2 Copyright (C) Intel Corp. 2006. All Rights Reserved.
3 Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
4 develop this 3D driver.
6 Permission is hereby granted, free of charge, to any person obtaining
7 a copy of this software and associated documentation files (the
8 "Software"), to deal in the Software without restriction, including
9 without limitation the rights to use, copy, modify, merge, publish,
10 distribute, sublicense, and/or sell copies of the Software, and to
11 permit persons to whom the Software is furnished to do so, subject to
12 the following conditions:
14 The above copyright notice and this permission notice (including the
15 next paragraph) shall be included in all copies or substantial
16 portions of the Software.
18 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
19 EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
21 IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
22 LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
23 OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
24 WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
26 **********************************************************************/
29 * Keith Whitwell <keith@tungstengraphics.com>
34 #include "brw_context.h"
35 #include "brw_state.h"
36 #include "drivers/common/meta.h"
37 #include "intel_batchbuffer.h"
38 #include "intel_buffers.h"
40 static const struct brw_tracked_state *gen4_atoms[] =
42 &brw_vs_prog, /* must do before GS prog, state base address. */
43 &brw_gs_prog, /* must do before state base address */
44 &brw_clip_prog, /* must do before state base address */
45 &brw_sf_prog, /* must do before state base address */
46 &brw_wm_prog, /* must do before state base address */
48 /* Once all the programs are done, we know how large urb entry
49 * sizes need to be and can decide if we need to change the urb
53 &brw_recalculate_urb_fence,
58 /* Surface state setup. Must come before the VS/WM unit. The binding
59 * table upload must be last.
61 &brw_vs_pull_constants,
62 &brw_wm_pull_constants,
63 &brw_renderbuffer_surfaces,
64 &brw_texture_surfaces,
65 &brw_vs_binding_table,
66 &brw_wm_binding_table,
70 /* These set up state for brw_psp_urb_cbs */
74 &brw_vs_unit, /* always required, enabled or not */
81 &brw_state_base_address,
83 &brw_binding_table_pointers,
84 &brw_blend_constant_color,
89 &brw_polygon_stipple_offset,
92 &brw_aa_line_parameters,
104 static const struct brw_tracked_state *gen6_atoms[] =
106 &brw_vs_prog, /* must do before state base address */
107 &brw_gs_prog, /* must do before state base address */
108 &brw_wm_prog, /* must do before state base address */
113 /* Command packets: */
115 /* must do before binding table pointers, cc state ptrs */
116 &brw_state_base_address,
119 &gen6_viewport_state, /* must do after *_vp stages */
122 &gen6_blend_state, /* must do before cc unit */
123 &gen6_color_calc_state, /* must do before cc unit */
124 &gen6_depth_stencil_state, /* must do before cc unit */
125 &gen6_cc_state_pointers,
127 &gen6_vs_push_constants, /* Before vs_state */
128 &gen6_wm_push_constants, /* Before wm_state */
130 /* Surface state setup. Must come before the VS/WM unit. The binding
131 * table upload must be last.
133 &brw_vs_pull_constants,
134 &brw_vs_ubo_surfaces,
135 &brw_wm_pull_constants,
136 &brw_wm_ubo_surfaces,
137 &gen6_renderbuffer_surfaces,
138 &brw_texture_surfaces,
140 &brw_vs_binding_table,
141 &gen6_gs_binding_table,
142 &brw_wm_binding_table,
146 &gen6_multisample_state,
156 &gen6_binding_table_pointers,
160 &brw_polygon_stipple,
161 &brw_polygon_stipple_offset,
164 &brw_aa_line_parameters,
173 static const struct brw_tracked_state *gen7_atoms[] =
178 /* Command packets: */
180 /* must do before binding table pointers, cc state ptrs */
181 &brw_state_base_address,
184 &gen7_cc_viewport_state_pointer, /* must do after brw_cc_vp */
185 &gen7_sf_clip_viewport,
188 &gen6_blend_state, /* must do before cc unit */
189 &gen6_color_calc_state, /* must do before cc unit */
190 &gen6_depth_stencil_state, /* must do before cc unit */
191 &gen7_depth_stencil_state_pointer,
193 &gen6_vs_push_constants, /* Before vs_state */
194 &gen6_wm_push_constants, /* Before wm_surfaces and constant_buffer */
196 /* Surface state setup. Must come before the VS/WM unit. The binding
197 * table upload must be last.
