDFACTOR_ZERO 17
+#define BLENDFACTOR_INV_SRC_COLOR 18
+#define BLENDFACTOR_INV_SRC_ALPHA 19
+#define BLENDFACTOR_INV_DST_ALPHA 20
+#define BLENDFACTOR_INV_DST_COLOR 21
+#define BLENDFACTOR_INV_CONST_COLOR 23
+#define BLENDFACTOR_INV_CONST_ALPHA 24
+#define BLENDFACTOR_INV_SRC1_COLOR 25
+#define BLENDFACTOR_INV_SRC1_ALPHA 26
+ uint32_t SourceAlphaBlendFactor;
+ uint32_t DestinationAlphaBlendFactor;
+#define BLENDFUNCTION_ADD 0
+#define BLENDFUNCTION_SUBTRACT 1
+#define BLENDFUNCTION_REVERSE_SUBTRACT 2
+#define BLENDFUNCTION_MIN 3
+#define BLENDFUNCTION_MAX 4
+ uint32_t ColorBlendFunction;
+ uint32_t SourceBlendFactor;
+ uint32_t DestinationBlendFactor;
+ bool AlphaToCoverageEnable;
+ bool AlphaToOneEnable;
+ bool AlphaToCoverageDitherEnable;
+ bool WriteDisableAlpha;
+ bool WriteDisableRed;
+ bool WriteDisableGreen;
+ bool WriteDisableBlue;
+ bool LogicOpEnable;
+#define LOGICOP_CLEAR 0
+#define LOGICOP_NOR 1
+#define LOGICOP_AND_INVERTED 2
+#define LOGICOP_COPY_INVERTED 3
+#define LOGICOP_AND_REVERSE 4
+#define LOGICOP_INVERT 5
+#define LOGICOP_XOR 6
+#define LOGICOP_NAND 7
+#define LOGICOP_AND 8
+#define LOGICOP_EQUIV 9
+#define LOGICOP_NOOP 10
+#define LOGICOP_OR_INVERTED 11
+#define LOGICOP_COPY 12
+#define LOGICOP_OR_REVERSE 13
+#define LOGICOP_OR 14
+#define LOGICOP_SET 15
+ uint32_t LogicOpFunction;
+ bool AlphaTestEnable;
+#define COMPAREFUNCTION_ALWAYS 0
+#define COMPAREFUNCTION_NEVER 1
+#define COMPAREFUNCTION_LESS 2
+#define COMPAREFUNCTION_EQUAL 3
+#define COMPAREFUNCTION_LEQUAL 4
+#define COMPAREFUNCTION_GREATER 5
+#define COMPAREFUNCTION_NOTEQUAL 6
+#define COMPAREFUNCTION_GEQUAL 7
+ uint32_t AlphaTestFunction;
+ bool ColorDitherEnable;
+ uint32_t XDitherOffset;
+ uint32_t YDitherOffset;
+#define COLORCLAMP_UNORM 0
+#define COLORCLAMP_SNORM 1
+#define COLORCLAMP_RTFORMAT 2
+ uint32_t ColorClampRange;
+ bool PreBlendColorClampEnable;
+ bool PostBlendColorClampEnable;
+};
+
+static inline void
+GEN7_BLEND_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_BLEND_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->ColorBufferBlendEnable, 31, 31) |
+ __gen_field(values->IndependentAlphaBlendEnable, 30, 30) |
+ __gen_field(values->AlphaBlendFunction, 26, 28) |
+ __gen_field(values->SourceAlphaBlendFactor, 20, 24) |
+ __gen_field(values->DestinationAlphaBlendFactor, 15, 19) |
+ __gen_field(values->ColorBlendFunction, 11, 13) |
+ __gen_field(values->SourceBlendFactor, 5, 9) |
+ __gen_field(values->DestinationBlendFactor, 0, 4) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->AlphaToCoverageEnable, 31, 31) |
+ __gen_field(values->AlphaToOneEnable, 30, 30) |
+ __gen_field(values->AlphaToCoverageDitherEnable, 29, 29) |
+ __gen_field(values->WriteDisableAlpha, 27, 27) |
+ __gen_field(values->WriteDisableRed, 26, 26) |
+ __gen_field(values->WriteDisableGreen, 25, 25) |
+ __gen_field(values->WriteDisableBlue, 24, 24) |
+ __gen_field(values->LogicOpEnable, 22, 22) |
+ __gen_field(values->LogicOpFunction, 18, 21) |
+ __gen_field(values->AlphaTestEnable, 16, 16) |
+ __gen_field(values->AlphaTestFunction, 13, 15) |
+ __gen_field(values->ColorDitherEnable, 12, 12) |
+ __gen_field(values->XDitherOffset, 10, 11) |
+ __gen_field(values->YDitherOffset, 8, 9) |
+ __gen_field(values->ColorClampRange, 2, 3) |
+ __gen_field(values->PreBlendColorClampEnable, 1, 1) |
+ __gen_field(values->PostBlendColorClampEnable, 0, 0) |
+ 0;
+
+}
+
+#define GEN7_CC_VIEWPORT_length 0x00000002
+
+struct GEN7_CC_VIEWPORT {
+ float MinimumDepth;
+ float MaximumDepth;
+};
+
+static inline void
+GEN7_CC_VIEWPORT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_CC_VIEWPORT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_float(values->MinimumDepth) |
+ 0;
+
+ dw[1] =
+ __gen_float(values->MaximumDepth) |
+ 0;
+
+}
+
+#define GEN7_COLOR_CALC_STATE_length 0x00000006
+
+struct GEN7_COLOR_CALC_STATE {
+ uint32_t StencilReferenceValue;
+ uint32_t BackFaceStencilReferenceValue;
+#define Cancelled 0
+#define NotCancelled 1
+ uint32_t RoundDisableFunctionDisable;
+#define ALPHATEST_UNORM8 0
+#define ALPHATEST_FLOAT32 1
+ uint32_t AlphaTestFormat;
+ uint32_t AlphaReferenceValueAsUNORM8;
+ float AlphaReferenceValueAsFLOAT32;
+ float BlendConstantColorRed;
+ float BlendConstantColorGreen;
+ float BlendConstantColorBlue;
+ float BlendConstantColorAlpha;
+};
+
+static inline void
+GEN7_COLOR_CALC_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_COLOR_CALC_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->StencilReferenceValue, 24, 31) |
+ __gen_field(values->BackFaceStencilReferenceValue, 16, 23) |
+ __gen_field(values->RoundDisableFunctionDisable, 15, 15) |
+ __gen_field(values->AlphaTestFormat, 0, 0) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->AlphaReferenceValueAsUNORM8, 0, 31) |
+ __gen_float(values->AlphaReferenceValueAsFLOAT32) |
+ 0;
+
+ dw[2] =
+ __gen_float(values->BlendConstantColorRed) |
+ 0;
+
+ dw[3] =
+ __gen_float(values->BlendConstantColorGreen) |
+ 0;
+
+ dw[4] =
+ __gen_float(values->BlendConstantColorBlue) |
+ 0;
+
+ dw[5] =
+ __gen_float(values->BlendConstantColorAlpha) |
+ 0;
+
+}
+
+#define GEN7_DEPTH_STENCIL_STATE_length 0x00000003
+
+struct GEN7_DEPTH_STENCIL_STATE {
+ bool StencilTestEnable;
+#define COMPAREFUNCTION_ALWAYS 0
+#define COMPAREFUNCTION_NEVER 1
+#define COMPAREFUNCTION_LESS 2
+#define COMPAREFUNCTION_EQUAL 3
+#define COMPAREFUNCTION_LEQUAL 4
+#define COMPAREFUNCTION_GREATER 5
+#define COMPAREFUNCTION_NOTEQUAL 6
+#define COMPAREFUNCTION_GEQUAL 7
+ uint32_t StencilTestFunction;
+#define STENCILOP_KEEP 0
+#define STENCILOP_ZERO 1
+#define STENCILOP_REPLACE 2
+#define STENCILOP_INCRSAT 3
+#define STENCILOP_DECRSAT 4
+#define STENCILOP_INCR 5
+#define STENCILOP_DECR 6
+#define STENCILOP_INVERT 7
+ uint32_t StencilFailOp;
+ uint32_t StencilPassDepthFailOp;
+ uint32_t StencilPassDepthPassOp;
+ bool StencilBufferWriteEnable;
+ bool DoubleSidedStencilEnable;
+#define COMPAREFUNCTION_ALWAYS 0
+#define COMPAREFUNCTION_NEVER 1
+#define COMPAREFUNCTION_LESS 2
+#define COMPAREFUNCTION_EQUAL 3
+#define COMPAREFUNCTION_LEQUAL 4
+#define COMPAREFUNCTION_GREATER 5
+#define COMPAREFUNCTION_NOTEQUAL 6
+#define COMPAREFUNCTION_GEQUAL 7
+ uint32_t BackFaceStencilTestFunction;
+#define STENCILOP_KEEP 0
+#define STENCILOP_ZERO 1
+#define STENCILOP_REPLACE 2
+#define STENCILOP_INCRSAT 3
+#define STENCILOP_DECRSAT 4
+#define STENCILOP_INCR 5
+#define STENCILOP_DECR 6
+#define STENCILOP_INVERT 7
+ uint32_t BackfaceStencilFailOp;
+ uint32_t BackfaceStencilPassDepthFailOp;
+ uint32_t BackfaceStencilPassDepthPassOp;
+ uint32_t StencilTestMask;
+ uint32_t StencilWriteMask;
+ uint32_t BackfaceStencilTestMask;
+ uint32_t BackfaceStencilWriteMask;
+ bool DepthTestEnable;
+#define COMPAREFUNCTION_ALWAYS 0
+#define COMPAREFUNCTION_NEVER 1
+#define COMPAREFUNCTION_LESS 2
+#define COMPAREFUNCTION_EQUAL 3
+#define COMPAREFUNCTION_LEQUAL 4
+#define COMPAREFUNCTION_GREATER 5
+#define COMPAREFUNCTION_NOTEQUAL 6
+#define COMPAREFUNCTION_GEQUAL 7
+ uint32_t DepthTestFunction;
+ bool DepthBufferWriteEnable;
+};
+
+static inline void
+GEN7_DEPTH_STENCIL_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_DEPTH_STENCIL_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->StencilTestEnable, 31, 31) |
+ __gen_field(values->StencilTestFunction, 28, 30) |
+ __gen_field(values->StencilFailOp, 25, 27) |
+ __gen_field(values->StencilPassDepthFailOp, 22, 24) |
+ __gen_field(values->StencilPassDepthPassOp, 19, 21) |
+ __gen_field(values->StencilBufferWriteEnable, 18, 18) |
+ __gen_field(values->DoubleSidedStencilEnable, 15, 15) |
+ __gen_field(values->BackFaceStencilTestFunction, 12, 14) |
+ __gen_field(values->BackfaceStencilFailOp, 9, 11) |
+ __gen_field(values->BackfaceStencilPassDepthFailOp, 6, 8) |
+ __gen_field(values->BackfaceStencilPassDepthPassOp, 3, 5) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->StencilTestMask, 24, 31) |
+ __gen_field(values->StencilWriteMask, 16, 23) |
+ __gen_field(values->BackfaceStencilTestMask, 8, 15) |
+ __gen_field(values->BackfaceStencilWriteMask, 0, 7) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->DepthTestEnable, 31, 31) |
+ __gen_field(values->DepthTestFunction, 27, 29) |
+ __gen_field(values->DepthBufferWriteEnable, 26, 26) |
+ 0;
+
+}
+
+#define GEN7_INTERFACE_DESCRIPTOR_DATA_length 0x00000008
+
+struct GEN7_INTERFACE_DESCRIPTOR_DATA {
+ uint32_t KernelStartPointer;
+#define Multiple 0
+#define Single 1
+ uint32_t SingleProgramFlow;
+#define NormalPriority 0
+#define HighPriority 1
+ uint32_t ThreadPriority;
+#define IEEE754 0
+#define Alternate 1
+ uint32_t FloatingPointMode;
+ bool IllegalOpcodeExceptionEnable;
+ bool MaskStackExceptionEnable;
+ bool SoftwareExceptionEnable;
+ uint32_t SamplerStatePointer;
+#define Nosamplersused 0
+#define Between1and4samplersused 1
+#define Between5and8samplersused 2
+#define Between9and12samplersused 3
+#define Between13and16samplersused 4
+ uint32_t SamplerCount;
+ uint32_t BindingTablePointer;
+ uint32_t BindingTableEntryCount;
+ uint32_t ConstantURBEntryReadLength;
+ uint32_t ConstantURBEntryReadOffset;
+#define RTNE 0
+#define RU 1
+#define RD 2
+#define RTZ 3
+ uint32_t RoundingMode;
+ bool BarrierEnable;
+ uint32_t SharedLocalMemorySize;
+ uint32_t NumberofThreadsinGPGPUThreadGroup;
+};
+
+static inline void
+GEN7_INTERFACE_DESCRIPTOR_DATA_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_INTERFACE_DESCRIPTOR_DATA * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_offset(values->KernelStartPointer, 6, 31) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->SingleProgramFlow, 18, 18) |
+ __gen_field(values->ThreadPriority, 17, 17) |
+ __gen_field(values->FloatingPointMode, 16, 16) |
+ __gen_field(values->IllegalOpcodeExceptionEnable, 13, 13) |
+ __gen_field(values->MaskStackExceptionEnable, 11, 11) |
+ __gen_field(values->SoftwareExceptionEnable, 7, 7) |
+ 0;
+
+ dw[2] =
+ __gen_offset(values->SamplerStatePointer, 5, 31) |
+ __gen_field(values->SamplerCount, 2, 4) |
+ 0;
+
+ dw[3] =
+ __gen_offset(values->BindingTablePointer, 5, 15) |
+ __gen_field(values->BindingTableEntryCount, 0, 4) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->ConstantURBEntryReadLength, 16, 31) |
+ __gen_field(values->ConstantURBEntryReadOffset, 0, 15) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->RoundingMode, 22, 23) |
+ __gen_field(values->BarrierEnable, 21, 21) |
+ __gen_field(values->SharedLocalMemorySize, 16, 20) |
+ __gen_field(values->NumberofThreadsinGPGPUThreadGroup, 0, 7) |
+ 0;
+
+ dw[6] =
+ 0;
+
+ dw[7] =
+ 0;
+
+}
+
+#define GEN7_BINDING_TABLE_STATE_length 0x00000001
+
+struct GEN7_BINDING_TABLE_STATE {
+ uint32_t SurfaceStatePointer;
+};
+
+static inline void
+GEN7_BINDING_TABLE_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_BINDING_TABLE_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_offset(values->SurfaceStatePointer, 5, 31) |
+ 0;
+
+}
+
+#define GEN7_RENDER_SURFACE_STATE_length 0x00000008
+
+struct GEN7_RENDER_SURFACE_STATE {
+#define SURFTYPE_1D 0
+#define SURFTYPE_2D 1
+#define SURFTYPE_3D 2
+#define SURFTYPE_CUBE 3
+#define SURFTYPE_BUFFER 4
+#define SURFTYPE_STRBUF 5
+#define SURFTYPE_NULL 7
+ uint32_t SurfaceType;
+ bool SurfaceArray;
+ uint32_t SurfaceFormat;
+#define VALIGN_2 0
+#define VALIGN_4 1
+ uint32_t SurfaceVerticalAlignment;
+#define HALIGN_4 0
+#define HALIGN_8 1
+ uint32_t SurfaceHorizontalAlignment;
+ uint32_t TiledSurface;
+#define TILEWALK_XMAJOR 0
+#define TILEWALK_YMAJOR 1
+ uint32_t TileWalk;
+ uint32_t VerticalLineStride;
+ uint32_t VerticalLineStrideOffset;
+#define ARYSPC_FULL 0
+#define ARYSPC_LOD0 1
+ uint32_t SurfaceArraySpacing;
+ uint32_t RenderCacheReadWriteMode;
+#define NORMAL_MODE 0
+#define PROGRESSIVE_FRAME 2
+#define INTERLACED_FRAME 3
+ uint32_t MediaBoundaryPixelMode;
+ uint32_t CubeFaceEnables;
+ __gen_address_type SurfaceBaseAddress;
+ uint32_t Height;
+ uint32_t Width;
+ uint32_t Depth;
+ uint32_t SurfacePitch;
+#define RTROTATE_0DEG 0
+#define RTROTATE_90DEG 1
+#define RTROTATE_270DEG 3
+ uint32_t RenderTargetRotation;
+ uint32_t MinimumArrayElement;
+ uint32_t RenderTargetViewExtent;
+#define MSFMT_MSS 0
+#define MSFMT_DEPTH_STENCIL 1
+ uint32_t MultisampledSurfaceStorageFormat;
+#define MULTISAMPLECOUNT_1 0
+#define MULTISAMPLECOUNT_4 2
+#define MULTISAMPLECOUNT_8 3
+ uint32_t NumberofMultisamples;
+ uint32_t MultisamplePositionPaletteIndex;
+ uint32_t MinimumArrayElement0;
+ uint32_t XOffset;
+ uint32_t YOffset;
+ struct GEN7_MEMORY_OBJECT_CONTROL_STATE SurfaceObjectControlState;
+ uint32_t SurfaceMinLOD;
+ uint32_t MIPCountLOD;
+ __gen_address_type MCSBaseAddress;
+ uint32_t MCSSurfacePitch;
+ __gen_address_type AppendCounterAddress;
+ bool AppendCounterEnable;
+ bool MCSEnable;
+ uint32_t XOffsetforUVPlane;
+ uint32_t YOffsetforUVPlane;
+#define CC_ZERO 0
+#define CC_ONE 1
+ uint32_t RedClearColor;
+#define CC_ZERO 0
+#define CC_ONE 1
+ uint32_t GreenClearColor;
+#define CC_ZERO 0
+#define CC_ONE 1
+ uint32_t BlueClearColor;
+#define CC_ZERO 0
+#define CC_ONE 1
+ uint32_t AlphaClearColor;
+ float ResourceMinLOD;
+};
+
+static inline void
+GEN7_RENDER_SURFACE_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_RENDER_SURFACE_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->SurfaceType, 29, 31) |
+ __gen_field(values->SurfaceArray, 28, 28) |
+ __gen_field(values->SurfaceFormat, 18, 26) |
+ __gen_field(values->SurfaceVerticalAlignment, 16, 17) |
+ __gen_field(values->SurfaceHorizontalAlignment, 15, 15) |
+ __gen_field(values->TiledSurface, 14, 14) |
+ __gen_field(values->TileWalk, 13, 13) |
+ __gen_field(values->VerticalLineStride, 12, 12) |
+ __gen_field(values->VerticalLineStrideOffset, 11, 11) |
+ __gen_field(values->SurfaceArraySpacing, 10, 10) |
+ __gen_field(values->RenderCacheReadWriteMode, 8, 8) |
+ __gen_field(values->MediaBoundaryPixelMode, 6, 7) |
+ __gen_field(values->CubeFaceEnables, 0, 5) |
+ 0;
+
+ uint32_t dw1 =
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->SurfaceBaseAddress, dw1);
+
+ dw[2] =
+ __gen_field(values->Height, 16, 29) |
+ __gen_field(values->Width, 0, 13) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->Depth, 21, 31) |
+ __gen_field(values->SurfacePitch, 0, 17) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->RenderTargetRotation, 29, 30) |
+ __gen_field(values->MinimumArrayElement, 18, 28) |
+ __gen_field(values->RenderTargetViewExtent, 7, 17) |
+ __gen_field(values->MultisampledSurfaceStorageFormat, 6, 6) |
+ __gen_field(values->NumberofMultisamples, 3, 5) |
+ __gen_field(values->MultisamplePositionPaletteIndex, 0, 2) |
+ __gen_field(values->MinimumArrayElement, 0, 26) |
+ 0;
+
+ uint32_t dw_SurfaceObjectControlState;
+ GEN7_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_SurfaceObjectControlState, &values->SurfaceObjectControlState);
+ dw[5] =
+ __gen_offset(values->XOffset, 25, 31) |
+ __gen_offset(values->YOffset, 20, 23) |
+ __gen_field(dw_SurfaceObjectControlState, 16, 19) |
+ __gen_field(values->SurfaceMinLOD, 4, 7) |
+ __gen_field(values->MIPCountLOD, 0, 3) |
+ 0;
+
+ uint32_t dw6 =
+ __gen_field(values->MCSSurfacePitch, 3, 11) |
+ __gen_field(values->AppendCounterEnable, 1, 1) |
+ __gen_field(values->MCSEnable, 0, 0) |
+ __gen_field(values->XOffsetforUVPlane, 16, 29) |
+ __gen_field(values->YOffsetforUVPlane, 0, 13) |
+ 0;
+
+ dw[6] =
+ __gen_combine_address(data, &dw[6], values->AppendCounterAddress, dw6);
+
+ dw[7] =
+ __gen_field(values->RedClearColor, 31, 31) |
+ __gen_field(values->GreenClearColor, 30, 30) |
+ __gen_field(values->BlueClearColor, 29, 29) |
+ __gen_field(values->AlphaClearColor, 28, 28) |
+ __gen_field(values->ResourceMinLOD * (1 << 8), 0, 11) |
+ 0;
+
+}
+
+#define GEN7_SAMPLER_BORDER_COLOR_STATE_length 0x00000004
+
+struct GEN7_SAMPLER_BORDER_COLOR_STATE {
+ float BorderColorRedDX100GL;
+ uint32_t BorderColorAlpha;
+ uint32_t BorderColorBlue;
+ uint32_t BorderColorGreen;
+ uint32_t BorderColorRedDX9;
+ float BorderColorGreen0;
+ float BorderColorBlue0;
+ float BorderColorAlpha0;
+};
+
+static inline void
+GEN7_SAMPLER_BORDER_COLOR_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_SAMPLER_BORDER_COLOR_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_float(values->BorderColorRedDX100GL) |
+ __gen_field(values->BorderColorAlpha, 24, 31) |
+ __gen_field(values->BorderColorBlue, 16, 23) |
+ __gen_field(values->BorderColorGreen, 8, 15) |
+ __gen_field(values->BorderColorRedDX9, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_float(values->BorderColorGreen) |
+ 0;
+
+ dw[2] =
+ __gen_float(values->BorderColorBlue) |
+ 0;
+
+ dw[3] =
+ __gen_float(values->BorderColorAlpha) |
+ 0;
+
+}
+
+#define GEN7_SAMPLER_STATE_length 0x00000004
+
+struct GEN7_SAMPLER_STATE {
+ bool SamplerDisable;
+#define DX10OGL 0
+#define DX9 1
+ uint32_t TextureBorderColorMode;
+#define OGL 1
+ uint32_t LODPreClampEnable;
+ float BaseMipLevel;
+#define MIPFILTER_NONE 0
+#define MIPFILTER_NEAREST 1
+#define MIPFILTER_LINEAR 3
+ uint32_t MipModeFilter;
+#define MAPFILTER_NEAREST 0
+#define MAPFILTER_LINEAR 1
+#define MAPFILTER_ANISOTROPIC 2
+#define MAPFILTER_MONO 6
+ uint32_t MagModeFilter;
+#define MAPFILTER_NEAREST 0
+#define MAPFILTER_LINEAR 1
+#define MAPFILTER_ANISOTROPIC 2
+#define MAPFILTER_MONO 6
+ uint32_t MinModeFilter;
+ float TextureLODBias;
+#define LEGACY 0
+#define EWAApproximation 1
+ uint32_t AnisotropicAlgorithm;
+ float MinLOD;
+ float MaxLOD;
+#define PREFILTEROPALWAYS 0
+#define PREFILTEROPNEVER 1
+#define PREFILTEROPLESS 2
+#define PREFILTEROPEQUAL 3
+#define PREFILTEROPLEQUAL 4
+#define PREFILTEROPGREATER 5
+#define PREFILTEROPNOTEQUAL 6
+#define PREFILTEROPGEQUAL 7
+ uint32_t ShadowFunction;
+#define PROGRAMMED 0
+#define OVERRIDE 1
+ uint32_t CubeSurfaceControlMode;
+ uint32_t BorderColorPointer;
+ bool ChromaKeyEnable;
+ uint32_t ChromaKeyIndex;
+#define KEYFILTER_KILL_ON_ANY_MATCH 0
+#define KEYFILTER_REPLACE_BLACK 1
+ uint32_t ChromaKeyMode;
+#define RATIO21 0
+#define RATIO41 1
+#define RATIO61 2
+#define RATIO81 3
+#define RATIO101 4
+#define RATIO121 5
+#define RATIO141 6
+#define RATIO161 7
+ uint32_t MaximumAnisotropy;
+ bool RAddressMinFilterRoundingEnable;
+ bool RAddressMagFilterRoundingEnable;
+ bool VAddressMinFilterRoundingEnable;
+ bool VAddressMagFilterRoundingEnable;
+ bool UAddressMinFilterRoundingEnable;
+ bool UAddressMagFilterRoundingEnable;
+#define FULL 0
+#define MED 2
+#define LOW 3
+ uint32_t TrilinearFilterQuality;
+ bool NonnormalizedCoordinateEnable;
+ uint32_t TCXAddressControlMode;
+ uint32_t TCYAddressControlMode;
+ uint32_t TCZAddressControlMode;
+};
+
+static inline void
+GEN7_SAMPLER_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN7_SAMPLER_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->SamplerDisable, 31, 31) |
+ __gen_field(values->TextureBorderColorMode, 29, 29) |
+ __gen_field(values->LODPreClampEnable, 28, 28) |
+ __gen_field(values->BaseMipLevel * (1 << 1), 22, 26) |
+ __gen_field(values->MipModeFilter, 20, 21) |
+ __gen_field(values->MagModeFilter, 17, 19) |
+ __gen_field(values->MinModeFilter, 14, 16) |
+ __gen_fixed(values->TextureLODBias, 1, 13, true, 8) |
+ __gen_field(values->AnisotropicAlgorithm, 0, 0) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->MinLOD * (1 << 8), 20, 31) |
+ __gen_field(values->MaxLOD * (1 << 8), 8, 19) |
+ __gen_field(values->ShadowFunction, 1, 3) |
+ __gen_field(values->CubeSurfaceControlMode, 0, 0) |
+ 0;
+
+ dw[2] =
+ __gen_offset(values->BorderColorPointer, 5, 31) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->ChromaKeyEnable, 25, 25) |
+ __gen_field(values->ChromaKeyIndex, 23, 24) |
+ __gen_field(values->ChromaKeyMode, 22, 22) |
+ __gen_field(values->MaximumAnisotropy, 19, 21) |
+ __gen_field(values->RAddressMinFilterRoundingEnable, 13, 13) |
+ __gen_field(values->RAddressMagFilterRoundingEnable, 14, 14) |
+ __gen_field(values->VAddressMinFilterRoundingEnable, 15, 15) |
+ __gen_field(values->VAddressMagFilterRoundingEnable, 16, 16) |
+ __gen_field(values->UAddressMinFilterRoundingEnable, 17, 17) |
+ __gen_field(values->UAddressMagFilterRoundingEnable, 18, 18) |
+ __gen_field(values->TrilinearFilterQuality, 11, 12) |
+ __gen_field(values->NonnormalizedCoordinateEnable, 10, 10) |
+ __gen_field(values->TCXAddressControlMode, 6, 8) |
+ __gen_field(values->TCYAddressControlMode, 3, 5) |
+ __gen_field(values->TCZAddressControlMode, 0, 2) |
+ 0;
+
+}
+
+/* Enum 3D_Prim_Topo_Type */
+#define _3DPRIM_POINTLIST 1
+#define _3DPRIM_LINELIST 2
+#define _3DPRIM_LINESTRIP 3
+#define _3DPRIM_TRILIST 4
+#define _3DPRIM_TRISTRIP 5
+#define _3DPRIM_TRIFAN 6
+#define _3DPRIM_QUADLIST 7
+#define _3DPRIM_QUADSTRIP 8
+#define _3DPRIM_LINELIST_ADJ 9
+#define _3DPRIM_LINESTRIP_ADJ 10
+#define _3DPRIM_TRILIST_ADJ 11
+#define _3DPRIM_TRISTRIP_ADJ 12
+#define _3DPRIM_TRISTRIP_REVERSE 13
+#define _3DPRIM_POLYGON 14
+#define _3DPRIM_RECTLIST 15
+#define _3DPRIM_LINELOOP 16
+#define _3DPRIM_POINTLIST_BF 17
+#define _3DPRIM_LINESTRIP_CONT 18
+#define _3DPRIM_LINESTRIP_BF 19
+#define _3DPRIM_LINESTRIP_CONT_BF 20
+#define _3DPRIM_TRIFAN_NOSTIPPLE 22
+#define _3DPRIM_PATCHLIST_1 32
+#define _3DPRIM_PATCHLIST_2 33
+#define _3DPRIM_PATCHLIST_3 34
+#define _3DPRIM_PATCHLIST_4 35
+#define _3DPRIM_PATCHLIST_5 36
+#define _3DPRIM_PATCHLIST_6 37
+#define _3DPRIM_PATCHLIST_7 38
+#define _3DPRIM_PATCHLIST_8 39
+#define _3DPRIM_PATCHLIST_9 40
+#define _3DPRIM_PATCHLIST_10 41
+#define _3DPRIM_PATCHLIST_11 42
+#define _3DPRIM_PATCHLIST_12 43
+#define _3DPRIM_PATCHLIST_13 44
+#define _3DPRIM_PATCHLIST_14 45
+#define _3DPRIM_PATCHLIST_15 46
+#define _3DPRIM_PATCHLIST_16 47
+#define _3DPRIM_PATCHLIST_17 48
+#define _3DPRIM_PATCHLIST_18 49
+#define _3DPRIM_PATCHLIST_19 50
+#define _3DPRIM_PATCHLIST_20 51
+#define _3DPRIM_PATCHLIST_21 52
+#define _3DPRIM_PATCHLIST_22 53
+#define _3DPRIM_PATCHLIST_23 54
+#define _3DPRIM_PATCHLIST_24 55
+#define _3DPRIM_PATCHLIST_25 56
+#define _3DPRIM_PATCHLIST_26 57
+#define _3DPRIM_PATCHLIST_27 58
+#define _3DPRIM_PATCHLIST_28 59
+#define _3DPRIM_PATCHLIST_29 60
+#define _3DPRIM_PATCHLIST_30 61
+#define _3DPRIM_PATCHLIST_31 62
+#define _3DPRIM_PATCHLIST_32 63
+
+/* Enum 3D_Vertex_Component_Control */
+#define VFCOMP_NOSTORE 0
+#define VFCOMP_STORE_SRC 1
+#define VFCOMP_STORE_0 2
+#define VFCOMP_STORE_1_FP 3
+#define VFCOMP_STORE_1_INT 4
+#define VFCOMP_STORE_VID 5
+#define VFCOMP_STORE_IID 6
+#define VFCOMP_STORE_PID 7
+
+/* Enum 3D_Compare_Function */
+#define COMPAREFUNCTION_ALWAYS 0
+#define COMPAREFUNCTION_NEVER 1
+#define COMPAREFUNCTION_LESS 2
+#define COMPAREFUNCTION_EQUAL 3
+#define COMPAREFUNCTION_LEQUAL 4
+#define COMPAREFUNCTION_GREATER 5
+#define COMPAREFUNCTION_NOTEQUAL 6
+#define COMPAREFUNCTION_GEQUAL 7
+
+/* Enum SURFACE_FORMAT */
+#define R32G32B32A32_FLOAT 0
+#define R32G32B32A32_SINT 1
+#define R32G32B32A32_UINT 2
+#define R32G32B32A32_UNORM 3
+#define R32G32B32A32_SNORM 4
+#define R64G64_FLOAT 5
+#define R32G32B32X32_FLOAT 6
+#define R32G32B32A32_SSCALED 7
+#define R32G32B32A32_USCALED 8
+#define R32G32B32A32_SFIXED 32
+#define R64G64_PASSTHRU 33
+#define R32G32B32_FLOAT 64
+#define R32G32B32_SINT 65
+#define R32G32B32_UINT 66
+#define R32G32B32_UNORM 67
+#define R32G32B32_SNORM 68
+#define R32G32B32_SSCALED 69
+#define R32G32B32_USCALED 70
+#define R32G32B32_SFIXED 80
+#define R16G16B16A16_UNORM 128
+#define R16G16B16A16_SNORM 129
+#define R16G16B16A16_SINT 130
+#define R16G16B16A16_UINT 131
+#define R16G16B16A16_FLOAT 132
+#define R32G32_FLOAT 133
+#define R32G32_SINT 134
+#define R32G32_UINT 135
+#define R32_FLOAT_X8X24_TYPELESS 136
+#define X32_TYPELESS_G8X24_UINT 137
+#define L32A32_FLOAT 138
+#define R32G32_UNORM 139
+#define R32G32_SNORM 140
+#define R64_FLOAT 141
+#define R16G16B16X16_UNORM 142
+#define R16G16B16X16_FLOAT 143
+#define A32X32_FLOAT 144
+#define L32X32_FLOAT 145
+#define I32X32_FLOAT 146
+#define R16G16B16A16_SSCALED 147
+#define R16G16B16A16_USCALED 148
+#define R32G32_SSCALED 149
+#define R32G32_USCALED 150
+#define R32G32_SFIXED 160
+#define R64_PASSTHRU 161
+#define B8G8R8A8_UNORM 192
+#define B8G8R8A8_UNORM_SRGB 193
+#define R10G10B10A2_UNORM 194
+#define R10G10B10A2_UNORM_SRGB 195
+#define R10G10B10A2_UINT 196
+#define R10G10B10_SNORM_A2_UNORM 197
+#define R8G8B8A8_UNORM 199
+#define R8G8B8A8_UNORM_SRGB 200
+#define R8G8B8A8_SNORM 201
+#define R8G8B8A8_SINT 202
+#define R8G8B8A8_UINT 203
+#define R16G16_UNORM 204
+#define R16G16_SNORM 205
+#define R16G16_SINT 206
+#define R16G16_UINT 207
+#define R16G16_FLOAT 208
+#define B10G10R10A2_UNORM 209
+#define B10G10R10A2_UNORM_SRGB 210
+#define R11G11B10_FLOAT 211
+#define R32_SINT 214
+#define R32_UINT 215
+#define R32_FLOAT 216
+#define R24_UNORM_X8_TYPELESS 217
+#define X24_TYPELESS_G8_UINT 218
+#define L32_UNORM 221
+#define A32_UNORM 222
+#define L16A16_UNORM 223
+#define I24X8_UNORM 224
+#define L24X8_UNORM 225
+#define A24X8_UNORM 226
+#define I32_FLOAT 227
+#define L32_FLOAT 228
+#define A32_FLOAT 229
+#define X8B8_UNORM_G8R8_SNORM 230
+#define A8X8_UNORM_G8R8_SNORM 231
+#define B8X8_UNORM_G8R8_SNORM 232
+#define B8G8R8X8_UNORM 233
+#define B8G8R8X8_UNORM_SRGB 234
+#define R8G8B8X8_UNORM 235
+#define R8G8B8X8_UNORM_SRGB 236
+#define R9G9B9E5_SHAREDEXP 237
+#define B10G10R10X2_UNORM 238
+#define L16A16_FLOAT 240
+#define R32_UNORM 241
+#define R32_SNORM 242
+#define R10G10B10X2_USCALED 243
+#define R8G8B8A8_SSCALED 244
+#define R8G8B8A8_USCALED 245
+#define R16G16_SSCALED 246
+#define R16G16_USCALED 247
+#define R32_SSCALED 248
+#define R32_USCALED 249
+#define B5G6R5_UNORM 256
+#define B5G6R5_UNORM_SRGB 257
+#define B5G5R5A1_UNORM 258
+#define B5G5R5A1_UNORM_SRGB 259
+#define B4G4R4A4_UNORM 260
+#define B4G4R4A4_UNORM_SRGB 261
+#define R8G8_UNORM 262
+#define R8G8_SNORM 263
+#define R8G8_SINT 264
+#define R8G8_UINT 265
+#define R16_UNORM 266
+#define R16_SNORM 267
+#define R16_SINT 268
+#define R16_UINT 269
+#define R16_FLOAT 270
+#define A8P8_UNORM_PALETTE0 271
+#define A8P8_UNORM_PALETTE1 272
+#define I16_UNORM 273
+#define L16_UNORM 274
+#define A16_UNORM 275
+#define L8A8_UNORM 276
+#define I16_FLOAT 277
+#define L16_FLOAT 278
+#define A16_FLOAT 279
+#define L8A8_UNORM_SRGB 280
+#define R5G5_SNORM_B6_UNORM 281
+#define B5G5R5X1_UNORM 282
+#define B5G5R5X1_UNORM_SRGB 283
+#define R8G8_SSCALED 284
+#define R8G8_USCALED 285
+#define R16_SSCALED 286
+#define R16_USCALED 287
+#define P8A8_UNORM_PALETTE0 290
+#define P8A8_UNORM_PALETTE1 291
+#define A1B5G5R5_UNORM 292
+#define A4B4G4R4_UNORM 293
+#define L8A8_UINT 294
+#define L8A8_SINT 295
+#define R8_UNORM 320
+#define R8_SNORM 321
+#define R8_SINT 322
+#define R8_UINT 323
+#define A8_UNORM 324
+#define I8_UNORM 325
+#define L8_UNORM 326
+#define P4A4_UNORM_PALETTE0 327
+#define A4P4_UNORM_PALETTE0 328
+#define R8_SSCALED 329
+#define R8_USCALED 330
+#define P8_UNORM_PALETTE0 331
+#define L8_UNORM_SRGB 332
+#define P8_UNORM_PALETTE1 333
+#define P4A4_UNORM_PALETTE1 334
+#define A4P4_UNORM_PALETTE1 335
+#define Y8_UNORM 336
+#define L8_UINT 338
+#define L8_SINT 339
+#define I8_UINT 340
+#define I8_SINT 341
+#define DXT1_RGB_SRGB 384
+#define R1_UNORM 385
+#define YCRCB_NORMAL 386
+#define YCRCB_SWAPUVY 387
+#define P2_UNORM_PALETTE0 388
+#define P2_UNORM_PALETTE1 389
+#define BC1_UNORM 390
+#define BC2_UNORM 391
+#define BC3_UNORM 392
+#define BC4_UNORM 393
+#define BC5_UNORM 394
+#define BC1_UNORM_SRGB 395
+#define BC2_UNORM_SRGB 396
+#define BC3_UNORM_SRGB 397
+#define MONO8 398
+#define YCRCB_SWAPUV 399
+#define YCRCB_SWAPY 400
+#define DXT1_RGB 401
+#define FXT1 402
+#define R8G8B8_UNORM 403
+#define R8G8B8_SNORM 404
+#define R8G8B8_SSCALED 405
+#define R8G8B8_USCALED 406
+#define R64G64B64A64_FLOAT 407
+#define R64G64B64_FLOAT 408
+#define BC4_SNORM 409
+#define BC5_SNORM 410
+#define R16G16B16_FLOAT 411
+#define R16G16B16_UNORM 412
+#define R16G16B16_SNORM 413
+#define R16G16B16_SSCALED 414
+#define R16G16B16_USCALED 415
+#define BC6H_SF16 417
+#define BC7_UNORM 418
+#define BC7_UNORM_SRGB 419
+#define BC6H_UF16 420
+#define PLANAR_420_8 421
+#define R8G8B8_UNORM_SRGB 424
+#define ETC1_RGB8 425
+#define ETC2_RGB8 426
+#define EAC_R11 427
+#define EAC_RG11 428
+#define EAC_SIGNED_R11 429
+#define EAC_SIGNED_RG11 430
+#define ETC2_SRGB8 431
+#define R16G16B16_UINT 432
+#define R16G16B16_SINT 433
+#define R32_SFIXED 434
+#define R10G10B10A2_SNORM 435
+#define R10G10B10A2_USCALED 436
+#define R10G10B10A2_SSCALED 437
+#define R10G10B10A2_SINT 438
+#define B10G10R10A2_SNORM 439
+#define B10G10R10A2_USCALED 440
+#define B10G10R10A2_SSCALED 441
+#define B10G10R10A2_UINT 442
+#define B10G10R10A2_SINT 443
+#define R64G64B64A64_PASSTHRU 444
+#define R64G64B64_PASSTHRU 445
+#define ETC2_RGB8_PTA 448
+#define ETC2_SRGB8_PTA 449
+#define ETC2_EAC_RGBA8 450
+#define ETC2_EAC_SRGB8_A8 451
+#define R8G8B8_UINT 456
+#define R8G8B8_SINT 457
+#define RAW 511
+
+/* Enum Texture Coordinate Mode */
+#define TCM_WRAP 0
+#define TCM_MIRROR 1
+#define TCM_CLAMP 2
+#define TCM_CUBE 3
+#define TCM_CLAMP_BORDER 4
+#define TCM_MIRROR_ONCE 5
+
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <assert.h>
+#include <stdbool.h>
+#include <string.h>
+#include <unistd.h>
+#include <fcntl.h>
+
+#include "anv_private.h"
+
+#include "gen7_pack.h"
+#include "gen75_pack.h"
+
+#include "genX_pipeline_util.h"
+
+static void
+emit_vertex_input(struct anv_pipeline *pipeline,
+ const VkPipelineVertexInputStateCreateInfo *info,
+ const struct anv_graphics_pipeline_create_info *extra)
+{
+ uint32_t elements;
+ if (extra && extra->disable_vs) {
+ /* If the VS is disabled, just assume the user knows what they're
+ * doing and apply the layout blindly. This can only come from
+ * meta, so this *should* be safe.
+ */
+ elements = 0;
+ for (uint32_t i = 0; i < info->vertexAttributeDescriptionCount; i++)
+ elements |= (1 << info->pVertexAttributeDescriptions[i].location);
+ } else {
+ /* Pull inputs_read out of the VS prog data */
+ uint64_t inputs_read = pipeline->vs_prog_data.inputs_read;
+ assert((inputs_read & ((1 << VERT_ATTRIB_GENERIC0) - 1)) == 0);
+ elements = inputs_read >> VERT_ATTRIB_GENERIC0;
+ }
+
+ uint32_t vb_count = __builtin_popcount(elements);
+
+ if (pipeline->vs_prog_data.uses_vertexid ||
+ pipeline->vs_prog_data.uses_instanceid)
+ vb_count++;
+
+ if (vb_count == 0)
+ return;
+
+ const uint32_t num_dwords = 1 + vb_count * 2;
+
+ uint32_t *p = anv_batch_emitn(&pipeline->batch, num_dwords,
+ GEN7_3DSTATE_VERTEX_ELEMENTS);
+ memset(p + 1, 0, (num_dwords - 1) * 4);
+
+ for (uint32_t i = 0; i < info->vertexAttributeDescriptionCount; i++) {
+ const VkVertexInputAttributeDescription *desc =
+ &info->pVertexAttributeDescriptions[i];
+ enum isl_format format = anv_get_isl_format(desc->format,
+ VK_IMAGE_ASPECT_COLOR_BIT,
+ VK_IMAGE_TILING_LINEAR);
+
+ assert(desc->binding < 32);
+
+ if ((elements & (1 << desc->location)) == 0)
+ continue; /* Binding unused */
+
+ uint32_t slot = __builtin_popcount(elements & ((1 << desc->location) - 1));
+
+ struct GEN7_VERTEX_ELEMENT_STATE element = {
+ .VertexBufferIndex = desc->binding,
+ .Valid = true,
+ .SourceElementFormat = format,
+ .EdgeFlagEnable = false,
+ .SourceElementOffset = desc->offset,
+ .Component0Control = vertex_element_comp_control(format, 0),
+ .Component1Control = vertex_element_comp_control(format, 1),
+ .Component2Control = vertex_element_comp_control(format, 2),
+ .Component3Control = vertex_element_comp_control(format, 3),
+ };
+ GEN7_VERTEX_ELEMENT_STATE_pack(NULL, &p[1 + slot * 2], &element);
+ }
+
+ if (pipeline->vs_prog_data.uses_vertexid ||
+ pipeline->vs_prog_data.uses_instanceid) {
+ struct GEN7_VERTEX_ELEMENT_STATE element = {
+ .Valid = true,
+ /* FIXME: Do we need to provide the base vertex as component 0 here
+ * to support the correct base vertex ID? */
+ .Component0Control = VFCOMP_STORE_0,
+ .Component1Control = VFCOMP_STORE_0,
+ .Component2Control = VFCOMP_STORE_VID,
+ .Component3Control = VFCOMP_STORE_IID
+ };
+ GEN7_VERTEX_ELEMENT_STATE_pack(NULL, &p[1 + (vb_count - 1) * 2], &element);
+ }
+}
+
+static void
+gen7_emit_rs_state(struct anv_pipeline *pipeline,
+ const VkPipelineRasterizationStateCreateInfo *info,
+ const struct anv_graphics_pipeline_create_info *extra)
+{
+ struct GEN7_3DSTATE_SF sf = {
+ GEN7_3DSTATE_SF_header,
+
+ /* FIXME: Get this from pass info */
+ .DepthBufferSurfaceFormat = D24_UNORM_X8_UINT,
+
+ /* LegacyGlobalDepthBiasEnable */
+
+ .StatisticsEnable = true,
+ .FrontFaceFillMode = vk_to_gen_fillmode[info->polygonMode],
+ .BackFaceFillMode = vk_to_gen_fillmode[info->polygonMode],
+ .ViewTransformEnable = !(extra && extra->disable_viewport),
+ .FrontWinding = vk_to_gen_front_face[info->frontFace],
+ /* bool AntiAliasingEnable; */
+
+ .CullMode = vk_to_gen_cullmode[info->cullMode],
+
+ /* uint32_t LineEndCapAntialiasingRegionWidth; */
+ .ScissorRectangleEnable = !(extra && extra->disable_scissor),
+
+ /* uint32_t MultisampleRasterizationMode; */
+ /* bool LastPixelEnable; */
+
+ .TriangleStripListProvokingVertexSelect = 0,
+ .LineStripListProvokingVertexSelect = 0,
+ .TriangleFanProvokingVertexSelect = 0,
+
+ /* uint32_t AALineDistanceMode; */
+ /* uint32_t VertexSubPixelPrecisionSelect; */
+ .UsePointWidthState = !pipeline->writes_point_size,
+ .PointWidth = 1.0,
+ };
+
+ GEN7_3DSTATE_SF_pack(NULL, &pipeline->gen7.sf, &sf);
+}
+
+static void
+gen7_emit_ds_state(struct anv_pipeline *pipeline,
+ const VkPipelineDepthStencilStateCreateInfo *info)
+{
+ if (info == NULL) {
+ /* We're going to OR this together with the dynamic state. We need
+ * to make sure it's initialized to something useful.
+ */
+ memset(pipeline->gen7.depth_stencil_state, 0,
+ sizeof(pipeline->gen7.depth_stencil_state));
+ return;
+ }
+
+ struct GEN7_DEPTH_STENCIL_STATE state = {
+ .DepthTestEnable = info->depthTestEnable,
+ .DepthBufferWriteEnable = info->depthWriteEnable,
+ .DepthTestFunction = vk_to_gen_compare_op[info->depthCompareOp],
+ .DoubleSidedStencilEnable = true,
+
+ .StencilTestEnable = info->stencilTestEnable,
+ .StencilFailOp = vk_to_gen_stencil_op[info->front.failOp],
+ .StencilPassDepthPassOp = vk_to_gen_stencil_op[info->front.passOp],
+ .StencilPassDepthFailOp = vk_to_gen_stencil_op[info->front.depthFailOp],
+ .StencilTestFunction = vk_to_gen_compare_op[info->front.compareOp],
+
+ .BackfaceStencilFailOp = vk_to_gen_stencil_op[info->back.failOp],
+ .BackfaceStencilPassDepthPassOp = vk_to_gen_stencil_op[info->back.passOp],
+ .BackfaceStencilPassDepthFailOp = vk_to_gen_stencil_op[info->back.depthFailOp],
+ .BackFaceStencilTestFunction = vk_to_gen_compare_op[info->back.compareOp],
+ };
+
+ GEN7_DEPTH_STENCIL_STATE_pack(NULL, &pipeline->gen7.depth_stencil_state, &state);
+}
+
+static void
+gen7_emit_cb_state(struct anv_pipeline *pipeline,
+ const VkPipelineColorBlendStateCreateInfo *info,
+ const VkPipelineMultisampleStateCreateInfo *ms_info)
+{
+ struct anv_device *device = pipeline->device;
+
+ if (info->pAttachments == NULL) {
+ pipeline->blend_state =
+ anv_state_pool_emit(&device->dynamic_state_pool,
+ GEN7_BLEND_STATE, 64,
+ .ColorBufferBlendEnable = false,
+ .WriteDisableAlpha = false,
+ .WriteDisableRed = false,
+ .WriteDisableGreen = false,
+ .WriteDisableBlue = false);
+ } else {
+ /* FIXME-GEN7: All render targets share blend state settings on gen7, we
+ * can't implement this.
+ */
+ const VkPipelineColorBlendAttachmentState *a = &info->pAttachments[0];
+ pipeline->blend_state =
+ anv_state_pool_emit(&device->dynamic_state_pool,
+ GEN7_BLEND_STATE, 64,
+
+ .ColorBufferBlendEnable = a->blendEnable,
+ .IndependentAlphaBlendEnable = true, /* FIXME: yes? */
+ .AlphaBlendFunction = vk_to_gen_blend_op[a->alphaBlendOp],
+
+ .SourceAlphaBlendFactor = vk_to_gen_blend[a->srcAlphaBlendFactor],
+ .DestinationAlphaBlendFactor = vk_to_gen_blend[a->dstAlphaBlendFactor],
+
+ .ColorBlendFunction = vk_to_gen_blend_op[a->colorBlendOp],
+ .SourceBlendFactor = vk_to_gen_blend[a->srcColorBlendFactor],
+ .DestinationBlendFactor = vk_to_gen_blend[a->dstColorBlendFactor],
+ .AlphaToCoverageEnable = ms_info && ms_info->alphaToCoverageEnable,
+
+# if 0
+ bool AlphaToOneEnable;
+ bool AlphaToCoverageDitherEnable;
+# endif
+
+ .WriteDisableAlpha = !(a->colorWriteMask & VK_COLOR_COMPONENT_A_BIT),
+ .WriteDisableRed = !(a->colorWriteMask & VK_COLOR_COMPONENT_R_BIT),
+ .WriteDisableGreen = !(a->colorWriteMask & VK_COLOR_COMPONENT_G_BIT),
+ .WriteDisableBlue = !(a->colorWriteMask & VK_COLOR_COMPONENT_B_BIT),
+
+ .LogicOpEnable = info->logicOpEnable,
+ .LogicOpFunction = vk_to_gen_logic_op[info->logicOp],
+
+# if 0
+ bool AlphaTestEnable;
+ uint32_t AlphaTestFunction;
+ bool ColorDitherEnable;
+ uint32_t XDitherOffset;
+ uint32_t YDitherOffset;
+ uint32_t ColorClampRange;
+ bool PreBlendColorClampEnable;
+ bool PostBlendColorClampEnable;
+# endif
+ );
+ }
+
+ anv_batch_emit(&pipeline->batch, GEN7_3DSTATE_BLEND_STATE_POINTERS,
+ .BlendStatePointer = pipeline->blend_state.offset);
+}
+
+static inline uint32_t
+scratch_space(const struct brw_stage_prog_data *prog_data)
+{
+ return ffs(prog_data->total_scratch / 1024);
+}
+
+GENX_FUNC(GEN7, GEN75) VkResult
+genX(graphics_pipeline_create)(
+ VkDevice _device,
+ struct anv_pipeline_cache * cache,
+ const VkGraphicsPipelineCreateInfo* pCreateInfo,
+ const struct anv_graphics_pipeline_create_info *extra,
+ const VkAllocationCallbacks* pAllocator,
+ VkPipeline* pPipeline)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ struct anv_pipeline *pipeline;
+ VkResult result;
+
+ assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO);
+
+ pipeline = anv_alloc2(&device->alloc, pAllocator, sizeof(*pipeline), 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (pipeline == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ result = anv_pipeline_init(pipeline, device, cache,
+ pCreateInfo, extra, pAllocator);
+ if (result != VK_SUCCESS) {
+ anv_free2(&device->alloc, pAllocator, pipeline);
+ return result;
+ }
+
+ assert(pCreateInfo->pVertexInputState);
+ emit_vertex_input(pipeline, pCreateInfo->pVertexInputState, extra);
+
+ assert(pCreateInfo->pRasterizationState);
+ gen7_emit_rs_state(pipeline, pCreateInfo->pRasterizationState, extra);
+
+ gen7_emit_ds_state(pipeline, pCreateInfo->pDepthStencilState);
+
+ gen7_emit_cb_state(pipeline, pCreateInfo->pColorBlendState,
+ pCreateInfo->pMultisampleState);
+
+ anv_batch_emit(&pipeline->batch, GEN7_3DSTATE_VF_STATISTICS,
+ .StatisticsEnable = true);
+ anv_batch_emit(&pipeline->batch, GEN7_3DSTATE_HS, .Enable = false);
+ anv_batch_emit(&pipeline->batch, GEN7_3DSTATE_TE, .TEEnable = false);
+ anv_batch_emit(&pipeline->batch, GEN7_3DSTATE_DS, .DSFunctionEnable = false);
+ anv_batch_emit(&pipeline->batch, GEN7_3DSTATE_STREAMOUT, .SOFunctionEnable = false);
+
+ /* From the IVB PRM Vol. 2, Part 1, Section 3.2.1:
+ *
+ * "A PIPE_CONTROL with Post-Sync Operation set to 1h and a depth stall
+ * needs to be sent just prior to any 3DSTATE_VS, 3DSTATE_URB_VS,
+ * 3DSTATE_CONSTANT_VS, 3DSTATE_BINDING_TABLE_POINTER_VS,
+ * 3DSTATE_SAMPLER_STATE_POINTER_VS command. Only one PIPE_CONTROL
+ * needs to be sent before any combination of VS associated 3DSTATE."
+ */
+ anv_batch_emit(&pipeline->batch, GEN7_PIPE_CONTROL,
+ .DepthStallEnable = true,
+ .PostSyncOperation = WriteImmediateData,
+ .Address = { &device->workaround_bo, 0 });
+
+ anv_batch_emit(&pipeline->batch, GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_VS,
+ .ConstantBufferOffset = 0,
+ .ConstantBufferSize = 4);
+ anv_batch_emit(&pipeline->batch, GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_GS,
+ .ConstantBufferOffset = 4,
+ .ConstantBufferSize = 4);
+ anv_batch_emit(&pipeline->batch, GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_PS,
+ .ConstantBufferOffset = 8,
+ .ConstantBufferSize = 4);
+
+ anv_batch_emit(&pipeline->batch, GEN7_3DSTATE_AA_LINE_PARAMETERS);
+
+ const VkPipelineRasterizationStateCreateInfo *rs_info =
+ pCreateInfo->pRasterizationState;
+
+ anv_batch_emit(&pipeline->batch, GEN7_3DSTATE_CLIP,
+ .FrontWinding = vk_to_gen_front_face[rs_info->frontFace],
+ .CullMode = vk_to_gen_cullmode[rs_info->cullMode],
+ .ClipEnable = true,
+ .APIMode = APIMODE_OGL,
+ .ViewportXYClipTestEnable = !(extra && extra->disable_viewport),
+ .ClipMode = CLIPMODE_NORMAL,
+ .TriangleStripListProvokingVertexSelect = 0,
+ .LineStripListProvokingVertexSelect = 0,
+ .TriangleFanProvokingVertexSelect = 0,
+ .MinimumPointWidth = 0.125,
+ .MaximumPointWidth = 255.875);
+
+ uint32_t samples = 1;
+ uint32_t log2_samples = __builtin_ffs(samples) - 1;
+
+ anv_batch_emit(&pipeline->batch, GEN7_3DSTATE_MULTISAMPLE,
+ .PixelLocation = PIXLOC_CENTER,
+ .NumberofMultisamples = log2_samples);
+
+ anv_batch_emit(&pipeline->batch, GEN7_3DSTATE_SAMPLE_MASK,
+ .SampleMask = 0xff);
+
+ anv_batch_emit(&pipeline->batch, GEN7_3DSTATE_URB_VS,
+ .VSURBStartingAddress = pipeline->urb.vs_start,
+ .VSURBEntryAllocationSize = pipeline->urb.vs_size - 1,
+ .VSNumberofURBEntries = pipeline->urb.nr_vs_entries);
+
+ anv_batch_emit(&pipeline->batch, GEN7_3DSTATE_URB_GS,
+ .GSURBStartingAddress = pipeline->urb.gs_start,
+ .GSURBEntryAllocationSize = pipeline->urb.gs_size - 1,
+ .GSNumberofURBEntries = pipeline->urb.nr_gs_entries);
+
+ anv_batch_emit(&pipeline->batch, GEN7_3DSTATE_URB_HS,
+ .HSURBStartingAddress = pipeline->urb.vs_start,
+ .HSURBEntryAllocationSize = 0,
+ .HSNumberofURBEntries = 0);
+
+ anv_batch_emit(&pipeline->batch, GEN7_3DSTATE_URB_DS,
+ .DSURBStartingAddress = pipeline->urb.vs_start,
+ .DSURBEntryAllocationSize = 0,
+ .DSNumberofURBEntries = 0);
+
+ const struct brw_vue_prog_data *vue_prog_data = &pipeline->vs_prog_data.base;
+ /* The last geometry producing stage will set urb_offset and urb_length,
+ * which we use in 3DSTATE_SBE. Skip the VUE header and position slots. */
+ uint32_t urb_offset = 1;
+ uint32_t urb_length = (vue_prog_data->vue_map.num_slots + 1) / 2 - urb_offset;
+
+#if 0
+ /* From gen7_vs_state.c */
+
+ /**
+ * From Graphics BSpec: 3D-Media-GPGPU Engine > 3D Pipeline Stages >
+ * Geometry > Geometry Shader > State:
+ *
+ * "Note: Because of corruption in IVB:GT2, software needs to flush the
+ * whole fixed function pipeline when the GS enable changes value in
+ * the 3DSTATE_GS."
+ *
+ * The hardware architects have clarified that in this context "flush the
+ * whole fixed function pipeline" means to emit a PIPE_CONTROL with the "CS
+ * Stall" bit set.
+ */
+ if (!brw->is_haswell && !brw->is_baytrail)
+ gen7_emit_vs_workaround_flush(brw);
+#endif
+
+ if (pipeline->vs_vec4 == NO_KERNEL || (extra && extra->disable_vs))
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_VS), .VSFunctionEnable = false);
+ else
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_VS),
+ .KernelStartPointer = pipeline->vs_vec4,
+ .ScratchSpaceBaseOffset = pipeline->scratch_start[MESA_SHADER_VERTEX],
+ .PerThreadScratchSpace = scratch_space(&vue_prog_data->base),
+
+ .DispatchGRFStartRegisterforURBData =
+ vue_prog_data->base.dispatch_grf_start_reg,
+ .VertexURBEntryReadLength = vue_prog_data->urb_read_length,
+ .VertexURBEntryReadOffset = 0,
+
+ .MaximumNumberofThreads = device->info.max_vs_threads - 1,
+ .StatisticsEnable = true,
+ .VSFunctionEnable = true);
+
+ const struct brw_gs_prog_data *gs_prog_data = &pipeline->gs_prog_data;
+
+ if (pipeline->gs_kernel == NO_KERNEL || (extra && extra->disable_vs)) {
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_GS), .GSEnable = false);
+ } else {
+ urb_offset = 1;
+ urb_length = (gs_prog_data->base.vue_map.num_slots + 1) / 2 - urb_offset;
+
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_GS),
+ .KernelStartPointer = pipeline->gs_kernel,
+ .ScratchSpaceBasePointer = pipeline->scratch_start[MESA_SHADER_GEOMETRY],
+ .PerThreadScratchSpace = scratch_space(&gs_prog_data->base.base),
+
+ .OutputVertexSize = gs_prog_data->output_vertex_size_hwords * 2 - 1,
+ .OutputTopology = gs_prog_data->output_topology,
+ .VertexURBEntryReadLength = gs_prog_data->base.urb_read_length,
+ .IncludeVertexHandles = gs_prog_data->base.include_vue_handles,
+ .DispatchGRFStartRegisterforURBData =
+ gs_prog_data->base.base.dispatch_grf_start_reg,
+
+ .MaximumNumberofThreads = device->info.max_gs_threads - 1,
+ /* This in the next dword on HSW. */
+ .ControlDataFormat = gs_prog_data->control_data_format,
+ .ControlDataHeaderSize = gs_prog_data->control_data_header_size_hwords,
+ .InstanceControl = MAX2(gs_prog_data->invocations, 1) - 1,
+ .DispatchMode = gs_prog_data->base.dispatch_mode,
+ .GSStatisticsEnable = true,
+ .IncludePrimitiveID = gs_prog_data->include_primitive_id,
+# if (ANV_IS_HASWELL)
+ .ReorderMode = REORDER_TRAILING,
+# else
+ .ReorderEnable = true,
+# endif
+ .GSEnable = true);
+ }
+
+ const struct brw_wm_prog_data *wm_prog_data = &pipeline->wm_prog_data;
+ if (wm_prog_data->urb_setup[VARYING_SLOT_BFC0] != -1 ||
+ wm_prog_data->urb_setup[VARYING_SLOT_BFC1] != -1)
+ anv_finishme("two-sided color needs sbe swizzling setup");
+ if (wm_prog_data->urb_setup[VARYING_SLOT_PRIMITIVE_ID] != -1)
+ anv_finishme("primitive_id needs sbe swizzling setup");
+
+ /* FIXME: generated header doesn't emit attr swizzle fields */
+ anv_batch_emit(&pipeline->batch, GEN7_3DSTATE_SBE,
+ .NumberofSFOutputAttributes = pipeline->wm_prog_data.num_varying_inputs,
+ .VertexURBEntryReadLength = urb_length,
+ .VertexURBEntryReadOffset = urb_offset,
+ .PointSpriteTextureCoordinateOrigin = UPPERLEFT);
+
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_PS),
+ .KernelStartPointer0 = pipeline->ps_ksp0,
+ .ScratchSpaceBasePointer = pipeline->scratch_start[MESA_SHADER_FRAGMENT],
+ .PerThreadScratchSpace = scratch_space(&wm_prog_data->base),
+
+ .MaximumNumberofThreads = device->info.max_wm_threads - 1,
+ .PushConstantEnable = wm_prog_data->base.nr_params > 0,
+ .AttributeEnable = wm_prog_data->num_varying_inputs > 0,
+ .oMaskPresenttoRenderTarget = wm_prog_data->uses_omask,
+
+ .RenderTargetFastClearEnable = false,
+ .DualSourceBlendEnable = false,
+ .RenderTargetResolveEnable = false,
+
+ .PositionXYOffsetSelect = wm_prog_data->uses_pos_offset ?
+ POSOFFSET_SAMPLE : POSOFFSET_NONE,
+
+ ._32PixelDispatchEnable = false,
+ ._16PixelDispatchEnable = pipeline->ps_simd16 != NO_KERNEL,
+ ._8PixelDispatchEnable = pipeline->ps_simd8 != NO_KERNEL,
+
+ .DispatchGRFStartRegisterforConstantSetupData0 = pipeline->ps_grf_start0,
+ .DispatchGRFStartRegisterforConstantSetupData1 = 0,
+ .DispatchGRFStartRegisterforConstantSetupData2 = pipeline->ps_grf_start2,
+
+#if 0
+ /* Haswell requires the sample mask to be set in this packet as well as
+ * in 3DSTATE_SAMPLE_MASK; the values should match. */
+ /* _NEW_BUFFERS, _NEW_MULTISAMPLE */
+#endif
+
+ .KernelStartPointer1 = 0,
+ .KernelStartPointer2 = pipeline->ps_ksp2);
+
+ /* FIXME-GEN7: This needs a lot more work, cf gen7 upload_wm_state(). */
+ anv_batch_emit(&pipeline->batch, GEN7_3DSTATE_WM,
+ .StatisticsEnable = true,
+ .ThreadDispatchEnable = true,
+ .LineEndCapAntialiasingRegionWidth = 0, /* 0.5 pixels */
+ .LineAntialiasingRegionWidth = 1, /* 1.0 pixels */
+ .EarlyDepthStencilControl = EDSC_NORMAL,
+ .PointRasterizationRule = RASTRULE_UPPER_RIGHT,
+ .PixelShaderComputedDepthMode = wm_prog_data->computed_depth_mode,
+ .BarycentricInterpolationMode = wm_prog_data->barycentric_interp_modes);
+
+ *pPipeline = anv_pipeline_to_handle(pipeline);
+
+ return VK_SUCCESS;
+}
+
+GENX_FUNC(GEN7, GEN75) VkResult
+genX(compute_pipeline_create)(
+ VkDevice _device,
+ struct anv_pipeline_cache * cache,
+ const VkComputePipelineCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkPipeline* pPipeline)
+{
+ anv_finishme("primitive_id needs sbe swizzling setup");
+ abort();
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <assert.h>
+#include <stdbool.h>
+#include <string.h>
+#include <unistd.h>
+#include <fcntl.h>
+
+#include "anv_private.h"
+
+#include "gen7_pack.h"
+#include "gen75_pack.h"
+
+#include "genX_state_util.h"
+
+GENX_FUNC(GEN7, GEN75) void
+genX(fill_buffer_surface_state)(void *state, enum isl_format format,
+ uint32_t offset, uint32_t range,
+ uint32_t stride)
+{
+ uint32_t num_elements = range / stride;
+
+ struct GENX(RENDER_SURFACE_STATE) surface_state = {
+ .SurfaceType = SURFTYPE_BUFFER,
+ .SurfaceFormat = format,
+ .SurfaceVerticalAlignment = VALIGN_4,
+ .SurfaceHorizontalAlignment = HALIGN_4,
+ .TiledSurface = false,
+ .RenderCacheReadWriteMode = false,
+ .SurfaceObjectControlState = GENX(MOCS),
+ .Height = (num_elements >> 7) & 0x3fff,
+ .Width = num_elements & 0x7f,
+ .Depth = (num_elements >> 21) & 0x3f,
+ .SurfacePitch = stride - 1,
+# if (ANV_IS_HASWELL)
+ .ShaderChannelSelectR = SCS_RED,
+ .ShaderChannelSelectG = SCS_GREEN,
+ .ShaderChannelSelectB = SCS_BLUE,
+ .ShaderChannelSelectA = SCS_ALPHA,
+# endif
+ .SurfaceBaseAddress = { NULL, offset },
+ };
+
+ GENX(RENDER_SURFACE_STATE_pack)(NULL, state, &surface_state);
+}
+
+static struct anv_state
+alloc_surface_state(struct anv_device *device,
+ struct anv_cmd_buffer *cmd_buffer)
+{
+ if (cmd_buffer) {
+ return anv_cmd_buffer_alloc_surface_state(cmd_buffer);
+ } else {
+ return anv_state_pool_alloc(&device->surface_state_pool, 64, 64);
+ }
+}
+
+VkResult genX(CreateSampler)(
+ VkDevice _device,
+ const VkSamplerCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkSampler* pSampler)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ struct anv_sampler *sampler;
+
+ assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO);
+
+ sampler = anv_alloc2(&device->alloc, pAllocator, sizeof(*sampler), 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (!sampler)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ uint32_t filter = vk_to_gen_tex_filter(pCreateInfo->magFilter,
+ pCreateInfo->anisotropyEnable);
+
+ struct GEN7_SAMPLER_STATE sampler_state = {
+ .SamplerDisable = false,
+ .TextureBorderColorMode = DX10OGL,
+ .LODPreClampEnable = OGL,
+ .BaseMipLevel = 0.0,
+ .MipModeFilter = vk_to_gen_mipmap_mode[pCreateInfo->mipmapMode],
+ .MagModeFilter = filter,
+ .MinModeFilter = filter,
+ .TextureLODBias = pCreateInfo->mipLodBias * 256,
+ .AnisotropicAlgorithm = EWAApproximation,
+ .MinLOD = pCreateInfo->minLod,
+ .MaxLOD = pCreateInfo->maxLod,
+ .ChromaKeyEnable = 0,
+ .ChromaKeyIndex = 0,
+ .ChromaKeyMode = 0,
+ .ShadowFunction = vk_to_gen_compare_op[pCreateInfo->compareOp],
+ .CubeSurfaceControlMode = 0,
+
+ .BorderColorPointer =
+ device->border_colors.offset +
+ pCreateInfo->borderColor * sizeof(float) * 4,
+
+ .MaximumAnisotropy = vk_to_gen_max_anisotropy(pCreateInfo->maxAnisotropy),
+ .RAddressMinFilterRoundingEnable = 0,
+ .RAddressMagFilterRoundingEnable = 0,
+ .VAddressMinFilterRoundingEnable = 0,
+ .VAddressMagFilterRoundingEnable = 0,
+ .UAddressMinFilterRoundingEnable = 0,
+ .UAddressMagFilterRoundingEnable = 0,
+ .TrilinearFilterQuality = 0,
+ .NonnormalizedCoordinateEnable = pCreateInfo->unnormalizedCoordinates,
+ .TCXAddressControlMode = vk_to_gen_tex_address[pCreateInfo->addressModeU],
+ .TCYAddressControlMode = vk_to_gen_tex_address[pCreateInfo->addressModeV],
+ .TCZAddressControlMode = vk_to_gen_tex_address[pCreateInfo->addressModeW],
+ };
+
+ GEN7_SAMPLER_STATE_pack(NULL, sampler->state, &sampler_state);
+
+ *pSampler = anv_sampler_to_handle(sampler);
+
+ return VK_SUCCESS;
+}
+
+static const uint8_t anv_halign[] = {
+ [4] = HALIGN_4,
+ [8] = HALIGN_8,
+};
+
+static const uint8_t anv_valign[] = {
+ [2] = VALIGN_2,
+ [4] = VALIGN_4,
+};
+
+GENX_FUNC(GEN7, GEN75) void
+genX(image_view_init)(struct anv_image_view *iview,
+ struct anv_device *device,
+ const VkImageViewCreateInfo* pCreateInfo,
+ struct anv_cmd_buffer *cmd_buffer)
+{
+ ANV_FROM_HANDLE(anv_image, image, pCreateInfo->image);
+
+ const VkImageSubresourceRange *range = &pCreateInfo->subresourceRange;
+
+ struct anv_surface *surface =
+ anv_image_get_surface_for_aspect_mask(image, range->aspectMask);
+
+ if (pCreateInfo->viewType != VK_IMAGE_VIEW_TYPE_2D)
+ anv_finishme("non-2D image views");
+
+ uint32_t depth = 1;
+ if (range->layerCount > 1) {
+ depth = range->layerCount;
+ } else if (image->extent.depth > 1) {
+ depth = image->extent.depth;
+ }
+
+ const struct isl_extent3d image_align_sa =
+ isl_surf_get_image_alignment_sa(&surface->isl);
+
+ const struct GENX(RENDER_SURFACE_STATE) template = {
+ .SurfaceType = anv_surftype(image, pCreateInfo->viewType, false),
+ .SurfaceArray = image->array_size > 1,
+ .SurfaceFormat = iview->format,
+ .SurfaceVerticalAlignment = anv_valign[image_align_sa.height],
+ .SurfaceHorizontalAlignment = anv_halign[image_align_sa.width],
+
+ /* From bspec (DevSNB, DevIVB): "Set Tile Walk to TILEWALK_XMAJOR if
+ * Tiled Surface is False."
+ */
+ .TiledSurface = surface->isl.tiling != ISL_TILING_LINEAR,
+ .TileWalk = surface->isl.tiling == ISL_TILING_Y0 ?
+ TILEWALK_YMAJOR : TILEWALK_XMAJOR,
+
+ .VerticalLineStride = 0,
+ .VerticalLineStrideOffset = 0,
+
+ .RenderCacheReadWriteMode = 0, /* TEMPLATE */
+
+ .Height = image->extent.height - 1,
+ .Width = image->extent.width - 1,
+ .Depth = depth - 1,
+ .SurfacePitch = surface->isl.row_pitch - 1,
+ .MinimumArrayElement = range->baseArrayLayer,
+ .NumberofMultisamples = MULTISAMPLECOUNT_1,
+ .XOffset = 0,
+ .YOffset = 0,
+
+ .SurfaceObjectControlState = GENX(MOCS),
+
+ .MIPCountLOD = 0, /* TEMPLATE */
+ .SurfaceMinLOD = 0, /* TEMPLATE */
+
+ .MCSEnable = false,
+# if (ANV_IS_HASWELL)
+ .ShaderChannelSelectR = vk_to_gen_swizzle(pCreateInfo->components.r,
+ VK_COMPONENT_SWIZZLE_R),
+ .ShaderChannelSelectG = vk_to_gen_swizzle(pCreateInfo->components.g,
+ VK_COMPONENT_SWIZZLE_G),
+ .ShaderChannelSelectB = vk_to_gen_swizzle(pCreateInfo->components.b,
+ VK_COMPONENT_SWIZZLE_B),
+ .ShaderChannelSelectA = vk_to_gen_swizzle(pCreateInfo->components.a,
+ VK_COMPONENT_SWIZZLE_A),
+# else /* XXX: Seriously? */
+ .RedClearColor = 0,
+ .GreenClearColor = 0,
+ .BlueClearColor = 0,
+ .AlphaClearColor = 0,
+# endif
+ .ResourceMinLOD = 0.0,
+ .SurfaceBaseAddress = { NULL, iview->offset },
+ };
+
+ if (image->needs_nonrt_surface_state) {
+ struct GENX(RENDER_SURFACE_STATE) surface_state = template;
+
+ iview->nonrt_surface_state = alloc_surface_state(device, cmd_buffer);
+
+ surface_state.RenderCacheReadWriteMode = false;
+
+ /* For non render target surfaces, the hardware interprets field
+ * MIPCount/LOD as MIPCount. The range of levels accessible by the
+ * sampler engine is [SurfaceMinLOD, SurfaceMinLOD + MIPCountLOD].
+ */
+ surface_state.SurfaceMinLOD = range->baseMipLevel;
+ surface_state.MIPCountLOD = MAX2(range->levelCount, 1) - 1;
+
+ GENX(RENDER_SURFACE_STATE_pack)(NULL, iview->nonrt_surface_state.map,
+ &surface_state);
+
+ if (!device->info.has_llc)
+ anv_state_clflush(iview->nonrt_surface_state);
+ } else {
+ iview->nonrt_surface_state.alloc_size = 0;
+ }
+
+ if (image->needs_color_rt_surface_state) {
+ struct GENX(RENDER_SURFACE_STATE) surface_state = template;
+
+ iview->color_rt_surface_state = alloc_surface_state(device, cmd_buffer);
+
+ surface_state.RenderCacheReadWriteMode = 0; /* Write only */
+
+ /* For render target surfaces, the hardware interprets field MIPCount/LOD as
+ * LOD. The Broadwell PRM says:
+ *
+ * MIPCountLOD defines the LOD that will be rendered into.
+ * SurfaceMinLOD is ignored.
+ */
+ surface_state.MIPCountLOD = range->baseMipLevel;
+ surface_state.SurfaceMinLOD = 0;
+
+ GENX(RENDER_SURFACE_STATE_pack)(NULL, iview->color_rt_surface_state.map,
+ &surface_state);
+ if (!device->info.has_llc)
+ anv_state_clflush(iview->color_rt_surface_state);
+ } else {
+ iview->color_rt_surface_state.alloc_size = 0;
+ }
+
+ if (image->needs_storage_surface_state) {
+ struct GENX(RENDER_SURFACE_STATE) surface_state = template;
+
+ iview->storage_surface_state = alloc_surface_state(device, cmd_buffer);
+
+ surface_state.SurfaceType =
+ anv_surftype(image, pCreateInfo->viewType, true),
+
+ surface_state.SurfaceFormat =
+ isl_lower_storage_image_format(&device->isl_dev, iview->format);
+
+ surface_state.SurfaceMinLOD = range->baseMipLevel;
+ surface_state.MIPCountLOD = MAX2(range->levelCount, 1) - 1;
+
+ GENX(RENDER_SURFACE_STATE_pack)(NULL, iview->storage_surface_state.map,
+ &surface_state);
+ } else {
+ iview->storage_surface_state.alloc_size = 0;
+ }
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <assert.h>
+#include <stdbool.h>
+#include <string.h>
+#include <unistd.h>
+#include <fcntl.h>
+
+#include "anv_private.h"
+
+#include "gen8_pack.h"
+#include "gen9_pack.h"
+
+static uint32_t
+cmd_buffer_flush_push_constants(struct anv_cmd_buffer *cmd_buffer)
+{
+ static const uint32_t push_constant_opcodes[] = {
+ [MESA_SHADER_VERTEX] = 21,
+ [MESA_SHADER_TESS_CTRL] = 25, /* HS */
+ [MESA_SHADER_TESS_EVAL] = 26, /* DS */
+ [MESA_SHADER_GEOMETRY] = 22,
+ [MESA_SHADER_FRAGMENT] = 23,
+ [MESA_SHADER_COMPUTE] = 0,
+ };
+
+ VkShaderStageFlags flushed = 0;
+
+ anv_foreach_stage(stage, cmd_buffer->state.push_constants_dirty) {
+ if (stage == MESA_SHADER_COMPUTE)
+ continue;
+
+ struct anv_state state = anv_cmd_buffer_push_constants(cmd_buffer, stage);
+
+ if (state.offset == 0)
+ continue;
+
+ anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_CONSTANT_VS),
+ ._3DCommandSubOpcode = push_constant_opcodes[stage],
+ .ConstantBody = {
+ .PointerToConstantBuffer0 = { .offset = state.offset },
+ .ConstantBuffer0ReadLength = DIV_ROUND_UP(state.alloc_size, 32),
+ });
+
+ flushed |= mesa_to_vk_shader_stage(stage);
+ }
+
+ cmd_buffer->state.push_constants_dirty &= ~flushed;
+
+ return flushed;
+}
+
+#if ANV_GEN == 8
+static void
+emit_viewport_state(struct anv_cmd_buffer *cmd_buffer,
+ uint32_t count, const VkViewport *viewports)
+{
+ struct anv_state sf_clip_state =
+ anv_cmd_buffer_alloc_dynamic_state(cmd_buffer, count * 64, 64);
+ struct anv_state cc_state =
+ anv_cmd_buffer_alloc_dynamic_state(cmd_buffer, count * 8, 32);
+
+ for (uint32_t i = 0; i < count; i++) {
+ const VkViewport *vp = &viewports[i];
+
+ /* The gen7 state struct has just the matrix and guardband fields, the
+ * gen8 struct adds the min/max viewport fields. */
+ struct GENX(SF_CLIP_VIEWPORT) sf_clip_viewport = {
+ .ViewportMatrixElementm00 = vp->width / 2,
+ .ViewportMatrixElementm11 = vp->height / 2,
+ .ViewportMatrixElementm22 = 1.0,
+ .ViewportMatrixElementm30 = vp->x + vp->width / 2,
+ .ViewportMatrixElementm31 = vp->y + vp->height / 2,
+ .ViewportMatrixElementm32 = 0.0,
+ .XMinClipGuardband = -1.0f,
+ .XMaxClipGuardband = 1.0f,
+ .YMinClipGuardband = -1.0f,
+ .YMaxClipGuardband = 1.0f,
+ .XMinViewPort = vp->x,
+ .XMaxViewPort = vp->x + vp->width - 1,
+ .YMinViewPort = vp->y,
+ .YMaxViewPort = vp->y + vp->height - 1,
+ };
+
+ struct GENX(CC_VIEWPORT) cc_viewport = {
+ .MinimumDepth = vp->minDepth,
+ .MaximumDepth = vp->maxDepth
+ };
+
+ GENX(SF_CLIP_VIEWPORT_pack)(NULL, sf_clip_state.map + i * 64,
+ &sf_clip_viewport);
+ GENX(CC_VIEWPORT_pack)(NULL, cc_state.map + i * 8, &cc_viewport);
+ }
+
+ if (!cmd_buffer->device->info.has_llc) {
+ anv_state_clflush(sf_clip_state);
+ anv_state_clflush(cc_state);
+ }
+
+ anv_batch_emit(&cmd_buffer->batch,
+ GENX(3DSTATE_VIEWPORT_STATE_POINTERS_CC),
+ .CCViewportPointer = cc_state.offset);
+ anv_batch_emit(&cmd_buffer->batch,
+ GENX(3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP),
+ .SFClipViewportPointer = sf_clip_state.offset);
+}
+
+void
+gen8_cmd_buffer_emit_viewport(struct anv_cmd_buffer *cmd_buffer)
+{
+ if (cmd_buffer->state.dynamic.viewport.count > 0) {
+ emit_viewport_state(cmd_buffer, cmd_buffer->state.dynamic.viewport.count,
+ cmd_buffer->state.dynamic.viewport.viewports);
+ } else {
+ /* If viewport count is 0, this is taken to mean "use the default" */
+ emit_viewport_state(cmd_buffer, 1,
+ &(VkViewport) {
+ .x = 0.0f,
+ .y = 0.0f,
+ .width = cmd_buffer->state.framebuffer->width,
+ .height = cmd_buffer->state.framebuffer->height,
+ .minDepth = 0.0f,
+ .maxDepth = 1.0f,
+ });
+ }
+}
+#endif
+
+static void
+emit_lrm(struct anv_batch *batch,
+ uint32_t reg, struct anv_bo *bo, uint32_t offset)
+{
+ anv_batch_emit(batch, GENX(MI_LOAD_REGISTER_MEM),
+ .RegisterAddress = reg,
+ .MemoryAddress = { bo, offset });
+}
+
+static void
+emit_lri(struct anv_batch *batch, uint32_t reg, uint32_t imm)
+{
+ anv_batch_emit(batch, GENX(MI_LOAD_REGISTER_IMM),
+ .RegisterOffset = reg,
+ .DataDWord = imm);
+}
+
+#define GEN8_L3CNTLREG 0x7034
+
+static void
+config_l3(struct anv_cmd_buffer *cmd_buffer, bool enable_slm)
+{
+ /* References for GL state:
+ *
+ * - commits e307cfa..228d5a3
+ * - src/mesa/drivers/dri/i965/gen7_l3_state.c
+ */
+
+ uint32_t val = enable_slm ?
+ /* All = 48 ways; URB = 16 ways; DC and RO = 0, SLM = 1 */
+ 0x60000021 :
+ /* All = 48 ways; URB = 48 ways; DC, RO and SLM = 0 */
+ 0x60000060;
+ bool changed = cmd_buffer->state.current_l3_config != val;
+
+ if (changed) {
+ /* According to the hardware docs, the L3 partitioning can only be changed
+ * while the pipeline is completely drained and the caches are flushed,
+ * which involves a first PIPE_CONTROL flush which stalls the pipeline and
+ * initiates invalidation of the relevant caches...
+ */
+ anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL),
+ .TextureCacheInvalidationEnable = true,
+ .ConstantCacheInvalidationEnable = true,
+ .InstructionCacheInvalidateEnable = true,
+ .DCFlushEnable = true,
+ .PostSyncOperation = NoWrite,
+ .CommandStreamerStallEnable = true);
+
+ /* ...followed by a second stalling flush which guarantees that
+ * invalidation is complete when the L3 configuration registers are
+ * modified.
+ */
+ anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL),
+ .DCFlushEnable = true,
+ .PostSyncOperation = NoWrite,
+ .CommandStreamerStallEnable = true);
+
+ emit_lri(&cmd_buffer->batch, GEN8_L3CNTLREG, val);
+ cmd_buffer->state.current_l3_config = val;
+ }
+}
+
+static void
+flush_pipeline_select_3d(struct anv_cmd_buffer *cmd_buffer)
+{
+ config_l3(cmd_buffer, false);
+
+ if (cmd_buffer->state.current_pipeline != _3D) {
+ anv_batch_emit(&cmd_buffer->batch, GENX(PIPELINE_SELECT),
+#if ANV_GEN >= 9
+ .MaskBits = 3,
+#endif
+ .PipelineSelection = _3D);
+ cmd_buffer->state.current_pipeline = _3D;
+ }
+}
+
+static void
+cmd_buffer_flush_state(struct anv_cmd_buffer *cmd_buffer)
+{
+ struct anv_pipeline *pipeline = cmd_buffer->state.pipeline;
+ uint32_t *p;
+
+ uint32_t vb_emit = cmd_buffer->state.vb_dirty & pipeline->vb_used;
+
+ assert((pipeline->active_stages & VK_SHADER_STAGE_COMPUTE_BIT) == 0);
+
+ flush_pipeline_select_3d(cmd_buffer);
+
+ if (vb_emit) {
+ const uint32_t num_buffers = __builtin_popcount(vb_emit);
+ const uint32_t num_dwords = 1 + num_buffers * 4;
+
+ p = anv_batch_emitn(&cmd_buffer->batch, num_dwords,
+ GENX(3DSTATE_VERTEX_BUFFERS));
+ uint32_t vb, i = 0;
+ for_each_bit(vb, vb_emit) {
+ struct anv_buffer *buffer = cmd_buffer->state.vertex_bindings[vb].buffer;
+ uint32_t offset = cmd_buffer->state.vertex_bindings[vb].offset;
+
+ struct GENX(VERTEX_BUFFER_STATE) state = {
+ .VertexBufferIndex = vb,
+ .MemoryObjectControlState = GENX(MOCS),
+ .AddressModifyEnable = true,
+ .BufferPitch = pipeline->binding_stride[vb],
+ .BufferStartingAddress = { buffer->bo, buffer->offset + offset },
+ .BufferSize = buffer->size - offset
+ };
+
+ GENX(VERTEX_BUFFER_STATE_pack)(&cmd_buffer->batch, &p[1 + i * 4], &state);
+ i++;
+ }
+ }
+
+ if (cmd_buffer->state.dirty & ANV_CMD_DIRTY_PIPELINE) {
+ /* If somebody compiled a pipeline after starting a command buffer the
+ * scratch bo may have grown since we started this cmd buffer (and
+ * emitted STATE_BASE_ADDRESS). If we're binding that pipeline now,
+ * reemit STATE_BASE_ADDRESS so that we use the bigger scratch bo. */
+ if (cmd_buffer->state.scratch_size < pipeline->total_scratch)
+ anv_cmd_buffer_emit_state_base_address(cmd_buffer);
+
+ anv_batch_emit_batch(&cmd_buffer->batch, &pipeline->batch);
+ }
+
+ /* We emit the binding tables and sampler tables first, then emit push
+ * constants and then finally emit binding table and sampler table
+ * pointers. It has to happen in this order, since emitting the binding
+ * tables may change the push constants (in case of storage images). After
+ * emitting push constants, on SKL+ we have to emit the corresponding
+ * 3DSTATE_BINDING_TABLE_POINTER_* for the push constants to take effect.
+ */
+ uint32_t dirty = 0;
+ if (cmd_buffer->state.descriptors_dirty)
+ dirty = gen7_cmd_buffer_flush_descriptor_sets(cmd_buffer);
+
+ if (cmd_buffer->state.push_constants_dirty)
+ dirty |= cmd_buffer_flush_push_constants(cmd_buffer);
+
+ if (dirty)
+ gen7_cmd_buffer_emit_descriptor_pointers(cmd_buffer, dirty);
+
+ if (cmd_buffer->state.dirty & ANV_CMD_DIRTY_DYNAMIC_VIEWPORT)
+ gen8_cmd_buffer_emit_viewport(cmd_buffer);
+
+ if (cmd_buffer->state.dirty & ANV_CMD_DIRTY_DYNAMIC_SCISSOR)
+ gen7_cmd_buffer_emit_scissor(cmd_buffer);
+
+ if (cmd_buffer->state.dirty & (ANV_CMD_DIRTY_PIPELINE |
+ ANV_CMD_DIRTY_DYNAMIC_LINE_WIDTH)) {
+ uint32_t sf_dw[GENX(3DSTATE_SF_length)];
+ struct GENX(3DSTATE_SF) sf = {
+ GENX(3DSTATE_SF_header),
+ .LineWidth = cmd_buffer->state.dynamic.line_width,
+ };
+ GENX(3DSTATE_SF_pack)(NULL, sf_dw, &sf);
+ /* FIXME: gen9.fs */
+ anv_batch_emit_merge(&cmd_buffer->batch, sf_dw, pipeline->gen8.sf);
+ }
+
+ if (cmd_buffer->state.dirty & (ANV_CMD_DIRTY_PIPELINE |
+ ANV_CMD_DIRTY_DYNAMIC_DEPTH_BIAS)){
+ bool enable_bias = cmd_buffer->state.dynamic.depth_bias.bias != 0.0f ||
+ cmd_buffer->state.dynamic.depth_bias.slope != 0.0f;
+
+ uint32_t raster_dw[GENX(3DSTATE_RASTER_length)];
+ struct GENX(3DSTATE_RASTER) raster = {
+ GENX(3DSTATE_RASTER_header),
+ .GlobalDepthOffsetEnableSolid = enable_bias,
+ .GlobalDepthOffsetEnableWireframe = enable_bias,
+ .GlobalDepthOffsetEnablePoint = enable_bias,
+ .GlobalDepthOffsetConstant = cmd_buffer->state.dynamic.depth_bias.bias,
+ .GlobalDepthOffsetScale = cmd_buffer->state.dynamic.depth_bias.slope,
+ .GlobalDepthOffsetClamp = cmd_buffer->state.dynamic.depth_bias.clamp
+ };
+ GENX(3DSTATE_RASTER_pack)(NULL, raster_dw, &raster);
+ anv_batch_emit_merge(&cmd_buffer->batch, raster_dw,
+ pipeline->gen8.raster);
+ }
+
+ /* Stencil reference values moved from COLOR_CALC_STATE in gen8 to
+ * 3DSTATE_WM_DEPTH_STENCIL in gen9. That means the dirty bits gets split
+ * across different state packets for gen8 and gen9. We handle that by
+ * using a big old #if switch here.
