OSDN Git Service
José Fonseca [Thu, 22 Oct 2009 18:02:04 +0000 (19:02 +0100)]
llvmpipe: Avoid variable size arrays.
Not really variable size, but MSVC still doesn't like them.
José Fonseca [Thu, 22 Oct 2009 17:28:57 +0000 (18:28 +0100)]
llvmpipe: Use the pack/unpack functions for 8bit unsigned norm multiplication.
José Fonseca [Thu, 22 Oct 2009 17:28:37 +0000 (18:28 +0100)]
llvmpipe: Utility function to double the bit width of a type.
José Fonseca [Thu, 22 Oct 2009 17:28:17 +0000 (18:28 +0100)]
llvmpipe: Factor vector packing/unpacking to a separate source file.
These functions will be needed to implement many of the 8bit operations,
and they are quite complex on its own.
José Fonseca [Thu, 22 Oct 2009 17:10:19 +0000 (18:10 +0100)]
llvmpipe: Call util_cpu_detect() from the unit tests.
José Fonseca [Thu, 22 Oct 2009 17:09:37 +0000 (18:09 +0100)]
util: Set cpu endianness too.
José Fonseca [Wed, 21 Oct 2009 16:03:52 +0000 (17:03 +0100)]
scons: Disable more MSVC pedantic security warnings.
Alex Deucher [Thu, 22 Oct 2009 16:29:36 +0000 (12:29 -0400)]
Merge branch 'master' of git+ssh://agd5f@git.freedesktop.org/git/mesa/mesa
Zack Rusin [Thu, 22 Oct 2009 16:13:02 +0000 (12:13 -0400)]
st/xorg: cleanup the checks for whether the op is accelerated
Alex Deucher [Thu, 22 Oct 2009 16:02:31 +0000 (12:02 -0400)]
Merge branch 'mesa_7_6_branch' of git+ssh://agd5f@git.freedesktop.org/git/mesa/mesa
Zack Rusin [Thu, 22 Oct 2009 15:49:19 +0000 (11:49 -0400)]
st/xorg: lots of render fixes
fixes all the blend modes, fixes flushing/finishing semantics, adds
acceleration for the component alpha modes that we can support, fixes
src in mask shader and general cleanups
Chia-I Wu [Sun, 18 Oct 2009 05:15:30 +0000 (13:15 +0800)]
progs/egl: Add support for pixmap and pbuffer surface to xeglgears.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Chia-I Wu [Fri, 2 Oct 2009 02:38:14 +0000 (10:38 +0800)]
egl_glx: Add support for eglWaitClient and eglWaitNative.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Chia-I Wu [Thu, 1 Oct 2009 10:23:58 +0000 (18:23 +0800)]
egl_glx: Clean up eglGetProcAddress.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Chia-I Wu [Thu, 1 Oct 2009 10:23:41 +0000 (18:23 +0800)]
egl_glx: Clean up context functions.
This lifts the requirement that a context must be direct.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Chia-I Wu [Thu, 1 Oct 2009 10:22:33 +0000 (18:22 +0800)]
egl_glx: Clean up surface functions.
Separete Drawable and GLXDrawable. Add support for pbuffer and pixmap
surfaces on GLX <= 1.3. Remove surface binding code that will never
work.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Chia-I Wu [Thu, 1 Oct 2009 10:16:10 +0000 (18:16 +0800)]
egl_glx: Clean up the initialization code.
Proper detection of GLX extensions. Convert fbconfigs or visuals in a
more unified way and validate the resulting configs.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Chia-I Wu [Fri, 16 Oct 2009 04:18:25 +0000 (12:18 +0800)]
egl: Correct conversion of native visual type.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Ben Skeggs [Thu, 22 Oct 2009 11:55:22 +0000 (21:55 +1000)]
nv50: support 3D class 0x8597, remove redundant unknown chipset detection
Corbin Simpson [Thu, 22 Oct 2009 07:21:08 +0000 (00:21 -0700)]
r300g: Check for NULL Draw during flush.
Split from the fastpath WIP.