199 &brw_vs_pull_constants,
200 &brw_vs_ubo_surfaces,
201 &brw_wm_pull_constants,
202 &brw_wm_ubo_surfaces,
203 &gen6_renderbuffer_surfaces,
204 &brw_texture_surfaces,
205 &brw_vs_binding_table,
206 &brw_wm_binding_table,
209 &gen6_multisample_state,
211 &gen7_disable_stages,
224 &brw_polygon_stipple,
225 &brw_polygon_stipple_offset,
228 &brw_aa_line_parameters,
240 brw_upload_initial_gpu_state(struct brw_context *brw)
242 struct intel_context *intel = &brw->intel;
244 /* On platforms with hardware contexts, we can set our initial GPU state
245 * right away rather than doing it via state atoms. This saves a small
246 * amount of overhead on every draw call.
251 brw_upload_invariant_state(brw);
253 if (intel->gen >= 7) {
254 gen7_allocate_push_constants(brw);
258 void brw_init_state( struct brw_context *brw )
260 const struct brw_tracked_state **atoms;
263 brw_init_caches(brw);
265 if (brw->intel.gen >= 7) {
267 num_atoms = ARRAY_SIZE(gen7_atoms);
268 } else if (brw->intel.gen == 6) {
270 num_atoms = ARRAY_SIZE(gen6_atoms);
273 num_atoms = ARRAY_SIZE(gen4_atoms);
277 brw->num_atoms = num_atoms;
279 while (num_atoms--) {
280 assert((*atoms)->dirty.mesa |
281 (*atoms)->dirty.brw |
282 (*atoms)->dirty.cache);
283 assert((*atoms)->emit);
287 brw_upload_initial_gpu_state(brw);
291 void brw_destroy_state( struct brw_context *brw )
293 brw_destroy_caches(brw);
296 /***********************************************************************
300 check_state(const struct brw_state_flags *a, const struct brw_state_flags *b)
302 return ((a->mesa & b->mesa) |
304 (a->cache & b->cache)) != 0;
307 static void accumulate_state( struct brw_state_flags *a,
308 const struct brw_state_flags *b )
312 a->cache |= b->cache;
316 static void xor_states( struct brw_state_flags *result,
317 const struct brw_state_flags *a,
318 const struct brw_state_flags *b )
320 result->mesa = a->mesa ^ b->mesa;
321 result->brw = a->brw ^ b->brw;
322 result->cache = a->cache ^ b->cache;
325 struct dirty_bit_map {
331 #define DEFINE_BIT(name) {name, #name, 0}
333 static struct dirty_bit_map mesa_bits[] = {
334 DEFINE_BIT(_NEW_MODELVIEW),
335 DEFINE_BIT(_NEW_PROJECTION),
336 DEFINE_BIT(_NEW_TEXTURE_MATRIX),
337 DEFINE_BIT(_NEW_COLOR),
338 DEFINE_BIT(_NEW_DEPTH),
339 DEFINE_BIT(_NEW_EVAL),
340 DEFINE_BIT(_NEW_FOG),
341 DEFINE_BIT(_NEW_HINT),
342 DEFINE_BIT(_NEW_LIGHT),
343 DEFINE_BIT(_NEW_LINE),
344 DEFINE_BIT(_NEW_PIXEL),
345 DEFINE_BIT(_NEW_POINT),