+ */
+#if ANV_GEN == 8
+ if (cmd_buffer->state.dirty & (ANV_CMD_DIRTY_DYNAMIC_BLEND_CONSTANTS |
+ ANV_CMD_DIRTY_DYNAMIC_STENCIL_REFERENCE)) {
+ struct anv_state cc_state =
+ anv_cmd_buffer_alloc_dynamic_state(cmd_buffer,
+ GEN8_COLOR_CALC_STATE_length * 4,
+ 64);
+ struct GEN8_COLOR_CALC_STATE cc = {
+ .BlendConstantColorRed = cmd_buffer->state.dynamic.blend_constants[0],
+ .BlendConstantColorGreen = cmd_buffer->state.dynamic.blend_constants[1],
+ .BlendConstantColorBlue = cmd_buffer->state.dynamic.blend_constants[2],
+ .BlendConstantColorAlpha = cmd_buffer->state.dynamic.blend_constants[3],
+ .StencilReferenceValue =
+ cmd_buffer->state.dynamic.stencil_reference.front,
+ .BackFaceStencilReferenceValue =
+ cmd_buffer->state.dynamic.stencil_reference.back,
+ };
+ GEN8_COLOR_CALC_STATE_pack(NULL, cc_state.map, &cc);
+
+ if (!cmd_buffer->device->info.has_llc)
+ anv_state_clflush(cc_state);
+
+ anv_batch_emit(&cmd_buffer->batch,
+ GEN8_3DSTATE_CC_STATE_POINTERS,
+ .ColorCalcStatePointer = cc_state.offset,
+ .ColorCalcStatePointerValid = true);
+ }
+
+ if (cmd_buffer->state.dirty & (ANV_CMD_DIRTY_PIPELINE |
+ ANV_CMD_DIRTY_DYNAMIC_STENCIL_COMPARE_MASK |
+ ANV_CMD_DIRTY_DYNAMIC_STENCIL_WRITE_MASK)) {
+ uint32_t wm_depth_stencil_dw[GEN8_3DSTATE_WM_DEPTH_STENCIL_length];
+
+ struct GEN8_3DSTATE_WM_DEPTH_STENCIL wm_depth_stencil = {
+ GEN8_3DSTATE_WM_DEPTH_STENCIL_header,
+
+ /* Is this what we need to do? */
+ .StencilBufferWriteEnable =
+ cmd_buffer->state.dynamic.stencil_write_mask.front != 0,
+
+ .StencilTestMask =
+ cmd_buffer->state.dynamic.stencil_compare_mask.front & 0xff,
+ .StencilWriteMask =
+ cmd_buffer->state.dynamic.stencil_write_mask.front & 0xff,
+
+ .BackfaceStencilTestMask =
+ cmd_buffer->state.dynamic.stencil_compare_mask.back & 0xff,
+ .BackfaceStencilWriteMask =
+ cmd_buffer->state.dynamic.stencil_write_mask.back & 0xff,
+ };
+ GEN8_3DSTATE_WM_DEPTH_STENCIL_pack(NULL, wm_depth_stencil_dw,
+ &wm_depth_stencil);
+
+ anv_batch_emit_merge(&cmd_buffer->batch, wm_depth_stencil_dw,
+ pipeline->gen8.wm_depth_stencil);
+ }
+#else
+ if (cmd_buffer->state.dirty & ANV_CMD_DIRTY_DYNAMIC_BLEND_CONSTANTS) {
+ struct anv_state cc_state =
+ anv_cmd_buffer_alloc_dynamic_state(cmd_buffer,
+ GEN9_COLOR_CALC_STATE_length * 4,
+ 64);
+ struct GEN9_COLOR_CALC_STATE cc = {
+ .BlendConstantColorRed = cmd_buffer->state.dynamic.blend_constants[0],
+ .BlendConstantColorGreen = cmd_buffer->state.dynamic.blend_constants[1],
+ .BlendConstantColorBlue = cmd_buffer->state.dynamic.blend_constants[2],
+ .BlendConstantColorAlpha = cmd_buffer->state.dynamic.blend_constants[3],
+ };
+ GEN9_COLOR_CALC_STATE_pack(NULL, cc_state.map, &cc);
+
+ if (!cmd_buffer->device->info.has_llc)
+ anv_state_clflush(cc_state);
+
+ anv_batch_emit(&cmd_buffer->batch,
+ GEN9_3DSTATE_CC_STATE_POINTERS,
+ .ColorCalcStatePointer = cc_state.offset,
+ .ColorCalcStatePointerValid = true);
+ }
+
+ if (cmd_buffer->state.dirty & (ANV_CMD_DIRTY_PIPELINE |
+ ANV_CMD_DIRTY_DYNAMIC_STENCIL_COMPARE_MASK |
+ ANV_CMD_DIRTY_DYNAMIC_STENCIL_WRITE_MASK |
+ ANV_CMD_DIRTY_DYNAMIC_STENCIL_REFERENCE)) {
+ uint32_t dwords[GEN9_3DSTATE_WM_DEPTH_STENCIL_length];
+ struct anv_dynamic_state *d = &cmd_buffer->state.dynamic;
+ struct GEN9_3DSTATE_WM_DEPTH_STENCIL wm_depth_stencil = {
+ GEN9_3DSTATE_WM_DEPTH_STENCIL_header,
+
+ .StencilBufferWriteEnable = d->stencil_write_mask.front != 0,
+
+ .StencilTestMask = d->stencil_compare_mask.front & 0xff,
+ .StencilWriteMask = d->stencil_write_mask.front & 0xff,
+
+ .BackfaceStencilTestMask = d->stencil_compare_mask.back & 0xff,
+ .BackfaceStencilWriteMask = d->stencil_write_mask.back & 0xff,
+
+ .StencilReferenceValue = d->stencil_reference.front,
+ .BackfaceStencilReferenceValue = d->stencil_reference.back
+ };
+ GEN9_3DSTATE_WM_DEPTH_STENCIL_pack(NULL, dwords, &wm_depth_stencil);
+
+ anv_batch_emit_merge(&cmd_buffer->batch, dwords,
+ pipeline->gen9.wm_depth_stencil);
+ }
+#endif
+
+ if (cmd_buffer->state.dirty & (ANV_CMD_DIRTY_PIPELINE |
+ ANV_CMD_DIRTY_INDEX_BUFFER)) {
+ anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_VF),
+ .IndexedDrawCutIndexEnable = pipeline->primitive_restart,
+ .CutIndex = cmd_buffer->state.restart_index,
+ );
+ }
+
+ cmd_buffer->state.vb_dirty &= ~vb_emit;
+ cmd_buffer->state.dirty = 0;
+}
+
+void genX(CmdDraw)(
+ VkCommandBuffer commandBuffer,
+ uint32_t vertexCount,
+ uint32_t instanceCount,
+ uint32_t firstVertex,
+ uint32_t firstInstance)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+
+ cmd_buffer_flush_state(cmd_buffer);
+
+ anv_batch_emit(&cmd_buffer->batch, GENX(3DPRIMITIVE),
+ .VertexAccessType = SEQUENTIAL,
+ .VertexCountPerInstance = vertexCount,
+ .StartVertexLocation = firstVertex,
+ .InstanceCount = instanceCount,
+ .StartInstanceLocation = firstInstance,
+ .BaseVertexLocation = 0);
+}
+
+void genX(CmdDrawIndexed)(
+ VkCommandBuffer commandBuffer,
+ uint32_t indexCount,
+ uint32_t instanceCount,
+ uint32_t firstIndex,
+ int32_t vertexOffset,
+ uint32_t firstInstance)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+
+ cmd_buffer_flush_state(cmd_buffer);
+
+ anv_batch_emit(&cmd_buffer->batch, GENX(3DPRIMITIVE),
+ .VertexAccessType = RANDOM,
+ .VertexCountPerInstance = indexCount,
+ .StartVertexLocation = firstIndex,
+ .InstanceCount = instanceCount,
+ .StartInstanceLocation = firstInstance,
+ .BaseVertexLocation = vertexOffset);
+}
+
+/* Auto-Draw / Indirect Registers */
+#define GEN7_3DPRIM_END_OFFSET 0x2420
+#define GEN7_3DPRIM_START_VERTEX 0x2430
+#define GEN7_3DPRIM_VERTEX_COUNT 0x2434
+#define GEN7_3DPRIM_INSTANCE_COUNT 0x2438
+#define GEN7_3DPRIM_START_INSTANCE 0x243C
+#define GEN7_3DPRIM_BASE_VERTEX 0x2440
+
+void genX(CmdDrawIndirect)(
+ VkCommandBuffer commandBuffer,
+ VkBuffer _buffer,
+ VkDeviceSize offset,
+ uint32_t drawCount,
+ uint32_t stride)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_buffer, buffer, _buffer);
+ struct anv_bo *bo = buffer->bo;
+ uint32_t bo_offset = buffer->offset + offset;
+
+ cmd_buffer_flush_state(cmd_buffer);
+
+ emit_lrm(&cmd_buffer->batch, GEN7_3DPRIM_VERTEX_COUNT, bo, bo_offset);
+ emit_lrm(&cmd_buffer->batch, GEN7_3DPRIM_INSTANCE_COUNT, bo, bo_offset + 4);
+ emit_lrm(&cmd_buffer->batch, GEN7_3DPRIM_START_VERTEX, bo, bo_offset + 8);
+ emit_lrm(&cmd_buffer->batch, GEN7_3DPRIM_START_INSTANCE, bo, bo_offset + 12);
+ emit_lri(&cmd_buffer->batch, GEN7_3DPRIM_BASE_VERTEX, 0);
+
+ anv_batch_emit(&cmd_buffer->batch, GENX(3DPRIMITIVE),
+ .IndirectParameterEnable = true,
+ .VertexAccessType = SEQUENTIAL);
+}
+
+void genX(CmdBindIndexBuffer)(
+ VkCommandBuffer commandBuffer,
+ VkBuffer _buffer,
+ VkDeviceSize offset,
+ VkIndexType indexType)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_buffer, buffer, _buffer);
+
+ static const uint32_t vk_to_gen_index_type[] = {
+ [VK_INDEX_TYPE_UINT16] = INDEX_WORD,
+ [VK_INDEX_TYPE_UINT32] = INDEX_DWORD,
+ };
+
+ static const uint32_t restart_index_for_type[] = {
+ [VK_INDEX_TYPE_UINT16] = UINT16_MAX,
+ [VK_INDEX_TYPE_UINT32] = UINT32_MAX,
+ };
+
+ cmd_buffer->state.restart_index = restart_index_for_type[indexType];
+
+ anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_INDEX_BUFFER),
+ .IndexFormat = vk_to_gen_index_type[indexType],
+ .MemoryObjectControlState = GENX(MOCS),
+ .BufferStartingAddress = { buffer->bo, buffer->offset + offset },
+ .BufferSize = buffer->size - offset);
+
+ cmd_buffer->state.dirty |= ANV_CMD_DIRTY_INDEX_BUFFER;
+}
+
+static VkResult
+flush_compute_descriptor_set(struct anv_cmd_buffer *cmd_buffer)
+{
+ struct anv_device *device = cmd_buffer->device;
+ struct anv_pipeline *pipeline = cmd_buffer->state.compute_pipeline;
+ struct anv_state surfaces = { 0, }, samplers = { 0, };
+ VkResult result;
+
+ result = anv_cmd_buffer_emit_samplers(cmd_buffer,
+ MESA_SHADER_COMPUTE, &samplers);
+ if (result != VK_SUCCESS)
+ return result;
+ result = anv_cmd_buffer_emit_binding_table(cmd_buffer,
+ MESA_SHADER_COMPUTE, &surfaces);
+ if (result != VK_SUCCESS)
+ return result;
+
+ struct anv_state push_state = anv_cmd_buffer_cs_push_constants(cmd_buffer);
+
+ const struct brw_cs_prog_data *cs_prog_data = &pipeline->cs_prog_data;
+ const struct brw_stage_prog_data *prog_data = &cs_prog_data->base;
+
+ unsigned local_id_dwords = cs_prog_data->local_invocation_id_regs * 8;
+ unsigned push_constant_data_size =
+ (prog_data->nr_params + local_id_dwords) * 4;
+ unsigned reg_aligned_constant_size = ALIGN(push_constant_data_size, 32);
+ unsigned push_constant_regs = reg_aligned_constant_size / 32;
+
+ if (push_state.alloc_size) {
+ anv_batch_emit(&cmd_buffer->batch, GENX(MEDIA_CURBE_LOAD),
+ .CURBETotalDataLength = push_state.alloc_size,
+ .CURBEDataStartAddress = push_state.offset);
+ }
+
+ assert(prog_data->total_shared <= 64 * 1024);
+ uint32_t slm_size = 0;
+ if (prog_data->total_shared > 0) {
+ /* slm_size is in 4k increments, but must be a power of 2. */
+ slm_size = 4 * 1024;
+ while (slm_size < prog_data->total_shared)
+ slm_size <<= 1;
+ slm_size /= 4 * 1024;
+ }
+
+ struct anv_state state =
+ anv_state_pool_emit(&device->dynamic_state_pool,
+ GENX(INTERFACE_DESCRIPTOR_DATA), 64,
+ .KernelStartPointer = pipeline->cs_simd,
+ .KernelStartPointerHigh = 0,
+ .BindingTablePointer = surfaces.offset,
+ .BindingTableEntryCount = 0,
+ .SamplerStatePointer = samplers.offset,
+ .SamplerCount = 0,
+ .ConstantIndirectURBEntryReadLength = push_constant_regs,
+ .ConstantURBEntryReadOffset = 0,
+ .BarrierEnable = cs_prog_data->uses_barrier,
+ .SharedLocalMemorySize = slm_size,
+ .NumberofThreadsinGPGPUThreadGroup =
+ pipeline->cs_thread_width_max);
+
+ uint32_t size = GENX(INTERFACE_DESCRIPTOR_DATA_length) * sizeof(uint32_t);
+ anv_batch_emit(&cmd_buffer->batch, GENX(MEDIA_INTERFACE_DESCRIPTOR_LOAD),
+ .InterfaceDescriptorTotalLength = size,
+ .InterfaceDescriptorDataStartAddress = state.offset);
+
+ return VK_SUCCESS;
+}
+
+static void
+cmd_buffer_flush_compute_state(struct anv_cmd_buffer *cmd_buffer)
+{
+ struct anv_pipeline *pipeline = cmd_buffer->state.compute_pipeline;
+ VkResult result;
+
+ assert(pipeline->active_stages == VK_SHADER_STAGE_COMPUTE_BIT);
+
+ bool needs_slm = pipeline->cs_prog_data.base.total_shared > 0;
+ config_l3(cmd_buffer, needs_slm);
+
+ if (cmd_buffer->state.current_pipeline != GPGPU) {
+#if ANV_GEN < 10
+ /* From the Broadwell PRM, Volume 2a: Instructions, PIPELINE_SELECT:
+ *
+ * Software must clear the COLOR_CALC_STATE Valid field in
+ * 3DSTATE_CC_STATE_POINTERS command prior to send a PIPELINE_SELECT
+ * with Pipeline Select set to GPGPU.
+ *
+ * The internal hardware docs recommend the same workaround for Gen9
+ * hardware too.
+ */
+ anv_batch_emit(&cmd_buffer->batch,
+ GENX(3DSTATE_CC_STATE_POINTERS));
+#endif
+
+ anv_batch_emit(&cmd_buffer->batch, GENX(PIPELINE_SELECT),
+#if ANV_GEN >= 9
+ .MaskBits = 3,
+#endif
+ .PipelineSelection = GPGPU);
+ cmd_buffer->state.current_pipeline = GPGPU;
+ }
+
+ if (cmd_buffer->state.compute_dirty & ANV_CMD_DIRTY_PIPELINE)
+ anv_batch_emit_batch(&cmd_buffer->batch, &pipeline->batch);
+
+ if ((cmd_buffer->state.descriptors_dirty & VK_SHADER_STAGE_COMPUTE_BIT) ||
+ (cmd_buffer->state.compute_dirty & ANV_CMD_DIRTY_PIPELINE)) {
+ result = flush_compute_descriptor_set(cmd_buffer);
+ assert(result == VK_SUCCESS);
+ cmd_buffer->state.descriptors_dirty &= ~VK_SHADER_STAGE_COMPUTE_BIT;
+ }
+
+ cmd_buffer->state.compute_dirty = 0;
+}
+
+void genX(CmdDrawIndexedIndirect)(
+ VkCommandBuffer commandBuffer,
+ VkBuffer _buffer,
+ VkDeviceSize offset,
+ uint32_t drawCount,
+ uint32_t stride)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_buffer, buffer, _buffer);
+ struct anv_bo *bo = buffer->bo;
+ uint32_t bo_offset = buffer->offset + offset;
+
+ cmd_buffer_flush_state(cmd_buffer);
+
+ emit_lrm(&cmd_buffer->batch, GEN7_3DPRIM_VERTEX_COUNT, bo, bo_offset);
+ emit_lrm(&cmd_buffer->batch, GEN7_3DPRIM_INSTANCE_COUNT, bo, bo_offset + 4);
+ emit_lrm(&cmd_buffer->batch, GEN7_3DPRIM_START_VERTEX, bo, bo_offset + 8);
+ emit_lrm(&cmd_buffer->batch, GEN7_3DPRIM_BASE_VERTEX, bo, bo_offset + 12);
+ emit_lrm(&cmd_buffer->batch, GEN7_3DPRIM_START_INSTANCE, bo, bo_offset + 16);
+
+ anv_batch_emit(&cmd_buffer->batch, GENX(3DPRIMITIVE),
+ .IndirectParameterEnable = true,
+ .VertexAccessType = RANDOM);
+}
+
+void genX(CmdDispatch)(
+ VkCommandBuffer commandBuffer,
+ uint32_t x,
+ uint32_t y,
+ uint32_t z)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ struct anv_pipeline *pipeline = cmd_buffer->state.compute_pipeline;
+ struct brw_cs_prog_data *prog_data = &pipeline->cs_prog_data;
+
+ if (prog_data->uses_num_work_groups) {
+ struct anv_state state =
+ anv_cmd_buffer_alloc_dynamic_state(cmd_buffer, 12, 4);
+ uint32_t *sizes = state.map;
+ sizes[0] = x;
+ sizes[1] = y;
+ sizes[2] = z;
+ if (!cmd_buffer->device->info.has_llc)
+ anv_state_clflush(state);
+ cmd_buffer->state.num_workgroups_offset = state.offset;
+ cmd_buffer->state.num_workgroups_bo =
+ &cmd_buffer->device->dynamic_state_block_pool.bo;
+ }
+
+ cmd_buffer_flush_compute_state(cmd_buffer);
+
+ anv_batch_emit(&cmd_buffer->batch, GENX(GPGPU_WALKER),
+ .SIMDSize = prog_data->simd_size / 16,
+ .ThreadDepthCounterMaximum = 0,
+ .ThreadHeightCounterMaximum = 0,
+ .ThreadWidthCounterMaximum = pipeline->cs_thread_width_max - 1,
+ .ThreadGroupIDXDimension = x,
+ .ThreadGroupIDYDimension = y,
+ .ThreadGroupIDZDimension = z,
+ .RightExecutionMask = pipeline->cs_right_mask,
+ .BottomExecutionMask = 0xffffffff);
+
+ anv_batch_emit(&cmd_buffer->batch, GENX(MEDIA_STATE_FLUSH));
+}
+
+#define GPGPU_DISPATCHDIMX 0x2500
+#define GPGPU_DISPATCHDIMY 0x2504
+#define GPGPU_DISPATCHDIMZ 0x2508
+
+void genX(CmdDispatchIndirect)(
+ VkCommandBuffer commandBuffer,
+ VkBuffer _buffer,
+ VkDeviceSize offset)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_buffer, buffer, _buffer);
+ struct anv_pipeline *pipeline = cmd_buffer->state.compute_pipeline;
+ struct brw_cs_prog_data *prog_data = &pipeline->cs_prog_data;
+ struct anv_bo *bo = buffer->bo;
+ uint32_t bo_offset = buffer->offset + offset;
+
+ if (prog_data->uses_num_work_groups) {
+ cmd_buffer->state.num_workgroups_offset = bo_offset;
+ cmd_buffer->state.num_workgroups_bo = bo;
+ }
+
+ cmd_buffer_flush_compute_state(cmd_buffer);
+
+ emit_lrm(&cmd_buffer->batch, GPGPU_DISPATCHDIMX, bo, bo_offset);
+ emit_lrm(&cmd_buffer->batch, GPGPU_DISPATCHDIMY, bo, bo_offset + 4);
+ emit_lrm(&cmd_buffer->batch, GPGPU_DISPATCHDIMZ, bo, bo_offset + 8);
+
+ anv_batch_emit(&cmd_buffer->batch, GENX(GPGPU_WALKER),
+ .IndirectParameterEnable = true,
+ .SIMDSize = prog_data->simd_size / 16,
+ .ThreadDepthCounterMaximum = 0,
+ .ThreadHeightCounterMaximum = 0,
+ .ThreadWidthCounterMaximum = pipeline->cs_thread_width_max - 1,
+ .RightExecutionMask = pipeline->cs_right_mask,
+ .BottomExecutionMask = 0xffffffff);
+
+ anv_batch_emit(&cmd_buffer->batch, GENX(MEDIA_STATE_FLUSH));
+}
+
+static void
+cmd_buffer_emit_depth_stencil(struct anv_cmd_buffer *cmd_buffer)
+{
+ const struct anv_framebuffer *fb = cmd_buffer->state.framebuffer;
+ const struct anv_image_view *iview =
+ anv_cmd_buffer_get_depth_stencil_view(cmd_buffer);
+ const struct anv_image *image = iview ? iview->image : NULL;
+
+ /* XXX: isl needs to grow depth format support */
+ const struct anv_format *anv_format =
+ iview ? anv_format_for_vk_format(iview->vk_format) : NULL;
+
+ const bool has_depth = iview && anv_format->depth_format;
+ const bool has_stencil = iview && anv_format->has_stencil;
+
+ /* FIXME: Implement the PMA stall W/A */
+ /* FIXME: Width and Height are wrong */
+
+ /* Emit 3DSTATE_DEPTH_BUFFER */
+ if (has_depth) {
+ anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_DEPTH_BUFFER),
+ .SurfaceType = SURFTYPE_2D,
+ .DepthWriteEnable = anv_format->depth_format,
+ .StencilWriteEnable = has_stencil,
+ .HierarchicalDepthBufferEnable = false,
+ .SurfaceFormat = anv_format->depth_format,
+ .SurfacePitch = image->depth_surface.isl.row_pitch - 1,
+ .SurfaceBaseAddress = {
+ .bo = image->bo,
+ .offset = image->depth_surface.offset,
+ },
+ .Height = fb->height - 1,
+ .Width = fb->width - 1,
+ .LOD = 0,
+ .Depth = 1 - 1,
+ .MinimumArrayElement = 0,
+ .DepthBufferObjectControlState = GENX(MOCS),
+ .RenderTargetViewExtent = 1 - 1,
+ .SurfaceQPitch = isl_surf_get_array_pitch_el_rows(&image->depth_surface.isl) >> 2);
+ } else {
+ /* Even when no depth buffer is present, the hardware requires that
+ * 3DSTATE_DEPTH_BUFFER be programmed correctly. The Broadwell PRM says:
+ *
+ * If a null depth buffer is bound, the driver must instead bind depth as:
+ * 3DSTATE_DEPTH.SurfaceType = SURFTYPE_2D
+ * 3DSTATE_DEPTH.Width = 1
+ * 3DSTATE_DEPTH.Height = 1
+ * 3DSTATE_DEPTH.SuraceFormat = D16_UNORM
+ * 3DSTATE_DEPTH.SurfaceBaseAddress = 0
+ * 3DSTATE_DEPTH.HierarchicalDepthBufferEnable = 0
+ * 3DSTATE_WM_DEPTH_STENCIL.DepthTestEnable = 0
+ * 3DSTATE_WM_DEPTH_STENCIL.DepthBufferWriteEnable = 0
+ *
+ * The PRM is wrong, though. The width and height must be programmed to
+ * actual framebuffer's width and height, even when neither depth buffer
+ * nor stencil buffer is present.
+ */
+ anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_DEPTH_BUFFER),
+ .SurfaceType = SURFTYPE_2D,
+ .SurfaceFormat = D16_UNORM,
+ .Width = fb->width - 1,
+ .Height = fb->height - 1,
+ .StencilWriteEnable = has_stencil);
+ }
+
+ /* Emit 3DSTATE_STENCIL_BUFFER */
+ if (has_stencil) {
+ anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_STENCIL_BUFFER),
+ .StencilBufferEnable = true,
+ .StencilBufferObjectControlState = GENX(MOCS),
+
+ /* Stencil buffers have strange pitch. The PRM says:
+ *
+ * The pitch must be set to 2x the value computed based on width,
+ * as the stencil buffer is stored with two rows interleaved.
+ */
+ .SurfacePitch = 2 * image->stencil_surface.isl.row_pitch - 1,
+
+ .SurfaceBaseAddress = {
+ .bo = image->bo,
+ .offset = image->offset + image->stencil_surface.offset,
+ },
+ .SurfaceQPitch = isl_surf_get_array_pitch_el_rows(&image->stencil_surface.isl) >> 2);
+ } else {
+ anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_STENCIL_BUFFER));
+ }
+
+ /* Disable hierarchial depth buffers. */
+ anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_HIER_DEPTH_BUFFER));
+
+ /* Clear the clear params. */
+ anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_CLEAR_PARAMS));
+}
+
+/**
+ * @see anv_cmd_buffer_set_subpass()
+ */
+void
+genX(cmd_buffer_set_subpass)(struct anv_cmd_buffer *cmd_buffer,
+ struct anv_subpass *subpass)
+{
+ cmd_buffer->state.subpass = subpass;
+
+ cmd_buffer->state.descriptors_dirty |= VK_SHADER_STAGE_FRAGMENT_BIT;
+
+ cmd_buffer_emit_depth_stencil(cmd_buffer);
+}
+
+void genX(CmdBeginRenderPass)(
+ VkCommandBuffer commandBuffer,
+ const VkRenderPassBeginInfo* pRenderPassBegin,
+ VkSubpassContents contents)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_render_pass, pass, pRenderPassBegin->renderPass);
+ ANV_FROM_HANDLE(anv_framebuffer, framebuffer, pRenderPassBegin->framebuffer);
+
+ cmd_buffer->state.framebuffer = framebuffer;
+ cmd_buffer->state.pass = pass;
+ anv_cmd_state_setup_attachments(cmd_buffer, pRenderPassBegin);
+
+ flush_pipeline_select_3d(cmd_buffer);
+
+ const VkRect2D *render_area = &pRenderPassBegin->renderArea;
+
+ anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_DRAWING_RECTANGLE),
+ .ClippedDrawingRectangleYMin = render_area->offset.y,
+ .ClippedDrawingRectangleXMin = render_area->offset.x,
+ .ClippedDrawingRectangleYMax =
+ render_area->offset.y + render_area->extent.height - 1,
+ .ClippedDrawingRectangleXMax =
+ render_area->offset.x + render_area->extent.width - 1,
+ .DrawingRectangleOriginY = 0,
+ .DrawingRectangleOriginX = 0);
+
+ genX(cmd_buffer_set_subpass)(cmd_buffer, pass->subpasses);
+ anv_cmd_buffer_clear_subpass(cmd_buffer);
+}
+
+void genX(CmdNextSubpass)(
+ VkCommandBuffer commandBuffer,
+ VkSubpassContents contents)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+
+ assert(cmd_buffer->level == VK_COMMAND_BUFFER_LEVEL_PRIMARY);
+
+ genX(cmd_buffer_set_subpass)(cmd_buffer, cmd_buffer->state.subpass + 1);
+ anv_cmd_buffer_clear_subpass(cmd_buffer);
+}
+
+void genX(CmdEndRenderPass)(
+ VkCommandBuffer commandBuffer)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+
+ /* Emit a flushing pipe control at the end of a pass. This is kind of a
+ * hack but it ensures that render targets always actually get written.
+ * Eventually, we should do flushing based on image format transitions
+ * or something of that nature.
+ */
+ anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL),
+ .PostSyncOperation = NoWrite,
+ .RenderTargetCacheFlushEnable = true,
+ .InstructionCacheInvalidateEnable = true,
+ .DepthCacheFlushEnable = true,
+ .VFCacheInvalidationEnable = true,
+ .TextureCacheInvalidationEnable = true,
+ .CommandStreamerStallEnable = true);
+}
+
+static void
+emit_ps_depth_count(struct anv_batch *batch,
+ struct anv_bo *bo, uint32_t offset)
+{
+ anv_batch_emit(batch, GENX(PIPE_CONTROL),
+ .DestinationAddressType = DAT_PPGTT,
+ .PostSyncOperation = WritePSDepthCount,
+ .DepthStallEnable = true,
+ .Address = { bo, offset });
+}
+
+static void
+emit_query_availability(struct anv_batch *batch,
+ struct anv_bo *bo, uint32_t offset)
+{
+ anv_batch_emit(batch, GENX(PIPE_CONTROL),
+ .DestinationAddressType = DAT_PPGTT,
+ .PostSyncOperation = WriteImmediateData,
+ .Address = { bo, offset },
+ .ImmediateData = 1);
+}
+
+void genX(CmdBeginQuery)(
+ VkCommandBuffer commandBuffer,
+ VkQueryPool queryPool,
+ uint32_t query,
+ VkQueryControlFlags flags)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_query_pool, pool, queryPool);
+
+ /* Workaround: When meta uses the pipeline with the VS disabled, it seems
+ * that the pipelining of the depth write breaks. What we see is that
+ * samples from the render pass clear leaks into the first query
+ * immediately after the clear. Doing a pipecontrol with a post-sync
+ * operation and DepthStallEnable seems to work around the issue.
+ */
+ if (cmd_buffer->state.need_query_wa) {
+ cmd_buffer->state.need_query_wa = false;
+ anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL),
+ .DepthCacheFlushEnable = true,
+ .DepthStallEnable = true);
+ }
+
+ switch (pool->type) {
+ case VK_QUERY_TYPE_OCCLUSION:
+ emit_ps_depth_count(&cmd_buffer->batch, &pool->bo,
+ query * sizeof(struct anv_query_pool_slot));
+ break;
+
+ case VK_QUERY_TYPE_PIPELINE_STATISTICS:
+ default:
+ unreachable("");
+ }
+}
+
+void genX(CmdEndQuery)(
+ VkCommandBuffer commandBuffer,
+ VkQueryPool queryPool,
+ uint32_t query)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_query_pool, pool, queryPool);
+
+ switch (pool->type) {
+ case VK_QUERY_TYPE_OCCLUSION:
+ emit_ps_depth_count(&cmd_buffer->batch, &pool->bo,
+ query * sizeof(struct anv_query_pool_slot) + 8);
+
+ emit_query_availability(&cmd_buffer->batch, &pool->bo,
+ query * sizeof(struct anv_query_pool_slot) + 16);
+ break;
+
+ case VK_QUERY_TYPE_PIPELINE_STATISTICS:
+ default:
+ unreachable("");
+ }
+}
+
+#define TIMESTAMP 0x2358
+
+void genX(CmdWriteTimestamp)(
+ VkCommandBuffer commandBuffer,
+ VkPipelineStageFlagBits pipelineStage,
+ VkQueryPool queryPool,
+ uint32_t query)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_query_pool, pool, queryPool);
+ uint32_t offset = query * sizeof(struct anv_query_pool_slot);
+
+ assert(pool->type == VK_QUERY_TYPE_TIMESTAMP);
+
+ switch (pipelineStage) {
+ case VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT:
+ anv_batch_emit(&cmd_buffer->batch, GENX(MI_STORE_REGISTER_MEM),
+ .RegisterAddress = TIMESTAMP,
+ .MemoryAddress = { &pool->bo, offset });
+ anv_batch_emit(&cmd_buffer->batch, GENX(MI_STORE_REGISTER_MEM),
+ .RegisterAddress = TIMESTAMP + 4,
+ .MemoryAddress = { &pool->bo, offset + 4 });
+ break;
+
+ default:
+ /* Everything else is bottom-of-pipe */
+ anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL),
+ .DestinationAddressType = DAT_PPGTT,
+ .PostSyncOperation = WriteTimestamp,
+ .Address = { &pool->bo, offset });
+ break;
+ }
+
+ emit_query_availability(&cmd_buffer->batch, &pool->bo, query + 16);
+}
+
+#define alu_opcode(v) __gen_field((v), 20, 31)
+#define alu_operand1(v) __gen_field((v), 10, 19)
+#define alu_operand2(v) __gen_field((v), 0, 9)
+#define alu(opcode, operand1, operand2) \
+ alu_opcode(opcode) | alu_operand1(operand1) | alu_operand2(operand2)
+
+#define OPCODE_NOOP 0x000
+#define OPCODE_LOAD 0x080
+#define OPCODE_LOADINV 0x480
+#define OPCODE_LOAD0 0x081
+#define OPCODE_LOAD1 0x481
+#define OPCODE_ADD 0x100
+#define OPCODE_SUB 0x101
+#define OPCODE_AND 0x102
+#define OPCODE_OR 0x103
+#define OPCODE_XOR 0x104
+#define OPCODE_STORE 0x180
+#define OPCODE_STOREINV 0x580
+
+#define OPERAND_R0 0x00
+#define OPERAND_R1 0x01
+#define OPERAND_R2 0x02
+#define OPERAND_R3 0x03
+#define OPERAND_R4 0x04
+#define OPERAND_SRCA 0x20
+#define OPERAND_SRCB 0x21
+#define OPERAND_ACCU 0x31
+#define OPERAND_ZF 0x32
+#define OPERAND_CF 0x33
+
+#define CS_GPR(n) (0x2600 + (n) * 8)
+
+static void
+emit_load_alu_reg_u64(struct anv_batch *batch, uint32_t reg,
+ struct anv_bo *bo, uint32_t offset)
+{
+ anv_batch_emit(batch, GENX(MI_LOAD_REGISTER_MEM),
+ .RegisterAddress = reg,
+ .MemoryAddress = { bo, offset });
+ anv_batch_emit(batch, GENX(MI_LOAD_REGISTER_MEM),
+ .RegisterAddress = reg + 4,
+ .MemoryAddress = { bo, offset + 4 });
+}
+
+static void
+store_query_result(struct anv_batch *batch, uint32_t reg,
+ struct anv_bo *bo, uint32_t offset, VkQueryResultFlags flags)
+{
+ anv_batch_emit(batch, GENX(MI_STORE_REGISTER_MEM),
+ .RegisterAddress = reg,
+ .MemoryAddress = { bo, offset });
+
+ if (flags & VK_QUERY_RESULT_64_BIT)
+ anv_batch_emit(batch, GENX(MI_STORE_REGISTER_MEM),
+ .RegisterAddress = reg + 4,
+ .MemoryAddress = { bo, offset + 4 });
+}
+
+void genX(CmdCopyQueryPoolResults)(
+ VkCommandBuffer commandBuffer,
+ VkQueryPool queryPool,
+ uint32_t firstQuery,
+ uint32_t queryCount,
+ VkBuffer destBuffer,
+ VkDeviceSize destOffset,
+ VkDeviceSize destStride,
+ VkQueryResultFlags flags)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_query_pool, pool, queryPool);
+ ANV_FROM_HANDLE(anv_buffer, buffer, destBuffer);
+ uint32_t slot_offset, dst_offset;
+
+ if (flags & VK_QUERY_RESULT_WAIT_BIT)
+ anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL),
+ .CommandStreamerStallEnable = true,
+ .StallAtPixelScoreboard = true);
+
+ dst_offset = buffer->offset + destOffset;
+ for (uint32_t i = 0; i < queryCount; i++) {
+
+ slot_offset = (firstQuery + i) * sizeof(struct anv_query_pool_slot);
+ switch (pool->type) {
+ case VK_QUERY_TYPE_OCCLUSION:
+ emit_load_alu_reg_u64(&cmd_buffer->batch,
+ CS_GPR(0), &pool->bo, slot_offset);
+ emit_load_alu_reg_u64(&cmd_buffer->batch,
+ CS_GPR(1), &pool->bo, slot_offset + 8);
+
+ /* FIXME: We need to clamp the result for 32 bit. */
+
+ uint32_t *dw = anv_batch_emitn(&cmd_buffer->batch, 5, GENX(MI_MATH));
+ dw[1] = alu(OPCODE_LOAD, OPERAND_SRCA, OPERAND_R1);
+ dw[2] = alu(OPCODE_LOAD, OPERAND_SRCB, OPERAND_R0);
+ dw[3] = alu(OPCODE_SUB, 0, 0);
+ dw[4] = alu(OPCODE_STORE, OPERAND_R2, OPERAND_ACCU);
+ break;
+
+ case VK_QUERY_TYPE_TIMESTAMP:
+ emit_load_alu_reg_u64(&cmd_buffer->batch,
+ CS_GPR(2), &pool->bo, slot_offset);
+ break;
+
+ default:
+ unreachable("unhandled query type");
+ }
+
+ store_query_result(&cmd_buffer->batch,
+ CS_GPR(2), buffer->bo, dst_offset, flags);
+
+ if (flags & VK_QUERY_RESULT_WITH_AVAILABILITY_BIT) {
+ emit_load_alu_reg_u64(&cmd_buffer->batch, CS_GPR(0),
+ &pool->bo, slot_offset + 16);
+ if (flags & VK_QUERY_RESULT_64_BIT)
+ store_query_result(&cmd_buffer->batch,
+ CS_GPR(0), buffer->bo, dst_offset + 8, flags);
+ else
+ store_query_result(&cmd_buffer->batch,
+ CS_GPR(0), buffer->bo, dst_offset + 4, flags);
+ }
+
+ dst_offset += destStride;
+ }
+}
+
+void genX(CmdSetEvent)(
+ VkCommandBuffer commandBuffer,
+ VkEvent _event,
+ VkPipelineStageFlags stageMask)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_event, event, _event);
+
+ anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL),
+ .DestinationAddressType = DAT_PPGTT,
+ .PostSyncOperation = WriteImmediateData,
+ .Address = {
+ &cmd_buffer->device->dynamic_state_block_pool.bo,
+ event->state.offset
+ },
+ .ImmediateData = VK_EVENT_SET);
+}
+
+void genX(CmdResetEvent)(
+ VkCommandBuffer commandBuffer,
+ VkEvent _event,
+ VkPipelineStageFlags stageMask)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_event, event, _event);
+
+ anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL),
+ .DestinationAddressType = DAT_PPGTT,
+ .PostSyncOperation = WriteImmediateData,
+ .Address = {
+ &cmd_buffer->device->dynamic_state_block_pool.bo,
+ event->state.offset
+ },
+ .ImmediateData = VK_EVENT_RESET);
+}
+
+void genX(CmdWaitEvents)(
+ VkCommandBuffer commandBuffer,
+ uint32_t eventCount,
+ const VkEvent* pEvents,
+ VkPipelineStageFlags srcStageMask,
+ VkPipelineStageFlags destStageMask,
+ uint32_t memoryBarrierCount,
+ const VkMemoryBarrier* pMemoryBarriers,
+ uint32_t bufferMemoryBarrierCount,
+ const VkBufferMemoryBarrier* pBufferMemoryBarriers,
+ uint32_t imageMemoryBarrierCount,
+ const VkImageMemoryBarrier* pImageMemoryBarriers)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ for (uint32_t i = 0; i < eventCount; i++) {
+ ANV_FROM_HANDLE(anv_event, event, pEvents[i]);
+
+ anv_batch_emit(&cmd_buffer->batch, GENX(MI_SEMAPHORE_WAIT),
+ .WaitMode = PollingMode,
+ .CompareOperation = SAD_EQUAL_SDD,
+ .SemaphoreDataDword = VK_EVENT_SET,
+ .SemaphoreAddress = {
+ &cmd_buffer->device->dynamic_state_block_pool.bo,
+ event->state.offset
+ });
+ }
+
+ genX(CmdPipelineBarrier)(commandBuffer, srcStageMask, destStageMask,
+ false, /* byRegion */
+ memoryBarrierCount, pMemoryBarriers,
+ bufferMemoryBarrierCount, pBufferMemoryBarriers,
+ imageMemoryBarrierCount, pImageMemoryBarriers);
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+
+/* Instructions, enums and structures for BDW.
+ *
+ * This file has been generated, do not hand edit.
+ */
+
+#pragma once
+
+#include <stdio.h>
+#include <assert.h>
+
+#ifndef __gen_validate_value
+#define __gen_validate_value(x)
+#endif
+
+#ifndef __gen_field_functions
+#define __gen_field_functions
+
+union __gen_value {
+ float f;
+ uint32_t dw;
+};
+
+static inline uint64_t
+__gen_mbo(uint32_t start, uint32_t end)
+{
+ return (~0ul >> (64 - (end - start + 1))) << start;
+}
+
+static inline uint64_t
+__gen_field(uint64_t v, uint32_t start, uint32_t end)
+{
+ __gen_validate_value(v);
+#if DEBUG
+ if (end - start + 1 < 64)
+ assert(v < 1ul << (end - start + 1));
+#endif
+
+ return v << start;
+}
+
+static inline uint64_t
+__gen_fixed(float v, uint32_t start, uint32_t end,
+ bool is_signed, uint32_t fract_bits)
+{
+ __gen_validate_value(v);
+
+ const float factor = (1 << fract_bits);
+
+ float max, min;
+ if (is_signed) {
+ max = ((1 << (end - start)) - 1) / factor;
+ min = -(1 << (end - start)) / factor;
+ } else {
+ max = ((1 << (end - start + 1)) - 1) / factor;
+ min = 0.0f;
+ }
+
+ if (v > max)
+ v = max;
+ else if (v < min)
+ v = min;
+
+ int32_t int_val = roundf(v * factor);
+
+ if (is_signed)
+ int_val &= (1 << (end - start + 1)) - 1;
+
+ return int_val << start;
+}
+
+static inline uint64_t
+__gen_offset(uint64_t v, uint32_t start, uint32_t end)
+{
+ __gen_validate_value(v);
+#if DEBUG
+ uint64_t mask = (~0ul >> (64 - (end - start + 1))) << start;
+
+ assert((v & ~mask) == 0);
+#endif
+
+ return v;
+}
+
+static inline uint32_t
+__gen_float(float v)
+{
+ __gen_validate_value(v);
+ return ((union __gen_value) { .f = (v) }).dw;
+}
+
+#ifndef __gen_address_type
+#error #define __gen_address_type before including this file
+#endif
+
+#ifndef __gen_user_data
+#error #define __gen_combine_address before including this file
+#endif
+
+#endif
+
+#define GEN8_3DSTATE_URB_VS_length_bias 0x00000002
+#define GEN8_3DSTATE_URB_VS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 48, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_URB_VS_length 0x00000002
+
+struct GEN8_3DSTATE_URB_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t VSURBStartingAddress;
+ uint32_t VSURBEntryAllocationSize;
+ uint32_t VSNumberofURBEntries;
+};
+
+static inline void
+GEN8_3DSTATE_URB_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_URB_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->VSURBStartingAddress, 25, 31) |
+ __gen_field(values->VSURBEntryAllocationSize, 16, 24) |
+ __gen_field(values->VSNumberofURBEntries, 0, 15) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_VS_length_bias 0x00000002
+#define GEN8_3DSTATE_VS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 16, \
+ .DwordLength = 7
+
+#define GEN8_3DSTATE_VS_length 0x00000009
+
+struct GEN8_3DSTATE_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint64_t KernelStartPointer;
+#define Multiple 0
+#define Single 1
+ uint32_t SingleVertexDispatch;
+#define Dmask 0
+#define Vmask 1
+ uint32_t VectorMaskEnable;
+#define NoSamplers 0
+#define _14Samplers 1
+#define _58Samplers 2
+#define _912Samplers 3
+#define _1316Samplers 4
+ uint32_t SamplerCount;
+ uint32_t BindingTableEntryCount;
+#define Normal 0
+#define High 1
+ uint32_t ThreadDispatchPriority;
+#define IEEE754 0
+#define Alternate 1
+ uint32_t FloatingPointMode;
+ bool IllegalOpcodeExceptionEnable;
+ bool AccessesUAV;
+ bool SoftwareExceptionEnable;
+ uint64_t ScratchSpaceBasePointer;
+ uint32_t PerThreadScratchSpace;
+ uint32_t DispatchGRFStartRegisterForURBData;
+ uint32_t VertexURBEntryReadLength;
+ uint32_t VertexURBEntryReadOffset;
+ uint32_t MaximumNumberofThreads;
+ bool StatisticsEnable;
+ bool SIMD8DispatchEnable;
+ bool VertexCacheDisable;
+ bool FunctionEnable;
+ uint32_t VertexURBEntryOutputReadOffset;
+ uint32_t VertexURBEntryOutputLength;
+ uint32_t UserClipDistanceClipTestEnableBitmask;
+ uint32_t UserClipDistanceCullTestEnableBitmask;
+};
+
+static inline void
+GEN8_3DSTATE_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint64_t qw1 =
+ __gen_offset(values->KernelStartPointer, 6, 63) |
+ 0;
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+ dw[3] =
+ __gen_field(values->SingleVertexDispatch, 31, 31) |
+ __gen_field(values->VectorMaskEnable, 30, 30) |
+ __gen_field(values->SamplerCount, 27, 29) |
+ __gen_field(values->BindingTableEntryCount, 18, 25) |
+ __gen_field(values->ThreadDispatchPriority, 17, 17) |
+ __gen_field(values->FloatingPointMode, 16, 16) |
+ __gen_field(values->IllegalOpcodeExceptionEnable, 13, 13) |
+ __gen_field(values->AccessesUAV, 12, 12) |
+ __gen_field(values->SoftwareExceptionEnable, 7, 7) |
+ 0;
+
+ uint64_t qw4 =
+ __gen_offset(values->ScratchSpaceBasePointer, 10, 63) |
+ __gen_field(values->PerThreadScratchSpace, 0, 3) |
+ 0;
+
+ dw[4] = qw4;
+ dw[5] = qw4 >> 32;
+
+ dw[6] =
+ __gen_field(values->DispatchGRFStartRegisterForURBData, 20, 24) |
+ __gen_field(values->VertexURBEntryReadLength, 11, 16) |
+ __gen_field(values->VertexURBEntryReadOffset, 4, 9) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->MaximumNumberofThreads, 23, 31) |
+ __gen_field(values->StatisticsEnable, 10, 10) |
+ __gen_field(values->SIMD8DispatchEnable, 2, 2) |
+ __gen_field(values->VertexCacheDisable, 1, 1) |
+ __gen_field(values->FunctionEnable, 0, 0) |
+ 0;
+
+ dw[8] =
+ __gen_field(values->VertexURBEntryOutputReadOffset, 21, 26) |
+ __gen_field(values->VertexURBEntryOutputLength, 16, 20) |
+ __gen_field(values->UserClipDistanceClipTestEnableBitmask, 8, 15) |
+ __gen_field(values->UserClipDistanceCullTestEnableBitmask, 0, 7) |
+ 0;
+
+}
+
+#define GEN8_GPGPU_CSR_BASE_ADDRESS_length_bias 0x00000002
+#define GEN8_GPGPU_CSR_BASE_ADDRESS_header \
+ .CommandType = 3, \
+ .CommandSubType = 0, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 4, \
+ .DwordLength = 1
+
+#define GEN8_GPGPU_CSR_BASE_ADDRESS_length 0x00000003
+
+struct GEN8_GPGPU_CSR_BASE_ADDRESS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ __gen_address_type GPGPUCSRBaseAddress;
+};
+
+static inline void
+GEN8_GPGPU_CSR_BASE_ADDRESS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_GPGPU_CSR_BASE_ADDRESS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw1 =
+ 0;
+
+ uint64_t qw1 =
+ __gen_combine_address(data, &dw[1], values->GPGPUCSRBaseAddress, dw1);
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+}
+
+#define GEN8_MI_ATOMIC_length_bias 0x00000002
+#define GEN8_MI_ATOMIC_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 47
+
+#define GEN8_MI_ATOMIC_length 0x00000003
+
+struct GEN8_MI_ATOMIC {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define PerProcessGraphicsAddress 0
+#define GlobalGraphicsAddress 1
+ uint32_t MemoryType;
+ uint32_t PostSyncOperation;
+#define DWORD 0
+#define QWORD 1
+#define OCTWORD 2
+#define RESERVED 3
+ uint32_t DataSize;
+ uint32_t InlineData;
+ uint32_t CSSTALL;
+ uint32_t ReturnDataControl;
+ uint32_t ATOMICOPCODE;
+ uint32_t DwordLength;
+ __gen_address_type MemoryAddress;
+ uint32_t Operand1DataDword0;
+ uint32_t Operand2DataDword0;
+ uint32_t Operand1DataDword1;
+ uint32_t Operand2DataDword1;
+ uint32_t Operand1DataDword2;
+ uint32_t Operand2DataDword2;
+ uint32_t Operand1DataDword3;
+ uint32_t Operand2DataDword3;
+};
+
+static inline void
+GEN8_MI_ATOMIC_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_ATOMIC * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->MemoryType, 22, 22) |
+ __gen_field(values->PostSyncOperation, 21, 21) |
+ __gen_field(values->DataSize, 19, 20) |
+ __gen_field(values->InlineData, 18, 18) |
+ __gen_field(values->CSSTALL, 17, 17) |
+ __gen_field(values->ReturnDataControl, 16, 16) |
+ __gen_field(values->ATOMICOPCODE, 8, 15) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw1 =
+ 0;
+
+ uint64_t qw1 =
+ __gen_combine_address(data, &dw[1], values->MemoryAddress, dw1);
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+ dw[3] =
+ __gen_field(values->Operand1DataDword0, 0, 31) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->Operand2DataDword0, 0, 31) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->Operand1DataDword1, 0, 31) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->Operand2DataDword1, 0, 31) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->Operand1DataDword2, 0, 31) |
+ 0;
+
+ dw[8] =
+ __gen_field(values->Operand2DataDword2, 0, 31) |
+ 0;
+
+ dw[9] =
+ __gen_field(values->Operand1DataDword3, 0, 31) |
+ 0;
+
+ dw[10] =
+ __gen_field(values->Operand2DataDword3, 0, 31) |
+ 0;
+
+}
+
+#define GEN8_MI_LOAD_REGISTER_REG_length_bias 0x00000002
+#define GEN8_MI_LOAD_REGISTER_REG_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 42, \
+ .DwordLength = 1
+
+#define GEN8_MI_LOAD_REGISTER_REG_length 0x00000003
+
+struct GEN8_MI_LOAD_REGISTER_REG {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t DwordLength;
+ uint32_t SourceRegisterAddress;
+ uint32_t DestinationRegisterAddress;
+};
+
+static inline void
+GEN8_MI_LOAD_REGISTER_REG_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_LOAD_REGISTER_REG * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->SourceRegisterAddress, 2, 22) |
+ 0;
+
+ dw[2] =
+ __gen_offset(values->DestinationRegisterAddress, 2, 22) |
+ 0;
+
+}
+
+#define GEN8_MI_SEMAPHORE_SIGNAL_length_bias 0x00000002
+#define GEN8_MI_SEMAPHORE_SIGNAL_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 27, \
+ .DwordLength = 0
+
+#define GEN8_MI_SEMAPHORE_SIGNAL_length 0x00000002
+
+struct GEN8_MI_SEMAPHORE_SIGNAL {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t PostSyncOperation;
+#define RCS 0
+#define VCS0 1
+#define BCS 2
+#define VECS 3
+#define VCS1 4
+ uint32_t TargetEngineSelect;
+ uint32_t DwordLength;
+ uint32_t TargetContextID;
+};
+
+static inline void
+GEN8_MI_SEMAPHORE_SIGNAL_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_SEMAPHORE_SIGNAL * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->PostSyncOperation, 21, 21) |
+ __gen_field(values->TargetEngineSelect, 15, 17) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->TargetContextID, 0, 31) |
+ 0;
+
+}
+
+#define GEN8_MI_SEMAPHORE_WAIT_length_bias 0x00000002
+#define GEN8_MI_SEMAPHORE_WAIT_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 28, \
+ .DwordLength = 2
+
+#define GEN8_MI_SEMAPHORE_WAIT_length 0x00000004
+
+struct GEN8_MI_SEMAPHORE_WAIT {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define PerProcessGraphicsAddress 0
+#define GlobalGraphicsAddress 1
+ uint32_t MemoryType;
+#define PollingMode 1
+#define SignalMode 0
+ uint32_t WaitMode;
+#define SAD_GREATER_THAN_SDD 0
+#define SAD_GREATER_THAN_OR_EQUAL_SDD 1
+#define SAD_LESS_THAN_SDD 2
+#define SAD_LESS_THAN_OR_EQUAL_SDD 3
+#define SAD_EQUAL_SDD 4
+#define SAD_NOT_EQUAL_SDD 5
+ uint32_t CompareOperation;
+ uint32_t DwordLength;
+ uint32_t SemaphoreDataDword;
+ __gen_address_type SemaphoreAddress;
+};
+
+static inline void
+GEN8_MI_SEMAPHORE_WAIT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_SEMAPHORE_WAIT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->MemoryType, 22, 22) |
+ __gen_field(values->WaitMode, 15, 15) |
+ __gen_field(values->CompareOperation, 12, 14) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->SemaphoreDataDword, 0, 31) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ uint64_t qw2 =
+ __gen_combine_address(data, &dw[2], values->SemaphoreAddress, dw2);
+
+ dw[2] = qw2;
+ dw[3] = qw2 >> 32;
+
+}
+
+#define GEN8_MI_STORE_REGISTER_MEM_length_bias 0x00000002
+#define GEN8_MI_STORE_REGISTER_MEM_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 36, \
+ .DwordLength = 2
+
+#define GEN8_MI_STORE_REGISTER_MEM_length 0x00000004
+
+struct GEN8_MI_STORE_REGISTER_MEM {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool UseGlobalGTT;
+ uint32_t PredicateEnable;
+ uint32_t DwordLength;
+ uint32_t RegisterAddress;
+ __gen_address_type MemoryAddress;
+};
+
+static inline void
+GEN8_MI_STORE_REGISTER_MEM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_STORE_REGISTER_MEM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->UseGlobalGTT, 22, 22) |
+ __gen_field(values->PredicateEnable, 21, 21) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->RegisterAddress, 2, 22) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ uint64_t qw2 =
+ __gen_combine_address(data, &dw[2], values->MemoryAddress, dw2);
+
+ dw[2] = qw2;
+ dw[3] = qw2 >> 32;
+
+}
+
+#define GEN8_PIPELINE_SELECT_length_bias 0x00000001
+#define GEN8_PIPELINE_SELECT_header \
+ .CommandType = 3, \
+ .CommandSubType = 1, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 4
+
+#define GEN8_PIPELINE_SELECT_length 0x00000001
+
+struct GEN8_PIPELINE_SELECT {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+#define _3D 0
+#define Media 1
+#define GPGPU 2
+ uint32_t PipelineSelection;
+};
+
+static inline void
+GEN8_PIPELINE_SELECT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_PIPELINE_SELECT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->PipelineSelection, 0, 1) |
+ 0;
+
+}
+
+#define GEN8_STATE_BASE_ADDRESS_length_bias 0x00000002
+#define GEN8_STATE_BASE_ADDRESS_header \
+ .CommandType = 3, \
+ .CommandSubType = 0, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 1, \
+ .DwordLength = 14
+
+#define GEN8_STATE_BASE_ADDRESS_length 0x00000010
+
+#define GEN8_MEMORY_OBJECT_CONTROL_STATE_length 0x00000001
+
+struct GEN8_MEMORY_OBJECT_CONTROL_STATE {
+#define UCwithFenceifcoherentcycle 0
+#define UCUncacheable 1
+#define WT 2
+#define WB 3
+ uint32_t MemoryTypeLLCeLLCCacheabilityControl;
+#define eLLCOnlywheneDRAMispresentelsegetsallocatedinLLC 0
+#define LLCOnly 1
+#define LLCeLLCAllowed 2
+#define L3DefertoPATforLLCeLLCselection 3
+ uint32_t TargetCache;
+ uint32_t AgeforQUADLRU;
+};
+
+static inline void
+GEN8_MEMORY_OBJECT_CONTROL_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MEMORY_OBJECT_CONTROL_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->MemoryTypeLLCeLLCCacheabilityControl, 5, 6) |
+ __gen_field(values->TargetCache, 3, 4) |
+ __gen_field(values->AgeforQUADLRU, 0, 1) |
+ 0;
+
+}
+
+struct GEN8_STATE_BASE_ADDRESS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ __gen_address_type GeneralStateBaseAddress;
+ struct GEN8_MEMORY_OBJECT_CONTROL_STATE GeneralStateMemoryObjectControlState;
+ bool GeneralStateBaseAddressModifyEnable;
+ struct GEN8_MEMORY_OBJECT_CONTROL_STATE StatelessDataPortAccessMemoryObjectControlState;
+ __gen_address_type SurfaceStateBaseAddress;
+ struct GEN8_MEMORY_OBJECT_CONTROL_STATE SurfaceStateMemoryObjectControlState;
+ bool SurfaceStateBaseAddressModifyEnable;
+ __gen_address_type DynamicStateBaseAddress;
+ struct GEN8_MEMORY_OBJECT_CONTROL_STATE DynamicStateMemoryObjectControlState;
+ bool DynamicStateBaseAddressModifyEnable;
+ __gen_address_type IndirectObjectBaseAddress;
+ struct GEN8_MEMORY_OBJECT_CONTROL_STATE IndirectObjectMemoryObjectControlState;
+ bool IndirectObjectBaseAddressModifyEnable;
+ __gen_address_type InstructionBaseAddress;
+ struct GEN8_MEMORY_OBJECT_CONTROL_STATE InstructionMemoryObjectControlState;
+ bool InstructionBaseAddressModifyEnable;
+ uint32_t GeneralStateBufferSize;
+ bool GeneralStateBufferSizeModifyEnable;
+ uint32_t DynamicStateBufferSize;
+ bool DynamicStateBufferSizeModifyEnable;
+ uint32_t IndirectObjectBufferSize;
+ bool IndirectObjectBufferSizeModifyEnable;
+ uint32_t InstructionBufferSize;
+ bool InstructionBuffersizeModifyEnable;
+};
+
+static inline void
+GEN8_STATE_BASE_ADDRESS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_STATE_BASE_ADDRESS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw_GeneralStateMemoryObjectControlState;
+ GEN8_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_GeneralStateMemoryObjectControlState, &values->GeneralStateMemoryObjectControlState);
+ uint32_t dw1 =
+ __gen_field(dw_GeneralStateMemoryObjectControlState, 4, 10) |
+ __gen_field(values->GeneralStateBaseAddressModifyEnable, 0, 0) |
+ 0;
+
+ uint64_t qw1 =
+ __gen_combine_address(data, &dw[1], values->GeneralStateBaseAddress, dw1);
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+ uint32_t dw_StatelessDataPortAccessMemoryObjectControlState;
+ GEN8_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_StatelessDataPortAccessMemoryObjectControlState, &values->StatelessDataPortAccessMemoryObjectControlState);
+ dw[3] =
+ __gen_field(dw_StatelessDataPortAccessMemoryObjectControlState, 16, 22) |
+ 0;
+
+ uint32_t dw_SurfaceStateMemoryObjectControlState;
+ GEN8_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_SurfaceStateMemoryObjectControlState, &values->SurfaceStateMemoryObjectControlState);
+ uint32_t dw4 =
+ __gen_field(dw_SurfaceStateMemoryObjectControlState, 4, 10) |
+ __gen_field(values->SurfaceStateBaseAddressModifyEnable, 0, 0) |
+ 0;
+
+ uint64_t qw4 =
+ __gen_combine_address(data, &dw[4], values->SurfaceStateBaseAddress, dw4);
+
+ dw[4] = qw4;
+ dw[5] = qw4 >> 32;
+
+ uint32_t dw_DynamicStateMemoryObjectControlState;
+ GEN8_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_DynamicStateMemoryObjectControlState, &values->DynamicStateMemoryObjectControlState);
+ uint32_t dw6 =
+ __gen_field(dw_DynamicStateMemoryObjectControlState, 4, 10) |
+ __gen_field(values->DynamicStateBaseAddressModifyEnable, 0, 0) |
+ 0;
+
+ uint64_t qw6 =
+ __gen_combine_address(data, &dw[6], values->DynamicStateBaseAddress, dw6);
+
+ dw[6] = qw6;
+ dw[7] = qw6 >> 32;
+
+ uint32_t dw_IndirectObjectMemoryObjectControlState;
+ GEN8_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_IndirectObjectMemoryObjectControlState, &values->IndirectObjectMemoryObjectControlState);
+ uint32_t dw8 =
+ __gen_field(dw_IndirectObjectMemoryObjectControlState, 4, 10) |
+ __gen_field(values->IndirectObjectBaseAddressModifyEnable, 0, 0) |
+ 0;
+
+ uint64_t qw8 =
+ __gen_combine_address(data, &dw[8], values->IndirectObjectBaseAddress, dw8);
+
+ dw[8] = qw8;
+ dw[9] = qw8 >> 32;
+
+ uint32_t dw_InstructionMemoryObjectControlState;
+ GEN8_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_InstructionMemoryObjectControlState, &values->InstructionMemoryObjectControlState);
+ uint32_t dw10 =
+ __gen_field(dw_InstructionMemoryObjectControlState, 4, 10) |
+ __gen_field(values->InstructionBaseAddressModifyEnable, 0, 0) |
+ 0;
+
+ uint64_t qw10 =
+ __gen_combine_address(data, &dw[10], values->InstructionBaseAddress, dw10);
+
+ dw[10] = qw10;
+ dw[11] = qw10 >> 32;
+
+ dw[12] =
+ __gen_field(values->GeneralStateBufferSize, 12, 31) |
+ __gen_field(values->GeneralStateBufferSizeModifyEnable, 0, 0) |
+ 0;
+
+ dw[13] =
+ __gen_field(values->DynamicStateBufferSize, 12, 31) |
+ __gen_field(values->DynamicStateBufferSizeModifyEnable, 0, 0) |
+ 0;
+
+ dw[14] =
+ __gen_field(values->IndirectObjectBufferSize, 12, 31) |
+ __gen_field(values->IndirectObjectBufferSizeModifyEnable, 0, 0) |
+ 0;
+
+ dw[15] =
+ __gen_field(values->InstructionBufferSize, 12, 31) |
+ __gen_field(values->InstructionBuffersizeModifyEnable, 0, 0) |
+ 0;
+
+}
+
+#define GEN8_STATE_PREFETCH_length_bias 0x00000002
+#define GEN8_STATE_PREFETCH_header \
+ .CommandType = 3, \
+ .CommandSubType = 0, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 3, \
+ .DwordLength = 0
+
+#define GEN8_STATE_PREFETCH_length 0x00000002
+
+struct GEN8_STATE_PREFETCH {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ __gen_address_type PrefetchPointer;
+ uint32_t PrefetchCount;
+};
+
+static inline void
+GEN8_STATE_PREFETCH_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_STATE_PREFETCH * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw1 =
+ __gen_field(values->PrefetchCount, 0, 2) |
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->PrefetchPointer, dw1);
+
+}
+
+#define GEN8_STATE_SIP_length_bias 0x00000002
+#define GEN8_STATE_SIP_header \
+ .CommandType = 3, \
+ .CommandSubType = 0, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 2, \
+ .DwordLength = 1
+
+#define GEN8_STATE_SIP_length 0x00000003
+
+struct GEN8_STATE_SIP {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint64_t SystemInstructionPointer;
+};
+
+static inline void
+GEN8_STATE_SIP_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_STATE_SIP * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint64_t qw1 =
+ __gen_offset(values->SystemInstructionPointer, 4, 63) |
+ 0;
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+}
+
+#define GEN8_SWTESS_BASE_ADDRESS_length_bias 0x00000002
+#define GEN8_SWTESS_BASE_ADDRESS_header \
+ .CommandType = 3, \
+ .CommandSubType = 0, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 3, \
+ .DwordLength = 0
+
+#define GEN8_SWTESS_BASE_ADDRESS_length 0x00000002
+
+struct GEN8_SWTESS_BASE_ADDRESS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ __gen_address_type SWTessellationBaseAddress;
+ struct GEN8_MEMORY_OBJECT_CONTROL_STATE SWTessellationMemoryObjectControlState;
+};
+
+static inline void
+GEN8_SWTESS_BASE_ADDRESS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_SWTESS_BASE_ADDRESS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw_SWTessellationMemoryObjectControlState;
+ GEN8_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_SWTessellationMemoryObjectControlState, &values->SWTessellationMemoryObjectControlState);
+ uint32_t dw1 =
+ __gen_field(dw_SWTessellationMemoryObjectControlState, 8, 11) |
+ 0;
+
+ uint64_t qw1 =
+ __gen_combine_address(data, &dw[1], values->SWTessellationBaseAddress, dw1);
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+}
+
+#define GEN8_3DPRIMITIVE_length_bias 0x00000002
+#define GEN8_3DPRIMITIVE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 3, \
+ ._3DCommandSubOpcode = 0, \
+ .DwordLength = 5
+
+#define GEN8_3DPRIMITIVE_length 0x00000007
+
+struct GEN8_3DPRIMITIVE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ bool IndirectParameterEnable;
+ uint32_t UAVCoherencyRequired;
+ bool PredicateEnable;
+ uint32_t DwordLength;
+ bool EndOffsetEnable;
+#define SEQUENTIAL 0
+#define RANDOM 1
+ uint32_t VertexAccessType;
+ uint32_t PrimitiveTopologyType;
+ uint32_t VertexCountPerInstance;
+ uint32_t StartVertexLocation;
+ uint32_t InstanceCount;
+ uint32_t StartInstanceLocation;
+ uint32_t BaseVertexLocation;
+};
+
+static inline void
+GEN8_3DPRIMITIVE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DPRIMITIVE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->IndirectParameterEnable, 10, 10) |
+ __gen_field(values->UAVCoherencyRequired, 9, 9) |
+ __gen_field(values->PredicateEnable, 8, 8) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->EndOffsetEnable, 9, 9) |
+ __gen_field(values->VertexAccessType, 8, 8) |
+ __gen_field(values->PrimitiveTopologyType, 0, 5) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->VertexCountPerInstance, 0, 31) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->StartVertexLocation, 0, 31) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->InstanceCount, 0, 31) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->StartInstanceLocation, 0, 31) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->BaseVertexLocation, 0, 31) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_AA_LINE_PARAMETERS_length_bias 0x00000002
+#define GEN8_3DSTATE_AA_LINE_PARAMETERS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 10, \
+ .DwordLength = 1
+
+#define GEN8_3DSTATE_AA_LINE_PARAMETERS_length 0x00000003
+
+struct GEN8_3DSTATE_AA_LINE_PARAMETERS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ float AAPointCoverageBias;
+ float AACoverageBias;
+ float AAPointCoverageSlope;
+ float AACoverageSlope;
+ float AAPointCoverageEndCapBias;
+ float AACoverageEndCapBias;
+ float AAPointCoverageEndCapSlope;
+ float AACoverageEndCapSlope;
+};
+
+static inline void
+GEN8_3DSTATE_AA_LINE_PARAMETERS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_AA_LINE_PARAMETERS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->AAPointCoverageBias * (1 << 8), 24, 31) |
+ __gen_field(values->AACoverageBias * (1 << 8), 16, 23) |
+ __gen_field(values->AAPointCoverageSlope * (1 << 8), 8, 15) |
+ __gen_field(values->AACoverageSlope * (1 << 8), 0, 7) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->AAPointCoverageEndCapBias * (1 << 8), 24, 31) |
+ __gen_field(values->AACoverageEndCapBias * (1 << 8), 16, 23) |
+ __gen_field(values->AAPointCoverageEndCapSlope * (1 << 8), 8, 15) |
+ __gen_field(values->AACoverageEndCapSlope * (1 << 8), 0, 7) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_BINDING_TABLE_EDIT_DS_length_bias 0x00000002
+#define GEN8_3DSTATE_BINDING_TABLE_EDIT_DS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 70
+
+#define GEN8_3DSTATE_BINDING_TABLE_EDIT_DS_length 0x00000000
+
+#define GEN8_BINDING_TABLE_EDIT_ENTRY_length 0x00000001
+
+struct GEN8_BINDING_TABLE_EDIT_ENTRY {
+ uint32_t BindingTableIndex;
+ uint32_t SurfaceStatePointer;
+};
+
+static inline void
+GEN8_BINDING_TABLE_EDIT_ENTRY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_BINDING_TABLE_EDIT_ENTRY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->BindingTableIndex, 16, 23) |
+ __gen_offset(values->SurfaceStatePointer, 0, 15) |
+ 0;
+
+}
+
+struct GEN8_3DSTATE_BINDING_TABLE_EDIT_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t BindingTableBlockClear;
+#define AllCores 3
+#define Core1 2
+#define Core0 1
+ uint32_t BindingTableEditTarget;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN8_3DSTATE_BINDING_TABLE_EDIT_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_BINDING_TABLE_EDIT_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 8) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->BindingTableBlockClear, 16, 31) |
+ __gen_field(values->BindingTableEditTarget, 0, 1) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN8_3DSTATE_BINDING_TABLE_EDIT_GS_length_bias 0x00000002
+#define GEN8_3DSTATE_BINDING_TABLE_EDIT_GS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 68
+
+#define GEN8_3DSTATE_BINDING_TABLE_EDIT_GS_length 0x00000000
+
+struct GEN8_3DSTATE_BINDING_TABLE_EDIT_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t BindingTableBlockClear;
+#define AllCores 3
+#define Core1 2
+#define Core0 1
+ uint32_t BindingTableEditTarget;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN8_3DSTATE_BINDING_TABLE_EDIT_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_BINDING_TABLE_EDIT_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 8) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->BindingTableBlockClear, 16, 31) |
+ __gen_field(values->BindingTableEditTarget, 0, 1) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN8_3DSTATE_BINDING_TABLE_EDIT_HS_length_bias 0x00000002
+#define GEN8_3DSTATE_BINDING_TABLE_EDIT_HS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 69
+
+#define GEN8_3DSTATE_BINDING_TABLE_EDIT_HS_length 0x00000000
+
+struct GEN8_3DSTATE_BINDING_TABLE_EDIT_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t BindingTableBlockClear;
+#define AllCores 3
+#define Core1 2
+#define Core0 1
+ uint32_t BindingTableEditTarget;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN8_3DSTATE_BINDING_TABLE_EDIT_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_BINDING_TABLE_EDIT_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 8) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->BindingTableBlockClear, 16, 31) |
+ __gen_field(values->BindingTableEditTarget, 0, 1) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN8_3DSTATE_BINDING_TABLE_EDIT_PS_length_bias 0x00000002
+#define GEN8_3DSTATE_BINDING_TABLE_EDIT_PS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 71
+
+#define GEN8_3DSTATE_BINDING_TABLE_EDIT_PS_length 0x00000000
+
+struct GEN8_3DSTATE_BINDING_TABLE_EDIT_PS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t BindingTableBlockClear;
+#define AllCores 3
+#define Core1 2
+#define Core0 1
+ uint32_t BindingTableEditTarget;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN8_3DSTATE_BINDING_TABLE_EDIT_PS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_BINDING_TABLE_EDIT_PS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 8) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->BindingTableBlockClear, 16, 31) |
+ __gen_field(values->BindingTableEditTarget, 0, 1) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN8_3DSTATE_BINDING_TABLE_EDIT_VS_length_bias 0x00000002
+#define GEN8_3DSTATE_BINDING_TABLE_EDIT_VS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 67
+
+#define GEN8_3DSTATE_BINDING_TABLE_EDIT_VS_length 0x00000000
+
+struct GEN8_3DSTATE_BINDING_TABLE_EDIT_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t BindingTableBlockClear;
+#define AllCores 3
+#define Core1 2
+#define Core0 1
+ uint32_t BindingTableEditTarget;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN8_3DSTATE_BINDING_TABLE_EDIT_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_BINDING_TABLE_EDIT_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 8) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->BindingTableBlockClear, 16, 31) |
+ __gen_field(values->BindingTableEditTarget, 0, 1) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN8_3DSTATE_BINDING_TABLE_POINTERS_DS_length_bias 0x00000002
+#define GEN8_3DSTATE_BINDING_TABLE_POINTERS_DS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 40, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_BINDING_TABLE_POINTERS_DS_length 0x00000002
+
+struct GEN8_3DSTATE_BINDING_TABLE_POINTERS_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoDSBindingTable;
+};
+
+static inline void
+GEN8_3DSTATE_BINDING_TABLE_POINTERS_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_BINDING_TABLE_POINTERS_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoDSBindingTable, 5, 15) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_BINDING_TABLE_POINTERS_GS_length_bias 0x00000002
+#define GEN8_3DSTATE_BINDING_TABLE_POINTERS_GS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 41, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_BINDING_TABLE_POINTERS_GS_length 0x00000002
+
+struct GEN8_3DSTATE_BINDING_TABLE_POINTERS_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoGSBindingTable;
+};
+
+static inline void
+GEN8_3DSTATE_BINDING_TABLE_POINTERS_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_BINDING_TABLE_POINTERS_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoGSBindingTable, 5, 15) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_BINDING_TABLE_POINTERS_HS_length_bias 0x00000002
+#define GEN8_3DSTATE_BINDING_TABLE_POINTERS_HS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 39, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_BINDING_TABLE_POINTERS_HS_length 0x00000002
+
+struct GEN8_3DSTATE_BINDING_TABLE_POINTERS_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoHSBindingTable;
+};
+
+static inline void
+GEN8_3DSTATE_BINDING_TABLE_POINTERS_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_BINDING_TABLE_POINTERS_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoHSBindingTable, 5, 15) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_BINDING_TABLE_POINTERS_PS_length_bias 0x00000002
+#define GEN8_3DSTATE_BINDING_TABLE_POINTERS_PS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 42, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_BINDING_TABLE_POINTERS_PS_length 0x00000002
+
+struct GEN8_3DSTATE_BINDING_TABLE_POINTERS_PS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoPSBindingTable;
+};
+
+static inline void
+GEN8_3DSTATE_BINDING_TABLE_POINTERS_PS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_BINDING_TABLE_POINTERS_PS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoPSBindingTable, 5, 15) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_BINDING_TABLE_POINTERS_VS_length_bias 0x00000002
+#define GEN8_3DSTATE_BINDING_TABLE_POINTERS_VS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 38, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_BINDING_TABLE_POINTERS_VS_length 0x00000002
+
+struct GEN8_3DSTATE_BINDING_TABLE_POINTERS_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoVSBindingTable;
+};
+
+static inline void
+GEN8_3DSTATE_BINDING_TABLE_POINTERS_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_BINDING_TABLE_POINTERS_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoVSBindingTable, 5, 15) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_BINDING_TABLE_POOL_ALLOC_length_bias 0x00000002
+#define GEN8_3DSTATE_BINDING_TABLE_POOL_ALLOC_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 25, \
+ .DwordLength = 2
+
+#define GEN8_3DSTATE_BINDING_TABLE_POOL_ALLOC_length 0x00000004
+
+struct GEN8_3DSTATE_BINDING_TABLE_POOL_ALLOC {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ __gen_address_type BindingTablePoolBaseAddress;
+ uint32_t BindingTablePoolEnable;
+ struct GEN8_MEMORY_OBJECT_CONTROL_STATE SurfaceObjectControlState;
+#define NoValidData 0
+ uint32_t BindingTablePoolBufferSize;
+};
+
+static inline void
+GEN8_3DSTATE_BINDING_TABLE_POOL_ALLOC_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_BINDING_TABLE_POOL_ALLOC * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw_SurfaceObjectControlState;
+ GEN8_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_SurfaceObjectControlState, &values->SurfaceObjectControlState);
+ uint32_t dw1 =
+ __gen_field(values->BindingTablePoolEnable, 11, 11) |
+ __gen_field(dw_SurfaceObjectControlState, 0, 6) |
+ 0;
+
+ uint64_t qw1 =
+ __gen_combine_address(data, &dw[1], values->BindingTablePoolBaseAddress, dw1);
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+ dw[3] =
+ __gen_field(values->BindingTablePoolBufferSize, 12, 31) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_BLEND_STATE_POINTERS_length_bias 0x00000002
+#define GEN8_3DSTATE_BLEND_STATE_POINTERS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 36, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_BLEND_STATE_POINTERS_length 0x00000002
+
+struct GEN8_3DSTATE_BLEND_STATE_POINTERS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t BlendStatePointer;
+ bool BlendStatePointerValid;
+};
+
+static inline void
+GEN8_3DSTATE_BLEND_STATE_POINTERS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_BLEND_STATE_POINTERS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->BlendStatePointer, 6, 31) |
+ __gen_field(values->BlendStatePointerValid, 0, 0) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_CC_STATE_POINTERS_length_bias 0x00000002
+#define GEN8_3DSTATE_CC_STATE_POINTERS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 14, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_CC_STATE_POINTERS_length 0x00000002
+
+struct GEN8_3DSTATE_CC_STATE_POINTERS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ColorCalcStatePointer;
+ bool ColorCalcStatePointerValid;
+};
+
+static inline void
+GEN8_3DSTATE_CC_STATE_POINTERS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_CC_STATE_POINTERS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->ColorCalcStatePointer, 6, 31) |
+ __gen_field(values->ColorCalcStatePointerValid, 0, 0) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_CHROMA_KEY_length_bias 0x00000002
+#define GEN8_3DSTATE_CHROMA_KEY_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 4, \
+ .DwordLength = 2
+
+#define GEN8_3DSTATE_CHROMA_KEY_length 0x00000004
+
+struct GEN8_3DSTATE_CHROMA_KEY {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ChromaKeyTableIndex;
+ uint32_t ChromaKeyLowValue;
+ uint32_t ChromaKeyHighValue;
+};
+
+static inline void
+GEN8_3DSTATE_CHROMA_KEY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_CHROMA_KEY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ChromaKeyTableIndex, 30, 31) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ChromaKeyLowValue, 0, 31) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->ChromaKeyHighValue, 0, 31) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_CLEAR_PARAMS_length_bias 0x00000002
+#define GEN8_3DSTATE_CLEAR_PARAMS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 4, \
+ .DwordLength = 1
+
+#define GEN8_3DSTATE_CLEAR_PARAMS_length 0x00000003
+
+struct GEN8_3DSTATE_CLEAR_PARAMS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ float DepthClearValue;
+ bool DepthClearValueValid;
+};
+
+static inline void
+GEN8_3DSTATE_CLEAR_PARAMS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_CLEAR_PARAMS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_float(values->DepthClearValue) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->DepthClearValueValid, 0, 0) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_CLIP_length_bias 0x00000002
+#define GEN8_3DSTATE_CLIP_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 18, \
+ .DwordLength = 2
+
+#define GEN8_3DSTATE_CLIP_length 0x00000004
+
+struct GEN8_3DSTATE_CLIP {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define Normal 0
+#define Force 1
+ bool ForceUserClipDistanceCullTestEnableBitmask;
+#define _8Bit 0
+#define _4Bit 1
+ uint32_t VertexSubPixelPrecisionSelect;
+ bool EarlyCullEnable;
+#define Normal 0
+#define Force 1
+ bool ForceUserClipDistanceClipTestEnableBitmask;
+#define Normal 0
+#define Force 1
+ bool ForceClipMode;
+ bool ClipperStatisticsEnable;
+ uint32_t UserClipDistanceCullTestEnableBitmask;
+ bool ClipEnable;
+#define API_OGL 0
+ uint32_t APIMode;
+ bool ViewportXYClipTestEnable;
+ bool GuardbandClipTestEnable;
+ uint32_t UserClipDistanceClipTestEnableBitmask;
+#define NORMAL 0
+#define REJECT_ALL 3
+#define ACCEPT_ALL 4
+ uint32_t ClipMode;
+ bool PerspectiveDivideDisable;
+ bool NonPerspectiveBarycentricEnable;
+ uint32_t TriangleStripListProvokingVertexSelect;
+ uint32_t LineStripListProvokingVertexSelect;
+ uint32_t TriangleFanProvokingVertexSelect;
+ float MinimumPointWidth;
+ float MaximumPointWidth;
+ bool ForceZeroRTAIndexEnable;
+ uint32_t MaximumVPIndex;
+};
+
+static inline void
+GEN8_3DSTATE_CLIP_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_CLIP * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ForceUserClipDistanceCullTestEnableBitmask, 20, 20) |
+ __gen_field(values->VertexSubPixelPrecisionSelect, 19, 19) |
+ __gen_field(values->EarlyCullEnable, 18, 18) |
+ __gen_field(values->ForceUserClipDistanceClipTestEnableBitmask, 17, 17) |
+ __gen_field(values->ForceClipMode, 16, 16) |
+ __gen_field(values->ClipperStatisticsEnable, 10, 10) |
+ __gen_field(values->UserClipDistanceCullTestEnableBitmask, 0, 7) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ClipEnable, 31, 31) |
+ __gen_field(values->APIMode, 30, 30) |
+ __gen_field(values->ViewportXYClipTestEnable, 28, 28) |
+ __gen_field(values->GuardbandClipTestEnable, 26, 26) |
+ __gen_field(values->UserClipDistanceClipTestEnableBitmask, 16, 23) |
+ __gen_field(values->ClipMode, 13, 15) |
+ __gen_field(values->PerspectiveDivideDisable, 9, 9) |
+ __gen_field(values->NonPerspectiveBarycentricEnable, 8, 8) |
+ __gen_field(values->TriangleStripListProvokingVertexSelect, 4, 5) |
+ __gen_field(values->LineStripListProvokingVertexSelect, 2, 3) |
+ __gen_field(values->TriangleFanProvokingVertexSelect, 0, 1) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->MinimumPointWidth * (1 << 3), 17, 27) |
+ __gen_field(values->MaximumPointWidth * (1 << 3), 6, 16) |
+ __gen_field(values->ForceZeroRTAIndexEnable, 5, 5) |
+ __gen_field(values->MaximumVPIndex, 0, 3) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_CONSTANT_DS_length_bias 0x00000002
+#define GEN8_3DSTATE_CONSTANT_DS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 26, \
+ .DwordLength = 9
+
+#define GEN8_3DSTATE_CONSTANT_DS_length 0x0000000b
+
+#define GEN8_3DSTATE_CONSTANT_BODY_length 0x0000000a
+
+struct GEN8_3DSTATE_CONSTANT_BODY {
+ uint32_t ConstantBuffer1ReadLength;
+ uint32_t ConstantBuffer0ReadLength;
+ uint32_t ConstantBuffer3ReadLength;
+ uint32_t ConstantBuffer2ReadLength;
+ __gen_address_type PointerToConstantBuffer0;
+ __gen_address_type PointerToConstantBuffer1;
+ __gen_address_type PointerToConstantBuffer2;
+ __gen_address_type PointerToConstantBuffer3;
+};
+
+static inline void
+GEN8_3DSTATE_CONSTANT_BODY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_CONSTANT_BODY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->ConstantBuffer1ReadLength, 16, 31) |
+ __gen_field(values->ConstantBuffer0ReadLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBuffer3ReadLength, 16, 31) |
+ __gen_field(values->ConstantBuffer2ReadLength, 0, 15) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ uint64_t qw2 =
+ __gen_combine_address(data, &dw[2], values->PointerToConstantBuffer0, dw2);
+
+ dw[2] = qw2;
+ dw[3] = qw2 >> 32;
+
+ uint32_t dw4 =
+ 0;
+
+ uint64_t qw4 =
+ __gen_combine_address(data, &dw[4], values->PointerToConstantBuffer1, dw4);
+
+ dw[4] = qw4;
+ dw[5] = qw4 >> 32;
+
+ uint32_t dw6 =
+ 0;
+
+ uint64_t qw6 =
+ __gen_combine_address(data, &dw[6], values->PointerToConstantBuffer2, dw6);
+
+ dw[6] = qw6;
+ dw[7] = qw6 >> 32;
+
+ uint32_t dw8 =
+ 0;
+
+ uint64_t qw8 =
+ __gen_combine_address(data, &dw[8], values->PointerToConstantBuffer3, dw8);
+
+ dw[8] = qw8;
+ dw[9] = qw8 >> 32;
+
+}
+
+struct GEN8_3DSTATE_CONSTANT_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ struct GEN8_MEMORY_OBJECT_CONTROL_STATE ConstantBufferObjectControlState;
+ uint32_t DwordLength;
+ struct GEN8_3DSTATE_CONSTANT_BODY ConstantBody;
+};
+
+static inline void
+GEN8_3DSTATE_CONSTANT_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_CONSTANT_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ uint32_t dw_ConstantBufferObjectControlState;
+ GEN8_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_ConstantBufferObjectControlState, &values->ConstantBufferObjectControlState);
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(dw_ConstantBufferObjectControlState, 8, 14) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ GEN8_3DSTATE_CONSTANT_BODY_pack(data, &dw[1], &values->ConstantBody);
+}
+
+#define GEN8_3DSTATE_CONSTANT_GS_length_bias 0x00000002
+#define GEN8_3DSTATE_CONSTANT_GS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 22, \
+ .DwordLength = 9
+
+#define GEN8_3DSTATE_CONSTANT_GS_length 0x0000000b
+
+struct GEN8_3DSTATE_CONSTANT_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ struct GEN8_MEMORY_OBJECT_CONTROL_STATE ConstantBufferObjectControlState;
+ uint32_t DwordLength;
+ struct GEN8_3DSTATE_CONSTANT_BODY ConstantBody;
+};
+
+static inline void
+GEN8_3DSTATE_CONSTANT_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_CONSTANT_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ uint32_t dw_ConstantBufferObjectControlState;
+ GEN8_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_ConstantBufferObjectControlState, &values->ConstantBufferObjectControlState);
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(dw_ConstantBufferObjectControlState, 8, 14) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ GEN8_3DSTATE_CONSTANT_BODY_pack(data, &dw[1], &values->ConstantBody);
+}
+
+#define GEN8_3DSTATE_CONSTANT_HS_length_bias 0x00000002
+#define GEN8_3DSTATE_CONSTANT_HS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 25, \
+ .DwordLength = 9
+
+#define GEN8_3DSTATE_CONSTANT_HS_length 0x0000000b
+
+struct GEN8_3DSTATE_CONSTANT_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ struct GEN8_MEMORY_OBJECT_CONTROL_STATE ConstantBufferObjectControlState;
+ uint32_t DwordLength;
+ struct GEN8_3DSTATE_CONSTANT_BODY ConstantBody;
+};
+
+static inline void
+GEN8_3DSTATE_CONSTANT_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_CONSTANT_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ uint32_t dw_ConstantBufferObjectControlState;
+ GEN8_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_ConstantBufferObjectControlState, &values->ConstantBufferObjectControlState);
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(dw_ConstantBufferObjectControlState, 8, 14) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ GEN8_3DSTATE_CONSTANT_BODY_pack(data, &dw[1], &values->ConstantBody);
+}
+
+#define GEN8_3DSTATE_CONSTANT_PS_length_bias 0x00000002
+#define GEN8_3DSTATE_CONSTANT_PS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 23, \
+ .DwordLength = 9
+
+#define GEN8_3DSTATE_CONSTANT_PS_length 0x0000000b
+
+struct GEN8_3DSTATE_CONSTANT_PS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ struct GEN8_MEMORY_OBJECT_CONTROL_STATE ConstantBufferObjectControlState;
+ uint32_t DwordLength;
+ struct GEN8_3DSTATE_CONSTANT_BODY ConstantBody;
+};
+
+static inline void
+GEN8_3DSTATE_CONSTANT_PS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_CONSTANT_PS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ uint32_t dw_ConstantBufferObjectControlState;
+ GEN8_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_ConstantBufferObjectControlState, &values->ConstantBufferObjectControlState);
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(dw_ConstantBufferObjectControlState, 8, 14) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ GEN8_3DSTATE_CONSTANT_BODY_pack(data, &dw[1], &values->ConstantBody);
+}
+
+#define GEN8_3DSTATE_CONSTANT_VS_length_bias 0x00000002
+#define GEN8_3DSTATE_CONSTANT_VS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 21, \
+ .DwordLength = 9
+
+#define GEN8_3DSTATE_CONSTANT_VS_length 0x0000000b
+
+struct GEN8_3DSTATE_CONSTANT_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ struct GEN8_MEMORY_OBJECT_CONTROL_STATE ConstantBufferObjectControlState;
+ uint32_t DwordLength;
+ struct GEN8_3DSTATE_CONSTANT_BODY ConstantBody;
+};
+
+static inline void
+GEN8_3DSTATE_CONSTANT_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_CONSTANT_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ uint32_t dw_ConstantBufferObjectControlState;
+ GEN8_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_ConstantBufferObjectControlState, &values->ConstantBufferObjectControlState);
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(dw_ConstantBufferObjectControlState, 8, 14) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ GEN8_3DSTATE_CONSTANT_BODY_pack(data, &dw[1], &values->ConstantBody);
+}
+
+#define GEN8_3DSTATE_DEPTH_BUFFER_length_bias 0x00000002
+#define GEN8_3DSTATE_DEPTH_BUFFER_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 5, \
+ .DwordLength = 6
+
+#define GEN8_3DSTATE_DEPTH_BUFFER_length 0x00000008
+
+struct GEN8_3DSTATE_DEPTH_BUFFER {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define SURFTYPE_1D 0
+#define SURFTYPE_2D 1
+#define SURFTYPE_3D 2
+#define SURFTYPE_CUBE 3
+#define SURFTYPE_NULL 7
+ uint32_t SurfaceType;
+ bool DepthWriteEnable;
+ bool StencilWriteEnable;
+ bool HierarchicalDepthBufferEnable;
+#define D32_FLOAT 1
+#define D24_UNORM_X8_UINT 3
+#define D16_UNORM 5
+ uint32_t SurfaceFormat;
+ uint32_t SurfacePitch;
+ __gen_address_type SurfaceBaseAddress;
+ uint32_t Height;
+ uint32_t Width;
+ uint32_t LOD;
+ uint32_t Depth;
+ uint32_t MinimumArrayElement;
+ struct GEN8_MEMORY_OBJECT_CONTROL_STATE DepthBufferObjectControlState;
+ uint32_t RenderTargetViewExtent;
+ uint32_t SurfaceQPitch;
+};
+
+static inline void
+GEN8_3DSTATE_DEPTH_BUFFER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_DEPTH_BUFFER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->SurfaceType, 29, 31) |
+ __gen_field(values->DepthWriteEnable, 28, 28) |
+ __gen_field(values->StencilWriteEnable, 27, 27) |
+ __gen_field(values->HierarchicalDepthBufferEnable, 22, 22) |
+ __gen_field(values->SurfaceFormat, 18, 20) |
+ __gen_field(values->SurfacePitch, 0, 17) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ uint64_t qw2 =
+ __gen_combine_address(data, &dw[2], values->SurfaceBaseAddress, dw2);
+
+ dw[2] = qw2;
+ dw[3] = qw2 >> 32;
+
+ dw[4] =
+ __gen_field(values->Height, 18, 31) |
+ __gen_field(values->Width, 4, 17) |
+ __gen_field(values->LOD, 0, 3) |
+ 0;
+
+ uint32_t dw_DepthBufferObjectControlState;
+ GEN8_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_DepthBufferObjectControlState, &values->DepthBufferObjectControlState);
+ dw[5] =
+ __gen_field(values->Depth, 21, 31) |
+ __gen_field(values->MinimumArrayElement, 10, 20) |
+ __gen_field(dw_DepthBufferObjectControlState, 0, 6) |
+ 0;
+
+ dw[6] =
+ 0;
+
+ dw[7] =
+ __gen_field(values->RenderTargetViewExtent, 21, 31) |
+ __gen_field(values->SurfaceQPitch, 0, 14) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_DRAWING_RECTANGLE_length_bias 0x00000002
+#define GEN8_3DSTATE_DRAWING_RECTANGLE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 0, \
+ .DwordLength = 2
+
+#define GEN8_3DSTATE_DRAWING_RECTANGLE_length 0x00000004
+
+struct GEN8_3DSTATE_DRAWING_RECTANGLE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+#define Legacy 0
+#define Core0Enabled 1
+#define Core1Enabled 2
+ uint32_t CoreModeSelect;
+ uint32_t DwordLength;
+ uint32_t ClippedDrawingRectangleYMin;
+ uint32_t ClippedDrawingRectangleXMin;
+ uint32_t ClippedDrawingRectangleYMax;
+ uint32_t ClippedDrawingRectangleXMax;
+ uint32_t DrawingRectangleOriginY;
+ uint32_t DrawingRectangleOriginX;
+};
+
+static inline void
+GEN8_3DSTATE_DRAWING_RECTANGLE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_DRAWING_RECTANGLE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->CoreModeSelect, 14, 15) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ClippedDrawingRectangleYMin, 16, 31) |
