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 "intel_batchbuffer.h"
37 #include "intel_buffers.h"
39 /* This is used to initialize brw->state.atoms[]. We could use this
40 * list directly except for a single atom, brw_constant_buffer, which
41 * has a .dirty value which changes according to the parameters of the
42 * current fragment and vertex programs, and so cannot be a static
45 static const struct brw_tracked_state
*gen4_atoms
[] =
56 /* Once all the programs are done, we know how large urb entry
57 * sizes need to be and can decide if we need to change the urb
61 &brw_recalculate_urb_fence
,
66 &brw_vs_constants
, /* Before vs_surfaces and constant_buffer */
67 &brw_wm_constants
, /* Before wm_surfaces and constant_buffer */
69 &brw_vs_surfaces
, /* must do before unit */
70 &brw_wm_constant_surface
, /* must do before wm surfaces/bind bo */
71 &brw_wm_surfaces
, /* must do before samplers and unit */
72 &brw_wm_binding_table
,
78 &brw_vs_unit
, /* always required, enabled or not */
85 &brw_state_base_address
,
87 &brw_binding_table_pointers
,
88 &brw_blend_constant_color
,
93 &brw_polygon_stipple_offset
,
96 &brw_aa_line_parameters
,
108 static const struct brw_tracked_state
*gen6_atoms
[] =
120 /* Command packets: */
121 &brw_invarient_state
,
123 /* must do before binding table pointers, cc state ptrs */
124 &brw_state_base_address
,
127 &gen6_viewport_state
, /* must do after *_vp stages */
130 &gen6_blend_state
, /* must do before cc unit */
131 &gen6_color_calc_state
, /* must do before cc unit */
132 &gen6_depth_stencil_state
, /* must do before cc unit */
133 &gen6_cc_state_pointers
,
135 &brw_vs_constants
, /* Before vs_surfaces and constant_buffer */
136 &brw_wm_constants
, /* Before wm_surfaces and constant_buffer */
137 &gen6_vs_constants
, /* Before vs_state */
138 &gen6_wm_constants
, /* Before wm_state */
140 &brw_vs_surfaces
, /* must do before unit */
141 &brw_wm_constant_surface
, /* must do before wm surfaces/bind bo */
142 &brw_wm_surfaces
, /* must do before samplers and unit */
143 &brw_wm_binding_table
,
156 &gen6_binding_table_pointers
,
160 &brw_polygon_stipple
,
161 &brw_polygon_stipple_offset
,
164 &brw_aa_line_parameters
,
173 const struct brw_tracked_state
*gen7_atoms
[] =
185 /* Command packets: */
186 &brw_invarient_state
,
188 /* must do before binding table pointers, cc state ptrs */
189 &brw_state_base_address
,
192 &gen6_viewport_state
, /* must do after *_vp stages */
195 &gen6_blend_state
, /* must do before cc unit */
196 &gen6_color_calc_state
, /* must do before cc unit */
197 &gen6_depth_stencil_state
, /* must do before cc unit */
198 &gen7_blend_state_pointer
,
199 &gen7_cc_state_pointer
,
200 &gen7_depth_stencil_state_pointer
,
202 &brw_vs_constants
, /* Before vs_surfaces and constant_buffer */