346 DEFINE_BIT(_NEW_POLYGON),
347 DEFINE_BIT(_NEW_POLYGONSTIPPLE),
348 DEFINE_BIT(_NEW_SCISSOR),
349 DEFINE_BIT(_NEW_STENCIL),
350 DEFINE_BIT(_NEW_TEXTURE),
351 DEFINE_BIT(_NEW_TRANSFORM),
352 DEFINE_BIT(_NEW_VIEWPORT),
353 DEFINE_BIT(_NEW_ARRAY),
354 DEFINE_BIT(_NEW_RENDERMODE),
355 DEFINE_BIT(_NEW_BUFFERS),
356 DEFINE_BIT(_NEW_MULTISAMPLE),
357 DEFINE_BIT(_NEW_TRACK_MATRIX),
358 DEFINE_BIT(_NEW_PROGRAM),
359 DEFINE_BIT(_NEW_PROGRAM_CONSTANTS),
360 DEFINE_BIT(_NEW_BUFFER_OBJECT),
361 DEFINE_BIT(_NEW_FRAG_CLAMP),
362 DEFINE_BIT(_NEW_VARYING_VP_INPUTS),
366 static struct dirty_bit_map brw_bits[] = {
367 DEFINE_BIT(BRW_NEW_URB_FENCE),
368 DEFINE_BIT(BRW_NEW_FRAGMENT_PROGRAM),
369 DEFINE_BIT(BRW_NEW_VERTEX_PROGRAM),
370 DEFINE_BIT(BRW_NEW_CURBE_OFFSETS),
371 DEFINE_BIT(BRW_NEW_REDUCED_PRIMITIVE),
372 DEFINE_BIT(BRW_NEW_PRIMITIVE),
373 DEFINE_BIT(BRW_NEW_CONTEXT),
374 DEFINE_BIT(BRW_NEW_PSP),
375 DEFINE_BIT(BRW_NEW_SURFACES),
376 DEFINE_BIT(BRW_NEW_VS_BINDING_TABLE),
377 DEFINE_BIT(BRW_NEW_GS_BINDING_TABLE),
378 DEFINE_BIT(BRW_NEW_PS_BINDING_TABLE),
379 DEFINE_BIT(BRW_NEW_INDICES),
380 DEFINE_BIT(BRW_NEW_VERTICES),
381 DEFINE_BIT(BRW_NEW_BATCH),
382 DEFINE_BIT(BRW_NEW_INDEX_BUFFER),
383 DEFINE_BIT(BRW_NEW_VS_CONSTBUF),
384 DEFINE_BIT(BRW_NEW_PROGRAM_CACHE),
385 DEFINE_BIT(BRW_NEW_STATE_BASE_ADDRESS),
386 DEFINE_BIT(BRW_NEW_VUE_MAP_GEOM_OUT),
387 DEFINE_BIT(BRW_NEW_TRANSFORM_FEEDBACK),
388 DEFINE_BIT(BRW_NEW_RASTERIZER_DISCARD),
389 DEFINE_BIT(BRW_NEW_UNIFORM_BUFFER),
390 DEFINE_BIT(BRW_NEW_META_IN_PROGRESS),
394 static struct dirty_bit_map cache_bits[] = {
395 DEFINE_BIT(CACHE_NEW_DEPTH_STENCIL_STATE),
396 DEFINE_BIT(CACHE_NEW_CC_VP),
397 DEFINE_BIT(CACHE_NEW_CC_UNIT),
398 DEFINE_BIT(CACHE_NEW_WM_PROG),
399 DEFINE_BIT(CACHE_NEW_SAMPLER),
400 DEFINE_BIT(CACHE_NEW_WM_UNIT),
401 DEFINE_BIT(CACHE_NEW_SF_PROG),
402 DEFINE_BIT(CACHE_NEW_SF_VP),
403 DEFINE_BIT(CACHE_NEW_SF_UNIT),
404 DEFINE_BIT(CACHE_NEW_VS_UNIT),
405 DEFINE_BIT(CACHE_NEW_VS_PROG),
406 DEFINE_BIT(CACHE_NEW_GS_UNIT),
407 DEFINE_BIT(CACHE_NEW_GS_PROG),
408 DEFINE_BIT(CACHE_NEW_CLIP_VP),
409 DEFINE_BIT(CACHE_NEW_CLIP_UNIT),
410 DEFINE_BIT(CACHE_NEW_CLIP_PROG),
416 brw_update_dirty_count(struct dirty_bit_map *bit_map, int32_t bits)
420 for (i = 0; i < 32; i++) {
421 if (bit_map[i].bit == 0)
424 if (bit_map[i].bit & bits)
430 brw_print_dirty_count(struct dirty_bit_map *bit_map)
434 for (i = 0; i < 32; i++) {
435 if (bit_map[i].bit == 0)