+ __gen_field(values->ClippedDrawingRectangleXMin, 0, 15) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ClippedDrawingRectangleYMax, 16, 31) |
+ __gen_field(values->ClippedDrawingRectangleXMax, 0, 15) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->DrawingRectangleOriginY, 16, 31) |
+ __gen_field(values->DrawingRectangleOriginX, 0, 15) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_DS_length_bias 0x00000002
+#define GEN8_3DSTATE_DS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 29, \
+ .DwordLength = 7
+
+#define GEN8_3DSTATE_DS_length 0x00000009
+
+struct GEN8_3DSTATE_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint64_t KernelStartPointer;
+#define Multiple 0
+#define Single 1
+ uint32_t SingleDomainPointDispatch;
+#define Dmask 0
+#define Vmask 1
+ uint32_t VectorMaskEnable;
+#define NoSamplers 0
+#define _14Samplers 1
+#define _58Samplers 2
+#define _912Samplers 3
+#define _1316Samplers 4
+ uint32_t SamplerCount;
+ uint32_t BindingTableEntryCount;
+#define Normal 0
+#define High 1
+ uint32_t ThreadDispatchPriority;
+#define IEEE754 0
+#define Alternate 1
+ uint32_t FloatingPointMode;
+ bool AccessesUAV;
+ bool IllegalOpcodeExceptionEnable;
+ bool SoftwareExceptionEnable;
+ uint64_t ScratchSpaceBasePointer;
+ uint32_t PerThreadScratchSpace;
+ uint32_t DispatchGRFStartRegisterForURBData;
+ uint32_t PatchURBEntryReadLength;
+ uint32_t PatchURBEntryReadOffset;
+ uint32_t MaximumNumberofThreads;
+ bool StatisticsEnable;
+ bool SIMD8DispatchEnable;
+ bool ComputeWCoordinateEnable;
+ bool CacheDisable;
+ bool FunctionEnable;
+ uint32_t VertexURBEntryOutputReadOffset;
+ uint32_t VertexURBEntryOutputLength;
+ uint32_t UserClipDistanceClipTestEnableBitmask;
+ uint32_t UserClipDistanceCullTestEnableBitmask;
+};
+
+static inline void
+GEN8_3DSTATE_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint64_t qw1 =
+ __gen_offset(values->KernelStartPointer, 6, 63) |
+ 0;
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+ dw[3] =
+ __gen_field(values->SingleDomainPointDispatch, 31, 31) |
+ __gen_field(values->VectorMaskEnable, 30, 30) |
+ __gen_field(values->SamplerCount, 27, 29) |
+ __gen_field(values->BindingTableEntryCount, 18, 25) |
+ __gen_field(values->ThreadDispatchPriority, 17, 17) |
+ __gen_field(values->FloatingPointMode, 16, 16) |
+ __gen_field(values->AccessesUAV, 14, 14) |
+ __gen_field(values->IllegalOpcodeExceptionEnable, 13, 13) |
+ __gen_field(values->SoftwareExceptionEnable, 7, 7) |
+ 0;
+
+ uint64_t qw4 =
+ __gen_offset(values->ScratchSpaceBasePointer, 10, 63) |
+ __gen_field(values->PerThreadScratchSpace, 0, 3) |
+ 0;
+
+ dw[4] = qw4;
+ dw[5] = qw4 >> 32;
+
+ dw[6] =
+ __gen_field(values->DispatchGRFStartRegisterForURBData, 20, 24) |
+ __gen_field(values->PatchURBEntryReadLength, 11, 17) |
+ __gen_field(values->PatchURBEntryReadOffset, 4, 9) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->MaximumNumberofThreads, 21, 29) |
+ __gen_field(values->StatisticsEnable, 10, 10) |
+ __gen_field(values->SIMD8DispatchEnable, 3, 3) |
+ __gen_field(values->ComputeWCoordinateEnable, 2, 2) |
+ __gen_field(values->CacheDisable, 1, 1) |
+ __gen_field(values->FunctionEnable, 0, 0) |
+ 0;
+
+ dw[8] =
+ __gen_field(values->VertexURBEntryOutputReadOffset, 21, 26) |
+ __gen_field(values->VertexURBEntryOutputLength, 16, 20) |
+ __gen_field(values->UserClipDistanceClipTestEnableBitmask, 8, 15) |
+ __gen_field(values->UserClipDistanceCullTestEnableBitmask, 0, 7) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_GATHER_CONSTANT_DS_length_bias 0x00000002
+#define GEN8_3DSTATE_GATHER_CONSTANT_DS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 55
+
+#define GEN8_3DSTATE_GATHER_CONSTANT_DS_length 0x00000000
+
+#define GEN8_GATHER_CONSTANT_ENTRY_length 0x00000001
+
+struct GEN8_GATHER_CONSTANT_ENTRY {
+ uint32_t ConstantBufferOffset;
+ uint32_t ChannelMask;
+ uint32_t BindingTableIndexOffset;
+};
+
+static inline void
+GEN8_GATHER_CONSTANT_ENTRY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_GATHER_CONSTANT_ENTRY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_offset(values->ConstantBufferOffset, 8, 15) |
+ __gen_field(values->ChannelMask, 4, 7) |
+ __gen_field(values->BindingTableIndexOffset, 0, 3) |
+ 0;
+
+}
+
+struct GEN8_3DSTATE_GATHER_CONSTANT_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferValid;
+ uint32_t ConstantBufferBindingTableBlock;
+ uint32_t GatherBufferOffset;
+ bool ConstantBufferDx9GenerateStall;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN8_3DSTATE_GATHER_CONSTANT_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_GATHER_CONSTANT_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferValid, 16, 31) |
+ __gen_field(values->ConstantBufferBindingTableBlock, 12, 15) |
+ 0;
+
+ dw[2] =
+ __gen_offset(values->GatherBufferOffset, 6, 22) |
+ __gen_field(values->ConstantBufferDx9GenerateStall, 5, 5) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN8_3DSTATE_GATHER_CONSTANT_GS_length_bias 0x00000002
+#define GEN8_3DSTATE_GATHER_CONSTANT_GS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 53
+
+#define GEN8_3DSTATE_GATHER_CONSTANT_GS_length 0x00000000
+
+struct GEN8_3DSTATE_GATHER_CONSTANT_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferValid;
+ uint32_t ConstantBufferBindingTableBlock;
+ uint32_t GatherBufferOffset;
+ bool ConstantBufferDx9GenerateStall;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN8_3DSTATE_GATHER_CONSTANT_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_GATHER_CONSTANT_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferValid, 16, 31) |
+ __gen_field(values->ConstantBufferBindingTableBlock, 12, 15) |
+ 0;
+
+ dw[2] =
+ __gen_offset(values->GatherBufferOffset, 6, 22) |
+ __gen_field(values->ConstantBufferDx9GenerateStall, 5, 5) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN8_3DSTATE_GATHER_CONSTANT_HS_length_bias 0x00000002
+#define GEN8_3DSTATE_GATHER_CONSTANT_HS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 54
+
+#define GEN8_3DSTATE_GATHER_CONSTANT_HS_length 0x00000000
+
+struct GEN8_3DSTATE_GATHER_CONSTANT_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferValid;
+ uint32_t ConstantBufferBindingTableBlock;
+ uint32_t GatherBufferOffset;
+ bool ConstantBufferDx9GenerateStall;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN8_3DSTATE_GATHER_CONSTANT_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_GATHER_CONSTANT_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferValid, 16, 31) |
+ __gen_field(values->ConstantBufferBindingTableBlock, 12, 15) |
+ 0;
+
+ dw[2] =
+ __gen_offset(values->GatherBufferOffset, 6, 22) |
+ __gen_field(values->ConstantBufferDx9GenerateStall, 5, 5) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN8_3DSTATE_GATHER_CONSTANT_PS_length_bias 0x00000002
+#define GEN8_3DSTATE_GATHER_CONSTANT_PS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 56
+
+#define GEN8_3DSTATE_GATHER_CONSTANT_PS_length 0x00000000
+
+struct GEN8_3DSTATE_GATHER_CONSTANT_PS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferValid;
+ uint32_t ConstantBufferBindingTableBlock;
+ uint32_t GatherBufferOffset;
+ bool ConstantBufferDx9GenerateStall;
+ bool ConstantBufferDx9Enable;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN8_3DSTATE_GATHER_CONSTANT_PS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_GATHER_CONSTANT_PS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferValid, 16, 31) |
+ __gen_field(values->ConstantBufferBindingTableBlock, 12, 15) |
+ 0;
+
+ dw[2] =
+ __gen_offset(values->GatherBufferOffset, 6, 22) |
+ __gen_field(values->ConstantBufferDx9GenerateStall, 5, 5) |
+ __gen_field(values->ConstantBufferDx9Enable, 4, 4) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN8_3DSTATE_GATHER_CONSTANT_VS_length_bias 0x00000002
+#define GEN8_3DSTATE_GATHER_CONSTANT_VS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 52
+
+#define GEN8_3DSTATE_GATHER_CONSTANT_VS_length 0x00000000
+
+struct GEN8_3DSTATE_GATHER_CONSTANT_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferValid;
+ uint32_t ConstantBufferBindingTableBlock;
+ uint32_t GatherBufferOffset;
+ bool ConstantBufferDx9GenerateStall;
+ bool ConstantBufferDx9Enable;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN8_3DSTATE_GATHER_CONSTANT_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_GATHER_CONSTANT_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferValid, 16, 31) |
+ __gen_field(values->ConstantBufferBindingTableBlock, 12, 15) |
+ 0;
+
+ dw[2] =
+ __gen_offset(values->GatherBufferOffset, 6, 22) |
+ __gen_field(values->ConstantBufferDx9GenerateStall, 5, 5) |
+ __gen_field(values->ConstantBufferDx9Enable, 4, 4) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN8_3DSTATE_GATHER_POOL_ALLOC_length_bias 0x00000002
+#define GEN8_3DSTATE_GATHER_POOL_ALLOC_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 26, \
+ .DwordLength = 2
+
+#define GEN8_3DSTATE_GATHER_POOL_ALLOC_length 0x00000004
+
+struct GEN8_3DSTATE_GATHER_POOL_ALLOC {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ __gen_address_type GatherPoolBaseAddress;
+ bool GatherPoolEnable;
+ struct GEN8_MEMORY_OBJECT_CONTROL_STATE MemoryObjectControlState;
+ uint32_t GatherPoolBufferSize;
+};
+
+static inline void
+GEN8_3DSTATE_GATHER_POOL_ALLOC_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_GATHER_POOL_ALLOC * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw_MemoryObjectControlState;
+ GEN8_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_MemoryObjectControlState, &values->MemoryObjectControlState);
+ uint32_t dw1 =
+ __gen_field(values->GatherPoolEnable, 11, 11) |
+ __gen_field(dw_MemoryObjectControlState, 0, 6) |
+ 0;
+
+ uint64_t qw1 =
+ __gen_combine_address(data, &dw[1], values->GatherPoolBaseAddress, dw1);
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+ dw[3] =
+ __gen_field(values->GatherPoolBufferSize, 12, 31) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_GS_length_bias 0x00000002
+#define GEN8_3DSTATE_GS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 17, \
+ .DwordLength = 8
+
+#define GEN8_3DSTATE_GS_length 0x0000000a
+
+struct GEN8_3DSTATE_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint64_t KernelStartPointer;
+ uint32_t SingleProgramFlow;
+#define Dmask 0
+#define Vmask 1
+ uint32_t VectorMaskEnable;
+#define NoSamplers 0
+#define _14Samplers 1
+#define _58Samplers 2
+#define _912Samplers 3
+#define _1316Samplers 4
+ uint32_t SamplerCount;
+ uint32_t BindingTableEntryCount;
+#define Normal 0
+#define High 1
+ uint32_t ThreadDispatchPriority;
+#define IEEE754 0
+#define Alternate 1
+ uint32_t FloatingPointMode;
+ bool IllegalOpcodeExceptionEnable;
+ bool AccessesUAV;
+ bool MaskStackExceptionEnable;
+ bool SoftwareExceptionEnable;
+ uint32_t ExpectedVertexCount;
+ uint64_t ScratchSpaceBasePointer;
+ uint32_t PerThreadScratchSpace;
+ uint32_t OutputVertexSize;
+ uint32_t OutputTopology;
+ uint32_t VertexURBEntryReadLength;
+ bool IncludeVertexHandles;
+ uint32_t VertexURBEntryReadOffset;
+ uint32_t DispatchGRFStartRegisterForURBData;
+ uint32_t MaximumNumberofThreads;
+ uint32_t ControlDataHeaderSize;
+ uint32_t InstanceControl;
+ uint32_t DefaultStreamId;
+#define DispatchModeSingle 0
+#define DispatchModeDualInstance 1
+#define DispatchModeDualObject 2
+#define DispatchModeSIMD8 3
+ uint32_t DispatchMode;
+ bool StatisticsEnable;
+ uint32_t InvocationsIncrementValue;
+ bool IncludePrimitiveID;
+ uint32_t Hint;
+#define LEADING 0
+#define TRAILING 1
+ uint32_t ReorderMode;
+ bool DiscardAdjacency;
+ bool Enable;
+#define CUT 0
+#define SID 1
+ uint32_t ControlDataFormat;
+ bool StaticOutput;
+ uint32_t StaticOutputVertexCount;
+ uint32_t VertexURBEntryOutputReadOffset;
+ uint32_t VertexURBEntryOutputLength;
+ uint32_t UserClipDistanceClipTestEnableBitmask;
+ uint32_t UserClipDistanceCullTestEnableBitmask;
+};
+
+static inline void
+GEN8_3DSTATE_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint64_t qw1 =
+ __gen_offset(values->KernelStartPointer, 6, 63) |
+ 0;
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+ dw[3] =
+ __gen_field(values->SingleProgramFlow, 31, 31) |
+ __gen_field(values->VectorMaskEnable, 30, 30) |
+ __gen_field(values->SamplerCount, 27, 29) |
+ __gen_field(values->BindingTableEntryCount, 18, 25) |
+ __gen_field(values->ThreadDispatchPriority, 17, 17) |
+ __gen_field(values->FloatingPointMode, 16, 16) |
+ __gen_field(values->IllegalOpcodeExceptionEnable, 13, 13) |
+ __gen_field(values->AccessesUAV, 12, 12) |
+ __gen_field(values->MaskStackExceptionEnable, 11, 11) |
+ __gen_field(values->SoftwareExceptionEnable, 7, 7) |
+ __gen_field(values->ExpectedVertexCount, 0, 5) |
+ 0;
+
+ uint64_t qw4 =
+ __gen_offset(values->ScratchSpaceBasePointer, 10, 63) |
+ __gen_field(values->PerThreadScratchSpace, 0, 3) |
+ 0;
+
+ dw[4] = qw4;
+ dw[5] = qw4 >> 32;
+
+ dw[6] =
+ __gen_field(values->OutputVertexSize, 23, 28) |
+ __gen_field(values->OutputTopology, 17, 22) |
+ __gen_field(values->VertexURBEntryReadLength, 11, 16) |
+ __gen_field(values->IncludeVertexHandles, 10, 10) |
+ __gen_field(values->VertexURBEntryReadOffset, 4, 9) |
+ __gen_field(values->DispatchGRFStartRegisterForURBData, 0, 3) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->MaximumNumberofThreads, 24, 31) |
+ __gen_field(values->ControlDataHeaderSize, 20, 23) |
+ __gen_field(values->InstanceControl, 15, 19) |
+ __gen_field(values->DefaultStreamId, 13, 14) |
+ __gen_field(values->DispatchMode, 11, 12) |
+ __gen_field(values->StatisticsEnable, 10, 10) |
+ __gen_field(values->InvocationsIncrementValue, 5, 9) |
+ __gen_field(values->IncludePrimitiveID, 4, 4) |
+ __gen_field(values->Hint, 3, 3) |
+ __gen_field(values->ReorderMode, 2, 2) |
+ __gen_field(values->DiscardAdjacency, 1, 1) |
+ __gen_field(values->Enable, 0, 0) |
+ 0;
+
+ dw[8] =
+ __gen_field(values->ControlDataFormat, 31, 31) |
+ __gen_field(values->StaticOutput, 30, 30) |
+ __gen_field(values->StaticOutputVertexCount, 16, 26) |
+ 0;
+
+ dw[9] =
+ __gen_field(values->VertexURBEntryOutputReadOffset, 21, 26) |
+ __gen_field(values->VertexURBEntryOutputLength, 16, 20) |
+ __gen_field(values->UserClipDistanceClipTestEnableBitmask, 8, 15) |
+ __gen_field(values->UserClipDistanceCullTestEnableBitmask, 0, 7) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_HIER_DEPTH_BUFFER_length_bias 0x00000002
+#define GEN8_3DSTATE_HIER_DEPTH_BUFFER_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 7, \
+ .DwordLength = 3
+
+#define GEN8_3DSTATE_HIER_DEPTH_BUFFER_length 0x00000005
+
+struct GEN8_3DSTATE_HIER_DEPTH_BUFFER {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ struct GEN8_MEMORY_OBJECT_CONTROL_STATE HierarchicalDepthBufferObjectControlState;
+ uint32_t SurfacePitch;
+ __gen_address_type SurfaceBaseAddress;
+ uint32_t SurfaceQPitch;
+};
+
+static inline void
+GEN8_3DSTATE_HIER_DEPTH_BUFFER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_HIER_DEPTH_BUFFER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw_HierarchicalDepthBufferObjectControlState;
+ GEN8_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_HierarchicalDepthBufferObjectControlState, &values->HierarchicalDepthBufferObjectControlState);
+ dw[1] =
+ __gen_field(dw_HierarchicalDepthBufferObjectControlState, 25, 31) |
+ __gen_field(values->SurfacePitch, 0, 16) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ uint64_t qw2 =
+ __gen_combine_address(data, &dw[2], values->SurfaceBaseAddress, dw2);
+
+ dw[2] = qw2;
+ dw[3] = qw2 >> 32;
+
+ dw[4] =
+ __gen_field(values->SurfaceQPitch, 0, 14) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_HS_length_bias 0x00000002
+#define GEN8_3DSTATE_HS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 27, \
+ .DwordLength = 7
+
+#define GEN8_3DSTATE_HS_length 0x00000009
+
+struct GEN8_3DSTATE_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define NoSamplers 0
+#define _14Samplers 1
+#define _58Samplers 2
+#define _912Samplers 3
+#define _1316Samplers 4
+ uint32_t SamplerCount;
+ uint32_t BindingTableEntryCount;
+#define Normal 0
+#define High 1
+ uint32_t ThreadDispatchPriority;
+#define IEEE754 0
+#define alternate 1
+ uint32_t FloatingPointMode;
+ bool IllegalOpcodeExceptionEnable;
+ bool SoftwareExceptionEnable;
+ bool Enable;
+ bool StatisticsEnable;
+ uint32_t MaximumNumberofThreads;
+ uint32_t InstanceCount;
+ uint64_t KernelStartPointer;
+ uint64_t ScratchSpaceBasePointer;
+ uint32_t PerThreadScratchSpace;
+ bool SingleProgramFlow;
+#define Dmask 0
+#define Vmask 1
+ uint32_t VectorMaskEnable;
+ bool AccessesUAV;
+ bool IncludeVertexHandles;
+ uint32_t DispatchGRFStartRegisterForURBData;
+ uint32_t VertexURBEntryReadLength;
+ uint32_t VertexURBEntryReadOffset;
+};
+
+static inline void
+GEN8_3DSTATE_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->SamplerCount, 27, 29) |
+ __gen_field(values->BindingTableEntryCount, 18, 25) |
+ __gen_field(values->ThreadDispatchPriority, 17, 17) |
+ __gen_field(values->FloatingPointMode, 16, 16) |
+ __gen_field(values->IllegalOpcodeExceptionEnable, 13, 13) |
+ __gen_field(values->SoftwareExceptionEnable, 12, 12) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->Enable, 31, 31) |
+ __gen_field(values->StatisticsEnable, 29, 29) |
+ __gen_field(values->MaximumNumberofThreads, 8, 16) |
+ __gen_field(values->InstanceCount, 0, 3) |
+ 0;
+
+ uint64_t qw3 =
+ __gen_offset(values->KernelStartPointer, 6, 63) |
+ 0;
+
+ dw[3] = qw3;
+ dw[4] = qw3 >> 32;
+
+ uint64_t qw5 =
+ __gen_offset(values->ScratchSpaceBasePointer, 10, 63) |
+ __gen_field(values->PerThreadScratchSpace, 0, 3) |
+ 0;
+
+ dw[5] = qw5;
+ dw[6] = qw5 >> 32;
+
+ dw[7] =
+ __gen_field(values->SingleProgramFlow, 27, 27) |
+ __gen_field(values->VectorMaskEnable, 26, 26) |
+ __gen_field(values->AccessesUAV, 25, 25) |
+ __gen_field(values->IncludeVertexHandles, 24, 24) |
+ __gen_field(values->DispatchGRFStartRegisterForURBData, 19, 23) |
+ __gen_field(values->VertexURBEntryReadLength, 11, 16) |
+ __gen_field(values->VertexURBEntryReadOffset, 4, 9) |
+ 0;
+
+ dw[8] =
+ 0;
+
+}
+
+#define GEN8_3DSTATE_INDEX_BUFFER_length_bias 0x00000002
+#define GEN8_3DSTATE_INDEX_BUFFER_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 10, \
+ .DwordLength = 3
+
+#define GEN8_3DSTATE_INDEX_BUFFER_length 0x00000005
+
+struct GEN8_3DSTATE_INDEX_BUFFER {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define INDEX_BYTE 0
+#define INDEX_WORD 1
+#define INDEX_DWORD 2
+ uint32_t IndexFormat;
+ struct GEN8_MEMORY_OBJECT_CONTROL_STATE MemoryObjectControlState;
+ __gen_address_type BufferStartingAddress;
+ uint32_t BufferSize;
+};
+
+static inline void
+GEN8_3DSTATE_INDEX_BUFFER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_INDEX_BUFFER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw_MemoryObjectControlState;
+ GEN8_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_MemoryObjectControlState, &values->MemoryObjectControlState);
+ dw[1] =
+ __gen_field(values->IndexFormat, 8, 9) |
+ __gen_field(dw_MemoryObjectControlState, 0, 6) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ uint64_t qw2 =
+ __gen_combine_address(data, &dw[2], values->BufferStartingAddress, dw2);
+
+ dw[2] = qw2;
+ dw[3] = qw2 >> 32;
+
+ dw[4] =
+ __gen_field(values->BufferSize, 0, 31) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_LINE_STIPPLE_length_bias 0x00000002
+#define GEN8_3DSTATE_LINE_STIPPLE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 8, \
+ .DwordLength = 1
+
+#define GEN8_3DSTATE_LINE_STIPPLE_length 0x00000003
+
+struct GEN8_3DSTATE_LINE_STIPPLE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool ModifyEnableCurrentRepeatCounterCurrentStippleIndex;
+ uint32_t CurrentRepeatCounter;
+ uint32_t CurrentStippleIndex;
+ uint32_t LineStipplePattern;
+ float LineStippleInverseRepeatCount;
+ uint32_t LineStippleRepeatCount;
+};
+
+static inline void
+GEN8_3DSTATE_LINE_STIPPLE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_LINE_STIPPLE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ModifyEnableCurrentRepeatCounterCurrentStippleIndex, 31, 31) |
+ __gen_field(values->CurrentRepeatCounter, 21, 29) |
+ __gen_field(values->CurrentStippleIndex, 16, 19) |
+ __gen_field(values->LineStipplePattern, 0, 15) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->LineStippleInverseRepeatCount * (1 << 16), 15, 31) |
+ __gen_field(values->LineStippleRepeatCount, 0, 8) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_MONOFILTER_SIZE_length_bias 0x00000002
+#define GEN8_3DSTATE_MONOFILTER_SIZE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 17, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_MONOFILTER_SIZE_length 0x00000002
+
+struct GEN8_3DSTATE_MONOFILTER_SIZE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t MonochromeFilterWidth;
+ uint32_t MonochromeFilterHeight;
+};
+
+static inline void
+GEN8_3DSTATE_MONOFILTER_SIZE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_MONOFILTER_SIZE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->MonochromeFilterWidth, 3, 5) |
+ __gen_field(values->MonochromeFilterHeight, 0, 2) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_MULTISAMPLE_length_bias 0x00000002
+#define GEN8_3DSTATE_MULTISAMPLE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 13, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_MULTISAMPLE_length 0x00000002
+
+struct GEN8_3DSTATE_MULTISAMPLE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PixelPositionOffsetEnable;
+#define CENTER 0
+#define UL_CORNER 1
+ uint32_t PixelLocation;
+ uint32_t NumberofMultisamples;
+};
+
+static inline void
+GEN8_3DSTATE_MULTISAMPLE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_MULTISAMPLE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->PixelPositionOffsetEnable, 5, 5) |
+ __gen_field(values->PixelLocation, 4, 4) |
+ __gen_field(values->NumberofMultisamples, 1, 3) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_POLY_STIPPLE_OFFSET_length_bias 0x00000002
+#define GEN8_3DSTATE_POLY_STIPPLE_OFFSET_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 6, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_POLY_STIPPLE_OFFSET_length 0x00000002
+
+struct GEN8_3DSTATE_POLY_STIPPLE_OFFSET {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PolygonStippleXOffset;
+ uint32_t PolygonStippleYOffset;
+};
+
+static inline void
+GEN8_3DSTATE_POLY_STIPPLE_OFFSET_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_POLY_STIPPLE_OFFSET * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->PolygonStippleXOffset, 8, 12) |
+ __gen_field(values->PolygonStippleYOffset, 0, 4) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_POLY_STIPPLE_PATTERN_length_bias 0x00000002
+#define GEN8_3DSTATE_POLY_STIPPLE_PATTERN_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 7, \
+ .DwordLength = 31
+
+#define GEN8_3DSTATE_POLY_STIPPLE_PATTERN_length 0x00000021
+
+struct GEN8_3DSTATE_POLY_STIPPLE_PATTERN {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PatternRow[32];
+};
+
+static inline void
+GEN8_3DSTATE_POLY_STIPPLE_PATTERN_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_POLY_STIPPLE_PATTERN * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ for (uint32_t i = 0, j = 1; i < 32; i += 1, j++) {
+ dw[j] =
+ __gen_field(values->PatternRow[i + 0], 0, 31) |
+ 0;
+ }
+
+}
+
+#define GEN8_3DSTATE_PS_length_bias 0x00000002
+#define GEN8_3DSTATE_PS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 32, \
+ .DwordLength = 10
+
+#define GEN8_3DSTATE_PS_length 0x0000000c
+
+struct GEN8_3DSTATE_PS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint64_t KernelStartPointer0;
+#define Multiple 0
+#define Single 1
+ uint32_t SingleProgramFlow;
+#define Dmask 0
+#define Vmask 1
+ uint32_t VectorMaskEnable;
+#define NoSamplers 0
+#define _14Samplers 1
+#define _58Samplers 2
+#define _912Samplers 3
+#define _1316Samplers 4
+ uint32_t SamplerCount;
+#define FlushedtoZero 0
+#define Retained 1
+ uint32_t SinglePrecisionDenormalMode;
+ uint32_t BindingTableEntryCount;
+#define Normal 0
+#define High 1
+ uint32_t ThreadDispatchPriority;
+#define IEEE754 0
+#define Alternate 1
+ uint32_t FloatingPointMode;
+#define RTNE 0
+#define RU 1
+#define RD 2
+#define RTZ 3
+ uint32_t RoundingMode;
+ bool IllegalOpcodeExceptionEnable;
+ bool MaskStackExceptionEnable;
+ bool SoftwareExceptionEnable;
+ uint64_t ScratchSpaceBasePointer;
+ uint32_t PerThreadScratchSpace;
+ uint32_t MaximumNumberofThreadsPerPSD;
+ bool PushConstantEnable;
+ bool RenderTargetFastClearEnable;
+ bool RenderTargetResolveEnable;
+#define POSOFFSET_NONE 0
+#define POSOFFSET_CENTROID 2
+#define POSOFFSET_SAMPLE 3
+ uint32_t PositionXYOffsetSelect;
+ bool _32PixelDispatchEnable;
+ bool _16PixelDispatchEnable;
+ bool _8PixelDispatchEnable;
+ uint32_t DispatchGRFStartRegisterForConstantSetupData0;
+ uint32_t DispatchGRFStartRegisterForConstantSetupData1;
+ uint32_t DispatchGRFStartRegisterForConstantSetupData2;
+ uint64_t KernelStartPointer1;
+ uint64_t KernelStartPointer2;
+};
+
+static inline void
+GEN8_3DSTATE_PS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_PS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint64_t qw1 =
+ __gen_offset(values->KernelStartPointer0, 6, 63) |
+ 0;
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+ dw[3] =
+ __gen_field(values->SingleProgramFlow, 31, 31) |
+ __gen_field(values->VectorMaskEnable, 30, 30) |
+ __gen_field(values->SamplerCount, 27, 29) |
+ __gen_field(values->SinglePrecisionDenormalMode, 26, 26) |
+ __gen_field(values->BindingTableEntryCount, 18, 25) |
+ __gen_field(values->ThreadDispatchPriority, 17, 17) |
+ __gen_field(values->FloatingPointMode, 16, 16) |
+ __gen_field(values->RoundingMode, 14, 15) |
+ __gen_field(values->IllegalOpcodeExceptionEnable, 13, 13) |
+ __gen_field(values->MaskStackExceptionEnable, 11, 11) |
+ __gen_field(values->SoftwareExceptionEnable, 7, 7) |
+ 0;
+
+ uint64_t qw4 =
+ __gen_offset(values->ScratchSpaceBasePointer, 10, 63) |
+ __gen_field(values->PerThreadScratchSpace, 0, 3) |
+ 0;
+
+ dw[4] = qw4;
+ dw[5] = qw4 >> 32;
+
+ dw[6] =
+ __gen_field(values->MaximumNumberofThreadsPerPSD, 23, 31) |
+ __gen_field(values->PushConstantEnable, 11, 11) |
+ __gen_field(values->RenderTargetFastClearEnable, 8, 8) |
+ __gen_field(values->RenderTargetResolveEnable, 6, 6) |
+ __gen_field(values->PositionXYOffsetSelect, 3, 4) |
+ __gen_field(values->_32PixelDispatchEnable, 2, 2) |
+ __gen_field(values->_16PixelDispatchEnable, 1, 1) |
+ __gen_field(values->_8PixelDispatchEnable, 0, 0) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->DispatchGRFStartRegisterForConstantSetupData0, 16, 22) |
+ __gen_field(values->DispatchGRFStartRegisterForConstantSetupData1, 8, 14) |
+ __gen_field(values->DispatchGRFStartRegisterForConstantSetupData2, 0, 6) |
+ 0;
+
+ uint64_t qw8 =
+ __gen_offset(values->KernelStartPointer1, 6, 63) |
+ 0;
+
+ dw[8] = qw8;
+ dw[9] = qw8 >> 32;
+
+ uint64_t qw10 =
+ __gen_offset(values->KernelStartPointer2, 6, 63) |
+ 0;
+
+ dw[10] = qw10;
+ dw[11] = qw10 >> 32;
+
+}
+
+#define GEN8_3DSTATE_PS_BLEND_length_bias 0x00000002
+#define GEN8_3DSTATE_PS_BLEND_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 77, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_PS_BLEND_length 0x00000002
+
+struct GEN8_3DSTATE_PS_BLEND {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool AlphaToCoverageEnable;
+ bool HasWriteableRT;
+ bool ColorBufferBlendEnable;
+ uint32_t SourceAlphaBlendFactor;
+ uint32_t DestinationAlphaBlendFactor;
+ uint32_t SourceBlendFactor;
+ uint32_t DestinationBlendFactor;
+ bool AlphaTestEnable;
+ bool IndependentAlphaBlendEnable;
+};
+
+static inline void
+GEN8_3DSTATE_PS_BLEND_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_PS_BLEND * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->AlphaToCoverageEnable, 31, 31) |
+ __gen_field(values->HasWriteableRT, 30, 30) |
+ __gen_field(values->ColorBufferBlendEnable, 29, 29) |
+ __gen_field(values->SourceAlphaBlendFactor, 24, 28) |
+ __gen_field(values->DestinationAlphaBlendFactor, 19, 23) |
+ __gen_field(values->SourceBlendFactor, 14, 18) |
+ __gen_field(values->DestinationBlendFactor, 9, 13) |
+ __gen_field(values->AlphaTestEnable, 8, 8) |
+ __gen_field(values->IndependentAlphaBlendEnable, 7, 7) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_PS_EXTRA_length_bias 0x00000002
+#define GEN8_3DSTATE_PS_EXTRA_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 79, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_PS_EXTRA_length 0x00000002
+
+struct GEN8_3DSTATE_PS_EXTRA {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool PixelShaderValid;
+ bool PixelShaderDoesnotwritetoRT;
+ bool oMaskPresenttoRenderTarget;
+ bool PixelShaderKillsPixel;
+#define PSCDEPTH_OFF 0
+#define PSCDEPTH_ON 1
+#define PSCDEPTH_ON_GE 2
+#define PSCDEPTH_ON_LE 3
+ uint32_t PixelShaderComputedDepthMode;
+ bool ForceComputedDepth;
+ bool PixelShaderUsesSourceDepth;
+ bool PixelShaderUsesSourceW;
+ uint32_t Removed;
+ bool AttributeEnable;
+ bool PixelShaderDisablesAlphaToCoverage;
+ bool PixelShaderIsPerSample;
+ bool PixelShaderHasUAV;
+ bool PixelShaderUsesInputCoverageMask;
+};
+
+static inline void
+GEN8_3DSTATE_PS_EXTRA_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_PS_EXTRA * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->PixelShaderValid, 31, 31) |
+ __gen_field(values->PixelShaderDoesnotwritetoRT, 30, 30) |
+ __gen_field(values->oMaskPresenttoRenderTarget, 29, 29) |
+ __gen_field(values->PixelShaderKillsPixel, 28, 28) |
+ __gen_field(values->PixelShaderComputedDepthMode, 26, 27) |
+ __gen_field(values->ForceComputedDepth, 25, 25) |
+ __gen_field(values->PixelShaderUsesSourceDepth, 24, 24) |
+ __gen_field(values->PixelShaderUsesSourceW, 23, 23) |
+ __gen_field(values->Removed, 17, 17) |
+ __gen_field(values->AttributeEnable, 8, 8) |
+ __gen_field(values->PixelShaderDisablesAlphaToCoverage, 7, 7) |
+ __gen_field(values->PixelShaderIsPerSample, 6, 6) |
+ __gen_field(values->PixelShaderHasUAV, 2, 2) |
+ __gen_field(values->PixelShaderUsesInputCoverageMask, 1, 1) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_DS_length_bias 0x00000002
+#define GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_DS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 20, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_DS_length 0x00000002
+
+struct GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferOffset;
+ uint32_t ConstantBufferSize;
+};
+
+static inline void
+GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferOffset, 16, 20) |
+ __gen_field(values->ConstantBufferSize, 0, 5) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_GS_length_bias 0x00000002
+#define GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_GS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 21, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_GS_length 0x00000002
+
+struct GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferOffset;
+ uint32_t ConstantBufferSize;
+};
+
+static inline void
+GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferOffset, 16, 20) |
+ __gen_field(values->ConstantBufferSize, 0, 5) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_HS_length_bias 0x00000002
+#define GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_HS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 19, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_HS_length 0x00000002
+
+struct GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferOffset;
+ uint32_t ConstantBufferSize;
+};
+
+static inline void
+GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferOffset, 16, 20) |
+ __gen_field(values->ConstantBufferSize, 0, 5) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_PS_length_bias 0x00000002
+#define GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_PS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 22, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_PS_length 0x00000002
+
+struct GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_PS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferOffset;
+ uint32_t ConstantBufferSize;
+};
+
+static inline void
+GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_PS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_PS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferOffset, 16, 20) |
+ __gen_field(values->ConstantBufferSize, 0, 5) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_VS_length_bias 0x00000002
+#define GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_VS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 18, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_VS_length 0x00000002
+
+struct GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferOffset;
+ uint32_t ConstantBufferSize;
+};
+
+static inline void
+GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferOffset, 16, 20) |
+ __gen_field(values->ConstantBufferSize, 0, 5) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_RASTER_length_bias 0x00000002
+#define GEN8_3DSTATE_RASTER_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 80, \
+ .DwordLength = 3
+
+#define GEN8_3DSTATE_RASTER_length 0x00000005
+
+struct GEN8_3DSTATE_RASTER {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define DX9OGL 0
+#define DX100 1
+#define DX101 2
+ uint32_t APIMode;
+#define Clockwise 0
+#define CounterClockwise 1
+ uint32_t FrontWinding;
+#define FSC_NUMRASTSAMPLES_0 0
+#define FSC_NUMRASTSAMPLES_1 1
+#define FSC_NUMRASTSAMPLES_2 2
+#define FSC_NUMRASTSAMPLES_4 3
+#define FSC_NUMRASTSAMPLES_8 4
+#define FSC_NUMRASTSAMPLES_16 5
+ uint32_t ForcedSampleCount;
+#define CULLMODE_BOTH 0
+#define CULLMODE_NONE 1
+#define CULLMODE_FRONT 2
+#define CULLMODE_BACK 3
+ uint32_t CullMode;
+#define Normal 0
+#define Force 1
+ uint32_t ForceMultisampling;
+ bool SmoothPointEnable;
+ bool DXMultisampleRasterizationEnable;
+#define MSRASTMODE_OFF_PIXEL 0
+#define MSRASTMODE_OFF_PATTERN 1
+#define MSRASTMODE_ON_PIXEL 2
+#define MSRASTMODE_ON_PATTERN 3
+ uint32_t DXMultisampleRasterizationMode;
+ bool GlobalDepthOffsetEnableSolid;
+ bool GlobalDepthOffsetEnableWireframe;
+ bool GlobalDepthOffsetEnablePoint;
+#define RASTER_SOLID 0
+#define RASTER_WIREFRAME 1
+#define RASTER_POINT 2
+ uint32_t FrontFaceFillMode;
+#define RASTER_SOLID 0
+#define RASTER_WIREFRAME 1
+#define RASTER_POINT 2
+ uint32_t BackFaceFillMode;
+ bool AntialiasingEnable;
+ bool ScissorRectangleEnable;
+ bool ViewportZClipTestEnable;
+ float GlobalDepthOffsetConstant;
+ float GlobalDepthOffsetScale;
+ float GlobalDepthOffsetClamp;
+};
+
+static inline void
+GEN8_3DSTATE_RASTER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_RASTER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->APIMode, 22, 23) |
+ __gen_field(values->FrontWinding, 21, 21) |
+ __gen_field(values->ForcedSampleCount, 18, 20) |
+ __gen_field(values->CullMode, 16, 17) |
+ __gen_field(values->ForceMultisampling, 14, 14) |
+ __gen_field(values->SmoothPointEnable, 13, 13) |
+ __gen_field(values->DXMultisampleRasterizationEnable, 12, 12) |
+ __gen_field(values->DXMultisampleRasterizationMode, 10, 11) |
+ __gen_field(values->GlobalDepthOffsetEnableSolid, 9, 9) |
+ __gen_field(values->GlobalDepthOffsetEnableWireframe, 8, 8) |
+ __gen_field(values->GlobalDepthOffsetEnablePoint, 7, 7) |
+ __gen_field(values->FrontFaceFillMode, 5, 6) |
+ __gen_field(values->BackFaceFillMode, 3, 4) |
+ __gen_field(values->AntialiasingEnable, 2, 2) |
+ __gen_field(values->ScissorRectangleEnable, 1, 1) |
+ __gen_field(values->ViewportZClipTestEnable, 0, 0) |
+ 0;
+
+ dw[2] =
+ __gen_float(values->GlobalDepthOffsetConstant) |
+ 0;
+
+ dw[3] =
+ __gen_float(values->GlobalDepthOffsetScale) |
+ 0;
+
+ dw[4] =
+ __gen_float(values->GlobalDepthOffsetClamp) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_SAMPLER_PALETTE_LOAD0_length_bias 0x00000002
+#define GEN8_3DSTATE_SAMPLER_PALETTE_LOAD0_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 2
+
+#define GEN8_3DSTATE_SAMPLER_PALETTE_LOAD0_length 0x00000000
+
+#define GEN8_PALETTE_ENTRY_length 0x00000001
+
+struct GEN8_PALETTE_ENTRY {
+ uint32_t Alpha;
+ uint32_t Red;
+ uint32_t Green;
+ uint32_t Blue;
+};
+
+static inline void
+GEN8_PALETTE_ENTRY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_PALETTE_ENTRY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->Alpha, 24, 31) |
+ __gen_field(values->Red, 16, 23) |
+ __gen_field(values->Green, 8, 15) |
+ __gen_field(values->Blue, 0, 7) |
+ 0;
+
+}
+
+struct GEN8_3DSTATE_SAMPLER_PALETTE_LOAD0 {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN8_3DSTATE_SAMPLER_PALETTE_LOAD0_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_SAMPLER_PALETTE_LOAD0 * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN8_3DSTATE_SAMPLER_PALETTE_LOAD1_length_bias 0x00000002
+#define GEN8_3DSTATE_SAMPLER_PALETTE_LOAD1_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 12
+
+#define GEN8_3DSTATE_SAMPLER_PALETTE_LOAD1_length 0x00000000
+
+struct GEN8_3DSTATE_SAMPLER_PALETTE_LOAD1 {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN8_3DSTATE_SAMPLER_PALETTE_LOAD1_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_SAMPLER_PALETTE_LOAD1 * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN8_3DSTATE_SAMPLER_STATE_POINTERS_DS_length_bias 0x00000002
+#define GEN8_3DSTATE_SAMPLER_STATE_POINTERS_DS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 45, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_SAMPLER_STATE_POINTERS_DS_length 0x00000002
+
+struct GEN8_3DSTATE_SAMPLER_STATE_POINTERS_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoDSSamplerState;
+};
+
+static inline void
+GEN8_3DSTATE_SAMPLER_STATE_POINTERS_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_SAMPLER_STATE_POINTERS_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoDSSamplerState, 5, 31) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_SAMPLER_STATE_POINTERS_GS_length_bias 0x00000002
+#define GEN8_3DSTATE_SAMPLER_STATE_POINTERS_GS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 46, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_SAMPLER_STATE_POINTERS_GS_length 0x00000002
+
+struct GEN8_3DSTATE_SAMPLER_STATE_POINTERS_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoGSSamplerState;
+};
+
+static inline void
+GEN8_3DSTATE_SAMPLER_STATE_POINTERS_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_SAMPLER_STATE_POINTERS_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoGSSamplerState, 5, 31) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_SAMPLER_STATE_POINTERS_HS_length_bias 0x00000002
+#define GEN8_3DSTATE_SAMPLER_STATE_POINTERS_HS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 44, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_SAMPLER_STATE_POINTERS_HS_length 0x00000002
+
+struct GEN8_3DSTATE_SAMPLER_STATE_POINTERS_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoHSSamplerState;
+};
+
+static inline void
+GEN8_3DSTATE_SAMPLER_STATE_POINTERS_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_SAMPLER_STATE_POINTERS_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoHSSamplerState, 5, 31) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_SAMPLER_STATE_POINTERS_PS_length_bias 0x00000002
+#define GEN8_3DSTATE_SAMPLER_STATE_POINTERS_PS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 47, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_SAMPLER_STATE_POINTERS_PS_length 0x00000002
+
+struct GEN8_3DSTATE_SAMPLER_STATE_POINTERS_PS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoPSSamplerState;
+};
+
+static inline void
+GEN8_3DSTATE_SAMPLER_STATE_POINTERS_PS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_SAMPLER_STATE_POINTERS_PS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoPSSamplerState, 5, 31) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_SAMPLER_STATE_POINTERS_VS_length_bias 0x00000002
+#define GEN8_3DSTATE_SAMPLER_STATE_POINTERS_VS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 43, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_SAMPLER_STATE_POINTERS_VS_length 0x00000002
+
+struct GEN8_3DSTATE_SAMPLER_STATE_POINTERS_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoVSSamplerState;
+};
+
+static inline void
+GEN8_3DSTATE_SAMPLER_STATE_POINTERS_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_SAMPLER_STATE_POINTERS_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoVSSamplerState, 5, 31) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_SAMPLE_MASK_length_bias 0x00000002
+#define GEN8_3DSTATE_SAMPLE_MASK_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 24, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_SAMPLE_MASK_length 0x00000002
+
+struct GEN8_3DSTATE_SAMPLE_MASK {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t SampleMask;
+};
+
+static inline void
+GEN8_3DSTATE_SAMPLE_MASK_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_SAMPLE_MASK * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->SampleMask, 0, 15) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_SAMPLE_PATTERN_length_bias 0x00000002
+#define GEN8_3DSTATE_SAMPLE_PATTERN_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 28, \
+ .DwordLength = 7
+
+#define GEN8_3DSTATE_SAMPLE_PATTERN_length 0x00000009
+
+struct GEN8_3DSTATE_SAMPLE_PATTERN {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ float _8xSample7XOffset;
+ float _8xSample7YOffset;
+ float _8xSample6XOffset;
+ float _8xSample6YOffset;
+ float _8xSample5XOffset;
+ float _8xSample5YOffset;
+ float _8xSample4XOffset;
+ float _8xSample4YOffset;
+ float _8xSample3XOffset;
+ float _8xSample3YOffset;
+ float _8xSample2XOffset;
+ float _8xSample2YOffset;
+ float _8xSample1XOffset;
+ float _8xSample1YOffset;
+ float _8xSample0XOffset;
+ float _8xSample0YOffset;
+ float _4xSample3XOffset;
+ float _4xSample3YOffset;
+ float _4xSample2XOffset;
+ float _4xSample2YOffset;
+ float _4xSample1XOffset;
+ float _4xSample1YOffset;
+ float _4xSample0XOffset;
+ float _4xSample0YOffset;
+ float _1xSample0XOffset;
+ float _1xSample0YOffset;
+ float _2xSample1XOffset;
+ float _2xSample1YOffset;
+ float _2xSample0XOffset;
+ float _2xSample0YOffset;
+};
+
+static inline void
+GEN8_3DSTATE_SAMPLE_PATTERN_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_SAMPLE_PATTERN * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ for (uint32_t i = 0, j = 1; i < 4; i += 1, j++) {
+ dw[j] =
+ 0;
+ }
+
+ dw[5] =
+ __gen_field(values->_8xSample7XOffset * (1 << 4), 28, 31) |
+ __gen_field(values->_8xSample7YOffset * (1 << 4), 24, 27) |
+ __gen_field(values->_8xSample6XOffset * (1 << 4), 20, 23) |
+ __gen_field(values->_8xSample6YOffset * (1 << 4), 16, 19) |
+ __gen_field(values->_8xSample5XOffset * (1 << 4), 12, 15) |
+ __gen_field(values->_8xSample5YOffset * (1 << 4), 8, 11) |
+ __gen_field(values->_8xSample4XOffset * (1 << 4), 4, 7) |
+ __gen_field(values->_8xSample4YOffset * (1 << 4), 0, 3) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->_8xSample3XOffset * (1 << 4), 28, 31) |
+ __gen_field(values->_8xSample3YOffset * (1 << 4), 24, 27) |
+ __gen_field(values->_8xSample2XOffset * (1 << 4), 20, 23) |
+ __gen_field(values->_8xSample2YOffset * (1 << 4), 16, 19) |
+ __gen_field(values->_8xSample1XOffset * (1 << 4), 12, 15) |
+ __gen_field(values->_8xSample1YOffset * (1 << 4), 8, 11) |
+ __gen_field(values->_8xSample0XOffset * (1 << 4), 4, 7) |
+ __gen_field(values->_8xSample0YOffset * (1 << 4), 0, 3) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->_4xSample3XOffset * (1 << 4), 28, 31) |
+ __gen_field(values->_4xSample3YOffset * (1 << 4), 24, 27) |
+ __gen_field(values->_4xSample2XOffset * (1 << 4), 20, 23) |
+ __gen_field(values->_4xSample2YOffset * (1 << 4), 16, 19) |
+ __gen_field(values->_4xSample1XOffset * (1 << 4), 12, 15) |
+ __gen_field(values->_4xSample1YOffset * (1 << 4), 8, 11) |
+ __gen_field(values->_4xSample0XOffset * (1 << 4), 4, 7) |
+ __gen_field(values->_4xSample0YOffset * (1 << 4), 0, 3) |
+ 0;
+
+ dw[8] =
+ __gen_field(values->_1xSample0XOffset * (1 << 4), 20, 23) |
+ __gen_field(values->_1xSample0YOffset * (1 << 4), 16, 19) |
+ __gen_field(values->_2xSample1XOffset * (1 << 4), 12, 15) |
+ __gen_field(values->_2xSample1YOffset * (1 << 4), 8, 11) |
+ __gen_field(values->_2xSample0XOffset * (1 << 4), 4, 7) |
+ __gen_field(values->_2xSample0YOffset * (1 << 4), 0, 3) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_SBE_length_bias 0x00000002
+#define GEN8_3DSTATE_SBE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 31, \
+ .DwordLength = 2
+
+#define GEN8_3DSTATE_SBE_length 0x00000004
+
+struct GEN8_3DSTATE_SBE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool ForceVertexURBEntryReadLength;
+ bool ForceVertexURBEntryReadOffset;
+ uint32_t NumberofSFOutputAttributes;
+ bool AttributeSwizzleEnable;
+#define UPPERLEFT 0
+#define LOWERLEFT 1
+ uint32_t PointSpriteTextureCoordinateOrigin;
+ bool PrimitiveIDOverrideComponentW;
+ bool PrimitiveIDOverrideComponentZ;
+ bool PrimitiveIDOverrideComponentY;
+ bool PrimitiveIDOverrideComponentX;
+ uint32_t VertexURBEntryReadLength;
+ uint32_t VertexURBEntryReadOffset;
+ uint32_t PrimitiveIDOverrideAttributeSelect;
+ uint32_t PointSpriteTextureCoordinateEnable;
+ uint32_t ConstantInterpolationEnable;
+};
+
+static inline void
+GEN8_3DSTATE_SBE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_SBE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ForceVertexURBEntryReadLength, 29, 29) |
+ __gen_field(values->ForceVertexURBEntryReadOffset, 28, 28) |
+ __gen_field(values->NumberofSFOutputAttributes, 22, 27) |
+ __gen_field(values->AttributeSwizzleEnable, 21, 21) |
+ __gen_field(values->PointSpriteTextureCoordinateOrigin, 20, 20) |
+ __gen_field(values->PrimitiveIDOverrideComponentW, 19, 19) |
+ __gen_field(values->PrimitiveIDOverrideComponentZ, 18, 18) |
+ __gen_field(values->PrimitiveIDOverrideComponentY, 17, 17) |
+ __gen_field(values->PrimitiveIDOverrideComponentX, 16, 16) |
+ __gen_field(values->VertexURBEntryReadLength, 11, 15) |
+ __gen_field(values->VertexURBEntryReadOffset, 5, 10) |
+ __gen_field(values->PrimitiveIDOverrideAttributeSelect, 0, 4) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->PointSpriteTextureCoordinateEnable, 0, 31) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->ConstantInterpolationEnable, 0, 31) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_SBE_SWIZ_length_bias 0x00000002
+#define GEN8_3DSTATE_SBE_SWIZ_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 81, \
+ .DwordLength = 9
+
+#define GEN8_3DSTATE_SBE_SWIZ_length 0x0000000b
+
+#define GEN8_SF_OUTPUT_ATTRIBUTE_DETAIL_length 0x00000001
+
+struct GEN8_SF_OUTPUT_ATTRIBUTE_DETAIL {
+ bool ComponentOverrideW;
+ bool ComponentOverrideZ;
+ bool ComponentOverrideY;
+ bool ComponentOverrideX;
+ uint32_t SwizzleControlMode;
+#define CONST_0000 0
+#define CONST_0001_FLOAT 1
+#define CONST_1111_FLOAT 2
+#define PRIM_ID 3
+ uint32_t ConstantSource;
+#define INPUTATTR 0
+#define INPUTATTR_FACING 1
+#define INPUTATTR_W 2
+#define INPUTATTR_FACING_W 3
+ uint32_t SwizzleSelect;
+ uint32_t SourceAttribute;
+};
+
+static inline void
+GEN8_SF_OUTPUT_ATTRIBUTE_DETAIL_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_SF_OUTPUT_ATTRIBUTE_DETAIL * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->ComponentOverrideW, 15, 15) |
+ __gen_field(values->ComponentOverrideZ, 14, 14) |
+ __gen_field(values->ComponentOverrideY, 13, 13) |
+ __gen_field(values->ComponentOverrideX, 12, 12) |
+ __gen_field(values->SwizzleControlMode, 11, 11) |
+ __gen_field(values->ConstantSource, 9, 10) |
+ __gen_field(values->SwizzleSelect, 6, 7) |
+ __gen_field(values->SourceAttribute, 0, 4) |
+ 0;
+
+}
+
+struct GEN8_3DSTATE_SBE_SWIZ {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ struct GEN8_SF_OUTPUT_ATTRIBUTE_DETAIL Attribute[16];
+ uint32_t AttributeWrapShortestEnables[16];
+};
+
+static inline void
+GEN8_3DSTATE_SBE_SWIZ_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_SBE_SWIZ * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ for (uint32_t i = 0, j = 1; i < 16; i += 2, j++) {
+ uint32_t dw_Attribute0;
+ GEN8_SF_OUTPUT_ATTRIBUTE_DETAIL_pack(data, &dw_Attribute0, &values->Attribute[i + 0]);
+ uint32_t dw_Attribute1;
+ GEN8_SF_OUTPUT_ATTRIBUTE_DETAIL_pack(data, &dw_Attribute1, &values->Attribute[i + 1]);
+ dw[j] =
+ __gen_field(dw_Attribute0, 0, 15) |
+ __gen_field(dw_Attribute1, 16, 31) |
+ 0;
+ }
+
+ for (uint32_t i = 0, j = 9; i < 16; i += 8, j++) {
+ dw[j] =
+ __gen_field(values->AttributeWrapShortestEnables[i + 0], 0, 3) |
+ __gen_field(values->AttributeWrapShortestEnables[i + 1], 4, 7) |
+ __gen_field(values->AttributeWrapShortestEnables[i + 2], 8, 11) |
+ __gen_field(values->AttributeWrapShortestEnables[i + 3], 12, 15) |
+ __gen_field(values->AttributeWrapShortestEnables[i + 4], 16, 19) |
+ __gen_field(values->AttributeWrapShortestEnables[i + 5], 20, 23) |
+ __gen_field(values->AttributeWrapShortestEnables[i + 6], 24, 27) |
+ __gen_field(values->AttributeWrapShortestEnables[i + 7], 28, 31) |
+ 0;
+ }
+
+}
+
+#define GEN8_3DSTATE_SCISSOR_STATE_POINTERS_length_bias 0x00000002
+#define GEN8_3DSTATE_SCISSOR_STATE_POINTERS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 15, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_SCISSOR_STATE_POINTERS_length 0x00000002
+
+struct GEN8_3DSTATE_SCISSOR_STATE_POINTERS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ScissorRectPointer;
+};
+
+static inline void
+GEN8_3DSTATE_SCISSOR_STATE_POINTERS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_SCISSOR_STATE_POINTERS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->ScissorRectPointer, 5, 31) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_SF_length_bias 0x00000002
+#define GEN8_3DSTATE_SF_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 19, \
+ .DwordLength = 2
+
+#define GEN8_3DSTATE_SF_length 0x00000004
+
+struct GEN8_3DSTATE_SF {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool LegacyGlobalDepthBiasEnable;
+ bool StatisticsEnable;
+ bool ViewportTransformEnable;
+ float LineWidth;
+#define _05pixels 0
+#define _10pixels 1
+#define _20pixels 2
+#define _40pixels 3
+ uint32_t LineEndCapAntialiasingRegionWidth;
+ bool LastPixelEnable;
+ uint32_t TriangleStripListProvokingVertexSelect;
+ uint32_t LineStripListProvokingVertexSelect;
+ uint32_t TriangleFanProvokingVertexSelect;
+#define AALINEDISTANCE_TRUE 1
+ uint32_t AALineDistanceMode;
+ bool SmoothPointEnable;
+ uint32_t VertexSubPixelPrecisionSelect;
+#define Vertex 0
+#define State 1
+ uint32_t PointWidthSource;
+ float PointWidth;
+};
+
+static inline void
+GEN8_3DSTATE_SF_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_SF * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->LegacyGlobalDepthBiasEnable, 11, 11) |
+ __gen_field(values->StatisticsEnable, 10, 10) |
+ __gen_field(values->ViewportTransformEnable, 1, 1) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->LineWidth * (1 << 7), 18, 27) |
+ __gen_field(values->LineEndCapAntialiasingRegionWidth, 16, 17) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->LastPixelEnable, 31, 31) |
+ __gen_field(values->TriangleStripListProvokingVertexSelect, 29, 30) |
+ __gen_field(values->LineStripListProvokingVertexSelect, 27, 28) |
+ __gen_field(values->TriangleFanProvokingVertexSelect, 25, 26) |
+ __gen_field(values->AALineDistanceMode, 14, 14) |
+ __gen_field(values->SmoothPointEnable, 13, 13) |
+ __gen_field(values->VertexSubPixelPrecisionSelect, 12, 12) |
+ __gen_field(values->PointWidthSource, 11, 11) |
+ __gen_field(values->PointWidth * (1 << 3), 0, 10) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_SO_BUFFER_length_bias 0x00000002
+#define GEN8_3DSTATE_SO_BUFFER_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 24, \
+ .DwordLength = 6
+
+#define GEN8_3DSTATE_SO_BUFFER_length 0x00000008
+
+struct GEN8_3DSTATE_SO_BUFFER {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool SOBufferEnable;
+ uint32_t SOBufferIndex;
+ struct GEN8_MEMORY_OBJECT_CONTROL_STATE SOBufferObjectControlState;
+ bool StreamOffsetWriteEnable;
+ bool StreamOutputBufferOffsetAddressEnable;
+ __gen_address_type SurfaceBaseAddress;
+ uint32_t SurfaceSize;
+ __gen_address_type StreamOutputBufferOffsetAddress;
+ uint32_t StreamOffset;
+};
+
+static inline void
+GEN8_3DSTATE_SO_BUFFER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_SO_BUFFER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw_SOBufferObjectControlState;
+ GEN8_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_SOBufferObjectControlState, &values->SOBufferObjectControlState);
+ dw[1] =
+ __gen_field(values->SOBufferEnable, 31, 31) |
+ __gen_field(values->SOBufferIndex, 29, 30) |
+ __gen_field(dw_SOBufferObjectControlState, 22, 28) |
+ __gen_field(values->StreamOffsetWriteEnable, 21, 21) |
+ __gen_field(values->StreamOutputBufferOffsetAddressEnable, 20, 20) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ uint64_t qw2 =
+ __gen_combine_address(data, &dw[2], values->SurfaceBaseAddress, dw2);
+
+ dw[2] = qw2;
+ dw[3] = qw2 >> 32;
+
+ dw[4] =
+ __gen_field(values->SurfaceSize, 0, 29) |
+ 0;
+
+ uint32_t dw5 =
+ 0;
+
+ uint64_t qw5 =
+ __gen_combine_address(data, &dw[5], values->StreamOutputBufferOffsetAddress, dw5);
+
+ dw[5] = qw5;
+ dw[6] = qw5 >> 32;
+
+ dw[7] =
+ __gen_field(values->StreamOffset, 0, 31) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_SO_DECL_LIST_length_bias 0x00000002
+#define GEN8_3DSTATE_SO_DECL_LIST_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 23
+
+#define GEN8_3DSTATE_SO_DECL_LIST_length 0x00000000
+
+#define GEN8_SO_DECL_ENTRY_length 0x00000002
+
+#define GEN8_SO_DECL_length 0x00000001
+
+struct GEN8_SO_DECL {
+ uint32_t OutputBufferSlot;
+ uint32_t HoleFlag;
+ uint32_t RegisterIndex;
+ uint32_t ComponentMask;
+};
+
+static inline void
+GEN8_SO_DECL_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_SO_DECL * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->OutputBufferSlot, 12, 13) |
+ __gen_field(values->HoleFlag, 11, 11) |
+ __gen_field(values->RegisterIndex, 4, 9) |
+ __gen_field(values->ComponentMask, 0, 3) |
+ 0;
+
+}
+
+struct GEN8_SO_DECL_ENTRY {
+ struct GEN8_SO_DECL Stream3Decl;
+ struct GEN8_SO_DECL Stream2Decl;
+ struct GEN8_SO_DECL Stream1Decl;
+ struct GEN8_SO_DECL Stream0Decl;
+};
+
+static inline void
+GEN8_SO_DECL_ENTRY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_SO_DECL_ENTRY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ uint32_t dw_Stream3Decl;
+ GEN8_SO_DECL_pack(data, &dw_Stream3Decl, &values->Stream3Decl);
+ uint32_t dw_Stream2Decl;
+ GEN8_SO_DECL_pack(data, &dw_Stream2Decl, &values->Stream2Decl);
+ uint32_t dw_Stream1Decl;
+ GEN8_SO_DECL_pack(data, &dw_Stream1Decl, &values->Stream1Decl);
+ uint32_t dw_Stream0Decl;
+ GEN8_SO_DECL_pack(data, &dw_Stream0Decl, &values->Stream0Decl);
+ uint64_t qw0 =
+ __gen_field(dw_Stream3Decl, 48, 63) |
+ __gen_field(dw_Stream2Decl, 32, 47) |
+ __gen_field(dw_Stream1Decl, 16, 31) |
+ __gen_field(dw_Stream0Decl, 0, 15) |
+ 0;
+
+ dw[0] = qw0;
+ dw[1] = qw0 >> 32;
+
+}
+
+struct GEN8_3DSTATE_SO_DECL_LIST {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t StreamtoBufferSelects3;
+ uint32_t StreamtoBufferSelects2;
+ uint32_t StreamtoBufferSelects1;
+ uint32_t StreamtoBufferSelects0;
+ uint32_t NumEntries3;
+ uint32_t NumEntries2;
+ uint32_t NumEntries1;
+ uint32_t NumEntries0;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN8_3DSTATE_SO_DECL_LIST_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_SO_DECL_LIST * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 8) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->StreamtoBufferSelects3, 12, 15) |
+ __gen_field(values->StreamtoBufferSelects2, 8, 11) |
+ __gen_field(values->StreamtoBufferSelects1, 4, 7) |
+ __gen_field(values->StreamtoBufferSelects0, 0, 3) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->NumEntries3, 24, 31) |
+ __gen_field(values->NumEntries2, 16, 23) |
+ __gen_field(values->NumEntries1, 8, 15) |
+ __gen_field(values->NumEntries0, 0, 7) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN8_3DSTATE_STENCIL_BUFFER_length_bias 0x00000002
+#define GEN8_3DSTATE_STENCIL_BUFFER_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 6, \
+ .DwordLength = 3
+
+#define GEN8_3DSTATE_STENCIL_BUFFER_length 0x00000005
+
+struct GEN8_3DSTATE_STENCIL_BUFFER {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t StencilBufferEnable;
+ struct GEN8_MEMORY_OBJECT_CONTROL_STATE StencilBufferObjectControlState;
+ uint32_t SurfacePitch;
+ __gen_address_type SurfaceBaseAddress;
+ uint32_t SurfaceQPitch;
+};
+
+static inline void
+GEN8_3DSTATE_STENCIL_BUFFER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_STENCIL_BUFFER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw_StencilBufferObjectControlState;
+ GEN8_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_StencilBufferObjectControlState, &values->StencilBufferObjectControlState);
+ dw[1] =
+ __gen_field(values->StencilBufferEnable, 31, 31) |
+ __gen_field(dw_StencilBufferObjectControlState, 22, 28) |
+ __gen_field(values->SurfacePitch, 0, 16) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ uint64_t qw2 =
+ __gen_combine_address(data, &dw[2], values->SurfaceBaseAddress, dw2);
+
+ dw[2] = qw2;
+ dw[3] = qw2 >> 32;
+
+ dw[4] =
+ __gen_field(values->SurfaceQPitch, 0, 14) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_STREAMOUT_length_bias 0x00000002
+#define GEN8_3DSTATE_STREAMOUT_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 30, \
+ .DwordLength = 3
+
+#define GEN8_3DSTATE_STREAMOUT_length 0x00000005
+
+struct GEN8_3DSTATE_STREAMOUT {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t SOFunctionEnable;
+ uint32_t APIRenderingDisable;
+ uint32_t RenderStreamSelect;
+#define LEADING 0
+#define TRAILING 1
+ uint32_t ReorderMode;
+ bool SOStatisticsEnable;
+#define Normal 0
+#define Resreved 1
+#define Force_Off 2
+#define Force_on 3
+ uint32_t ForceRendering;
+ uint32_t Stream3VertexReadOffset;
+ uint32_t Stream3VertexReadLength;
+ uint32_t Stream2VertexReadOffset;
+ uint32_t Stream2VertexReadLength;
+ uint32_t Stream1VertexReadOffset;
+ uint32_t Stream1VertexReadLength;
+ uint32_t Stream0VertexReadOffset;
+ uint32_t Stream0VertexReadLength;
+ uint32_t Buffer1SurfacePitch;
+ uint32_t Buffer0SurfacePitch;
+ uint32_t Buffer3SurfacePitch;
+ uint32_t Buffer2SurfacePitch;
+};
+
+static inline void
+GEN8_3DSTATE_STREAMOUT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_STREAMOUT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->SOFunctionEnable, 31, 31) |
+ __gen_field(values->APIRenderingDisable, 30, 30) |
+ __gen_field(values->RenderStreamSelect, 27, 28) |
+ __gen_field(values->ReorderMode, 26, 26) |
+ __gen_field(values->SOStatisticsEnable, 25, 25) |
+ __gen_field(values->ForceRendering, 23, 24) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->Stream3VertexReadOffset, 29, 29) |
+ __gen_field(values->Stream3VertexReadLength, 24, 28) |
+ __gen_field(values->Stream2VertexReadOffset, 21, 21) |
+ __gen_field(values->Stream2VertexReadLength, 16, 20) |
+ __gen_field(values->Stream1VertexReadOffset, 13, 13) |
+ __gen_field(values->Stream1VertexReadLength, 8, 12) |
+ __gen_field(values->Stream0VertexReadOffset, 5, 5) |
+ __gen_field(values->Stream0VertexReadLength, 0, 4) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->Buffer1SurfacePitch, 16, 27) |
+ __gen_field(values->Buffer0SurfacePitch, 0, 11) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->Buffer3SurfacePitch, 16, 27) |
+ __gen_field(values->Buffer2SurfacePitch, 0, 11) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_TE_length_bias 0x00000002
+#define GEN8_3DSTATE_TE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 28, \
+ .DwordLength = 2
+
+#define GEN8_3DSTATE_TE_length 0x00000004
+
+struct GEN8_3DSTATE_TE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define INTEGER 0
+#define ODD_FRACTIONAL 1
+#define EVEN_FRACTIONAL 2
+ uint32_t Partitioning;
+#define POINT 0
+#define OUTPUT_LINE 1
+#define OUTPUT_TRI_CW 2
+#define OUTPUT_TRI_CCW 3
+ uint32_t OutputTopology;
+#define QUAD 0
+#define TRI 1
+#define ISOLINE 2
+ uint32_t TEDomain;
+#define HW_TESS 0
+#define SW_TESS 1
+ uint32_t TEMode;
+ bool TEEnable;
+ float MaximumTessellationFactorOdd;
+ float MaximumTessellationFactorNotOdd;
+};
+
+static inline void
+GEN8_3DSTATE_TE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_TE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->Partitioning, 12, 13) |
+ __gen_field(values->OutputTopology, 8, 9) |
+ __gen_field(values->TEDomain, 4, 5) |
+ __gen_field(values->TEMode, 1, 2) |
+ __gen_field(values->TEEnable, 0, 0) |
+ 0;
+
+ dw[2] =
+ __gen_float(values->MaximumTessellationFactorOdd) |
+ 0;
+
+ dw[3] =
+ __gen_float(values->MaximumTessellationFactorNotOdd) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_URB_DS_length_bias 0x00000002
+#define GEN8_3DSTATE_URB_DS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 50, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_URB_DS_length 0x00000002
+
+struct GEN8_3DSTATE_URB_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t DSURBStartingAddress;
+ uint32_t DSURBEntryAllocationSize;
+ uint32_t DSNumberofURBEntries;
+};
+
+static inline void
+GEN8_3DSTATE_URB_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_URB_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->DSURBStartingAddress, 25, 31) |
+ __gen_field(values->DSURBEntryAllocationSize, 16, 24) |
+ __gen_field(values->DSNumberofURBEntries, 0, 15) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_URB_GS_length_bias 0x00000002
+#define GEN8_3DSTATE_URB_GS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 51, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_URB_GS_length 0x00000002
+
+struct GEN8_3DSTATE_URB_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t GSURBStartingAddress;
+ uint32_t GSURBEntryAllocationSize;
+ uint32_t GSNumberofURBEntries;
+};
+
+static inline void
+GEN8_3DSTATE_URB_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_URB_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->GSURBStartingAddress, 25, 31) |
+ __gen_field(values->GSURBEntryAllocationSize, 16, 24) |
+ __gen_field(values->GSNumberofURBEntries, 0, 15) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_URB_HS_length_bias 0x00000002
+#define GEN8_3DSTATE_URB_HS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 49, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_URB_HS_length 0x00000002
+
+struct GEN8_3DSTATE_URB_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t HSURBStartingAddress;
+ uint32_t HSURBEntryAllocationSize;
+ uint32_t HSNumberofURBEntries;
+};
+
+static inline void
+GEN8_3DSTATE_URB_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_URB_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->HSURBStartingAddress, 25, 31) |
+ __gen_field(values->HSURBEntryAllocationSize, 16, 24) |
+ __gen_field(values->HSNumberofURBEntries, 0, 15) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_VERTEX_BUFFERS_length_bias 0x00000002
+#define GEN8_3DSTATE_VERTEX_BUFFERS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 8
+
+#define GEN8_3DSTATE_VERTEX_BUFFERS_length 0x00000000
+
+#define GEN8_VERTEX_BUFFER_STATE_length 0x00000004
+
+struct GEN8_VERTEX_BUFFER_STATE {
+ uint32_t VertexBufferIndex;
+ struct GEN8_MEMORY_OBJECT_CONTROL_STATE MemoryObjectControlState;
+ uint32_t AddressModifyEnable;
+ bool NullVertexBuffer;
+ uint32_t BufferPitch;
+ __gen_address_type BufferStartingAddress;
+ uint32_t BufferSize;
+};
+
+static inline void
+GEN8_VERTEX_BUFFER_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_VERTEX_BUFFER_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ uint32_t dw_MemoryObjectControlState;
+ GEN8_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_MemoryObjectControlState, &values->MemoryObjectControlState);
+ dw[0] =
+ __gen_field(values->VertexBufferIndex, 26, 31) |
+ __gen_field(dw_MemoryObjectControlState, 16, 22) |
+ __gen_field(values->AddressModifyEnable, 14, 14) |
+ __gen_field(values->NullVertexBuffer, 13, 13) |
+ __gen_field(values->BufferPitch, 0, 11) |
+ 0;
+
+ uint32_t dw1 =
+ 0;
+
+ uint64_t qw1 =
+ __gen_combine_address(data, &dw[1], values->BufferStartingAddress, dw1);
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+ dw[3] =
+ __gen_field(values->BufferSize, 0, 31) |
+ 0;
+
+}
+
+struct GEN8_3DSTATE_VERTEX_BUFFERS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN8_3DSTATE_VERTEX_BUFFERS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_VERTEX_BUFFERS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN8_3DSTATE_VERTEX_ELEMENTS_length_bias 0x00000002
+#define GEN8_3DSTATE_VERTEX_ELEMENTS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 9
+
+#define GEN8_3DSTATE_VERTEX_ELEMENTS_length 0x00000000
+
+#define GEN8_VERTEX_ELEMENT_STATE_length 0x00000002
+
+struct GEN8_VERTEX_ELEMENT_STATE {
+ uint32_t VertexBufferIndex;
+ bool Valid;
+ uint32_t SourceElementFormat;
+ bool EdgeFlagEnable;
+ uint32_t SourceElementOffset;
+ uint32_t Component0Control;
+ uint32_t Component1Control;
+ uint32_t Component2Control;
+ uint32_t Component3Control;
+};
+
+static inline void
+GEN8_VERTEX_ELEMENT_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_VERTEX_ELEMENT_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->VertexBufferIndex, 26, 31) |
+ __gen_field(values->Valid, 25, 25) |
+ __gen_field(values->SourceElementFormat, 16, 24) |
+ __gen_field(values->EdgeFlagEnable, 15, 15) |
+ __gen_field(values->SourceElementOffset, 0, 11) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->Component0Control, 28, 30) |
+ __gen_field(values->Component1Control, 24, 26) |
+ __gen_field(values->Component2Control, 20, 22) |
+ __gen_field(values->Component3Control, 16, 18) |
+ 0;
+
+}
+
+struct GEN8_3DSTATE_VERTEX_ELEMENTS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN8_3DSTATE_VERTEX_ELEMENTS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_VERTEX_ELEMENTS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN8_3DSTATE_VF_length_bias 0x00000002
+#define GEN8_3DSTATE_VF_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 12, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_VF_length 0x00000002
+
+struct GEN8_3DSTATE_VF {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ bool IndexedDrawCutIndexEnable;
+ uint32_t DwordLength;
+ uint32_t CutIndex;
+};
+
+static inline void
+GEN8_3DSTATE_VF_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_VF * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->IndexedDrawCutIndexEnable, 8, 8) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->CutIndex, 0, 31) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_VF_INSTANCING_length_bias 0x00000002
+#define GEN8_3DSTATE_VF_INSTANCING_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 73, \
+ .DwordLength = 1
+
+#define GEN8_3DSTATE_VF_INSTANCING_length 0x00000003
+
+struct GEN8_3DSTATE_VF_INSTANCING {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool InstancingEnable;
+ uint32_t VertexElementIndex;
+ uint32_t InstanceDataStepRate;
+};
+
+static inline void
+GEN8_3DSTATE_VF_INSTANCING_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_VF_INSTANCING * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->InstancingEnable, 8, 8) |
+ __gen_field(values->VertexElementIndex, 0, 5) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->InstanceDataStepRate, 0, 31) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_VF_SGVS_length_bias 0x00000002
+#define GEN8_3DSTATE_VF_SGVS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 74, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_VF_SGVS_length 0x00000002
+
+struct GEN8_3DSTATE_VF_SGVS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool InstanceIDEnable;
+#define COMP_0 0
+#define COMP_1 1
+#define COMP_2 2
+#define COMP_3 3
+ uint32_t InstanceIDComponentNumber;
+ uint32_t InstanceIDElementOffset;
+ bool VertexIDEnable;
+#define COMP_0 0
+#define COMP_1 1
+#define COMP_2 2
+#define COMP_3 3
+ uint32_t VertexIDComponentNumber;
+ uint32_t VertexIDElementOffset;
+};
+
+static inline void
+GEN8_3DSTATE_VF_SGVS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_VF_SGVS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->InstanceIDEnable, 31, 31) |
+ __gen_field(values->InstanceIDComponentNumber, 29, 30) |
+ __gen_field(values->InstanceIDElementOffset, 16, 21) |
+ __gen_field(values->VertexIDEnable, 15, 15) |
+ __gen_field(values->VertexIDComponentNumber, 13, 14) |
+ __gen_field(values->VertexIDElementOffset, 0, 5) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_VF_STATISTICS_length_bias 0x00000001
+#define GEN8_3DSTATE_VF_STATISTICS_header \
+ .CommandType = 3, \
+ .CommandSubType = 1, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 11
+
+#define GEN8_3DSTATE_VF_STATISTICS_length 0x00000001
+
+struct GEN8_3DSTATE_VF_STATISTICS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ bool StatisticsEnable;
+};
+
+static inline void
+GEN8_3DSTATE_VF_STATISTICS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_VF_STATISTICS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->StatisticsEnable, 0, 0) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_VF_TOPOLOGY_length_bias 0x00000002
+#define GEN8_3DSTATE_VF_TOPOLOGY_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 75, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_VF_TOPOLOGY_length 0x00000002
+
+struct GEN8_3DSTATE_VF_TOPOLOGY {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PrimitiveTopologyType;
+};
+
+static inline void
+GEN8_3DSTATE_VF_TOPOLOGY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_VF_TOPOLOGY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->PrimitiveTopologyType, 0, 5) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_VIEWPORT_STATE_POINTERS_CC_length_bias 0x00000002
+#define GEN8_3DSTATE_VIEWPORT_STATE_POINTERS_CC_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 35, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_VIEWPORT_STATE_POINTERS_CC_length 0x00000002
+
+struct GEN8_3DSTATE_VIEWPORT_STATE_POINTERS_CC {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t CCViewportPointer;
+};
+
+static inline void
+GEN8_3DSTATE_VIEWPORT_STATE_POINTERS_CC_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_VIEWPORT_STATE_POINTERS_CC * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->CCViewportPointer, 5, 31) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP_length_bias 0x00000002
+#define GEN8_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 33, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP_length 0x00000002
+
+struct GEN8_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t SFClipViewportPointer;
+};
+
+static inline void
+GEN8_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->SFClipViewportPointer, 6, 31) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_WM_length_bias 0x00000002
+#define GEN8_3DSTATE_WM_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 20, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_WM_length 0x00000002
+
+struct GEN8_3DSTATE_WM {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool StatisticsEnable;
+ bool LegacyDepthBufferClearEnable;
+ bool LegacyDepthBufferResolveEnable;
+ bool LegacyHierarchicalDepthBufferResolveEnable;
+ bool LegacyDiamondLineRasterization;
+#define NORMAL 0
+#define PSEXEC 1
+#define PREPS 2
+ uint32_t EarlyDepthStencilControl;
+#define Normal 0
+#define ForceOff 1
+#define ForceON 2
+ uint32_t ForceThreadDispatchEnable;
+#define INTERP_PIXEL 0
+#define INTERP_CENTROID 2
+#define INTERP_SAMPLE 3
+ uint32_t PositionZWInterpolationMode;
+ uint32_t BarycentricInterpolationMode;
+#define _05pixels 0
+#define _10pixels 1
+#define _20pixels 2
+#define _40pixels 3
+ uint32_t LineEndCapAntialiasingRegionWidth;
+#define _05pixels 0
+#define _10pixels 1
+#define _20pixels 2
+#define _40pixels 3
+ uint32_t LineAntialiasingRegionWidth;
+ bool PolygonStippleEnable;
+ bool LineStippleEnable;
+#define RASTRULE_UPPER_LEFT 0
+#define RASTRULE_UPPER_RIGHT 1
+ uint32_t PointRasterizationRule;
+#define Normal 0
+#define ForceOff 1
+#define ForceON 2
+ uint32_t ForceKillPixelEnable;
+};
+
+static inline void
+GEN8_3DSTATE_WM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_WM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->StatisticsEnable, 31, 31) |
+ __gen_field(values->LegacyDepthBufferClearEnable, 30, 30) |
+ __gen_field(values->LegacyDepthBufferResolveEnable, 28, 28) |
+ __gen_field(values->LegacyHierarchicalDepthBufferResolveEnable, 27, 27) |
+ __gen_field(values->LegacyDiamondLineRasterization, 26, 26) |
+ __gen_field(values->EarlyDepthStencilControl, 21, 22) |
+ __gen_field(values->ForceThreadDispatchEnable, 19, 20) |
+ __gen_field(values->PositionZWInterpolationMode, 17, 18) |
+ __gen_field(values->BarycentricInterpolationMode, 11, 16) |
+ __gen_field(values->LineEndCapAntialiasingRegionWidth, 8, 9) |
+ __gen_field(values->LineAntialiasingRegionWidth, 6, 7) |
+ __gen_field(values->PolygonStippleEnable, 4, 4) |
+ __gen_field(values->LineStippleEnable, 3, 3) |
+ __gen_field(values->PointRasterizationRule, 2, 2) |
+ __gen_field(values->ForceKillPixelEnable, 0, 1) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_WM_CHROMAKEY_length_bias 0x00000002
+#define GEN8_3DSTATE_WM_CHROMAKEY_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 76, \
+ .DwordLength = 0
+
+#define GEN8_3DSTATE_WM_CHROMAKEY_length 0x00000002
+
+struct GEN8_3DSTATE_WM_CHROMAKEY {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool ChromaKeyKillEnable;
+};
+
+static inline void
+GEN8_3DSTATE_WM_CHROMAKEY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_WM_CHROMAKEY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ChromaKeyKillEnable, 31, 31) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_WM_DEPTH_STENCIL_length_bias 0x00000002
+#define GEN8_3DSTATE_WM_DEPTH_STENCIL_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 78, \
+ .DwordLength = 1
+
+#define GEN8_3DSTATE_WM_DEPTH_STENCIL_length 0x00000003
+
+struct GEN8_3DSTATE_WM_DEPTH_STENCIL {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t StencilFailOp;
+ uint32_t StencilPassDepthFailOp;
+ uint32_t StencilPassDepthPassOp;
+ uint32_t BackfaceStencilTestFunction;
+ uint32_t BackfaceStencilFailOp;
+ uint32_t BackfaceStencilPassDepthFailOp;
+ uint32_t BackfaceStencilPassDepthPassOp;
+ uint32_t StencilTestFunction;
+ uint32_t DepthTestFunction;
+ bool DoubleSidedStencilEnable;
+ bool StencilTestEnable;
+ bool StencilBufferWriteEnable;
+ bool DepthTestEnable;
+ bool DepthBufferWriteEnable;
+ uint32_t StencilTestMask;
+ uint32_t StencilWriteMask;
+ uint32_t BackfaceStencilTestMask;
+ uint32_t BackfaceStencilWriteMask;
+};
+
+static inline void
+GEN8_3DSTATE_WM_DEPTH_STENCIL_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_WM_DEPTH_STENCIL * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->StencilFailOp, 29, 31) |
+ __gen_field(values->StencilPassDepthFailOp, 26, 28) |
+ __gen_field(values->StencilPassDepthPassOp, 23, 25) |
+ __gen_field(values->BackfaceStencilTestFunction, 20, 22) |
+ __gen_field(values->BackfaceStencilFailOp, 17, 19) |
+ __gen_field(values->BackfaceStencilPassDepthFailOp, 14, 16) |
+ __gen_field(values->BackfaceStencilPassDepthPassOp, 11, 13) |
+ __gen_field(values->StencilTestFunction, 8, 10) |
+ __gen_field(values->DepthTestFunction, 5, 7) |
+ __gen_field(values->DoubleSidedStencilEnable, 4, 4) |
+ __gen_field(values->StencilTestEnable, 3, 3) |
+ __gen_field(values->StencilBufferWriteEnable, 2, 2) |
+ __gen_field(values->DepthTestEnable, 1, 1) |
+ __gen_field(values->DepthBufferWriteEnable, 0, 0) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->StencilTestMask, 24, 31) |
+ __gen_field(values->StencilWriteMask, 16, 23) |
+ __gen_field(values->BackfaceStencilTestMask, 8, 15) |
+ __gen_field(values->BackfaceStencilWriteMask, 0, 7) |
+ 0;
+
+}
+
+#define GEN8_3DSTATE_WM_HZ_OP_length_bias 0x00000002
+#define GEN8_3DSTATE_WM_HZ_OP_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 82, \
+ .DwordLength = 3
+
+#define GEN8_3DSTATE_WM_HZ_OP_length 0x00000005
+
+struct GEN8_3DSTATE_WM_HZ_OP {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool StencilBufferClearEnable;
+ bool DepthBufferClearEnable;
+ bool ScissorRectangleEnable;
+ bool DepthBufferResolveEnable;
+ bool HierarchicalDepthBufferResolveEnable;
+ uint32_t PixelPositionOffsetEnable;
+ bool FullSurfaceDepthClear;
+ uint32_t StencilClearValue;
+ uint32_t NumberofMultisamples;
+ uint32_t ClearRectangleYMin;
+ uint32_t ClearRectangleXMin;
+ uint32_t ClearRectangleYMax;
+ uint32_t ClearRectangleXMax;
+ uint32_t SampleMask;
+};
+
+static inline void
+GEN8_3DSTATE_WM_HZ_OP_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_3DSTATE_WM_HZ_OP * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->StencilBufferClearEnable, 31, 31) |
+ __gen_field(values->DepthBufferClearEnable, 30, 30) |
+ __gen_field(values->ScissorRectangleEnable, 29, 29) |
+ __gen_field(values->DepthBufferResolveEnable, 28, 28) |
+ __gen_field(values->HierarchicalDepthBufferResolveEnable, 27, 27) |
+ __gen_field(values->PixelPositionOffsetEnable, 26, 26) |
+ __gen_field(values->FullSurfaceDepthClear, 25, 25) |
+ __gen_field(values->StencilClearValue, 16, 23) |
+ __gen_field(values->NumberofMultisamples, 13, 15) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ClearRectangleYMin, 16, 31) |
+ __gen_field(values->ClearRectangleXMin, 0, 15) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->ClearRectangleYMax, 16, 31) |
+ __gen_field(values->ClearRectangleXMax, 0, 15) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->SampleMask, 0, 15) |
+ 0;
+
+}
+
+#define GEN8_GPGPU_WALKER_length_bias 0x00000002
+#define GEN8_GPGPU_WALKER_header \
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 1, \
+ .SubOpcode = 5, \
+ .DwordLength = 13
+
+#define GEN8_GPGPU_WALKER_length 0x0000000f
+
+struct GEN8_GPGPU_WALKER {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ bool IndirectParameterEnable;
+ bool PredicateEnable;
+ uint32_t DwordLength;
+ uint32_t InterfaceDescriptorOffset;
+ uint32_t IndirectDataLength;
+ uint32_t IndirectDataStartAddress;
+#define SIMD8 0
+#define SIMD16 1
+#define SIMD32 2
+ uint32_t SIMDSize;
+ uint32_t ThreadDepthCounterMaximum;
+ uint32_t ThreadHeightCounterMaximum;
+ uint32_t ThreadWidthCounterMaximum;
+ uint32_t ThreadGroupIDStartingX;
+ uint32_t ThreadGroupIDXDimension;
+ uint32_t ThreadGroupIDStartingY;
+ uint32_t ThreadGroupIDYDimension;
+ uint32_t ThreadGroupIDStartingResumeZ;
+ uint32_t ThreadGroupIDZDimension;
+ uint32_t RightExecutionMask;
+ uint32_t BottomExecutionMask;
+};
+
+static inline void
+GEN8_GPGPU_WALKER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_GPGPU_WALKER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->IndirectParameterEnable, 10, 10) |
+ __gen_field(values->PredicateEnable, 8, 8) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->InterfaceDescriptorOffset, 0, 5) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->IndirectDataLength, 0, 16) |