Andre Maasikas [Wed, 21 Oct 2009 16:04:21 +0000 (19:04 +0300)]
r600: set barrier for tex inst if dst is used earlier, might overwrite it otherwise
Andre Maasikas [Wed, 21 Oct 2009 09:23:27 +0000 (12:23 +0300)]
r600: need to export something from PS
Also avoids empty shader for "END" - seems to be somewhat valid fp
Maybe this can be done differently in the future (fake FRAG_RESULT_COLOR
already in Map_Fragment_Program() or is there a way to program the chip
to not hang in case of no exports.
Andre Maasikas [Tue, 20 Oct 2009 08:45:39 +0000 (11:45 +0300)]
r600: add beginnings of ARL instruction
Corbin Simpson [Thu, 22 Oct 2009 06:26:02 +0000 (23:26 -0700)]
r300g: Update comments, function names in r300_state_inlines.
Corbin Simpson [Thu, 22 Oct 2009 06:23:37 +0000 (23:23 -0700)]
r300g: Update comments, asserts, indents in r300_texture.
I wish I knew enough about textures to really really REALLY fix that file.
Corbin Simpson [Thu, 22 Oct 2009 05:54:52 +0000 (22:54 -0700)]
r300g: Remove unused debug flag.
Corbin Simpson [Thu, 22 Oct 2009 05:53:37 +0000 (22:53 -0700)]
r300g: Cleanup old static shader state.
Marc Dietrich [Sun, 18 Oct 2009 15:28:33 +0000 (08:28 -0700)]
gallium/util: fix cpu detection on ppc
As we are compiling with -D_BSD_SOURCE, sigjmp_buf and siglongjmp
should be replaced by the non-sig functions (see man 3 setjmp).
Tested on linux/cell.
Brian Paul [Tue, 20 Oct 2009 22:13:08 +0000 (16:13 -0600)]
mesa: added _mesa_dump_renderbuffers() debug code
Patrice Mandin [Wed, 21 Oct 2009 20:01:03 +0000 (22:01 +0200)]
nouveau: nv30: Do not use assert to return NULL
Patrice Mandin [Wed, 21 Oct 2009 19:56:09 +0000 (21:56 +0200)]
nouveau: nv30: check number of colour buffers to bind
José Fonseca [Tue, 20 Oct 2009 09:54:21 +0000 (10:54 +0100)]
llvmpipe: Remove extraneous name in lp_type pre-declaration.
Corbin Simpson [Wed, 21 Oct 2009 13:49:16 +0000 (06:49 -0700)]
r300g: No debug in r300_state.
Corbin Simpson [Wed, 21 Oct 2009 13:31:36 +0000 (06:31 -0700)]
r300g: Cleanup header includes.
Corbin Simpson [Wed, 21 Oct 2009 12:54:49 +0000 (05:54 -0700)]
r300g: Don't use getenv; use debug_get_*_option instead.
Corbin Simpson [Wed, 21 Oct 2009 12:48:45 +0000 (05:48 -0700)]
r300g: Don't use the hashtable internally.
As osiris pointed out, glxgears slowly gets slower for some reason
when it's enabled, and it's not helping at the moment, so just turn
it off.
Corbin Simpson [Wed, 21 Oct 2009 12:45:05 +0000 (05:45 -0700)]
r300g: Examine vertex attribute type on HW TCL too.
Corbin Simpson [Wed, 21 Oct 2009 13:47:05 +0000 (06:47 -0700)]
r300g: Also kill r300_shader_inlines with fire.
Corbin Simpson [Wed, 21 Oct 2009 13:44:16 +0000 (06:44 -0700)]
r300g: Kill r300_surface with fire.
If you really want to see it again, check the history.
Brian Paul [Tue, 20 Oct 2009 22:22:03 +0000 (16:22 -0600)]
progs/trivial: check if GL_ARB_occlusion_query is supported
Eric Anholt [Tue, 20 Oct 2009 21:51:53 +0000 (14:51 -0700)]
intel: Fix flipped condition in ARB_sync GetSYnciv(GL_SYNC_STATUS).
Bug #24435
Eric Anholt [Tue, 20 Oct 2009 00:54:42 +0000 (17:54 -0700)]
meta: Fix the BufferSubData in meta clear to be BufferData.