438 fprintf(stderr, "0x%08x: %12d (%s)\n",
439 bit_map[i].bit, bit_map[i].count, bit_map[i].name);
443 /***********************************************************************
446 void brw_upload_state(struct brw_context *brw)
448 struct gl_context *ctx = &brw->intel.ctx;
449 struct intel_context *intel = &brw->intel;
450 struct brw_state_flags *state = &brw->state.dirty;
452 static int dirty_count = 0;
454 state->mesa |= brw->intel.NewGLState;
455 brw->intel.NewGLState = 0;
457 state->brw |= ctx->NewDriverState;
458 ctx->NewDriverState = 0;
460 if (brw->emit_state_always) {
466 if (brw->fragment_program != ctx->FragmentProgram._Current) {
467 brw->fragment_program = ctx->FragmentProgram._Current;
468 brw->state.dirty.brw |= BRW_NEW_FRAGMENT_PROGRAM;
471 if (brw->vertex_program != ctx->VertexProgram._Current) {
472 brw->vertex_program = ctx->VertexProgram._Current;
473 brw->state.dirty.brw |= BRW_NEW_VERTEX_PROGRAM;
476 if (brw->meta_in_progress != _mesa_meta_in_progress(ctx)) {
477 brw->meta_in_progress = _mesa_meta_in_progress(ctx);
478 brw->state.dirty.brw |= BRW_NEW_META_IN_PROGRESS;
481 if ((state->mesa | state->cache | state->brw) == 0)
484 intel_check_front_buffer_rendering(intel);
486 if (unlikely(INTEL_DEBUG)) {
487 /* Debug version which enforces various sanity checks on the
488 * state flags which are generated and checked to help ensure
489 * state atoms are ordered correctly in the list.
491 struct brw_state_flags examined, prev;
492 memset(&examined, 0, sizeof(examined));
495 for (i = 0; i < brw->num_atoms; i++) {
496 const struct brw_tracked_state *atom = brw->atoms[i];
497 struct brw_state_flags generated;
499 if (check_state(state, &atom->dirty)) {
503 accumulate_state(&examined, &atom->dirty);
505 /* generated = (prev ^ state)
506 * if (examined & generated)
509 xor_states(&generated, &prev, state);
510 assert(!check_state(&examined, &generated));
515 for (i = 0; i < brw->num_atoms; i++) {
516 const struct brw_tracked_state *atom = brw->atoms[i];
518 if (check_state(state, &atom->dirty)) {
524 if (unlikely(INTEL_DEBUG & DEBUG_STATE)) {
525 brw_update_dirty_count(mesa_bits, state->mesa);
526 brw_update_dirty_count(brw_bits, state->brw);
527 brw_update_dirty_count(cache_bits, state->cache);
528 if (dirty_count++ % 1000 == 0) {
529 brw_print_dirty_count(mesa_bits);
530 brw_print_dirty_count(brw_bits);
531 brw_print_dirty_count(cache_bits);
532 fprintf(stderr, "\n");
536 memset(state, 0, sizeof(*state));