+ 0;
+
+ dw[3] =
+ __gen_offset(values->IndirectDataStartAddress, 6, 31) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->SIMDSize, 30, 31) |
+ __gen_field(values->ThreadDepthCounterMaximum, 16, 21) |
+ __gen_field(values->ThreadHeightCounterMaximum, 8, 13) |
+ __gen_field(values->ThreadWidthCounterMaximum, 0, 5) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->ThreadGroupIDStartingX, 0, 31) |
+ 0;
+
+ dw[6] =
+ 0;
+
+ dw[7] =
+ __gen_field(values->ThreadGroupIDXDimension, 0, 31) |
+ 0;
+
+ dw[8] =
+ __gen_field(values->ThreadGroupIDStartingY, 0, 31) |
+ 0;
+
+ dw[9] =
+ 0;
+
+ dw[10] =
+ __gen_field(values->ThreadGroupIDYDimension, 0, 31) |
+ 0;
+
+ dw[11] =
+ __gen_field(values->ThreadGroupIDStartingResumeZ, 0, 31) |
+ 0;
+
+ dw[12] =
+ __gen_field(values->ThreadGroupIDZDimension, 0, 31) |
+ 0;
+
+ dw[13] =
+ __gen_field(values->RightExecutionMask, 0, 31) |
+ 0;
+
+ dw[14] =
+ __gen_field(values->BottomExecutionMask, 0, 31) |
+ 0;
+
+}
+
+#define GEN8_MEDIA_CURBE_LOAD_length_bias 0x00000002
+#define GEN8_MEDIA_CURBE_LOAD_header \
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 0, \
+ .SubOpcode = 1, \
+ .DwordLength = 2
+
+#define GEN8_MEDIA_CURBE_LOAD_length 0x00000004
+
+struct GEN8_MEDIA_CURBE_LOAD {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ uint32_t DwordLength;
+ uint32_t CURBETotalDataLength;
+ uint32_t CURBEDataStartAddress;
+};
+
+static inline void
+GEN8_MEDIA_CURBE_LOAD_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MEDIA_CURBE_LOAD * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ 0;
+
+ dw[2] =
+ __gen_field(values->CURBETotalDataLength, 0, 16) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->CURBEDataStartAddress, 0, 31) |
+ 0;
+
+}
+
+#define GEN8_MEDIA_INTERFACE_DESCRIPTOR_LOAD_length_bias 0x00000002
+#define GEN8_MEDIA_INTERFACE_DESCRIPTOR_LOAD_header\
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 0, \
+ .SubOpcode = 2, \
+ .DwordLength = 2
+
+#define GEN8_MEDIA_INTERFACE_DESCRIPTOR_LOAD_length 0x00000004
+
+struct GEN8_MEDIA_INTERFACE_DESCRIPTOR_LOAD {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ uint32_t DwordLength;
+ uint32_t InterfaceDescriptorTotalLength;
+ uint32_t InterfaceDescriptorDataStartAddress;
+};
+
+static inline void
+GEN8_MEDIA_INTERFACE_DESCRIPTOR_LOAD_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MEDIA_INTERFACE_DESCRIPTOR_LOAD * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ 0;
+
+ dw[2] =
+ __gen_field(values->InterfaceDescriptorTotalLength, 0, 16) |
+ 0;
+
+ dw[3] =
+ __gen_offset(values->InterfaceDescriptorDataStartAddress, 0, 31) |
+ 0;
+
+}
+
+#define GEN8_MEDIA_OBJECT_length_bias 0x00000002
+#define GEN8_MEDIA_OBJECT_header \
+ .CommandType = 3, \
+ .MediaCommandPipeline = 2, \
+ .MediaCommandOpcode = 1, \
+ .MediaCommandSubOpcode = 0
+
+#define GEN8_MEDIA_OBJECT_length 0x00000000
+
+struct GEN8_MEDIA_OBJECT {
+ uint32_t CommandType;
+ uint32_t MediaCommandPipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t MediaCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t InterfaceDescriptorOffset;
+ bool ChildrenPresent;
+#define Nothreadsynchronization 0
+#define Threaddispatchissynchronizedbythespawnrootthreadmessage 1
+ uint32_t ThreadSynchronization;
+ uint32_t ForceDestination;
+#define Notusingscoreboard 0
+#define Usingscoreboard 1
+ uint32_t UseScoreboard;
+#define Slice0 0
+#define Slice1 1
+#define Slice2 2
+ uint32_t SliceDestinationSelect;
+#define SubSlice2 2
+#define SubSlice1 1
+#define SubSlice0 0
+ uint32_t SubSliceDestinationSelect;
+ uint32_t IndirectDataLength;
+ __gen_address_type IndirectDataStartAddress;
+ uint32_t ScoredboardY;
+ uint32_t ScoreboardX;
+ uint32_t ScoreboardColor;
+ bool ScoreboardMask;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN8_MEDIA_OBJECT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MEDIA_OBJECT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MediaCommandPipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->MediaCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->InterfaceDescriptorOffset, 0, 5) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ChildrenPresent, 31, 31) |
+ __gen_field(values->ThreadSynchronization, 24, 24) |
+ __gen_field(values->ForceDestination, 22, 22) |
+ __gen_field(values->UseScoreboard, 21, 21) |
+ __gen_field(values->SliceDestinationSelect, 19, 20) |
+ __gen_field(values->SubSliceDestinationSelect, 17, 18) |
+ __gen_field(values->IndirectDataLength, 0, 16) |
+ 0;
+
+ uint32_t dw3 =
+ 0;
+
+ dw[3] =
+ __gen_combine_address(data, &dw[3], values->IndirectDataStartAddress, dw3);
+
+ dw[4] =
+ __gen_field(values->ScoredboardY, 16, 24) |
+ __gen_field(values->ScoreboardX, 0, 8) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->ScoreboardColor, 16, 19) |
+ __gen_field(values->ScoreboardMask, 0, 7) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN8_MEDIA_OBJECT_GRPID_length_bias 0x00000002
+#define GEN8_MEDIA_OBJECT_GRPID_header \
+ .CommandType = 3, \
+ .MediaCommandPipeline = 2, \
+ .MediaCommandOpcode = 1, \
+ .MediaCommandSubOpcode = 6
+
+#define GEN8_MEDIA_OBJECT_GRPID_length 0x00000000
+
+struct GEN8_MEDIA_OBJECT_GRPID {
+ uint32_t CommandType;
+ uint32_t MediaCommandPipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t MediaCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t InterfaceDescriptorOffset;
+ uint32_t EndofThreadGroup;
+ uint32_t ForceDestination;
+#define Notusingscoreboard 0
+#define Usingscoreboard 1
+ uint32_t UseScoreboard;
+#define Slice0 0
+#define Slice1 1
+#define Slice2 2
+ uint32_t SliceDestinationSelect;
+#define SubSlice2 2
+#define SubSlice1 1
+#define SubSlice0 0
+ uint32_t SubSliceDestinationSelect;
+ uint32_t IndirectDataLength;
+ __gen_address_type IndirectDataStartAddress;
+ uint32_t ScoreboardY;
+ uint32_t ScoreboardX;
+ uint32_t ScoreboardColor;
+ bool ScoreboardMask;
+ uint32_t GroupID;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN8_MEDIA_OBJECT_GRPID_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MEDIA_OBJECT_GRPID * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MediaCommandPipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->MediaCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->InterfaceDescriptorOffset, 0, 5) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->EndofThreadGroup, 23, 23) |
+ __gen_field(values->ForceDestination, 22, 22) |
+ __gen_field(values->UseScoreboard, 21, 21) |
+ __gen_field(values->SliceDestinationSelect, 19, 20) |
+ __gen_field(values->SubSliceDestinationSelect, 17, 18) |
+ __gen_field(values->IndirectDataLength, 0, 16) |
+ 0;
+
+ uint32_t dw3 =
+ 0;
+
+ dw[3] =
+ __gen_combine_address(data, &dw[3], values->IndirectDataStartAddress, dw3);
+
+ dw[4] =
+ __gen_field(values->ScoreboardY, 16, 24) |
+ __gen_field(values->ScoreboardX, 0, 8) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->ScoreboardColor, 16, 19) |
+ __gen_field(values->ScoreboardMask, 0, 7) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->GroupID, 0, 31) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN8_MEDIA_OBJECT_PRT_length_bias 0x00000002
+#define GEN8_MEDIA_OBJECT_PRT_header \
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 1, \
+ .SubOpcode = 2, \
+ .DwordLength = 14
+
+#define GEN8_MEDIA_OBJECT_PRT_length 0x00000010
+
+struct GEN8_MEDIA_OBJECT_PRT {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ uint32_t DwordLength;
+ uint32_t InterfaceDescriptorOffset;
+ bool ChildrenPresent;
+ bool PRT_FenceNeeded;
+#define Rootthreadqueue 0
+#define VFEstateflush 1
+ uint32_t PRT_FenceType;
+ uint32_t InlineData[12];
+};
+
+static inline void
+GEN8_MEDIA_OBJECT_PRT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MEDIA_OBJECT_PRT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->InterfaceDescriptorOffset, 0, 5) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ChildrenPresent, 31, 31) |
+ __gen_field(values->PRT_FenceNeeded, 23, 23) |
+ __gen_field(values->PRT_FenceType, 22, 22) |
+ 0;
+
+ dw[3] =
+ 0;
+
+ for (uint32_t i = 0, j = 4; i < 12; i += 1, j++) {
+ dw[j] =
+ __gen_field(values->InlineData[i + 0], 0, 31) |
+ 0;
+ }
+
+}
+
+#define GEN8_MEDIA_OBJECT_WALKER_length_bias 0x00000002
+#define GEN8_MEDIA_OBJECT_WALKER_header \
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 1, \
+ .SubOpcode = 3
+
+#define GEN8_MEDIA_OBJECT_WALKER_length 0x00000000
+
+struct GEN8_MEDIA_OBJECT_WALKER {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ uint32_t DwordLength;
+ uint32_t InterfaceDescriptorOffset;
+ bool ChildrenPresent;
+#define Nothreadsynchronization 0
+#define Threaddispatchissynchronizedbythespawnrootthreadmessage 1
+ uint32_t ThreadSynchronization;
+#define Notusingscoreboard 0
+#define Usingscoreboard 1
+ uint32_t UseScoreboard;
+ uint32_t IndirectDataLength;
+ uint32_t IndirectDataStartAddress;
+ uint32_t GroupIDLoopSelect;
+ bool ScoreboardMask;
+ uint32_t ColorCountMinusOne;
+ uint32_t MiddleLoopExtraSteps;
+ uint32_t LocalMidLoopUnitY;
+ uint32_t MidLoopUnitX;
+ uint32_t GlobalLoopExecCount;
+ uint32_t LocalLoopExecCount;
+ uint32_t BlockResolutionY;
+ uint32_t BlockResolutionX;
+ uint32_t LocalStartY;
+ uint32_t LocalStartX;
+ uint32_t LocalOuterLoopStrideY;
+ uint32_t LocalOuterLoopStrideX;
+ uint32_t LocalInnerLoopUnitY;
+ uint32_t LocalInnerLoopUnitX;
+ uint32_t GlobalResolutionY;
+ uint32_t GlobalResolutionX;
+ uint32_t GlobalStartY;
+ uint32_t GlobalStartX;
+ uint32_t GlobalOuterLoopStrideY;
+ uint32_t GlobalOuterLoopStrideX;
+ uint32_t GlobalInnerLoopUnitY;
+ uint32_t GlobalInnerLoopUnitX;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN8_MEDIA_OBJECT_WALKER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MEDIA_OBJECT_WALKER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->InterfaceDescriptorOffset, 0, 5) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ChildrenPresent, 31, 31) |
+ __gen_field(values->ThreadSynchronization, 24, 24) |
+ __gen_field(values->UseScoreboard, 21, 21) |
+ __gen_field(values->IndirectDataLength, 0, 16) |
+ 0;
+
+ dw[3] =
+ __gen_offset(values->IndirectDataStartAddress, 0, 31) |
+ 0;
+
+ dw[4] =
+ 0;
+
+ dw[5] =
+ __gen_field(values->GroupIDLoopSelect, 8, 31) |
+ __gen_field(values->ScoreboardMask, 0, 7) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->ColorCountMinusOne, 24, 27) |
+ __gen_field(values->MiddleLoopExtraSteps, 16, 20) |
+ __gen_field(values->LocalMidLoopUnitY, 12, 13) |
+ __gen_field(values->MidLoopUnitX, 8, 9) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->GlobalLoopExecCount, 16, 25) |
+ __gen_field(values->LocalLoopExecCount, 0, 9) |
+ 0;
+
+ dw[8] =
+ __gen_field(values->BlockResolutionY, 16, 24) |
+ __gen_field(values->BlockResolutionX, 0, 8) |
+ 0;
+
+ dw[9] =
+ __gen_field(values->LocalStartY, 16, 24) |
+ __gen_field(values->LocalStartX, 0, 8) |
+ 0;
+
+ dw[10] =
+ 0;
+
+ dw[11] =
+ __gen_field(values->LocalOuterLoopStrideY, 16, 25) |
+ __gen_field(values->LocalOuterLoopStrideX, 0, 9) |
+ 0;
+
+ dw[12] =
+ __gen_field(values->LocalInnerLoopUnitY, 16, 25) |
+ __gen_field(values->LocalInnerLoopUnitX, 0, 9) |
+ 0;
+
+ dw[13] =
+ __gen_field(values->GlobalResolutionY, 16, 24) |
+ __gen_field(values->GlobalResolutionX, 0, 8) |
+ 0;
+
+ dw[14] =
+ __gen_field(values->GlobalStartY, 16, 25) |
+ __gen_field(values->GlobalStartX, 0, 9) |
+ 0;
+
+ dw[15] =
+ __gen_field(values->GlobalOuterLoopStrideY, 16, 25) |
+ __gen_field(values->GlobalOuterLoopStrideX, 0, 9) |
+ 0;
+
+ dw[16] =
+ __gen_field(values->GlobalInnerLoopUnitY, 16, 25) |
+ __gen_field(values->GlobalInnerLoopUnitX, 0, 9) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN8_MEDIA_STATE_FLUSH_length_bias 0x00000002
+#define GEN8_MEDIA_STATE_FLUSH_header \
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 0, \
+ .SubOpcode = 4, \
+ .DwordLength = 0
+
+#define GEN8_MEDIA_STATE_FLUSH_length 0x00000002
+
+struct GEN8_MEDIA_STATE_FLUSH {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ uint32_t DwordLength;
+ bool FlushtoGO;
+ uint32_t WatermarkRequired;
+ uint32_t InterfaceDescriptorOffset;
+};
+
+static inline void
+GEN8_MEDIA_STATE_FLUSH_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MEDIA_STATE_FLUSH * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->FlushtoGO, 7, 7) |
+ __gen_field(values->WatermarkRequired, 6, 6) |
+ __gen_field(values->InterfaceDescriptorOffset, 0, 5) |
+ 0;
+
+}
+
+#define GEN8_MEDIA_VFE_STATE_length_bias 0x00000002
+#define GEN8_MEDIA_VFE_STATE_header \
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 0, \
+ .SubOpcode = 0, \
+ .DwordLength = 7
+
+#define GEN8_MEDIA_VFE_STATE_length 0x00000009
+
+struct GEN8_MEDIA_VFE_STATE {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ uint32_t DwordLength;
+ uint32_t ScratchSpaceBasePointer;
+ uint32_t StackSize;
+ uint32_t PerThreadScratchSpace;
+ uint32_t ScratchSpaceBasePointerHigh;
+ uint32_t MaximumNumberofThreads;
+ uint32_t NumberofURBEntries;
+#define Maintainingtheexistingtimestampstate 0
+#define Resettingrelativetimerandlatchingtheglobaltimestamp 1
+ uint32_t ResetGatewayTimer;
+#define MaintainingOpenGatewayForwardMsgCloseGatewayprotocollegacymode 0
+#define BypassingOpenGatewayCloseGatewayprotocol 1
+ uint32_t BypassGatewayControl;
+ uint32_t SliceDisable;
+ uint32_t URBEntryAllocationSize;
+ uint32_t CURBEAllocationSize;
+#define Scoreboarddisabled 0
+#define Scoreboardenabled 1
+ uint32_t ScoreboardEnable;
+#define StallingScoreboard 0
+#define NonStallingScoreboard 1
+ uint32_t ScoreboardType;
+ uint32_t ScoreboardMask;
+ uint32_t Scoreboard3DeltaY;
+ uint32_t Scoreboard3DeltaX;
+ uint32_t Scoreboard2DeltaY;
+ uint32_t Scoreboard2DeltaX;
+ uint32_t Scoreboard1DeltaY;
+ uint32_t Scoreboard1DeltaX;
+ uint32_t Scoreboard0DeltaY;
+ uint32_t Scoreboard0DeltaX;
+ uint32_t Scoreboard7DeltaY;
+ uint32_t Scoreboard7DeltaX;
+ uint32_t Scoreboard6DeltaY;
+ uint32_t Scoreboard6DeltaX;
+ uint32_t Scoreboard5DeltaY;
+ uint32_t Scoreboard5DeltaX;
+ uint32_t Scoreboard4DeltaY;
+ uint32_t Scoreboard4DeltaX;
+};
+
+static inline void
+GEN8_MEDIA_VFE_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MEDIA_VFE_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->ScratchSpaceBasePointer, 10, 31) |
+ __gen_field(values->StackSize, 4, 7) |
+ __gen_field(values->PerThreadScratchSpace, 0, 3) |
+ 0;
+
+ dw[2] =
+ __gen_offset(values->ScratchSpaceBasePointerHigh, 0, 15) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->MaximumNumberofThreads, 16, 31) |
+ __gen_field(values->NumberofURBEntries, 8, 15) |
+ __gen_field(values->ResetGatewayTimer, 7, 7) |
+ __gen_field(values->BypassGatewayControl, 6, 6) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->SliceDisable, 0, 1) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->URBEntryAllocationSize, 16, 31) |
+ __gen_field(values->CURBEAllocationSize, 0, 15) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->ScoreboardEnable, 31, 31) |
+ __gen_field(values->ScoreboardType, 30, 30) |
+ __gen_field(values->ScoreboardMask, 0, 7) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->Scoreboard3DeltaY, 28, 31) |
+ __gen_field(values->Scoreboard3DeltaX, 24, 27) |
+ __gen_field(values->Scoreboard2DeltaY, 20, 23) |
+ __gen_field(values->Scoreboard2DeltaX, 16, 19) |
+ __gen_field(values->Scoreboard1DeltaY, 12, 15) |
+ __gen_field(values->Scoreboard1DeltaX, 8, 11) |
+ __gen_field(values->Scoreboard0DeltaY, 4, 7) |
+ __gen_field(values->Scoreboard0DeltaX, 0, 3) |
+ 0;
+
+ dw[8] =
+ __gen_field(values->Scoreboard7DeltaY, 28, 31) |
+ __gen_field(values->Scoreboard7DeltaX, 24, 27) |
+ __gen_field(values->Scoreboard6DeltaY, 20, 23) |
+ __gen_field(values->Scoreboard6DeltaX, 16, 19) |
+ __gen_field(values->Scoreboard5DeltaY, 12, 15) |
+ __gen_field(values->Scoreboard5DeltaX, 8, 11) |
+ __gen_field(values->Scoreboard4DeltaY, 4, 7) |
+ __gen_field(values->Scoreboard4DeltaX, 0, 3) |
+ 0;
+
+}
+
+#define GEN8_MI_ARB_CHECK_length_bias 0x00000001
+#define GEN8_MI_ARB_CHECK_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 5
+
+#define GEN8_MI_ARB_CHECK_length 0x00000001
+
+struct GEN8_MI_ARB_CHECK {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+};
+
+static inline void
+GEN8_MI_ARB_CHECK_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_ARB_CHECK * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ 0;
+
+}
+
+#define GEN8_MI_BATCH_BUFFER_END_length_bias 0x00000001
+#define GEN8_MI_BATCH_BUFFER_END_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 10
+
+#define GEN8_MI_BATCH_BUFFER_END_length 0x00000001
+
+struct GEN8_MI_BATCH_BUFFER_END {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+};
+
+static inline void
+GEN8_MI_BATCH_BUFFER_END_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_BATCH_BUFFER_END * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ 0;
+
+}
+
+#define GEN8_MI_BATCH_BUFFER_START_length_bias 0x00000002
+#define GEN8_MI_BATCH_BUFFER_START_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 49, \
+ .DwordLength = 1
+
+#define GEN8_MI_BATCH_BUFFER_START_length 0x00000003
+
+struct GEN8_MI_BATCH_BUFFER_START {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define _1stlevelbatch 0
+#define _2ndlevelbatch 1
+ uint32_t _2ndLevelBatchBuffer;
+ bool AddOffsetEnable;
+ uint32_t PredicationEnable;
+ bool ResourceStreamerEnable;
+#define ASI_GGTT 0
+#define ASI_PPGTT 1
+ uint32_t AddressSpaceIndicator;
+ uint32_t DwordLength;
+ __gen_address_type BatchBufferStartAddress;
+};
+
+static inline void
+GEN8_MI_BATCH_BUFFER_START_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_BATCH_BUFFER_START * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->_2ndLevelBatchBuffer, 22, 22) |
+ __gen_field(values->AddOffsetEnable, 16, 16) |
+ __gen_field(values->PredicationEnable, 15, 15) |
+ __gen_field(values->ResourceStreamerEnable, 10, 10) |
+ __gen_field(values->AddressSpaceIndicator, 8, 8) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw1 =
+ 0;
+
+ uint64_t qw1 =
+ __gen_combine_address(data, &dw[1], values->BatchBufferStartAddress, dw1);
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+}
+
+#define GEN8_MI_CLFLUSH_length_bias 0x00000002
+#define GEN8_MI_CLFLUSH_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 39
+
+#define GEN8_MI_CLFLUSH_length 0x00000000
+
+struct GEN8_MI_CLFLUSH {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define PerProcessGraphicsAddress 0
+#define GlobalGraphicsAddress 1
+ uint32_t UseGlobalGTT;
+ uint32_t DwordLength;
+ __gen_address_type PageBaseAddress;
+ uint32_t StartingCachelineOffset;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN8_MI_CLFLUSH_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_CLFLUSH * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->UseGlobalGTT, 22, 22) |
+ __gen_field(values->DwordLength, 0, 9) |
+ 0;
+
+ uint32_t dw1 =
+ __gen_field(values->StartingCachelineOffset, 6, 11) |
+ 0;
+
+ uint64_t qw1 =
+ __gen_combine_address(data, &dw[1], values->PageBaseAddress, dw1);
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+ /* variable length fields follow */
+}
+
+#define GEN8_MI_CONDITIONAL_BATCH_BUFFER_END_length_bias 0x00000002
+#define GEN8_MI_CONDITIONAL_BATCH_BUFFER_END_header\
+ .CommandType = 0, \
+ .MICommandOpcode = 54, \
+ .UseGlobalGTT = 0, \
+ .CompareSemaphore = 0, \
+ .DwordLength = 1
+
+#define GEN8_MI_CONDITIONAL_BATCH_BUFFER_END_length 0x00000003
+
+struct GEN8_MI_CONDITIONAL_BATCH_BUFFER_END {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t UseGlobalGTT;
+ uint32_t CompareSemaphore;
+ uint32_t DwordLength;
+ uint32_t CompareDataDword;
+ __gen_address_type CompareAddress;
+};
+
+static inline void
+GEN8_MI_CONDITIONAL_BATCH_BUFFER_END_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_CONDITIONAL_BATCH_BUFFER_END * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->UseGlobalGTT, 22, 22) |
+ __gen_field(values->CompareSemaphore, 21, 21) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->CompareDataDword, 0, 31) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ uint64_t qw2 =
+ __gen_combine_address(data, &dw[2], values->CompareAddress, dw2);
+
+ dw[2] = qw2;
+ dw[3] = qw2 >> 32;
+
+}
+
+#define GEN8_MI_COPY_MEM_MEM_length_bias 0x00000002
+#define GEN8_MI_COPY_MEM_MEM_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 46, \
+ .DwordLength = 3
+
+#define GEN8_MI_COPY_MEM_MEM_length 0x00000005
+
+struct GEN8_MI_COPY_MEM_MEM {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define PerProcessGraphicsAddress 0
+#define GlobalGraphicsAddress 1
+ uint32_t UseGlobalGTTSource;
+#define PerProcessGraphicsAddress 0
+#define GlobalGraphicsAddress 1
+ uint32_t UseGlobalGTTDestination;
+ uint32_t DwordLength;
+ __gen_address_type DestinationMemoryAddress;
+ __gen_address_type SourceMemoryAddress;
+};
+
+static inline void
+GEN8_MI_COPY_MEM_MEM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_COPY_MEM_MEM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->UseGlobalGTTSource, 22, 22) |
+ __gen_field(values->UseGlobalGTTDestination, 21, 21) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw1 =
+ 0;
+
+ uint64_t qw1 =
+ __gen_combine_address(data, &dw[1], values->DestinationMemoryAddress, dw1);
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+ uint32_t dw3 =
+ 0;
+
+ uint64_t qw3 =
+ __gen_combine_address(data, &dw[3], values->SourceMemoryAddress, dw3);
+
+ dw[3] = qw3;
+ dw[4] = qw3 >> 32;
+
+}
+
+#define GEN8_MI_LOAD_REGISTER_IMM_length_bias 0x00000002
+#define GEN8_MI_LOAD_REGISTER_IMM_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 34, \
+ .DwordLength = 1
+
+#define GEN8_MI_LOAD_REGISTER_IMM_length 0x00000003
+
+struct GEN8_MI_LOAD_REGISTER_IMM {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t ByteWriteDisables;
+ uint32_t DwordLength;
+ uint32_t RegisterOffset;
+ uint32_t DataDWord;
+};
+
+static inline void
+GEN8_MI_LOAD_REGISTER_IMM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_LOAD_REGISTER_IMM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->ByteWriteDisables, 8, 11) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->RegisterOffset, 2, 22) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->DataDWord, 0, 31) |
+ 0;
+
+}
+
+#define GEN8_MI_LOAD_REGISTER_MEM_length_bias 0x00000002
+#define GEN8_MI_LOAD_REGISTER_MEM_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 41, \
+ .DwordLength = 2
+
+#define GEN8_MI_LOAD_REGISTER_MEM_length 0x00000004
+
+struct GEN8_MI_LOAD_REGISTER_MEM {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool UseGlobalGTT;
+ uint32_t AsyncModeEnable;
+ uint32_t DwordLength;
+ uint32_t RegisterAddress;
+ __gen_address_type MemoryAddress;
+};
+
+static inline void
+GEN8_MI_LOAD_REGISTER_MEM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_LOAD_REGISTER_MEM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->UseGlobalGTT, 22, 22) |
+ __gen_field(values->AsyncModeEnable, 21, 21) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->RegisterAddress, 2, 22) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ uint64_t qw2 =
+ __gen_combine_address(data, &dw[2], values->MemoryAddress, dw2);
+
+ dw[2] = qw2;
+ dw[3] = qw2 >> 32;
+
+}
+
+#define GEN8_MI_LOAD_SCAN_LINES_EXCL_length_bias 0x00000002
+#define GEN8_MI_LOAD_SCAN_LINES_EXCL_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 19, \
+ .DwordLength = 0
+
+#define GEN8_MI_LOAD_SCAN_LINES_EXCL_length 0x00000002
+
+struct GEN8_MI_LOAD_SCAN_LINES_EXCL {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define DisplayPlaneA 0
+#define DisplayPlaneB 1
+#define DisplayPlaneC 4
+ uint32_t DisplayPlaneSelect;
+ uint32_t DwordLength;
+ uint32_t StartScanLineNumber;
+ uint32_t EndScanLineNumber;
+};
+
+static inline void
+GEN8_MI_LOAD_SCAN_LINES_EXCL_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_LOAD_SCAN_LINES_EXCL * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DisplayPlaneSelect, 19, 21) |
+ __gen_field(values->DwordLength, 0, 5) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->StartScanLineNumber, 16, 28) |
+ __gen_field(values->EndScanLineNumber, 0, 12) |
+ 0;
+
+}
+
+#define GEN8_MI_LOAD_SCAN_LINES_INCL_length_bias 0x00000002
+#define GEN8_MI_LOAD_SCAN_LINES_INCL_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 18, \
+ .DwordLength = 0
+
+#define GEN8_MI_LOAD_SCAN_LINES_INCL_length 0x00000002
+
+struct GEN8_MI_LOAD_SCAN_LINES_INCL {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define DisplayPlaneA 0
+#define DisplayPlaneB 1
+#define DisplayPlaneC 4
+ uint32_t DisplayPlaneSelect;
+#define NeverForward 0
+#define AlwaysForward 1
+#define ConditionallyForward 2
+ bool ScanLineEventDoneForward;
+ uint32_t DwordLength;
+ uint32_t StartScanLineNumber;
+ uint32_t EndScanLineNumber;
+};
+
+static inline void
+GEN8_MI_LOAD_SCAN_LINES_INCL_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_LOAD_SCAN_LINES_INCL * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DisplayPlaneSelect, 19, 21) |
+ __gen_field(values->ScanLineEventDoneForward, 17, 18) |
+ __gen_field(values->DwordLength, 0, 5) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->StartScanLineNumber, 16, 28) |
+ __gen_field(values->EndScanLineNumber, 0, 12) |
+ 0;
+
+}
+
+#define GEN8_MI_LOAD_URB_MEM_length_bias 0x00000002
+#define GEN8_MI_LOAD_URB_MEM_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 44, \
+ .DwordLength = 2
+
+#define GEN8_MI_LOAD_URB_MEM_length 0x00000004
+
+struct GEN8_MI_LOAD_URB_MEM {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t DwordLength;
+ uint32_t URBAddress;
+ __gen_address_type MemoryAddress;
+};
+
+static inline void
+GEN8_MI_LOAD_URB_MEM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_LOAD_URB_MEM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->URBAddress, 2, 14) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ uint64_t qw2 =
+ __gen_combine_address(data, &dw[2], values->MemoryAddress, dw2);
+
+ dw[2] = qw2;
+ dw[3] = qw2 >> 32;
+
+}
+
+#define GEN8_MI_MATH_length_bias 0x00000002
+#define GEN8_MI_MATH_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 26
+
+#define GEN8_MI_MATH_length 0x00000000
+
+struct GEN8_MI_MATH {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t DwordLength;
+ uint32_t ALUINSTRUCTION1;
+ uint32_t ALUINSTRUCTION2;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN8_MI_MATH_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_MATH * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DwordLength, 0, 5) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ALUINSTRUCTION1, 0, 31) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ALUINSTRUCTION2, 0, 31) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN8_MI_NOOP_length_bias 0x00000001
+#define GEN8_MI_NOOP_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 0
+
+#define GEN8_MI_NOOP_length 0x00000001
+
+struct GEN8_MI_NOOP {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool IdentificationNumberRegisterWriteEnable;
+ uint32_t IdentificationNumber;
+};
+
+static inline void
+GEN8_MI_NOOP_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_NOOP * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->IdentificationNumberRegisterWriteEnable, 22, 22) |
+ __gen_field(values->IdentificationNumber, 0, 21) |
+ 0;
+
+}
+
+#define GEN8_MI_PREDICATE_length_bias 0x00000001
+#define GEN8_MI_PREDICATE_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 12
+
+#define GEN8_MI_PREDICATE_length 0x00000001
+
+struct GEN8_MI_PREDICATE {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define LOAD_KEEP 0
+#define LOAD_LOAD 2
+#define LOAD_LOADINV 3
+ uint32_t LoadOperation;
+#define COMBINE_SET 0
+#define COMBINE_AND 1
+#define COMBINE_OR 2
+#define COMBINE_XOR 3
+ uint32_t CombineOperation;
+#define COMPARE_SRCS_EQUAL 2
+#define COMPARE_DELTAS_EQUAL 3
+ uint32_t CompareOperation;
+};
+
+static inline void
+GEN8_MI_PREDICATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_PREDICATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->LoadOperation, 6, 7) |
+ __gen_field(values->CombineOperation, 3, 4) |
+ __gen_field(values->CompareOperation, 0, 1) |
+ 0;
+
+}
+
+#define GEN8_MI_REPORT_HEAD_length_bias 0x00000001
+#define GEN8_MI_REPORT_HEAD_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 7
+
+#define GEN8_MI_REPORT_HEAD_length 0x00000001
+
+struct GEN8_MI_REPORT_HEAD {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+};
+
+static inline void
+GEN8_MI_REPORT_HEAD_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_REPORT_HEAD * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ 0;
+
+}
+
+#define GEN8_MI_RS_CONTEXT_length_bias 0x00000001
+#define GEN8_MI_RS_CONTEXT_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 15
+
+#define GEN8_MI_RS_CONTEXT_length 0x00000001
+
+struct GEN8_MI_RS_CONTEXT {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define RS_RESTORE 0
+#define RS_SAVE 1
+ uint32_t ResourceStreamerSave;
+};
+
+static inline void
+GEN8_MI_RS_CONTEXT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_RS_CONTEXT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->ResourceStreamerSave, 0, 0) |
+ 0;
+
+}
+
+#define GEN8_MI_RS_CONTROL_length_bias 0x00000001
+#define GEN8_MI_RS_CONTROL_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 6
+
+#define GEN8_MI_RS_CONTROL_length 0x00000001
+
+struct GEN8_MI_RS_CONTROL {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define RS_STOP 0
+#define RS_START 1
+ uint32_t ResourceStreamerControl;
+};
+
+static inline void
+GEN8_MI_RS_CONTROL_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_RS_CONTROL * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->ResourceStreamerControl, 0, 0) |
+ 0;
+
+}
+
+#define GEN8_MI_RS_STORE_DATA_IMM_length_bias 0x00000002
+#define GEN8_MI_RS_STORE_DATA_IMM_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 43, \
+ .DwordLength = 2
+
+#define GEN8_MI_RS_STORE_DATA_IMM_length 0x00000004
+
+struct GEN8_MI_RS_STORE_DATA_IMM {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t DwordLength;
+ __gen_address_type DestinationAddress;
+ uint32_t CoreModeEnable;
+ uint32_t DataDWord0;
+};
+
+static inline void
+GEN8_MI_RS_STORE_DATA_IMM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_RS_STORE_DATA_IMM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw1 =
+ __gen_field(values->CoreModeEnable, 0, 0) |
+ 0;
+
+ uint64_t qw1 =
+ __gen_combine_address(data, &dw[1], values->DestinationAddress, dw1);
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+ dw[3] =
+ __gen_field(values->DataDWord0, 0, 31) |
+ 0;
+
+}
+
+#define GEN8_MI_SET_CONTEXT_length_bias 0x00000002
+#define GEN8_MI_SET_CONTEXT_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 24, \
+ .DwordLength = 0
+
+#define GEN8_MI_SET_CONTEXT_length 0x00000002
+
+struct GEN8_MI_SET_CONTEXT {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t DwordLength;
+ __gen_address_type LogicalContextAddress;
+ uint32_t ReservedMustbe1;
+ bool CoreModeEnable;
+ bool ResourceStreamerStateSaveEnable;
+ bool ResourceStreamerStateRestoreEnable;
+ uint32_t ForceRestore;
+ uint32_t RestoreInhibit;
+};
+
+static inline void
+GEN8_MI_SET_CONTEXT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_SET_CONTEXT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw1 =
+ __gen_field(values->ReservedMustbe1, 8, 8) |
+ __gen_field(values->CoreModeEnable, 4, 4) |
+ __gen_field(values->ResourceStreamerStateSaveEnable, 3, 3) |
+ __gen_field(values->ResourceStreamerStateRestoreEnable, 2, 2) |
+ __gen_field(values->ForceRestore, 1, 1) |
+ __gen_field(values->RestoreInhibit, 0, 0) |
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->LogicalContextAddress, dw1);
+
+}
+
+#define GEN8_MI_SET_PREDICATE_length_bias 0x00000001
+#define GEN8_MI_SET_PREDICATE_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 1
+
+#define GEN8_MI_SET_PREDICATE_length 0x00000001
+
+struct GEN8_MI_SET_PREDICATE {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define NOOPNever 0
+#define NOOPonResult2clear 1
+#define NOOPonResult2set 2
+#define NOOPonResultclear 3
+#define NOOPonResultset 4
+#define Executewhenonesliceenabled 5
+#define Executewhentwoslicesareenabled 6
+#define Executewhenthreeslicesareenabled 7
+#define NOOPAlways 15
+ uint32_t PREDICATEENABLE;
+};
+
+static inline void
+GEN8_MI_SET_PREDICATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_SET_PREDICATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->PREDICATEENABLE, 0, 3) |
+ 0;
+
+}
+
+#define GEN8_MI_STORE_DATA_IMM_length_bias 0x00000002
+#define GEN8_MI_STORE_DATA_IMM_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 32, \
+ .DwordLength = 2
+
+#define GEN8_MI_STORE_DATA_IMM_length 0x00000004
+
+struct GEN8_MI_STORE_DATA_IMM {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool UseGlobalGTT;
+ bool StoreQword;
+ uint32_t DwordLength;
+ __gen_address_type Address;
+ uint32_t CoreModeEnable;
+ uint32_t DataDWord0;
+ uint32_t DataDWord1;
+};
+
+static inline void
+GEN8_MI_STORE_DATA_IMM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_STORE_DATA_IMM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->UseGlobalGTT, 22, 22) |
+ __gen_field(values->StoreQword, 21, 21) |
+ __gen_field(values->DwordLength, 0, 9) |
+ 0;
+
+ uint32_t dw1 =
+ __gen_field(values->CoreModeEnable, 0, 0) |
+ 0;
+
+ uint64_t qw1 =
+ __gen_combine_address(data, &dw[1], values->Address, dw1);
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+ dw[3] =
+ __gen_field(values->DataDWord0, 0, 31) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->DataDWord1, 0, 31) |
+ 0;
+
+}
+
+#define GEN8_MI_STORE_DATA_INDEX_length_bias 0x00000002
+#define GEN8_MI_STORE_DATA_INDEX_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 33, \
+ .DwordLength = 1
+
+#define GEN8_MI_STORE_DATA_INDEX_length 0x00000003
+
+struct GEN8_MI_STORE_DATA_INDEX {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t UsePerProcessHardwareStatusPage;
+ uint32_t DwordLength;
+ uint32_t Offset;
+ uint32_t DataDWord0;
+ uint32_t DataDWord1;
+};
+
+static inline void
+GEN8_MI_STORE_DATA_INDEX_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_STORE_DATA_INDEX * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->UsePerProcessHardwareStatusPage, 21, 21) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->Offset, 2, 11) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->DataDWord0, 0, 31) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->DataDWord1, 0, 31) |
+ 0;
+
+}
+
+#define GEN8_MI_STORE_URB_MEM_length_bias 0x00000002
+#define GEN8_MI_STORE_URB_MEM_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 45, \
+ .DwordLength = 2
+
+#define GEN8_MI_STORE_URB_MEM_length 0x00000004
+
+struct GEN8_MI_STORE_URB_MEM {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t DwordLength;
+ uint32_t URBAddress;
+ __gen_address_type MemoryAddress;
+};
+
+static inline void
+GEN8_MI_STORE_URB_MEM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_STORE_URB_MEM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->URBAddress, 2, 14) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ uint64_t qw2 =
+ __gen_combine_address(data, &dw[2], values->MemoryAddress, dw2);
+
+ dw[2] = qw2;
+ dw[3] = qw2 >> 32;
+
+}
+
+#define GEN8_MI_SUSPEND_FLUSH_length_bias 0x00000001
+#define GEN8_MI_SUSPEND_FLUSH_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 11
+
+#define GEN8_MI_SUSPEND_FLUSH_length 0x00000001
+
+struct GEN8_MI_SUSPEND_FLUSH {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool SuspendFlush;
+};
+
+static inline void
+GEN8_MI_SUSPEND_FLUSH_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_SUSPEND_FLUSH * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->SuspendFlush, 0, 0) |
+ 0;
+
+}
+
+#define GEN8_MI_TOPOLOGY_FILTER_length_bias 0x00000001
+#define GEN8_MI_TOPOLOGY_FILTER_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 13
+
+#define GEN8_MI_TOPOLOGY_FILTER_length 0x00000001
+
+struct GEN8_MI_TOPOLOGY_FILTER {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t TopologyFilterValue;
+};
+
+static inline void
+GEN8_MI_TOPOLOGY_FILTER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_TOPOLOGY_FILTER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->TopologyFilterValue, 0, 5) |
+ 0;
+
+}
+
+#define GEN8_MI_UPDATE_GTT_length_bias 0x00000002
+#define GEN8_MI_UPDATE_GTT_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 35
+
+#define GEN8_MI_UPDATE_GTT_length 0x00000000
+
+struct GEN8_MI_UPDATE_GTT {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t DwordLength;
+ __gen_address_type EntryAddress;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN8_MI_UPDATE_GTT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_UPDATE_GTT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DwordLength, 0, 9) |
+ 0;
+
+ uint32_t dw1 =
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->EntryAddress, dw1);
+
+ /* variable length fields follow */
+}
+
+#define GEN8_MI_URB_ATOMIC_ALLOC_length_bias 0x00000001
+#define GEN8_MI_URB_ATOMIC_ALLOC_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 9
+
+#define GEN8_MI_URB_ATOMIC_ALLOC_length 0x00000001
+
+struct GEN8_MI_URB_ATOMIC_ALLOC {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t URBAtomicStorageOffset;
+ uint32_t URBAtomicStorageSize;
+};
+
+static inline void
+GEN8_MI_URB_ATOMIC_ALLOC_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_URB_ATOMIC_ALLOC * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->URBAtomicStorageOffset, 12, 19) |
+ __gen_field(values->URBAtomicStorageSize, 0, 8) |
+ 0;
+
+}
+
+#define GEN8_MI_URB_CLEAR_length_bias 0x00000002
+#define GEN8_MI_URB_CLEAR_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 25, \
+ .DwordLength = 0
+
+#define GEN8_MI_URB_CLEAR_length 0x00000002
+
+struct GEN8_MI_URB_CLEAR {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t DwordLength;
+ uint32_t URBClearLength;
+ uint32_t URBAddress;
+};
+
+static inline void
+GEN8_MI_URB_CLEAR_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_URB_CLEAR * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->URBClearLength, 16, 29) |
+ __gen_offset(values->URBAddress, 0, 14) |
+ 0;
+
+}
+
+#define GEN8_MI_USER_INTERRUPT_length_bias 0x00000001
+#define GEN8_MI_USER_INTERRUPT_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 2
+
+#define GEN8_MI_USER_INTERRUPT_length 0x00000001
+
+struct GEN8_MI_USER_INTERRUPT {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+};
+
+static inline void
+GEN8_MI_USER_INTERRUPT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_USER_INTERRUPT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ 0;
+
+}
+
+#define GEN8_MI_WAIT_FOR_EVENT_length_bias 0x00000001
+#define GEN8_MI_WAIT_FOR_EVENT_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 3
+
+#define GEN8_MI_WAIT_FOR_EVENT_length 0x00000001
+
+struct GEN8_MI_WAIT_FOR_EVENT {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool DisplayPipeCVerticalBlankWaitEnable;
+ bool DisplaySpriteCFlipPendingWaitEnable;
+ bool DisplayPlaneCFlipPendingWaitEnable;
+ bool DisplayPipeCScanLineWaitEnable;
+ bool DisplayPipeBVerticalBlankWaitEnable;
+ bool DisplaySpriteBFlipPendingWaitEnable;
+ bool DisplayPlaneBFlipPendingWaitEnable;
+ bool DisplayPipeBScanLineWaitEnable;
+ bool DisplayPipeAVerticalBlankWaitEnable;
+ bool DisplaySpriteAFlipPendingWaitEnable;
+ bool DisplayPlaneAFlipPendingWaitEnable;
+ bool DisplayPipeAScanLineWaitEnable;
+};
+
+static inline void
+GEN8_MI_WAIT_FOR_EVENT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_MI_WAIT_FOR_EVENT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DisplayPipeCVerticalBlankWaitEnable, 21, 21) |
+ __gen_field(values->DisplaySpriteCFlipPendingWaitEnable, 20, 20) |
+ __gen_field(values->DisplayPlaneCFlipPendingWaitEnable, 15, 15) |
+ __gen_field(values->DisplayPipeCScanLineWaitEnable, 14, 14) |
+ __gen_field(values->DisplayPipeBVerticalBlankWaitEnable, 11, 11) |
+ __gen_field(values->DisplaySpriteBFlipPendingWaitEnable, 10, 10) |
+ __gen_field(values->DisplayPlaneBFlipPendingWaitEnable, 9, 9) |
+ __gen_field(values->DisplayPipeBScanLineWaitEnable, 8, 8) |
+ __gen_field(values->DisplayPipeAVerticalBlankWaitEnable, 3, 3) |
+ __gen_field(values->DisplaySpriteAFlipPendingWaitEnable, 2, 2) |
+ __gen_field(values->DisplayPlaneAFlipPendingWaitEnable, 1, 1) |
+ __gen_field(values->DisplayPipeAScanLineWaitEnable, 0, 0) |
+ 0;
+
+}
+
+#define GEN8_PIPE_CONTROL_length_bias 0x00000002
+#define GEN8_PIPE_CONTROL_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 2, \
+ ._3DCommandSubOpcode = 0, \
+ .DwordLength = 4
+
+#define GEN8_PIPE_CONTROL_length 0x00000006
+
+struct GEN8_PIPE_CONTROL {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define DAT_PPGTT 0
+#define DAT_GGTT 1
+ uint32_t DestinationAddressType;
+#define NoLRIOperation 0
+#define MMIOWriteImmediateData 1
+ uint32_t LRIPostSyncOperation;
+ uint32_t StoreDataIndex;
+ uint32_t CommandStreamerStallEnable;
+#define DontReset 0
+#define Reset 1
+ uint32_t GlobalSnapshotCountReset;
+ uint32_t TLBInvalidate;
+ bool GenericMediaStateClear;
+#define NoWrite 0
+#define WriteImmediateData 1
+#define WritePSDepthCount 2
+#define WriteTimestamp 3
+ uint32_t PostSyncOperation;
+ bool DepthStallEnable;
+#define DisableFlush 0
+#define EnableFlush 1
+ bool RenderTargetCacheFlushEnable;
+ bool InstructionCacheInvalidateEnable;
+ bool TextureCacheInvalidationEnable;
+ bool IndirectStatePointersDisable;
+ bool NotifyEnable;
+ bool PipeControlFlushEnable;
+ bool DCFlushEnable;
+ bool VFCacheInvalidationEnable;
+ bool ConstantCacheInvalidationEnable;
+ bool StateCacheInvalidationEnable;
+ bool StallAtPixelScoreboard;
+#define FlushDisabled 0
+#define FlushEnabled 1
+ bool DepthCacheFlushEnable;
+ __gen_address_type Address;
+ uint64_t ImmediateData;
+};
+
+static inline void
+GEN8_PIPE_CONTROL_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_PIPE_CONTROL * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->DestinationAddressType, 24, 24) |
+ __gen_field(values->LRIPostSyncOperation, 23, 23) |
+ __gen_field(values->StoreDataIndex, 21, 21) |
+ __gen_field(values->CommandStreamerStallEnable, 20, 20) |
+ __gen_field(values->GlobalSnapshotCountReset, 19, 19) |
+ __gen_field(values->TLBInvalidate, 18, 18) |
+ __gen_field(values->GenericMediaStateClear, 16, 16) |
+ __gen_field(values->PostSyncOperation, 14, 15) |
+ __gen_field(values->DepthStallEnable, 13, 13) |
+ __gen_field(values->RenderTargetCacheFlushEnable, 12, 12) |
+ __gen_field(values->InstructionCacheInvalidateEnable, 11, 11) |
+ __gen_field(values->TextureCacheInvalidationEnable, 10, 10) |
+ __gen_field(values->IndirectStatePointersDisable, 9, 9) |
+ __gen_field(values->NotifyEnable, 8, 8) |
+ __gen_field(values->PipeControlFlushEnable, 7, 7) |
+ __gen_field(values->DCFlushEnable, 5, 5) |
+ __gen_field(values->VFCacheInvalidationEnable, 4, 4) |
+ __gen_field(values->ConstantCacheInvalidationEnable, 3, 3) |
+ __gen_field(values->StateCacheInvalidationEnable, 2, 2) |
+ __gen_field(values->StallAtPixelScoreboard, 1, 1) |
+ __gen_field(values->DepthCacheFlushEnable, 0, 0) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ uint64_t qw2 =
+ __gen_combine_address(data, &dw[2], values->Address, dw2);
+
+ dw[2] = qw2;
+ dw[3] = qw2 >> 32;
+
+ uint64_t qw4 =
+ __gen_field(values->ImmediateData, 0, 63) |
+ 0;
+
+ dw[4] = qw4;
+ dw[5] = qw4 >> 32;
+
+}
+
+#define GEN8_SCISSOR_RECT_length 0x00000002
+
+struct GEN8_SCISSOR_RECT {
+ uint32_t ScissorRectangleYMin;
+ uint32_t ScissorRectangleXMin;
+ uint32_t ScissorRectangleYMax;
+ uint32_t ScissorRectangleXMax;
+};
+
+static inline void
+GEN8_SCISSOR_RECT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_SCISSOR_RECT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->ScissorRectangleYMin, 16, 31) |
+ __gen_field(values->ScissorRectangleXMin, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ScissorRectangleYMax, 16, 31) |
+ __gen_field(values->ScissorRectangleXMax, 0, 15) |
+ 0;
+
+}
+
+#define GEN8_SF_CLIP_VIEWPORT_length 0x00000010
+
+struct GEN8_SF_CLIP_VIEWPORT {
+ float ViewportMatrixElementm00;
+ float ViewportMatrixElementm11;
+ float ViewportMatrixElementm22;
+ float ViewportMatrixElementm30;
+ float ViewportMatrixElementm31;
+ float ViewportMatrixElementm32;
+ float XMinClipGuardband;
+ float XMaxClipGuardband;
+ float YMinClipGuardband;
+ float YMaxClipGuardband;
+ float XMinViewPort;
+ float XMaxViewPort;
+ float YMinViewPort;
+ float YMaxViewPort;
+};
+
+static inline void
+GEN8_SF_CLIP_VIEWPORT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_SF_CLIP_VIEWPORT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_float(values->ViewportMatrixElementm00) |
+ 0;
+
+ dw[1] =
+ __gen_float(values->ViewportMatrixElementm11) |
+ 0;
+
+ dw[2] =
+ __gen_float(values->ViewportMatrixElementm22) |
+ 0;
+
+ dw[3] =
+ __gen_float(values->ViewportMatrixElementm30) |
+ 0;
+
+ dw[4] =
+ __gen_float(values->ViewportMatrixElementm31) |
+ 0;
+
+ dw[5] =
+ __gen_float(values->ViewportMatrixElementm32) |
+ 0;
+
+ dw[6] =
+ 0;
+
+ dw[7] =
+ 0;
+
+ dw[8] =
+ __gen_float(values->XMinClipGuardband) |
+ 0;
+
+ dw[9] =
+ __gen_float(values->XMaxClipGuardband) |
+ 0;
+
+ dw[10] =
+ __gen_float(values->YMinClipGuardband) |
+ 0;
+
+ dw[11] =
+ __gen_float(values->YMaxClipGuardband) |
+ 0;
+
+ dw[12] =
+ __gen_float(values->XMinViewPort) |
+ 0;
+
+ dw[13] =
+ __gen_float(values->XMaxViewPort) |
+ 0;
+
+ dw[14] =
+ __gen_float(values->YMinViewPort) |
+ 0;
+
+ dw[15] =
+ __gen_float(values->YMaxViewPort) |
+ 0;
+
+}
+
+#define GEN8_BLEND_STATE_length 0x00000011
+
+#define GEN8_BLEND_STATE_ENTRY_length 0x00000002
+
+struct GEN8_BLEND_STATE_ENTRY {
+ bool LogicOpEnable;
+ uint32_t LogicOpFunction;
+ uint32_t PreBlendSourceOnlyClampEnable;
+#define COLORCLAMP_UNORM 0
+#define COLORCLAMP_SNORM 1
+#define COLORCLAMP_RTFORMAT 2
+ uint32_t ColorClampRange;
+ bool PreBlendColorClampEnable;
+ bool PostBlendColorClampEnable;
+ bool ColorBufferBlendEnable;
+ uint32_t SourceBlendFactor;
+ uint32_t DestinationBlendFactor;
+ uint32_t ColorBlendFunction;
+ uint32_t SourceAlphaBlendFactor;
+ uint32_t DestinationAlphaBlendFactor;
+ uint32_t AlphaBlendFunction;
+ bool WriteDisableAlpha;
+ bool WriteDisableRed;
+ bool WriteDisableGreen;
+ bool WriteDisableBlue;
+};
+
+static inline void
+GEN8_BLEND_STATE_ENTRY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_BLEND_STATE_ENTRY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ uint64_t qw0 =
+ __gen_field(values->LogicOpEnable, 63, 63) |
+ __gen_field(values->LogicOpFunction, 59, 62) |
+ __gen_field(values->PreBlendSourceOnlyClampEnable, 36, 36) |
+ __gen_field(values->ColorClampRange, 34, 35) |
+ __gen_field(values->PreBlendColorClampEnable, 33, 33) |
+ __gen_field(values->PostBlendColorClampEnable, 32, 32) |
+ __gen_field(values->ColorBufferBlendEnable, 31, 31) |
+ __gen_field(values->SourceBlendFactor, 26, 30) |
+ __gen_field(values->DestinationBlendFactor, 21, 25) |
+ __gen_field(values->ColorBlendFunction, 18, 20) |
+ __gen_field(values->SourceAlphaBlendFactor, 13, 17) |
+ __gen_field(values->DestinationAlphaBlendFactor, 8, 12) |
+ __gen_field(values->AlphaBlendFunction, 5, 7) |
+ __gen_field(values->WriteDisableAlpha, 3, 3) |
+ __gen_field(values->WriteDisableRed, 2, 2) |
+ __gen_field(values->WriteDisableGreen, 1, 1) |
+ __gen_field(values->WriteDisableBlue, 0, 0) |
+ 0;
+
+ dw[0] = qw0;
+ dw[1] = qw0 >> 32;
+
+}
+
+struct GEN8_BLEND_STATE {
+ bool AlphaToCoverageEnable;
+ bool IndependentAlphaBlendEnable;
+ bool AlphaToOneEnable;
+ bool AlphaToCoverageDitherEnable;
+ bool AlphaTestEnable;
+ uint32_t AlphaTestFunction;
+ bool ColorDitherEnable;
+ uint32_t XDitherOffset;
+ uint32_t YDitherOffset;
+ struct GEN8_BLEND_STATE_ENTRY Entry[8];
+};
+
+static inline void
+GEN8_BLEND_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_BLEND_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->AlphaToCoverageEnable, 31, 31) |
+ __gen_field(values->IndependentAlphaBlendEnable, 30, 30) |
+ __gen_field(values->AlphaToOneEnable, 29, 29) |
+ __gen_field(values->AlphaToCoverageDitherEnable, 28, 28) |
+ __gen_field(values->AlphaTestEnable, 27, 27) |
+ __gen_field(values->AlphaTestFunction, 24, 26) |
+ __gen_field(values->ColorDitherEnable, 23, 23) |
+ __gen_field(values->XDitherOffset, 21, 22) |
+ __gen_field(values->YDitherOffset, 19, 20) |
+ 0;
+
+ for (uint32_t i = 0, j = 1; i < 8; i++, j += 2)
+ GEN8_BLEND_STATE_ENTRY_pack(data, &dw[j], &values->Entry[i]);
+}
+
+#define GEN8_CC_VIEWPORT_length 0x00000002
+
+struct GEN8_CC_VIEWPORT {
+ float MinimumDepth;
+ float MaximumDepth;
+};
+
+static inline void
+GEN8_CC_VIEWPORT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_CC_VIEWPORT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_float(values->MinimumDepth) |
+ 0;
+
+ dw[1] =
+ __gen_float(values->MaximumDepth) |
+ 0;
+
+}
+
+#define GEN8_COLOR_CALC_STATE_length 0x00000006
+
+struct GEN8_COLOR_CALC_STATE {
+ uint32_t StencilReferenceValue;
+ uint32_t BackFaceStencilReferenceValue;
+#define Cancelled 0
+#define NotCancelled 1
+ uint32_t RoundDisableFunctionDisable;
+#define ALPHATEST_UNORM8 0
+#define ALPHATEST_FLOAT32 1
+ uint32_t AlphaTestFormat;
+ uint32_t AlphaReferenceValueAsUNORM8;
+ float AlphaReferenceValueAsFLOAT32;
+ float BlendConstantColorRed;
+ float BlendConstantColorGreen;
+ float BlendConstantColorBlue;
+ float BlendConstantColorAlpha;
+};
+
+static inline void
+GEN8_COLOR_CALC_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_COLOR_CALC_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->StencilReferenceValue, 24, 31) |
+ __gen_field(values->BackFaceStencilReferenceValue, 16, 23) |
+ __gen_field(values->RoundDisableFunctionDisable, 15, 15) |
+ __gen_field(values->AlphaTestFormat, 0, 0) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->AlphaReferenceValueAsUNORM8, 0, 31) |
+ __gen_float(values->AlphaReferenceValueAsFLOAT32) |
+ 0;
+
+ dw[2] =
+ __gen_float(values->BlendConstantColorRed) |
+ 0;
+
+ dw[3] =
+ __gen_float(values->BlendConstantColorGreen) |
+ 0;
+
+ dw[4] =
+ __gen_float(values->BlendConstantColorBlue) |
+ 0;
+
+ dw[5] =
+ __gen_float(values->BlendConstantColorAlpha) |
+ 0;
+
+}
+
+#define GEN8_BLACK_LEVEL_CORRECTION_STATE__DW7576_length 0x00000002
+
+struct GEN8_BLACK_LEVEL_CORRECTION_STATE__DW7576 {
+ uint32_t BlackPointOffsetR;
+ uint32_t BlackPointOffsetG;
+ uint32_t BlackPointOffsetB;
+};
+
+static inline void
+GEN8_BLACK_LEVEL_CORRECTION_STATE__DW7576_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_BLACK_LEVEL_CORRECTION_STATE__DW7576 * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->BlackPointOffsetR, 0, 12) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->BlackPointOffsetG, 13, 25) |
+ __gen_field(values->BlackPointOffsetB, 0, 12) |
+ 0;
+
+}
+
+#define GEN8_INTERFACE_DESCRIPTOR_DATA_length 0x00000008
+
+struct GEN8_INTERFACE_DESCRIPTOR_DATA {
+ uint32_t KernelStartPointer;
+ uint32_t KernelStartPointerHigh;
+#define Ftz 0
+#define SetByKernel 1
+ uint32_t DenormMode;
+#define Multiple 0
+#define Single 1
+ uint32_t SingleProgramFlow;
+#define NormalPriority 0
+#define HighPriority 1
+ uint32_t ThreadPriority;
+#define IEEE754 0
+#define Alternate 1
+ uint32_t FloatingPointMode;
+ bool IllegalOpcodeExceptionEnable;
+ bool MaskStackExceptionEnable;
+ bool SoftwareExceptionEnable;
+ uint32_t SamplerStatePointer;
+#define Nosamplersused 0
+#define Between1and4samplersused 1
+#define Between5and8samplersused 2
+#define Between9and12samplersused 3
+#define Between13and16samplersused 4
+ uint32_t SamplerCount;
+ uint32_t BindingTablePointer;
+ uint32_t BindingTableEntryCount;
+ uint32_t ConstantIndirectURBEntryReadLength;
+ uint32_t ConstantURBEntryReadOffset;
+#define RTNE 0
+#define RU 1
+#define RD 2
+#define RTZ 3
+ uint32_t RoundingMode;
+ bool BarrierEnable;
+#define Encodes0k 0
+#define Encodes4k 1
+#define Encodes8k 2
+#define Encodes16k 4
+#define Encodes32k 8
+#define Encodes64k 16
+ uint32_t SharedLocalMemorySize;
+ uint32_t NumberofThreadsinGPGPUThreadGroup;
+ uint32_t CrossThreadConstantDataReadLength;
+};
+
+static inline void
+GEN8_INTERFACE_DESCRIPTOR_DATA_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_INTERFACE_DESCRIPTOR_DATA * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_offset(values->KernelStartPointer, 6, 31) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->KernelStartPointerHigh, 0, 15) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->DenormMode, 19, 19) |
+ __gen_field(values->SingleProgramFlow, 18, 18) |
+ __gen_field(values->ThreadPriority, 17, 17) |
+ __gen_field(values->FloatingPointMode, 16, 16) |
+ __gen_field(values->IllegalOpcodeExceptionEnable, 13, 13) |
+ __gen_field(values->MaskStackExceptionEnable, 11, 11) |
+ __gen_field(values->SoftwareExceptionEnable, 7, 7) |
+ 0;
+
+ dw[3] =
+ __gen_offset(values->SamplerStatePointer, 5, 31) |
+ __gen_field(values->SamplerCount, 2, 4) |
+ 0;
+
+ dw[4] =
+ __gen_offset(values->BindingTablePointer, 5, 15) |
+ __gen_field(values->BindingTableEntryCount, 0, 4) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->ConstantIndirectURBEntryReadLength, 16, 31) |
+ __gen_field(values->ConstantURBEntryReadOffset, 0, 15) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->RoundingMode, 22, 23) |
+ __gen_field(values->BarrierEnable, 21, 21) |
+ __gen_field(values->SharedLocalMemorySize, 16, 20) |
+ __gen_field(values->NumberofThreadsinGPGPUThreadGroup, 0, 9) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->CrossThreadConstantDataReadLength, 0, 7) |
+ 0;
+
+}
+
+#define GEN8_BINDING_TABLE_STATE_length 0x00000001
+
+struct GEN8_BINDING_TABLE_STATE {
+ uint32_t SurfaceStatePointer;
+};
+
+static inline void
+GEN8_BINDING_TABLE_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_BINDING_TABLE_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_offset(values->SurfaceStatePointer, 6, 31) |
+ 0;
+
+}
+
+#define GEN8_RENDER_SURFACE_STATE_length 0x00000010
+
+struct GEN8_RENDER_SURFACE_STATE {
+#define SURFTYPE_1D 0
+#define SURFTYPE_2D 1
+#define SURFTYPE_3D 2
+#define SURFTYPE_CUBE 3
+#define SURFTYPE_BUFFER 4
+#define SURFTYPE_STRBUF 5
+#define SURFTYPE_NULL 7
+ uint32_t SurfaceType;
+ bool SurfaceArray;
+ uint32_t SurfaceFormat;
+#define VALIGN4 1
+#define VALIGN8 2
+#define VALIGN16 3
+ uint32_t SurfaceVerticalAlignment;
+#define HALIGN4 1
+#define HALIGN8 2
+#define HALIGN16 3
+ uint32_t SurfaceHorizontalAlignment;
+#define LINEAR 0
+#define WMAJOR 1
+#define XMAJOR 2
+#define YMAJOR 3
+ uint32_t TileMode;
+ uint32_t VerticalLineStride;
+ uint32_t VerticalLineStrideOffset;
+ bool SamplerL2BypassModeDisable;
+#define WriteOnlyCache 0
+#define ReadWriteCache 1
+ uint32_t RenderCacheReadWriteMode;
+#define NORMAL_MODE 0
+#define PROGRESSIVE_FRAME 2
+#define INTERLACED_FRAME 3
+ uint32_t MediaBoundaryPixelMode;
+ bool CubeFaceEnablePositiveZ;
+ bool CubeFaceEnableNegativeZ;
+ bool CubeFaceEnablePositiveY;
+ bool CubeFaceEnableNegativeY;
+ bool CubeFaceEnablePositiveX;
+ bool CubeFaceEnableNegativeX;
+ struct GEN8_MEMORY_OBJECT_CONTROL_STATE MemoryObjectControlState;
+ float BaseMipLevel;
+ uint32_t SurfaceQPitch;
+ uint32_t Height;
+ uint32_t Width;
+ uint32_t Depth;
+ uint32_t SurfacePitch;
+#define _0DEG 0
+#define _90DEG 1
+#define _270DEG 3
+ uint32_t RenderTargetAndSampleUnormRotation;
+ uint32_t MinimumArrayElement;
+ uint32_t RenderTargetViewExtent;
+#define MSS 0
+#define DEPTH_STENCIL 1
+ uint32_t MultisampledSurfaceStorageFormat;
+#define MULTISAMPLECOUNT_1 0
+#define MULTISAMPLECOUNT_2 1
+#define MULTISAMPLECOUNT_4 2
+#define MULTISAMPLECOUNT_8 3
+ uint32_t NumberofMultisamples;
+ uint32_t MultisamplePositionPaletteIndex;
+ uint32_t XOffset;
+ uint32_t YOffset;
+ bool EWADisableForCube;
+#define GPUcoherent 0
+#define IAcoherent 1
+ uint32_t CoherencyType;
+ uint32_t SurfaceMinLOD;
+ uint32_t MIPCountLOD;
+ uint32_t AuxiliarySurfaceQPitch;
+ uint32_t AuxiliarySurfacePitch;
+#define AUX_NONE 0
+#define AUX_MCS 1
+#define AUX_APPEND 2
+#define AUX_HIZ 3
+ uint32_t AuxiliarySurfaceMode;
+ bool SeparateUVPlaneEnable;
+ uint32_t XOffsetforUorUVPlane;
+ uint32_t YOffsetforUorUVPlane;
+ uint32_t RedClearColor;
+ uint32_t GreenClearColor;
+ uint32_t BlueClearColor;
+ uint32_t AlphaClearColor;
+ uint32_t ShaderChannelSelectRed;
+ uint32_t ShaderChannelSelectGreen;
+ uint32_t ShaderChannelSelectBlue;
+ uint32_t ShaderChannelSelectAlpha;
+ float ResourceMinLOD;
+ __gen_address_type SurfaceBaseAddress;
+ uint32_t XOffsetforVPlane;
+ uint32_t YOffsetforVPlane;
+ uint32_t AuxiliaryTableIndexforMediaCompressedSurface;
+ __gen_address_type AuxiliarySurfaceBaseAddress;
+};
+
+static inline void
+GEN8_RENDER_SURFACE_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_RENDER_SURFACE_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->SurfaceType, 29, 31) |
+ __gen_field(values->SurfaceArray, 28, 28) |
+ __gen_field(values->SurfaceFormat, 18, 26) |
+ __gen_field(values->SurfaceVerticalAlignment, 16, 17) |
+ __gen_field(values->SurfaceHorizontalAlignment, 14, 15) |
+ __gen_field(values->TileMode, 12, 13) |
+ __gen_field(values->VerticalLineStride, 11, 11) |
+ __gen_field(values->VerticalLineStrideOffset, 10, 10) |
+ __gen_field(values->SamplerL2BypassModeDisable, 9, 9) |
+ __gen_field(values->RenderCacheReadWriteMode, 8, 8) |
+ __gen_field(values->MediaBoundaryPixelMode, 6, 7) |
+ __gen_field(values->CubeFaceEnablePositiveZ, 0, 0) |
+ __gen_field(values->CubeFaceEnableNegativeZ, 1, 1) |
+ __gen_field(values->CubeFaceEnablePositiveY, 2, 2) |
+ __gen_field(values->CubeFaceEnableNegativeY, 3, 3) |
+ __gen_field(values->CubeFaceEnablePositiveX, 4, 4) |
+ __gen_field(values->CubeFaceEnableNegativeX, 5, 5) |
+ 0;
+
+ uint32_t dw_MemoryObjectControlState;
+ GEN8_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_MemoryObjectControlState, &values->MemoryObjectControlState);
+ dw[1] =
+ __gen_field(dw_MemoryObjectControlState, 24, 30) |
+ __gen_field(values->BaseMipLevel * (1 << 1), 19, 23) |
+ __gen_field(values->SurfaceQPitch, 0, 14) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->Height, 16, 29) |
+ __gen_field(values->Width, 0, 13) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->Depth, 21, 31) |
+ __gen_field(values->SurfacePitch, 0, 17) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->RenderTargetAndSampleUnormRotation, 29, 30) |
+ __gen_field(values->MinimumArrayElement, 18, 28) |
+ __gen_field(values->RenderTargetViewExtent, 7, 17) |
+ __gen_field(values->MultisampledSurfaceStorageFormat, 6, 6) |
+ __gen_field(values->NumberofMultisamples, 3, 5) |
+ __gen_field(values->MultisamplePositionPaletteIndex, 0, 2) |
+ 0;
+
+ dw[5] =
+ __gen_offset(values->XOffset, 25, 31) |
+ __gen_offset(values->YOffset, 21, 23) |
+ __gen_field(values->EWADisableForCube, 20, 20) |
+ __gen_field(values->CoherencyType, 14, 14) |
+ __gen_field(values->SurfaceMinLOD, 4, 7) |
+ __gen_field(values->MIPCountLOD, 0, 3) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->AuxiliarySurfaceQPitch, 16, 30) |
+ __gen_field(values->AuxiliarySurfacePitch, 3, 11) |
+ __gen_field(values->AuxiliarySurfaceMode, 0, 2) |
+ __gen_field(values->SeparateUVPlaneEnable, 31, 31) |
+ __gen_field(values->XOffsetforUorUVPlane, 16, 29) |
+ __gen_field(values->YOffsetforUorUVPlane, 0, 13) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->RedClearColor, 31, 31) |
+ __gen_field(values->GreenClearColor, 30, 30) |
+ __gen_field(values->BlueClearColor, 29, 29) |
+ __gen_field(values->AlphaClearColor, 28, 28) |
+ __gen_field(values->ShaderChannelSelectRed, 25, 27) |
+ __gen_field(values->ShaderChannelSelectGreen, 22, 24) |
+ __gen_field(values->ShaderChannelSelectBlue, 19, 21) |
+ __gen_field(values->ShaderChannelSelectAlpha, 16, 18) |
+ __gen_field(values->ResourceMinLOD * (1 << 8), 0, 11) |
+ 0;
+
+ uint32_t dw8 =
+ 0;
+
+ uint64_t qw8 =
+ __gen_combine_address(data, &dw[8], values->SurfaceBaseAddress, dw8);
+
+ dw[8] = qw8;
+ dw[9] = qw8 >> 32;
+
+ uint32_t dw10 =
+ __gen_field(values->XOffsetforVPlane, 48, 61) |
+ __gen_field(values->YOffsetforVPlane, 32, 45) |
+ __gen_field(values->AuxiliaryTableIndexforMediaCompressedSurface, 21, 31) |
+ 0;
+
+ uint64_t qw10 =
+ __gen_combine_address(data, &dw[10], values->AuxiliarySurfaceBaseAddress, dw10);
+
+ dw[10] = qw10;
+ dw[11] = qw10 >> 32;
+
+ dw[12] =
+ 0;
+
+ dw[13] =
+ 0;
+
+ dw[14] =
+ 0;
+
+ dw[15] =
+ 0;
+
+}
+
+#define GEN8_FILTER_COEFFICIENT_length 0x00000001
+
+struct GEN8_FILTER_COEFFICIENT {
+ uint32_t FilterCoefficient;
+};
+
+static inline void
+GEN8_FILTER_COEFFICIENT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_FILTER_COEFFICIENT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->FilterCoefficient, 0, 7) |
+ 0;
+
+}
+
+#define GEN8_SAMPLER_STATE_length 0x00000004
+
+struct GEN8_SAMPLER_STATE {
+ bool SamplerDisable;
+#define DX10OGL 0
+#define DX9 1
+ uint32_t TextureBorderColorMode;
+#define CLAMP_NONE 0
+#define CLAMP_OGL 2
+ uint32_t LODPreClampMode;
+ float BaseMipLevel;
+#define MIPFILTER_NONE 0
+#define MIPFILTER_NEAREST 1
+#define MIPFILTER_LINEAR 3
+ uint32_t MipModeFilter;
+#define MAPFILTER_NEAREST 0
+#define MAPFILTER_LINEAR 1
+#define MAPFILTER_ANISOTROPIC 2
+#define MAPFILTER_MONO 6
+ uint32_t MagModeFilter;
+#define MAPFILTER_NEAREST 0
+#define MAPFILTER_LINEAR 1
+#define MAPFILTER_ANISOTROPIC 2
+#define MAPFILTER_MONO 6
+ uint32_t MinModeFilter;
+ float TextureLODBias;
+#define LEGACY 0
+#define EWAApproximation 1
+ uint32_t AnisotropicAlgorithm;
+ float MinLOD;
+ float MaxLOD;
+ bool ChromaKeyEnable;
+ uint32_t ChromaKeyIndex;
+#define KEYFILTER_KILL_ON_ANY_MATCH 0
+#define KEYFILTER_REPLACE_BLACK 1
+ uint32_t ChromaKeyMode;
+#define PREFILTEROPALWAYS 0
+#define PREFILTEROPNEVER 1
+#define PREFILTEROPLESS 2
+#define PREFILTEROPEQUAL 3
+#define PREFILTEROPLEQUAL 4
+#define PREFILTEROPGREATER 5
+#define PREFILTEROPNOTEQUAL 6
+#define PREFILTEROPGEQUAL 7
+ uint32_t ShadowFunction;
+#define PROGRAMMED 0
+#define OVERRIDE 1
+ uint32_t CubeSurfaceControlMode;
+ uint32_t IndirectStatePointer;
+#define MIPNONE 0
+#define MIPFILTER 1
+ uint32_t LODClampMagnificationMode;
+#define RATIO21 0
+#define RATIO41 1
+#define RATIO61 2
+#define RATIO81 3
+#define RATIO101 4
+#define RATIO121 5
+#define RATIO141 6
+#define RATIO161 7
+ uint32_t MaximumAnisotropy;
+ bool RAddressMinFilterRoundingEnable;
+ bool RAddressMagFilterRoundingEnable;
+ bool VAddressMinFilterRoundingEnable;
+ bool VAddressMagFilterRoundingEnable;
+ bool UAddressMinFilterRoundingEnable;
+ bool UAddressMagFilterRoundingEnable;
+#define FULL 0
+#define HIGH 1
+#define MED 2
+#define LOW 3
+ uint32_t TrilinearFilterQuality;
+ bool NonnormalizedCoordinateEnable;
+ uint32_t TCXAddressControlMode;
+ uint32_t TCYAddressControlMode;
+ uint32_t TCZAddressControlMode;
+};
+
+static inline void
+GEN8_SAMPLER_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_SAMPLER_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->SamplerDisable, 31, 31) |
+ __gen_field(values->TextureBorderColorMode, 29, 29) |
+ __gen_field(values->LODPreClampMode, 27, 28) |
+ __gen_field(values->BaseMipLevel * (1 << 1), 22, 26) |
+ __gen_field(values->MipModeFilter, 20, 21) |
+ __gen_field(values->MagModeFilter, 17, 19) |
+ __gen_field(values->MinModeFilter, 14, 16) |
+ __gen_fixed(values->TextureLODBias, 1, 13, true, 8) |
+ __gen_field(values->AnisotropicAlgorithm, 0, 0) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->MinLOD * (1 << 8), 20, 31) |
+ __gen_field(values->MaxLOD * (1 << 8), 8, 19) |
+ __gen_field(values->ChromaKeyEnable, 7, 7) |
+ __gen_field(values->ChromaKeyIndex, 5, 6) |
+ __gen_field(values->ChromaKeyMode, 4, 4) |
+ __gen_field(values->ShadowFunction, 1, 3) |
+ __gen_field(values->CubeSurfaceControlMode, 0, 0) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->IndirectStatePointer, 6, 23) |
+ __gen_field(values->LODClampMagnificationMode, 0, 0) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->MaximumAnisotropy, 19, 21) |
+ __gen_field(values->RAddressMinFilterRoundingEnable, 13, 13) |
+ __gen_field(values->RAddressMagFilterRoundingEnable, 14, 14) |
+ __gen_field(values->VAddressMinFilterRoundingEnable, 15, 15) |
+ __gen_field(values->VAddressMagFilterRoundingEnable, 16, 16) |
+ __gen_field(values->UAddressMinFilterRoundingEnable, 17, 17) |
+ __gen_field(values->UAddressMagFilterRoundingEnable, 18, 18) |
+ __gen_field(values->TrilinearFilterQuality, 11, 12) |
+ __gen_field(values->NonnormalizedCoordinateEnable, 10, 10) |
+ __gen_field(values->TCXAddressControlMode, 6, 8) |
+ __gen_field(values->TCYAddressControlMode, 3, 5) |
+ __gen_field(values->TCZAddressControlMode, 0, 2) |
+ 0;
+
+}
+
+#define GEN8_SAMPLER_STATE_8X8_AVS_COEFFICIENTS_length 0x00000008
+
+struct GEN8_SAMPLER_STATE_8X8_AVS_COEFFICIENTS {
+ uint32_t Table0YFilterCoefficientn1;
+ uint32_t Table0XFilterCoefficientn1;
+ uint32_t Table0YFilterCoefficientn0;
+ uint32_t Table0XFilterCoefficientn0;
+ uint32_t Table0YFilterCoefficientn3;
+ uint32_t Table0XFilterCoefficientn3;
+ uint32_t Table0YFilterCoefficientn2;
+ uint32_t Table0XFilterCoefficientn2;
+ uint32_t Table0YFilterCoefficientn5;
+ uint32_t Table0XFilterCoefficientn5;
+ uint32_t Table0YFilterCoefficientn4;
+ uint32_t Table0XFilterCoefficientn4;
+ uint32_t Table0YFilterCoefficientn7;
+ uint32_t Table0XFilterCoefficientn7;
+ uint32_t Table0YFilterCoefficientn6;
+ uint32_t Table0XFilterCoefficientn6;
+ uint32_t Table1XFilterCoefficientn3;
+ uint32_t Table1XFilterCoefficientn2;
+ uint32_t Table1XFilterCoefficientn5;
+ uint32_t Table1XFilterCoefficientn4;
+ uint32_t Table1YFilterCoefficientn3;
+ uint32_t Table1YFilterCoefficientn2;
+ uint32_t Table1YFilterCoefficientn5;
+ uint32_t Table1YFilterCoefficientn4;
+};
+
+static inline void
+GEN8_SAMPLER_STATE_8X8_AVS_COEFFICIENTS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN8_SAMPLER_STATE_8X8_AVS_COEFFICIENTS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->Table0YFilterCoefficientn1, 24, 31) |
+ __gen_field(values->Table0XFilterCoefficientn1, 16, 23) |
+ __gen_field(values->Table0YFilterCoefficientn0, 8, 15) |
+ __gen_field(values->Table0XFilterCoefficientn0, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->Table0YFilterCoefficientn3, 24, 31) |
+ __gen_field(values->Table0XFilterCoefficientn3, 16, 23) |
+ __gen_field(values->Table0YFilterCoefficientn2, 8, 15) |
+ __gen_field(values->Table0XFilterCoefficientn2, 0, 7) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->Table0YFilterCoefficientn5, 24, 31) |
+ __gen_field(values->Table0XFilterCoefficientn5, 16, 23) |
+ __gen_field(values->Table0YFilterCoefficientn4, 8, 15) |
+ __gen_field(values->Table0XFilterCoefficientn4, 0, 7) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->Table0YFilterCoefficientn7, 24, 31) |
+ __gen_field(values->Table0XFilterCoefficientn7, 16, 23) |
+ __gen_field(values->Table0YFilterCoefficientn6, 8, 15) |
+ __gen_field(values->Table0XFilterCoefficientn6, 0, 7) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->Table1XFilterCoefficientn3, 24, 31) |
+ __gen_field(values->Table1XFilterCoefficientn2, 16, 23) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->Table1XFilterCoefficientn5, 8, 15) |
+ __gen_field(values->Table1XFilterCoefficientn4, 0, 7) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->Table1YFilterCoefficientn3, 24, 31) |
+ __gen_field(values->Table1YFilterCoefficientn2, 16, 23) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->Table1YFilterCoefficientn5, 8, 15) |
+ __gen_field(values->Table1YFilterCoefficientn4, 0, 7) |
+ 0;
+
+}
+
+/* Enum 3D_Prim_Topo_Type */
+#define _3DPRIM_POINTLIST 1
+#define _3DPRIM_LINELIST 2
+#define _3DPRIM_LINESTRIP 3
+#define _3DPRIM_TRILIST 4
+#define _3DPRIM_TRISTRIP 5
+#define _3DPRIM_TRIFAN 6
+#define _3DPRIM_QUADLIST 7
+#define _3DPRIM_QUADSTRIP 8
+#define _3DPRIM_LINELIST_ADJ 9
+#define _3DPRIM_LINESTRIP_ADJ 10
+#define _3DPRIM_TRILIST_ADJ 11
+#define _3DPRIM_TRISTRIP_ADJ 12
+#define _3DPRIM_TRISTRIP_REVERSE 13
+#define _3DPRIM_POLYGON 14
+#define _3DPRIM_RECTLIST 15
+#define _3DPRIM_LINELOOP 16
+#define _3DPRIM_POINTLIST_BF 17
+#define _3DPRIM_LINESTRIP_CONT 18
+#define _3DPRIM_LINESTRIP_BF 19
+#define _3DPRIM_LINESTRIP_CONT_BF 20
+#define _3DPRIM_TRIFAN_NOSTIPPLE 22
+#define _3DPRIM_PATCHLIST_1 32
+#define _3DPRIM_PATCHLIST_2 33
+#define _3DPRIM_PATCHLIST_3 34
+#define _3DPRIM_PATCHLIST_4 35
+#define _3DPRIM_PATCHLIST_5 36
+#define _3DPRIM_PATCHLIST_6 37
+#define _3DPRIM_PATCHLIST_7 38
+#define _3DPRIM_PATCHLIST_8 39
+#define _3DPRIM_PATCHLIST_9 40
+#define _3DPRIM_PATCHLIST_10 41
+#define _3DPRIM_PATCHLIST_11 42
+#define _3DPRIM_PATCHLIST_12 43
+#define _3DPRIM_PATCHLIST_13 44
+#define _3DPRIM_PATCHLIST_14 45
+#define _3DPRIM_PATCHLIST_15 46
+#define _3DPRIM_PATCHLIST_16 47
+#define _3DPRIM_PATCHLIST_17 48
+#define _3DPRIM_PATCHLIST_18 49
+#define _3DPRIM_PATCHLIST_19 50
+#define _3DPRIM_PATCHLIST_20 51
+#define _3DPRIM_PATCHLIST_21 52
+#define _3DPRIM_PATCHLIST_22 53
+#define _3DPRIM_PATCHLIST_23 54
+#define _3DPRIM_PATCHLIST_24 55
+#define _3DPRIM_PATCHLIST_25 56
+#define _3DPRIM_PATCHLIST_26 57
+#define _3DPRIM_PATCHLIST_27 58
+#define _3DPRIM_PATCHLIST_28 59
+#define _3DPRIM_PATCHLIST_29 60
+#define _3DPRIM_PATCHLIST_30 61
+#define _3DPRIM_PATCHLIST_31 62
+#define _3DPRIM_PATCHLIST_32 63
+
+/* Enum 3D_Vertex_Component_Control */
+#define VFCOMP_NOSTORE 0
+#define VFCOMP_STORE_SRC 1
+#define VFCOMP_STORE_0 2
+#define VFCOMP_STORE_1_FP 3
+#define VFCOMP_STORE_1_INT 4
+#define VFCOMP_STORE_PID 7
+
+/* Enum WRAP_SHORTEST_ENABLE */
+#define WSE_X 1
+#define WSE_Y 2
+#define WSE_XY 3
+#define WSE_Z 4
+#define WSE_XZ 5
+#define WSE_YZ 6
+#define WSE_XYZ 7
+#define WSE_W 8
+#define WSE_XW 9
+#define WSE_YW 10
+#define WSE_XYW 11
+#define WSE_ZW 12
+#define WSE_XZW 13
+#define WSE_YZW 14
+#define WSE_XYZW 15
+
+/* Enum 3D_Stencil_Operation */
+#define STENCILOP_KEEP 0
+#define STENCILOP_ZERO 1
+#define STENCILOP_REPLACE 2
+#define STENCILOP_INCRSAT 3
+#define STENCILOP_DECRSAT 4
+#define STENCILOP_INCR 5
+#define STENCILOP_DECR 6
+#define STENCILOP_INVERT 7
+
+/* Enum 3D_Color_Buffer_Blend_Factor */
+#define BLENDFACTOR_ONE 1
+#define BLENDFACTOR_SRC_COLOR 2
+#define BLENDFACTOR_SRC_ALPHA 3
+#define BLENDFACTOR_DST_ALPHA 4
+#define BLENDFACTOR_DST_COLOR 5
+#define BLENDFACTOR_SRC_ALPHA_SATURATE 6
+#define BLENDFACTOR_CONST_COLOR 7
+#define BLENDFACTOR_CONST_ALPHA 8
+#define BLENDFACTOR_SRC1_COLOR 9
+#define BLENDFACTOR_SRC1_ALPHA 10
+#define BLENDFACTOR_ZERO 17
+#define BLENDFACTOR_INV_SRC_COLOR 18
+#define BLENDFACTOR_INV_SRC_ALPHA 19
+#define BLENDFACTOR_INV_DST_ALPHA 20
+#define BLENDFACTOR_INV_DST_COLOR 21
+#define BLENDFACTOR_INV_CONST_COLOR 23
+#define BLENDFACTOR_INV_CONST_ALPHA 24
+#define BLENDFACTOR_INV_SRC1_COLOR 25
+#define BLENDFACTOR_INV_SRC1_ALPHA 26
+
+/* Enum 3D_Color_Buffer_Blend_Function */
+#define BLENDFUNCTION_ADD 0
+#define BLENDFUNCTION_SUBTRACT 1
+#define BLENDFUNCTION_REVERSE_SUBTRACT 2
+#define BLENDFUNCTION_MIN 3
+#define BLENDFUNCTION_MAX 4
+
+/* Enum 3D_Compare_Function */
+#define COMPAREFUNCTION_ALWAYS 0
+#define COMPAREFUNCTION_NEVER 1
+#define COMPAREFUNCTION_LESS 2
+#define COMPAREFUNCTION_EQUAL 3
+#define COMPAREFUNCTION_LEQUAL 4
+#define COMPAREFUNCTION_GREATER 5
+#define COMPAREFUNCTION_NOTEQUAL 6
+#define COMPAREFUNCTION_GEQUAL 7
+
+/* Enum 3D_Logic_Op_Function */
+#define LOGICOP_CLEAR 0
+#define LOGICOP_NOR 1
+#define LOGICOP_AND_INVERTED 2
+#define LOGICOP_COPY_INVERTED 3
+#define LOGICOP_AND_REVERSE 4
+#define LOGICOP_INVERT 5
+#define LOGICOP_XOR 6
+#define LOGICOP_NAND 7
+#define LOGICOP_AND 8
+#define LOGICOP_EQUIV 9
+#define LOGICOP_NOOP 10
+#define LOGICOP_OR_INVERTED 11
+#define LOGICOP_COPY 12
+#define LOGICOP_OR_REVERSE 13
+#define LOGICOP_OR 14
+#define LOGICOP_SET 15
+
+/* Enum SURFACE_FORMAT */
+#define R32G32B32A32_FLOAT 0
+#define R32G32B32A32_SINT 1
+#define R32G32B32A32_UINT 2
+#define R32G32B32A32_UNORM 3
+#define R32G32B32A32_SNORM 4
+#define R64G64_FLOAT 5
+#define R32G32B32X32_FLOAT 6
+#define R32G32B32A32_SSCALED 7
+#define R32G32B32A32_USCALED 8
+#define R32G32B32A32_SFIXED 32
+#define R64G64_PASSTHRU 33
+#define R32G32B32_FLOAT 64
+#define R32G32B32_SINT 65
+#define R32G32B32_UINT 66
+#define R32G32B32_UNORM 67
+#define R32G32B32_SNORM 68
+#define R32G32B32_SSCALED 69
+#define R32G32B32_USCALED 70
+#define R32G32B32_SFIXED 80
+#define R16G16B16A16_UNORM 128
+#define R16G16B16A16_SNORM 129
+#define R16G16B16A16_SINT 130
+#define R16G16B16A16_UINT 131
+#define R16G16B16A16_FLOAT 132
+#define R32G32_FLOAT 133
+#define R32G32_SINT 134
+#define R32G32_UINT 135
+#define R32_FLOAT_X8X24_TYPELESS 136
+#define X32_TYPELESS_G8X24_UINT 137
+#define L32A32_FLOAT 138
+#define R32G32_UNORM 139
+#define R32G32_SNORM 140
+#define R64_FLOAT 141
+#define R16G16B16X16_UNORM 142
+#define R16G16B16X16_FLOAT 143
+#define A32X32_FLOAT 144
+#define L32X32_FLOAT 145
+#define I32X32_FLOAT 146
+#define R16G16B16A16_SSCALED 147
+#define R16G16B16A16_USCALED 148
+#define R32G32_SSCALED 149
+#define R32G32_USCALED 150
+#define R32G32_SFIXED 160
+#define R64_PASSTHRU 161
+#define B8G8R8A8_UNORM 192
+#define B8G8R8A8_UNORM_SRGB 193
+#define R10G10B10A2_UNORM 194
+#define R10G10B10A2_UNORM_SRGB 195
+#define R10G10B10A2_UINT 196
+#define R10G10B10_SNORM_A2_UNORM 197
+#define R8G8B8A8_UNORM 199
+#define R8G8B8A8_UNORM_SRGB 200
+#define R8G8B8A8_SNORM 201
+#define R8G8B8A8_SINT 202
+#define R8G8B8A8_UINT 203
+#define R16G16_UNORM 204
+#define R16G16_SNORM 205
+#define R16G16_SINT 206
+#define R16G16_UINT 207
+#define R16G16_FLOAT 208
+#define B10G10R10A2_UNORM 209
+#define B10G10R10A2_UNORM_SRGB 210
+#define R11G11B10_FLOAT 211
+#define R32_SINT 214
+#define R32_UINT 215
+#define R32_FLOAT 216
+#define R24_UNORM_X8_TYPELESS 217
+#define X24_TYPELESS_G8_UINT 218
+#define L32_UNORM 221
+#define A32_UNORM 222
+#define L16A16_UNORM 223
+#define I24X8_UNORM 224
+#define L24X8_UNORM 225
+#define A24X8_UNORM 226
+#define I32_FLOAT 227
+#define L32_FLOAT 228
+#define A32_FLOAT 229
+#define X8B8_UNORM_G8R8_SNORM 230
+#define A8X8_UNORM_G8R8_SNORM 231
+#define B8X8_UNORM_G8R8_SNORM 232
+#define B8G8R8X8_UNORM 233
+#define B8G8R8X8_UNORM_SRGB 234
+#define R8G8B8X8_UNORM 235
+#define R8G8B8X8_UNORM_SRGB 236
+#define R9G9B9E5_SHAREDEXP 237
+#define B10G10R10X2_UNORM 238
+#define L16A16_FLOAT 240
+#define R32_UNORM 241
+#define R32_SNORM 242
+#define R10G10B10X2_USCALED 243
+#define R8G8B8A8_SSCALED 244
+#define R8G8B8A8_USCALED 245
+#define R16G16_SSCALED 246
+#define R16G16_USCALED 247
+#define R32_SSCALED 248
+#define R32_USCALED 249
+#define B5G6R5_UNORM 256
+#define B5G6R5_UNORM_SRGB 257
+#define B5G5R5A1_UNORM 258
+#define B5G5R5A1_UNORM_SRGB 259
+#define B4G4R4A4_UNORM 260
+#define B4G4R4A4_UNORM_SRGB 261
+#define R8G8_UNORM 262
+#define R8G8_SNORM 263
+#define R8G8_SINT 264
+#define R8G8_UINT 265
+#define R16_UNORM 266
+#define R16_SNORM 267
+#define R16_SINT 268
+#define R16_UINT 269
+#define R16_FLOAT 270
+#define A8P8_UNORM_PALETTE0 271
+#define A8P8_UNORM_PALETTE1 272
+#define I16_UNORM 273
+#define L16_UNORM 274
+#define A16_UNORM 275
+#define L8A8_UNORM 276
+#define I16_FLOAT 277
+#define L16_FLOAT 278
+#define A16_FLOAT 279
+#define L8A8_UNORM_SRGB 280
+#define R5G5_SNORM_B6_UNORM 281
+#define B5G5R5X1_UNORM 282
+#define B5G5R5X1_UNORM_SRGB 283
+#define R8G8_SSCALED 284
+#define R8G8_USCALED 285
+#define R16_SSCALED 286
+#define R16_USCALED 287
+#define P8A8_UNORM_PALETTE0 290
+#define P8A8_UNORM_PALETTE1 291
+#define A1B5G5R5_UNORM 292
+#define A4B4G4R4_UNORM 293
+#define L8A8_UINT 294
+#define L8A8_SINT 295
+#define R8_UNORM 320
+#define R8_SNORM 321
+#define R8_SINT 322
+#define R8_UINT 323
+#define A8_UNORM 324
+#define I8_UNORM 325
+#define L8_UNORM 326
+#define P4A4_UNORM_PALETTE0 327
+#define A4P4_UNORM_PALETTE0 328
+#define R8_SSCALED 329
+#define R8_USCALED 330
+#define P8_UNORM_PALETTE0 331
+#define L8_UNORM_SRGB 332
+#define P8_UNORM_PALETTE1 333
+#define P4A4_UNORM_PALETTE1 334
+#define A4P4_UNORM_PALETTE1 335
+#define Y8_UNORM 336
+#define L8_UINT 338
+#define L8_SINT 339
+#define I8_UINT 340
+#define I8_SINT 341
+#define DXT1_RGB_SRGB 384
+#define R1_UNORM 385
+#define YCRCB_NORMAL 386
+#define YCRCB_SWAPUVY 387
+#define P2_UNORM_PALETTE0 388
+#define P2_UNORM_PALETTE1 389
+#define BC1_UNORM 390
+#define BC2_UNORM 391
+#define BC3_UNORM 392
+#define BC4_UNORM 393
+#define BC5_UNORM 394
+#define BC1_UNORM_SRGB 395
+#define BC2_UNORM_SRGB 396
+#define BC3_UNORM_SRGB 397
+#define MONO8 398
+#define YCRCB_SWAPUV 399
+#define YCRCB_SWAPY 400
+#define DXT1_RGB 401
+#define FXT1 402
+#define R8G8B8_UNORM 403
+#define R8G8B8_SNORM 404
+#define R8G8B8_SSCALED 405
+#define R8G8B8_USCALED 406
+#define R64G64B64A64_FLOAT 407
+#define R64G64B64_FLOAT 408
+#define BC4_SNORM 409
+#define BC5_SNORM 410
+#define R16G16B16_FLOAT 411
+#define R16G16B16_UNORM 412
+#define R16G16B16_SNORM 413
+#define R16G16B16_SSCALED 414
+#define R16G16B16_USCALED 415
+#define BC6H_SF16 417
+#define BC7_UNORM 418
+#define BC7_UNORM_SRGB 419
+#define BC6H_UF16 420
+#define PLANAR_420_8 421
+#define R8G8B8_UNORM_SRGB 424
+#define ETC1_RGB8 425
+#define ETC2_RGB8 426
+#define EAC_R11 427
+#define EAC_RG11 428
+#define EAC_SIGNED_R11 429
+#define EAC_SIGNED_RG11 430
+#define ETC2_SRGB8 431
+#define R16G16B16_UINT 432
+#define R16G16B16_SINT 433
+#define R32_SFIXED 434
+#define R10G10B10A2_SNORM 435
+#define R10G10B10A2_USCALED 436
+#define R10G10B10A2_SSCALED 437
+#define R10G10B10A2_SINT 438
+#define B10G10R10A2_SNORM 439
+#define B10G10R10A2_USCALED 440
+#define B10G10R10A2_SSCALED 441
+#define B10G10R10A2_UINT 442
+#define B10G10R10A2_SINT 443
+#define R64G64B64A64_PASSTHRU 444
+#define R64G64B64_PASSTHRU 445
+#define ETC2_RGB8_PTA 448
+#define ETC2_SRGB8_PTA 449
+#define ETC2_EAC_RGBA8 450
+#define ETC2_EAC_SRGB8_A8 451
+#define R8G8B8_UINT 456
+#define R8G8B8_SINT 457
+#define RAW 511
+
+/* Enum Shader Channel Select */
+#define SCS_ZERO 0
+#define SCS_ONE 1
+#define SCS_RED 4
+#define SCS_GREEN 5
+#define SCS_BLUE 6
+#define SCS_ALPHA 7
+
+/* Enum Clear Color */
+#define CC_ZERO 0
+#define CC_ONE 1
+
+/* Enum Texture Coordinate Mode */
+#define TCM_WRAP 0
+#define TCM_MIRROR 1
+#define TCM_CLAMP 2
+#define TCM_CUBE 3
+#define TCM_CLAMP_BORDER 4
+#define TCM_MIRROR_ONCE 5
+#define TCM_HALF_BORDER 6
+
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <assert.h>
+#include <stdbool.h>
+#include <string.h>
+#include <unistd.h>
+#include <fcntl.h>
+
+#include "anv_private.h"
+
+#include "gen8_pack.h"
+#include "gen9_pack.h"
+
+#include "genX_pipeline_util.h"
+
+static void
+emit_vertex_input(struct anv_pipeline *pipeline,
+ const VkPipelineVertexInputStateCreateInfo *info,
+ const struct anv_graphics_pipeline_create_info *extra)
+{
+ static_assert(ANV_GEN >= 8, "should be compiling this for gen < 8");
+
+ uint32_t elements;
+ if (extra && extra->disable_vs) {
+ /* If the VS is disabled, just assume the user knows what they're
+ * doing and apply the layout blindly. This can only come from
+ * meta, so this *should* be safe.