Fixes a 3.4% +/- 1.3% performance regression in my GL demo (n=3). The
other meta code could probably also use the same treatment.
Marc Dietrich [Sun, 18 Oct 2009 15:28:34 +0000 (08:28 -0700)]
cell: fix compilation on cell
s/LERP/LRP/
Vinson Lee [Tue, 20 Oct 2009 16:43:46 +0000 (10:43 -0600)]
mesa: Fix Mac OS build.
strtod_l needs the xlocale.h header on Mac OS. It's possible other
non-Linux OSes would need this header too.
Jakob Bornecrantz [Tue, 20 Oct 2009 15:52:33 +0000 (17:52 +0200)]
st/xorg: Move drm init to own function
Jakob Bornecrantz [Tue, 20 Oct 2009 15:17:41 +0000 (17:17 +0200)]
trace: Check for destroy before calling it
Jakob Bornecrantz [Tue, 20 Oct 2009 15:18:25 +0000 (17:18 +0200)]
st/xorg: Cleanly shutdown
Mathias Fröhlich [Mon, 19 Oct 2009 15:47:21 +0000 (11:47 -0400)]
r300: fix texture size handling with size > 2048
The in kernel texture check fails because of both
bit11 flags being set on 16x16 textures. It tuns out
that these bits are still set and not cleared in the
pp_txpitch field of the texture. The attached patch
at least helps for this case on my machine. It clears
the bit 11 from the pitch field if the texture is smaller
and masks out that hight bits on the conventional width
and height field.
Fixes bug 24584
Robert Noland [Mon, 19 Oct 2009 14:47:39 +0000 (09:47 -0500)]
r600: Fix size calculation for 24 bit depth
size was being calculated based on 3 bytes per pixel with 24 bit depth
instead of 4 bytes. This caused corruption in the bottom 25% of objects.
This finishes fixing the menu/text corruption in compiz/kde4.
Signed-off-by: Robert Noland <rnoland@2hip.net>
Jakob Bornecrantz [Mon, 19 Oct 2009 05:49:40 +0000 (07:49 +0200)]
st/xorg: Remove output functions not used
Zack Rusin [Tue, 20 Oct 2009 03:48:38 +0000 (23:48 -0400)]
st/xorg: makefile garbage
Zack Rusin [Tue, 13 Oct 2009 17:06:39 +0000 (13:06 -0400)]
st/xorg: implement basic src/mask transformations
plus fix filters
Corbin Simpson [Mon, 19 Oct 2009 20:51:28 +0000 (13:51 -0700)]
r300g: Demonstratory kludge to unbreak glxgears.
We *must* recalculate something in vformat every rebind; let's see
if we can't narrow it down a bit.
Corbin Simpson [Mon, 19 Oct 2009 20:06:45 +0000 (13:06 -0700)]
dri-st: Add EXT_provoking_vertex.
Hm, I could have sworn I did this before?
Jakob Bornecrantz [Sun, 18 Oct 2009 09:49:44 +0000 (11:49 +0200)]
st/xorg: Switch to set_mode_major
Jakob Bornecrantz [Sun, 18 Oct 2009 13:42:25 +0000 (15:42 +0200)]
st/xorg: Massivly redo root pixmap creation
Jakob Bornecrantz [Sun, 18 Oct 2009 09:46:19 +0000 (11:46 +0200)]
st/xorg: Clean up cursor functions a bit
Jakob Bornecrantz [Sun, 18 Oct 2009 13:25:30 +0000 (15:25 +0200)]
st/xorg: Support more then one output of a given type
Jakob Bornecrantz [Sun, 18 Oct 2009 05:50:14 +0000 (07:50 +0200)]
st/xorg: Change how disable accel is handled
Christoph Bumiller [Mon, 19 Oct 2009 16:17:45 +0000 (18:17 +0200)]
nv50: add support for address regs
Allow indirect uniform access and increase the
limit on parameters from 128 to 512.