203 &brw_wm_constants
, /* Before wm_surfaces and constant_buffer */
204 &gen6_vs_constants
, /* Before vs_state */
205 &gen7_wm_constants
, /* Before wm_state */
207 &brw_vs_surfaces
, /* must do before unit */
208 &brw_wm_constant_surface
, /* must do before wm surfaces/bind bo */
209 &brw_wm_surfaces
, /* must do before samplers and unit */
210 &brw_wm_binding_table
,
225 &gen6_binding_table_pointers
,
229 &brw_polygon_stipple
,
230 &brw_polygon_stipple_offset
,
233 &brw_aa_line_parameters
,
243 void brw_init_state( struct brw_context
*brw
)
245 const struct brw_tracked_state
**atoms
;
248 brw_init_caches(brw
);
250 if (brw
->intel
.gen
>= 7) {
252 num_atoms
= ARRAY_SIZE(gen7_atoms
);
253 } else if (brw
->intel
.gen
== 6) {
255 num_atoms
= ARRAY_SIZE(gen6_atoms
);
258 num_atoms
= ARRAY_SIZE(gen4_atoms
);
261 while (num_atoms
--) {
262 assert((*atoms
)->dirty
.mesa
|
263 (*atoms
)->dirty
.brw
|
264 (*atoms
)->dirty
.cache
);
266 if ((*atoms
)->prepare
)
267 brw
->prepare_atoms
[brw
->num_prepare_atoms
++] = **atoms
;
269 brw
->emit_atoms
[brw
->num_emit_atoms
++] = **atoms
;
272 assert(brw
->num_emit_atoms
<= ARRAY_SIZE(brw
->emit_atoms
));
273 assert(brw
->num_prepare_atoms
<= ARRAY_SIZE(brw
->prepare_atoms
));
277 void brw_destroy_state( struct brw_context
*brw
)
279 brw_destroy_caches(brw
);
282 /***********************************************************************
285 static GLuint
check_state( const struct brw_state_flags
*a
,
286 const struct brw_state_flags
*b
)
288 return ((a
->mesa
& b
->mesa
) |
290 (a
->cache
& b
->cache
)) != 0;
293 static void accumulate_state( struct brw_state_flags
*a
,
294 const struct brw_state_flags
*b
)
298 a
->cache
|= b
->cache
;
302 static void xor_states( struct brw_state_flags
*result
,
303 const struct brw_state_flags
*a
,
304 const struct brw_state_flags
*b
)
306 result
->mesa
= a
->mesa
^ b
->mesa
;
307 result
->brw
= a
->brw
^ b
->brw
;
308 result
->cache
= a
->cache
^ b
->cache
;
312 brw_clear_validated_bos(struct brw_context
*brw
)
316 /* Clear the last round of validated bos */
317 for (i
= 0; i
< brw
->state
.validated_bo_count
; i
++) {
318 drm_intel_bo_unreference(brw
->state
.validated_bos
[i
]);
319 brw
->state
.validated_bos
[i
] = NULL
;
321 brw
->state
.validated_bo_count
= 0;
324 struct dirty_bit_map
{
330 #define DEFINE_BIT(name) {name, #name, 0}
332 static struct dirty_bit_map mesa_bits
[] = {
333 DEFINE_BIT(_NEW_MODELVIEW
),
334 DEFINE_BIT(_NEW_PROJECTION
),
335 DEFINE_BIT(_NEW_TEXTURE_MATRIX
),
336 DEFINE_BIT(_NEW_COLOR
),
337 DEFINE_BIT(_NEW_DEPTH
),
338 DEFINE_BIT(_NEW_EVAL
),
339 DEFINE_BIT(_NEW_FOG
),
340 DEFINE_BIT(_NEW_HINT
),
341 DEFINE_BIT(_NEW_LIGHT
),
342 DEFINE_BIT(_NEW_LINE
),
343 DEFINE_BIT(_NEW_PIXEL
),
344 DEFINE_BIT(_NEW_POINT
),
345 DEFINE_BIT(_NEW_POLYGON
),
346 DEFINE_BIT(_NEW_POLYGONSTIPPLE