+ */
+ elements = 0;
+ for (uint32_t i = 0; i < info->vertexAttributeDescriptionCount; i++)
+ elements |= (1 << info->pVertexAttributeDescriptions[i].location);
+ } else {
+ /* Pull inputs_read out of the VS prog data */
+ uint64_t inputs_read = pipeline->vs_prog_data.inputs_read;
+ assert((inputs_read & ((1 << VERT_ATTRIB_GENERIC0) - 1)) == 0);
+ elements = inputs_read >> VERT_ATTRIB_GENERIC0;
+ }
+
+ const uint32_t num_dwords = 1 + __builtin_popcount(elements) * 2;
+
+ uint32_t *p;
+ if (elements != 0) {
+ p = anv_batch_emitn(&pipeline->batch, num_dwords,
+ GENX(3DSTATE_VERTEX_ELEMENTS));
+ memset(p + 1, 0, (num_dwords - 1) * 4);
+ }
+
+ for (uint32_t i = 0; i < info->vertexAttributeDescriptionCount; i++) {
+ const VkVertexInputAttributeDescription *desc =
+ &info->pVertexAttributeDescriptions[i];
+ enum isl_format format = anv_get_isl_format(desc->format,
+ VK_IMAGE_ASPECT_COLOR_BIT,
+ VK_IMAGE_TILING_LINEAR);
+
+ assert(desc->binding < 32);
+
+ if ((elements & (1 << desc->location)) == 0)
+ continue; /* Binding unused */
+
+ uint32_t slot = __builtin_popcount(elements & ((1 << desc->location) - 1));
+
+ struct GENX(VERTEX_ELEMENT_STATE) element = {
+ .VertexBufferIndex = desc->binding,
+ .Valid = true,
+ .SourceElementFormat = format,
+ .EdgeFlagEnable = false,
+ .SourceElementOffset = desc->offset,
+ .Component0Control = vertex_element_comp_control(format, 0),
+ .Component1Control = vertex_element_comp_control(format, 1),
+ .Component2Control = vertex_element_comp_control(format, 2),
+ .Component3Control = vertex_element_comp_control(format, 3),
+ };
+ GENX(VERTEX_ELEMENT_STATE_pack)(NULL, &p[1 + slot * 2], &element);
+
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_VF_INSTANCING),
+ .InstancingEnable = pipeline->instancing_enable[desc->binding],
+ .VertexElementIndex = slot,
+ /* Vulkan so far doesn't have an instance divisor, so
+ * this is always 1 (ignored if not instancing). */
+ .InstanceDataStepRate = 1);
+ }
+
+ const uint32_t id_slot = __builtin_popcount(elements);
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_VF_SGVS),
+ .VertexIDEnable = pipeline->vs_prog_data.uses_vertexid,
+ .VertexIDComponentNumber = 2,
+ .VertexIDElementOffset = id_slot,
+ .InstanceIDEnable = pipeline->vs_prog_data.uses_instanceid,
+ .InstanceIDComponentNumber = 3,
+ .InstanceIDElementOffset = id_slot);
+}
+
+static void
+emit_ia_state(struct anv_pipeline *pipeline,
+ const VkPipelineInputAssemblyStateCreateInfo *info,
+ const struct anv_graphics_pipeline_create_info *extra)
+{
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_VF_TOPOLOGY),
+ .PrimitiveTopologyType = pipeline->topology);
+}
+
+static void
+emit_rs_state(struct anv_pipeline *pipeline,
+ const VkPipelineRasterizationStateCreateInfo *info,
+ const struct anv_graphics_pipeline_create_info *extra)
+{
+ struct GENX(3DSTATE_SF) sf = {
+ GENX(3DSTATE_SF_header),
+ .ViewportTransformEnable = !(extra && extra->disable_viewport),
+ .TriangleStripListProvokingVertexSelect = 0,
+ .LineStripListProvokingVertexSelect = 0,
+ .TriangleFanProvokingVertexSelect = 0,
+ .PointWidthSource = pipeline->writes_point_size ? Vertex : State,
+ .PointWidth = 1.0,
+ };
+
+ /* FINISHME: VkBool32 rasterizerDiscardEnable; */
+
+ GENX(3DSTATE_SF_pack)(NULL, pipeline->gen8.sf, &sf);
+
+ struct GENX(3DSTATE_RASTER) raster = {
+ GENX(3DSTATE_RASTER_header),
+ .FrontWinding = vk_to_gen_front_face[info->frontFace],
+ .CullMode = vk_to_gen_cullmode[info->cullMode],
+ .FrontFaceFillMode = vk_to_gen_fillmode[info->polygonMode],
+ .BackFaceFillMode = vk_to_gen_fillmode[info->polygonMode],
+ .ScissorRectangleEnable = !(extra && extra->disable_scissor),
+#if ANV_GEN == 8
+ .ViewportZClipTestEnable = true,
+#else
+ /* GEN9+ splits ViewportZClipTestEnable into near and far enable bits */
+ .ViewportZFarClipTestEnable = true,
+ .ViewportZNearClipTestEnable = true,
+#endif
+ };
+
+ GENX(3DSTATE_RASTER_pack)(NULL, pipeline->gen8.raster, &raster);
+}
+
+static void
+emit_cb_state(struct anv_pipeline *pipeline,
+ const VkPipelineColorBlendStateCreateInfo *info,
+ const VkPipelineMultisampleStateCreateInfo *ms_info)
+{
+ struct anv_device *device = pipeline->device;
+
+ uint32_t num_dwords = GENX(BLEND_STATE_length);
+ pipeline->blend_state =
+ anv_state_pool_alloc(&device->dynamic_state_pool, num_dwords * 4, 64);
+
+ struct GENX(BLEND_STATE) blend_state = {
+ .AlphaToCoverageEnable = ms_info && ms_info->alphaToCoverageEnable,
+ .AlphaToOneEnable = ms_info && ms_info->alphaToOneEnable,
+ };
+
+ for (uint32_t i = 0; i < info->attachmentCount; i++) {
+ const VkPipelineColorBlendAttachmentState *a = &info->pAttachments[i];
+
+ if (a->srcColorBlendFactor != a->srcAlphaBlendFactor ||
+ a->dstColorBlendFactor != a->dstAlphaBlendFactor ||
+ a->colorBlendOp != a->alphaBlendOp) {
+ blend_state.IndependentAlphaBlendEnable = true;
+ }
+
+ blend_state.Entry[i] = (struct GENX(BLEND_STATE_ENTRY)) {
+ .LogicOpEnable = info->logicOpEnable,
+ .LogicOpFunction = vk_to_gen_logic_op[info->logicOp],
+ .ColorBufferBlendEnable = a->blendEnable,
+ .PreBlendSourceOnlyClampEnable = false,
+ .ColorClampRange = COLORCLAMP_RTFORMAT,
+ .PreBlendColorClampEnable = true,
+ .PostBlendColorClampEnable = true,
+ .SourceBlendFactor = vk_to_gen_blend[a->srcColorBlendFactor],
+ .DestinationBlendFactor = vk_to_gen_blend[a->dstColorBlendFactor],
+ .ColorBlendFunction = vk_to_gen_blend_op[a->colorBlendOp],
+ .SourceAlphaBlendFactor = vk_to_gen_blend[a->srcAlphaBlendFactor],
+ .DestinationAlphaBlendFactor = vk_to_gen_blend[a->dstAlphaBlendFactor],
+ .AlphaBlendFunction = vk_to_gen_blend_op[a->alphaBlendOp],
+ .WriteDisableAlpha = !(a->colorWriteMask & VK_COLOR_COMPONENT_A_BIT),
+ .WriteDisableRed = !(a->colorWriteMask & VK_COLOR_COMPONENT_R_BIT),
+ .WriteDisableGreen = !(a->colorWriteMask & VK_COLOR_COMPONENT_G_BIT),
+ .WriteDisableBlue = !(a->colorWriteMask & VK_COLOR_COMPONENT_B_BIT),
+ };
+
+ /* Our hardware applies the blend factor prior to the blend function
+ * regardless of what function is used. Technically, this means the
+ * hardware can do MORE than GL or Vulkan specify. However, it also
+ * means that, for MIN and MAX, we have to stomp the blend factor to
+ * ONE to make it a no-op.
+ */
+ if (a->colorBlendOp == VK_BLEND_OP_MIN ||
+ a->colorBlendOp == VK_BLEND_OP_MAX) {
+ blend_state.Entry[i].SourceBlendFactor = BLENDFACTOR_ONE;
+ blend_state.Entry[i].DestinationBlendFactor = BLENDFACTOR_ONE;
+ }
+ if (a->alphaBlendOp == VK_BLEND_OP_MIN ||
+ a->alphaBlendOp == VK_BLEND_OP_MAX) {
+ blend_state.Entry[i].SourceAlphaBlendFactor = BLENDFACTOR_ONE;
+ blend_state.Entry[i].DestinationAlphaBlendFactor = BLENDFACTOR_ONE;
+ }
+ }
+
+ GENX(BLEND_STATE_pack)(NULL, pipeline->blend_state.map, &blend_state);
+ if (!device->info.has_llc)
+ anv_state_clflush(pipeline->blend_state);
+
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_BLEND_STATE_POINTERS),
+ .BlendStatePointer = pipeline->blend_state.offset,
+ .BlendStatePointerValid = true);
+}
+
+static void
+emit_ds_state(struct anv_pipeline *pipeline,
+ const VkPipelineDepthStencilStateCreateInfo *info)
+{
+ uint32_t *dw = ANV_GEN == 8 ?
+ pipeline->gen8.wm_depth_stencil : pipeline->gen9.wm_depth_stencil;
+
+ if (info == NULL) {
+ /* We're going to OR this together with the dynamic state. We need
+ * to make sure it's initialized to something useful.
+ */
+ memset(pipeline->gen8.wm_depth_stencil, 0,
+ sizeof(pipeline->gen8.wm_depth_stencil));
+ memset(pipeline->gen9.wm_depth_stencil, 0,
+ sizeof(pipeline->gen9.wm_depth_stencil));
+ return;
+ }
+
+ /* VkBool32 depthBoundsTestEnable; // optional (depth_bounds_test) */
+
+ struct GENX(3DSTATE_WM_DEPTH_STENCIL) wm_depth_stencil = {
+ .DepthTestEnable = info->depthTestEnable,
+ .DepthBufferWriteEnable = info->depthWriteEnable,
+ .DepthTestFunction = vk_to_gen_compare_op[info->depthCompareOp],
+ .DoubleSidedStencilEnable = true,
+
+ .StencilTestEnable = info->stencilTestEnable,
+ .StencilFailOp = vk_to_gen_stencil_op[info->front.failOp],
+ .StencilPassDepthPassOp = vk_to_gen_stencil_op[info->front.passOp],
+ .StencilPassDepthFailOp = vk_to_gen_stencil_op[info->front.depthFailOp],
+ .StencilTestFunction = vk_to_gen_compare_op[info->front.compareOp],
+ .BackfaceStencilFailOp = vk_to_gen_stencil_op[info->back.failOp],
+ .BackfaceStencilPassDepthPassOp = vk_to_gen_stencil_op[info->back.passOp],
+ .BackfaceStencilPassDepthFailOp =vk_to_gen_stencil_op[info->back.depthFailOp],
+ .BackfaceStencilTestFunction = vk_to_gen_compare_op[info->back.compareOp],
+ };
+
+ GENX(3DSTATE_WM_DEPTH_STENCIL_pack)(NULL, dw, &wm_depth_stencil);
+}
+
+VkResult
+genX(graphics_pipeline_create)(
+ VkDevice _device,
+ struct anv_pipeline_cache * cache,
+ const VkGraphicsPipelineCreateInfo* pCreateInfo,
+ const struct anv_graphics_pipeline_create_info *extra,
+ const VkAllocationCallbacks* pAllocator,
+ VkPipeline* pPipeline)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ struct anv_pipeline *pipeline;
+ VkResult result;
+ uint32_t offset, length;
+
+ assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO);
+
+ pipeline = anv_alloc2(&device->alloc, pAllocator, sizeof(*pipeline), 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (pipeline == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ result = anv_pipeline_init(pipeline, device, cache,
+ pCreateInfo, extra, pAllocator);
+ if (result != VK_SUCCESS) {
+ anv_free2(&device->alloc, pAllocator, pipeline);
+ return result;
+ }
+
+ assert(pCreateInfo->pVertexInputState);
+ emit_vertex_input(pipeline, pCreateInfo->pVertexInputState, extra);
+ assert(pCreateInfo->pInputAssemblyState);
+ emit_ia_state(pipeline, pCreateInfo->pInputAssemblyState, extra);
+ assert(pCreateInfo->pRasterizationState);
+ emit_rs_state(pipeline, pCreateInfo->pRasterizationState, extra);
+ emit_ds_state(pipeline, pCreateInfo->pDepthStencilState);
+ emit_cb_state(pipeline, pCreateInfo->pColorBlendState,
+ pCreateInfo->pMultisampleState);
+
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_VF_STATISTICS),
+ .StatisticsEnable = true);
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_HS), .Enable = false);
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_TE), .TEEnable = false);
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_DS), .FunctionEnable = false);
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_STREAMOUT), .SOFunctionEnable = false);
+
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_PUSH_CONSTANT_ALLOC_VS),
+ .ConstantBufferOffset = 0,
+ .ConstantBufferSize = 4);
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_PUSH_CONSTANT_ALLOC_GS),
+ .ConstantBufferOffset = 4,
+ .ConstantBufferSize = 4);
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_PUSH_CONSTANT_ALLOC_PS),
+ .ConstantBufferOffset = 8,
+ .ConstantBufferSize = 4);
+
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_WM_CHROMAKEY),
+ .ChromaKeyKillEnable = false);
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_AA_LINE_PARAMETERS));
+
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_CLIP),
+ .ClipEnable = true,
+ .ViewportXYClipTestEnable = !(extra && extra->disable_viewport),
+ .MinimumPointWidth = 0.125,
+ .MaximumPointWidth = 255.875);
+
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_WM),
+ .StatisticsEnable = true,
+ .LineEndCapAntialiasingRegionWidth = _05pixels,
+ .LineAntialiasingRegionWidth = _10pixels,
+ .EarlyDepthStencilControl = NORMAL,
+ .ForceThreadDispatchEnable = NORMAL,
+ .PointRasterizationRule = RASTRULE_UPPER_RIGHT,
+ .BarycentricInterpolationMode =
+ pipeline->wm_prog_data.barycentric_interp_modes);
+
+ uint32_t samples = 1;
+ uint32_t log2_samples = __builtin_ffs(samples) - 1;
+ bool enable_sampling = samples > 1 ? true : false;
+
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_MULTISAMPLE),
+ .PixelPositionOffsetEnable = enable_sampling,
+ .PixelLocation = CENTER,
+ .NumberofMultisamples = log2_samples);
+
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_SAMPLE_MASK),
+ .SampleMask = 0xffff);
+
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_URB_VS),
+ .VSURBStartingAddress = pipeline->urb.vs_start,
+ .VSURBEntryAllocationSize = pipeline->urb.vs_size - 1,
+ .VSNumberofURBEntries = pipeline->urb.nr_vs_entries);
+
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_URB_GS),
+ .GSURBStartingAddress = pipeline->urb.gs_start,
+ .GSURBEntryAllocationSize = pipeline->urb.gs_size - 1,
+ .GSNumberofURBEntries = pipeline->urb.nr_gs_entries);
+
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_URB_HS),
+ .HSURBStartingAddress = pipeline->urb.vs_start,
+ .HSURBEntryAllocationSize = 0,
+ .HSNumberofURBEntries = 0);
+
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_URB_DS),
+ .DSURBStartingAddress = pipeline->urb.vs_start,
+ .DSURBEntryAllocationSize = 0,
+ .DSNumberofURBEntries = 0);
+
+ const struct brw_gs_prog_data *gs_prog_data = &pipeline->gs_prog_data;
+ offset = 1;
+ length = (gs_prog_data->base.vue_map.num_slots + 1) / 2 - offset;
+
+ if (pipeline->gs_kernel == NO_KERNEL)
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_GS), .Enable = false);
+ else
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_GS),
+ .SingleProgramFlow = false,
+ .KernelStartPointer = pipeline->gs_kernel,
+ .VectorMaskEnable = Dmask,
+ .SamplerCount = 0,
+ .BindingTableEntryCount = 0,
+ .ExpectedVertexCount = pipeline->gs_vertex_count,
+
+ .ScratchSpaceBasePointer = pipeline->scratch_start[MESA_SHADER_GEOMETRY],
+ .PerThreadScratchSpace = ffs(gs_prog_data->base.base.total_scratch / 2048),
+
+ .OutputVertexSize = gs_prog_data->output_vertex_size_hwords * 2 - 1,
+ .OutputTopology = gs_prog_data->output_topology,
+ .VertexURBEntryReadLength = gs_prog_data->base.urb_read_length,
+ .IncludeVertexHandles = gs_prog_data->base.include_vue_handles,
+ .DispatchGRFStartRegisterForURBData =
+ gs_prog_data->base.base.dispatch_grf_start_reg,
+
+ .MaximumNumberofThreads = device->info.max_gs_threads / 2 - 1,
+ .ControlDataHeaderSize = gs_prog_data->control_data_header_size_hwords,
+ .DispatchMode = gs_prog_data->base.dispatch_mode,
+ .StatisticsEnable = true,
+ .IncludePrimitiveID = gs_prog_data->include_primitive_id,
+ .ReorderMode = TRAILING,
+ .Enable = true,
+
+ .ControlDataFormat = gs_prog_data->control_data_format,
+
+ .StaticOutput = gs_prog_data->static_vertex_count >= 0,
+ .StaticOutputVertexCount =
+ gs_prog_data->static_vertex_count >= 0 ?
+ gs_prog_data->static_vertex_count : 0,
+
+ /* FIXME: mesa sets this based on ctx->Transform.ClipPlanesEnabled:
+ * UserClipDistanceClipTestEnableBitmask_3DSTATE_GS(v)
+ * UserClipDistanceCullTestEnableBitmask(v)
+ */
+
+ .VertexURBEntryOutputReadOffset = offset,
+ .VertexURBEntryOutputLength = length);
+
+ const struct brw_vue_prog_data *vue_prog_data = &pipeline->vs_prog_data.base;
+ /* Skip the VUE header and position slots */
+ offset = 1;
+ length = (vue_prog_data->vue_map.num_slots + 1) / 2 - offset;
+
+ uint32_t vs_start = pipeline->vs_simd8 != NO_KERNEL ? pipeline->vs_simd8 :
+ pipeline->vs_vec4;
+
+ if (vs_start == NO_KERNEL || (extra && extra->disable_vs))
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_VS),
+ .FunctionEnable = false,
+ /* Even if VS is disabled, SBE still gets the amount of
+ * vertex data to read from this field. */
+ .VertexURBEntryOutputReadOffset = offset,
+ .VertexURBEntryOutputLength = length);
+ else
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_VS),
+ .KernelStartPointer = vs_start,
+ .SingleVertexDispatch = Multiple,
+ .VectorMaskEnable = Dmask,
+ .SamplerCount = 0,
+ .BindingTableEntryCount =
+ vue_prog_data->base.binding_table.size_bytes / 4,
+ .ThreadDispatchPriority = Normal,
+ .FloatingPointMode = IEEE754,
+ .IllegalOpcodeExceptionEnable = false,
+ .AccessesUAV = false,
+ .SoftwareExceptionEnable = false,
+
+ .ScratchSpaceBasePointer = pipeline->scratch_start[MESA_SHADER_VERTEX],
+ .PerThreadScratchSpace = ffs(vue_prog_data->base.total_scratch / 2048),
+
+ .DispatchGRFStartRegisterForURBData =
+ vue_prog_data->base.dispatch_grf_start_reg,
+ .VertexURBEntryReadLength = vue_prog_data->urb_read_length,
+ .VertexURBEntryReadOffset = 0,
+
+ .MaximumNumberofThreads = device->info.max_vs_threads - 1,
+ .StatisticsEnable = false,
+ .SIMD8DispatchEnable = pipeline->vs_simd8 != NO_KERNEL,
+ .VertexCacheDisable = false,
+ .FunctionEnable = true,
+
+ .VertexURBEntryOutputReadOffset = offset,
+ .VertexURBEntryOutputLength = length,
+ .UserClipDistanceClipTestEnableBitmask = 0,
+ .UserClipDistanceCullTestEnableBitmask = 0);
+
+ const struct brw_wm_prog_data *wm_prog_data = &pipeline->wm_prog_data;
+
+ /* TODO: We should clean this up. Among other things, this is mostly
+ * shared with other gens.
+ */
+ const struct brw_vue_map *fs_input_map;
+ if (pipeline->gs_kernel == NO_KERNEL)
+ fs_input_map = &vue_prog_data->vue_map;
+ else
+ fs_input_map = &gs_prog_data->base.vue_map;
+
+ struct GENX(3DSTATE_SBE_SWIZ) swiz = {
+ GENX(3DSTATE_SBE_SWIZ_header),
+ };
+
+ int max_source_attr = 0;
+ for (int attr = 0; attr < VARYING_SLOT_MAX; attr++) {
+ int input_index = wm_prog_data->urb_setup[attr];
+
+ if (input_index < 0)
+ continue;
+
+ int source_attr = fs_input_map->varying_to_slot[attr];
+ max_source_attr = MAX2(max_source_attr, source_attr);
+
+ if (input_index >= 16)
+ continue;
+
+ if (source_attr == -1) {
+ /* This attribute does not exist in the VUE--that means that the
+ * vertex shader did not write to it. It could be that it's a
+ * regular varying read by the fragment shader but not written by the
+ * vertex shader or it's gl_PrimitiveID. In the first case the value
+ * is undefined, in the second it needs to be gl_PrimitiveID.
+ */
+ swiz.Attribute[input_index].ConstantSource = PRIM_ID;
+ swiz.Attribute[input_index].ComponentOverrideX = true;
+ swiz.Attribute[input_index].ComponentOverrideY = true;
+ swiz.Attribute[input_index].ComponentOverrideZ = true;
+ swiz.Attribute[input_index].ComponentOverrideW = true;
+ } else {
+ /* We have to subtract two slots to accout for the URB entry output
+ * read offset in the VS and GS stages.
+ */
+ swiz.Attribute[input_index].SourceAttribute = source_attr - 2;
+ }
+ }
+
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_SBE),
+ .AttributeSwizzleEnable = true,
+ .ForceVertexURBEntryReadLength = false,
+ .ForceVertexURBEntryReadOffset = false,
+ .VertexURBEntryReadLength = DIV_ROUND_UP(max_source_attr + 1, 2),
+ .PointSpriteTextureCoordinateOrigin = UPPERLEFT,
+ .NumberofSFOutputAttributes =
+ wm_prog_data->num_varying_inputs,
+
+#if ANV_GEN >= 9
+ .Attribute0ActiveComponentFormat = ACF_XYZW,
+ .Attribute1ActiveComponentFormat = ACF_XYZW,
+ .Attribute2ActiveComponentFormat = ACF_XYZW,
+ .Attribute3ActiveComponentFormat = ACF_XYZW,
+ .Attribute4ActiveComponentFormat = ACF_XYZW,
+ .Attribute5ActiveComponentFormat = ACF_XYZW,
+ .Attribute6ActiveComponentFormat = ACF_XYZW,
+ .Attribute7ActiveComponentFormat = ACF_XYZW,
+ .Attribute8ActiveComponentFormat = ACF_XYZW,
+ .Attribute9ActiveComponentFormat = ACF_XYZW,
+ .Attribute10ActiveComponentFormat = ACF_XYZW,
+ .Attribute11ActiveComponentFormat = ACF_XYZW,
+ .Attribute12ActiveComponentFormat = ACF_XYZW,
+ .Attribute13ActiveComponentFormat = ACF_XYZW,
+ .Attribute14ActiveComponentFormat = ACF_XYZW,
+ .Attribute15ActiveComponentFormat = ACF_XYZW,
+ /* wow, much field, very attribute */
+ .Attribute16ActiveComponentFormat = ACF_XYZW,
+ .Attribute17ActiveComponentFormat = ACF_XYZW,
+ .Attribute18ActiveComponentFormat = ACF_XYZW,
+ .Attribute19ActiveComponentFormat = ACF_XYZW,
+ .Attribute20ActiveComponentFormat = ACF_XYZW,
+ .Attribute21ActiveComponentFormat = ACF_XYZW,
+ .Attribute22ActiveComponentFormat = ACF_XYZW,
+ .Attribute23ActiveComponentFormat = ACF_XYZW,
+ .Attribute24ActiveComponentFormat = ACF_XYZW,
+ .Attribute25ActiveComponentFormat = ACF_XYZW,
+ .Attribute26ActiveComponentFormat = ACF_XYZW,
+ .Attribute27ActiveComponentFormat = ACF_XYZW,
+ .Attribute28ActiveComponentFormat = ACF_XYZW,
+ .Attribute29ActiveComponentFormat = ACF_XYZW,
+ .Attribute28ActiveComponentFormat = ACF_XYZW,
+ .Attribute29ActiveComponentFormat = ACF_XYZW,
+ .Attribute30ActiveComponentFormat = ACF_XYZW,
+#endif
+ );
+
+ uint32_t *dw = anv_batch_emit_dwords(&pipeline->batch,
+ GENX(3DSTATE_SBE_SWIZ_length));
+ GENX(3DSTATE_SBE_SWIZ_pack)(&pipeline->batch, dw, &swiz);
+
+ const int num_thread_bias = ANV_GEN == 8 ? 2 : 1;
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_PS),
+ .KernelStartPointer0 = pipeline->ps_ksp0,
+
+ .SingleProgramFlow = false,
+ .VectorMaskEnable = true,
+ .SamplerCount = 1,
+
+ .ScratchSpaceBasePointer = pipeline->scratch_start[MESA_SHADER_FRAGMENT],
+ .PerThreadScratchSpace = ffs(wm_prog_data->base.total_scratch / 2048),
+
+ .MaximumNumberofThreadsPerPSD = 64 - num_thread_bias,
+ .PositionXYOffsetSelect = wm_prog_data->uses_pos_offset ?
+ POSOFFSET_SAMPLE: POSOFFSET_NONE,
+ .PushConstantEnable = wm_prog_data->base.nr_params > 0,
+ ._8PixelDispatchEnable = pipeline->ps_simd8 != NO_KERNEL,
+ ._16PixelDispatchEnable = pipeline->ps_simd16 != NO_KERNEL,
+ ._32PixelDispatchEnable = false,
+
+ .DispatchGRFStartRegisterForConstantSetupData0 = pipeline->ps_grf_start0,
+ .DispatchGRFStartRegisterForConstantSetupData1 = 0,
+ .DispatchGRFStartRegisterForConstantSetupData2 = pipeline->ps_grf_start2,
+
+ .KernelStartPointer1 = 0,
+ .KernelStartPointer2 = pipeline->ps_ksp2);
+
+ bool per_sample_ps = false;
+ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_PS_EXTRA),
+ .PixelShaderValid = true,
+ .PixelShaderKillsPixel = wm_prog_data->uses_kill,
+ .PixelShaderComputedDepthMode = wm_prog_data->computed_depth_mode,
+ .AttributeEnable = wm_prog_data->num_varying_inputs > 0,
+ .oMaskPresenttoRenderTarget = wm_prog_data->uses_omask,
+ .PixelShaderIsPerSample = per_sample_ps,
+#if ANV_GEN >= 9
+ .PixelShaderPullsBary = wm_prog_data->pulls_bary,
+ .InputCoverageMaskState = ICMS_NONE
+#endif
+ );
+
+ *pPipeline = anv_pipeline_to_handle(pipeline);
+
+ return VK_SUCCESS;
+}
+
+VkResult genX(compute_pipeline_create)(
+ VkDevice _device,
+ struct anv_pipeline_cache * cache,
+ const VkComputePipelineCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkPipeline* pPipeline)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ struct anv_pipeline *pipeline;
+ VkResult result;
+
+ assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO);
+
+ pipeline = anv_alloc2(&device->alloc, pAllocator, sizeof(*pipeline), 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (pipeline == NULL)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ pipeline->device = device;
+ pipeline->layout = anv_pipeline_layout_from_handle(pCreateInfo->layout);
+
+ pipeline->blend_state.map = NULL;
+
+ result = anv_reloc_list_init(&pipeline->batch_relocs,
+ pAllocator ? pAllocator : &device->alloc);
+ if (result != VK_SUCCESS) {
+ anv_free2(&device->alloc, pAllocator, pipeline);
+ return result;
+ }
+ pipeline->batch.next = pipeline->batch.start = pipeline->batch_data;
+ pipeline->batch.end = pipeline->batch.start + sizeof(pipeline->batch_data);
+ pipeline->batch.relocs = &pipeline->batch_relocs;
+
+ /* When we free the pipeline, we detect stages based on the NULL status
+ * of various prog_data pointers. Make them NULL by default.
+ */
+ memset(pipeline->prog_data, 0, sizeof(pipeline->prog_data));
+ memset(pipeline->scratch_start, 0, sizeof(pipeline->scratch_start));
+
+ pipeline->vs_simd8 = NO_KERNEL;
+ pipeline->vs_vec4 = NO_KERNEL;
+ pipeline->gs_kernel = NO_KERNEL;
+
+ pipeline->active_stages = 0;
+ pipeline->total_scratch = 0;
+
+ assert(pCreateInfo->stage.stage == VK_SHADER_STAGE_COMPUTE_BIT);
+ ANV_FROM_HANDLE(anv_shader_module, module, pCreateInfo->stage.module);
+ anv_pipeline_compile_cs(pipeline, cache, pCreateInfo, module,
+ pCreateInfo->stage.pName,
+ pCreateInfo->stage.pSpecializationInfo);
+
+ pipeline->use_repclear = false;
+
+ const struct brw_cs_prog_data *cs_prog_data = &pipeline->cs_prog_data;
+
+ anv_batch_emit(&pipeline->batch, GENX(MEDIA_VFE_STATE),
+ .ScratchSpaceBasePointer = pipeline->scratch_start[MESA_SHADER_COMPUTE],
+ .PerThreadScratchSpace = ffs(cs_prog_data->base.total_scratch / 2048),
+ .ScratchSpaceBasePointerHigh = 0,
+ .StackSize = 0,
+
+ .MaximumNumberofThreads = device->info.max_cs_threads - 1,
+ .NumberofURBEntries = 2,
+ .ResetGatewayTimer = true,
+#if ANV_GEN == 8
+ .BypassGatewayControl = true,
+#endif
+ .URBEntryAllocationSize = 2,
+ .CURBEAllocationSize = 0);
+
+ struct brw_cs_prog_data *prog_data = &pipeline->cs_prog_data;
+ uint32_t group_size = prog_data->local_size[0] *
+ prog_data->local_size[1] * prog_data->local_size[2];
+ pipeline->cs_thread_width_max = DIV_ROUND_UP(group_size, prog_data->simd_size);
+ uint32_t remainder = group_size & (prog_data->simd_size - 1);
+
+ if (remainder > 0)
+ pipeline->cs_right_mask = ~0u >> (32 - remainder);
+ else
+ pipeline->cs_right_mask = ~0u >> (32 - prog_data->simd_size);
+
+
+ *pPipeline = anv_pipeline_to_handle(pipeline);
+
+ return VK_SUCCESS;
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <assert.h>
+#include <stdbool.h>
+#include <string.h>
+#include <unistd.h>
+#include <fcntl.h>
+
+#include "anv_private.h"
+
+#include "gen8_pack.h"
+#include "gen9_pack.h"
+
+#include "genX_state_util.h"
+
+void
+genX(fill_buffer_surface_state)(void *state, enum isl_format format,
+ uint32_t offset, uint32_t range, uint32_t stride)
+{
+ uint32_t num_elements = range / stride;
+
+ struct GENX(RENDER_SURFACE_STATE) surface_state = {
+ .SurfaceType = SURFTYPE_BUFFER,
+ .SurfaceArray = false,
+ .SurfaceFormat = format,
+ .SurfaceVerticalAlignment = VALIGN4,
+ .SurfaceHorizontalAlignment = HALIGN4,
+ .TileMode = LINEAR,
+ .SamplerL2BypassModeDisable = true,
+ .RenderCacheReadWriteMode = WriteOnlyCache,
+ .MemoryObjectControlState = GENX(MOCS),
+ .Height = ((num_elements - 1) >> 7) & 0x3fff,
+ .Width = (num_elements - 1) & 0x7f,
+ .Depth = ((num_elements - 1) >> 21) & 0x3f,
+ .SurfacePitch = stride - 1,
+ .NumberofMultisamples = MULTISAMPLECOUNT_1,
+ .ShaderChannelSelectRed = SCS_RED,
+ .ShaderChannelSelectGreen = SCS_GREEN,
+ .ShaderChannelSelectBlue = SCS_BLUE,
+ .ShaderChannelSelectAlpha = SCS_ALPHA,
+ /* FIXME: We assume that the image must be bound at this time. */
+ .SurfaceBaseAddress = { NULL, offset },
+ };
+
+ GENX(RENDER_SURFACE_STATE_pack)(NULL, state, &surface_state);
+}
+
+static const uint8_t anv_halign[] = {
+ [4] = HALIGN4,
+ [8] = HALIGN8,
+ [16] = HALIGN16,
+};
+
+static const uint8_t anv_valign[] = {
+ [4] = VALIGN4,
+ [8] = VALIGN8,
+ [16] = VALIGN16,
+};
+
+static struct anv_state
+alloc_surface_state(struct anv_device *device,
+ struct anv_cmd_buffer *cmd_buffer)
+{
+ if (cmd_buffer) {
+ return anv_cmd_buffer_alloc_surface_state(cmd_buffer);
+ } else {
+ return anv_state_pool_alloc(&device->surface_state_pool, 64, 64);
+ }
+}
+
+/**
+ * Get the values to pack into RENDER_SUFFACE_STATE.SurfaceHorizontalAlignment
+ * and SurfaceVerticalAlignment.
+ */
+static void
+get_halign_valign(const struct isl_surf *surf, uint32_t *halign, uint32_t *valign)
+{
+ #if ANV_GENx10 >= 90
+ if (isl_tiling_is_std_y(surf->tiling) ||
+ surf->dim_layout == ISL_DIM_LAYOUT_GEN9_1D) {
+ /* The hardware ignores the alignment values. Anyway, the surface's
+ * true alignment is likely outside the enum range of HALIGN* and
+ * VALIGN*.
+ */
+ *halign = 0;
+ *valign = 0;
+ } else {
+ /* In Skylake, RENDER_SUFFACE_STATE.SurfaceVerticalAlignment is in units
+ * of surface elements (not pixels nor samples). For compressed formats,
+ * a "surface element" is defined as a compression block. For example,
+ * if SurfaceVerticalAlignment is VALIGN_4 and SurfaceFormat is an ETC2
+ * format (ETC2 has a block height of 4), then the vertical alignment is
+ * 4 compression blocks or, equivalently, 16 pixels.
+ */
+ struct isl_extent3d image_align_el
+ = isl_surf_get_image_alignment_el(surf);
+
+ *halign = anv_halign[image_align_el.width];
+ *valign = anv_valign[image_align_el.height];
+ }
+ #else
+ /* Pre-Skylake, RENDER_SUFFACE_STATE.SurfaceVerticalAlignment is in
+ * units of surface samples. For example, if SurfaceVerticalAlignment
+ * is VALIGN_4 and the surface is singlesampled, then for any surface
+ * format (compressed or not) the vertical alignment is
+ * 4 pixels.
+ */
+ struct isl_extent3d image_align_sa
+ = isl_surf_get_image_alignment_sa(surf);
+
+ *halign = anv_halign[image_align_sa.width];
+ *valign = anv_valign[image_align_sa.height];
+ #endif
+}
+
+static uint32_t
+get_qpitch(const struct isl_surf *surf)
+{
+ switch (surf->dim) {
+ default:
+ unreachable(!"bad isl_surf_dim");
+ case ISL_SURF_DIM_1D:
+ #if ANV_GENx10 >= 90
+ /* QPitch is usually expressed as rows of surface elements (where
+ * a surface element is an compression block or a single surface
+ * sample). Skylake 1D is an outlier.
+ *
+ * From the Skylake BSpec >> Memory Views >> Common Surface
+ * Formats >> Surface Layout and Tiling >> 1D Surfaces:
+ *
+ * Surface QPitch specifies the distance in pixels between array
+ * slices.
+ */
+ return isl_surf_get_array_pitch_el(surf);
+ #else
+ return isl_surf_get_array_pitch_el_rows(surf);
+ #endif
+ case ISL_SURF_DIM_2D:
+ case ISL_SURF_DIM_3D:
+ return isl_surf_get_array_pitch_el_rows(surf);
+ }
+}
+
+void
+genX(image_view_init)(struct anv_image_view *iview,
+ struct anv_device *device,
+ const VkImageViewCreateInfo* pCreateInfo,
+ struct anv_cmd_buffer *cmd_buffer)
+{
+ ANV_FROM_HANDLE(anv_image, image, pCreateInfo->image);
+
+ const VkImageSubresourceRange *range = &pCreateInfo->subresourceRange;
+
+ struct anv_surface *surface =
+ anv_image_get_surface_for_aspect_mask(image, range->aspectMask);
+
+ static const uint8_t isl_to_gen_tiling[] = {
+ [ISL_TILING_LINEAR] = LINEAR,
+ [ISL_TILING_X] = XMAJOR,
+ [ISL_TILING_Y0] = YMAJOR,
+ [ISL_TILING_Yf] = YMAJOR,
+ [ISL_TILING_Ys] = YMAJOR,
+ [ISL_TILING_W] = WMAJOR,
+ };
+
+ uint32_t halign, valign;
+ get_halign_valign(&surface->isl, &halign, &valign);
+
+ struct GENX(RENDER_SURFACE_STATE) template = {
+ .SurfaceType = anv_surftype(image, pCreateInfo->viewType, false),
+ .SurfaceArray = image->array_size > 1,
+ .SurfaceFormat = iview->format,
+ .SurfaceVerticalAlignment = valign,
+ .SurfaceHorizontalAlignment = halign,
+ .TileMode = isl_to_gen_tiling[surface->isl.tiling],
+ .VerticalLineStride = 0,
+ .VerticalLineStrideOffset = 0,
+ .SamplerL2BypassModeDisable = true,
+ .RenderCacheReadWriteMode = WriteOnlyCache,
+ .MemoryObjectControlState = GENX(MOCS),
+
+ /* The driver sets BaseMipLevel in SAMPLER_STATE, not here in
+ * RENDER_SURFACE_STATE. The Broadwell PRM says "it is illegal to have
+ * both Base Mip Level fields nonzero".
+ */
+ .BaseMipLevel = 0.0,
+
+ .SurfaceQPitch = get_qpitch(&surface->isl) >> 2,
+ .Height = image->extent.height - 1,
+ .Width = image->extent.width - 1,
+ .Depth = 0, /* TEMPLATE */
+ .SurfacePitch = surface->isl.row_pitch - 1,
+ .RenderTargetViewExtent = 0, /* TEMPLATE */
+ .MinimumArrayElement = 0, /* TEMPLATE */
+ .NumberofMultisamples = MULTISAMPLECOUNT_1,
+ .XOffset = 0,
+ .YOffset = 0,
+
+ .MIPCountLOD = 0, /* TEMPLATE */
+ .SurfaceMinLOD = 0, /* TEMPLATE */
+
+ .AuxiliarySurfaceMode = AUX_NONE,
+ .RedClearColor = 0,
+ .GreenClearColor = 0,
+ .BlueClearColor = 0,
+ .AlphaClearColor = 0,
+ .ShaderChannelSelectRed = vk_to_gen_swizzle(pCreateInfo->components.r,
+ VK_COMPONENT_SWIZZLE_R),
+ .ShaderChannelSelectGreen = vk_to_gen_swizzle(pCreateInfo->components.g,
+ VK_COMPONENT_SWIZZLE_G),
+ .ShaderChannelSelectBlue = vk_to_gen_swizzle(pCreateInfo->components.b,
+ VK_COMPONENT_SWIZZLE_B),
+ .ShaderChannelSelectAlpha = vk_to_gen_swizzle(pCreateInfo->components.a,
+ VK_COMPONENT_SWIZZLE_A),
+ .ResourceMinLOD = 0.0,
+ .SurfaceBaseAddress = { NULL, iview->offset },
+ };
+
+ switch (template.SurfaceType) {
+ case SURFTYPE_1D:
+ case SURFTYPE_2D:
+ template.MinimumArrayElement = range->baseArrayLayer;
+
+ /* From the Broadwell PRM >> RENDER_SURFACE_STATE::Depth:
+ *
+ * For SURFTYPE_1D, 2D, and CUBE: The range of this field is reduced
+ * by one for each increase from zero of Minimum Array Element. For
+ * example, if Minimum Array Element is set to 1024 on a 2D surface,
+ * the range of this field is reduced to [0,1023].
+ *
+ * In other words, 'Depth' is the number of array layers.
+ */
+ template.Depth = range->layerCount - 1;
+
+ /* From the Broadwell PRM >> RENDER_SURFACE_STATE::RenderTargetViewExtent:
+ *
+ * For Render Target and Typed Dataport 1D and 2D Surfaces:
+ * This field must be set to the same value as the Depth field.
+ */
+ template.RenderTargetViewExtent = template.Depth;
+ break;
+ case SURFTYPE_CUBE:
+ #if ANV_GENx10 >= 90
+ /* Like SURFTYPE_2D, but divided by 6. */
+ template.MinimumArrayElement = range->baseArrayLayer / 6;
+ template.Depth = range->layerCount / 6 - 1;
+ template.RenderTargetViewExtent = template.Depth;
+ #else
+ /* Same as SURFTYPE_2D */
+ template.MinimumArrayElement = range->baseArrayLayer;
+ template.Depth = range->layerCount - 1;
+ template.RenderTargetViewExtent = template.Depth;
+ #endif
+ break;
+ case SURFTYPE_3D:
+ template.MinimumArrayElement = range->baseArrayLayer;
+
+ /* From the Broadwell PRM >> RENDER_SURFACE_STATE::Depth:
+ *
+ * If the volume texture is MIP-mapped, this field specifies the
+ * depth of the base MIP level.
+ */
+ template.Depth = image->extent.depth - 1;
+
+ /* From the Broadwell PRM >> RENDER_SURFACE_STATE::RenderTargetViewExtent:
+ *
+ * For Render Target and Typed Dataport 3D Surfaces: This field
+ * indicates the extent of the accessible 'R' coordinates minus 1 on
+ * the LOD currently being rendered to.
+ */
+ template.RenderTargetViewExtent = iview->extent.depth - 1;
+ break;
+ default:
+ unreachable(!"bad SurfaceType");
+ }
+
+ if (image->needs_nonrt_surface_state) {
+ struct GENX(RENDER_SURFACE_STATE) surface_state = template;
+
+ iview->nonrt_surface_state =
+ alloc_surface_state(device, cmd_buffer);
+
+ /* For non render target surfaces, the hardware interprets field
+ * MIPCount/LOD as MIPCount. The range of levels accessible by the
+ * sampler engine is [SurfaceMinLOD, SurfaceMinLOD + MIPCountLOD].
+ */
+ surface_state.SurfaceMinLOD = range->baseMipLevel;
+ surface_state.MIPCountLOD = MAX2(range->levelCount, 1) - 1;
+
+ GENX(RENDER_SURFACE_STATE_pack)(NULL, iview->nonrt_surface_state.map,
+ &surface_state);
+ if (!device->info.has_llc)
+ anv_state_clflush(iview->nonrt_surface_state);
+ } else {
+ iview->nonrt_surface_state.alloc_size = 0;
+ }
+
+ if (image->needs_color_rt_surface_state) {
+ struct GENX(RENDER_SURFACE_STATE) surface_state = template;
+
+ iview->color_rt_surface_state =
+ alloc_surface_state(device, cmd_buffer);
+
+ /* For render target surfaces, the hardware interprets field
+ * MIPCount/LOD as LOD. The Broadwell PRM says:
+ *
+ * MIPCountLOD defines the LOD that will be rendered into.
+ * SurfaceMinLOD is ignored.
+ */
+ surface_state.MIPCountLOD = range->baseMipLevel;
+ surface_state.SurfaceMinLOD = 0;
+
+ GENX(RENDER_SURFACE_STATE_pack)(NULL, iview->color_rt_surface_state.map,
+ &surface_state);
+ if (!device->info.has_llc)
+ anv_state_clflush(iview->color_rt_surface_state);
+ } else {
+ iview->color_rt_surface_state.alloc_size = 0;
+ }
+
+ if (image->needs_storage_surface_state) {
+ struct GENX(RENDER_SURFACE_STATE) surface_state = template;
+
+ iview->storage_surface_state =
+ alloc_surface_state(device, cmd_buffer);
+
+ surface_state.SurfaceType =
+ anv_surftype(image, pCreateInfo->viewType, true),
+
+ surface_state.SurfaceFormat =
+ isl_lower_storage_image_format(&device->isl_dev, iview->format);
+
+ surface_state.SurfaceMinLOD = range->baseMipLevel;
+ surface_state.MIPCountLOD = MAX2(range->levelCount, 1) - 1;
+
+ GENX(RENDER_SURFACE_STATE_pack)(NULL, iview->storage_surface_state.map,
+ &surface_state);
+ } else {
+ iview->storage_surface_state.alloc_size = 0;
+ }
+}
+
+VkResult genX(CreateSampler)(
+ VkDevice _device,
+ const VkSamplerCreateInfo* pCreateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkSampler* pSampler)
+{
+ ANV_FROM_HANDLE(anv_device, device, _device);
+ struct anv_sampler *sampler;
+
+ assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO);
+
+ sampler = anv_alloc2(&device->alloc, pAllocator, sizeof(*sampler), 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ if (!sampler)
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+
+ uint32_t filter = vk_to_gen_tex_filter(pCreateInfo->magFilter,
+ pCreateInfo->anisotropyEnable);
+
+ uint32_t border_color_offset = device->border_colors.offset +
+ pCreateInfo->borderColor * 64;
+
+ struct GENX(SAMPLER_STATE) sampler_state = {
+ .SamplerDisable = false,
+ .TextureBorderColorMode = DX10OGL,
+ .LODPreClampMode = CLAMP_OGL,
+#if ANV_GEN == 8
+ .BaseMipLevel = 0.0,
+#endif
+ .MipModeFilter = vk_to_gen_mipmap_mode[pCreateInfo->mipmapMode],
+ .MagModeFilter = filter,
+ .MinModeFilter = filter,
+ .TextureLODBias = anv_clamp_f(pCreateInfo->mipLodBias, -16, 15.996),
+ .AnisotropicAlgorithm = EWAApproximation,
+ .MinLOD = anv_clamp_f(pCreateInfo->minLod, 0, 14),
+ .MaxLOD = anv_clamp_f(pCreateInfo->maxLod, 0, 14),
+ .ChromaKeyEnable = 0,
+ .ChromaKeyIndex = 0,
+ .ChromaKeyMode = 0,
+ .ShadowFunction = vk_to_gen_compare_op[pCreateInfo->compareOp],
+ .CubeSurfaceControlMode = 0,
+
+ .IndirectStatePointer = border_color_offset >> 6,
+
+ .LODClampMagnificationMode = MIPNONE,
+ .MaximumAnisotropy = vk_to_gen_max_anisotropy(pCreateInfo->maxAnisotropy),
+ .RAddressMinFilterRoundingEnable = 0,
+ .RAddressMagFilterRoundingEnable = 0,
+ .VAddressMinFilterRoundingEnable = 0,
+ .VAddressMagFilterRoundingEnable = 0,
+ .UAddressMinFilterRoundingEnable = 0,
+ .UAddressMagFilterRoundingEnable = 0,
+ .TrilinearFilterQuality = 0,
+ .NonnormalizedCoordinateEnable = pCreateInfo->unnormalizedCoordinates,
+ .TCXAddressControlMode = vk_to_gen_tex_address[pCreateInfo->addressModeU],
+ .TCYAddressControlMode = vk_to_gen_tex_address[pCreateInfo->addressModeV],
+ .TCZAddressControlMode = vk_to_gen_tex_address[pCreateInfo->addressModeW],
+ };
+
+ GENX(SAMPLER_STATE_pack)(NULL, sampler->state, &sampler_state);
+
+ *pSampler = anv_sampler_to_handle(sampler);
+
+ return VK_SUCCESS;
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+
+/* Instructions, enums and structures for SKL.
+ *
+ * This file has been generated, do not hand edit.
+ */
+
+#pragma once
+
+#include <stdio.h>
+#include <assert.h>
+
+#ifndef __gen_validate_value
+#define __gen_validate_value(x)
+#endif
+
+#ifndef __gen_field_functions
+#define __gen_field_functions
+
+union __gen_value {
+ float f;
+ uint32_t dw;
+};
+
+static inline uint64_t
+__gen_mbo(uint32_t start, uint32_t end)
+{
+ return (~0ul >> (64 - (end - start + 1))) << start;
+}
+
+static inline uint64_t
+__gen_field(uint64_t v, uint32_t start, uint32_t end)
+{
+ __gen_validate_value(v);
+#if DEBUG
+ if (end - start + 1 < 64)
+ assert(v < 1ul << (end - start + 1));
+#endif
+
+ return v << start;
+}
+
+static inline uint64_t
+__gen_fixed(float v, uint32_t start, uint32_t end,
+ bool is_signed, uint32_t fract_bits)
+{
+ __gen_validate_value(v);
+
+ const float factor = (1 << fract_bits);
+
+ float max, min;
+ if (is_signed) {
+ max = ((1 << (end - start)) - 1) / factor;
+ min = -(1 << (end - start)) / factor;
+ } else {
+ max = ((1 << (end - start + 1)) - 1) / factor;
+ min = 0.0f;
+ }
+
+ if (v > max)
+ v = max;
+ else if (v < min)
+ v = min;
+
+ int32_t int_val = roundf(v * factor);
+
+ if (is_signed)
+ int_val &= (1 << (end - start + 1)) - 1;
+
+ return int_val << start;
+}
+
+static inline uint64_t
+__gen_offset(uint64_t v, uint32_t start, uint32_t end)
+{
+ __gen_validate_value(v);
+#if DEBUG
+ uint64_t mask = (~0ul >> (64 - (end - start + 1))) << start;
+
+ assert((v & ~mask) == 0);
+#endif
+
+ return v;
+}
+
+static inline uint32_t
+__gen_float(float v)
+{
+ __gen_validate_value(v);
+ return ((union __gen_value) { .f = (v) }).dw;
+}
+
+#ifndef __gen_address_type
+#error #define __gen_address_type before including this file
+#endif
+
+#ifndef __gen_user_data
+#error #define __gen_combine_address before including this file
+#endif
+
+#endif
+
+#define GEN9_3DSTATE_URB_VS_length_bias 0x00000002
+#define GEN9_3DSTATE_URB_VS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 48, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_URB_VS_length 0x00000002
+
+struct GEN9_3DSTATE_URB_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t VSURBStartingAddress;
+ uint32_t VSURBEntryAllocationSize;
+ uint32_t VSNumberofURBEntries;
+};
+
+static inline void
+GEN9_3DSTATE_URB_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_URB_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->VSURBStartingAddress, 25, 31) |
+ __gen_field(values->VSURBEntryAllocationSize, 16, 24) |
+ __gen_field(values->VSNumberofURBEntries, 0, 15) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_VS_length_bias 0x00000002
+#define GEN9_3DSTATE_VS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 16, \
+ .DwordLength = 7
+
+#define GEN9_3DSTATE_VS_length 0x00000009
+
+#define __gen_prefix(name) GEN9_ ## name
+
+struct __gen_prefix(3DSTATE_VS) {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint64_t KernelStartPointer;
+#define Multiple 0
+#define Single 1
+ uint32_t SingleVertexDispatch;
+#define Dmask 0
+#define Vmask 1
+ uint32_t VectorMaskEnable;
+#define NoSamplers 0
+#define _14Samplers 1
+#define _58Samplers 2
+#define _912Samplers 3
+#define _1316Samplers 4
+ uint32_t SamplerCount;
+ uint32_t BindingTableEntryCount;
+#define Normal 0
+#define High 1
+ uint32_t ThreadDispatchPriority;
+#define IEEE754 0
+#define Alternate 1
+ uint32_t FloatingPointMode;
+ bool IllegalOpcodeExceptionEnable;
+ bool AccessesUAV;
+ bool SoftwareExceptionEnable;
+ uint64_t ScratchSpaceBasePointer;
+ uint32_t PerThreadScratchSpace;
+ uint32_t DispatchGRFStartRegisterForURBData;
+ uint32_t VertexURBEntryReadLength;
+ uint32_t VertexURBEntryReadOffset;
+ uint32_t MaximumNumberofThreads;
+ bool StatisticsEnable;
+ bool SIMD8DispatchEnable;
+ bool VertexCacheDisable;
+ bool FunctionEnable;
+ uint32_t VertexURBEntryOutputReadOffset;
+ uint32_t VertexURBEntryOutputLength;
+ uint32_t UserClipDistanceClipTestEnableBitmask;
+ uint32_t UserClipDistanceCullTestEnableBitmask;
+};
+
+static inline void
+GEN9_3DSTATE_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint64_t qw1 =
+ __gen_offset(values->KernelStartPointer, 6, 63) |
+ 0;
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+ dw[3] =
+ __gen_field(values->SingleVertexDispatch, 31, 31) |
+ __gen_field(values->VectorMaskEnable, 30, 30) |
+ __gen_field(values->SamplerCount, 27, 29) |
+ __gen_field(values->BindingTableEntryCount, 18, 25) |
+ __gen_field(values->ThreadDispatchPriority, 17, 17) |
+ __gen_field(values->FloatingPointMode, 16, 16) |
+ __gen_field(values->IllegalOpcodeExceptionEnable, 13, 13) |
+ __gen_field(values->AccessesUAV, 12, 12) |
+ __gen_field(values->SoftwareExceptionEnable, 7, 7) |
+ 0;
+
+ uint64_t qw4 =
+ __gen_offset(values->ScratchSpaceBasePointer, 10, 63) |
+ __gen_field(values->PerThreadScratchSpace, 0, 3) |
+ 0;
+
+ dw[4] = qw4;
+ dw[5] = qw4 >> 32;
+
+ dw[6] =
+ __gen_field(values->DispatchGRFStartRegisterForURBData, 20, 24) |
+ __gen_field(values->VertexURBEntryReadLength, 11, 16) |
+ __gen_field(values->VertexURBEntryReadOffset, 4, 9) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->MaximumNumberofThreads, 23, 31) |
+ __gen_field(values->StatisticsEnable, 10, 10) |
+ __gen_field(values->SIMD8DispatchEnable, 2, 2) |
+ __gen_field(values->VertexCacheDisable, 1, 1) |
+ __gen_field(values->FunctionEnable, 0, 0) |
+ 0;
+
+ dw[8] =
+ __gen_field(values->VertexURBEntryOutputReadOffset, 21, 26) |
+ __gen_field(values->VertexURBEntryOutputLength, 16, 20) |
+ __gen_field(values->UserClipDistanceClipTestEnableBitmask, 8, 15) |
+ __gen_field(values->UserClipDistanceCullTestEnableBitmask, 0, 7) |
+ 0;
+
+}
+
+#define GEN9_GPGPU_CSR_BASE_ADDRESS_length_bias 0x00000002
+#define GEN9_GPGPU_CSR_BASE_ADDRESS_header \
+ .CommandType = 3, \
+ .CommandSubType = 0, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 4, \
+ .DwordLength = 1
+
+#define GEN9_GPGPU_CSR_BASE_ADDRESS_length 0x00000003
+
+struct GEN9_GPGPU_CSR_BASE_ADDRESS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ __gen_address_type GPGPUCSRBaseAddress;
+};
+
+static inline void
+GEN9_GPGPU_CSR_BASE_ADDRESS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_GPGPU_CSR_BASE_ADDRESS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw1 =
+ 0;
+
+ uint64_t qw1 =
+ __gen_combine_address(data, &dw[1], values->GPGPUCSRBaseAddress, dw1);
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+}
+
+#define GEN9_MI_ATOMIC_length_bias 0x00000002
+#define GEN9_MI_ATOMIC_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 47
+
+#define GEN9_MI_ATOMIC_length 0x00000003
+
+struct __gen_prefix(MI_ATOMIC) {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define PerProcessGraphicsAddress 0
+#define GlobalGraphicsAddress 1
+ uint32_t MemoryType;
+ uint32_t PostSyncOperation;
+#define DWORD 0
+#define QWORD 1
+#define OCTWORD 2
+#define RESERVED 3
+ uint32_t DataSize;
+ uint32_t InlineData;
+ uint32_t CSSTALL;
+ uint32_t ReturnDataControl;
+ uint32_t ATOMICOPCODE;
+ uint32_t DwordLength;
+ __gen_address_type MemoryAddress;
+ uint32_t Operand1DataDword0;
+ uint32_t Operand2DataDword0;
+ uint32_t Operand1DataDword1;
+ uint32_t Operand2DataDword1;
+ uint32_t Operand1DataDword2;
+ uint32_t Operand2DataDword2;
+ uint32_t Operand1DataDword3;
+ uint32_t Operand2DataDword3;
+};
+
+static inline void
+GEN9_MI_ATOMIC_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_ATOMIC * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->MemoryType, 22, 22) |
+ __gen_field(values->PostSyncOperation, 21, 21) |
+ __gen_field(values->DataSize, 19, 20) |
+ __gen_field(values->InlineData, 18, 18) |
+ __gen_field(values->CSSTALL, 17, 17) |
+ __gen_field(values->ReturnDataControl, 16, 16) |
+ __gen_field(values->ATOMICOPCODE, 8, 15) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw1 =
+ 0;
+
+ uint64_t qw1 =
+ __gen_combine_address(data, &dw[1], values->MemoryAddress, dw1);
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+ dw[3] =
+ __gen_field(values->Operand1DataDword0, 0, 31) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->Operand2DataDword0, 0, 31) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->Operand1DataDword1, 0, 31) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->Operand2DataDword1, 0, 31) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->Operand1DataDword2, 0, 31) |
+ 0;
+
+ dw[8] =
+ __gen_field(values->Operand2DataDword2, 0, 31) |
+ 0;
+
+ dw[9] =
+ __gen_field(values->Operand1DataDword3, 0, 31) |
+ 0;
+
+ dw[10] =
+ __gen_field(values->Operand2DataDword3, 0, 31) |
+ 0;
+
+}
+
+#define GEN9_MI_BATCH_BUFFER_START_length_bias 0x00000002
+#define GEN9_MI_BATCH_BUFFER_START_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 49, \
+ .DwordLength = 1
+
+#define GEN9_MI_BATCH_BUFFER_START_length 0x00000003
+
+struct GEN9_MI_BATCH_BUFFER_START {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define Firstlevelbatch 0
+#define Secondlevelbatch 1
+ uint32_t SecondLevelBatchBuffer;
+ bool AddOffsetEnable;
+ uint32_t PredicationEnable;
+ bool ResourceStreamerEnable;
+#define ASI_GGTT 0
+#define ASI_PPGTT 1
+ uint32_t AddressSpaceIndicator;
+ uint32_t DwordLength;
+ __gen_address_type BatchBufferStartAddress;
+};
+
+static inline void
+GEN9_MI_BATCH_BUFFER_START_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_BATCH_BUFFER_START * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->SecondLevelBatchBuffer, 22, 22) |
+ __gen_field(values->AddOffsetEnable, 16, 16) |
+ __gen_field(values->PredicationEnable, 15, 15) |
+ __gen_field(values->ResourceStreamerEnable, 10, 10) |
+ __gen_field(values->AddressSpaceIndicator, 8, 8) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw1 =
+ 0;
+
+ uint64_t qw1 =
+ __gen_combine_address(data, &dw[1], values->BatchBufferStartAddress, dw1);
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+}
+
+#define GEN9_MI_CONDITIONAL_BATCH_BUFFER_END_length_bias 0x00000002
+#define GEN9_MI_CONDITIONAL_BATCH_BUFFER_END_header\
+ .CommandType = 0, \
+ .MICommandOpcode = 54, \
+ .UseGlobalGTT = 0, \
+ .CompareSemaphore = 0, \
+ .DwordLength = 2
+
+#define GEN9_MI_CONDITIONAL_BATCH_BUFFER_END_length 0x00000004
+
+struct GEN9_MI_CONDITIONAL_BATCH_BUFFER_END {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t UseGlobalGTT;
+ uint32_t CompareSemaphore;
+#define CompareMaskModeDisabled 0
+#define CompareMaskModeEnabled 1
+ uint32_t CompareMaskMode;
+ uint32_t DwordLength;
+ uint32_t CompareDataDword;
+ __gen_address_type CompareAddress;
+};
+
+static inline void
+GEN9_MI_CONDITIONAL_BATCH_BUFFER_END_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_CONDITIONAL_BATCH_BUFFER_END * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->UseGlobalGTT, 22, 22) |
+ __gen_field(values->CompareSemaphore, 21, 21) |
+ __gen_field(values->CompareMaskMode, 19, 19) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->CompareDataDword, 0, 31) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ uint64_t qw2 =
+ __gen_combine_address(data, &dw[2], values->CompareAddress, dw2);
+
+ dw[2] = qw2;
+ dw[3] = qw2 >> 32;
+
+}
+
+#define GEN9_MI_FORCE_WAKEUP_length_bias 0x00000002
+#define GEN9_MI_FORCE_WAKEUP_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 29, \
+ .DwordLength = 0
+
+#define GEN9_MI_FORCE_WAKEUP_length 0x00000002
+
+struct GEN9_MI_FORCE_WAKEUP {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t DwordLength;
+ uint32_t MaskBits;
+ uint32_t ForceRenderAwake;
+ uint32_t ForceMediaAwake;
+};
+
+static inline void
+GEN9_MI_FORCE_WAKEUP_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_FORCE_WAKEUP * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->MaskBits, 16, 31) |
+ __gen_field(values->ForceRenderAwake, 1, 1) |
+ __gen_field(values->ForceMediaAwake, 0, 0) |
+ 0;
+
+}
+
+#define GEN9_MI_LOAD_REGISTER_IMM_length_bias 0x00000002
+#define GEN9_MI_LOAD_REGISTER_IMM_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 34, \
+ .DwordLength = 1
+
+#define GEN9_MI_LOAD_REGISTER_IMM_length 0x00000003
+
+struct GEN9_MI_LOAD_REGISTER_IMM {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t ByteWriteDisables;
+ uint32_t DwordLength;
+ uint32_t RegisterOffset;
+ uint32_t DataDWord;
+};
+
+static inline void
+GEN9_MI_LOAD_REGISTER_IMM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_LOAD_REGISTER_IMM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->ByteWriteDisables, 8, 11) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->RegisterOffset, 2, 22) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->DataDWord, 0, 31) |
+ 0;
+
+}
+
+#define GEN9_MI_LOAD_REGISTER_REG_length_bias 0x00000002
+#define GEN9_MI_LOAD_REGISTER_REG_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 42, \
+ .DwordLength = 1
+
+#define GEN9_MI_LOAD_REGISTER_REG_length 0x00000003
+
+struct GEN9_MI_LOAD_REGISTER_REG {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t DwordLength;
+ uint32_t SourceRegisterAddress;
+ uint32_t DestinationRegisterAddress;
+};
+
+static inline void
+GEN9_MI_LOAD_REGISTER_REG_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_LOAD_REGISTER_REG * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->SourceRegisterAddress, 2, 22) |
+ 0;
+
+ dw[2] =
+ __gen_offset(values->DestinationRegisterAddress, 2, 22) |
+ 0;
+
+}
+
+#define GEN9_MI_SEMAPHORE_SIGNAL_length_bias 0x00000002
+#define GEN9_MI_SEMAPHORE_SIGNAL_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 27, \
+ .DwordLength = 0
+
+#define GEN9_MI_SEMAPHORE_SIGNAL_length 0x00000002
+
+struct GEN9_MI_SEMAPHORE_SIGNAL {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t PostSyncOperation;
+#define RCS 0
+#define VCS0 1
+#define BCS 2
+#define VECS 3
+#define VCS1 4
+ uint32_t TargetEngineSelect;
+ uint32_t DwordLength;
+ uint32_t TargetContextID;
+};
+
+static inline void
+GEN9_MI_SEMAPHORE_SIGNAL_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_SEMAPHORE_SIGNAL * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->PostSyncOperation, 21, 21) |
+ __gen_field(values->TargetEngineSelect, 15, 17) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->TargetContextID, 0, 31) |
+ 0;
+
+}
+
+#define GEN9_MI_SEMAPHORE_WAIT_length_bias 0x00000002
+#define GEN9_MI_SEMAPHORE_WAIT_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 28, \
+ .DwordLength = 2
+
+#define GEN9_MI_SEMAPHORE_WAIT_length 0x00000004
+
+struct GEN9_MI_SEMAPHORE_WAIT {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define PerProcessGraphicsAddress 0
+#define GlobalGraphicsAddress 1
+ uint32_t MemoryType;
+#define PollingMode 1
+#define SignalMode 0
+ uint32_t WaitMode;
+#define SAD_GREATER_THAN_SDD 0
+#define SAD_GREATER_THAN_OR_EQUAL_SDD 1
+#define SAD_LESS_THAN_SDD 2
+#define SAD_LESS_THAN_OR_EQUAL_SDD 3
+#define SAD_EQUAL_SDD 4
+#define SAD_NOT_EQUAL_SDD 5
+ uint32_t CompareOperation;
+ uint32_t DwordLength;
+ uint32_t SemaphoreDataDword;
+ __gen_address_type SemaphoreAddress;
+};
+
+static inline void
+GEN9_MI_SEMAPHORE_WAIT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_SEMAPHORE_WAIT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->MemoryType, 22, 22) |
+ __gen_field(values->WaitMode, 15, 15) |
+ __gen_field(values->CompareOperation, 12, 14) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->SemaphoreDataDword, 0, 31) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ uint64_t qw2 =
+ __gen_combine_address(data, &dw[2], values->SemaphoreAddress, dw2);
+
+ dw[2] = qw2;
+ dw[3] = qw2 >> 32;
+
+}
+
+#define GEN9_MI_STORE_DATA_IMM_length_bias 0x00000002
+#define GEN9_MI_STORE_DATA_IMM_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 32, \
+ .DwordLength = 2
+
+#define GEN9_MI_STORE_DATA_IMM_length 0x00000004
+
+struct GEN9_MI_STORE_DATA_IMM {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool UseGlobalGTT;
+ bool StoreQword;
+ uint32_t DwordLength;
+ __gen_address_type Address;
+ uint32_t CoreModeEnable;
+ uint32_t DataDWord0;
+ uint32_t DataDWord1;
+};
+
+static inline void
+GEN9_MI_STORE_DATA_IMM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_STORE_DATA_IMM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->UseGlobalGTT, 22, 22) |
+ __gen_field(values->StoreQword, 21, 21) |
+ __gen_field(values->DwordLength, 0, 9) |
+ 0;
+
+ uint32_t dw1 =
+ __gen_field(values->CoreModeEnable, 0, 0) |
+ 0;
+
+ uint64_t qw1 =
+ __gen_combine_address(data, &dw[1], values->Address, dw1);
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+ dw[3] =
+ __gen_field(values->DataDWord0, 0, 31) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->DataDWord1, 0, 31) |
+ 0;
+
+}
+
+#define GEN9_MI_STORE_REGISTER_MEM_length_bias 0x00000002
+#define GEN9_MI_STORE_REGISTER_MEM_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 36, \
+ .DwordLength = 2
+
+#define GEN9_MI_STORE_REGISTER_MEM_length 0x00000004
+
+struct GEN9_MI_STORE_REGISTER_MEM {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool UseGlobalGTT;
+ uint32_t PredicateEnable;
+ uint32_t DwordLength;
+ uint32_t RegisterAddress;
+ __gen_address_type MemoryAddress;
+};
+
+static inline void
+GEN9_MI_STORE_REGISTER_MEM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_STORE_REGISTER_MEM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->UseGlobalGTT, 22, 22) |
+ __gen_field(values->PredicateEnable, 21, 21) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->RegisterAddress, 2, 22) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ uint64_t qw2 =
+ __gen_combine_address(data, &dw[2], values->MemoryAddress, dw2);
+
+ dw[2] = qw2;
+ dw[3] = qw2 >> 32;
+
+}
+
+#define GEN9_PIPELINE_SELECT_length_bias 0x00000001
+#define GEN9_PIPELINE_SELECT_header \
+ .CommandType = 3, \
+ .CommandSubType = 1, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 4
+
+#define GEN9_PIPELINE_SELECT_length 0x00000001
+
+struct GEN9_PIPELINE_SELECT {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t MaskBits;
+ uint32_t ForceMediaAwake;
+ uint32_t MediaSamplerDOPClockGateEnable;
+#define _3D 0
+#define Media 1
+#define GPGPU 2
+ uint32_t PipelineSelection;
+};
+
+static inline void
+GEN9_PIPELINE_SELECT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_PIPELINE_SELECT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->MaskBits, 8, 15) |
+ __gen_field(values->ForceMediaAwake, 5, 5) |
+ __gen_field(values->MediaSamplerDOPClockGateEnable, 4, 4) |
+ __gen_field(values->PipelineSelection, 0, 1) |
+ 0;
+
+}
+
+#define GEN9_STATE_BASE_ADDRESS_length_bias 0x00000002
+#define GEN9_STATE_BASE_ADDRESS_header \
+ .CommandType = 3, \
+ .CommandSubType = 0, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 1, \
+ .DwordLength = 17
+
+#define GEN9_STATE_BASE_ADDRESS_length 0x00000013
+
+#define GEN9_MEMORY_OBJECT_CONTROL_STATE_length 0x00000001
+
+struct GEN9_MEMORY_OBJECT_CONTROL_STATE {
+ uint32_t IndextoMOCSTables;
+};
+
+static inline void
+GEN9_MEMORY_OBJECT_CONTROL_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MEMORY_OBJECT_CONTROL_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->IndextoMOCSTables, 1, 6) |
+ 0;
+
+}
+
+struct GEN9_STATE_BASE_ADDRESS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ __gen_address_type GeneralStateBaseAddress;
+ struct GEN9_MEMORY_OBJECT_CONTROL_STATE GeneralStateMemoryObjectControlState;
+ bool GeneralStateBaseAddressModifyEnable;
+ struct GEN9_MEMORY_OBJECT_CONTROL_STATE StatelessDataPortAccessMemoryObjectControlState;
+ __gen_address_type SurfaceStateBaseAddress;
+ struct GEN9_MEMORY_OBJECT_CONTROL_STATE SurfaceStateMemoryObjectControlState;
+ bool SurfaceStateBaseAddressModifyEnable;
+ __gen_address_type DynamicStateBaseAddress;
+ struct GEN9_MEMORY_OBJECT_CONTROL_STATE DynamicStateMemoryObjectControlState;
+ bool DynamicStateBaseAddressModifyEnable;
+ __gen_address_type IndirectObjectBaseAddress;
+ struct GEN9_MEMORY_OBJECT_CONTROL_STATE IndirectObjectMemoryObjectControlState;
+ bool IndirectObjectBaseAddressModifyEnable;
+ __gen_address_type InstructionBaseAddress;
+ struct GEN9_MEMORY_OBJECT_CONTROL_STATE InstructionMemoryObjectControlState;
+ bool InstructionBaseAddressModifyEnable;
+ uint32_t GeneralStateBufferSize;
+ bool GeneralStateBufferSizeModifyEnable;
+ uint32_t DynamicStateBufferSize;
+ bool DynamicStateBufferSizeModifyEnable;
+ uint32_t IndirectObjectBufferSize;
+ bool IndirectObjectBufferSizeModifyEnable;
+ uint32_t InstructionBufferSize;
+ bool InstructionBuffersizeModifyEnable;
+ __gen_address_type BindlessSurfaceStateBaseAddress;
+ struct GEN9_MEMORY_OBJECT_CONTROL_STATE BindlessSurfaceStateMemoryObjectControlState;
+ bool BindlessSurfaceStateBaseAddressModifyEnable;
+ uint32_t BindlessSurfaceStateSize;
+};
+
+static inline void
+GEN9_STATE_BASE_ADDRESS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_STATE_BASE_ADDRESS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw_GeneralStateMemoryObjectControlState;
+ GEN9_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_GeneralStateMemoryObjectControlState, &values->GeneralStateMemoryObjectControlState);
+ uint32_t dw1 =
+ __gen_field(dw_GeneralStateMemoryObjectControlState, 4, 10) |
+ __gen_field(values->GeneralStateBaseAddressModifyEnable, 0, 0) |
+ 0;
+
+ uint64_t qw1 =
+ __gen_combine_address(data, &dw[1], values->GeneralStateBaseAddress, dw1);
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+ uint32_t dw_StatelessDataPortAccessMemoryObjectControlState;
+ GEN9_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_StatelessDataPortAccessMemoryObjectControlState, &values->StatelessDataPortAccessMemoryObjectControlState);
+ dw[3] =
+ __gen_field(dw_StatelessDataPortAccessMemoryObjectControlState, 16, 22) |
+ 0;
+
+ uint32_t dw_SurfaceStateMemoryObjectControlState;
+ GEN9_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_SurfaceStateMemoryObjectControlState, &values->SurfaceStateMemoryObjectControlState);
+ uint32_t dw4 =
+ __gen_field(dw_SurfaceStateMemoryObjectControlState, 4, 10) |
+ __gen_field(values->SurfaceStateBaseAddressModifyEnable, 0, 0) |
+ 0;
+
+ uint64_t qw4 =
+ __gen_combine_address(data, &dw[4], values->SurfaceStateBaseAddress, dw4);
+
+ dw[4] = qw4;
+ dw[5] = qw4 >> 32;
+
+ uint32_t dw_DynamicStateMemoryObjectControlState;
+ GEN9_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_DynamicStateMemoryObjectControlState, &values->DynamicStateMemoryObjectControlState);
+ uint32_t dw6 =
+ __gen_field(dw_DynamicStateMemoryObjectControlState, 4, 10) |
+ __gen_field(values->DynamicStateBaseAddressModifyEnable, 0, 0) |
+ 0;
+
+ uint64_t qw6 =
+ __gen_combine_address(data, &dw[6], values->DynamicStateBaseAddress, dw6);
+
+ dw[6] = qw6;
+ dw[7] = qw6 >> 32;
+
+ uint32_t dw_IndirectObjectMemoryObjectControlState;
+ GEN9_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_IndirectObjectMemoryObjectControlState, &values->IndirectObjectMemoryObjectControlState);
+ uint32_t dw8 =
+ __gen_field(dw_IndirectObjectMemoryObjectControlState, 4, 10) |
+ __gen_field(values->IndirectObjectBaseAddressModifyEnable, 0, 0) |
+ 0;
+
+ uint64_t qw8 =
+ __gen_combine_address(data, &dw[8], values->IndirectObjectBaseAddress, dw8);
+
+ dw[8] = qw8;
+ dw[9] = qw8 >> 32;
+
+ uint32_t dw_InstructionMemoryObjectControlState;
+ GEN9_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_InstructionMemoryObjectControlState, &values->InstructionMemoryObjectControlState);
+ uint32_t dw10 =
+ __gen_field(dw_InstructionMemoryObjectControlState, 4, 10) |
+ __gen_field(values->InstructionBaseAddressModifyEnable, 0, 0) |
+ 0;
+
+ uint64_t qw10 =
+ __gen_combine_address(data, &dw[10], values->InstructionBaseAddress, dw10);
+
+ dw[10] = qw10;
+ dw[11] = qw10 >> 32;
+
+ dw[12] =
+ __gen_field(values->GeneralStateBufferSize, 12, 31) |
+ __gen_field(values->GeneralStateBufferSizeModifyEnable, 0, 0) |
+ 0;
+
+ dw[13] =
+ __gen_field(values->DynamicStateBufferSize, 12, 31) |
+ __gen_field(values->DynamicStateBufferSizeModifyEnable, 0, 0) |
+ 0;
+
+ dw[14] =
+ __gen_field(values->IndirectObjectBufferSize, 12, 31) |
+ __gen_field(values->IndirectObjectBufferSizeModifyEnable, 0, 0) |
+ 0;
+
+ dw[15] =
+ __gen_field(values->InstructionBufferSize, 12, 31) |
+ __gen_field(values->InstructionBuffersizeModifyEnable, 0, 0) |
+ 0;
+
+ uint32_t dw_BindlessSurfaceStateMemoryObjectControlState;
+ GEN9_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_BindlessSurfaceStateMemoryObjectControlState, &values->BindlessSurfaceStateMemoryObjectControlState);
+ uint32_t dw16 =
+ __gen_field(dw_BindlessSurfaceStateMemoryObjectControlState, 4, 10) |
+ __gen_field(values->BindlessSurfaceStateBaseAddressModifyEnable, 0, 0) |
+ 0;
+
+ uint64_t qw16 =
+ __gen_combine_address(data, &dw[16], values->BindlessSurfaceStateBaseAddress, dw16);
+
+ dw[16] = qw16;
+ dw[17] = qw16 >> 32;
+
+ dw[18] =
+ __gen_field(values->BindlessSurfaceStateSize, 12, 31) |
+ 0;
+
+}
+
+#define GEN9_STATE_PREFETCH_length_bias 0x00000002
+#define GEN9_STATE_PREFETCH_header \
+ .CommandType = 3, \
+ .CommandSubType = 0, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 3, \
+ .DwordLength = 0
+
+#define GEN9_STATE_PREFETCH_length 0x00000002
+
+struct GEN9_STATE_PREFETCH {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ __gen_address_type PrefetchPointer;
+ uint32_t PrefetchCount;
+};
+
+static inline void
+GEN9_STATE_PREFETCH_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_STATE_PREFETCH * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw1 =
+ __gen_field(values->PrefetchCount, 0, 2) |
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->PrefetchPointer, dw1);
+
+}
+
+#define GEN9_STATE_SIP_length_bias 0x00000002
+#define GEN9_STATE_SIP_header \
+ .CommandType = 3, \
+ .CommandSubType = 0, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 2, \
+ .DwordLength = 1
+
+#define GEN9_STATE_SIP_length 0x00000003
+
+struct GEN9_STATE_SIP {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint64_t SystemInstructionPointer;
+};
+
+static inline void
+GEN9_STATE_SIP_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_STATE_SIP * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint64_t qw1 =
+ __gen_offset(values->SystemInstructionPointer, 4, 63) |
+ 0;
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+}
+
+#define GEN9_3DPRIMITIVE_length_bias 0x00000002
+#define GEN9_3DPRIMITIVE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 3, \
+ ._3DCommandSubOpcode = 0, \
+ .DwordLength = 5
+
+#define GEN9_3DPRIMITIVE_length 0x00000007
+
+struct GEN9_3DPRIMITIVE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ bool IndirectParameterEnable;
+ uint32_t UAVCoherencyRequired;
+ bool PredicateEnable;
+ uint32_t DwordLength;
+ bool EndOffsetEnable;
+#define SEQUENTIAL 0
+#define RANDOM 1
+ uint32_t VertexAccessType;
+ uint32_t PrimitiveTopologyType;
+ uint32_t VertexCountPerInstance;
+ uint32_t StartVertexLocation;
+ uint32_t InstanceCount;
+ uint32_t StartInstanceLocation;
+ uint32_t BaseVertexLocation;
+};
+
+static inline void
+GEN9_3DPRIMITIVE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DPRIMITIVE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->IndirectParameterEnable, 10, 10) |
+ __gen_field(values->UAVCoherencyRequired, 9, 9) |
+ __gen_field(values->PredicateEnable, 8, 8) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->EndOffsetEnable, 9, 9) |
+ __gen_field(values->VertexAccessType, 8, 8) |
+ __gen_field(values->PrimitiveTopologyType, 0, 5) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->VertexCountPerInstance, 0, 31) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->StartVertexLocation, 0, 31) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->InstanceCount, 0, 31) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->StartInstanceLocation, 0, 31) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->BaseVertexLocation, 0, 31) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_AA_LINE_PARAMETERS_length_bias 0x00000002
+#define GEN9_3DSTATE_AA_LINE_PARAMETERS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 10, \
+ .DwordLength = 1
+
+#define GEN9_3DSTATE_AA_LINE_PARAMETERS_length 0x00000003
+
+struct GEN9_3DSTATE_AA_LINE_PARAMETERS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ float AAPointCoverageBias;
+ float AACoverageBias;
+ float AAPointCoverageSlope;
+ float AACoverageSlope;
+ float AAPointCoverageEndCapBias;
+ float AACoverageEndCapBias;
+ float AAPointCoverageEndCapSlope;
+ float AACoverageEndCapSlope;
+};
+
+static inline void
+GEN9_3DSTATE_AA_LINE_PARAMETERS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_AA_LINE_PARAMETERS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->AAPointCoverageBias * (1 << 8), 24, 31) |
+ __gen_field(values->AACoverageBias * (1 << 8), 16, 23) |
+ __gen_field(values->AAPointCoverageSlope * (1 << 8), 8, 15) |
+ __gen_field(values->AACoverageSlope * (1 << 8), 0, 7) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->AAPointCoverageEndCapBias * (1 << 8), 24, 31) |
+ __gen_field(values->AACoverageEndCapBias * (1 << 8), 16, 23) |
+ __gen_field(values->AAPointCoverageEndCapSlope * (1 << 8), 8, 15) |
+ __gen_field(values->AACoverageEndCapSlope * (1 << 8), 0, 7) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_BINDING_TABLE_EDIT_DS_length_bias 0x00000002
+#define GEN9_3DSTATE_BINDING_TABLE_EDIT_DS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 70
+
+#define GEN9_3DSTATE_BINDING_TABLE_EDIT_DS_length 0x00000000
+
+#define GEN9_BINDING_TABLE_EDIT_ENTRY_length 0x00000001
+
+struct GEN9_BINDING_TABLE_EDIT_ENTRY {
+ uint32_t BindingTableIndex;
+ uint32_t SurfaceStatePointer;
+};
+
+static inline void
+GEN9_BINDING_TABLE_EDIT_ENTRY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_BINDING_TABLE_EDIT_ENTRY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->BindingTableIndex, 16, 23) |
+ __gen_offset(values->SurfaceStatePointer, 0, 15) |
+ 0;
+
+}
+
+struct GEN9_3DSTATE_BINDING_TABLE_EDIT_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t BindingTableBlockClear;
+#define AllCores 3
+#define Core1 2
+#define Core0 1
+ uint32_t BindingTableEditTarget;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN9_3DSTATE_BINDING_TABLE_EDIT_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_BINDING_TABLE_EDIT_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 8) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->BindingTableBlockClear, 16, 31) |
+ __gen_field(values->BindingTableEditTarget, 0, 1) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN9_3DSTATE_BINDING_TABLE_EDIT_GS_length_bias 0x00000002
+#define GEN9_3DSTATE_BINDING_TABLE_EDIT_GS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 68
+
+#define GEN9_3DSTATE_BINDING_TABLE_EDIT_GS_length 0x00000000
+
+struct GEN9_3DSTATE_BINDING_TABLE_EDIT_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t BindingTableBlockClear;
+#define AllCores 3
+#define Core1 2
+#define Core0 1
+ uint32_t BindingTableEditTarget;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN9_3DSTATE_BINDING_TABLE_EDIT_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_BINDING_TABLE_EDIT_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 8) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->BindingTableBlockClear, 16, 31) |
+ __gen_field(values->BindingTableEditTarget, 0, 1) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN9_3DSTATE_BINDING_TABLE_EDIT_HS_length_bias 0x00000002
+#define GEN9_3DSTATE_BINDING_TABLE_EDIT_HS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 69
+
+#define GEN9_3DSTATE_BINDING_TABLE_EDIT_HS_length 0x00000000
+
+struct GEN9_3DSTATE_BINDING_TABLE_EDIT_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t BindingTableBlockClear;
+#define AllCores 3
+#define Core1 2
+#define Core0 1
+ uint32_t BindingTableEditTarget;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN9_3DSTATE_BINDING_TABLE_EDIT_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_BINDING_TABLE_EDIT_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 8) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->BindingTableBlockClear, 16, 31) |
+ __gen_field(values->BindingTableEditTarget, 0, 1) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN9_3DSTATE_BINDING_TABLE_EDIT_PS_length_bias 0x00000002
+#define GEN9_3DSTATE_BINDING_TABLE_EDIT_PS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 71
+
+#define GEN9_3DSTATE_BINDING_TABLE_EDIT_PS_length 0x00000000
+
+struct GEN9_3DSTATE_BINDING_TABLE_EDIT_PS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t BindingTableBlockClear;
+#define AllCores 3
+#define Core1 2
+#define Core0 1
+ uint32_t BindingTableEditTarget;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN9_3DSTATE_BINDING_TABLE_EDIT_PS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_BINDING_TABLE_EDIT_PS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 8) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->BindingTableBlockClear, 16, 31) |
+ __gen_field(values->BindingTableEditTarget, 0, 1) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN9_3DSTATE_BINDING_TABLE_EDIT_VS_length_bias 0x00000002
+#define GEN9_3DSTATE_BINDING_TABLE_EDIT_VS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 67
+
+#define GEN9_3DSTATE_BINDING_TABLE_EDIT_VS_length 0x00000000
+
+struct GEN9_3DSTATE_BINDING_TABLE_EDIT_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t BindingTableBlockClear;
+#define AllCores 3
+#define Core1 2
+#define Core0 1
+ uint32_t BindingTableEditTarget;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN9_3DSTATE_BINDING_TABLE_EDIT_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_BINDING_TABLE_EDIT_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 8) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->BindingTableBlockClear, 16, 31) |
+ __gen_field(values->BindingTableEditTarget, 0, 1) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN9_3DSTATE_BINDING_TABLE_POINTERS_DS_length_bias 0x00000002
+#define GEN9_3DSTATE_BINDING_TABLE_POINTERS_DS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 40, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_BINDING_TABLE_POINTERS_DS_length 0x00000002
+
+struct GEN9_3DSTATE_BINDING_TABLE_POINTERS_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoDSBindingTable;
+};
+
+static inline void
+GEN9_3DSTATE_BINDING_TABLE_POINTERS_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_BINDING_TABLE_POINTERS_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoDSBindingTable, 5, 15) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_BINDING_TABLE_POINTERS_GS_length_bias 0x00000002
+#define GEN9_3DSTATE_BINDING_TABLE_POINTERS_GS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 41, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_BINDING_TABLE_POINTERS_GS_length 0x00000002