Christoph Bumiller [Mon, 19 Oct 2009 15:53:31 +0000 (17:53 +0200)]
nv50: cleanup emit_kil
Christoph Bumiller [Mon, 19 Oct 2009 15:47:29 +0000 (17:47 +0200)]
nv50: implement TGSI_OPCODE_CMP
Christoph Bumiller [Tue, 13 Oct 2009 13:09:13 +0000 (15:09 +0200)]
nv50: quick fix for insn src negation
We only have a per nv50_reg negation flag, if an
nv50_reg is used more than once in a TGSI op with
different sign modes, we'd generate wrong code.
We probably can't do much better without more
invasive changes.
Christoph Bumiller [Wed, 14 Oct 2009 19:27:35 +0000 (21:27 +0200)]
nv50: add support for DDX and DDY opcodes
Christoph Bumiller [Wed, 14 Oct 2009 19:23:29 +0000 (21:23 +0200)]
nv50: submit user vbo data through the fifo
Requesting a new real buffer from the kernel and
copying all the data is wasteful e.g. if only a
few (but widely spread) vertices are accessed.
Christoph Bumiller [Sat, 10 Oct 2009 11:13:16 +0000 (13:13 +0200)]
nv50: use SIFC for TIC, TSC upload
Add proper flushes for TIC and TSC and remove
the costly 2D.0110 flush in nv50_flush.
Correct TIC and TSC bo sizes.
Christoph Bumiller [Sat, 10 Oct 2009 11:18:07 +0000 (13:18 +0200)]
nv50: nicer texture format switch
Similar to nv40.
Alex Deucher [Mon, 19 Oct 2009 16:10:59 +0000 (12:10 -0400)]
Merge branch 'mesa_7_6_branch' of git+ssh://agd5f@git.freedesktop.org/git/mesa/mesa
regenerated lex.yy.c
Ben Skeggs [Sun, 18 Oct 2009 23:28:59 +0000 (09:28 +1000)]
nouveau: implement is_{texture,buffer}_referenced properly
Ben Skeggs [Wed, 7 Oct 2009 06:07:34 +0000 (16:07 +1000)]
drm/nv50: write tic/tsc setup to correct slots when skipping units
Corbin Simpson [Sun, 18 Oct 2009 22:54:39 +0000 (15:54 -0700)]
r300g: Fix up a bunch of warnings.
Corbin Simpson [Sun, 18 Oct 2009 17:30:18 +0000 (10:30 -0700)]
r300g: Add another ZTOP condition.
I don't even know if texkill works right now.
José Fonseca [Sun, 18 Oct 2009 13:31:58 +0000 (14:31 +0100)]
llvmpipe: Allocate texture storage for whole quads.
Corbin Simpson [Sun, 18 Oct 2009 04:32:56 +0000 (21:32 -0700)]
gallium: Permit surface_copy and surface_fill to be NULL.
Uf. Lots of files touched. Would people with working vega, xorg, dri1, etc.
please make sure you are not broken, and fix yourself up if you are.
There were only two or three places where the code did not have painful
fallbacks, so I would advise st maintainers to find less painful workarounds,
or consider overhauling util_surface_copy and util_surface_fill.
Per ymanton, darktama, and Dr_Jakob's suggestions, clear has been left as-is.
I will not add PIPE_CAP_BLITTER unless it is deemed necessary.
Corbin Simpson [Sun, 18 Oct 2009 03:53:19 +0000 (20:53 -0700)]
r300g: Squash format warning.
Won't ever be supported.
Corbin Simpson [Sun, 18 Oct 2009 03:47:45 +0000 (20:47 -0700)]
r300g: Use a dirty test to bring framerate back up.
This is just split out from the next commit, that's all.
Corbin Simpson [Sun, 18 Oct 2009 03:29:27 +0000 (20:29 -0700)]
r300g: Also have rs_block keyed to the current shader combo.
Eliminates part of the glxgears corruption here. Need to clean up PSC more,
to get rid of the rest of it.
Corbin Simpson [Sun, 18 Oct 2009 03:05:23 +0000 (20:05 -0700)]
r300g: Fix u_hash_table rename.