),
347 DEFINE_BIT(_NEW_SCISSOR
),
348 DEFINE_BIT(_NEW_STENCIL
),
349 DEFINE_BIT(_NEW_TEXTURE
),
350 DEFINE_BIT(_NEW_TRANSFORM
),
351 DEFINE_BIT(_NEW_VIEWPORT
),
352 DEFINE_BIT(_NEW_PACKUNPACK
),
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
),
363 static struct dirty_bit_map brw_bits
[] = {
364 DEFINE_BIT(BRW_NEW_URB_FENCE
),
365 DEFINE_BIT(BRW_NEW_FRAGMENT_PROGRAM
),
366 DEFINE_BIT(BRW_NEW_VERTEX_PROGRAM
),
367 DEFINE_BIT(BRW_NEW_INPUT_DIMENSIONS
),
368 DEFINE_BIT(BRW_NEW_CURBE_OFFSETS
),
369 DEFINE_BIT(BRW_NEW_REDUCED_PRIMITIVE
),
370 DEFINE_BIT(BRW_NEW_PRIMITIVE
),
371 DEFINE_BIT(BRW_NEW_CONTEXT
),
372 DEFINE_BIT(BRW_NEW_WM_INPUT_DIMENSIONS
),
373 DEFINE_BIT(BRW_NEW_PSP
),
374 DEFINE_BIT(BRW_NEW_WM_SURFACES
),
375 DEFINE_BIT(BRW_NEW_BINDING_TABLE
),
376 DEFINE_BIT(BRW_NEW_INDICES
),
377 DEFINE_BIT(BRW_NEW_INDEX_BUFFER
),
378 DEFINE_BIT(BRW_NEW_VERTICES
),
379 DEFINE_BIT(BRW_NEW_BATCH
),
380 DEFINE_BIT(BRW_NEW_DEPTH_BUFFER
),
381 DEFINE_BIT(BRW_NEW_NR_WM_SURFACES
),
382 DEFINE_BIT(BRW_NEW_NR_VS_SURFACES
),
383 DEFINE_BIT(BRW_NEW_VS_CONSTBUF
),
384 DEFINE_BIT(BRW_NEW_WM_CONSTBUF
),
388 static struct dirty_bit_map cache_bits
[] = {
389 DEFINE_BIT(CACHE_NEW_BLEND_STATE
),
390 DEFINE_BIT(CACHE_NEW_CC_VP
),
391 DEFINE_BIT(CACHE_NEW_CC_UNIT
),
392 DEFINE_BIT(CACHE_NEW_WM_PROG
),
393 DEFINE_BIT(CACHE_NEW_SAMPLER
),
394 DEFINE_BIT(CACHE_NEW_WM_UNIT
),
395 DEFINE_BIT(CACHE_NEW_SF_PROG
),
396 DEFINE_BIT(CACHE_NEW_SF_VP
),
397 DEFINE_BIT(CACHE_NEW_SF_UNIT
),
398 DEFINE_BIT(CACHE_NEW_VS_UNIT
),
399 DEFINE_BIT(CACHE_NEW_VS_PROG
),
400 DEFINE_BIT(CACHE_NEW_GS_UNIT
),
401 DEFINE_BIT(CACHE_NEW_GS_PROG
),
402 DEFINE_BIT(CACHE_NEW_CLIP_VP
),
403 DEFINE_BIT(CACHE_NEW_CLIP_UNIT
),
404 DEFINE_BIT(CACHE_NEW_CLIP_PROG
),
410 brw_update_dirty_count(struct dirty_bit_map
*bit_map
, int32_t bits
)
414 for (i
= 0; i
< 32; i
++) {
415 if (bit_map
[i
].bit
== 0)
418 if (bit_map
[i
].bit
& bits
)
424 brw_print_dirty_count(struct dirty_bit_map
*bit_map
, int32_t bits
)
428 for (i
= 0; i
< 32; i
++) {
429 if (bit_map
[i
].bit
== 0)
432 fprintf(stderr
, "0x%08x: %12d (%s)\n",
433 bit_map
[i
].bit
, bit_map
[i
].count
, bit_map
[i
].name
);
437 /***********************************************************************
440 void brw_validate_state( struct brw_context
*brw
)
442 struct gl_context
*ctx
= &brw
->intel
.ctx
;
443 struct intel_context
*intel
= &brw
->intel
;
444 struct brw_state_flags
*state
= &brw
->state
.dirty
;
445 const struct brw_tracked_state
*atoms
= brw
->prepare_atoms
;
446 int num_atoms
= brw
->num_prepare_atoms
;
449 brw_clear_validated_bos(brw
);
451 state
->mesa
|= brw
->intel
.NewGLState
;
452 brw
->intel
.NewGLState
= 0;
454 brw_add_validated_bo(brw
, intel
->batch
.bo
);
456 if (brw
->emit_state_always
) {
462 if (brw
->fragment_program
!= ctx