+
+struct GEN9_3DSTATE_BINDING_TABLE_POINTERS_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoGSBindingTable;
+};
+
+static inline void
+GEN9_3DSTATE_BINDING_TABLE_POINTERS_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_BINDING_TABLE_POINTERS_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoGSBindingTable, 5, 15) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_BINDING_TABLE_POINTERS_HS_length_bias 0x00000002
+#define GEN9_3DSTATE_BINDING_TABLE_POINTERS_HS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 39, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_BINDING_TABLE_POINTERS_HS_length 0x00000002
+
+struct GEN9_3DSTATE_BINDING_TABLE_POINTERS_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoHSBindingTable;
+};
+
+static inline void
+GEN9_3DSTATE_BINDING_TABLE_POINTERS_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_BINDING_TABLE_POINTERS_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoHSBindingTable, 5, 15) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_BINDING_TABLE_POINTERS_PS_length_bias 0x00000002
+#define GEN9_3DSTATE_BINDING_TABLE_POINTERS_PS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 42, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_BINDING_TABLE_POINTERS_PS_length 0x00000002
+
+struct GEN9_3DSTATE_BINDING_TABLE_POINTERS_PS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoPSBindingTable;
+};
+
+static inline void
+GEN9_3DSTATE_BINDING_TABLE_POINTERS_PS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_BINDING_TABLE_POINTERS_PS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoPSBindingTable, 5, 15) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_BINDING_TABLE_POINTERS_VS_length_bias 0x00000002
+#define GEN9_3DSTATE_BINDING_TABLE_POINTERS_VS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 38, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_BINDING_TABLE_POINTERS_VS_length 0x00000002
+
+struct GEN9_3DSTATE_BINDING_TABLE_POINTERS_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoVSBindingTable;
+};
+
+static inline void
+GEN9_3DSTATE_BINDING_TABLE_POINTERS_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_BINDING_TABLE_POINTERS_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoVSBindingTable, 5, 15) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_BINDING_TABLE_POOL_ALLOC_length_bias 0x00000002
+#define GEN9_3DSTATE_BINDING_TABLE_POOL_ALLOC_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 25, \
+ .DwordLength = 2
+
+#define GEN9_3DSTATE_BINDING_TABLE_POOL_ALLOC_length 0x00000004
+
+struct GEN9_3DSTATE_BINDING_TABLE_POOL_ALLOC {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ __gen_address_type BindingTablePoolBaseAddress;
+ uint32_t BindingTablePoolEnable;
+ struct GEN9_MEMORY_OBJECT_CONTROL_STATE SurfaceObjectControlState;
+#define NoValidData 0
+ uint32_t BindingTablePoolBufferSize;
+};
+
+static inline void
+GEN9_3DSTATE_BINDING_TABLE_POOL_ALLOC_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_BINDING_TABLE_POOL_ALLOC * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw_SurfaceObjectControlState;
+ GEN9_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_SurfaceObjectControlState, &values->SurfaceObjectControlState);
+ uint32_t dw1 =
+ __gen_field(values->BindingTablePoolEnable, 11, 11) |
+ __gen_field(dw_SurfaceObjectControlState, 0, 6) |
+ 0;
+
+ uint64_t qw1 =
+ __gen_combine_address(data, &dw[1], values->BindingTablePoolBaseAddress, dw1);
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+ dw[3] =
+ __gen_field(values->BindingTablePoolBufferSize, 12, 31) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_BLEND_STATE_POINTERS_length_bias 0x00000002
+#define GEN9_3DSTATE_BLEND_STATE_POINTERS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 36, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_BLEND_STATE_POINTERS_length 0x00000002
+
+struct GEN9_3DSTATE_BLEND_STATE_POINTERS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t BlendStatePointer;
+ bool BlendStatePointerValid;
+};
+
+static inline void
+GEN9_3DSTATE_BLEND_STATE_POINTERS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_BLEND_STATE_POINTERS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->BlendStatePointer, 6, 31) |
+ __gen_field(values->BlendStatePointerValid, 0, 0) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_CC_STATE_POINTERS_length_bias 0x00000002
+#define GEN9_3DSTATE_CC_STATE_POINTERS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 14, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_CC_STATE_POINTERS_length 0x00000002
+
+struct GEN9_3DSTATE_CC_STATE_POINTERS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ColorCalcStatePointer;
+ bool ColorCalcStatePointerValid;
+};
+
+static inline void
+GEN9_3DSTATE_CC_STATE_POINTERS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_CC_STATE_POINTERS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->ColorCalcStatePointer, 6, 31) |
+ __gen_field(values->ColorCalcStatePointerValid, 0, 0) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_CHROMA_KEY_length_bias 0x00000002
+#define GEN9_3DSTATE_CHROMA_KEY_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 4, \
+ .DwordLength = 2
+
+#define GEN9_3DSTATE_CHROMA_KEY_length 0x00000004
+
+struct GEN9_3DSTATE_CHROMA_KEY {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ChromaKeyTableIndex;
+ uint32_t ChromaKeyLowValue;
+ uint32_t ChromaKeyHighValue;
+};
+
+static inline void
+GEN9_3DSTATE_CHROMA_KEY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_CHROMA_KEY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ChromaKeyTableIndex, 30, 31) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ChromaKeyLowValue, 0, 31) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->ChromaKeyHighValue, 0, 31) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_CLEAR_PARAMS_length_bias 0x00000002
+#define GEN9_3DSTATE_CLEAR_PARAMS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 4, \
+ .DwordLength = 1
+
+#define GEN9_3DSTATE_CLEAR_PARAMS_length 0x00000003
+
+struct GEN9_3DSTATE_CLEAR_PARAMS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ float DepthClearValue;
+ bool DepthClearValueValid;
+};
+
+static inline void
+GEN9_3DSTATE_CLEAR_PARAMS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_CLEAR_PARAMS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_float(values->DepthClearValue) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->DepthClearValueValid, 0, 0) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_CLIP_length_bias 0x00000002
+#define GEN9_3DSTATE_CLIP_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 18, \
+ .DwordLength = 2
+
+#define GEN9_3DSTATE_CLIP_length 0x00000004
+
+struct GEN9_3DSTATE_CLIP {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define Normal 0
+#define Force 1
+ bool ForceUserClipDistanceCullTestEnableBitmask;
+#define _8Bit 0
+#define _4Bit 1
+ uint32_t VertexSubPixelPrecisionSelect;
+ bool EarlyCullEnable;
+#define Normal 0
+#define Force 1
+ bool ForceUserClipDistanceClipTestEnableBitmask;
+#define Normal 0
+#define Force 1
+ bool ForceClipMode;
+ bool ClipperStatisticsEnable;
+ uint32_t UserClipDistanceCullTestEnableBitmask;
+ bool ClipEnable;
+#define API_OGL 0
+ uint32_t APIMode;
+ bool ViewportXYClipTestEnable;
+ bool GuardbandClipTestEnable;
+ uint32_t UserClipDistanceClipTestEnableBitmask;
+#define NORMAL 0
+#define REJECT_ALL 3
+#define ACCEPT_ALL 4
+ uint32_t ClipMode;
+ bool PerspectiveDivideDisable;
+ bool NonPerspectiveBarycentricEnable;
+ uint32_t TriangleStripListProvokingVertexSelect;
+ uint32_t LineStripListProvokingVertexSelect;
+ uint32_t TriangleFanProvokingVertexSelect;
+ float MinimumPointWidth;
+ float MaximumPointWidth;
+ bool ForceZeroRTAIndexEnable;
+ uint32_t MaximumVPIndex;
+};
+
+static inline void
+GEN9_3DSTATE_CLIP_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_CLIP * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ForceUserClipDistanceCullTestEnableBitmask, 20, 20) |
+ __gen_field(values->VertexSubPixelPrecisionSelect, 19, 19) |
+ __gen_field(values->EarlyCullEnable, 18, 18) |
+ __gen_field(values->ForceUserClipDistanceClipTestEnableBitmask, 17, 17) |
+ __gen_field(values->ForceClipMode, 16, 16) |
+ __gen_field(values->ClipperStatisticsEnable, 10, 10) |
+ __gen_field(values->UserClipDistanceCullTestEnableBitmask, 0, 7) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ClipEnable, 31, 31) |
+ __gen_field(values->APIMode, 30, 30) |
+ __gen_field(values->ViewportXYClipTestEnable, 28, 28) |
+ __gen_field(values->GuardbandClipTestEnable, 26, 26) |
+ __gen_field(values->UserClipDistanceClipTestEnableBitmask, 16, 23) |
+ __gen_field(values->ClipMode, 13, 15) |
+ __gen_field(values->PerspectiveDivideDisable, 9, 9) |
+ __gen_field(values->NonPerspectiveBarycentricEnable, 8, 8) |
+ __gen_field(values->TriangleStripListProvokingVertexSelect, 4, 5) |
+ __gen_field(values->LineStripListProvokingVertexSelect, 2, 3) |
+ __gen_field(values->TriangleFanProvokingVertexSelect, 0, 1) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->MinimumPointWidth * (1 << 3), 17, 27) |
+ __gen_field(values->MaximumPointWidth * (1 << 3), 6, 16) |
+ __gen_field(values->ForceZeroRTAIndexEnable, 5, 5) |
+ __gen_field(values->MaximumVPIndex, 0, 3) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_CONSTANT_DS_length_bias 0x00000002
+#define GEN9_3DSTATE_CONSTANT_DS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 26, \
+ .DwordLength = 9
+
+#define GEN9_3DSTATE_CONSTANT_DS_length 0x0000000b
+
+#define GEN9_3DSTATE_CONSTANT_BODY_length 0x0000000a
+
+struct GEN9_3DSTATE_CONSTANT_BODY {
+ uint32_t ConstantBuffer1ReadLength;
+ uint32_t ConstantBuffer0ReadLength;
+ uint32_t ConstantBuffer3ReadLength;
+ uint32_t ConstantBuffer2ReadLength;
+ __gen_address_type PointerToConstantBuffer0;
+ __gen_address_type PointerToConstantBuffer1;
+ __gen_address_type PointerToConstantBuffer2;
+ __gen_address_type PointerToConstantBuffer3;
+};
+
+static inline void
+GEN9_3DSTATE_CONSTANT_BODY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_CONSTANT_BODY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->ConstantBuffer1ReadLength, 16, 31) |
+ __gen_field(values->ConstantBuffer0ReadLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBuffer3ReadLength, 16, 31) |
+ __gen_field(values->ConstantBuffer2ReadLength, 0, 15) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ uint64_t qw2 =
+ __gen_combine_address(data, &dw[2], values->PointerToConstantBuffer0, dw2);
+
+ dw[2] = qw2;
+ dw[3] = qw2 >> 32;
+
+ uint32_t dw4 =
+ 0;
+
+ uint64_t qw4 =
+ __gen_combine_address(data, &dw[4], values->PointerToConstantBuffer1, dw4);
+
+ dw[4] = qw4;
+ dw[5] = qw4 >> 32;
+
+ uint32_t dw6 =
+ 0;
+
+ uint64_t qw6 =
+ __gen_combine_address(data, &dw[6], values->PointerToConstantBuffer2, dw6);
+
+ dw[6] = qw6;
+ dw[7] = qw6 >> 32;
+
+ uint32_t dw8 =
+ 0;
+
+ uint64_t qw8 =
+ __gen_combine_address(data, &dw[8], values->PointerToConstantBuffer3, dw8);
+
+ dw[8] = qw8;
+ dw[9] = qw8 >> 32;
+
+}
+
+struct GEN9_3DSTATE_CONSTANT_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ struct GEN9_MEMORY_OBJECT_CONTROL_STATE ConstantBufferObjectControlState;
+ uint32_t DwordLength;
+ struct GEN9_3DSTATE_CONSTANT_BODY ConstantBody;
+};
+
+static inline void
+GEN9_3DSTATE_CONSTANT_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_CONSTANT_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ uint32_t dw_ConstantBufferObjectControlState;
+ GEN9_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_ConstantBufferObjectControlState, &values->ConstantBufferObjectControlState);
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(dw_ConstantBufferObjectControlState, 8, 14) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ GEN9_3DSTATE_CONSTANT_BODY_pack(data, &dw[1], &values->ConstantBody);
+}
+
+#define GEN9_3DSTATE_CONSTANT_GS_length_bias 0x00000002
+#define GEN9_3DSTATE_CONSTANT_GS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 22, \
+ .DwordLength = 9
+
+#define GEN9_3DSTATE_CONSTANT_GS_length 0x0000000b
+
+struct GEN9_3DSTATE_CONSTANT_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ struct GEN9_MEMORY_OBJECT_CONTROL_STATE ConstantBufferObjectControlState;
+ uint32_t DwordLength;
+ struct GEN9_3DSTATE_CONSTANT_BODY ConstantBody;
+};
+
+static inline void
+GEN9_3DSTATE_CONSTANT_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_CONSTANT_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ uint32_t dw_ConstantBufferObjectControlState;
+ GEN9_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_ConstantBufferObjectControlState, &values->ConstantBufferObjectControlState);
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(dw_ConstantBufferObjectControlState, 8, 14) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ GEN9_3DSTATE_CONSTANT_BODY_pack(data, &dw[1], &values->ConstantBody);
+}
+
+#define GEN9_3DSTATE_CONSTANT_HS_length_bias 0x00000002
+#define GEN9_3DSTATE_CONSTANT_HS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 25, \
+ .DwordLength = 9
+
+#define GEN9_3DSTATE_CONSTANT_HS_length 0x0000000b
+
+struct GEN9_3DSTATE_CONSTANT_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ struct GEN9_MEMORY_OBJECT_CONTROL_STATE ConstantBufferObjectControlState;
+ uint32_t DwordLength;
+ struct GEN9_3DSTATE_CONSTANT_BODY ConstantBody;
+};
+
+static inline void
+GEN9_3DSTATE_CONSTANT_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_CONSTANT_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ uint32_t dw_ConstantBufferObjectControlState;
+ GEN9_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_ConstantBufferObjectControlState, &values->ConstantBufferObjectControlState);
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(dw_ConstantBufferObjectControlState, 8, 14) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ GEN9_3DSTATE_CONSTANT_BODY_pack(data, &dw[1], &values->ConstantBody);
+}
+
+#define GEN9_3DSTATE_CONSTANT_PS_length_bias 0x00000002
+#define GEN9_3DSTATE_CONSTANT_PS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 23, \
+ .DwordLength = 9
+
+#define GEN9_3DSTATE_CONSTANT_PS_length 0x0000000b
+
+struct GEN9_3DSTATE_CONSTANT_PS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ struct GEN9_MEMORY_OBJECT_CONTROL_STATE ConstantBufferObjectControlState;
+ uint32_t DwordLength;
+ struct GEN9_3DSTATE_CONSTANT_BODY ConstantBody;
+};
+
+static inline void
+GEN9_3DSTATE_CONSTANT_PS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_CONSTANT_PS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ uint32_t dw_ConstantBufferObjectControlState;
+ GEN9_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_ConstantBufferObjectControlState, &values->ConstantBufferObjectControlState);
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(dw_ConstantBufferObjectControlState, 8, 14) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ GEN9_3DSTATE_CONSTANT_BODY_pack(data, &dw[1], &values->ConstantBody);
+}
+
+#define GEN9_3DSTATE_CONSTANT_VS_length_bias 0x00000002
+#define GEN9_3DSTATE_CONSTANT_VS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 21, \
+ .DwordLength = 9
+
+#define GEN9_3DSTATE_CONSTANT_VS_length 0x0000000b
+
+struct GEN9_3DSTATE_CONSTANT_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ struct GEN9_MEMORY_OBJECT_CONTROL_STATE ConstantBufferObjectControlState;
+ uint32_t DwordLength;
+ struct GEN9_3DSTATE_CONSTANT_BODY ConstantBody;
+};
+
+static inline void
+GEN9_3DSTATE_CONSTANT_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_CONSTANT_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ uint32_t dw_ConstantBufferObjectControlState;
+ GEN9_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_ConstantBufferObjectControlState, &values->ConstantBufferObjectControlState);
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(dw_ConstantBufferObjectControlState, 8, 14) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ GEN9_3DSTATE_CONSTANT_BODY_pack(data, &dw[1], &values->ConstantBody);
+}
+
+#define GEN9_3DSTATE_DEPTH_BUFFER_length_bias 0x00000002
+#define GEN9_3DSTATE_DEPTH_BUFFER_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 5, \
+ .DwordLength = 6
+
+#define GEN9_3DSTATE_DEPTH_BUFFER_length 0x00000008
+
+struct GEN9_3DSTATE_DEPTH_BUFFER {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define SURFTYPE_2D 1
+#define SURFTYPE_CUBE 3
+#define SURFTYPE_NULL 7
+ uint32_t SurfaceType;
+ bool DepthWriteEnable;
+ bool StencilWriteEnable;
+ bool HierarchicalDepthBufferEnable;
+#define D32_FLOAT 1
+#define D24_UNORM_X8_UINT 3
+#define D16_UNORM 5
+ uint32_t SurfaceFormat;
+ uint32_t SurfacePitch;
+ __gen_address_type SurfaceBaseAddress;
+ uint32_t Height;
+ uint32_t Width;
+ uint32_t LOD;
+ uint32_t Depth;
+ uint32_t MinimumArrayElement;
+ struct GEN9_MEMORY_OBJECT_CONTROL_STATE DepthBufferObjectControlState;
+#define NONE 0
+#define TILEYF 1
+#define TILEYS 2
+ uint32_t TiledResourceMode;
+ uint32_t MipTailStartLOD;
+ uint32_t RenderTargetViewExtent;
+ uint32_t SurfaceQPitch;
+};
+
+static inline void
+GEN9_3DSTATE_DEPTH_BUFFER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_DEPTH_BUFFER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->SurfaceType, 29, 31) |
+ __gen_field(values->DepthWriteEnable, 28, 28) |
+ __gen_field(values->StencilWriteEnable, 27, 27) |
+ __gen_field(values->HierarchicalDepthBufferEnable, 22, 22) |
+ __gen_field(values->SurfaceFormat, 18, 20) |
+ __gen_field(values->SurfacePitch, 0, 17) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ uint64_t qw2 =
+ __gen_combine_address(data, &dw[2], values->SurfaceBaseAddress, dw2);
+
+ dw[2] = qw2;
+ dw[3] = qw2 >> 32;
+
+ dw[4] =
+ __gen_field(values->Height, 18, 31) |
+ __gen_field(values->Width, 4, 17) |
+ __gen_field(values->LOD, 0, 3) |
+ 0;
+
+ uint32_t dw_DepthBufferObjectControlState;
+ GEN9_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_DepthBufferObjectControlState, &values->DepthBufferObjectControlState);
+ dw[5] =
+ __gen_field(values->Depth, 21, 31) |
+ __gen_field(values->MinimumArrayElement, 10, 20) |
+ __gen_field(dw_DepthBufferObjectControlState, 0, 6) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->TiledResourceMode, 30, 31) |
+ __gen_field(values->MipTailStartLOD, 26, 29) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->RenderTargetViewExtent, 21, 31) |
+ __gen_field(values->SurfaceQPitch, 0, 14) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_DRAWING_RECTANGLE_length_bias 0x00000002
+#define GEN9_3DSTATE_DRAWING_RECTANGLE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 0, \
+ .DwordLength = 2
+
+#define GEN9_3DSTATE_DRAWING_RECTANGLE_length 0x00000004
+
+struct GEN9_3DSTATE_DRAWING_RECTANGLE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+#define Legacy 0
+#define Core0Enabled 1
+#define Core1Enabled 2
+ uint32_t CoreModeSelect;
+ uint32_t DwordLength;
+ uint32_t ClippedDrawingRectangleYMin;
+ uint32_t ClippedDrawingRectangleXMin;
+ uint32_t ClippedDrawingRectangleYMax;
+ uint32_t ClippedDrawingRectangleXMax;
+ uint32_t DrawingRectangleOriginY;
+ uint32_t DrawingRectangleOriginX;
+};
+
+static inline void
+GEN9_3DSTATE_DRAWING_RECTANGLE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_DRAWING_RECTANGLE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->CoreModeSelect, 14, 15) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ClippedDrawingRectangleYMin, 16, 31) |
+ __gen_field(values->ClippedDrawingRectangleXMin, 0, 15) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ClippedDrawingRectangleYMax, 16, 31) |
+ __gen_field(values->ClippedDrawingRectangleXMax, 0, 15) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->DrawingRectangleOriginY, 16, 31) |
+ __gen_field(values->DrawingRectangleOriginX, 0, 15) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_DS_length_bias 0x00000002
+#define GEN9_3DSTATE_DS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 29, \
+ .DwordLength = 9
+
+#define GEN9_3DSTATE_DS_length 0x0000000b
+
+struct GEN9_3DSTATE_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint64_t KernelStartPointer;
+#define Dmask 0
+#define Vmask 1
+ uint32_t VectorMaskEnable;
+#define NoSamplers 0
+#define _14Samplers 1
+#define _58Samplers 2
+#define _912Samplers 3
+#define _1316Samplers 4
+ uint32_t SamplerCount;
+ uint32_t BindingTableEntryCount;
+#define Normal 0
+#define High 1
+ uint32_t ThreadDispatchPriority;
+#define IEEE754 0
+#define Alternate 1
+ uint32_t FloatingPointMode;
+ bool AccessesUAV;
+ bool IllegalOpcodeExceptionEnable;
+ bool SoftwareExceptionEnable;
+ uint64_t ScratchSpaceBasePointer;
+ uint32_t PerThreadScratchSpace;
+ uint32_t DispatchGRFStartRegisterForURBData;
+ uint32_t PatchURBEntryReadLength;
+ uint32_t PatchURBEntryReadOffset;
+ uint32_t MaximumNumberofThreads;
+ bool StatisticsEnable;
+#define SIMD4X2 0
+#define SIMD8_SINGLE_PATCH 1
+#define SIMD8_SINGLE_OR_DUAL_PATCH 2
+ uint32_t DispatchMode;
+ bool ComputeWCoordinateEnable;
+ bool CacheDisable;
+ bool FunctionEnable;
+ uint32_t VertexURBEntryOutputReadOffset;
+ uint32_t VertexURBEntryOutputLength;
+ uint32_t UserClipDistanceClipTestEnableBitmask;
+ uint32_t UserClipDistanceCullTestEnableBitmask;
+ uint64_t DUAL_PATCHKernelStartPointer;
+};
+
+static inline void
+GEN9_3DSTATE_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint64_t qw1 =
+ __gen_offset(values->KernelStartPointer, 6, 63) |
+ 0;
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+ dw[3] =
+ __gen_field(values->VectorMaskEnable, 30, 30) |
+ __gen_field(values->SamplerCount, 27, 29) |
+ __gen_field(values->BindingTableEntryCount, 18, 25) |
+ __gen_field(values->ThreadDispatchPriority, 17, 17) |
+ __gen_field(values->FloatingPointMode, 16, 16) |
+ __gen_field(values->AccessesUAV, 14, 14) |
+ __gen_field(values->IllegalOpcodeExceptionEnable, 13, 13) |
+ __gen_field(values->SoftwareExceptionEnable, 7, 7) |
+ 0;
+
+ uint64_t qw4 =
+ __gen_offset(values->ScratchSpaceBasePointer, 10, 63) |
+ __gen_field(values->PerThreadScratchSpace, 0, 3) |
+ 0;
+
+ dw[4] = qw4;
+ dw[5] = qw4 >> 32;
+
+ dw[6] =
+ __gen_field(values->DispatchGRFStartRegisterForURBData, 20, 24) |
+ __gen_field(values->PatchURBEntryReadLength, 11, 17) |
+ __gen_field(values->PatchURBEntryReadOffset, 4, 9) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->MaximumNumberofThreads, 21, 29) |
+ __gen_field(values->StatisticsEnable, 10, 10) |
+ __gen_field(values->DispatchMode, 3, 4) |
+ __gen_field(values->ComputeWCoordinateEnable, 2, 2) |
+ __gen_field(values->CacheDisable, 1, 1) |
+ __gen_field(values->FunctionEnable, 0, 0) |
+ 0;
+
+ dw[8] =
+ __gen_field(values->VertexURBEntryOutputReadOffset, 21, 26) |
+ __gen_field(values->VertexURBEntryOutputLength, 16, 20) |
+ __gen_field(values->UserClipDistanceClipTestEnableBitmask, 8, 15) |
+ __gen_field(values->UserClipDistanceCullTestEnableBitmask, 0, 7) |
+ 0;
+
+ uint64_t qw9 =
+ __gen_offset(values->DUAL_PATCHKernelStartPointer, 6, 63) |
+ 0;
+
+ dw[9] = qw9;
+ dw[10] = qw9 >> 32;
+
+}
+
+#define GEN9_3DSTATE_GATHER_CONSTANT_DS_length_bias 0x00000002
+#define GEN9_3DSTATE_GATHER_CONSTANT_DS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 55
+
+#define GEN9_3DSTATE_GATHER_CONSTANT_DS_length 0x00000000
+
+#define GEN9_GATHER_CONSTANT_ENTRY_length 0x00000001
+
+struct GEN9_GATHER_CONSTANT_ENTRY {
+ uint32_t ConstantBufferOffset;
+ uint32_t ChannelMask;
+ uint32_t BindingTableIndexOffset;
+};
+
+static inline void
+GEN9_GATHER_CONSTANT_ENTRY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_GATHER_CONSTANT_ENTRY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_offset(values->ConstantBufferOffset, 8, 15) |
+ __gen_field(values->ChannelMask, 4, 7) |
+ __gen_field(values->BindingTableIndexOffset, 0, 3) |
+ 0;
+
+}
+
+struct GEN9_3DSTATE_GATHER_CONSTANT_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferValid;
+ uint32_t ConstantBufferBindingTableBlock;
+#define CommitGather 0
+#define NonCommitGather 1
+ uint32_t UpdateGatherTableOnly;
+ uint32_t GatherBufferOffset;
+ bool ConstantBufferDx9GenerateStall;
+#define Load 0
+#define Read 1
+ uint32_t OnDieTable;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN9_3DSTATE_GATHER_CONSTANT_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_GATHER_CONSTANT_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferValid, 16, 31) |
+ __gen_field(values->ConstantBufferBindingTableBlock, 12, 15) |
+ __gen_field(values->UpdateGatherTableOnly, 1, 1) |
+ 0;
+
+ dw[2] =
+ __gen_offset(values->GatherBufferOffset, 6, 22) |
+ __gen_field(values->ConstantBufferDx9GenerateStall, 5, 5) |
+ __gen_field(values->OnDieTable, 3, 3) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN9_3DSTATE_GATHER_CONSTANT_GS_length_bias 0x00000002
+#define GEN9_3DSTATE_GATHER_CONSTANT_GS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 53
+
+#define GEN9_3DSTATE_GATHER_CONSTANT_GS_length 0x00000000
+
+struct GEN9_3DSTATE_GATHER_CONSTANT_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferValid;
+ uint32_t ConstantBufferBindingTableBlock;
+#define CommitGather 0
+#define NonCommitGather 1
+ uint32_t UpdateGatherTableOnly;
+ uint32_t GatherBufferOffset;
+ bool ConstantBufferDx9GenerateStall;
+#define Load 0
+#define Read 1
+ uint32_t OnDieTable;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN9_3DSTATE_GATHER_CONSTANT_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_GATHER_CONSTANT_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferValid, 16, 31) |
+ __gen_field(values->ConstantBufferBindingTableBlock, 12, 15) |
+ __gen_field(values->UpdateGatherTableOnly, 1, 1) |
+ 0;
+
+ dw[2] =
+ __gen_offset(values->GatherBufferOffset, 6, 22) |
+ __gen_field(values->ConstantBufferDx9GenerateStall, 5, 5) |
+ __gen_field(values->OnDieTable, 3, 3) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN9_3DSTATE_GATHER_CONSTANT_HS_length_bias 0x00000002
+#define GEN9_3DSTATE_GATHER_CONSTANT_HS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 54
+
+#define GEN9_3DSTATE_GATHER_CONSTANT_HS_length 0x00000000
+
+struct GEN9_3DSTATE_GATHER_CONSTANT_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferValid;
+ uint32_t ConstantBufferBindingTableBlock;
+#define CommitGather 0
+#define NonCommitGather 1
+ uint32_t UpdateGatherTableOnly;
+ uint32_t GatherBufferOffset;
+ bool ConstantBufferDx9GenerateStall;
+#define Load 0
+#define Read 1
+ uint32_t OnDieTable;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN9_3DSTATE_GATHER_CONSTANT_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_GATHER_CONSTANT_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferValid, 16, 31) |
+ __gen_field(values->ConstantBufferBindingTableBlock, 12, 15) |
+ __gen_field(values->UpdateGatherTableOnly, 1, 1) |
+ 0;
+
+ dw[2] =
+ __gen_offset(values->GatherBufferOffset, 6, 22) |
+ __gen_field(values->ConstantBufferDx9GenerateStall, 5, 5) |
+ __gen_field(values->OnDieTable, 3, 3) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN9_3DSTATE_GATHER_CONSTANT_PS_length_bias 0x00000002
+#define GEN9_3DSTATE_GATHER_CONSTANT_PS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 56
+
+#define GEN9_3DSTATE_GATHER_CONSTANT_PS_length 0x00000000
+
+struct GEN9_3DSTATE_GATHER_CONSTANT_PS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferValid;
+ uint32_t ConstantBufferBindingTableBlock;
+#define CommitGather 0
+#define NonCommitGather 1
+ uint32_t UpdateGatherTableOnly;
+ bool DX9OnDieRegisterReadEnable;
+ uint32_t GatherBufferOffset;
+ bool ConstantBufferDx9GenerateStall;
+ bool ConstantBufferDx9Enable;
+#define Load 0
+#define Read 1
+ uint32_t OnDieTable;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN9_3DSTATE_GATHER_CONSTANT_PS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_GATHER_CONSTANT_PS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferValid, 16, 31) |
+ __gen_field(values->ConstantBufferBindingTableBlock, 12, 15) |
+ __gen_field(values->UpdateGatherTableOnly, 1, 1) |
+ __gen_field(values->DX9OnDieRegisterReadEnable, 0, 0) |
+ 0;
+
+ dw[2] =
+ __gen_offset(values->GatherBufferOffset, 6, 22) |
+ __gen_field(values->ConstantBufferDx9GenerateStall, 5, 5) |
+ __gen_field(values->ConstantBufferDx9Enable, 4, 4) |
+ __gen_field(values->OnDieTable, 3, 3) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN9_3DSTATE_GATHER_CONSTANT_VS_length_bias 0x00000002
+#define GEN9_3DSTATE_GATHER_CONSTANT_VS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 52
+
+#define GEN9_3DSTATE_GATHER_CONSTANT_VS_length 0x00000000
+
+struct GEN9_3DSTATE_GATHER_CONSTANT_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferValid;
+ uint32_t ConstantBufferBindingTableBlock;
+#define CommitGather 0
+#define NonCommitGather 1
+ uint32_t UpdateGatherTableOnly;
+ bool DX9OnDieRegisterReadEnable;
+ uint32_t GatherBufferOffset;
+ bool ConstantBufferDx9GenerateStall;
+ bool ConstantBufferDx9Enable;
+#define Load 0
+#define Read 1
+ uint32_t OnDieTable;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN9_3DSTATE_GATHER_CONSTANT_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_GATHER_CONSTANT_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferValid, 16, 31) |
+ __gen_field(values->ConstantBufferBindingTableBlock, 12, 15) |
+ __gen_field(values->UpdateGatherTableOnly, 1, 1) |
+ __gen_field(values->DX9OnDieRegisterReadEnable, 0, 0) |
+ 0;
+
+ dw[2] =
+ __gen_offset(values->GatherBufferOffset, 6, 22) |
+ __gen_field(values->ConstantBufferDx9GenerateStall, 5, 5) |
+ __gen_field(values->ConstantBufferDx9Enable, 4, 4) |
+ __gen_field(values->OnDieTable, 3, 3) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN9_3DSTATE_GATHER_POOL_ALLOC_length_bias 0x00000002
+#define GEN9_3DSTATE_GATHER_POOL_ALLOC_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 26, \
+ .DwordLength = 2
+
+#define GEN9_3DSTATE_GATHER_POOL_ALLOC_length 0x00000004
+
+struct GEN9_3DSTATE_GATHER_POOL_ALLOC {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ __gen_address_type GatherPoolBaseAddress;
+ bool GatherPoolEnable;
+ struct GEN9_MEMORY_OBJECT_CONTROL_STATE MemoryObjectControlState;
+ uint32_t GatherPoolBufferSize;
+};
+
+static inline void
+GEN9_3DSTATE_GATHER_POOL_ALLOC_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_GATHER_POOL_ALLOC * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw_MemoryObjectControlState;
+ GEN9_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_MemoryObjectControlState, &values->MemoryObjectControlState);
+ uint32_t dw1 =
+ __gen_field(values->GatherPoolEnable, 11, 11) |
+ __gen_field(dw_MemoryObjectControlState, 0, 6) |
+ 0;
+
+ uint64_t qw1 =
+ __gen_combine_address(data, &dw[1], values->GatherPoolBaseAddress, dw1);
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+ dw[3] =
+ __gen_field(values->GatherPoolBufferSize, 12, 31) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_GS_length_bias 0x00000002
+#define GEN9_3DSTATE_GS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 17, \
+ .DwordLength = 8
+
+#define GEN9_3DSTATE_GS_length 0x0000000a
+
+struct GEN9_3DSTATE_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint64_t KernelStartPointer;
+ uint32_t SingleProgramFlow;
+#define Dmask 0
+#define Vmask 1
+ uint32_t VectorMaskEnable;
+#define NoSamplers 0
+#define _14Samplers 1
+#define _58Samplers 2
+#define _912Samplers 3
+#define _1316Samplers 4
+ uint32_t SamplerCount;
+ uint32_t BindingTableEntryCount;
+#define Normal 0
+#define High 1
+ uint32_t ThreadDispatchPriority;
+#define IEEE754 0
+#define Alternate 1
+ uint32_t FloatingPointMode;
+ bool IllegalOpcodeExceptionEnable;
+ bool AccessesUAV;
+ bool MaskStackExceptionEnable;
+ bool SoftwareExceptionEnable;
+ uint32_t ExpectedVertexCount;
+ uint64_t ScratchSpaceBasePointer;
+ uint32_t PerThreadScratchSpace;
+ uint32_t DispatchGRFStartRegisterForURBData54;
+ uint32_t OutputVertexSize;
+ uint32_t OutputTopology;
+ uint32_t VertexURBEntryReadLength;
+ bool IncludeVertexHandles;
+ uint32_t VertexURBEntryReadOffset;
+ uint32_t DispatchGRFStartRegisterForURBData;
+ uint32_t ControlDataHeaderSize;
+ uint32_t InstanceControl;
+ uint32_t DefaultStreamId;
+#define DispatchModeSingle 0
+#define DispatchModeDualInstance 1
+#define DispatchModeDualObject 2
+#define DispatchModeSIMD8 3
+ uint32_t DispatchMode;
+ bool StatisticsEnable;
+ uint32_t InvocationsIncrementValue;
+ bool IncludePrimitiveID;
+ uint32_t Hint;
+#define LEADING 0
+#define TRAILING 1
+ uint32_t ReorderMode;
+ bool DiscardAdjacency;
+ bool Enable;
+#define CUT 0
+#define SID 1
+ uint32_t ControlDataFormat;
+ bool StaticOutput;
+ uint32_t StaticOutputVertexCount;
+ uint32_t MaximumNumberofThreads;
+ uint32_t VertexURBEntryOutputReadOffset;
+ uint32_t VertexURBEntryOutputLength;
+ uint32_t UserClipDistanceClipTestEnableBitmask;
+ uint32_t UserClipDistanceCullTestEnableBitmask;
+};
+
+static inline void
+GEN9_3DSTATE_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint64_t qw1 =
+ __gen_offset(values->KernelStartPointer, 6, 63) |
+ 0;
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+ dw[3] =
+ __gen_field(values->SingleProgramFlow, 31, 31) |
+ __gen_field(values->VectorMaskEnable, 30, 30) |
+ __gen_field(values->SamplerCount, 27, 29) |
+ __gen_field(values->BindingTableEntryCount, 18, 25) |
+ __gen_field(values->ThreadDispatchPriority, 17, 17) |
+ __gen_field(values->FloatingPointMode, 16, 16) |
+ __gen_field(values->IllegalOpcodeExceptionEnable, 13, 13) |
+ __gen_field(values->AccessesUAV, 12, 12) |
+ __gen_field(values->MaskStackExceptionEnable, 11, 11) |
+ __gen_field(values->SoftwareExceptionEnable, 7, 7) |
+ __gen_field(values->ExpectedVertexCount, 0, 5) |
+ 0;
+
+ uint64_t qw4 =
+ __gen_offset(values->ScratchSpaceBasePointer, 10, 63) |
+ __gen_field(values->PerThreadScratchSpace, 0, 3) |
+ 0;
+
+ dw[4] = qw4;
+ dw[5] = qw4 >> 32;
+
+ dw[6] =
+ __gen_field(values->DispatchGRFStartRegisterForURBData54, 29, 30) |
+ __gen_field(values->OutputVertexSize, 23, 28) |
+ __gen_field(values->OutputTopology, 17, 22) |
+ __gen_field(values->VertexURBEntryReadLength, 11, 16) |
+ __gen_field(values->IncludeVertexHandles, 10, 10) |
+ __gen_field(values->VertexURBEntryReadOffset, 4, 9) |
+ __gen_field(values->DispatchGRFStartRegisterForURBData, 0, 3) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->ControlDataHeaderSize, 20, 23) |
+ __gen_field(values->InstanceControl, 15, 19) |
+ __gen_field(values->DefaultStreamId, 13, 14) |
+ __gen_field(values->DispatchMode, 11, 12) |
+ __gen_field(values->StatisticsEnable, 10, 10) |
+ __gen_field(values->InvocationsIncrementValue, 5, 9) |
+ __gen_field(values->IncludePrimitiveID, 4, 4) |
+ __gen_field(values->Hint, 3, 3) |
+ __gen_field(values->ReorderMode, 2, 2) |
+ __gen_field(values->DiscardAdjacency, 1, 1) |
+ __gen_field(values->Enable, 0, 0) |
+ 0;
+
+ dw[8] =
+ __gen_field(values->ControlDataFormat, 31, 31) |
+ __gen_field(values->StaticOutput, 30, 30) |
+ __gen_field(values->StaticOutputVertexCount, 16, 26) |
+ __gen_field(values->MaximumNumberofThreads, 0, 8) |
+ 0;
+
+ dw[9] =
+ __gen_field(values->VertexURBEntryOutputReadOffset, 21, 26) |
+ __gen_field(values->VertexURBEntryOutputLength, 16, 20) |
+ __gen_field(values->UserClipDistanceClipTestEnableBitmask, 8, 15) |
+ __gen_field(values->UserClipDistanceCullTestEnableBitmask, 0, 7) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_HIER_DEPTH_BUFFER_length_bias 0x00000002
+#define GEN9_3DSTATE_HIER_DEPTH_BUFFER_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 7, \
+ .DwordLength = 3
+
+#define GEN9_3DSTATE_HIER_DEPTH_BUFFER_length 0x00000005
+
+struct GEN9_3DSTATE_HIER_DEPTH_BUFFER {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ struct GEN9_MEMORY_OBJECT_CONTROL_STATE HierarchicalDepthBufferObjectControlState;
+ uint32_t SurfacePitch;
+ __gen_address_type SurfaceBaseAddress;
+ uint32_t SurfaceQPitch;
+};
+
+static inline void
+GEN9_3DSTATE_HIER_DEPTH_BUFFER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_HIER_DEPTH_BUFFER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw_HierarchicalDepthBufferObjectControlState;
+ GEN9_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_HierarchicalDepthBufferObjectControlState, &values->HierarchicalDepthBufferObjectControlState);
+ dw[1] =
+ __gen_field(dw_HierarchicalDepthBufferObjectControlState, 25, 31) |
+ __gen_field(values->SurfacePitch, 0, 16) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ uint64_t qw2 =
+ __gen_combine_address(data, &dw[2], values->SurfaceBaseAddress, dw2);
+
+ dw[2] = qw2;
+ dw[3] = qw2 >> 32;
+
+ dw[4] =
+ __gen_field(values->SurfaceQPitch, 0, 14) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_HS_length_bias 0x00000002
+#define GEN9_3DSTATE_HS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 27, \
+ .DwordLength = 7
+
+#define GEN9_3DSTATE_HS_length 0x00000009
+
+struct GEN9_3DSTATE_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define NoSamplers 0
+#define _14Samplers 1
+#define _58Samplers 2
+#define _912Samplers 3
+#define _1316Samplers 4
+ uint32_t SamplerCount;
+ uint32_t BindingTableEntryCount;
+#define Normal 0
+#define High 1
+ uint32_t ThreadDispatchPriority;
+#define IEEE754 0
+#define alternate 1
+ uint32_t FloatingPointMode;
+ bool IllegalOpcodeExceptionEnable;
+ bool SoftwareExceptionEnable;
+ bool Enable;
+ bool StatisticsEnable;
+ uint32_t MaximumNumberofThreads;
+ uint32_t InstanceCount;
+ uint64_t KernelStartPointer;
+ uint64_t ScratchSpaceBasePointer;
+ uint32_t PerThreadScratchSpace;
+ uint32_t DispatchGRFStartRegisterForURBData5;
+ bool SingleProgramFlow;
+#define Dmask 0
+#define Vmask 1
+ uint32_t VectorMaskEnable;
+ bool AccessesUAV;
+ bool IncludeVertexHandles;
+ uint32_t DispatchGRFStartRegisterForURBData;
+#define SINGLE_PATCH 0
+#define DUAL_PATCH 1
+#define _8_PATCH 2
+ uint32_t DispatchMode;
+ uint32_t VertexURBEntryReadLength;
+ uint32_t VertexURBEntryReadOffset;
+ bool IncludePrimitiveID;
+};
+
+static inline void
+GEN9_3DSTATE_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->SamplerCount, 27, 29) |
+ __gen_field(values->BindingTableEntryCount, 18, 25) |
+ __gen_field(values->ThreadDispatchPriority, 17, 17) |
+ __gen_field(values->FloatingPointMode, 16, 16) |
+ __gen_field(values->IllegalOpcodeExceptionEnable, 13, 13) |
+ __gen_field(values->SoftwareExceptionEnable, 12, 12) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->Enable, 31, 31) |
+ __gen_field(values->StatisticsEnable, 29, 29) |
+ __gen_field(values->MaximumNumberofThreads, 8, 16) |
+ __gen_field(values->InstanceCount, 0, 3) |
+ 0;
+
+ uint64_t qw3 =
+ __gen_offset(values->KernelStartPointer, 6, 63) |
+ 0;
+
+ dw[3] = qw3;
+ dw[4] = qw3 >> 32;
+
+ uint64_t qw5 =
+ __gen_offset(values->ScratchSpaceBasePointer, 10, 63) |
+ __gen_field(values->PerThreadScratchSpace, 0, 3) |
+ 0;
+
+ dw[5] = qw5;
+ dw[6] = qw5 >> 32;
+
+ dw[7] =
+ __gen_field(values->DispatchGRFStartRegisterForURBData5, 28, 28) |
+ __gen_field(values->SingleProgramFlow, 27, 27) |
+ __gen_field(values->VectorMaskEnable, 26, 26) |
+ __gen_field(values->AccessesUAV, 25, 25) |
+ __gen_field(values->IncludeVertexHandles, 24, 24) |
+ __gen_field(values->DispatchGRFStartRegisterForURBData, 19, 23) |
+ __gen_field(values->DispatchMode, 17, 18) |
+ __gen_field(values->VertexURBEntryReadLength, 11, 16) |
+ __gen_field(values->VertexURBEntryReadOffset, 4, 9) |
+ __gen_field(values->IncludePrimitiveID, 0, 0) |
+ 0;
+
+ dw[8] =
+ 0;
+
+}
+
+#define GEN9_3DSTATE_INDEX_BUFFER_length_bias 0x00000002
+#define GEN9_3DSTATE_INDEX_BUFFER_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 10, \
+ .DwordLength = 3
+
+#define GEN9_3DSTATE_INDEX_BUFFER_length 0x00000005
+
+struct GEN9_3DSTATE_INDEX_BUFFER {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define INDEX_BYTE 0
+#define INDEX_WORD 1
+#define INDEX_DWORD 2
+ uint32_t IndexFormat;
+ struct GEN9_MEMORY_OBJECT_CONTROL_STATE MemoryObjectControlState;
+ __gen_address_type BufferStartingAddress;
+ uint32_t BufferSize;
+};
+
+static inline void
+GEN9_3DSTATE_INDEX_BUFFER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_INDEX_BUFFER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw_MemoryObjectControlState;
+ GEN9_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_MemoryObjectControlState, &values->MemoryObjectControlState);
+ dw[1] =
+ __gen_field(values->IndexFormat, 8, 9) |
+ __gen_field(dw_MemoryObjectControlState, 0, 6) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ uint64_t qw2 =
+ __gen_combine_address(data, &dw[2], values->BufferStartingAddress, dw2);
+
+ dw[2] = qw2;
+ dw[3] = qw2 >> 32;
+
+ dw[4] =
+ __gen_field(values->BufferSize, 0, 31) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_LINE_STIPPLE_length_bias 0x00000002
+#define GEN9_3DSTATE_LINE_STIPPLE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 8, \
+ .DwordLength = 1
+
+#define GEN9_3DSTATE_LINE_STIPPLE_length 0x00000003
+
+struct GEN9_3DSTATE_LINE_STIPPLE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool ModifyEnableCurrentRepeatCounterCurrentStippleIndex;
+ uint32_t CurrentRepeatCounter;
+ uint32_t CurrentStippleIndex;
+ uint32_t LineStipplePattern;
+ float LineStippleInverseRepeatCount;
+ uint32_t LineStippleRepeatCount;
+};
+
+static inline void
+GEN9_3DSTATE_LINE_STIPPLE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_LINE_STIPPLE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ModifyEnableCurrentRepeatCounterCurrentStippleIndex, 31, 31) |
+ __gen_field(values->CurrentRepeatCounter, 21, 29) |
+ __gen_field(values->CurrentStippleIndex, 16, 19) |
+ __gen_field(values->LineStipplePattern, 0, 15) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->LineStippleInverseRepeatCount * (1 << 16), 15, 31) |
+ __gen_field(values->LineStippleRepeatCount, 0, 8) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_MONOFILTER_SIZE_length_bias 0x00000002
+#define GEN9_3DSTATE_MONOFILTER_SIZE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 17, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_MONOFILTER_SIZE_length 0x00000002
+
+struct GEN9_3DSTATE_MONOFILTER_SIZE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t MonochromeFilterWidth;
+ uint32_t MonochromeFilterHeight;
+};
+
+static inline void
+GEN9_3DSTATE_MONOFILTER_SIZE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_MONOFILTER_SIZE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->MonochromeFilterWidth, 3, 5) |
+ __gen_field(values->MonochromeFilterHeight, 0, 2) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_MULTISAMPLE_length_bias 0x00000002
+#define GEN9_3DSTATE_MULTISAMPLE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 13, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_MULTISAMPLE_length 0x00000002
+
+struct GEN9_3DSTATE_MULTISAMPLE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PixelPositionOffsetEnable;
+#define CENTER 0
+#define UL_CORNER 1
+ uint32_t PixelLocation;
+ uint32_t NumberofMultisamples;
+};
+
+static inline void
+GEN9_3DSTATE_MULTISAMPLE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_MULTISAMPLE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->PixelPositionOffsetEnable, 5, 5) |
+ __gen_field(values->PixelLocation, 4, 4) |
+ __gen_field(values->NumberofMultisamples, 1, 3) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_POLY_STIPPLE_OFFSET_length_bias 0x00000002
+#define GEN9_3DSTATE_POLY_STIPPLE_OFFSET_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 6, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_POLY_STIPPLE_OFFSET_length 0x00000002
+
+struct GEN9_3DSTATE_POLY_STIPPLE_OFFSET {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PolygonStippleXOffset;
+ uint32_t PolygonStippleYOffset;
+};
+
+static inline void
+GEN9_3DSTATE_POLY_STIPPLE_OFFSET_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_POLY_STIPPLE_OFFSET * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->PolygonStippleXOffset, 8, 12) |
+ __gen_field(values->PolygonStippleYOffset, 0, 4) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_POLY_STIPPLE_PATTERN_length_bias 0x00000002
+#define GEN9_3DSTATE_POLY_STIPPLE_PATTERN_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 7, \
+ .DwordLength = 31
+
+#define GEN9_3DSTATE_POLY_STIPPLE_PATTERN_length 0x00000021
+
+struct GEN9_3DSTATE_POLY_STIPPLE_PATTERN {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PatternRow[32];
+};
+
+static inline void
+GEN9_3DSTATE_POLY_STIPPLE_PATTERN_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_POLY_STIPPLE_PATTERN * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ for (uint32_t i = 0, j = 1; i < 32; i += 1, j++) {
+ dw[j] =
+ __gen_field(values->PatternRow[i + 0], 0, 31) |
+ 0;
+ }
+
+}
+
+#define GEN9_3DSTATE_PS_length_bias 0x00000002
+#define GEN9_3DSTATE_PS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 32, \
+ .DwordLength = 10
+
+#define GEN9_3DSTATE_PS_length 0x0000000c
+
+struct GEN9_3DSTATE_PS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint64_t KernelStartPointer0;
+#define Multiple 0
+#define Single 1
+ uint32_t SingleProgramFlow;
+#define Dmask 0
+#define Vmask 1
+ uint32_t VectorMaskEnable;
+#define NoSamplers 0
+#define _14Samplers 1
+#define _58Samplers 2
+#define _912Samplers 3
+#define _1316Samplers 4
+ uint32_t SamplerCount;
+#define FlushedtoZero 0
+#define Retained 1
+ uint32_t SinglePrecisionDenormalMode;
+ uint32_t BindingTableEntryCount;
+#define Normal 0
+#define High 1
+ uint32_t ThreadDispatchPriority;
+#define IEEE754 0
+#define Alternate 1
+ uint32_t FloatingPointMode;
+#define RTNE 0
+#define RU 1
+#define RD 2
+#define RTZ 3
+ uint32_t RoundingMode;
+ bool IllegalOpcodeExceptionEnable;
+ bool MaskStackExceptionEnable;
+ bool SoftwareExceptionEnable;
+ uint64_t ScratchSpaceBasePointer;
+ uint32_t PerThreadScratchSpace;
+ uint32_t MaximumNumberofThreadsPerPSD;
+ bool PushConstantEnable;
+ bool RenderTargetFastClearEnable;
+#define RESOLVE_DISABLED 0
+#define RESOLVE_PARTIAL 1
+#define RESOLVE_FULL 3
+ uint32_t RenderTargetResolveType;
+#define POSOFFSET_NONE 0
+#define POSOFFSET_CENTROID 2
+#define POSOFFSET_SAMPLE 3
+ uint32_t PositionXYOffsetSelect;
+ bool _32PixelDispatchEnable;
+ bool _16PixelDispatchEnable;
+ bool _8PixelDispatchEnable;
+ uint32_t DispatchGRFStartRegisterForConstantSetupData0;
+ uint32_t DispatchGRFStartRegisterForConstantSetupData1;
+ uint32_t DispatchGRFStartRegisterForConstantSetupData2;
+ uint64_t KernelStartPointer1;
+ uint64_t KernelStartPointer2;
+};
+
+static inline void
+GEN9_3DSTATE_PS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_PS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint64_t qw1 =
+ __gen_offset(values->KernelStartPointer0, 6, 63) |
+ 0;
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+ dw[3] =
+ __gen_field(values->SingleProgramFlow, 31, 31) |
+ __gen_field(values->VectorMaskEnable, 30, 30) |
+ __gen_field(values->SamplerCount, 27, 29) |
+ __gen_field(values->SinglePrecisionDenormalMode, 26, 26) |
+ __gen_field(values->BindingTableEntryCount, 18, 25) |
+ __gen_field(values->ThreadDispatchPriority, 17, 17) |
+ __gen_field(values->FloatingPointMode, 16, 16) |
+ __gen_field(values->RoundingMode, 14, 15) |
+ __gen_field(values->IllegalOpcodeExceptionEnable, 13, 13) |
+ __gen_field(values->MaskStackExceptionEnable, 11, 11) |
+ __gen_field(values->SoftwareExceptionEnable, 7, 7) |
+ 0;
+
+ uint64_t qw4 =
+ __gen_offset(values->ScratchSpaceBasePointer, 10, 63) |
+ __gen_field(values->PerThreadScratchSpace, 0, 3) |
+ 0;
+
+ dw[4] = qw4;
+ dw[5] = qw4 >> 32;
+
+ dw[6] =
+ __gen_field(values->MaximumNumberofThreadsPerPSD, 23, 31) |
+ __gen_field(values->PushConstantEnable, 11, 11) |
+ __gen_field(values->RenderTargetFastClearEnable, 8, 8) |
+ __gen_field(values->RenderTargetResolveType, 6, 7) |
+ __gen_field(values->PositionXYOffsetSelect, 3, 4) |
+ __gen_field(values->_32PixelDispatchEnable, 2, 2) |
+ __gen_field(values->_16PixelDispatchEnable, 1, 1) |
+ __gen_field(values->_8PixelDispatchEnable, 0, 0) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->DispatchGRFStartRegisterForConstantSetupData0, 16, 22) |
+ __gen_field(values->DispatchGRFStartRegisterForConstantSetupData1, 8, 14) |
+ __gen_field(values->DispatchGRFStartRegisterForConstantSetupData2, 0, 6) |
+ 0;
+
+ uint64_t qw8 =
+ __gen_offset(values->KernelStartPointer1, 6, 63) |
+ 0;
+
+ dw[8] = qw8;
+ dw[9] = qw8 >> 32;
+
+ uint64_t qw10 =
+ __gen_offset(values->KernelStartPointer2, 6, 63) |
+ 0;
+
+ dw[10] = qw10;
+ dw[11] = qw10 >> 32;
+
+}
+
+#define GEN9_3DSTATE_PS_BLEND_length_bias 0x00000002
+#define GEN9_3DSTATE_PS_BLEND_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 77, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_PS_BLEND_length 0x00000002
+
+struct GEN9_3DSTATE_PS_BLEND {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool AlphaToCoverageEnable;
+ bool HasWriteableRT;
+ bool ColorBufferBlendEnable;
+ uint32_t SourceAlphaBlendFactor;
+ uint32_t DestinationAlphaBlendFactor;
+ uint32_t SourceBlendFactor;
+ uint32_t DestinationBlendFactor;
+ bool AlphaTestEnable;
+ bool IndependentAlphaBlendEnable;
+};
+
+static inline void
+GEN9_3DSTATE_PS_BLEND_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_PS_BLEND * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->AlphaToCoverageEnable, 31, 31) |
+ __gen_field(values->HasWriteableRT, 30, 30) |
+ __gen_field(values->ColorBufferBlendEnable, 29, 29) |
+ __gen_field(values->SourceAlphaBlendFactor, 24, 28) |
+ __gen_field(values->DestinationAlphaBlendFactor, 19, 23) |
+ __gen_field(values->SourceBlendFactor, 14, 18) |
+ __gen_field(values->DestinationBlendFactor, 9, 13) |
+ __gen_field(values->AlphaTestEnable, 8, 8) |
+ __gen_field(values->IndependentAlphaBlendEnable, 7, 7) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_PS_EXTRA_length_bias 0x00000002
+#define GEN9_3DSTATE_PS_EXTRA_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 79, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_PS_EXTRA_length 0x00000002
+
+struct GEN9_3DSTATE_PS_EXTRA {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool PixelShaderValid;
+ bool PixelShaderDoesnotwritetoRT;
+ bool oMaskPresenttoRenderTarget;
+ bool PixelShaderKillsPixel;
+#define PSCDEPTH_OFF 0
+#define PSCDEPTH_ON 1
+#define PSCDEPTH_ON_GE 2
+#define PSCDEPTH_ON_LE 3
+ uint32_t PixelShaderComputedDepthMode;
+ bool ForceComputedDepth;
+ bool PixelShaderUsesSourceDepth;
+ bool PixelShaderUsesSourceW;
+ uint32_t Removed;
+ bool AttributeEnable;
+ bool PixelShaderDisablesAlphaToCoverage;
+ bool PixelShaderIsPerSample;
+ bool PixelShaderComputesStencil;
+ bool PixelShaderPullsBary;
+ bool PixelShaderHasUAV;
+#define ICMS_NONE 0
+#define ICMS_NORMAL 1
+#define ICMS_INNER_CONSERVATIVE 2
+#define ICMS_DEPTH_COVERAGE 3
+ uint32_t InputCoverageMaskState;
+};
+
+static inline void
+GEN9_3DSTATE_PS_EXTRA_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_PS_EXTRA * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->PixelShaderValid, 31, 31) |
+ __gen_field(values->PixelShaderDoesnotwritetoRT, 30, 30) |
+ __gen_field(values->oMaskPresenttoRenderTarget, 29, 29) |
+ __gen_field(values->PixelShaderKillsPixel, 28, 28) |
+ __gen_field(values->PixelShaderComputedDepthMode, 26, 27) |
+ __gen_field(values->ForceComputedDepth, 25, 25) |
+ __gen_field(values->PixelShaderUsesSourceDepth, 24, 24) |
+ __gen_field(values->PixelShaderUsesSourceW, 23, 23) |
+ __gen_field(values->Removed, 17, 17) |
+ __gen_field(values->AttributeEnable, 8, 8) |
+ __gen_field(values->PixelShaderDisablesAlphaToCoverage, 7, 7) |
+ __gen_field(values->PixelShaderIsPerSample, 6, 6) |
+ __gen_field(values->PixelShaderComputesStencil, 5, 5) |
+ __gen_field(values->PixelShaderPullsBary, 3, 3) |
+ __gen_field(values->PixelShaderHasUAV, 2, 2) |
+ __gen_field(values->InputCoverageMaskState, 0, 1) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_DS_length_bias 0x00000002
+#define GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_DS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 20, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_DS_length 0x00000002
+
+struct GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferOffset;
+ uint32_t ConstantBufferSize;
+};
+
+static inline void
+GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferOffset, 16, 20) |
+ __gen_field(values->ConstantBufferSize, 0, 5) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_GS_length_bias 0x00000002
+#define GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_GS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 21, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_GS_length 0x00000002
+
+struct GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferOffset;
+ uint32_t ConstantBufferSize;
+};
+
+static inline void
+GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferOffset, 16, 20) |
+ __gen_field(values->ConstantBufferSize, 0, 5) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_HS_length_bias 0x00000002
+#define GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_HS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 19, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_HS_length 0x00000002
+
+struct GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferOffset;
+ uint32_t ConstantBufferSize;
+};
+
+static inline void
+GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferOffset, 16, 20) |
+ __gen_field(values->ConstantBufferSize, 0, 5) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_PS_length_bias 0x00000002
+#define GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_PS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 22, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_PS_length 0x00000002
+
+struct GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_PS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferOffset;
+ uint32_t ConstantBufferSize;
+};
+
+static inline void
+GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_PS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_PS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferOffset, 16, 20) |
+ __gen_field(values->ConstantBufferSize, 0, 5) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_VS_length_bias 0x00000002
+#define GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_VS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 18, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_VS_length 0x00000002
+
+struct GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ConstantBufferOffset;
+ uint32_t ConstantBufferSize;
+};
+
+static inline void
+GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_PUSH_CONSTANT_ALLOC_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ConstantBufferOffset, 16, 20) |
+ __gen_field(values->ConstantBufferSize, 0, 5) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_RASTER_length_bias 0x00000002
+#define GEN9_3DSTATE_RASTER_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 80, \
+ .DwordLength = 3
+
+#define GEN9_3DSTATE_RASTER_length 0x00000005
+
+struct GEN9_3DSTATE_RASTER {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool ViewportZFarClipTestEnable;
+ bool ConservativeRasterizationEnable;
+#define DX9OGL 0
+#define DX100 1
+#define DX101 2
+ uint32_t APIMode;
+#define Clockwise 0
+#define CounterClockwise 1
+ uint32_t FrontWinding;
+#define FSC_NUMRASTSAMPLES_0 0
+#define FSC_NUMRASTSAMPLES_1 1
+#define FSC_NUMRASTSAMPLES_2 2
+#define FSC_NUMRASTSAMPLES_4 3
+#define FSC_NUMRASTSAMPLES_8 4
+#define FSC_NUMRASTSAMPLES_16 5
+ uint32_t ForcedSampleCount;
+#define CULLMODE_BOTH 0
+#define CULLMODE_NONE 1
+#define CULLMODE_FRONT 2
+#define CULLMODE_BACK 3
+ uint32_t CullMode;
+#define Normal 0
+#define Force 1
+ uint32_t ForceMultisampling;
+ bool SmoothPointEnable;
+ bool DXMultisampleRasterizationEnable;
+#define MSRASTMODE_OFF_PIXEL 0
+#define MSRASTMODE_OFF_PATTERN 1
+#define MSRASTMODE_ON_PIXEL 2
+#define MSRASTMODE_ON_PATTERN 3
+ uint32_t DXMultisampleRasterizationMode;
+ bool GlobalDepthOffsetEnableSolid;
+ bool GlobalDepthOffsetEnableWireframe;
+ bool GlobalDepthOffsetEnablePoint;
+#define RASTER_SOLID 0
+#define RASTER_WIREFRAME 1
+#define RASTER_POINT 2
+ uint32_t FrontFaceFillMode;
+#define RASTER_SOLID 0
+#define RASTER_WIREFRAME 1
+#define RASTER_POINT 2
+ uint32_t BackFaceFillMode;
+ bool AntialiasingEnable;
+ bool ScissorRectangleEnable;
+ bool ViewportZNearClipTestEnable;
+ float GlobalDepthOffsetConstant;
+ float GlobalDepthOffsetScale;
+ float GlobalDepthOffsetClamp;
+};
+
+static inline void
+GEN9_3DSTATE_RASTER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_RASTER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ViewportZFarClipTestEnable, 26, 26) |
+ __gen_field(values->ConservativeRasterizationEnable, 24, 24) |
+ __gen_field(values->APIMode, 22, 23) |
+ __gen_field(values->FrontWinding, 21, 21) |
+ __gen_field(values->ForcedSampleCount, 18, 20) |
+ __gen_field(values->CullMode, 16, 17) |
+ __gen_field(values->ForceMultisampling, 14, 14) |
+ __gen_field(values->SmoothPointEnable, 13, 13) |
+ __gen_field(values->DXMultisampleRasterizationEnable, 12, 12) |
+ __gen_field(values->DXMultisampleRasterizationMode, 10, 11) |
+ __gen_field(values->GlobalDepthOffsetEnableSolid, 9, 9) |
+ __gen_field(values->GlobalDepthOffsetEnableWireframe, 8, 8) |
+ __gen_field(values->GlobalDepthOffsetEnablePoint, 7, 7) |
+ __gen_field(values->FrontFaceFillMode, 5, 6) |
+ __gen_field(values->BackFaceFillMode, 3, 4) |
+ __gen_field(values->AntialiasingEnable, 2, 2) |
+ __gen_field(values->ScissorRectangleEnable, 1, 1) |
+ __gen_field(values->ViewportZNearClipTestEnable, 0, 0) |
+ 0;
+
+ dw[2] =
+ __gen_float(values->GlobalDepthOffsetConstant) |
+ 0;
+
+ dw[3] =
+ __gen_float(values->GlobalDepthOffsetScale) |
+ 0;
+
+ dw[4] =
+ __gen_float(values->GlobalDepthOffsetClamp) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_RS_CONSTANT_POINTER_length_bias 0x00000002
+#define GEN9_3DSTATE_RS_CONSTANT_POINTER_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 84, \
+ .DwordLength = 2
+
+#define GEN9_3DSTATE_RS_CONSTANT_POINTER_length 0x00000004
+
+struct GEN9_3DSTATE_RS_CONSTANT_POINTER {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define VS 0
+#define PS 4
+ uint32_t ShaderSelect;
+#define RS_STORE 0
+#define RS_LOAD 1
+ uint32_t OperationLoadorStore;
+ __gen_address_type GlobalConstantBufferAddress;
+ __gen_address_type GlobalConstantBufferAddressHigh;
+};
+
+static inline void
+GEN9_3DSTATE_RS_CONSTANT_POINTER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_RS_CONSTANT_POINTER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ShaderSelect, 28, 30) |
+ __gen_field(values->OperationLoadorStore, 12, 12) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->GlobalConstantBufferAddress, dw2);
+
+ uint32_t dw3 =
+ 0;
+
+ dw[3] =
+ __gen_combine_address(data, &dw[3], values->GlobalConstantBufferAddressHigh, dw3);
+
+}
+
+#define GEN9_3DSTATE_SAMPLER_PALETTE_LOAD0_length_bias 0x00000002
+#define GEN9_3DSTATE_SAMPLER_PALETTE_LOAD0_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 2
+
+#define GEN9_3DSTATE_SAMPLER_PALETTE_LOAD0_length 0x00000000
+
+#define GEN9_PALETTE_ENTRY_length 0x00000001
+
+struct GEN9_PALETTE_ENTRY {
+ uint32_t Alpha;
+ uint32_t Red;
+ uint32_t Green;
+ uint32_t Blue;
+};
+
+static inline void
+GEN9_PALETTE_ENTRY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_PALETTE_ENTRY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->Alpha, 24, 31) |
+ __gen_field(values->Red, 16, 23) |
+ __gen_field(values->Green, 8, 15) |
+ __gen_field(values->Blue, 0, 7) |
+ 0;
+
+}
+
+struct GEN9_3DSTATE_SAMPLER_PALETTE_LOAD0 {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN9_3DSTATE_SAMPLER_PALETTE_LOAD0_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_SAMPLER_PALETTE_LOAD0 * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN9_3DSTATE_SAMPLER_PALETTE_LOAD1_length_bias 0x00000002
+#define GEN9_3DSTATE_SAMPLER_PALETTE_LOAD1_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 12
+
+#define GEN9_3DSTATE_SAMPLER_PALETTE_LOAD1_length 0x00000000
+
+struct GEN9_3DSTATE_SAMPLER_PALETTE_LOAD1 {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN9_3DSTATE_SAMPLER_PALETTE_LOAD1_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_SAMPLER_PALETTE_LOAD1 * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN9_3DSTATE_SAMPLER_STATE_POINTERS_DS_length_bias 0x00000002
+#define GEN9_3DSTATE_SAMPLER_STATE_POINTERS_DS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 45, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_SAMPLER_STATE_POINTERS_DS_length 0x00000002
+
+struct GEN9_3DSTATE_SAMPLER_STATE_POINTERS_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoDSSamplerState;
+};
+
+static inline void
+GEN9_3DSTATE_SAMPLER_STATE_POINTERS_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_SAMPLER_STATE_POINTERS_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoDSSamplerState, 5, 31) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_SAMPLER_STATE_POINTERS_GS_length_bias 0x00000002
+#define GEN9_3DSTATE_SAMPLER_STATE_POINTERS_GS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 46, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_SAMPLER_STATE_POINTERS_GS_length 0x00000002
+
+struct GEN9_3DSTATE_SAMPLER_STATE_POINTERS_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoGSSamplerState;
+};
+
+static inline void
+GEN9_3DSTATE_SAMPLER_STATE_POINTERS_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_SAMPLER_STATE_POINTERS_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoGSSamplerState, 5, 31) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_SAMPLER_STATE_POINTERS_HS_length_bias 0x00000002
+#define GEN9_3DSTATE_SAMPLER_STATE_POINTERS_HS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 44, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_SAMPLER_STATE_POINTERS_HS_length 0x00000002
+
+struct GEN9_3DSTATE_SAMPLER_STATE_POINTERS_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoHSSamplerState;
+};
+
+static inline void
+GEN9_3DSTATE_SAMPLER_STATE_POINTERS_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_SAMPLER_STATE_POINTERS_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoHSSamplerState, 5, 31) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_SAMPLER_STATE_POINTERS_PS_length_bias 0x00000002
+#define GEN9_3DSTATE_SAMPLER_STATE_POINTERS_PS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 47, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_SAMPLER_STATE_POINTERS_PS_length 0x00000002
+
+struct GEN9_3DSTATE_SAMPLER_STATE_POINTERS_PS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoPSSamplerState;
+};
+
+static inline void
+GEN9_3DSTATE_SAMPLER_STATE_POINTERS_PS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_SAMPLER_STATE_POINTERS_PS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoPSSamplerState, 5, 31) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_SAMPLER_STATE_POINTERS_VS_length_bias 0x00000002
+#define GEN9_3DSTATE_SAMPLER_STATE_POINTERS_VS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 43, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_SAMPLER_STATE_POINTERS_VS_length 0x00000002
+
+struct GEN9_3DSTATE_SAMPLER_STATE_POINTERS_VS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PointertoVSSamplerState;
+};
+
+static inline void
+GEN9_3DSTATE_SAMPLER_STATE_POINTERS_VS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_SAMPLER_STATE_POINTERS_VS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->PointertoVSSamplerState, 5, 31) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_SAMPLE_MASK_length_bias 0x00000002
+#define GEN9_3DSTATE_SAMPLE_MASK_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 24, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_SAMPLE_MASK_length 0x00000002
+
+struct GEN9_3DSTATE_SAMPLE_MASK {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t SampleMask;
+};
+
+static inline void
+GEN9_3DSTATE_SAMPLE_MASK_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_SAMPLE_MASK * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->SampleMask, 0, 15) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_SAMPLE_PATTERN_length_bias 0x00000002
+#define GEN9_3DSTATE_SAMPLE_PATTERN_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 28, \
+ .DwordLength = 7
+
+#define GEN9_3DSTATE_SAMPLE_PATTERN_length 0x00000009
+
+struct GEN9_3DSTATE_SAMPLE_PATTERN {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ float _16xSample3XOffset;
+ float _16xSample3YOffset;
+ float _16xSample2XOffset;
+ float _16xSample2YOffset;
+ float _16xSample1XOffset;
+ float _16xSample1YOffset;
+ float _16xSample0XOffset;
+ float _16xSample0YOffset;
+ float _16xSample7XOffset;
+ float _16xSample7YOffset;
+ float _16xSample6XOffset;
+ float _16xSample6YOffset;
+ float _16xSample5XOffset;
+ float _16xSample5YOffset;
+ float _16xSample4XOffset;
+ float _16xSample4YOffset;
+ float _16xSample11XOffset;
+ float _16xSample11YOffset;
+ float _16xSample10XOffset;
+ float _16xSample10YOffset;
+ float _16xSample9XOffset;
+ float _16xSample9YOffset;
+ float _16xSample8XOffset;
+ float _16xSample8YOffset;
+ float _16xSample15XOffset;
+ float _16xSample15YOffset;
+ float _16xSample14XOffset;
+ float _16xSample14YOffset;
+ float _16xSample13XOffset;
+ float _16xSample13YOffset;
+ float _16xSample12XOffset;
+ float _16xSample12YOffset;
+ float _8xSample7XOffset;
+ float _8xSample7YOffset;
+ float _8xSample6XOffset;
+ float _8xSample6YOffset;
+ float _8xSample5XOffset;
+ float _8xSample5YOffset;
+ float _8xSample4XOffset;
+ float _8xSample4YOffset;
+ float _8xSample3XOffset;
+ float _8xSample3YOffset;
+ float _8xSample2XOffset;
+ float _8xSample2YOffset;
+ float _8xSample1XOffset;
+ float _8xSample1YOffset;
+ float _8xSample0XOffset;
+ float _8xSample0YOffset;
+ float _4xSample3XOffset;
+ float _4xSample3YOffset;
+ float _4xSample2XOffset;
+ float _4xSample2YOffset;
+ float _4xSample1XOffset;
+ float _4xSample1YOffset;
+ float _4xSample0XOffset;
+ float _4xSample0YOffset;
+ float _1xSample0XOffset;
+ float _1xSample0YOffset;
+ float _2xSample1XOffset;
+ float _2xSample1YOffset;
+ float _2xSample0XOffset;
+ float _2xSample0YOffset;
+};
+
+static inline void
+GEN9_3DSTATE_SAMPLE_PATTERN_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_SAMPLE_PATTERN * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->_16xSample3XOffset * (1 << 4), 28, 31) |
+ __gen_field(values->_16xSample3YOffset * (1 << 4), 24, 27) |
+ __gen_field(values->_16xSample2XOffset * (1 << 4), 20, 23) |
+ __gen_field(values->_16xSample2YOffset * (1 << 4), 16, 19) |
+ __gen_field(values->_16xSample1XOffset * (1 << 4), 12, 15) |
+ __gen_field(values->_16xSample1YOffset * (1 << 4), 8, 11) |
+ __gen_field(values->_16xSample0XOffset * (1 << 4), 4, 7) |
+ __gen_field(values->_16xSample0YOffset * (1 << 4), 0, 3) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->_16xSample7XOffset * (1 << 4), 28, 31) |
+ __gen_field(values->_16xSample7YOffset * (1 << 4), 24, 27) |
+ __gen_field(values->_16xSample6XOffset * (1 << 4), 20, 23) |
+ __gen_field(values->_16xSample6YOffset * (1 << 4), 16, 19) |
+ __gen_field(values->_16xSample5XOffset * (1 << 4), 12, 15) |
+ __gen_field(values->_16xSample5YOffset * (1 << 4), 8, 11) |
+ __gen_field(values->_16xSample4XOffset * (1 << 4), 4, 7) |
+ __gen_field(values->_16xSample4YOffset * (1 << 4), 0, 3) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->_16xSample11XOffset * (1 << 4), 28, 31) |
+ __gen_field(values->_16xSample11YOffset * (1 << 4), 24, 27) |
+ __gen_field(values->_16xSample10XOffset * (1 << 4), 20, 23) |
+ __gen_field(values->_16xSample10YOffset * (1 << 4), 16, 19) |
+ __gen_field(values->_16xSample9XOffset * (1 << 4), 12, 15) |
+ __gen_field(values->_16xSample9YOffset * (1 << 4), 8, 11) |
+ __gen_field(values->_16xSample8XOffset * (1 << 4), 4, 7) |
+ __gen_field(values->_16xSample8YOffset * (1 << 4), 0, 3) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->_16xSample15XOffset * (1 << 4), 28, 31) |
+ __gen_field(values->_16xSample15YOffset * (1 << 4), 24, 27) |
+ __gen_field(values->_16xSample14XOffset * (1 << 4), 20, 23) |
+ __gen_field(values->_16xSample14YOffset * (1 << 4), 16, 19) |
+ __gen_field(values->_16xSample13XOffset * (1 << 4), 12, 15) |
+ __gen_field(values->_16xSample13YOffset * (1 << 4), 8, 11) |
+ __gen_field(values->_16xSample12XOffset * (1 << 4), 4, 7) |
+ __gen_field(values->_16xSample12YOffset * (1 << 4), 0, 3) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->_8xSample7XOffset * (1 << 4), 28, 31) |
+ __gen_field(values->_8xSample7YOffset * (1 << 4), 24, 27) |
+ __gen_field(values->_8xSample6XOffset * (1 << 4), 20, 23) |
+ __gen_field(values->_8xSample6YOffset * (1 << 4), 16, 19) |
+ __gen_field(values->_8xSample5XOffset * (1 << 4), 12, 15) |
+ __gen_field(values->_8xSample5YOffset * (1 << 4), 8, 11) |
+ __gen_field(values->_8xSample4XOffset * (1 << 4), 4, 7) |
+ __gen_field(values->_8xSample4YOffset * (1 << 4), 0, 3) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->_8xSample3XOffset * (1 << 4), 28, 31) |
+ __gen_field(values->_8xSample3YOffset * (1 << 4), 24, 27) |
+ __gen_field(values->_8xSample2XOffset * (1 << 4), 20, 23) |
+ __gen_field(values->_8xSample2YOffset * (1 << 4), 16, 19) |
+ __gen_field(values->_8xSample1XOffset * (1 << 4), 12, 15) |
+ __gen_field(values->_8xSample1YOffset * (1 << 4), 8, 11) |
+ __gen_field(values->_8xSample0XOffset * (1 << 4), 4, 7) |
+ __gen_field(values->_8xSample0YOffset * (1 << 4), 0, 3) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->_4xSample3XOffset * (1 << 4), 28, 31) |
+ __gen_field(values->_4xSample3YOffset * (1 << 4), 24, 27) |
+ __gen_field(values->_4xSample2XOffset * (1 << 4), 20, 23) |
+ __gen_field(values->_4xSample2YOffset * (1 << 4), 16, 19) |
+ __gen_field(values->_4xSample1XOffset * (1 << 4), 12, 15) |
+ __gen_field(values->_4xSample1YOffset * (1 << 4), 8, 11) |
+ __gen_field(values->_4xSample0XOffset * (1 << 4), 4, 7) |
+ __gen_field(values->_4xSample0YOffset * (1 << 4), 0, 3) |
+ 0;
+
+ dw[8] =
+ __gen_field(values->_1xSample0XOffset * (1 << 4), 20, 23) |
+ __gen_field(values->_1xSample0YOffset * (1 << 4), 16, 19) |
+ __gen_field(values->_2xSample1XOffset * (1 << 4), 12, 15) |
+ __gen_field(values->_2xSample1YOffset * (1 << 4), 8, 11) |
+ __gen_field(values->_2xSample0XOffset * (1 << 4), 4, 7) |
+ __gen_field(values->_2xSample0YOffset * (1 << 4), 0, 3) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_SBE_length_bias 0x00000002
+#define GEN9_3DSTATE_SBE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 31, \
+ .DwordLength = 4
+
+#define GEN9_3DSTATE_SBE_length 0x00000006
+
+struct GEN9_3DSTATE_SBE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool ForceVertexURBEntryReadLength;
+ bool ForceVertexURBEntryReadOffset;
+ uint32_t NumberofSFOutputAttributes;
+ bool AttributeSwizzleEnable;
+#define UPPERLEFT 0
+#define LOWERLEFT 1
+ uint32_t PointSpriteTextureCoordinateOrigin;
+ bool PrimitiveIDOverrideComponentW;
+ bool PrimitiveIDOverrideComponentZ;
+ bool PrimitiveIDOverrideComponentY;
+ bool PrimitiveIDOverrideComponentX;
+ uint32_t VertexURBEntryReadLength;
+ uint32_t VertexURBEntryReadOffset;
+ uint32_t PrimitiveIDOverrideAttributeSelect;
+ uint32_t PointSpriteTextureCoordinateEnable;
+ uint32_t ConstantInterpolationEnable;
+ uint32_t Attribute15ActiveComponentFormat;
+ uint32_t Attribute14ActiveComponentFormat;
+ uint32_t Attribute13ActiveComponentFormat;
+ uint32_t Attribute12ActiveComponentFormat;
+ uint32_t Attribute11ActiveComponentFormat;
+ uint32_t Attribute10ActiveComponentFormat;
+ uint32_t Attribute9ActiveComponentFormat;
+ uint32_t Attribute8ActiveComponentFormat;
+ uint32_t Attribute7ActiveComponentFormat;
+ uint32_t Attribute6ActiveComponentFormat;
+ uint32_t Attribute5ActiveComponentFormat;
+ uint32_t Attribute4ActiveComponentFormat;
+ uint32_t Attribute3ActiveComponentFormat;
+ uint32_t Attribute2ActiveComponentFormat;
+ uint32_t Attribute1ActiveComponentFormat;
+ uint32_t Attribute0ActiveComponentFormat;
+ uint32_t Attribute31ActiveComponentFormat;
+ uint32_t Attribute30ActiveComponentFormat;
+ uint32_t Attribute29ActiveComponentFormat;
+ uint32_t Attribute28ActiveComponentFormat;
+ uint32_t Attribute27ActiveComponentFormat;
+ uint32_t Attribute26ActiveComponentFormat;
+ uint32_t Attribute25ActiveComponentFormat;
+ uint32_t Attribute24ActiveComponentFormat;
+ uint32_t Attribute23ActiveComponentFormat;
+ uint32_t Attribute22ActiveComponentFormat;
+ uint32_t Attribute21ActiveComponentFormat;
+ uint32_t Attribute20ActiveComponentFormat;
+ uint32_t Attribute19ActiveComponentFormat;
+ uint32_t Attribute18ActiveComponentFormat;
+ uint32_t Attribute17ActiveComponentFormat;
+ uint32_t Attribute16ActiveComponentFormat;
+};
+
+static inline void
+GEN9_3DSTATE_SBE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_SBE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ForceVertexURBEntryReadLength, 29, 29) |