Corbin Simpson [Sun, 18 Oct 2009 02:58:23 +0000 (19:58 -0700)]
Merge branch 'master' of ssh://git.freedesktop.org/git/mesa/mesa
Patrice Mandin [Sat, 17 Oct 2009 18:49:18 +0000 (20:49 +0200)]
nouveau: nv40: Use same workaround as i915 for segfault related to vbo
Patrice Mandin [Sat, 17 Oct 2009 18:46:19 +0000 (20:46 +0200)]
nouveau: nv30: Remove duplicate case. Was a typo for X8R8G8B8, but that will never be use for front buffer.
Patrice Mandin [Sat, 17 Oct 2009 18:27:24 +0000 (20:27 +0200)]
nouveau: nv30: Use same workaround as i915 for segfault related to vbo
José Fonseca [Sat, 17 Oct 2009 10:45:04 +0000 (11:45 +0100)]
util: Rename from u_* to util_* while we're at it.
To be consistent with the rest.
Chris Wilson [Sat, 17 Oct 2009 07:25:48 +0000 (08:25 +0100)]
intel: Disallow relocations to the byte beyond the end of the buffer
Brian Paul [Fri, 16 Oct 2009 15:42:30 +0000 (09:42 -0600)]
mesa: fix/update some comments
Brian Paul [Fri, 16 Oct 2009 15:33:11 +0000 (09:33 -0600)]
mesa: lift default symlinks target into Makefile.template
Driver Makefiles can still add symlink dependencies/rules if needed.
Brian Paul [Fri, 16 Oct 2009 15:25:05 +0000 (09:25 -0600)]
mesa: use EXTRA_MODULES and SUBDIRS to build r300 compiler
This is a bit cleaner and avoids rebuilding the r300_dri.so library all
the time.
Brian Paul [Fri, 16 Oct 2009 15:21:45 +0000 (09:21 -0600)]
mesa: added SUBDIRS support in dri/Makefile.template
Brian Paul [Fri, 16 Oct 2009 14:12:47 +0000 (08:12 -0600)]
mesa: move a comma
Brian Paul [Thu, 15 Oct 2009 21:25:52 +0000 (15:25 -0600)]
mesa: added MESA_GLSL=useprog debug flag
This logs glUseProgram() calls to stderr.
Maciej Cencora [Fri, 16 Oct 2009 18:21:17 +0000 (20:21 +0200)]
r300: fix vertex program parameters limits
Corbin Simpson [Fri, 16 Oct 2009 16:45:07 +0000 (09:45 -0700)]
radeon-gallium: Use debug_get_bool_option instead of getenv.
Corbin Simpson [Fri, 16 Oct 2009 15:39:59 +0000 (08:39 -0700)]
r300g: Use a hash table to look up vertex info.
Need to move rs_block to this, too.
Also, I'm getting massive amounts of flicker for some reason; I bet we've gotta
re-re-examine PSC and friends. :C
Corbin Simpson [Fri, 16 Oct 2009 15:27:56 +0000 (08:27 -0700)]
util: Change function names to begin with u_.
Avoids link-time clashes with Mesa's internal hash table.
Brian Paul [Fri, 16 Oct 2009 13:38:39 +0000 (07:38 -0600)]
docs: document GLSL float parsing bug fix
Brian Paul [Wed, 14 Oct 2009 20:19:03 +0000 (14:19 -0600)]
mesa: use C locale for _mesa_strtod()
_mesa_strtod() is used for shader/program parsing where the decimal
point character is always '.' Use strtod_l() with a "C" locale to
ensure correct string->double conversion when the actual locale uses
another character such as ',' for the decimal point.
Fixes bug 24531.
José Fonseca [Fri, 16 Oct 2009 10:39:29 +0000 (11:39 +0100)]
progs/tests: Use rand() instead of random().
Forgot these on previous commit.
José Fonseca [Fri, 16 Oct 2009 10:39:29 +0000 (11:39 +0100)]
progs/tests: Use rand() instead of random().
More portable. Same implementation on Linux.
Vinson Lee [Fri, 16 Oct 2009 10:36:09 +0000 (11:36 +0100)]
prog/tests: Fix MSVC build.
Corbin Simpson [Thu, 15 Oct 2009 23:58:26 +0000 (16:58 -0700)]
r300g: Set logical ID for each emitted texture/sampler.
multitexarray works on my r300, but texrect doesn't.