->FragmentProgram
._Current
) {
463 brw
->fragment_program
= ctx
->FragmentProgram
._Current
;
464 brw
->state
.dirty
.brw
|= BRW_NEW_FRAGMENT_PROGRAM
;
467 if (brw
->vertex_program
!= ctx
->VertexProgram
._Current
) {
468 brw
->vertex_program
= ctx
->VertexProgram
._Current
;
469 brw
->state
.dirty
.brw
|= BRW_NEW_VERTEX_PROGRAM
;
472 if ((state
->mesa
| state
->cache
| state
->brw
) == 0)
475 brw
->intel
.Fallback
= GL_FALSE
; /* boolean, not bitfield */
477 /* do prepare stage for all atoms */
478 for (i
= 0; i
< num_atoms
; i
++) {
479 const struct brw_tracked_state
*atom
= &atoms
[i
];
481 if (check_state(state
, &atom
->dirty
)) {
484 if (brw
->intel
.Fallback
)
489 intel_check_front_buffer_rendering(intel
);
491 /* Make sure that the textures which are referenced by the current
492 * brw fragment program are actually present/valid.
493 * If this fails, we can experience GPU lock-ups.
496 const struct brw_fragment_program
*fp
;
497 fp
= brw_fragment_program_const(brw
->fragment_program
);
499 assert((fp
->tex_units_used
& ctx
->Texture
._EnabledUnits
)
500 == fp
->tex_units_used
);
506 void brw_upload_state(struct brw_context
*brw
)
508 struct brw_state_flags
*state
= &brw
->state
.dirty
;
509 const struct brw_tracked_state
*atoms
= brw
->emit_atoms
;
510 int num_atoms
= brw
->num_emit_atoms
;
512 static int dirty_count
= 0;
514 brw_clear_validated_bos(brw
);
516 if (unlikely(INTEL_DEBUG
)) {
517 /* Debug version which enforces various sanity checks on the
518 * state flags which are generated and checked to help ensure
519 * state atoms are ordered correctly in the list.
521 struct brw_state_flags examined
, prev
;
522 memset(&examined
, 0, sizeof(examined
));
525 for (i
= 0; i
< num_atoms
; i
++) {
526 const struct brw_tracked_state
*atom
= &atoms
[i
];
527 struct brw_state_flags generated
;
529 if (brw
->intel
.Fallback
)
532 if (check_state(state
, &atom
->dirty
)) {
536 accumulate_state(&examined
, &atom
->dirty
);
538 /* generated = (prev ^ state)
539 * if (examined & generated)
542 xor_states(&generated
, &prev
, state
);
543 assert(!check_state(&examined
, &generated
));
548 for (i
= 0; i
< num_atoms
; i
++) {
549 const struct brw_tracked_state
*atom
= &atoms
[i
];
551 if (brw
->intel
.Fallback
)
554 if (check_state(state
, &atom
->dirty
)) {
560 if (unlikely(INTEL_DEBUG
& DEBUG_STATE
)) {
561 brw_update_dirty_count(mesa_bits
, state
->mesa
);
562 brw_update_dirty_count(brw_bits
, state
->brw
);
563 brw_update_dirty_count(cache_bits
, state
->cache
);
564 if (dirty_count
++ % 1000 == 0) {
565 brw_print_dirty_count(mesa_bits
, state
->mesa
);
566 brw_print_dirty_count(brw_bits
, state
->brw
);
567 brw_print_dirty_count(cache_bits
, state
->cache
);
568 fprintf(stderr
, "\n");
572 if (!brw
->intel
.Fallback
)
573 memset(state
, 0, sizeof(*state
));