+ __gen_field(values->ForceVertexURBEntryReadOffset, 28, 28) |
+ __gen_field(values->NumberofSFOutputAttributes, 22, 27) |
+ __gen_field(values->AttributeSwizzleEnable, 21, 21) |
+ __gen_field(values->PointSpriteTextureCoordinateOrigin, 20, 20) |
+ __gen_field(values->PrimitiveIDOverrideComponentW, 19, 19) |
+ __gen_field(values->PrimitiveIDOverrideComponentZ, 18, 18) |
+ __gen_field(values->PrimitiveIDOverrideComponentY, 17, 17) |
+ __gen_field(values->PrimitiveIDOverrideComponentX, 16, 16) |
+ __gen_field(values->VertexURBEntryReadLength, 11, 15) |
+ __gen_field(values->VertexURBEntryReadOffset, 5, 10) |
+ __gen_field(values->PrimitiveIDOverrideAttributeSelect, 0, 4) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->PointSpriteTextureCoordinateEnable, 0, 31) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->ConstantInterpolationEnable, 0, 31) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->Attribute15ActiveComponentFormat, 30, 31) |
+ __gen_field(values->Attribute14ActiveComponentFormat, 28, 29) |
+ __gen_field(values->Attribute13ActiveComponentFormat, 26, 27) |
+ __gen_field(values->Attribute12ActiveComponentFormat, 24, 25) |
+ __gen_field(values->Attribute11ActiveComponentFormat, 22, 23) |
+ __gen_field(values->Attribute10ActiveComponentFormat, 20, 21) |
+ __gen_field(values->Attribute9ActiveComponentFormat, 18, 19) |
+ __gen_field(values->Attribute8ActiveComponentFormat, 16, 17) |
+ __gen_field(values->Attribute7ActiveComponentFormat, 14, 15) |
+ __gen_field(values->Attribute6ActiveComponentFormat, 12, 13) |
+ __gen_field(values->Attribute5ActiveComponentFormat, 10, 11) |
+ __gen_field(values->Attribute4ActiveComponentFormat, 8, 9) |
+ __gen_field(values->Attribute3ActiveComponentFormat, 6, 7) |
+ __gen_field(values->Attribute2ActiveComponentFormat, 4, 5) |
+ __gen_field(values->Attribute1ActiveComponentFormat, 2, 3) |
+ __gen_field(values->Attribute0ActiveComponentFormat, 0, 1) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->Attribute31ActiveComponentFormat, 30, 31) |
+ __gen_field(values->Attribute30ActiveComponentFormat, 28, 29) |
+ __gen_field(values->Attribute29ActiveComponentFormat, 26, 27) |
+ __gen_field(values->Attribute28ActiveComponentFormat, 24, 25) |
+ __gen_field(values->Attribute27ActiveComponentFormat, 22, 23) |
+ __gen_field(values->Attribute26ActiveComponentFormat, 20, 21) |
+ __gen_field(values->Attribute25ActiveComponentFormat, 18, 19) |
+ __gen_field(values->Attribute24ActiveComponentFormat, 16, 17) |
+ __gen_field(values->Attribute23ActiveComponentFormat, 14, 15) |
+ __gen_field(values->Attribute22ActiveComponentFormat, 12, 13) |
+ __gen_field(values->Attribute21ActiveComponentFormat, 10, 11) |
+ __gen_field(values->Attribute20ActiveComponentFormat, 8, 9) |
+ __gen_field(values->Attribute19ActiveComponentFormat, 6, 7) |
+ __gen_field(values->Attribute18ActiveComponentFormat, 4, 5) |
+ __gen_field(values->Attribute17ActiveComponentFormat, 2, 3) |
+ __gen_field(values->Attribute16ActiveComponentFormat, 0, 1) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_SBE_SWIZ_length_bias 0x00000002
+#define GEN9_3DSTATE_SBE_SWIZ_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 81, \
+ .DwordLength = 9
+
+#define GEN9_3DSTATE_SBE_SWIZ_length 0x0000000b
+
+#define GEN9_SF_OUTPUT_ATTRIBUTE_DETAIL_length 0x00000001
+
+struct GEN9_SF_OUTPUT_ATTRIBUTE_DETAIL {
+ bool ComponentOverrideW;
+ bool ComponentOverrideZ;
+ bool ComponentOverrideY;
+ bool ComponentOverrideX;
+ uint32_t SwizzleControlMode;
+#define CONST_0000 0
+#define CONST_0001_FLOAT 1
+#define CONST_1111_FLOAT 2
+#define PRIM_ID 3
+ uint32_t ConstantSource;
+#define INPUTATTR 0
+#define INPUTATTR_FACING 1
+#define INPUTATTR_W 2
+#define INPUTATTR_FACING_W 3
+ uint32_t SwizzleSelect;
+ uint32_t SourceAttribute;
+};
+
+static inline void
+GEN9_SF_OUTPUT_ATTRIBUTE_DETAIL_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_SF_OUTPUT_ATTRIBUTE_DETAIL * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->ComponentOverrideW, 15, 15) |
+ __gen_field(values->ComponentOverrideZ, 14, 14) |
+ __gen_field(values->ComponentOverrideY, 13, 13) |
+ __gen_field(values->ComponentOverrideX, 12, 12) |
+ __gen_field(values->SwizzleControlMode, 11, 11) |
+ __gen_field(values->ConstantSource, 9, 10) |
+ __gen_field(values->SwizzleSelect, 6, 7) |
+ __gen_field(values->SourceAttribute, 0, 4) |
+ 0;
+
+}
+
+struct GEN9_3DSTATE_SBE_SWIZ {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ struct GEN9_SF_OUTPUT_ATTRIBUTE_DETAIL Attribute[16];
+ uint32_t AttributeWrapShortestEnables[16];
+};
+
+static inline void
+GEN9_3DSTATE_SBE_SWIZ_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_SBE_SWIZ * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ for (uint32_t i = 0, j = 1; i < 16; i += 2, j++) {
+ uint32_t dw_Attribute0;
+ GEN9_SF_OUTPUT_ATTRIBUTE_DETAIL_pack(data, &dw_Attribute0, &values->Attribute[i + 0]);
+ uint32_t dw_Attribute1;
+ GEN9_SF_OUTPUT_ATTRIBUTE_DETAIL_pack(data, &dw_Attribute1, &values->Attribute[i + 1]);
+ dw[j] =
+ __gen_field(dw_Attribute0, 0, 15) |
+ __gen_field(dw_Attribute1, 16, 31) |
+ 0;
+ }
+
+ for (uint32_t i = 0, j = 9; i < 16; i += 8, j++) {
+ dw[j] =
+ __gen_field(values->AttributeWrapShortestEnables[i + 0], 0, 3) |
+ __gen_field(values->AttributeWrapShortestEnables[i + 1], 4, 7) |
+ __gen_field(values->AttributeWrapShortestEnables[i + 2], 8, 11) |
+ __gen_field(values->AttributeWrapShortestEnables[i + 3], 12, 15) |
+ __gen_field(values->AttributeWrapShortestEnables[i + 4], 16, 19) |
+ __gen_field(values->AttributeWrapShortestEnables[i + 5], 20, 23) |
+ __gen_field(values->AttributeWrapShortestEnables[i + 6], 24, 27) |
+ __gen_field(values->AttributeWrapShortestEnables[i + 7], 28, 31) |
+ 0;
+ }
+
+}
+
+#define GEN9_3DSTATE_SCISSOR_STATE_POINTERS_length_bias 0x00000002
+#define GEN9_3DSTATE_SCISSOR_STATE_POINTERS_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 15, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_SCISSOR_STATE_POINTERS_length 0x00000002
+
+struct GEN9_3DSTATE_SCISSOR_STATE_POINTERS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t ScissorRectPointer;
+};
+
+static inline void
+GEN9_3DSTATE_SCISSOR_STATE_POINTERS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_SCISSOR_STATE_POINTERS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->ScissorRectPointer, 5, 31) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_SF_length_bias 0x00000002
+#define GEN9_3DSTATE_SF_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 19, \
+ .DwordLength = 2
+
+#define GEN9_3DSTATE_SF_length 0x00000004
+
+struct GEN9_3DSTATE_SF {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ float LineWidth;
+ bool LegacyGlobalDepthBiasEnable;
+ bool StatisticsEnable;
+ bool ViewportTransformEnable;
+#define _05pixels 0
+#define _10pixels 1
+#define _20pixels 2
+#define _40pixels 3
+ uint32_t LineEndCapAntialiasingRegionWidth;
+ bool LastPixelEnable;
+ uint32_t TriangleStripListProvokingVertexSelect;
+ uint32_t LineStripListProvokingVertexSelect;
+ uint32_t TriangleFanProvokingVertexSelect;
+#define AALINEDISTANCE_TRUE 1
+ uint32_t AALineDistanceMode;
+ bool SmoothPointEnable;
+ uint32_t VertexSubPixelPrecisionSelect;
+#define Vertex 0
+#define State 1
+ uint32_t PointWidthSource;
+ float PointWidth;
+};
+
+static inline void
+GEN9_3DSTATE_SF_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_SF * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->LineWidth * (1 << 7), 12, 29) |
+ __gen_field(values->LegacyGlobalDepthBiasEnable, 11, 11) |
+ __gen_field(values->StatisticsEnable, 10, 10) |
+ __gen_field(values->ViewportTransformEnable, 1, 1) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->LineEndCapAntialiasingRegionWidth, 16, 17) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->LastPixelEnable, 31, 31) |
+ __gen_field(values->TriangleStripListProvokingVertexSelect, 29, 30) |
+ __gen_field(values->LineStripListProvokingVertexSelect, 27, 28) |
+ __gen_field(values->TriangleFanProvokingVertexSelect, 25, 26) |
+ __gen_field(values->AALineDistanceMode, 14, 14) |
+ __gen_field(values->SmoothPointEnable, 13, 13) |
+ __gen_field(values->VertexSubPixelPrecisionSelect, 12, 12) |
+ __gen_field(values->PointWidthSource, 11, 11) |
+ __gen_field(values->PointWidth * (1 << 3), 0, 10) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_SO_BUFFER_length_bias 0x00000002
+#define GEN9_3DSTATE_SO_BUFFER_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 24, \
+ .DwordLength = 6
+
+#define GEN9_3DSTATE_SO_BUFFER_length 0x00000008
+
+struct GEN9_3DSTATE_SO_BUFFER {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool SOBufferEnable;
+ uint32_t SOBufferIndex;
+ struct GEN9_MEMORY_OBJECT_CONTROL_STATE SOBufferObjectControlState;
+ bool StreamOffsetWriteEnable;
+ bool StreamOutputBufferOffsetAddressEnable;
+ __gen_address_type SurfaceBaseAddress;
+ uint32_t SurfaceSize;
+ __gen_address_type StreamOutputBufferOffsetAddress;
+ uint32_t StreamOffset;
+};
+
+static inline void
+GEN9_3DSTATE_SO_BUFFER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_SO_BUFFER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw_SOBufferObjectControlState;
+ GEN9_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_SOBufferObjectControlState, &values->SOBufferObjectControlState);
+ dw[1] =
+ __gen_field(values->SOBufferEnable, 31, 31) |
+ __gen_field(values->SOBufferIndex, 29, 30) |
+ __gen_field(dw_SOBufferObjectControlState, 22, 28) |
+ __gen_field(values->StreamOffsetWriteEnable, 21, 21) |
+ __gen_field(values->StreamOutputBufferOffsetAddressEnable, 20, 20) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ uint64_t qw2 =
+ __gen_combine_address(data, &dw[2], values->SurfaceBaseAddress, dw2);
+
+ dw[2] = qw2;
+ dw[3] = qw2 >> 32;
+
+ dw[4] =
+ __gen_field(values->SurfaceSize, 0, 29) |
+ 0;
+
+ uint32_t dw5 =
+ 0;
+
+ uint64_t qw5 =
+ __gen_combine_address(data, &dw[5], values->StreamOutputBufferOffsetAddress, dw5);
+
+ dw[5] = qw5;
+ dw[6] = qw5 >> 32;
+
+ dw[7] =
+ __gen_field(values->StreamOffset, 0, 31) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_SO_DECL_LIST_length_bias 0x00000002
+#define GEN9_3DSTATE_SO_DECL_LIST_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 23
+
+#define GEN9_3DSTATE_SO_DECL_LIST_length 0x00000000
+
+#define GEN9_SO_DECL_ENTRY_length 0x00000002
+
+#define GEN9_SO_DECL_length 0x00000001
+
+struct GEN9_SO_DECL {
+ uint32_t OutputBufferSlot;
+ uint32_t HoleFlag;
+ uint32_t RegisterIndex;
+ uint32_t ComponentMask;
+};
+
+static inline void
+GEN9_SO_DECL_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_SO_DECL * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->OutputBufferSlot, 12, 13) |
+ __gen_field(values->HoleFlag, 11, 11) |
+ __gen_field(values->RegisterIndex, 4, 9) |
+ __gen_field(values->ComponentMask, 0, 3) |
+ 0;
+
+}
+
+struct GEN9_SO_DECL_ENTRY {
+ struct GEN9_SO_DECL Stream3Decl;
+ struct GEN9_SO_DECL Stream2Decl;
+ struct GEN9_SO_DECL Stream1Decl;
+ struct GEN9_SO_DECL Stream0Decl;
+};
+
+static inline void
+GEN9_SO_DECL_ENTRY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_SO_DECL_ENTRY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ uint32_t dw_Stream3Decl;
+ GEN9_SO_DECL_pack(data, &dw_Stream3Decl, &values->Stream3Decl);
+ uint32_t dw_Stream2Decl;
+ GEN9_SO_DECL_pack(data, &dw_Stream2Decl, &values->Stream2Decl);
+ uint32_t dw_Stream1Decl;
+ GEN9_SO_DECL_pack(data, &dw_Stream1Decl, &values->Stream1Decl);
+ uint32_t dw_Stream0Decl;
+ GEN9_SO_DECL_pack(data, &dw_Stream0Decl, &values->Stream0Decl);
+ uint64_t qw0 =
+ __gen_field(dw_Stream3Decl, 48, 63) |
+ __gen_field(dw_Stream2Decl, 32, 47) |
+ __gen_field(dw_Stream1Decl, 16, 31) |
+ __gen_field(dw_Stream0Decl, 0, 15) |
+ 0;
+
+ dw[0] = qw0;
+ dw[1] = qw0 >> 32;
+
+}
+
+struct GEN9_3DSTATE_SO_DECL_LIST {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t StreamtoBufferSelects3;
+ uint32_t StreamtoBufferSelects2;
+ uint32_t StreamtoBufferSelects1;
+ uint32_t StreamtoBufferSelects0;
+ uint32_t NumEntries3;
+ uint32_t NumEntries2;
+ uint32_t NumEntries1;
+ uint32_t NumEntries0;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN9_3DSTATE_SO_DECL_LIST_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_SO_DECL_LIST * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 8) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->StreamtoBufferSelects3, 12, 15) |
+ __gen_field(values->StreamtoBufferSelects2, 8, 11) |
+ __gen_field(values->StreamtoBufferSelects1, 4, 7) |
+ __gen_field(values->StreamtoBufferSelects0, 0, 3) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->NumEntries3, 24, 31) |
+ __gen_field(values->NumEntries2, 16, 23) |
+ __gen_field(values->NumEntries1, 8, 15) |
+ __gen_field(values->NumEntries0, 0, 7) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN9_3DSTATE_STENCIL_BUFFER_length_bias 0x00000002
+#define GEN9_3DSTATE_STENCIL_BUFFER_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 6, \
+ .DwordLength = 3
+
+#define GEN9_3DSTATE_STENCIL_BUFFER_length 0x00000005
+
+struct GEN9_3DSTATE_STENCIL_BUFFER {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t StencilBufferEnable;
+ struct GEN9_MEMORY_OBJECT_CONTROL_STATE StencilBufferObjectControlState;
+ uint32_t SurfacePitch;
+ __gen_address_type SurfaceBaseAddress;
+ uint32_t SurfaceQPitch;
+};
+
+static inline void
+GEN9_3DSTATE_STENCIL_BUFFER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_STENCIL_BUFFER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw_StencilBufferObjectControlState;
+ GEN9_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_StencilBufferObjectControlState, &values->StencilBufferObjectControlState);
+ dw[1] =
+ __gen_field(values->StencilBufferEnable, 31, 31) |
+ __gen_field(dw_StencilBufferObjectControlState, 22, 28) |
+ __gen_field(values->SurfacePitch, 0, 16) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ uint64_t qw2 =
+ __gen_combine_address(data, &dw[2], values->SurfaceBaseAddress, dw2);
+
+ dw[2] = qw2;
+ dw[3] = qw2 >> 32;
+
+ dw[4] =
+ __gen_field(values->SurfaceQPitch, 0, 14) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_STREAMOUT_length_bias 0x00000002
+#define GEN9_3DSTATE_STREAMOUT_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 30, \
+ .DwordLength = 3
+
+#define GEN9_3DSTATE_STREAMOUT_length 0x00000005
+
+struct GEN9_3DSTATE_STREAMOUT {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t SOFunctionEnable;
+ uint32_t APIRenderingDisable;
+ uint32_t RenderStreamSelect;
+#define LEADING 0
+#define TRAILING 1
+ uint32_t ReorderMode;
+ bool SOStatisticsEnable;
+#define Normal 0
+#define Resreved 1
+#define Force_Off 2
+#define Force_on 3
+ uint32_t ForceRendering;
+ uint32_t Stream3VertexReadOffset;
+ uint32_t Stream3VertexReadLength;
+ uint32_t Stream2VertexReadOffset;
+ uint32_t Stream2VertexReadLength;
+ uint32_t Stream1VertexReadOffset;
+ uint32_t Stream1VertexReadLength;
+ uint32_t Stream0VertexReadOffset;
+ uint32_t Stream0VertexReadLength;
+ uint32_t Buffer1SurfacePitch;
+ uint32_t Buffer0SurfacePitch;
+ uint32_t Buffer3SurfacePitch;
+ uint32_t Buffer2SurfacePitch;
+};
+
+static inline void
+GEN9_3DSTATE_STREAMOUT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_STREAMOUT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->SOFunctionEnable, 31, 31) |
+ __gen_field(values->APIRenderingDisable, 30, 30) |
+ __gen_field(values->RenderStreamSelect, 27, 28) |
+ __gen_field(values->ReorderMode, 26, 26) |
+ __gen_field(values->SOStatisticsEnable, 25, 25) |
+ __gen_field(values->ForceRendering, 23, 24) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->Stream3VertexReadOffset, 29, 29) |
+ __gen_field(values->Stream3VertexReadLength, 24, 28) |
+ __gen_field(values->Stream2VertexReadOffset, 21, 21) |
+ __gen_field(values->Stream2VertexReadLength, 16, 20) |
+ __gen_field(values->Stream1VertexReadOffset, 13, 13) |
+ __gen_field(values->Stream1VertexReadLength, 8, 12) |
+ __gen_field(values->Stream0VertexReadOffset, 5, 5) |
+ __gen_field(values->Stream0VertexReadLength, 0, 4) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->Buffer1SurfacePitch, 16, 27) |
+ __gen_field(values->Buffer0SurfacePitch, 0, 11) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->Buffer3SurfacePitch, 16, 27) |
+ __gen_field(values->Buffer2SurfacePitch, 0, 11) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_TE_length_bias 0x00000002
+#define GEN9_3DSTATE_TE_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 28, \
+ .DwordLength = 2
+
+#define GEN9_3DSTATE_TE_length 0x00000004
+
+struct GEN9_3DSTATE_TE {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+#define INTEGER 0
+#define ODD_FRACTIONAL 1
+#define EVEN_FRACTIONAL 2
+ uint32_t Partitioning;
+#define POINT 0
+#define OUTPUT_LINE 1
+#define OUTPUT_TRI_CW 2
+#define OUTPUT_TRI_CCW 3
+ uint32_t OutputTopology;
+#define QUAD 0
+#define TRI 1
+#define ISOLINE 2
+ uint32_t TEDomain;
+#define HW_TESS 0
+ uint32_t TEMode;
+ bool TEEnable;
+ float MaximumTessellationFactorOdd;
+ float MaximumTessellationFactorNotOdd;
+};
+
+static inline void
+GEN9_3DSTATE_TE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_TE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->Partitioning, 12, 13) |
+ __gen_field(values->OutputTopology, 8, 9) |
+ __gen_field(values->TEDomain, 4, 5) |
+ __gen_field(values->TEMode, 1, 2) |
+ __gen_field(values->TEEnable, 0, 0) |
+ 0;
+
+ dw[2] =
+ __gen_float(values->MaximumTessellationFactorOdd) |
+ 0;
+
+ dw[3] =
+ __gen_float(values->MaximumTessellationFactorNotOdd) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_URB_CLEAR_length_bias 0x00000002
+#define GEN9_3DSTATE_URB_CLEAR_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 1, \
+ ._3DCommandSubOpcode = 29, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_URB_CLEAR_length 0x00000002
+
+struct GEN9_3DSTATE_URB_CLEAR {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t URBClearLength;
+ uint32_t URBAddress;
+};
+
+static inline void
+GEN9_3DSTATE_URB_CLEAR_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_URB_CLEAR * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->URBClearLength, 16, 29) |
+ __gen_offset(values->URBAddress, 0, 14) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_URB_DS_length_bias 0x00000002
+#define GEN9_3DSTATE_URB_DS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 50, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_URB_DS_length 0x00000002
+
+struct GEN9_3DSTATE_URB_DS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t DSURBStartingAddress;
+ uint32_t DSURBEntryAllocationSize;
+ uint32_t DSNumberofURBEntries;
+};
+
+static inline void
+GEN9_3DSTATE_URB_DS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_URB_DS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->DSURBStartingAddress, 25, 31) |
+ __gen_field(values->DSURBEntryAllocationSize, 16, 24) |
+ __gen_field(values->DSNumberofURBEntries, 0, 15) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_URB_GS_length_bias 0x00000002
+#define GEN9_3DSTATE_URB_GS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 51, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_URB_GS_length 0x00000002
+
+struct GEN9_3DSTATE_URB_GS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t GSURBStartingAddress;
+ uint32_t GSURBEntryAllocationSize;
+ uint32_t GSNumberofURBEntries;
+};
+
+static inline void
+GEN9_3DSTATE_URB_GS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_URB_GS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->GSURBStartingAddress, 25, 31) |
+ __gen_field(values->GSURBEntryAllocationSize, 16, 24) |
+ __gen_field(values->GSNumberofURBEntries, 0, 15) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_URB_HS_length_bias 0x00000002
+#define GEN9_3DSTATE_URB_HS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 49, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_URB_HS_length 0x00000002
+
+struct GEN9_3DSTATE_URB_HS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t HSURBStartingAddress;
+ uint32_t HSURBEntryAllocationSize;
+ uint32_t HSNumberofURBEntries;
+};
+
+static inline void
+GEN9_3DSTATE_URB_HS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_URB_HS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->HSURBStartingAddress, 25, 31) |
+ __gen_field(values->HSURBEntryAllocationSize, 16, 24) |
+ __gen_field(values->HSNumberofURBEntries, 0, 15) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_VERTEX_BUFFERS_length_bias 0x00000002
+#define GEN9_3DSTATE_VERTEX_BUFFERS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 8
+
+#define GEN9_3DSTATE_VERTEX_BUFFERS_length 0x00000000
+
+#define GEN9_VERTEX_BUFFER_STATE_length 0x00000004
+
+struct GEN9_VERTEX_BUFFER_STATE {
+ uint32_t VertexBufferIndex;
+ struct GEN9_MEMORY_OBJECT_CONTROL_STATE MemoryObjectControlState;
+ uint32_t AddressModifyEnable;
+ bool NullVertexBuffer;
+ uint32_t BufferPitch;
+ __gen_address_type BufferStartingAddress;
+ uint32_t BufferSize;
+};
+
+static inline void
+GEN9_VERTEX_BUFFER_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_VERTEX_BUFFER_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ uint32_t dw_MemoryObjectControlState;
+ GEN9_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_MemoryObjectControlState, &values->MemoryObjectControlState);
+ dw[0] =
+ __gen_field(values->VertexBufferIndex, 26, 31) |
+ __gen_field(dw_MemoryObjectControlState, 16, 22) |
+ __gen_field(values->AddressModifyEnable, 14, 14) |
+ __gen_field(values->NullVertexBuffer, 13, 13) |
+ __gen_field(values->BufferPitch, 0, 11) |
+ 0;
+
+ uint32_t dw1 =
+ 0;
+
+ uint64_t qw1 =
+ __gen_combine_address(data, &dw[1], values->BufferStartingAddress, dw1);
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+ dw[3] =
+ __gen_field(values->BufferSize, 0, 31) |
+ 0;
+
+}
+
+struct GEN9_3DSTATE_VERTEX_BUFFERS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN9_3DSTATE_VERTEX_BUFFERS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_VERTEX_BUFFERS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN9_3DSTATE_VERTEX_ELEMENTS_length_bias 0x00000002
+#define GEN9_3DSTATE_VERTEX_ELEMENTS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 9
+
+#define GEN9_3DSTATE_VERTEX_ELEMENTS_length 0x00000000
+
+#define GEN9_VERTEX_ELEMENT_STATE_length 0x00000002
+
+struct GEN9_VERTEX_ELEMENT_STATE {
+ uint32_t VertexBufferIndex;
+ bool Valid;
+ uint32_t SourceElementFormat;
+ bool EdgeFlagEnable;
+ uint32_t SourceElementOffset;
+ uint32_t Component0Control;
+ uint32_t Component1Control;
+ uint32_t Component2Control;
+ uint32_t Component3Control;
+};
+
+static inline void
+GEN9_VERTEX_ELEMENT_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_VERTEX_ELEMENT_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->VertexBufferIndex, 26, 31) |
+ __gen_field(values->Valid, 25, 25) |
+ __gen_field(values->SourceElementFormat, 16, 24) |
+ __gen_field(values->EdgeFlagEnable, 15, 15) |
+ __gen_field(values->SourceElementOffset, 0, 11) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->Component0Control, 28, 30) |
+ __gen_field(values->Component1Control, 24, 26) |
+ __gen_field(values->Component2Control, 20, 22) |
+ __gen_field(values->Component3Control, 16, 18) |
+ 0;
+
+}
+
+struct GEN9_3DSTATE_VERTEX_ELEMENTS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN9_3DSTATE_VERTEX_ELEMENTS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_VERTEX_ELEMENTS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN9_3DSTATE_VF_length_bias 0x00000002
+#define GEN9_3DSTATE_VF_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 12, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_VF_length 0x00000002
+
+struct GEN9_3DSTATE_VF {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ bool SequentialDrawCutIndexEnable;
+ bool ComponentPackingEnable;
+ bool IndexedDrawCutIndexEnable;
+ uint32_t DwordLength;
+ uint32_t CutIndex;
+};
+
+static inline void
+GEN9_3DSTATE_VF_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_VF * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->SequentialDrawCutIndexEnable, 10, 10) |
+ __gen_field(values->ComponentPackingEnable, 9, 9) |
+ __gen_field(values->IndexedDrawCutIndexEnable, 8, 8) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->CutIndex, 0, 31) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_VF_COMPONENT_PACKING_length_bias 0x00000002
+#define GEN9_3DSTATE_VF_COMPONENT_PACKING_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 85, \
+ .DwordLength = 3
+
+#define GEN9_3DSTATE_VF_COMPONENT_PACKING_length 0x00000005
+
+struct GEN9_3DSTATE_VF_COMPONENT_PACKING {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t VertexElement07Enables;
+ uint32_t VertexElement06Enables;
+ uint32_t VertexElement05Enables;
+ uint32_t VertexElement04Enables;
+ uint32_t VertexElement03Enables;
+ uint32_t VertexElement02Enables;
+ uint32_t VertexElement01Enables;
+ uint32_t VertexElement00Enables;
+ uint32_t VertexElement15Enables;
+ uint32_t VertexElement14Enables;
+ uint32_t VertexElement13Enables;
+ uint32_t VertexElement12Enables;
+ uint32_t VertexElement11Enables;
+ uint32_t VertexElement10Enables;
+ uint32_t VertexElement09Enables;
+ uint32_t VertexElement08Enables;
+ uint32_t VertexElement23Enables;
+ uint32_t VertexElement22Enables;
+ uint32_t VertexElement21Enables;
+ uint32_t VertexElement20Enables;
+ uint32_t VertexElement19Enables;
+ uint32_t VertexElement18Enables;
+ uint32_t VertexElement17Enables;
+ uint32_t VertexElement16Enables;
+ uint32_t VertexElement31Enables;
+ uint32_t VertexElement30Enables;
+ uint32_t VertexElement29Enables;
+ uint32_t VertexElement28Enables;
+ uint32_t VertexElement27Enables;
+ uint32_t VertexElement26Enables;
+ uint32_t VertexElement25Enables;
+ uint32_t VertexElement24Enables;
+};
+
+static inline void
+GEN9_3DSTATE_VF_COMPONENT_PACKING_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_VF_COMPONENT_PACKING * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->VertexElement07Enables, 28, 31) |
+ __gen_field(values->VertexElement06Enables, 24, 27) |
+ __gen_field(values->VertexElement05Enables, 20, 23) |
+ __gen_field(values->VertexElement04Enables, 16, 19) |
+ __gen_field(values->VertexElement03Enables, 12, 15) |
+ __gen_field(values->VertexElement02Enables, 8, 11) |
+ __gen_field(values->VertexElement01Enables, 4, 7) |
+ __gen_field(values->VertexElement00Enables, 0, 3) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->VertexElement15Enables, 28, 31) |
+ __gen_field(values->VertexElement14Enables, 24, 27) |
+ __gen_field(values->VertexElement13Enables, 20, 23) |
+ __gen_field(values->VertexElement12Enables, 16, 19) |
+ __gen_field(values->VertexElement11Enables, 12, 15) |
+ __gen_field(values->VertexElement10Enables, 8, 11) |
+ __gen_field(values->VertexElement09Enables, 4, 7) |
+ __gen_field(values->VertexElement08Enables, 0, 3) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->VertexElement23Enables, 28, 31) |
+ __gen_field(values->VertexElement22Enables, 24, 27) |
+ __gen_field(values->VertexElement21Enables, 20, 23) |
+ __gen_field(values->VertexElement20Enables, 16, 19) |
+ __gen_field(values->VertexElement19Enables, 12, 15) |
+ __gen_field(values->VertexElement18Enables, 8, 11) |
+ __gen_field(values->VertexElement17Enables, 4, 7) |
+ __gen_field(values->VertexElement16Enables, 0, 3) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->VertexElement31Enables, 28, 31) |
+ __gen_field(values->VertexElement30Enables, 24, 27) |
+ __gen_field(values->VertexElement29Enables, 20, 23) |
+ __gen_field(values->VertexElement28Enables, 16, 19) |
+ __gen_field(values->VertexElement27Enables, 12, 15) |
+ __gen_field(values->VertexElement26Enables, 8, 11) |
+ __gen_field(values->VertexElement25Enables, 4, 7) |
+ __gen_field(values->VertexElement24Enables, 0, 3) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_VF_INSTANCING_length_bias 0x00000002
+#define GEN9_3DSTATE_VF_INSTANCING_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 73, \
+ .DwordLength = 1
+
+#define GEN9_3DSTATE_VF_INSTANCING_length 0x00000003
+
+struct GEN9_3DSTATE_VF_INSTANCING {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool InstancingEnable;
+ uint32_t VertexElementIndex;
+ uint32_t InstanceDataStepRate;
+};
+
+static inline void
+GEN9_3DSTATE_VF_INSTANCING_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_VF_INSTANCING * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->InstancingEnable, 8, 8) |
+ __gen_field(values->VertexElementIndex, 0, 5) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->InstanceDataStepRate, 0, 31) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_VF_SGVS_length_bias 0x00000002
+#define GEN9_3DSTATE_VF_SGVS_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 74, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_VF_SGVS_length 0x00000002
+
+struct GEN9_3DSTATE_VF_SGVS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool InstanceIDEnable;
+#define COMP_0 0
+#define COMP_1 1
+#define COMP_2 2
+#define COMP_3 3
+ uint32_t InstanceIDComponentNumber;
+ uint32_t InstanceIDElementOffset;
+ bool VertexIDEnable;
+#define COMP_0 0
+#define COMP_1 1
+#define COMP_2 2
+#define COMP_3 3
+ uint32_t VertexIDComponentNumber;
+ uint32_t VertexIDElementOffset;
+};
+
+static inline void
+GEN9_3DSTATE_VF_SGVS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_VF_SGVS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->InstanceIDEnable, 31, 31) |
+ __gen_field(values->InstanceIDComponentNumber, 29, 30) |
+ __gen_field(values->InstanceIDElementOffset, 16, 21) |
+ __gen_field(values->VertexIDEnable, 15, 15) |
+ __gen_field(values->VertexIDComponentNumber, 13, 14) |
+ __gen_field(values->VertexIDElementOffset, 0, 5) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_VF_STATISTICS_length_bias 0x00000001
+#define GEN9_3DSTATE_VF_STATISTICS_header \
+ .CommandType = 3, \
+ .CommandSubType = 1, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 11
+
+#define GEN9_3DSTATE_VF_STATISTICS_length 0x00000001
+
+struct GEN9_3DSTATE_VF_STATISTICS {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ bool StatisticsEnable;
+};
+
+static inline void
+GEN9_3DSTATE_VF_STATISTICS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_VF_STATISTICS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->StatisticsEnable, 0, 0) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_VF_TOPOLOGY_length_bias 0x00000002
+#define GEN9_3DSTATE_VF_TOPOLOGY_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 75, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_VF_TOPOLOGY_length 0x00000002
+
+struct GEN9_3DSTATE_VF_TOPOLOGY {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t PrimitiveTopologyType;
+};
+
+static inline void
+GEN9_3DSTATE_VF_TOPOLOGY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_VF_TOPOLOGY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->PrimitiveTopologyType, 0, 5) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_VIEWPORT_STATE_POINTERS_CC_length_bias 0x00000002
+#define GEN9_3DSTATE_VIEWPORT_STATE_POINTERS_CC_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 35, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_VIEWPORT_STATE_POINTERS_CC_length 0x00000002
+
+struct GEN9_3DSTATE_VIEWPORT_STATE_POINTERS_CC {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t CCViewportPointer;
+};
+
+static inline void
+GEN9_3DSTATE_VIEWPORT_STATE_POINTERS_CC_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_VIEWPORT_STATE_POINTERS_CC * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->CCViewportPointer, 5, 31) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP_length_bias 0x00000002
+#define GEN9_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP_header\
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 33, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP_length 0x00000002
+
+struct GEN9_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t SFClipViewportPointer;
+};
+
+static inline void
+GEN9_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->SFClipViewportPointer, 6, 31) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_WM_length_bias 0x00000002
+#define GEN9_3DSTATE_WM_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 20, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_WM_length 0x00000002
+
+struct GEN9_3DSTATE_WM {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool StatisticsEnable;
+ bool LegacyDepthBufferClearEnable;
+ bool LegacyDepthBufferResolveEnable;
+ bool LegacyHierarchicalDepthBufferResolveEnable;
+ bool LegacyDiamondLineRasterization;
+#define NORMAL 0
+#define PSEXEC 1
+#define PREPS 2
+ uint32_t EarlyDepthStencilControl;
+#define Normal 0
+#define ForceOff 1
+#define ForceON 2
+ uint32_t ForceThreadDispatchEnable;
+#define INTERP_PIXEL 0
+#define INTERP_CENTROID 2
+#define INTERP_SAMPLE 3
+ uint32_t PositionZWInterpolationMode;
+ uint32_t BarycentricInterpolationMode;
+#define _05pixels 0
+#define _10pixels 1
+#define _20pixels 2
+#define _40pixels 3
+ uint32_t LineEndCapAntialiasingRegionWidth;
+#define _05pixels 0
+#define _10pixels 1
+#define _20pixels 2
+#define _40pixels 3
+ uint32_t LineAntialiasingRegionWidth;
+ bool PolygonStippleEnable;
+ bool LineStippleEnable;
+#define RASTRULE_UPPER_LEFT 0
+#define RASTRULE_UPPER_RIGHT 1
+ uint32_t PointRasterizationRule;
+#define Normal 0
+#define ForceOff 1
+#define ForceON 2
+ uint32_t ForceKillPixelEnable;
+};
+
+static inline void
+GEN9_3DSTATE_WM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_WM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->StatisticsEnable, 31, 31) |
+ __gen_field(values->LegacyDepthBufferClearEnable, 30, 30) |
+ __gen_field(values->LegacyDepthBufferResolveEnable, 28, 28) |
+ __gen_field(values->LegacyHierarchicalDepthBufferResolveEnable, 27, 27) |
+ __gen_field(values->LegacyDiamondLineRasterization, 26, 26) |
+ __gen_field(values->EarlyDepthStencilControl, 21, 22) |
+ __gen_field(values->ForceThreadDispatchEnable, 19, 20) |
+ __gen_field(values->PositionZWInterpolationMode, 17, 18) |
+ __gen_field(values->BarycentricInterpolationMode, 11, 16) |
+ __gen_field(values->LineEndCapAntialiasingRegionWidth, 8, 9) |
+ __gen_field(values->LineAntialiasingRegionWidth, 6, 7) |
+ __gen_field(values->PolygonStippleEnable, 4, 4) |
+ __gen_field(values->LineStippleEnable, 3, 3) |
+ __gen_field(values->PointRasterizationRule, 2, 2) |
+ __gen_field(values->ForceKillPixelEnable, 0, 1) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_WM_CHROMAKEY_length_bias 0x00000002
+#define GEN9_3DSTATE_WM_CHROMAKEY_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 76, \
+ .DwordLength = 0
+
+#define GEN9_3DSTATE_WM_CHROMAKEY_length 0x00000002
+
+struct GEN9_3DSTATE_WM_CHROMAKEY {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool ChromaKeyKillEnable;
+};
+
+static inline void
+GEN9_3DSTATE_WM_CHROMAKEY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_WM_CHROMAKEY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ChromaKeyKillEnable, 31, 31) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_WM_DEPTH_STENCIL_length_bias 0x00000002
+#define GEN9_3DSTATE_WM_DEPTH_STENCIL_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 78, \
+ .DwordLength = 2
+
+#define GEN9_3DSTATE_WM_DEPTH_STENCIL_length 0x00000004
+
+struct GEN9_3DSTATE_WM_DEPTH_STENCIL {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t StencilFailOp;
+ uint32_t StencilPassDepthFailOp;
+ uint32_t StencilPassDepthPassOp;
+ uint32_t BackfaceStencilTestFunction;
+ uint32_t BackfaceStencilFailOp;
+ uint32_t BackfaceStencilPassDepthFailOp;
+ uint32_t BackfaceStencilPassDepthPassOp;
+ uint32_t StencilTestFunction;
+ uint32_t DepthTestFunction;
+ bool DoubleSidedStencilEnable;
+ bool StencilTestEnable;
+ bool StencilBufferWriteEnable;
+ bool DepthTestEnable;
+ bool DepthBufferWriteEnable;
+ uint32_t StencilTestMask;
+ uint32_t StencilWriteMask;
+ uint32_t BackfaceStencilTestMask;
+ uint32_t BackfaceStencilWriteMask;
+ uint32_t StencilReferenceValue;
+ uint32_t BackfaceStencilReferenceValue;
+};
+
+static inline void
+GEN9_3DSTATE_WM_DEPTH_STENCIL_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_WM_DEPTH_STENCIL * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->StencilFailOp, 29, 31) |
+ __gen_field(values->StencilPassDepthFailOp, 26, 28) |
+ __gen_field(values->StencilPassDepthPassOp, 23, 25) |
+ __gen_field(values->BackfaceStencilTestFunction, 20, 22) |
+ __gen_field(values->BackfaceStencilFailOp, 17, 19) |
+ __gen_field(values->BackfaceStencilPassDepthFailOp, 14, 16) |
+ __gen_field(values->BackfaceStencilPassDepthPassOp, 11, 13) |
+ __gen_field(values->StencilTestFunction, 8, 10) |
+ __gen_field(values->DepthTestFunction, 5, 7) |
+ __gen_field(values->DoubleSidedStencilEnable, 4, 4) |
+ __gen_field(values->StencilTestEnable, 3, 3) |
+ __gen_field(values->StencilBufferWriteEnable, 2, 2) |
+ __gen_field(values->DepthTestEnable, 1, 1) |
+ __gen_field(values->DepthBufferWriteEnable, 0, 0) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->StencilTestMask, 24, 31) |
+ __gen_field(values->StencilWriteMask, 16, 23) |
+ __gen_field(values->BackfaceStencilTestMask, 8, 15) |
+ __gen_field(values->BackfaceStencilWriteMask, 0, 7) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->StencilReferenceValue, 8, 15) |
+ __gen_field(values->BackfaceStencilReferenceValue, 0, 7) |
+ 0;
+
+}
+
+#define GEN9_3DSTATE_WM_HZ_OP_length_bias 0x00000002
+#define GEN9_3DSTATE_WM_HZ_OP_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 0, \
+ ._3DCommandSubOpcode = 82, \
+ .DwordLength = 3
+
+#define GEN9_3DSTATE_WM_HZ_OP_length 0x00000005
+
+struct GEN9_3DSTATE_WM_HZ_OP {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool StencilBufferClearEnable;
+ bool DepthBufferClearEnable;
+ bool ScissorRectangleEnable;
+ bool DepthBufferResolveEnable;
+ bool HierarchicalDepthBufferResolveEnable;
+ uint32_t PixelPositionOffsetEnable;
+ bool FullSurfaceDepthClear;
+ uint32_t StencilClearValue;
+ uint32_t NumberofMultisamples;
+ uint32_t ClearRectangleYMin;
+ uint32_t ClearRectangleXMin;
+ uint32_t ClearRectangleYMax;
+ uint32_t ClearRectangleXMax;
+ uint32_t SampleMask;
+};
+
+static inline void
+GEN9_3DSTATE_WM_HZ_OP_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_3DSTATE_WM_HZ_OP * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->StencilBufferClearEnable, 31, 31) |
+ __gen_field(values->DepthBufferClearEnable, 30, 30) |
+ __gen_field(values->ScissorRectangleEnable, 29, 29) |
+ __gen_field(values->DepthBufferResolveEnable, 28, 28) |
+ __gen_field(values->HierarchicalDepthBufferResolveEnable, 27, 27) |
+ __gen_field(values->PixelPositionOffsetEnable, 26, 26) |
+ __gen_field(values->FullSurfaceDepthClear, 25, 25) |
+ __gen_field(values->StencilClearValue, 16, 23) |
+ __gen_field(values->NumberofMultisamples, 13, 15) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ClearRectangleYMin, 16, 31) |
+ __gen_field(values->ClearRectangleXMin, 0, 15) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->ClearRectangleYMax, 16, 31) |
+ __gen_field(values->ClearRectangleXMax, 0, 15) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->SampleMask, 0, 15) |
+ 0;
+
+}
+
+#define GEN9_GPGPU_WALKER_length_bias 0x00000002
+#define GEN9_GPGPU_WALKER_header \
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 1, \
+ .SubOpcode = 5, \
+ .DwordLength = 13
+
+#define GEN9_GPGPU_WALKER_length 0x0000000f
+
+struct GEN9_GPGPU_WALKER {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ bool IndirectParameterEnable;
+ bool PredicateEnable;
+ uint32_t DwordLength;
+ uint32_t InterfaceDescriptorOffset;
+ uint32_t IndirectDataLength;
+ uint32_t IndirectDataStartAddress;
+#define SIMD8 0
+#define SIMD16 1
+#define SIMD32 2
+ uint32_t SIMDSize;
+ uint32_t ThreadDepthCounterMaximum;
+ uint32_t ThreadHeightCounterMaximum;
+ uint32_t ThreadWidthCounterMaximum;
+ uint32_t ThreadGroupIDStartingX;
+ uint32_t ThreadGroupIDXDimension;
+ uint32_t ThreadGroupIDStartingY;
+ uint32_t ThreadGroupIDYDimension;
+ uint32_t ThreadGroupIDStartingResumeZ;
+ uint32_t ThreadGroupIDZDimension;
+ uint32_t RightExecutionMask;
+ uint32_t BottomExecutionMask;
+};
+
+static inline void
+GEN9_GPGPU_WALKER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_GPGPU_WALKER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->IndirectParameterEnable, 10, 10) |
+ __gen_field(values->PredicateEnable, 8, 8) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->InterfaceDescriptorOffset, 0, 5) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->IndirectDataLength, 0, 16) |
+ 0;
+
+ dw[3] =
+ __gen_offset(values->IndirectDataStartAddress, 6, 31) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->SIMDSize, 30, 31) |
+ __gen_field(values->ThreadDepthCounterMaximum, 16, 21) |
+ __gen_field(values->ThreadHeightCounterMaximum, 8, 13) |
+ __gen_field(values->ThreadWidthCounterMaximum, 0, 5) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->ThreadGroupIDStartingX, 0, 31) |
+ 0;
+
+ dw[6] =
+ 0;
+
+ dw[7] =
+ __gen_field(values->ThreadGroupIDXDimension, 0, 31) |
+ 0;
+
+ dw[8] =
+ __gen_field(values->ThreadGroupIDStartingY, 0, 31) |
+ 0;
+
+ dw[9] =
+ 0;
+
+ dw[10] =
+ __gen_field(values->ThreadGroupIDYDimension, 0, 31) |
+ 0;
+
+ dw[11] =
+ __gen_field(values->ThreadGroupIDStartingResumeZ, 0, 31) |
+ 0;
+
+ dw[12] =
+ __gen_field(values->ThreadGroupIDZDimension, 0, 31) |
+ 0;
+
+ dw[13] =
+ __gen_field(values->RightExecutionMask, 0, 31) |
+ 0;
+
+ dw[14] =
+ __gen_field(values->BottomExecutionMask, 0, 31) |
+ 0;
+
+}
+
+#define GEN9_MEDIA_CURBE_LOAD_length_bias 0x00000002
+#define GEN9_MEDIA_CURBE_LOAD_header \
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 0, \
+ .SubOpcode = 1, \
+ .DwordLength = 2
+
+#define GEN9_MEDIA_CURBE_LOAD_length 0x00000004
+
+struct GEN9_MEDIA_CURBE_LOAD {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ uint32_t DwordLength;
+ uint32_t CURBETotalDataLength;
+ uint32_t CURBEDataStartAddress;
+};
+
+static inline void
+GEN9_MEDIA_CURBE_LOAD_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MEDIA_CURBE_LOAD * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ 0;
+
+ dw[2] =
+ __gen_field(values->CURBETotalDataLength, 0, 16) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->CURBEDataStartAddress, 0, 31) |
+ 0;
+
+}
+
+#define GEN9_MEDIA_INTERFACE_DESCRIPTOR_LOAD_length_bias 0x00000002
+#define GEN9_MEDIA_INTERFACE_DESCRIPTOR_LOAD_header\
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 0, \
+ .SubOpcode = 2, \
+ .DwordLength = 2
+
+#define GEN9_MEDIA_INTERFACE_DESCRIPTOR_LOAD_length 0x00000004
+
+struct GEN9_MEDIA_INTERFACE_DESCRIPTOR_LOAD {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ uint32_t DwordLength;
+ uint32_t InterfaceDescriptorTotalLength;
+ uint32_t InterfaceDescriptorDataStartAddress;
+};
+
+static inline void
+GEN9_MEDIA_INTERFACE_DESCRIPTOR_LOAD_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MEDIA_INTERFACE_DESCRIPTOR_LOAD * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ 0;
+
+ dw[2] =
+ __gen_field(values->InterfaceDescriptorTotalLength, 0, 16) |
+ 0;
+
+ dw[3] =
+ __gen_offset(values->InterfaceDescriptorDataStartAddress, 0, 31) |
+ 0;
+
+}
+
+#define GEN9_MEDIA_OBJECT_length_bias 0x00000002
+#define GEN9_MEDIA_OBJECT_header \
+ .CommandType = 3, \
+ .MediaCommandPipeline = 2, \
+ .MediaCommandOpcode = 1, \
+ .MediaCommandSubOpcode = 0
+
+#define GEN9_MEDIA_OBJECT_length 0x00000000
+
+struct GEN9_MEDIA_OBJECT {
+ uint32_t CommandType;
+ uint32_t MediaCommandPipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t MediaCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t InterfaceDescriptorOffset;
+ bool ChildrenPresent;
+ uint32_t SliceDestinationSelectMSBs;
+#define Nothreadsynchronization 0
+#define Threaddispatchissynchronizedbythespawnrootthreadmessage 1
+ uint32_t ThreadSynchronization;
+ uint32_t ForceDestination;
+#define Notusingscoreboard 0
+#define Usingscoreboard 1
+ uint32_t UseScoreboard;
+#define Slice0 0
+#define Slice1 1
+#define Slice2 2
+ uint32_t SliceDestinationSelect;
+#define Subslice3 3
+#define SubSlice2 2
+#define SubSlice1 1
+#define SubSlice0 0
+ uint32_t SubSliceDestinationSelect;
+ uint32_t IndirectDataLength;
+ __gen_address_type IndirectDataStartAddress;
+ uint32_t ScoredboardY;
+ uint32_t ScoreboardX;
+ uint32_t ScoreboardColor;
+ bool ScoreboardMask;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN9_MEDIA_OBJECT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MEDIA_OBJECT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MediaCommandPipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->MediaCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->InterfaceDescriptorOffset, 0, 5) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ChildrenPresent, 31, 31) |
+ __gen_field(values->SliceDestinationSelectMSBs, 25, 26) |
+ __gen_field(values->ThreadSynchronization, 24, 24) |
+ __gen_field(values->ForceDestination, 22, 22) |
+ __gen_field(values->UseScoreboard, 21, 21) |
+ __gen_field(values->SliceDestinationSelect, 19, 20) |
+ __gen_field(values->SubSliceDestinationSelect, 17, 18) |
+ __gen_field(values->IndirectDataLength, 0, 16) |
+ 0;
+
+ uint32_t dw3 =
+ 0;
+
+ dw[3] =
+ __gen_combine_address(data, &dw[3], values->IndirectDataStartAddress, dw3);
+
+ dw[4] =
+ __gen_field(values->ScoredboardY, 16, 24) |
+ __gen_field(values->ScoreboardX, 0, 8) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->ScoreboardColor, 16, 19) |
+ __gen_field(values->ScoreboardMask, 0, 7) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN9_MEDIA_OBJECT_GRPID_length_bias 0x00000002
+#define GEN9_MEDIA_OBJECT_GRPID_header \
+ .CommandType = 3, \
+ .MediaCommandPipeline = 2, \
+ .MediaCommandOpcode = 1, \
+ .MediaCommandSubOpcode = 6
+
+#define GEN9_MEDIA_OBJECT_GRPID_length 0x00000000
+
+struct GEN9_MEDIA_OBJECT_GRPID {
+ uint32_t CommandType;
+ uint32_t MediaCommandPipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t MediaCommandSubOpcode;
+ uint32_t DwordLength;
+ uint32_t InterfaceDescriptorOffset;
+ uint32_t SliceDestinationSelectMSB;
+ uint32_t EndofThreadGroup;
+ uint32_t ForceDestination;
+#define Notusingscoreboard 0
+#define Usingscoreboard 1
+ uint32_t UseScoreboard;
+#define Slice0 0
+#define Slice1 1
+#define Slice2 2
+ uint32_t SliceDestinationSelect;
+#define Subslice3 3
+#define SubSlice2 2
+#define SubSlice1 1
+#define SubSlice0 0
+ uint32_t SubSliceDestinationSelect;
+ uint32_t IndirectDataLength;
+ __gen_address_type IndirectDataStartAddress;
+ uint32_t ScoreboardY;
+ uint32_t ScoreboardX;
+ uint32_t ScoreboardColor;
+ bool ScoreboardMask;
+ uint32_t GroupID;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN9_MEDIA_OBJECT_GRPID_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MEDIA_OBJECT_GRPID * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MediaCommandPipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->MediaCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->InterfaceDescriptorOffset, 0, 5) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->SliceDestinationSelectMSB, 24, 24) |
+ __gen_field(values->EndofThreadGroup, 23, 23) |
+ __gen_field(values->ForceDestination, 22, 22) |
+ __gen_field(values->UseScoreboard, 21, 21) |
+ __gen_field(values->SliceDestinationSelect, 19, 20) |
+ __gen_field(values->SubSliceDestinationSelect, 17, 18) |
+ __gen_field(values->IndirectDataLength, 0, 16) |
+ 0;
+
+ uint32_t dw3 =
+ 0;
+
+ dw[3] =
+ __gen_combine_address(data, &dw[3], values->IndirectDataStartAddress, dw3);
+
+ dw[4] =
+ __gen_field(values->ScoreboardY, 16, 24) |
+ __gen_field(values->ScoreboardX, 0, 8) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->ScoreboardColor, 16, 19) |
+ __gen_field(values->ScoreboardMask, 0, 7) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->GroupID, 0, 31) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN9_MEDIA_OBJECT_PRT_length_bias 0x00000002
+#define GEN9_MEDIA_OBJECT_PRT_header \
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 1, \
+ .SubOpcode = 2, \
+ .DwordLength = 14
+
+#define GEN9_MEDIA_OBJECT_PRT_length 0x00000010
+
+struct GEN9_MEDIA_OBJECT_PRT {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ uint32_t DwordLength;
+ uint32_t InterfaceDescriptorOffset;
+ bool ChildrenPresent;
+ bool PRT_FenceNeeded;
+#define Rootthreadqueue 0
+#define VFEstateflush 1
+ uint32_t PRT_FenceType;
+ uint32_t InlineData[12];
+};
+
+static inline void
+GEN9_MEDIA_OBJECT_PRT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MEDIA_OBJECT_PRT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->InterfaceDescriptorOffset, 0, 5) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ChildrenPresent, 31, 31) |
+ __gen_field(values->PRT_FenceNeeded, 23, 23) |
+ __gen_field(values->PRT_FenceType, 22, 22) |
+ 0;
+
+ dw[3] =
+ 0;
+
+ for (uint32_t i = 0, j = 4; i < 12; i += 1, j++) {
+ dw[j] =
+ __gen_field(values->InlineData[i + 0], 0, 31) |
+ 0;
+ }
+
+}
+
+#define GEN9_MEDIA_OBJECT_WALKER_length_bias 0x00000002
+#define GEN9_MEDIA_OBJECT_WALKER_header \
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 1, \
+ .SubOpcode = 3
+
+#define GEN9_MEDIA_OBJECT_WALKER_length 0x00000000
+
+struct GEN9_MEDIA_OBJECT_WALKER {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ uint32_t DwordLength;
+ uint32_t InterfaceDescriptorOffset;
+#define Nothreadsynchronization 0
+#define Threaddispatchissynchronizedbythespawnrootthreadmessage 1
+ uint32_t ThreadSynchronization;
+ uint32_t MaskedDispatch;
+#define Notusingscoreboard 0
+#define Usingscoreboard 1
+ uint32_t UseScoreboard;
+ uint32_t IndirectDataLength;
+ uint32_t IndirectDataStartAddress;
+ uint32_t GroupIDLoopSelect;
+ bool ScoreboardMask;
+ uint32_t ColorCountMinusOne;
+ uint32_t MiddleLoopExtraSteps;
+ uint32_t LocalMidLoopUnitY;
+ uint32_t MidLoopUnitX;
+ uint32_t GlobalLoopExecCount;
+ uint32_t LocalLoopExecCount;
+ uint32_t BlockResolutionY;
+ uint32_t BlockResolutionX;
+ uint32_t LocalStartY;
+ uint32_t LocalStartX;
+ uint32_t LocalOuterLoopStrideY;
+ uint32_t LocalOuterLoopStrideX;
+ uint32_t LocalInnerLoopUnitY;
+ uint32_t LocalInnerLoopUnitX;
+ uint32_t GlobalResolutionY;
+ uint32_t GlobalResolutionX;
+ uint32_t GlobalStartY;
+ uint32_t GlobalStartX;
+ uint32_t GlobalOuterLoopStrideY;
+ uint32_t GlobalOuterLoopStrideX;
+ uint32_t GlobalInnerLoopUnitY;
+ uint32_t GlobalInnerLoopUnitX;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN9_MEDIA_OBJECT_WALKER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MEDIA_OBJECT_WALKER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->InterfaceDescriptorOffset, 0, 5) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ThreadSynchronization, 24, 24) |
+ __gen_field(values->MaskedDispatch, 22, 23) |
+ __gen_field(values->UseScoreboard, 21, 21) |
+ __gen_field(values->IndirectDataLength, 0, 16) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->IndirectDataStartAddress, 0, 31) |
+ 0;
+
+ dw[4] =
+ 0;
+
+ dw[5] =
+ __gen_field(values->GroupIDLoopSelect, 8, 31) |
+ __gen_field(values->ScoreboardMask, 0, 7) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->ColorCountMinusOne, 24, 27) |
+ __gen_field(values->MiddleLoopExtraSteps, 16, 20) |
+ __gen_field(values->LocalMidLoopUnitY, 12, 13) |
+ __gen_field(values->MidLoopUnitX, 8, 9) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->GlobalLoopExecCount, 16, 27) |
+ __gen_field(values->LocalLoopExecCount, 0, 11) |
+ 0;
+
+ dw[8] =
+ __gen_field(values->BlockResolutionY, 16, 26) |
+ __gen_field(values->BlockResolutionX, 0, 10) |
+ 0;
+
+ dw[9] =
+ __gen_field(values->LocalStartY, 16, 26) |
+ __gen_field(values->LocalStartX, 0, 10) |
+ 0;
+
+ dw[10] =
+ 0;
+
+ dw[11] =
+ __gen_field(values->LocalOuterLoopStrideY, 16, 27) |
+ __gen_field(values->LocalOuterLoopStrideX, 0, 11) |
+ 0;
+
+ dw[12] =
+ __gen_field(values->LocalInnerLoopUnitY, 16, 27) |
+ __gen_field(values->LocalInnerLoopUnitX, 0, 11) |
+ 0;
+
+ dw[13] =
+ __gen_field(values->GlobalResolutionY, 16, 26) |
+ __gen_field(values->GlobalResolutionX, 0, 10) |
+ 0;
+
+ dw[14] =
+ __gen_field(values->GlobalStartY, 16, 27) |
+ __gen_field(values->GlobalStartX, 0, 11) |
+ 0;
+
+ dw[15] =
+ __gen_field(values->GlobalOuterLoopStrideY, 16, 27) |
+ __gen_field(values->GlobalOuterLoopStrideX, 0, 11) |
+ 0;
+
+ dw[16] =
+ __gen_field(values->GlobalInnerLoopUnitY, 16, 27) |
+ __gen_field(values->GlobalInnerLoopUnitX, 0, 11) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN9_MEDIA_STATE_FLUSH_length_bias 0x00000002
+#define GEN9_MEDIA_STATE_FLUSH_header \
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 0, \
+ .SubOpcode = 4, \
+ .DwordLength = 0
+
+#define GEN9_MEDIA_STATE_FLUSH_length 0x00000002
+
+struct GEN9_MEDIA_STATE_FLUSH {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ uint32_t DwordLength;
+ bool FlushtoGO;
+ uint32_t WatermarkRequired;
+ uint32_t InterfaceDescriptorOffset;
+};
+
+static inline void
+GEN9_MEDIA_STATE_FLUSH_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MEDIA_STATE_FLUSH * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->FlushtoGO, 7, 7) |
+ __gen_field(values->WatermarkRequired, 6, 6) |
+ __gen_field(values->InterfaceDescriptorOffset, 0, 5) |
+ 0;
+
+}
+
+#define GEN9_MEDIA_VFE_STATE_length_bias 0x00000002
+#define GEN9_MEDIA_VFE_STATE_header \
+ .CommandType = 3, \
+ .Pipeline = 2, \
+ .MediaCommandOpcode = 0, \
+ .SubOpcode = 0, \
+ .DwordLength = 7
+
+#define GEN9_MEDIA_VFE_STATE_length 0x00000009
+
+struct GEN9_MEDIA_VFE_STATE {
+ uint32_t CommandType;
+ uint32_t Pipeline;
+ uint32_t MediaCommandOpcode;
+ uint32_t SubOpcode;
+ uint32_t DwordLength;
+ uint32_t ScratchSpaceBasePointer;
+ uint32_t StackSize;
+ uint32_t PerThreadScratchSpace;
+ uint32_t ScratchSpaceBasePointerHigh;
+ uint32_t MaximumNumberofThreads;
+ uint32_t NumberofURBEntries;
+#define Maintainingtheexistingtimestampstate 0
+#define Resettingrelativetimerandlatchingtheglobaltimestamp 1
+ uint32_t ResetGatewayTimer;
+ uint32_t SliceDisable;
+ uint32_t URBEntryAllocationSize;
+ uint32_t CURBEAllocationSize;
+#define Scoreboarddisabled 0
+#define Scoreboardenabled 1
+ uint32_t ScoreboardEnable;
+#define StallingScoreboard 0
+#define NonStallingScoreboard 1
+ uint32_t ScoreboardType;
+ uint32_t ScoreboardMask;
+ uint32_t Scoreboard3DeltaY;
+ uint32_t Scoreboard3DeltaX;
+ uint32_t Scoreboard2DeltaY;
+ uint32_t Scoreboard2DeltaX;
+ uint32_t Scoreboard1DeltaY;
+ uint32_t Scoreboard1DeltaX;
+ uint32_t Scoreboard0DeltaY;
+ uint32_t Scoreboard0DeltaX;
+ uint32_t Scoreboard7DeltaY;
+ uint32_t Scoreboard7DeltaX;
+ uint32_t Scoreboard6DeltaY;
+ uint32_t Scoreboard6DeltaX;
+ uint32_t Scoreboard5DeltaY;
+ uint32_t Scoreboard5DeltaX;
+ uint32_t Scoreboard4DeltaY;
+ uint32_t Scoreboard4DeltaX;
+};
+
+static inline void
+GEN9_MEDIA_VFE_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MEDIA_VFE_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->Pipeline, 27, 28) |
+ __gen_field(values->MediaCommandOpcode, 24, 26) |
+ __gen_field(values->SubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->ScratchSpaceBasePointer, 10, 31) |
+ __gen_field(values->StackSize, 4, 7) |
+ __gen_field(values->PerThreadScratchSpace, 0, 3) |
+ 0;
+
+ dw[2] =
+ __gen_offset(values->ScratchSpaceBasePointerHigh, 0, 15) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->MaximumNumberofThreads, 16, 31) |
+ __gen_field(values->NumberofURBEntries, 8, 15) |
+ __gen_field(values->ResetGatewayTimer, 7, 7) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->SliceDisable, 0, 1) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->URBEntryAllocationSize, 16, 31) |
+ __gen_field(values->CURBEAllocationSize, 0, 15) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->ScoreboardEnable, 31, 31) |
+ __gen_field(values->ScoreboardType, 30, 30) |
+ __gen_field(values->ScoreboardMask, 0, 7) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->Scoreboard3DeltaY, 28, 31) |
+ __gen_field(values->Scoreboard3DeltaX, 24, 27) |
+ __gen_field(values->Scoreboard2DeltaY, 20, 23) |
+ __gen_field(values->Scoreboard2DeltaX, 16, 19) |
+ __gen_field(values->Scoreboard1DeltaY, 12, 15) |
+ __gen_field(values->Scoreboard1DeltaX, 8, 11) |
+ __gen_field(values->Scoreboard0DeltaY, 4, 7) |
+ __gen_field(values->Scoreboard0DeltaX, 0, 3) |
+ 0;
+
+ dw[8] =
+ __gen_field(values->Scoreboard7DeltaY, 28, 31) |
+ __gen_field(values->Scoreboard7DeltaX, 24, 27) |
+ __gen_field(values->Scoreboard6DeltaY, 20, 23) |
+ __gen_field(values->Scoreboard6DeltaX, 16, 19) |
+ __gen_field(values->Scoreboard5DeltaY, 12, 15) |
+ __gen_field(values->Scoreboard5DeltaX, 8, 11) |
+ __gen_field(values->Scoreboard4DeltaY, 4, 7) |
+ __gen_field(values->Scoreboard4DeltaX, 0, 3) |
+ 0;
+
+}
+
+#define GEN9_MI_ARB_CHECK_length_bias 0x00000001
+#define GEN9_MI_ARB_CHECK_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 5
+
+#define GEN9_MI_ARB_CHECK_length 0x00000001
+
+struct GEN9_MI_ARB_CHECK {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+};
+
+static inline void
+GEN9_MI_ARB_CHECK_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_ARB_CHECK * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ 0;
+
+}
+
+#define GEN9_MI_BATCH_BUFFER_END_length_bias 0x00000001
+#define GEN9_MI_BATCH_BUFFER_END_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 10
+
+#define GEN9_MI_BATCH_BUFFER_END_length 0x00000001
+
+struct GEN9_MI_BATCH_BUFFER_END {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+};
+
+static inline void
+GEN9_MI_BATCH_BUFFER_END_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_BATCH_BUFFER_END * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ 0;
+
+}
+
+#define GEN9_MI_CLFLUSH_length_bias 0x00000002
+#define GEN9_MI_CLFLUSH_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 39
+
+#define GEN9_MI_CLFLUSH_length 0x00000000
+
+struct GEN9_MI_CLFLUSH {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define PerProcessGraphicsAddress 0
+#define GlobalGraphicsAddress 1
+ uint32_t UseGlobalGTT;
+ uint32_t DwordLength;
+ __gen_address_type PageBaseAddress;
+ uint32_t StartingCachelineOffset;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN9_MI_CLFLUSH_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_CLFLUSH * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->UseGlobalGTT, 22, 22) |
+ __gen_field(values->DwordLength, 0, 9) |
+ 0;
+
+ uint32_t dw1 =
+ __gen_field(values->StartingCachelineOffset, 6, 11) |
+ 0;
+
+ uint64_t qw1 =
+ __gen_combine_address(data, &dw[1], values->PageBaseAddress, dw1);
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+ /* variable length fields follow */
+}
+
+#define GEN9_MI_COPY_MEM_MEM_length_bias 0x00000002
+#define GEN9_MI_COPY_MEM_MEM_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 46, \
+ .DwordLength = 3
+
+#define GEN9_MI_COPY_MEM_MEM_length 0x00000005
+
+struct GEN9_MI_COPY_MEM_MEM {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define PerProcessGraphicsAddress 0
+#define GlobalGraphicsAddress 1
+ uint32_t UseGlobalGTTSource;
+#define PerProcessGraphicsAddress 0
+#define GlobalGraphicsAddress 1
+ uint32_t UseGlobalGTTDestination;
+ uint32_t DwordLength;
+ __gen_address_type DestinationMemoryAddress;
+ __gen_address_type SourceMemoryAddress;
+};
+
+static inline void
+GEN9_MI_COPY_MEM_MEM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_COPY_MEM_MEM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->UseGlobalGTTSource, 22, 22) |
+ __gen_field(values->UseGlobalGTTDestination, 21, 21) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw1 =
+ 0;
+
+ uint64_t qw1 =
+ __gen_combine_address(data, &dw[1], values->DestinationMemoryAddress, dw1);
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+ uint32_t dw3 =
+ 0;
+
+ uint64_t qw3 =
+ __gen_combine_address(data, &dw[3], values->SourceMemoryAddress, dw3);
+
+ dw[3] = qw3;
+ dw[4] = qw3 >> 32;
+
+}
+
+#define GEN9_MI_DISPLAY_FLIP_length_bias 0x00000002
+#define GEN9_MI_DISPLAY_FLIP_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 20
+
+#define GEN9_MI_DISPLAY_FLIP_length 0x00000003
+
+struct GEN9_MI_DISPLAY_FLIP {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool AsyncFlipIndicator;
+#define DisplayPlane1 0
+#define DisplayPlane2 1
+#define DisplayPlane3 2
+#define DisplayPlane4 4
+#define DisplayPlane5 5
+#define DisplayPlane6 6
+#define DisplayPlane7 7
+#define DisplayPlane8 8
+#define DisplayPlane9 9
+#define DisplayPlane10 10
+#define DisplayPlane11 11
+#define DisplayPlane12 12
+ uint32_t DisplayPlaneSelect;
+ uint32_t DwordLength;
+ bool Stereoscopic3DMode;
+ uint32_t DisplayBufferPitch;
+#define Linear 0
+#define TiledX 1
+#define TiledYLegacyYB 4
+#define TiledYF 5
+ bool TileParameter;
+ __gen_address_type DisplayBufferBaseAddress;
+#define SyncFlip 0
+#define AsyncFlip 1
+#define Stereo3DFlip 2
+ uint32_t FlipType;
+ __gen_address_type LeftEyeDisplayBufferBaseAddress;
+};
+
+static inline void
+GEN9_MI_DISPLAY_FLIP_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_DISPLAY_FLIP * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->AsyncFlipIndicator, 22, 22) |
+ __gen_field(values->DisplayPlaneSelect, 8, 12) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->Stereoscopic3DMode, 31, 31) |
+ __gen_field(values->DisplayBufferPitch, 6, 15) |
+ __gen_field(values->TileParameter, 0, 2) |
+ 0;
+
+ uint32_t dw2 =
+ __gen_field(values->FlipType, 0, 1) |
+ 0;
+
+ dw[2] =
+ __gen_combine_address(data, &dw[2], values->DisplayBufferBaseAddress, dw2);
+
+ uint32_t dw3 =
+ 0;
+
+ dw[3] =
+ __gen_combine_address(data, &dw[3], values->LeftEyeDisplayBufferBaseAddress, dw3);
+
+}
+
+#define GEN9_MI_LOAD_REGISTER_MEM_length_bias 0x00000002
+#define GEN9_MI_LOAD_REGISTER_MEM_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 41, \
+ .DwordLength = 2
+
+#define GEN9_MI_LOAD_REGISTER_MEM_length 0x00000004
+
+struct GEN9_MI_LOAD_REGISTER_MEM {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool UseGlobalGTT;
+ uint32_t AsyncModeEnable;
+ uint32_t DwordLength;
+ uint32_t RegisterAddress;
+ __gen_address_type MemoryAddress;
+};
+
+static inline void
+GEN9_MI_LOAD_REGISTER_MEM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_LOAD_REGISTER_MEM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->UseGlobalGTT, 22, 22) |
+ __gen_field(values->AsyncModeEnable, 21, 21) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->RegisterAddress, 2, 22) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ uint64_t qw2 =
+ __gen_combine_address(data, &dw[2], values->MemoryAddress, dw2);
+
+ dw[2] = qw2;
+ dw[3] = qw2 >> 32;
+
+}
+
+#define GEN9_MI_LOAD_SCAN_LINES_EXCL_length_bias 0x00000002
+#define GEN9_MI_LOAD_SCAN_LINES_EXCL_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 19, \
+ .DwordLength = 0
+
+#define GEN9_MI_LOAD_SCAN_LINES_EXCL_length 0x00000002
+
+struct GEN9_MI_LOAD_SCAN_LINES_EXCL {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define DisplayPlaneA 0
+#define DisplayPlaneB 1
+#define DisplayPlaneC 4
+ uint32_t DisplayPlaneSelect;
+ uint32_t DwordLength;
+ uint32_t StartScanLineNumber;
+ uint32_t EndScanLineNumber;
+};
+
+static inline void
+GEN9_MI_LOAD_SCAN_LINES_EXCL_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_LOAD_SCAN_LINES_EXCL * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DisplayPlaneSelect, 19, 21) |
+ __gen_field(values->DwordLength, 0, 5) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->StartScanLineNumber, 16, 28) |
+ __gen_field(values->EndScanLineNumber, 0, 12) |
+ 0;
+
+}
+
+#define GEN9_MI_LOAD_SCAN_LINES_INCL_length_bias 0x00000002
+#define GEN9_MI_LOAD_SCAN_LINES_INCL_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 18, \
+ .DwordLength = 0
+
+#define GEN9_MI_LOAD_SCAN_LINES_INCL_length 0x00000002
+
+struct GEN9_MI_LOAD_SCAN_LINES_INCL {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define DisplayPlane1A 0
+#define DisplayPlane1B 1
+#define DisplayPlane1C 4
+ uint32_t DisplayPlaneSelect;
+#define NeverForward 0
+#define AlwaysForward 1
+#define ConditionallyForward 2
+ bool ScanLineEventDoneForward;
+ uint32_t DwordLength;
+ uint32_t StartScanLineNumber;
+ uint32_t EndScanLineNumber;
+};
+
+static inline void
+GEN9_MI_LOAD_SCAN_LINES_INCL_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_LOAD_SCAN_LINES_INCL * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DisplayPlaneSelect, 19, 21) |
+ __gen_field(values->ScanLineEventDoneForward, 17, 18) |
+ __gen_field(values->DwordLength, 0, 5) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->StartScanLineNumber, 16, 28) |
+ __gen_field(values->EndScanLineNumber, 0, 12) |
+ 0;
+
+}
+
+#define GEN9_MI_LOAD_URB_MEM_length_bias 0x00000002
+#define GEN9_MI_LOAD_URB_MEM_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 44, \
+ .DwordLength = 2
+
+#define GEN9_MI_LOAD_URB_MEM_length 0x00000004
+
+struct GEN9_MI_LOAD_URB_MEM {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t DwordLength;
+ uint32_t URBAddress;
+ __gen_address_type MemoryAddress;
+};
+
+static inline void
+GEN9_MI_LOAD_URB_MEM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_LOAD_URB_MEM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->URBAddress, 2, 14) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ uint64_t qw2 =
+ __gen_combine_address(data, &dw[2], values->MemoryAddress, dw2);
+
+ dw[2] = qw2;
+ dw[3] = qw2 >> 32;
+
+}
+
+#define GEN9_MI_MATH_length_bias 0x00000002
+#define GEN9_MI_MATH_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 26
+
+#define GEN9_MI_MATH_length 0x00000000
+
+struct GEN9_MI_MATH {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t DwordLength;
+ uint32_t ALUINSTRUCTION1;
+ uint32_t ALUINSTRUCTION2;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN9_MI_MATH_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_MATH * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ALUINSTRUCTION1, 0, 31) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->ALUINSTRUCTION2, 0, 31) |
+ 0;
+
+ /* variable length fields follow */
+}
+
+#define GEN9_MI_NOOP_length_bias 0x00000001
+#define GEN9_MI_NOOP_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 0
+
+#define GEN9_MI_NOOP_length 0x00000001
+
+struct GEN9_MI_NOOP {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool IdentificationNumberRegisterWriteEnable;
+ uint32_t IdentificationNumber;
+};
+
+static inline void
+GEN9_MI_NOOP_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_NOOP * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->IdentificationNumberRegisterWriteEnable, 22, 22) |
+ __gen_field(values->IdentificationNumber, 0, 21) |
+ 0;
+
+}
+
+#define GEN9_MI_PREDICATE_length_bias 0x00000001
+#define GEN9_MI_PREDICATE_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 12
+
+#define GEN9_MI_PREDICATE_length 0x00000001
+
+struct GEN9_MI_PREDICATE {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define LOAD_KEEP 0
+#define LOAD_LOAD 2
+#define LOAD_LOADINV 3
+ uint32_t LoadOperation;
+#define COMBINE_SET 0
+#define COMBINE_AND 1
+#define COMBINE_OR 2
+#define COMBINE_XOR 3
+ uint32_t CombineOperation;
+#define COMPARE_SRCS_EQUAL 2
+#define COMPARE_DELTAS_EQUAL 3
+ uint32_t CompareOperation;
+};
+
+static inline void
+GEN9_MI_PREDICATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_PREDICATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->LoadOperation, 6, 7) |
+ __gen_field(values->CombineOperation, 3, 4) |
+ __gen_field(values->CompareOperation, 0, 1) |
+ 0;
+
+}
+
+#define GEN9_MI_REPORT_HEAD_length_bias 0x00000001
+#define GEN9_MI_REPORT_HEAD_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 7
+
+#define GEN9_MI_REPORT_HEAD_length 0x00000001
+
+struct GEN9_MI_REPORT_HEAD {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+};
+
+static inline void
+GEN9_MI_REPORT_HEAD_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_REPORT_HEAD * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ 0;
+
+}
+
+#define GEN9_MI_RS_CONTEXT_length_bias 0x00000001
+#define GEN9_MI_RS_CONTEXT_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 15
+
+#define GEN9_MI_RS_CONTEXT_length 0x00000001
+
+struct GEN9_MI_RS_CONTEXT {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define RS_RESTORE 0
+#define RS_SAVE 1
+ uint32_t ResourceStreamerSave;
+};
+
+static inline void
+GEN9_MI_RS_CONTEXT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_RS_CONTEXT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->ResourceStreamerSave, 0, 0) |
+ 0;
+
+}
+
+#define GEN9_MI_RS_CONTROL_length_bias 0x00000001
+#define GEN9_MI_RS_CONTROL_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 6
+
+#define GEN9_MI_RS_CONTROL_length 0x00000001
+
+struct GEN9_MI_RS_CONTROL {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define RS_STOP 0
+#define RS_START 1
+ uint32_t ResourceStreamerControl;
+};
+
+static inline void
+GEN9_MI_RS_CONTROL_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_RS_CONTROL * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->ResourceStreamerControl, 0, 0) |
+ 0;
+
+}
+
+#define GEN9_MI_RS_STORE_DATA_IMM_length_bias 0x00000002
+#define GEN9_MI_RS_STORE_DATA_IMM_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 43, \
+ .DwordLength = 2
+
+#define GEN9_MI_RS_STORE_DATA_IMM_length 0x00000004
+
+struct GEN9_MI_RS_STORE_DATA_IMM {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t DwordLength;
+ __gen_address_type DestinationAddress;
+ uint32_t CoreModeEnable;
+ uint32_t DataDWord0;
+};
+
+static inline void
+GEN9_MI_RS_STORE_DATA_IMM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_RS_STORE_DATA_IMM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw1 =
+ __gen_field(values->CoreModeEnable, 0, 0) |
+ 0;
+
+ uint64_t qw1 =
+ __gen_combine_address(data, &dw[1], values->DestinationAddress, dw1);
+
+ dw[1] = qw1;
+ dw[2] = qw1 >> 32;
+
+ dw[3] =
+ __gen_field(values->DataDWord0, 0, 31) |
+ 0;
+
+}
+
+#define GEN9_MI_SET_CONTEXT_length_bias 0x00000002
+#define GEN9_MI_SET_CONTEXT_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 24, \
+ .DwordLength = 0
+
+#define GEN9_MI_SET_CONTEXT_length 0x00000002
+
+struct GEN9_MI_SET_CONTEXT {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t DwordLength;
+ __gen_address_type LogicalContextAddress;
+ uint32_t ReservedMustbe1;
+ bool CoreModeEnable;
+ bool ResourceStreamerStateSaveEnable;
+ bool ResourceStreamerStateRestoreEnable;
+ uint32_t ForceRestore;
+ uint32_t RestoreInhibit;
+};
+
+static inline void
+GEN9_MI_SET_CONTEXT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_SET_CONTEXT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ uint32_t dw1 =
+ __gen_field(values->ReservedMustbe1, 8, 8) |
+ __gen_field(values->CoreModeEnable, 4, 4) |
+ __gen_field(values->ResourceStreamerStateSaveEnable, 3, 3) |
+ __gen_field(values->ResourceStreamerStateRestoreEnable, 2, 2) |
+ __gen_field(values->ForceRestore, 1, 1) |
+ __gen_field(values->RestoreInhibit, 0, 0) |
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->LogicalContextAddress, dw1);
+
+}
+
+#define GEN9_MI_SET_PREDICATE_length_bias 0x00000001
+#define GEN9_MI_SET_PREDICATE_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 1
+
+#define GEN9_MI_SET_PREDICATE_length 0x00000001
+
+struct GEN9_MI_SET_PREDICATE {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+#define NOOPNever 0
+#define NOOPonResult2clear 1
+#define NOOPonResult2set 2
+#define NOOPonResultclear 3
+#define NOOPonResultset 4
+#define Executewhenonesliceenabled 5
+#define Executewhentwoslicesareenabled 6
+#define Executewhenthreeslicesareenabled 7
+#define NOOPAlways 15
+ uint32_t PREDICATEENABLE;
+};
+
+static inline void
+GEN9_MI_SET_PREDICATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_SET_PREDICATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->PREDICATEENABLE, 0, 3) |
+ 0;
+
+}
+
+#define GEN9_MI_STORE_DATA_INDEX_length_bias 0x00000002
+#define GEN9_MI_STORE_DATA_INDEX_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 33, \
+ .DwordLength = 1
+
+#define GEN9_MI_STORE_DATA_INDEX_length 0x00000003
+
+struct GEN9_MI_STORE_DATA_INDEX {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t UsePerProcessHardwareStatusPage;
+ uint32_t DwordLength;
+ uint32_t Offset;
+ uint32_t DataDWord0;
+ uint32_t DataDWord1;
+};
+
+static inline void
+GEN9_MI_STORE_DATA_INDEX_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_STORE_DATA_INDEX * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->UsePerProcessHardwareStatusPage, 21, 21) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->Offset, 2, 11) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->DataDWord0, 0, 31) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->DataDWord1, 0, 31) |
+ 0;
+
+}
+
+#define GEN9_MI_STORE_URB_MEM_length_bias 0x00000002
+#define GEN9_MI_STORE_URB_MEM_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 45, \
+ .DwordLength = 2
+
+#define GEN9_MI_STORE_URB_MEM_length 0x00000004
+
+struct GEN9_MI_STORE_URB_MEM {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t DwordLength;
+ uint32_t URBAddress;
+ __gen_address_type MemoryAddress;
+};
+
+static inline void
+GEN9_MI_STORE_URB_MEM_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_STORE_URB_MEM * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->URBAddress, 2, 14) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ uint64_t qw2 =
+ __gen_combine_address(data, &dw[2], values->MemoryAddress, dw2);
+
+ dw[2] = qw2;
+ dw[3] = qw2 >> 32;
+
+}
+
+#define GEN9_MI_SUSPEND_FLUSH_length_bias 0x00000001
+#define GEN9_MI_SUSPEND_FLUSH_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 11
+
+#define GEN9_MI_SUSPEND_FLUSH_length 0x00000001
+
+struct GEN9_MI_SUSPEND_FLUSH {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool SuspendFlush;
+};
+
+static inline void
+GEN9_MI_SUSPEND_FLUSH_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_SUSPEND_FLUSH * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->SuspendFlush, 0, 0) |
+ 0;
+
+}
+
+#define GEN9_MI_TOPOLOGY_FILTER_length_bias 0x00000001
+#define GEN9_MI_TOPOLOGY_FILTER_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 13
+
+#define GEN9_MI_TOPOLOGY_FILTER_length 0x00000001
+
+struct GEN9_MI_TOPOLOGY_FILTER {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t TopologyFilterValue;
+};
+
+static inline void
+GEN9_MI_TOPOLOGY_FILTER_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_TOPOLOGY_FILTER * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->TopologyFilterValue, 0, 5) |
+ 0;
+
+}
+
+#define GEN9_MI_UPDATE_GTT_length_bias 0x00000002
+#define GEN9_MI_UPDATE_GTT_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 35
+
+#define GEN9_MI_UPDATE_GTT_length 0x00000000
+
+struct GEN9_MI_UPDATE_GTT {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t DwordLength;
+ __gen_address_type EntryAddress;
+ /* variable length fields follow */
+};
+
+static inline void
+GEN9_MI_UPDATE_GTT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_UPDATE_GTT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DwordLength, 0, 9) |
+ 0;
+
+ uint32_t dw1 =
+ 0;
+
+ dw[1] =
+ __gen_combine_address(data, &dw[1], values->EntryAddress, dw1);
+
+ /* variable length fields follow */
+}
+
+#define GEN9_MI_URB_ATOMIC_ALLOC_length_bias 0x00000001
+#define GEN9_MI_URB_ATOMIC_ALLOC_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 9
+
+#define GEN9_MI_URB_ATOMIC_ALLOC_length 0x00000001
+
+struct GEN9_MI_URB_ATOMIC_ALLOC {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ uint32_t URBAtomicStorageOffset;
+ uint32_t URBAtomicStorageSize;
+};
+
+static inline void
+GEN9_MI_URB_ATOMIC_ALLOC_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_URB_ATOMIC_ALLOC * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->URBAtomicStorageOffset, 12, 19) |
+ __gen_field(values->URBAtomicStorageSize, 0, 8) |
+ 0;
+
+}
+
+#define GEN9_MI_USER_INTERRUPT_length_bias 0x00000001
+#define GEN9_MI_USER_INTERRUPT_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 2
+
+#define GEN9_MI_USER_INTERRUPT_length 0x00000001
+
+struct GEN9_MI_USER_INTERRUPT {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+};
+
+static inline void
+GEN9_MI_USER_INTERRUPT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_USER_INTERRUPT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ 0;
+
+}
+
+#define GEN9_MI_WAIT_FOR_EVENT_length_bias 0x00000001
+#define GEN9_MI_WAIT_FOR_EVENT_header \
+ .CommandType = 0, \
+ .MICommandOpcode = 3
+
+#define GEN9_MI_WAIT_FOR_EVENT_length 0x00000001
+
+struct GEN9_MI_WAIT_FOR_EVENT {
+ uint32_t CommandType;
+ uint32_t MICommandOpcode;
+ bool DisplayPlane1CVerticalBlankWaitEnable;
+ bool DisplayPlane6FlipPendingWaitEnable;
+ bool DisplayPlane12FlipPendingWaitEnable;
+ bool DisplayPlane11FlipPendingWaitEnable;
+ bool DisplayPlane10FlipPendingWaitEnable;
+ bool DisplayPlane9FlipPendingWaitEnable;
+ bool DisplayPlane3FlipPendingWaitEnable;
+ bool DisplayPlane1CScanLineWaitEnable;
+ bool DisplayPlane1BVerticalBlankWaitEnable;
+ bool DisplayPlane5FlipPendingWaitEnable;
+ bool DisplayPlane2FlipPendingWaitEnable;
+ bool DisplayPlane1BScanLineWaitEnable;
+ bool DisplayPlane8FlipPendingWaitEnable;
+ bool DisplayPlane7FlipPendingWaitEnable;
+ bool DisplayPlane1AVerticalBlankWaitEnable;
+ bool DisplayPlane4FlipPendingWaitEnable;
+ bool DisplayPlane1FlipPendingWaitEnable;
+ bool DisplayPlnae1AScanLineWaitEnable;
+};
+
+static inline void
+GEN9_MI_WAIT_FOR_EVENT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_MI_WAIT_FOR_EVENT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->MICommandOpcode, 23, 28) |
+ __gen_field(values->DisplayPlane1CVerticalBlankWaitEnable, 21, 21) |
+ __gen_field(values->DisplayPlane6FlipPendingWaitEnable, 20, 20) |
+ __gen_field(values->DisplayPlane12FlipPendingWaitEnable, 19, 19) |
+ __gen_field(values->DisplayPlane11FlipPendingWaitEnable, 18, 18) |
+ __gen_field(values->DisplayPlane10FlipPendingWaitEnable, 17, 17) |
+ __gen_field(values->DisplayPlane9FlipPendingWaitEnable, 16, 16) |
+ __gen_field(values->DisplayPlane3FlipPendingWaitEnable, 15, 15) |
+ __gen_field(values->DisplayPlane1CScanLineWaitEnable, 14, 14) |
+ __gen_field(values->DisplayPlane1BVerticalBlankWaitEnable, 11, 11) |
+ __gen_field(values->DisplayPlane5FlipPendingWaitEnable, 10, 10) |
+ __gen_field(values->DisplayPlane2FlipPendingWaitEnable, 9, 9) |
+ __gen_field(values->DisplayPlane1BScanLineWaitEnable, 8, 8) |
+ __gen_field(values->DisplayPlane8FlipPendingWaitEnable, 7, 7) |
+ __gen_field(values->DisplayPlane7FlipPendingWaitEnable, 6, 6) |
+ __gen_field(values->DisplayPlane1AVerticalBlankWaitEnable, 3, 3) |
+ __gen_field(values->DisplayPlane4FlipPendingWaitEnable, 2, 2) |
+ __gen_field(values->DisplayPlane1FlipPendingWaitEnable, 1, 1) |
+ __gen_field(values->DisplayPlnae1AScanLineWaitEnable, 0, 0) |
+ 0;
+
+}
+
+#define GEN9_PIPE_CONTROL_length_bias 0x00000002
+#define GEN9_PIPE_CONTROL_header \
+ .CommandType = 3, \
+ .CommandSubType = 3, \
+ ._3DCommandOpcode = 2, \
+ ._3DCommandSubOpcode = 0, \
+ .DwordLength = 4
+
+#define GEN9_PIPE_CONTROL_length 0x00000006
+
+struct GEN9_PIPE_CONTROL {
+ uint32_t CommandType;
+ uint32_t CommandSubType;
+ uint32_t _3DCommandOpcode;
+ uint32_t _3DCommandSubOpcode;
+ uint32_t DwordLength;
+ bool FlushLLC;
+#define DAT_PPGTT 0
+#define DAT_GGTT 1
+ uint32_t DestinationAddressType;
+#define NoLRIOperation 0
+#define MMIOWriteImmediateData 1
+ uint32_t LRIPostSyncOperation;
+ uint32_t StoreDataIndex;
+ uint32_t CommandStreamerStallEnable;
+#define DontReset 0
+#define Reset 1
+ uint32_t GlobalSnapshotCountReset;
+ uint32_t TLBInvalidate;
+ bool GenericMediaStateClear;
+#define NoWrite 0
+#define WriteImmediateData 1
+#define WritePSDepthCount 2
+#define WriteTimestamp 3
+ uint32_t PostSyncOperation;
+ bool DepthStallEnable;
+#define DisableFlush 0
+#define EnableFlush 1
+ bool RenderTargetCacheFlushEnable;
+ bool InstructionCacheInvalidateEnable;
+ bool TextureCacheInvalidationEnable;
+ bool IndirectStatePointersDisable;
+ bool NotifyEnable;
+ bool PipeControlFlushEnable;
+ bool DCFlushEnable;
+ bool VFCacheInvalidationEnable;
+ bool ConstantCacheInvalidationEnable;
+ bool StateCacheInvalidationEnable;
+ bool StallAtPixelScoreboard;
+#define FlushDisabled 0
+#define FlushEnabled 1
+ bool DepthCacheFlushEnable;
+ __gen_address_type Address;
+ uint64_t ImmediateData;
+};
+
+static inline void
+GEN9_PIPE_CONTROL_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_PIPE_CONTROL * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->CommandType, 29, 31) |
+ __gen_field(values->CommandSubType, 27, 28) |
+ __gen_field(values->_3DCommandOpcode, 24, 26) |
+ __gen_field(values->_3DCommandSubOpcode, 16, 23) |
+ __gen_field(values->DwordLength, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->FlushLLC, 26, 26) |
+ __gen_field(values->DestinationAddressType, 24, 24) |
+ __gen_field(values->LRIPostSyncOperation, 23, 23) |
+ __gen_field(values->StoreDataIndex, 21, 21) |
+ __gen_field(values->CommandStreamerStallEnable, 20, 20) |
+ __gen_field(values->GlobalSnapshotCountReset, 19, 19) |
+ __gen_field(values->TLBInvalidate, 18, 18) |
+ __gen_field(values->GenericMediaStateClear, 16, 16) |
+ __gen_field(values->PostSyncOperation, 14, 15) |
+ __gen_field(values->DepthStallEnable, 13, 13) |
+ __gen_field(values->RenderTargetCacheFlushEnable, 12, 12) |
+ __gen_field(values->InstructionCacheInvalidateEnable, 11, 11) |
+ __gen_field(values->TextureCacheInvalidationEnable, 10, 10) |
+ __gen_field(values->IndirectStatePointersDisable, 9, 9) |
+ __gen_field(values->NotifyEnable, 8, 8) |
+ __gen_field(values->PipeControlFlushEnable, 7, 7) |
+ __gen_field(values->DCFlushEnable, 5, 5) |
+ __gen_field(values->VFCacheInvalidationEnable, 4, 4) |
+ __gen_field(values->ConstantCacheInvalidationEnable, 3, 3) |
+ __gen_field(values->StateCacheInvalidationEnable, 2, 2) |
+ __gen_field(values->StallAtPixelScoreboard, 1, 1) |
+ __gen_field(values->DepthCacheFlushEnable, 0, 0) |
+ 0;
+
+ uint32_t dw2 =
+ 0;
+
+ uint64_t qw2 =
+ __gen_combine_address(data, &dw[2], values->Address, dw2);
+
+ dw[2] = qw2;
+ dw[3] = qw2 >> 32;
+
+ uint64_t qw4 =
+ __gen_field(values->ImmediateData, 0, 63) |
+ 0;
+
+ dw[4] = qw4;
+ dw[5] = qw4 >> 32;
+
+}
+
+#define GEN9_SCISSOR_RECT_length 0x00000002
+
+struct GEN9_SCISSOR_RECT {
+ uint32_t ScissorRectangleYMin;
+ uint32_t ScissorRectangleXMin;
+ uint32_t ScissorRectangleYMax;
+ uint32_t ScissorRectangleXMax;
+};
+
+static inline void
+GEN9_SCISSOR_RECT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_SCISSOR_RECT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->ScissorRectangleYMin, 16, 31) |
+ __gen_field(values->ScissorRectangleXMin, 0, 15) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->ScissorRectangleYMax, 16, 31) |
+ __gen_field(values->ScissorRectangleXMax, 0, 15) |
+ 0;
+
+}
+
+#define GEN9_SF_CLIP_VIEWPORT_length 0x00000010
+
+struct GEN9_SF_CLIP_VIEWPORT {
+ float ViewportMatrixElementm00;
+ float ViewportMatrixElementm11;
+ float ViewportMatrixElementm22;
+ float ViewportMatrixElementm30;
+ float ViewportMatrixElementm31;
+ float ViewportMatrixElementm32;
+ float XMinClipGuardband;
+ float XMaxClipGuardband;
+ float YMinClipGuardband;
+ float YMaxClipGuardband;
+ float XMinViewPort;
+ float XMaxViewPort;
+ float YMinViewPort;
+ float YMaxViewPort;
+};
+
+static inline void
+GEN9_SF_CLIP_VIEWPORT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_SF_CLIP_VIEWPORT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_float(values->ViewportMatrixElementm00) |
+ 0;
+
+ dw[1] =
+ __gen_float(values->ViewportMatrixElementm11) |
+ 0;
+
+ dw[2] =
+ __gen_float(values->ViewportMatrixElementm22) |
+ 0;
+
+ dw[3] =
+ __gen_float(values->ViewportMatrixElementm30) |
+ 0;
+
+ dw[4] =
+ __gen_float(values->ViewportMatrixElementm31) |
+ 0;
+
+ dw[5] =
+ __gen_float(values->ViewportMatrixElementm32) |
+ 0;
+
+ dw[6] =
+ 0;
+
+ dw[7] =
+ 0;
+
+ dw[8] =
+ __gen_float(values->XMinClipGuardband) |
+ 0;
+
+ dw[9] =
+ __gen_float(values->XMaxClipGuardband) |
+ 0;
+
+ dw[10] =
+ __gen_float(values->YMinClipGuardband) |
+ 0;
+
+ dw[11] =
+ __gen_float(values->YMaxClipGuardband) |
+ 0;
+
+ dw[12] =
+ __gen_float(values->XMinViewPort) |
+ 0;
+
+ dw[13] =
+ __gen_float(values->XMaxViewPort) |
+ 0;
+
+ dw[14] =
+ __gen_float(values->YMinViewPort) |
+ 0;
+
+ dw[15] =
+ __gen_float(values->YMaxViewPort) |
+ 0;
+
+}
+
+#define GEN9_BLEND_STATE_length 0x00000011
+
+#define GEN9_BLEND_STATE_ENTRY_length 0x00000002
+
+struct GEN9_BLEND_STATE_ENTRY {
+ bool LogicOpEnable;
+ uint32_t LogicOpFunction;
+ uint32_t PreBlendSourceOnlyClampEnable;
+#define COLORCLAMP_UNORM 0
+#define COLORCLAMP_SNORM 1
+#define COLORCLAMP_RTFORMAT 2
+ uint32_t ColorClampRange;
+ bool PreBlendColorClampEnable;
+ bool PostBlendColorClampEnable;
+ bool ColorBufferBlendEnable;
+ uint32_t SourceBlendFactor;
+ uint32_t DestinationBlendFactor;
+ uint32_t ColorBlendFunction;
+ uint32_t SourceAlphaBlendFactor;
+ uint32_t DestinationAlphaBlendFactor;
+ uint32_t AlphaBlendFunction;
+ bool WriteDisableAlpha;
+ bool WriteDisableRed;
+ bool WriteDisableGreen;
+ bool WriteDisableBlue;
+};
+
+static inline void
+GEN9_BLEND_STATE_ENTRY_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_BLEND_STATE_ENTRY * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ uint64_t qw0 =
+ __gen_field(values->LogicOpEnable, 63, 63) |
+ __gen_field(values->LogicOpFunction, 59, 62) |
+ __gen_field(values->PreBlendSourceOnlyClampEnable, 36, 36) |
+ __gen_field(values->ColorClampRange, 34, 35) |
+ __gen_field(values->PreBlendColorClampEnable, 33, 33) |
+ __gen_field(values->PostBlendColorClampEnable, 32, 32) |
+ __gen_field(values->ColorBufferBlendEnable, 31, 31) |
+ __gen_field(values->SourceBlendFactor, 26, 30) |
+ __gen_field(values->DestinationBlendFactor, 21, 25) |
+ __gen_field(values->ColorBlendFunction, 18, 20) |
+ __gen_field(values->SourceAlphaBlendFactor, 13, 17) |
+ __gen_field(values->DestinationAlphaBlendFactor, 8, 12) |
+ __gen_field(values->AlphaBlendFunction, 5, 7) |
+ __gen_field(values->WriteDisableAlpha, 3, 3) |
+ __gen_field(values->WriteDisableRed, 2, 2) |
+ __gen_field(values->WriteDisableGreen, 1, 1) |
+ __gen_field(values->WriteDisableBlue, 0, 0) |
+ 0;
+
+ dw[0] = qw0;
+ dw[1] = qw0 >> 32;
+
+}
+
+struct GEN9_BLEND_STATE {
+ bool AlphaToCoverageEnable;
+ bool IndependentAlphaBlendEnable;
+ bool AlphaToOneEnable;
+ bool AlphaToCoverageDitherEnable;
+ bool AlphaTestEnable;
+ uint32_t AlphaTestFunction;
+ bool ColorDitherEnable;
+ uint32_t XDitherOffset;
+ uint32_t YDitherOffset;
+ struct GEN9_BLEND_STATE_ENTRY Entry[8];
+};
+
+static inline void
+GEN9_BLEND_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_BLEND_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->AlphaToCoverageEnable, 31, 31) |
+ __gen_field(values->IndependentAlphaBlendEnable, 30, 30) |
+ __gen_field(values->AlphaToOneEnable, 29, 29) |
+ __gen_field(values->AlphaToCoverageDitherEnable, 28, 28) |
+ __gen_field(values->AlphaTestEnable, 27, 27) |
+ __gen_field(values->AlphaTestFunction, 24, 26) |
+ __gen_field(values->ColorDitherEnable, 23, 23) |
+ __gen_field(values->XDitherOffset, 21, 22) |
+ __gen_field(values->YDitherOffset, 19, 20) |
+ 0;
+
+ for (uint32_t i = 0, j = 1; i < 8; i++, j += 2)
+ GEN9_BLEND_STATE_ENTRY_pack(data, &dw[j], &values->Entry[i]);
+}
+
+#define GEN9_CC_VIEWPORT_length 0x00000002
+
+struct GEN9_CC_VIEWPORT {
+ float MinimumDepth;
+ float MaximumDepth;
+};
+
+static inline void
+GEN9_CC_VIEWPORT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_CC_VIEWPORT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_float(values->MinimumDepth) |
+ 0;
+
+ dw[1] =
+ __gen_float(values->MaximumDepth) |
+ 0;
+
+}
+
+#define GEN9_COLOR_CALC_STATE_length 0x00000006
+
+struct GEN9_COLOR_CALC_STATE {
+#define Cancelled 0
+#define NotCancelled 1
+ uint32_t RoundDisableFunctionDisable;
+#define ALPHATEST_UNORM8 0
+#define ALPHATEST_FLOAT32 1
+ uint32_t AlphaTestFormat;
+ uint32_t AlphaReferenceValueAsUNORM8;
+ float AlphaReferenceValueAsFLOAT32;
+ float BlendConstantColorRed;
+ float BlendConstantColorGreen;
+ float BlendConstantColorBlue;
+ float BlendConstantColorAlpha;
+};
+
+static inline void
+GEN9_COLOR_CALC_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_COLOR_CALC_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->RoundDisableFunctionDisable, 15, 15) |
+ __gen_field(values->AlphaTestFormat, 0, 0) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->AlphaReferenceValueAsUNORM8, 0, 31) |
+ __gen_float(values->AlphaReferenceValueAsFLOAT32) |
+ 0;
+
+ dw[2] =
+ __gen_float(values->BlendConstantColorRed) |
+ 0;
+
+ dw[3] =
+ __gen_float(values->BlendConstantColorGreen) |
+ 0;
+
+ dw[4] =
+ __gen_float(values->BlendConstantColorBlue) |
+ 0;
+
+ dw[5] =
+ __gen_float(values->BlendConstantColorAlpha) |
+ 0;
+
+}
+
+#define GEN9_EXECUTION_UNIT_EXTENDED_MESSAGE_DESCRIPTOR_length 0x00000001
+
+struct GEN9_EXECUTION_UNIT_EXTENDED_MESSAGE_DESCRIPTOR {
+ uint32_t ExtendedMessageLength;
+#define NoTermination 0
+#define EOT 1
+ uint32_t EndOfThread;
+ uint32_t TargetFunctionID;
+};
+
+static inline void
+GEN9_EXECUTION_UNIT_EXTENDED_MESSAGE_DESCRIPTOR_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_EXECUTION_UNIT_EXTENDED_MESSAGE_DESCRIPTOR * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->ExtendedMessageLength, 6, 9) |
+ __gen_field(values->EndOfThread, 5, 5) |
+ __gen_field(values->TargetFunctionID, 0, 3) |
+ 0;
+
+}
+
+#define GEN9_INTERFACE_DESCRIPTOR_DATA_length 0x00000008
+
+struct GEN9_INTERFACE_DESCRIPTOR_DATA {
+ uint32_t KernelStartPointer;
+ uint32_t KernelStartPointerHigh;
+#define Ftz 0
+#define SetByKernel 1
+ uint32_t DenormMode;
+#define Multiple 0
+#define Single 1
+ uint32_t SingleProgramFlow;
+#define NormalPriority 0
+#define HighPriority 1
+ uint32_t ThreadPriority;
+#define IEEE754 0
+#define Alternate 1
+ uint32_t FloatingPointMode;
+ bool IllegalOpcodeExceptionEnable;
+ bool MaskStackExceptionEnable;
+ bool SoftwareExceptionEnable;
+ uint32_t SamplerStatePointer;
+#define Nosamplersused 0
+#define Between1and4samplersused 1
+#define Between5and8samplersused 2
+#define Between9and12samplersused 3
+#define Between13and16samplersused 4
+ uint32_t SamplerCount;
+ uint32_t BindingTablePointer;
+ uint32_t BindingTableEntryCount;
+ uint32_t ConstantIndirectURBEntryReadLength;
+ uint32_t ConstantURBEntryReadOffset;
+#define RTNE 0
+#define RU 1
+#define RD 2
+#define RTZ 3
+ uint32_t RoundingMode;
+ bool BarrierEnable;
+#define Encodes0K 0
+#define Encodes1K 1
+#define Encodes2K 2
+#define Encodes4K 3
+#define Encodes8K 4
+#define Encodes16K 5
+#define Encodes32K 6
+#define Encodes64K 7
+ uint32_t SharedLocalMemorySize;
+ bool GlobalBarrierEnable;
+ uint32_t NumberofThreadsinGPGPUThreadGroup;
+ uint32_t CrossThreadConstantDataReadLength;
+};
+
+static inline void
+GEN9_INTERFACE_DESCRIPTOR_DATA_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_INTERFACE_DESCRIPTOR_DATA * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_offset(values->KernelStartPointer, 6, 31) |
+ 0;
+
+ dw[1] =
+ __gen_offset(values->KernelStartPointerHigh, 0, 15) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->DenormMode, 19, 19) |
+ __gen_field(values->SingleProgramFlow, 18, 18) |
+ __gen_field(values->ThreadPriority, 17, 17) |
+ __gen_field(values->FloatingPointMode, 16, 16) |
+ __gen_field(values->IllegalOpcodeExceptionEnable, 13, 13) |
+ __gen_field(values->MaskStackExceptionEnable, 11, 11) |
+ __gen_field(values->SoftwareExceptionEnable, 7, 7) |
+ 0;
+
+ dw[3] =
+ __gen_offset(values->SamplerStatePointer, 5, 31) |
+ __gen_field(values->SamplerCount, 2, 4) |
+ 0;
+
+ dw[4] =
+ __gen_offset(values->BindingTablePointer, 5, 15) |
+ __gen_field(values->BindingTableEntryCount, 0, 4) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->ConstantIndirectURBEntryReadLength, 16, 31) |
+ __gen_field(values->ConstantURBEntryReadOffset, 0, 15) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->RoundingMode, 22, 23) |
+ __gen_field(values->BarrierEnable, 21, 21) |
+ __gen_field(values->SharedLocalMemorySize, 16, 20) |
+ __gen_field(values->GlobalBarrierEnable, 15, 15) |
+ __gen_field(values->NumberofThreadsinGPGPUThreadGroup, 0, 9) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->CrossThreadConstantDataReadLength, 0, 7) |
+ 0;
+
+}
+
+#define GEN9_ROUNDINGPRECISIONTABLE_3_BITS_length 0x00000001
+
+struct GEN9_ROUNDINGPRECISIONTABLE_3_BITS {
+#define _116 0
+#define _216 1
+#define _316 2
+#define _416 3
+#define _516 4
+#define _616 5
+#define _716 6
+#define _816 7
+ uint32_t RoundingPrecision;
+};
+
+static inline void
+GEN9_ROUNDINGPRECISIONTABLE_3_BITS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_ROUNDINGPRECISIONTABLE_3_BITS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->RoundingPrecision, 0, 2) |
+ 0;
+
+}
+
+#define GEN9_BINDING_TABLE_STATE_length 0x00000001
+
+struct GEN9_BINDING_TABLE_STATE {
+ uint32_t SurfaceStatePointer;
+};
+
+static inline void
+GEN9_BINDING_TABLE_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_BINDING_TABLE_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_offset(values->SurfaceStatePointer, 6, 31) |
+ 0;
+
+}
+
+#define GEN9_RENDER_SURFACE_STATE_length 0x00000010
+
+struct GEN9_RENDER_SURFACE_STATE {
+#define SURFTYPE_1D 0
+#define SURFTYPE_2D 1
+#define SURFTYPE_3D 2
+#define SURFTYPE_CUBE 3
+#define SURFTYPE_BUFFER 4
+#define SURFTYPE_STRBUF 5
+#define SURFTYPE_NULL 7
+ uint32_t SurfaceType;
+ bool SurfaceArray;
+ bool ASTC_Enable;
+ uint32_t SurfaceFormat;
+#define VALIGN4 1
+#define VALIGN8 2
+#define VALIGN16 3
+ uint32_t SurfaceVerticalAlignment;
+#define HALIGN4 1
+#define HALIGN8 2
+#define HALIGN16 3
+ uint32_t SurfaceHorizontalAlignment;
+#define LINEAR 0
+#define WMAJOR 1
+#define XMAJOR 2
+#define YMAJOR 3
+ uint32_t TileMode;
+ uint32_t VerticalLineStride;
+ uint32_t VerticalLineStrideOffset;
+ bool SamplerL2BypassModeDisable;
+#define WriteOnlyCache 0
+#define ReadWriteCache 1
+ uint32_t RenderCacheReadWriteMode;
+#define NORMAL_MODE 0
+#define PROGRESSIVE_FRAME 2
+#define INTERLACED_FRAME 3
+ uint32_t MediaBoundaryPixelMode;
+ bool CubeFaceEnablePositiveZ;
+ bool CubeFaceEnableNegativeZ;
+ bool CubeFaceEnablePositiveY;
+ bool CubeFaceEnableNegativeY;
+ bool CubeFaceEnablePositiveX;
+ bool CubeFaceEnableNegativeX;
+ struct GEN9_MEMORY_OBJECT_CONTROL_STATE MemoryObjectControlState;
+ float BaseMipLevel;
+ uint32_t SurfaceQPitch;
+ uint32_t Height;
+ uint32_t Width;
+ uint32_t Depth;
+ uint32_t SurfacePitch;
+#define _0DEG 0
+#define _90DEG 1
+#define _180DEG 2
+#define _270DEG 3
+ uint32_t RenderTargetAndSampleUnormRotation;
+ uint32_t MinimumArrayElement;
+ uint32_t RenderTargetViewExtent;
+#define MSS 0
+#define DEPTH_STENCIL 1
+ uint32_t MultisampledSurfaceStorageFormat;
+#define MULTISAMPLECOUNT_1 0
+#define MULTISAMPLECOUNT_2 1
+#define MULTISAMPLECOUNT_4 2
+#define MULTISAMPLECOUNT_8 3
+#define MULTISAMPLECOUNT_16 4
+ uint32_t NumberofMultisamples;
+ uint32_t MultisamplePositionPaletteIndex;
+ uint32_t XOffset;
+ uint32_t YOffset;
+ bool EWADisableForCube;
+#define NONE 0
+#define _4KB 1
+#define _64KB 2
+#define TILEYF 1
+#define TILEYS 2
+ uint32_t TiledResourceMode;
+#define GPUcoherent 0
+#define IAcoherent 1
+ uint32_t CoherencyType;
+ uint32_t MipTailStartLOD;
+ uint32_t SurfaceMinLOD;
+ uint32_t MIPCountLOD;
+ uint32_t AuxiliarySurfaceQPitch;
+ uint32_t AuxiliarySurfacePitch;
+#define AUX_NONE 0
+#define AUX_CCS_D 1
+#define AUX_APPEND 2
+#define AUX_HIZ 3
+#define AUX_CCS_E 5
+ uint32_t AuxiliarySurfaceMode;
+ bool SeparateUVPlaneEnable;
+ uint32_t XOffsetforUorUVPlane;
+ uint32_t YOffsetforUorUVPlane;
+#define Horizontal 0
+#define Vertical 1
+ uint32_t MemoryCompressionMode;
+ bool MemoryCompressionEnable;
+ uint32_t ShaderChannelSelectRed;
+ uint32_t ShaderChannelSelectGreen;
+ uint32_t ShaderChannelSelectBlue;
+ uint32_t ShaderChannelSelectAlpha;
+ float ResourceMinLOD;
+ __gen_address_type SurfaceBaseAddress;
+ uint32_t XOffsetforVPlane;
+ uint32_t YOffsetforVPlane;
+ uint32_t AuxiliaryTableIndexforMediaCompressedSurface;
+ __gen_address_type AuxiliarySurfaceBaseAddress;
+ uint32_t QuiltHeight;
+ uint32_t QuiltWidth;
+ float HierarchicalDepthClearValue;
+ uint32_t RedClearColor;
+ uint32_t GreenClearColor;
+ uint32_t BlueClearColor;
+ uint32_t AlphaClearColor;
+};
+
+static inline void
+GEN9_RENDER_SURFACE_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_RENDER_SURFACE_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->SurfaceType, 29, 31) |
+ __gen_field(values->SurfaceArray, 28, 28) |
+ __gen_field(values->ASTC_Enable, 27, 27) |
+ __gen_field(values->SurfaceFormat, 18, 26) |
+ __gen_field(values->SurfaceVerticalAlignment, 16, 17) |
+ __gen_field(values->SurfaceHorizontalAlignment, 14, 15) |
+ __gen_field(values->TileMode, 12, 13) |
+ __gen_field(values->VerticalLineStride, 11, 11) |
+ __gen_field(values->VerticalLineStrideOffset, 10, 10) |
+ __gen_field(values->SamplerL2BypassModeDisable, 9, 9) |
+ __gen_field(values->RenderCacheReadWriteMode, 8, 8) |
+ __gen_field(values->MediaBoundaryPixelMode, 6, 7) |
+ __gen_field(values->CubeFaceEnablePositiveZ, 0, 0) |
+ __gen_field(values->CubeFaceEnableNegativeZ, 1, 1) |
+ __gen_field(values->CubeFaceEnablePositiveY, 2, 2) |
+ __gen_field(values->CubeFaceEnableNegativeY, 3, 3) |
+ __gen_field(values->CubeFaceEnablePositiveX, 4, 4) |
+ __gen_field(values->CubeFaceEnableNegativeX, 5, 5) |
+ 0;
+
+ uint32_t dw_MemoryObjectControlState;
+ GEN9_MEMORY_OBJECT_CONTROL_STATE_pack(data, &dw_MemoryObjectControlState, &values->MemoryObjectControlState);
+ dw[1] =
+ __gen_field(dw_MemoryObjectControlState, 24, 30) |
+ __gen_field(values->BaseMipLevel * (1 << 1), 19, 23) |
+ __gen_field(values->SurfaceQPitch, 0, 14) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->Height, 16, 29) |
+ __gen_field(values->Width, 0, 13) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->Depth, 21, 31) |
+ __gen_field(values->SurfacePitch, 0, 17) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->RenderTargetAndSampleUnormRotation, 29, 30) |
+ __gen_field(values->MinimumArrayElement, 18, 28) |
+ __gen_field(values->RenderTargetViewExtent, 7, 17) |
+ __gen_field(values->MultisampledSurfaceStorageFormat, 6, 6) |
+ __gen_field(values->NumberofMultisamples, 3, 5) |
+ __gen_field(values->MultisamplePositionPaletteIndex, 0, 2) |
+ 0;
+
+ dw[5] =
+ __gen_offset(values->XOffset, 25, 31) |
+ __gen_offset(values->YOffset, 21, 23) |
+ __gen_field(values->EWADisableForCube, 20, 20) |
+ __gen_field(values->TiledResourceMode, 18, 19) |
+ __gen_field(values->CoherencyType, 14, 14) |
+ __gen_field(values->MipTailStartLOD, 8, 11) |
+ __gen_field(values->SurfaceMinLOD, 4, 7) |
+ __gen_field(values->MIPCountLOD, 0, 3) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->AuxiliarySurfaceQPitch, 16, 30) |
+ __gen_field(values->AuxiliarySurfacePitch, 3, 11) |
+ __gen_field(values->AuxiliarySurfaceMode, 0, 2) |
+ __gen_field(values->SeparateUVPlaneEnable, 31, 31) |
+ __gen_field(values->XOffsetforUorUVPlane, 16, 29) |
+ __gen_field(values->YOffsetforUorUVPlane, 0, 13) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->MemoryCompressionMode, 31, 31) |
+ __gen_field(values->MemoryCompressionEnable, 30, 30) |
+ __gen_field(values->ShaderChannelSelectRed, 25, 27) |
+ __gen_field(values->ShaderChannelSelectGreen, 22, 24) |
+ __gen_field(values->ShaderChannelSelectBlue, 19, 21) |
+ __gen_field(values->ShaderChannelSelectAlpha, 16, 18) |
+ __gen_field(values->ResourceMinLOD * (1 << 8), 0, 11) |
+ 0;
+
+ uint32_t dw8 =
+ 0;
+
+ uint64_t qw8 =
+ __gen_combine_address(data, &dw[8], values->SurfaceBaseAddress, dw8);
+
+ dw[8] = qw8;
+ dw[9] = qw8 >> 32;
+
+ uint32_t dw10 =
+ __gen_field(values->XOffsetforVPlane, 48, 61) |
+ __gen_field(values->YOffsetforVPlane, 32, 45) |
+ __gen_field(values->AuxiliaryTableIndexforMediaCompressedSurface, 21, 31) |
+ __gen_field(values->QuiltHeight, 5, 9) |
+ __gen_field(values->QuiltWidth, 0, 4) |
+ 0;
+
+ uint64_t qw10 =
+ __gen_combine_address(data, &dw[10], values->AuxiliarySurfaceBaseAddress, dw10);
+
+ dw[10] = qw10;
+ dw[11] = qw10 >> 32;
+
+ dw[12] =
+ __gen_float(values->HierarchicalDepthClearValue) |
+ __gen_field(values->RedClearColor, 0, 31) |
+ 0;
+
+ dw[13] =
+ __gen_field(values->GreenClearColor, 0, 31) |
+ 0;
+
+ dw[14] =
+ __gen_field(values->BlueClearColor, 0, 31) |
+ 0;
+
+ dw[15] =
+ __gen_field(values->AlphaClearColor, 0, 31) |
+ 0;
+
+}
+
+#define GEN9_FILTER_COEFFICIENT_length 0x00000001
+
+struct GEN9_FILTER_COEFFICIENT {
+ uint32_t FilterCoefficient;
+};
+
+static inline void
+GEN9_FILTER_COEFFICIENT_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_FILTER_COEFFICIENT * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->FilterCoefficient, 0, 7) |
+ 0;
+
+}
+
+#define GEN9_SAMPLER_STATE_length 0x00000004
+
+struct GEN9_SAMPLER_STATE {
+ bool SamplerDisable;
+#define DX10OGL 0
+#define DX9 1
+ uint32_t TextureBorderColorMode;
+#define CLAMP_NONE 0
+#define CLAMP_OGL 2
+ uint32_t LODPreClampMode;
+ uint32_t CoarseLODQualityMode;
+#define MIPFILTER_NONE 0
+#define MIPFILTER_NEAREST 1
+#define MIPFILTER_LINEAR 3
+ uint32_t MipModeFilter;
+#define MAPFILTER_NEAREST 0
+#define MAPFILTER_LINEAR 1
+#define MAPFILTER_ANISOTROPIC 2
+#define MAPFILTER_MONO 6
+ uint32_t MagModeFilter;
+#define MAPFILTER_NEAREST 0
+#define MAPFILTER_LINEAR 1
+#define MAPFILTER_ANISOTROPIC 2
+#define MAPFILTER_MONO 6
+ uint32_t MinModeFilter;
+ float TextureLODBias;
+#define LEGACY 0
+#define EWAApproximation 1
+ uint32_t AnisotropicAlgorithm;
+ float MinLOD;
+ float MaxLOD;
+ bool ChromaKeyEnable;
+ uint32_t ChromaKeyIndex;
+#define KEYFILTER_KILL_ON_ANY_MATCH 0
+#define KEYFILTER_REPLACE_BLACK 1
+ uint32_t ChromaKeyMode;
+#define PREFILTEROPALWAYS 0
+#define PREFILTEROPNEVER 1
+#define PREFILTEROPLESS 2
+#define PREFILTEROPEQUAL 3
+#define PREFILTEROPLEQUAL 4
+#define PREFILTEROPGREATER 5
+#define PREFILTEROPNOTEQUAL 6
+#define PREFILTEROPGEQUAL 7
+ uint32_t ShadowFunction;
+#define PROGRAMMED 0
+#define OVERRIDE 1
+ uint32_t CubeSurfaceControlMode;
+ uint32_t IndirectStatePointer;
+#define MIPNONE 0
+#define MIPFILTER 1
+ uint32_t LODClampMagnificationMode;
+#define STD_FILTER 0
+#define COMPARISON 1
+#define MINIMUM 2
+#define MAXIMUM 3
+ uint32_t ReductionType;
+#define RATIO21 0
+#define RATIO41 1
+#define RATIO61 2
+#define RATIO81 3
+#define RATIO101 4
+#define RATIO121 5
+#define RATIO141 6
+#define RATIO161 7
+ uint32_t MaximumAnisotropy;
+ bool RAddressMinFilterRoundingEnable;
+ bool RAddressMagFilterRoundingEnable;
+ bool VAddressMinFilterRoundingEnable;
+ bool VAddressMagFilterRoundingEnable;
+ bool UAddressMinFilterRoundingEnable;
+ bool UAddressMagFilterRoundingEnable;
+#define FULL 0
+#define HIGH 1
+#define MED 2
+#define LOW 3
+ uint32_t TrilinearFilterQuality;
+ bool NonnormalizedCoordinateEnable;
+ bool ReductionTypeEnable;
+ uint32_t TCXAddressControlMode;
+ uint32_t TCYAddressControlMode;
+ uint32_t TCZAddressControlMode;
+};
+
+static inline void
+GEN9_SAMPLER_STATE_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_SAMPLER_STATE * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->SamplerDisable, 31, 31) |
+ __gen_field(values->TextureBorderColorMode, 29, 29) |
+ __gen_field(values->LODPreClampMode, 27, 28) |
+ __gen_field(values->CoarseLODQualityMode, 22, 26) |
+ __gen_field(values->MipModeFilter, 20, 21) |
+ __gen_field(values->MagModeFilter, 17, 19) |
+ __gen_field(values->MinModeFilter, 14, 16) |
+ __gen_fixed(values->TextureLODBias, 1, 13, true, 8) |
+ __gen_field(values->AnisotropicAlgorithm, 0, 0) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->MinLOD * (1 << 8), 20, 31) |
+ __gen_field(values->MaxLOD * (1 << 8), 8, 19) |
+ __gen_field(values->ChromaKeyEnable, 7, 7) |
+ __gen_field(values->ChromaKeyIndex, 5, 6) |
+ __gen_field(values->ChromaKeyMode, 4, 4) |
+ __gen_field(values->ShadowFunction, 1, 3) |
+ __gen_field(values->CubeSurfaceControlMode, 0, 0) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->IndirectStatePointer, 6, 23) |
+ __gen_field(values->LODClampMagnificationMode, 0, 0) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->ReductionType, 22, 23) |
+ __gen_field(values->MaximumAnisotropy, 19, 21) |
+ __gen_field(values->RAddressMinFilterRoundingEnable, 13, 13) |
+ __gen_field(values->RAddressMagFilterRoundingEnable, 14, 14) |
+ __gen_field(values->VAddressMinFilterRoundingEnable, 15, 15) |
+ __gen_field(values->VAddressMagFilterRoundingEnable, 16, 16) |
+ __gen_field(values->UAddressMinFilterRoundingEnable, 17, 17) |
+ __gen_field(values->UAddressMagFilterRoundingEnable, 18, 18) |
+ __gen_field(values->TrilinearFilterQuality, 11, 12) |
+ __gen_field(values->NonnormalizedCoordinateEnable, 10, 10) |
+ __gen_field(values->ReductionTypeEnable, 9, 9) |
+ __gen_field(values->TCXAddressControlMode, 6, 8) |
+ __gen_field(values->TCYAddressControlMode, 3, 5) |
+ __gen_field(values->TCZAddressControlMode, 0, 2) |
+ 0;
+
+}
+
+#define GEN9_SAMPLER_STATE_8X8_AVS_COEFFICIENTS_length 0x00000008
+
+struct GEN9_SAMPLER_STATE_8X8_AVS_COEFFICIENTS {
+ uint32_t Table0YFilterCoefficientn1;
+ uint32_t Table0XFilterCoefficientn1;
+ uint32_t Table0YFilterCoefficientn0;
+ uint32_t Table0XFilterCoefficientn0;
+ uint32_t Table0YFilterCoefficientn3;
+ uint32_t Table0XFilterCoefficientn3;
+ uint32_t Table0YFilterCoefficientn2;
+ uint32_t Table0XFilterCoefficientn2;
+ uint32_t Table0YFilterCoefficientn5;
+ uint32_t Table0XFilterCoefficientn5;
+ uint32_t Table0YFilterCoefficientn4;
+ uint32_t Table0XFilterCoefficientn4;
+ uint32_t Table0YFilterCoefficientn7;
+ uint32_t Table0XFilterCoefficientn7;
+ uint32_t Table0YFilterCoefficientn6;
+ uint32_t Table0XFilterCoefficientn6;
+ uint32_t Table1XFilterCoefficientn3;
+ uint32_t Table1XFilterCoefficientn2;
+ uint32_t Table1XFilterCoefficientn5;
+ uint32_t Table1XFilterCoefficientn4;
+ uint32_t Table1YFilterCoefficientn3;
+ uint32_t Table1YFilterCoefficientn2;
+ uint32_t Table1YFilterCoefficientn5;
+ uint32_t Table1YFilterCoefficientn4;
+};
+
+static inline void
+GEN9_SAMPLER_STATE_8X8_AVS_COEFFICIENTS_pack(__gen_user_data *data, void * restrict dst,
+ const struct GEN9_SAMPLER_STATE_8X8_AVS_COEFFICIENTS * restrict values)
+{
+ uint32_t *dw = (uint32_t * restrict) dst;
+
+ dw[0] =
+ __gen_field(values->Table0YFilterCoefficientn1, 24, 31) |
+ __gen_field(values->Table0XFilterCoefficientn1, 16, 23) |
+ __gen_field(values->Table0YFilterCoefficientn0, 8, 15) |
+ __gen_field(values->Table0XFilterCoefficientn0, 0, 7) |
+ 0;
+
+ dw[1] =
+ __gen_field(values->Table0YFilterCoefficientn3, 24, 31) |
+ __gen_field(values->Table0XFilterCoefficientn3, 16, 23) |
+ __gen_field(values->Table0YFilterCoefficientn2, 8, 15) |
+ __gen_field(values->Table0XFilterCoefficientn2, 0, 7) |
+ 0;
+
+ dw[2] =
+ __gen_field(values->Table0YFilterCoefficientn5, 24, 31) |
+ __gen_field(values->Table0XFilterCoefficientn5, 16, 23) |
+ __gen_field(values->Table0YFilterCoefficientn4, 8, 15) |
+ __gen_field(values->Table0XFilterCoefficientn4, 0, 7) |
+ 0;
+
+ dw[3] =
+ __gen_field(values->Table0YFilterCoefficientn7, 24, 31) |
+ __gen_field(values->Table0XFilterCoefficientn7, 16, 23) |
+ __gen_field(values->Table0YFilterCoefficientn6, 8, 15) |
+ __gen_field(values->Table0XFilterCoefficientn6, 0, 7) |
+ 0;
+
+ dw[4] =
+ __gen_field(values->Table1XFilterCoefficientn3, 24, 31) |
+ __gen_field(values->Table1XFilterCoefficientn2, 16, 23) |
+ 0;
+
+ dw[5] =
+ __gen_field(values->Table1XFilterCoefficientn5, 8, 15) |
+ __gen_field(values->Table1XFilterCoefficientn4, 0, 7) |
+ 0;
+
+ dw[6] =
+ __gen_field(values->Table1YFilterCoefficientn3, 24, 31) |
+ __gen_field(values->Table1YFilterCoefficientn2, 16, 23) |
+ 0;
+
+ dw[7] =
+ __gen_field(values->Table1YFilterCoefficientn5, 8, 15) |
+ __gen_field(values->Table1YFilterCoefficientn4, 0, 7) |
+ 0;
+
+}
+
+/* Enum 3D_Prim_Topo_Type */
+#define _3DPRIM_POINTLIST 1
+#define _3DPRIM_LINELIST 2
+#define _3DPRIM_LINESTRIP 3
+#define _3DPRIM_TRILIST 4
+#define _3DPRIM_TRISTRIP 5
+#define _3DPRIM_TRIFAN 6
+#define _3DPRIM_QUADLIST 7
+#define _3DPRIM_QUADSTRIP 8
+#define _3DPRIM_LINELIST_ADJ 9
+#define _3DPRIM_LINESTRIP_ADJ 10
+#define _3DPRIM_TRILIST_ADJ 11
+#define _3DPRIM_TRISTRIP_ADJ 12
+#define _3DPRIM_TRISTRIP_REVERSE 13
+#define _3DPRIM_POLYGON 14
+#define _3DPRIM_RECTLIST 15
+#define _3DPRIM_LINELOOP 16
+#define _3DPRIM_POINTLIST_BF 17
+#define _3DPRIM_LINESTRIP_CONT 18
+#define _3DPRIM_LINESTRIP_BF 19
+#define _3DPRIM_LINESTRIP_CONT_BF 20
+#define _3DPRIM_TRIFAN_NOSTIPPLE 22
+#define _3DPRIM_PATCHLIST_1 32
+#define _3DPRIM_PATCHLIST_2 33
+#define _3DPRIM_PATCHLIST_3 34
+#define _3DPRIM_PATCHLIST_4 35
+#define _3DPRIM_PATCHLIST_5 36
+#define _3DPRIM_PATCHLIST_6 37
+#define _3DPRIM_PATCHLIST_7 38
+#define _3DPRIM_PATCHLIST_8 39
+#define _3DPRIM_PATCHLIST_9 40
+#define _3DPRIM_PATCHLIST_10 41
+#define _3DPRIM_PATCHLIST_11 42
+#define _3DPRIM_PATCHLIST_12 43
+#define _3DPRIM_PATCHLIST_13 44
+#define _3DPRIM_PATCHLIST_14 45
+#define _3DPRIM_PATCHLIST_15 46
+#define _3DPRIM_PATCHLIST_16 47
+#define _3DPRIM_PATCHLIST_17 48
+#define _3DPRIM_PATCHLIST_18 49
+#define _3DPRIM_PATCHLIST_19 50
+#define _3DPRIM_PATCHLIST_20 51
+#define _3DPRIM_PATCHLIST_21 52
+#define _3DPRIM_PATCHLIST_22 53
+#define _3DPRIM_PATCHLIST_23 54
+#define _3DPRIM_PATCHLIST_24 55
+#define _3DPRIM_PATCHLIST_25 56
+#define _3DPRIM_PATCHLIST_26 57
+#define _3DPRIM_PATCHLIST_27 58
+#define _3DPRIM_PATCHLIST_28 59
+#define _3DPRIM_PATCHLIST_29 60
+#define _3DPRIM_PATCHLIST_30 61
+#define _3DPRIM_PATCHLIST_31 62
+#define _3DPRIM_PATCHLIST_32 63
+
+/* Enum 3D_Vertex_Component_Control */
+#define VFCOMP_NOSTORE 0
+#define VFCOMP_STORE_SRC 1
+#define VFCOMP_STORE_0 2
+#define VFCOMP_STORE_1_FP 3
+#define VFCOMP_STORE_1_INT 4
+#define VFCOMP_STORE_PID 7
+
+/* Enum COMPONENT_ENABLES */
+#define CE_NONE 0
+#define CE_X 1
+#define CE_Y 2
+#define CE_XY 3
+#define CE_Z 4
+#define CE_XZ 5
+#define CE_YZ 6
+#define CE_XYZ 7
+#define CE_W 8
+#define CE_XW 9
+#define CE_YW 10
+#define CE_XYW 11
+#define CE_ZW 12
+#define CE_XZW 13
+#define CE_YZW 14
+#define CE_XYZW 15
+
+/* Enum Attribute_Component_Format */
+#define ACF_DISABLED 0
+#define ACF_XY 1
+#define ACF_XYZ 2
+#define ACF_XYZW 3
+
+/* Enum WRAP_SHORTEST_ENABLE */
+#define WSE_X 1
+#define WSE_Y 2
+#define WSE_XY 3
+#define WSE_Z 4
+#define WSE_XZ 5
+#define WSE_YZ 6
+#define WSE_XYZ 7
+#define WSE_W 8
+#define WSE_XW 9
+#define WSE_YW 10
+#define WSE_XYW 11
+#define WSE_ZW 12
+#define WSE_XZW 13
+#define WSE_YZW 14
+#define WSE_XYZW 15
+
+/* Enum 3D_Stencil_Operation */
+#define STENCILOP_KEEP 0
+#define STENCILOP_ZERO 1
+#define STENCILOP_REPLACE 2
+#define STENCILOP_INCRSAT 3
+#define STENCILOP_DECRSAT 4
+#define STENCILOP_INCR 5
+#define STENCILOP_DECR 6
+#define STENCILOP_INVERT 7
+
+/* Enum 3D_Color_Buffer_Blend_Factor */
+#define BLENDFACTOR_ONE 1
+#define BLENDFACTOR_SRC_COLOR 2
+#define BLENDFACTOR_SRC_ALPHA 3
+#define BLENDFACTOR_DST_ALPHA 4
+#define BLENDFACTOR_DST_COLOR 5
+#define BLENDFACTOR_SRC_ALPHA_SATURATE 6
+#define BLENDFACTOR_CONST_COLOR 7
+#define BLENDFACTOR_CONST_ALPHA 8
+#define BLENDFACTOR_SRC1_COLOR 9
+#define BLENDFACTOR_SRC1_ALPHA 10
+#define BLENDFACTOR_ZERO 17
+#define BLENDFACTOR_INV_SRC_COLOR 18
+#define BLENDFACTOR_INV_SRC_ALPHA 19
+#define BLENDFACTOR_INV_DST_ALPHA 20
+#define BLENDFACTOR_INV_DST_COLOR 21
+#define BLENDFACTOR_INV_CONST_COLOR 23
+#define BLENDFACTOR_INV_CONST_ALPHA 24
+#define BLENDFACTOR_INV_SRC1_COLOR 25
+#define BLENDFACTOR_INV_SRC1_ALPHA 26
+
+/* Enum 3D_Color_Buffer_Blend_Function */
+#define BLENDFUNCTION_ADD 0
+#define BLENDFUNCTION_SUBTRACT 1
+#define BLENDFUNCTION_REVERSE_SUBTRACT 2
+#define BLENDFUNCTION_MIN 3
+#define BLENDFUNCTION_MAX 4
+
+/* Enum 3D_Compare_Function */
+#define COMPAREFUNCTION_ALWAYS 0
+#define COMPAREFUNCTION_NEVER 1
+#define COMPAREFUNCTION_LESS 2
+#define COMPAREFUNCTION_EQUAL 3
+#define COMPAREFUNCTION_LEQUAL 4
+#define COMPAREFUNCTION_GREATER 5
+#define COMPAREFUNCTION_NOTEQUAL 6
+#define COMPAREFUNCTION_GEQUAL 7
+
+/* Enum 3D_Logic_Op_Function */
+#define LOGICOP_CLEAR 0
+#define LOGICOP_NOR 1
+#define LOGICOP_AND_INVERTED 2
+#define LOGICOP_COPY_INVERTED 3
+#define LOGICOP_AND_REVERSE 4
+#define LOGICOP_INVERT 5
+#define LOGICOP_XOR 6
+#define LOGICOP_NAND 7
+#define LOGICOP_AND 8
+#define LOGICOP_EQUIV 9
+#define LOGICOP_NOOP 10
+#define LOGICOP_OR_INVERTED 11
+#define LOGICOP_COPY 12
+#define LOGICOP_OR_REVERSE 13
+#define LOGICOP_OR 14
+#define LOGICOP_SET 15
+
+/* Enum SURFACE_FORMAT */
+#define R32G32B32A32_FLOAT 0
+#define R32G32B32A32_SINT 1
+#define R32G32B32A32_UINT 2
+#define R32G32B32A32_UNORM 3
+#define R32G32B32A32_SNORM 4
+#define R64G64_FLOAT 5
+#define R32G32B32X32_FLOAT 6
+#define R32G32B32A32_SSCALED 7
+#define R32G32B32A32_USCALED 8
+#define R32G32B32A32_SFIXED 32
+#define R64G64_PASSTHRU 33
+#define R32G32B32_FLOAT 64
+#define R32G32B32_SINT 65
+#define R32G32B32_UINT 66
+#define R32G32B32_UNORM 67
+#define R32G32B32_SNORM 68
+#define R32G32B32_SSCALED 69
+#define R32G32B32_USCALED 70
+#define R32G32B32_SFIXED 80
+#define R16G16B16A16_UNORM 128
+#define R16G16B16A16_SNORM 129
+#define R16G16B16A16_SINT 130
+#define R16G16B16A16_UINT 131
+#define R16G16B16A16_FLOAT 132
+#define R32G32_FLOAT 133
+#define R32G32_SINT 134
+#define R32G32_UINT 135
+#define R32_FLOAT_X8X24_TYPELESS 136
+#define X32_TYPELESS_G8X24_UINT 137
+#define L32A32_FLOAT 138
+#define R32G32_UNORM 139
+#define R32G32_SNORM 140
+#define R64_FLOAT 141
+#define R16G16B16X16_UNORM 142
+#define R16G16B16X16_FLOAT 143
+#define A32X32_FLOAT 144
+#define L32X32_FLOAT 145
+#define I32X32_FLOAT 146
+#define R16G16B16A16_SSCALED 147
+#define R16G16B16A16_USCALED 148
+#define R32G32_SSCALED 149
+#define R32G32_USCALED 150
+#define R32G32_SFIXED 160
+#define R64_PASSTHRU 161
+#define B8G8R8A8_UNORM 192
+#define B8G8R8A8_UNORM_SRGB 193
+#define R10G10B10A2_UNORM 194
+#define R10G10B10A2_UNORM_SRGB 195
+#define R10G10B10A2_UINT 196
+#define R10G10B10_SNORM_A2_UNORM 197
+#define R8G8B8A8_UNORM 199
+#define R8G8B8A8_UNORM_SRGB 200
+#define R8G8B8A8_SNORM 201
+#define R8G8B8A8_SINT 202
+#define R8G8B8A8_UINT 203
+#define R16G16_UNORM 204
+#define R16G16_SNORM 205
+#define R16G16_SINT 206
+#define R16G16_UINT 207
+#define R16G16_FLOAT 208
+#define B10G10R10A2_UNORM 209
+#define B10G10R10A2_UNORM_SRGB 210
+#define R11G11B10_FLOAT 211
+#define R32_SINT 214
+#define R32_UINT 215
+#define R32_FLOAT 216
+#define R24_UNORM_X8_TYPELESS 217
+#define X24_TYPELESS_G8_UINT 218
+#define L32_UNORM 221
+#define A32_UNORM 222
+#define L16A16_UNORM 223
+#define I24X8_UNORM 224
+#define L24X8_UNORM 225
+#define A24X8_UNORM 226
+#define I32_FLOAT 227
+#define L32_FLOAT 228
+#define A32_FLOAT 229
+#define X8B8_UNORM_G8R8_SNORM 230
+#define A8X8_UNORM_G8R8_SNORM 231
+#define B8X8_UNORM_G8R8_SNORM 232
+#define B8G8R8X8_UNORM 233
+#define B8G8R8X8_UNORM_SRGB 234
+#define R8G8B8X8_UNORM 235
+#define R8G8B8X8_UNORM_SRGB 236
+#define R9G9B9E5_SHAREDEXP 237
+#define B10G10R10X2_UNORM 238
+#define L16A16_FLOAT 240
+#define R32_UNORM 241
+#define R32_SNORM 242
+#define R10G10B10X2_USCALED 243
+#define R8G8B8A8_SSCALED 244
+#define R8G8B8A8_USCALED 245
+#define R16G16_SSCALED 246
+#define R16G16_USCALED 247
+#define R32_SSCALED 248
+#define R32_USCALED 249
+#define B5G6R5_UNORM 256
+#define B5G6R5_UNORM_SRGB 257
+#define B5G5R5A1_UNORM 258
+#define B5G5R5A1_UNORM_SRGB 259
+#define B4G4R4A4_UNORM 260
+#define B4G4R4A4_UNORM_SRGB 261
+#define R8G8_UNORM 262
+#define R8G8_SNORM 263
+#define R8G8_SINT 264
+#define R8G8_UINT 265
+#define R16_UNORM 266
+#define R16_SNORM 267
+#define R16_SINT 268
+#define R16_UINT 269
+#define R16_FLOAT 270
+#define A8P8_UNORM_PALETTE0 271
+#define A8P8_UNORM_PALETTE1 272
+#define I16_UNORM 273
+#define L16_UNORM 274
+#define A16_UNORM 275
+#define L8A8_UNORM 276
+#define I16_FLOAT 277
+#define L16_FLOAT 278
+#define A16_FLOAT 279
+#define L8A8_UNORM_SRGB 280
+#define R5G5_SNORM_B6_UNORM 281
+#define B5G5R5X1_UNORM 282
+#define B5G5R5X1_UNORM_SRGB 283
+#define R8G8_SSCALED 284
+#define R8G8_USCALED 285
+#define R16_SSCALED 286
+#define R16_USCALED 287
+#define P8A8_UNORM_PALETTE0 290
+#define P8A8_UNORM_PALETTE1 291
+#define A1B5G5R5_UNORM 292
+#define A4B4G4R4_UNORM 293
+#define L8A8_UINT 294
+#define L8A8_SINT 295
+#define R8_UNORM 320
+#define R8_SNORM 321
+#define R8_SINT 322
+#define R8_UINT 323
+#define A8_UNORM 324
+#define I8_UNORM 325
+#define L8_UNORM 326
+#define P4A4_UNORM_PALETTE0 327
+#define A4P4_UNORM_PALETTE0 328
+#define R8_SSCALED 329
+#define R8_USCALED 330
+#define P8_UNORM_PALETTE0 331
+#define L8_UNORM_SRGB 332
+#define P8_UNORM_PALETTE1 333
+#define P4A4_UNORM_PALETTE1 334
+#define A4P4_UNORM_PALETTE1 335
+#define Y8_UNORM 336
+#define L8_UINT 338
+#define L8_SINT 339
+#define I8_UINT 340
+#define I8_SINT 341
+#define DXT1_RGB_SRGB 384
+#define R1_UNORM 385
+#define YCRCB_NORMAL 386
+#define YCRCB_SWAPUVY 387
+#define P2_UNORM_PALETTE0 388
+#define P2_UNORM_PALETTE1 389
+#define BC1_UNORM 390
+#define BC2_UNORM 391
+#define BC3_UNORM 392
+#define BC4_UNORM 393
+#define BC5_UNORM 394
+#define BC1_UNORM_SRGB 395
+#define BC2_UNORM_SRGB 396
+#define BC3_UNORM_SRGB 397
+#define MONO8 398
+#define YCRCB_SWAPUV 399
+#define YCRCB_SWAPY 400
+#define DXT1_RGB 401
+#define FXT1 402
+#define R8G8B8_UNORM 403
+#define R8G8B8_SNORM 404
+#define R8G8B8_SSCALED 405
+#define R8G8B8_USCALED 406
+#define R64G64B64A64_FLOAT 407
+#define R64G64B64_FLOAT 408
+#define BC4_SNORM 409
+#define BC5_SNORM 410
+#define R16G16B16_FLOAT 411
+#define R16G16B16_UNORM 412
+#define R16G16B16_SNORM 413
+#define R16G16B16_SSCALED 414
+#define R16G16B16_USCALED 415
+#define BC6H_SF16 417
+#define BC7_UNORM 418
+#define BC7_UNORM_SRGB 419
+#define BC6H_UF16 420
+#define PLANAR_420_8 421
+#define R8G8B8_UNORM_SRGB 424
+#define ETC1_RGB8 425
+#define ETC2_RGB8 426
+#define EAC_R11 427
+#define EAC_RG11 428
+#define EAC_SIGNED_R11 429
+#define EAC_SIGNED_RG11 430
+#define ETC2_SRGB8 431
+#define R16G16B16_UINT 432
+#define R16G16B16_SINT 433
+#define R32_SFIXED 434
+#define R10G10B10A2_SNORM 435
+#define R10G10B10A2_USCALED 436
+#define R10G10B10A2_SSCALED 437
+#define R10G10B10A2_SINT 438
+#define B10G10R10A2_SNORM 439
+#define B10G10R10A2_USCALED 440
+#define B10G10R10A2_SSCALED 441
+#define B10G10R10A2_UINT 442
+#define B10G10R10A2_SINT 443
+#define R64G64B64A64_PASSTHRU 444
+#define R64G64B64_PASSTHRU 445
+#define ETC2_RGB8_PTA 448
+#define ETC2_SRGB8_PTA 449
+#define ETC2_EAC_RGBA8 450
+#define ETC2_EAC_SRGB8_A8 451
+#define R8G8B8_UINT 456
+#define R8G8B8_SINT 457
+#define RAW 511
+
+/* Enum Shader Channel Select */
+#define SCS_ZERO 0
+#define SCS_ONE 1
+#define SCS_RED 4
+#define SCS_GREEN 5
+#define SCS_BLUE 6
+#define SCS_ALPHA 7
+
+/* Enum Texture Coordinate Mode */
+#define TCM_WRAP 0
+#define TCM_MIRROR 1
+#define TCM_CLAMP 2
+#define TCM_CUBE 3
+#define TCM_CLAMP_BORDER 4
+#define TCM_MIRROR_ONCE 5
+#define TCM_HALF_BORDER 6
+
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <assert.h>
+#include <stdbool.h>
+
+#include "anv_private.h"
+
+#if (ANV_GEN == 9)
+# include "gen9_pack.h"
+#elif (ANV_GEN == 8)
+# include "gen8_pack.h"
+#elif (ANV_IS_HASWELL)
+# include "gen75_pack.h"
+#elif (ANV_GEN == 7)
+# include "gen7_pack.h"
+#endif
+
+void
+genX(cmd_buffer_emit_state_base_address)(struct anv_cmd_buffer *cmd_buffer)
+{
+ struct anv_device *device = cmd_buffer->device;
+ struct anv_bo *scratch_bo = NULL;
+
+ cmd_buffer->state.scratch_size =
+ anv_block_pool_size(&device->scratch_block_pool);
+ if (cmd_buffer->state.scratch_size > 0)
+ scratch_bo = &device->scratch_block_pool.bo;
+
+/* XXX: Do we need this on more than just BDW? */
+#if (ANV_GEN == 8)
+ /* Emit a render target cache flush.
+ *
+ * This isn't documented anywhere in the PRM. However, it seems to be
+ * necessary prior to changing the surface state base adress. Without
+ * this, we get GPU hangs when using multi-level command buffers which
+ * clear depth, reset state base address, and then go render stuff.
+ */
+ anv_batch_emit(&cmd_buffer->batch, GEN8_PIPE_CONTROL,
+ .RenderTargetCacheFlushEnable = true);
+#endif
+
+ anv_batch_emit(&cmd_buffer->batch, GENX(STATE_BASE_ADDRESS),
+ .GeneralStateBaseAddress = { scratch_bo, 0 },
+ .GeneralStateMemoryObjectControlState = GENX(MOCS),
+ .GeneralStateBaseAddressModifyEnable = true,
+
+ .SurfaceStateBaseAddress = anv_cmd_buffer_surface_base_address(cmd_buffer),
+ .SurfaceStateMemoryObjectControlState = GENX(MOCS),
+ .SurfaceStateBaseAddressModifyEnable = true,
+
+ .DynamicStateBaseAddress = { &device->dynamic_state_block_pool.bo, 0 },
+ .DynamicStateMemoryObjectControlState = GENX(MOCS),
+ .DynamicStateBaseAddressModifyEnable = true,
+
+ .IndirectObjectBaseAddress = { NULL, 0 },
+ .IndirectObjectMemoryObjectControlState = GENX(MOCS),
+ .IndirectObjectBaseAddressModifyEnable = true,
+
+ .InstructionBaseAddress = { &device->instruction_block_pool.bo, 0 },
+ .InstructionMemoryObjectControlState = GENX(MOCS),
+ .InstructionBaseAddressModifyEnable = true,
+
+# if (ANV_GEN >= 8)
+ /* Broadwell requires that we specify a buffer size for a bunch of
+ * these fields. However, since we will be growing the BO's live, we
+ * just set them all to the maximum.
+ */
+ .GeneralStateBufferSize = 0xfffff,
+ .GeneralStateBufferSizeModifyEnable = true,
+ .DynamicStateBufferSize = 0xfffff,
+ .DynamicStateBufferSizeModifyEnable = true,
+ .IndirectObjectBufferSize = 0xfffff,
+ .IndirectObjectBufferSizeModifyEnable = true,
+ .InstructionBufferSize = 0xfffff,
+ .InstructionBuffersizeModifyEnable = true,
+# endif
+ );
+
+ /* After re-setting the surface state base address, we have to do some
+ * cache flusing so that the sampler engine will pick up the new
+ * SURFACE_STATE objects and binding tables. From the Broadwell PRM,
+ * Shared Function > 3D Sampler > State > State Caching (page 96):
+ *
+ * Coherency with system memory in the state cache, like the texture
+ * cache is handled partially by software. It is expected that the
+ * command stream or shader will issue Cache Flush operation or
+ * Cache_Flush sampler message to ensure that the L1 cache remains
+ * coherent with system memory.
+ *
+ * [...]
+ *
+ * Whenever the value of the Dynamic_State_Base_Addr,
+ * Surface_State_Base_Addr are altered, the L1 state cache must be
+ * invalidated to ensure the new surface or sampler state is fetched
+ * from system memory.
+ *
+ * The PIPE_CONTROL command has a "State Cache Invalidation Enable" bit
+ * which, according the PIPE_CONTROL instruction documentation in the
+ * Broadwell PRM:
+ *
+ * Setting this bit is independent of any other bit in this packet.
+ * This bit controls the invalidation of the L1 and L2 state caches
+ * at the top of the pipe i.e. at the parsing time.
+ *
+ * Unfortunately, experimentation seems to indicate that state cache
+ * invalidation through a PIPE_CONTROL does nothing whatsoever in
+ * regards to surface state and binding tables. In stead, it seems that
+ * invalidating the texture cache is what is actually needed.
+ *
+ * XXX: As far as we have been able to determine through
+ * experimentation, shows that flush the texture cache appears to be
+ * sufficient. The theory here is that all of the sampling/rendering
+ * units cache the binding table in the texture cache. However, we have
+ * yet to be able to actually confirm this.
+ */
+ anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL),
+ .TextureCacheInvalidationEnable = true);
+}
+
+void genX(CmdPipelineBarrier)(
+ VkCommandBuffer commandBuffer,
+ VkPipelineStageFlags srcStageMask,
+ VkPipelineStageFlags destStageMask,
+ VkBool32 byRegion,
+ uint32_t memoryBarrierCount,
+ const VkMemoryBarrier* pMemoryBarriers,
+ uint32_t bufferMemoryBarrierCount,
+ const VkBufferMemoryBarrier* pBufferMemoryBarriers,
+ uint32_t imageMemoryBarrierCount,
+ const VkImageMemoryBarrier* pImageMemoryBarriers)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ uint32_t b, *dw;
+
+ struct GENX(PIPE_CONTROL) cmd = {
+ GENX(PIPE_CONTROL_header),
+ .PostSyncOperation = NoWrite,
+ };
+
+ /* XXX: I think waitEvent is a no-op on our HW. We should verify that. */
+
+ if (anv_clear_mask(&srcStageMask, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT)) {
+ /* This is just what PIPE_CONTROL does */
+ }
+
+ if (anv_clear_mask(&srcStageMask,
+ VK_PIPELINE_STAGE_DRAW_INDIRECT_BIT |
+ VK_PIPELINE_STAGE_VERTEX_INPUT_BIT |
+ VK_PIPELINE_STAGE_VERTEX_SHADER_BIT |
+ VK_PIPELINE_STAGE_TESSELLATION_CONTROL_SHADER_BIT |
+ VK_PIPELINE_STAGE_TESSELLATION_EVALUATION_SHADER_BIT |
+ VK_PIPELINE_STAGE_GEOMETRY_SHADER_BIT |
+ VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT |
+ VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT |
+ VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT |
+ VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT)) {
+ cmd.StallAtPixelScoreboard = true;
+ }
+
+ if (anv_clear_mask(&srcStageMask,
+ VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT |
+ VK_PIPELINE_STAGE_TRANSFER_BIT)) {
+ cmd.CommandStreamerStallEnable = true;
+ }
+
+ if (anv_clear_mask(&srcStageMask, VK_PIPELINE_STAGE_HOST_BIT)) {
+ anv_finishme("VK_PIPE_EVENT_CPU_SIGNAL_BIT");
+ }
+
+ /* On our hardware, all stages will wait for execution as needed. */
+ (void)destStageMask;
+
+ /* We checked all known VkPipeEventFlags. */
+ anv_assert(srcStageMask == 0);
+
+ /* XXX: Right now, we're really dumb and just flush whatever categories
+ * the app asks for. One of these days we may make this a bit better
+ * but right now that's all the hardware allows for in most areas.
+ */
+ VkAccessFlags src_flags = 0;
+ VkAccessFlags dst_flags = 0;
+
+ for (uint32_t i = 0; i < memoryBarrierCount; i++) {
+ src_flags |= pMemoryBarriers[i].srcAccessMask;
+ dst_flags |= pMemoryBarriers[i].dstAccessMask;
+ }
+
+ for (uint32_t i = 0; i < bufferMemoryBarrierCount; i++) {
+ src_flags |= pBufferMemoryBarriers[i].srcAccessMask;
+ dst_flags |= pBufferMemoryBarriers[i].dstAccessMask;
+ }
+
+ for (uint32_t i = 0; i < imageMemoryBarrierCount; i++) {
+ src_flags |= pImageMemoryBarriers[i].srcAccessMask;
+ dst_flags |= pImageMemoryBarriers[i].dstAccessMask;
+ }
+
+ /* The src flags represent how things were used previously. This is
+ * what we use for doing flushes.
+ */
+ for_each_bit(b, src_flags) {
+ switch ((VkAccessFlagBits)(1 << b)) {
+ case VK_ACCESS_SHADER_WRITE_BIT:
+ cmd.DCFlushEnable = true;
+ break;
+ case VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT:
+ cmd.RenderTargetCacheFlushEnable = true;
+ break;
+ case VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT:
+ cmd.DepthCacheFlushEnable = true;
+ break;
+ case VK_ACCESS_TRANSFER_WRITE_BIT:
+ cmd.RenderTargetCacheFlushEnable = true;
+ cmd.DepthCacheFlushEnable = true;
+ break;
+ default:
+ /* Doesn't require a flush */
+ break;
+ }
+ }
+
+ /* The dst flags represent how things will be used in the fugure. This
+ * is what we use for doing cache invalidations.
+ */
+ for_each_bit(b, dst_flags) {
+ switch ((VkAccessFlagBits)(1 << b)) {
+ case VK_ACCESS_INDIRECT_COMMAND_READ_BIT:
+ case VK_ACCESS_INDEX_READ_BIT:
+ case VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT:
+ cmd.VFCacheInvalidationEnable = true;
+ break;
+ case VK_ACCESS_UNIFORM_READ_BIT:
+ cmd.ConstantCacheInvalidationEnable = true;
+ /* fallthrough */
+ case VK_ACCESS_SHADER_READ_BIT:
+ cmd.TextureCacheInvalidationEnable = true;
+ break;
+ case VK_ACCESS_COLOR_ATTACHMENT_READ_BIT:
+ cmd.TextureCacheInvalidationEnable = true;
+ break;
+ case VK_ACCESS_TRANSFER_READ_BIT:
+ cmd.TextureCacheInvalidationEnable = true;
+ break;
+ case VK_ACCESS_MEMORY_READ_BIT:
+ break; /* XXX: What is this? */
+ default:
+ /* Doesn't require a flush */
+ break;
+ }
+ }
+
+ dw = anv_batch_emit_dwords(&cmd_buffer->batch, GENX(PIPE_CONTROL_length));
+ GENX(PIPE_CONTROL_pack)(&cmd_buffer->batch, dw, &cmd);
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+static uint32_t
+vertex_element_comp_control(enum isl_format format, unsigned comp)
+{
+ uint8_t bits;
+ switch (comp) {
+ case 0: bits = isl_format_layouts[format].channels.r.bits; break;
+ case 1: bits = isl_format_layouts[format].channels.g.bits; break;
+ case 2: bits = isl_format_layouts[format].channels.b.bits; break;
+ case 3: bits = isl_format_layouts[format].channels.a.bits; break;
+ default: unreachable("Invalid component");
+ }
+
+ if (bits) {
+ return VFCOMP_STORE_SRC;
+ } else if (comp < 3) {
+ return VFCOMP_STORE_0;
+ } else if (isl_format_layouts[format].channels.r.type == ISL_UINT ||
+ isl_format_layouts[format].channels.r.type == ISL_SINT) {
+ assert(comp == 3);
+ return VFCOMP_STORE_1_INT;
+ } else {
+ assert(comp == 3);
+ return VFCOMP_STORE_1_FP;
+ }
+}
+
+static const uint32_t vk_to_gen_cullmode[] = {
+ [VK_CULL_MODE_NONE] = CULLMODE_NONE,
+ [VK_CULL_MODE_FRONT_BIT] = CULLMODE_FRONT,
+ [VK_CULL_MODE_BACK_BIT] = CULLMODE_BACK,
+ [VK_CULL_MODE_FRONT_AND_BACK] = CULLMODE_BOTH
+};
+
+static const uint32_t vk_to_gen_fillmode[] = {
+ [VK_POLYGON_MODE_FILL] = RASTER_SOLID,
+ [VK_POLYGON_MODE_LINE] = RASTER_WIREFRAME,
+ [VK_POLYGON_MODE_POINT] = RASTER_POINT,
+};
+
+static const uint32_t vk_to_gen_front_face[] = {
+ [VK_FRONT_FACE_COUNTER_CLOCKWISE] = 1,
+ [VK_FRONT_FACE_CLOCKWISE] = 0
+};
+
+static const uint32_t vk_to_gen_logic_op[] = {
+ [VK_LOGIC_OP_COPY] = LOGICOP_COPY,
+ [VK_LOGIC_OP_CLEAR] = LOGICOP_CLEAR,
+ [VK_LOGIC_OP_AND] = LOGICOP_AND,
+ [VK_LOGIC_OP_AND_REVERSE] = LOGICOP_AND_REVERSE,
+ [VK_LOGIC_OP_AND_INVERTED] = LOGICOP_AND_INVERTED,
+ [VK_LOGIC_OP_NO_OP] = LOGICOP_NOOP,
+ [VK_LOGIC_OP_XOR] = LOGICOP_XOR,
+ [VK_LOGIC_OP_OR] = LOGICOP_OR,
+ [VK_LOGIC_OP_NOR] = LOGICOP_NOR,
+ [VK_LOGIC_OP_EQUIVALENT] = LOGICOP_EQUIV,
+ [VK_LOGIC_OP_INVERT] = LOGICOP_INVERT,
+ [VK_LOGIC_OP_OR_REVERSE] = LOGICOP_OR_REVERSE,
+ [VK_LOGIC_OP_COPY_INVERTED] = LOGICOP_COPY_INVERTED,
+ [VK_LOGIC_OP_OR_INVERTED] = LOGICOP_OR_INVERTED,
+ [VK_LOGIC_OP_NAND] = LOGICOP_NAND,
+ [VK_LOGIC_OP_SET] = LOGICOP_SET,
+};
+
+static const uint32_t vk_to_gen_blend[] = {
+ [VK_BLEND_FACTOR_ZERO] = BLENDFACTOR_ZERO,
+ [VK_BLEND_FACTOR_ONE] = BLENDFACTOR_ONE,
+ [VK_BLEND_FACTOR_SRC_COLOR] = BLENDFACTOR_SRC_COLOR,
+ [VK_BLEND_FACTOR_ONE_MINUS_SRC_COLOR] = BLENDFACTOR_INV_SRC_COLOR,
+ [VK_BLEND_FACTOR_DST_COLOR] = BLENDFACTOR_DST_COLOR,
+ [VK_BLEND_FACTOR_ONE_MINUS_DST_COLOR] = BLENDFACTOR_INV_DST_COLOR,
+ [VK_BLEND_FACTOR_SRC_ALPHA] = BLENDFACTOR_SRC_ALPHA,
+ [VK_BLEND_FACTOR_ONE_MINUS_SRC_ALPHA] = BLENDFACTOR_INV_SRC_ALPHA,
+ [VK_BLEND_FACTOR_DST_ALPHA] = BLENDFACTOR_DST_ALPHA,
+ [VK_BLEND_FACTOR_ONE_MINUS_DST_ALPHA] = BLENDFACTOR_INV_DST_ALPHA,
+ [VK_BLEND_FACTOR_CONSTANT_COLOR] = BLENDFACTOR_CONST_COLOR,
+ [VK_BLEND_FACTOR_ONE_MINUS_CONSTANT_COLOR]= BLENDFACTOR_INV_CONST_COLOR,
+ [VK_BLEND_FACTOR_CONSTANT_ALPHA] = BLENDFACTOR_CONST_ALPHA,
+ [VK_BLEND_FACTOR_ONE_MINUS_CONSTANT_ALPHA]= BLENDFACTOR_INV_CONST_ALPHA,
+ [VK_BLEND_FACTOR_SRC_ALPHA_SATURATE] = BLENDFACTOR_SRC_ALPHA_SATURATE,
+ [VK_BLEND_FACTOR_SRC1_COLOR] = BLENDFACTOR_SRC1_COLOR,
+ [VK_BLEND_FACTOR_ONE_MINUS_SRC1_COLOR] = BLENDFACTOR_INV_SRC1_COLOR,
+ [VK_BLEND_FACTOR_SRC1_ALPHA] = BLENDFACTOR_SRC1_ALPHA,
+ [VK_BLEND_FACTOR_ONE_MINUS_SRC1_ALPHA] = BLENDFACTOR_INV_SRC1_ALPHA,
+};
+
+static const uint32_t vk_to_gen_blend_op[] = {
+ [VK_BLEND_OP_ADD] = BLENDFUNCTION_ADD,
+ [VK_BLEND_OP_SUBTRACT] = BLENDFUNCTION_SUBTRACT,
+ [VK_BLEND_OP_REVERSE_SUBTRACT] = BLENDFUNCTION_REVERSE_SUBTRACT,
+ [VK_BLEND_OP_MIN] = BLENDFUNCTION_MIN,
+ [VK_BLEND_OP_MAX] = BLENDFUNCTION_MAX,
+};
+
+static const uint32_t vk_to_gen_compare_op[] = {
+ [VK_COMPARE_OP_NEVER] = PREFILTEROPNEVER,
+ [VK_COMPARE_OP_LESS] = PREFILTEROPLESS,
+ [VK_COMPARE_OP_EQUAL] = PREFILTEROPEQUAL,
+ [VK_COMPARE_OP_LESS_OR_EQUAL] = PREFILTEROPLEQUAL,
+ [VK_COMPARE_OP_GREATER] = PREFILTEROPGREATER,
+ [VK_COMPARE_OP_NOT_EQUAL] = PREFILTEROPNOTEQUAL,
+ [VK_COMPARE_OP_GREATER_OR_EQUAL] = PREFILTEROPGEQUAL,
+ [VK_COMPARE_OP_ALWAYS] = PREFILTEROPALWAYS,
+};
+
+static const uint32_t vk_to_gen_stencil_op[] = {
+ [VK_STENCIL_OP_KEEP] = STENCILOP_KEEP,
+ [VK_STENCIL_OP_ZERO] = STENCILOP_ZERO,
+ [VK_STENCIL_OP_REPLACE] = STENCILOP_REPLACE,
+ [VK_STENCIL_OP_INCREMENT_AND_CLAMP] = STENCILOP_INCRSAT,
+ [VK_STENCIL_OP_DECREMENT_AND_CLAMP] = STENCILOP_DECRSAT,
+ [VK_STENCIL_OP_INVERT] = STENCILOP_INVERT,
+ [VK_STENCIL_OP_INCREMENT_AND_WRAP] = STENCILOP_INCR,
+ [VK_STENCIL_OP_DECREMENT_AND_WRAP] = STENCILOP_DECR,
+};
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+static const uint8_t
+anv_surftype(const struct anv_image *image, VkImageViewType view_type,
+ bool storage)
+{
+ switch (view_type) {
+ default:
+ unreachable("bad VkImageViewType");
+ case VK_IMAGE_VIEW_TYPE_1D:
+ case VK_IMAGE_VIEW_TYPE_1D_ARRAY:
+ assert(image->type == VK_IMAGE_TYPE_1D);
+ return SURFTYPE_1D;
+ case VK_IMAGE_VIEW_TYPE_CUBE:
+ case VK_IMAGE_VIEW_TYPE_CUBE_ARRAY:
+ assert(image->type == VK_IMAGE_TYPE_2D);
+ return storage ? SURFTYPE_2D : SURFTYPE_CUBE;
+ case VK_IMAGE_VIEW_TYPE_2D:
+ case VK_IMAGE_VIEW_TYPE_2D_ARRAY:
+ assert(image->type == VK_IMAGE_TYPE_2D);
+ return SURFTYPE_2D;
+ case VK_IMAGE_VIEW_TYPE_3D:
+ assert(image->type == VK_IMAGE_TYPE_3D);
+ return SURFTYPE_3D;
+ }
+}
+
+#if ANV_GEN > 7 || ANV_IS_HASWELL
+static const uint32_t vk_to_gen_swizzle_map[] = {
+ [VK_COMPONENT_SWIZZLE_ZERO] = SCS_ZERO,
+ [VK_COMPONENT_SWIZZLE_ONE] = SCS_ONE,
+ [VK_COMPONENT_SWIZZLE_R] = SCS_RED,
+ [VK_COMPONENT_SWIZZLE_G] = SCS_GREEN,
+ [VK_COMPONENT_SWIZZLE_B] = SCS_BLUE,
+ [VK_COMPONENT_SWIZZLE_A] = SCS_ALPHA
+};
+
+static inline uint32_t
+vk_to_gen_swizzle(VkComponentSwizzle swizzle, VkComponentSwizzle component)
+{
+ if (swizzle == VK_COMPONENT_SWIZZLE_IDENTITY)
+ return vk_to_gen_swizzle_map[component];
+ else
+ return vk_to_gen_swizzle_map[swizzle];
+}
+#endif
+
+static inline uint32_t
+vk_to_gen_tex_filter(VkFilter filter, bool anisotropyEnable)
+{
+ switch (filter) {
+ default:
+ assert(!"Invalid filter");
+ case VK_FILTER_NEAREST:
+ return MAPFILTER_NEAREST;
+ case VK_FILTER_LINEAR:
+ return anisotropyEnable ? MAPFILTER_ANISOTROPIC : MAPFILTER_LINEAR;
+ }
+}
+
+static inline uint32_t
+vk_to_gen_max_anisotropy(float ratio)
+{
+ return (anv_clamp_f(ratio, 2, 16) - 2) / 2;
+}
+
+static const uint32_t vk_to_gen_mipmap_mode[] = {
+ [VK_SAMPLER_MIPMAP_MODE_NEAREST] = MIPFILTER_NEAREST,
+ [VK_SAMPLER_MIPMAP_MODE_LINEAR] = MIPFILTER_LINEAR
+};
+
+static const uint32_t vk_to_gen_tex_address[] = {
+ [VK_SAMPLER_ADDRESS_MODE_REPEAT] = TCM_WRAP,
+ [VK_SAMPLER_ADDRESS_MODE_MIRRORED_REPEAT] = TCM_MIRROR,
+ [VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE] = TCM_CLAMP,
+ [VK_SAMPLER_ADDRESS_MODE_MIRROR_CLAMP_TO_EDGE] = TCM_MIRROR_ONCE,
+ [VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER] = TCM_CLAMP_BORDER,
+};
+
+static const uint32_t vk_to_gen_compare_op[] = {
+ [VK_COMPARE_OP_NEVER] = PREFILTEROPNEVER,
+ [VK_COMPARE_OP_LESS] = PREFILTEROPLESS,
+ [VK_COMPARE_OP_EQUAL] = PREFILTEROPEQUAL,
+ [VK_COMPARE_OP_LESS_OR_EQUAL] = PREFILTEROPLEQUAL,
+ [VK_COMPARE_OP_GREATER] = PREFILTEROPGREATER,
+ [VK_COMPARE_OP_NOT_EQUAL] = PREFILTEROPNOTEQUAL,
+ [VK_COMPARE_OP_GREATER_OR_EQUAL] = PREFILTEROPGEQUAL,
+ [VK_COMPARE_OP_ALWAYS] = PREFILTEROPALWAYS,
+};
--- /dev/null
+block_pool
+block_pool_no_free
+state_pool
+state_pool_free_list_only
+state_pool_no_free
--- /dev/null
+# Copyright © 2009 Intel Corporation
+#
+# Permission is hereby granted, free of charge, to any person obtaining a
+# copy of this software and associated documentation files (the "Software"),
+# to deal in the Software without restriction, including without limitation
+# on the rights to use, copy, modify, merge, publish, distribute, sub
+# license, and/or sell copies of the Software, and to permit persons to whom
+# the Software is furnished to do so, subject to the following conditions:
+#
+# The above copyright notice and this permission notice (including the next
+# paragraph) shall be included in all copies or substantial portions of the
+# Software.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+# FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+# ADAM JACKSON BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER
+# IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+# CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+AM_CPPFLAGS = \
+ $(INTEL_CFLAGS) \
+ $(VALGRIND_CFLAGS) \
+ $(DEFINES) \
+ -I$(top_srcdir)/include \
+ -I$(top_srcdir)/src \
+ -I$(top_srcdir)/src/mapi \
+ -I$(top_srcdir)/src/mesa \
+ -I$(top_srcdir)/src/mesa/drivers/dri/common \
+ -I$(top_srcdir)/src/mesa/drivers/dri/i965 \
+ -I$(top_srcdir)/src/gallium/auxiliary \
+ -I$(top_srcdir)/src/gallium/include \
+ -I$(top_srcdir)/src/isl/ \
+ -I$(top_srcdir)/src/vulkan
+
+LDADD = \
+ $(top_builddir)/src/vulkan/libvulkan-test.la \
+ $(PTHREAD_LIBS) -lm -lstdc++
+
+check_PROGRAMS = \
+ block_pool_no_free \
+ state_pool_no_free \
+ state_pool_free_list_only \
+ state_pool
+
+TESTS = $(check_PROGRAMS)
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <pthread.h>
+
+#include "anv_private.h"
+
+#define NUM_THREADS 16
+#define BLOCKS_PER_THREAD 1024
+#define NUM_RUNS 64
+
+struct job {
+ pthread_t thread;
+ unsigned id;
+ struct anv_block_pool *pool;
+ uint32_t blocks[BLOCKS_PER_THREAD];
+ uint32_t back_blocks[BLOCKS_PER_THREAD];
+} jobs[NUM_THREADS];
+
+
+static void *alloc_blocks(void *_job)
+{
+ struct job *job = _job;
+ int32_t block, *data;
+
+ for (unsigned i = 0; i < BLOCKS_PER_THREAD; i++) {
+ block = anv_block_pool_alloc(job->pool);
+ data = job->pool->map + block;
+ *data = block;
+ assert(block >= 0);
+ job->blocks[i] = block;
+
+ block = anv_block_pool_alloc_back(job->pool);
+ data = job->pool->map + block;
+ *data = block;
+ assert(block < 0);
+ job->back_blocks[i] = -block;
+ }
+
+ for (unsigned i = 0; i < BLOCKS_PER_THREAD; i++) {
+ block = job->blocks[i];
+ data = job->pool->map + block;
+ assert(*data == block);
+
+ block = -job->back_blocks[i];
+ data = job->pool->map + block;
+ assert(*data == block);
+ }
+
+ return NULL;
+}
+
+static void validate_monotonic(uint32_t **blocks)
+{
+ /* A list of indices, one per thread */
+ unsigned next[NUM_THREADS];
+ memset(next, 0, sizeof(next));
+
+ int highest = -1;
+ while (true) {
+ /* First, we find which thread has the highest next element */
+ int thread_max = -1;
+ int max_thread_idx = -1;
+ for (unsigned i = 0; i < NUM_THREADS; i++) {
+ if (next[i] >= BLOCKS_PER_THREAD)
+ continue;
+
+ if (thread_max < blocks[i][next[i]]) {
+ thread_max = blocks[i][next[i]];
+ max_thread_idx = i;
+ }
+ }
+
+ /* The only way this can happen is if all of the next[] values are at
+ * BLOCKS_PER_THREAD, in which case, we're done.
+ */
+ if (thread_max == -1)
+ break;
+
+ /* That next element had better be higher than the previous highest */
+ assert(blocks[max_thread_idx][next[max_thread_idx]] > highest);
+
+ highest = blocks[max_thread_idx][next[max_thread_idx]];
+ next[max_thread_idx]++;
+ }
+}
+
+static void run_test()
+{
+ struct anv_device device;
+ struct anv_block_pool pool;
+
+ pthread_mutex_init(&device.mutex, NULL);
+ anv_block_pool_init(&pool, &device, 16);
+
+ for (unsigned i = 0; i < NUM_THREADS; i++) {
+ jobs[i].pool = &pool;
+ jobs[i].id = i;
+ pthread_create(&jobs[i].thread, NULL, alloc_blocks, &jobs[i]);
+ }
+
+ for (unsigned i = 0; i < NUM_THREADS; i++)
+ pthread_join(jobs[i].thread, NULL);
+
+ /* Validate that the block allocations were monotonic */
+ uint32_t *block_ptrs[NUM_THREADS];
+ for (unsigned i = 0; i < NUM_THREADS; i++)
+ block_ptrs[i] = jobs[i].blocks;
+ validate_monotonic(block_ptrs);
+
+ /* Validate that the back block allocations were monotonic */
+ for (unsigned i = 0; i < NUM_THREADS; i++)
+ block_ptrs[i] = jobs[i].back_blocks;
+ validate_monotonic(block_ptrs);
+
+ anv_block_pool_finish(&pool);
+ pthread_mutex_destroy(&device.mutex);
+}
+
+int main(int argc, char **argv)
+{
+ for (unsigned i = 0; i < NUM_RUNS; i++)
+ run_test();
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <pthread.h>
+
+#include "anv_private.h"
+
+#define NUM_THREADS 8
+#define STATES_PER_THREAD_LOG2 10
+#define STATES_PER_THREAD (1 << STATES_PER_THREAD_LOG2)
+#define NUM_RUNS 64
+
+#include "state_pool_test_helper.h"
+
+int main(int argc, char **argv)
+{
+ struct anv_device device;
+ struct anv_block_pool block_pool;
+ struct anv_state_pool state_pool;
+
+ pthread_mutex_init(&device.mutex, NULL);
+
+ for (unsigned i = 0; i < NUM_RUNS; i++) {
+ anv_block_pool_init(&block_pool, &device, 256);
+ anv_state_pool_init(&state_pool, &block_pool);
+
+ /* Grab one so a zero offset is impossible */
+ anv_state_pool_alloc(&state_pool, 16, 16);
+
+ run_state_pool_test(&state_pool);
+
+ anv_state_pool_finish(&state_pool);
+ anv_block_pool_finish(&block_pool);
+ }
+
+ pthread_mutex_destroy(&device.mutex);
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <pthread.h>
+
+#include "anv_private.h"
+
+#define NUM_THREADS 8
+#define STATES_PER_THREAD_LOG2 12
+#define STATES_PER_THREAD (1 << STATES_PER_THREAD_LOG2)
+
+#include "state_pool_test_helper.h"
+
+int main(int argc, char **argv)
+{
+ struct anv_device device;
+ struct anv_block_pool block_pool;
+ struct anv_state_pool state_pool;
+
+ pthread_mutex_init(&device.mutex, NULL);
+ anv_block_pool_init(&block_pool, &device, 4096);
+ anv_state_pool_init(&state_pool, &block_pool);
+
+ /* Grab one so a zero offset is impossible */
+ anv_state_pool_alloc(&state_pool, 16, 16);
+
+ /* Grab and return enough states that the state pool test below won't
+ * actually ever resize anything.
+ */
+ {
+ struct anv_state states[NUM_THREADS * STATES_PER_THREAD];
+ for (unsigned i = 0; i < NUM_THREADS * STATES_PER_THREAD; i++) {
+ states[i] = anv_state_pool_alloc(&state_pool, 16, 16);
+ assert(states[i].offset != 0);
+ }
+
+ for (unsigned i = 0; i < NUM_THREADS * STATES_PER_THREAD; i++)
+ anv_state_pool_free(&state_pool, states[i]);
+ }
+
+ run_state_pool_test(&state_pool);
+
+ anv_state_pool_finish(&state_pool);
+ anv_block_pool_finish(&block_pool);
+ pthread_mutex_destroy(&device.mutex);
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <pthread.h>
+
+#include "anv_private.h"
+
+#define NUM_THREADS 16
+#define STATES_PER_THREAD 1024
+#define NUM_RUNS 64
+
+struct job {
+ pthread_t thread;
+ unsigned id;
+ struct anv_state_pool *pool;
+ uint32_t offsets[STATES_PER_THREAD];
+} jobs[NUM_THREADS];
+
+pthread_barrier_t barrier;
+
+static void *alloc_states(void *_job)
+{
+ struct job *job = _job;
+
+ pthread_barrier_wait(&barrier);
+
+ for (unsigned i = 0; i < STATES_PER_THREAD; i++) {
+ struct anv_state state = anv_state_pool_alloc(job->pool, 16, 16);
+ job->offsets[i] = state.offset;
+ }
+
+ return NULL;
+}
+
+static void run_test()
+{
+ struct anv_device device;
+ struct anv_block_pool block_pool;
+ struct anv_state_pool state_pool;
+
+ pthread_mutex_init(&device.mutex, NULL);
+ anv_block_pool_init(&block_pool, &device, 64);
+ anv_state_pool_init(&state_pool, &block_pool);
+
+ pthread_barrier_init(&barrier, NULL, NUM_THREADS);
+
+ for (unsigned i = 0; i < NUM_THREADS; i++) {
+ jobs[i].pool = &state_pool;
+ jobs[i].id = i;
+ pthread_create(&jobs[i].thread, NULL, alloc_states, &jobs[i]);
+ }
+
+ for (unsigned i = 0; i < NUM_THREADS; i++)
+ pthread_join(jobs[i].thread, NULL);
+
+ /* A list of indices, one per thread */
+ unsigned next[NUM_THREADS];
+ memset(next, 0, sizeof(next));
+
+ int highest = -1;
+ while (true) {
+ /* First, we find which thread has the highest next element */
+ int thread_max = -1;
+ int max_thread_idx = -1;
+ for (unsigned i = 0; i < NUM_THREADS; i++) {
+ if (next[i] >= STATES_PER_THREAD)
+ continue;
+
+ if (thread_max < jobs[i].offsets[next[i]]) {
+ thread_max = jobs[i].offsets[next[i]];
+ max_thread_idx = i;
+ }
+ }
+
+ /* The only way this can happen is if all of the next[] values are at
+ * BLOCKS_PER_THREAD, in which case, we're done.
+ */
+ if (thread_max == -1)
+ break;
+
+ /* That next element had better be higher than the previous highest */
+ assert(jobs[max_thread_idx].offsets[next[max_thread_idx]] > highest);
+
+ highest = jobs[max_thread_idx].offsets[next[max_thread_idx]];
+ next[max_thread_idx]++;
+ }
+
+ anv_state_pool_finish(&state_pool);
+ anv_block_pool_finish(&block_pool);
+ pthread_mutex_destroy(&device.mutex);
+}
+
+int main(int argc, char **argv)
+{
+ for (unsigned i = 0; i < NUM_RUNS; i++)
+ run_test();
+}
--- /dev/null
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <pthread.h>
+
+struct job {
+ struct anv_state_pool *pool;
+ unsigned id;
+ pthread_t thread;
+} jobs[NUM_THREADS];
+
+pthread_barrier_t barrier;
+
+static void *alloc_states(void *void_job)
+{
+ struct job *job = void_job;
+
+ const unsigned chunk_size = 1 << (job->id % STATES_PER_THREAD_LOG2);
+ const unsigned num_chunks = STATES_PER_THREAD / chunk_size;
+
+ struct anv_state states[chunk_size];
+
+ pthread_barrier_wait(&barrier);
+
+ for (unsigned c = 0; c < num_chunks; c++) {
+ for (unsigned i = 0; i < chunk_size; i++) {
+ states[i] = anv_state_pool_alloc(job->pool, 16, 16);
+ memset(states[i].map, 139, 16);
+ assert(states[i].offset != 0);
+ }
+
+ for (unsigned i = 0; i < chunk_size; i++)
+ anv_state_pool_free(job->pool, states[i]);
+ }
+
+ return NULL;
+}
+
+static void run_state_pool_test(struct anv_state_pool *state_pool)
+{
+ pthread_barrier_init(&barrier, NULL, NUM_THREADS);
+
+ for (unsigned i = 0; i < NUM_THREADS; i++) {
+ jobs[i].pool = state_pool;
+ jobs[i].id = i;
+ pthread_create(&jobs[i].thread, NULL, alloc_states, &jobs[i]);
+ }
+
+ for (unsigned i = 0; i < NUM_THREADS; i++)
+ pthread_join(jobs[i].thread, NULL);
+}