2 Copyright (C) Intel Corp. 2006. All Rights Reserved.
3 Intel funded Tungsten Graphics 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 <keithw@vmware.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_ff_gs_prog
, /* must do before state base address */
45 &brw_interpolation_map
,
47 &brw_clip_prog
, /* must do before state base address */
48 &brw_sf_prog
, /* must do before state base address */
49 &brw_wm_prog
, /* must do before state base address */
51 /* Once all the programs are done, we know how large urb entry
52 * sizes need to be and can decide if we need to change the urb
56 &brw_recalculate_urb_fence
,
61 /* Surface state setup. Must come before the VS/WM unit. The binding
62 * table upload must be last.
64 &brw_vs_pull_constants
,
65 &brw_wm_pull_constants
,
66 &brw_renderbuffer_surfaces
,
67 &brw_texture_surfaces
,
68 &brw_vs_binding_table
,
69 &brw_wm_binding_table
,
74 /* These set up state for brw_psp_urb_cbs */
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
[] =
110 &brw_vs_prog
, /* must do before state base address */
111 &brw_ff_gs_prog
, /* must do before state base address */
112 &brw_wm_prog
, /* must do before state base address */
117 /* Command packets: */
119 /* must do before binding table pointers, cc state ptrs */
120 &brw_state_base_address
,
123 &gen6_viewport_state
, /* must do after *_vp stages */
126 &gen6_blend_state
, /* must do before cc unit */
127 &gen6_color_calc_state
, /* must do before cc unit */
128 &gen6_depth_stencil_state
, /* must do before cc unit */
130 &gen6_vs_push_constants
, /* Before vs_state */
131 &gen6_wm_push_constants
, /* Before wm_state */
133 /* Surface state setup. Must come before the VS/WM unit. The binding
134 * table upload must be last.
136 &brw_vs_pull_constants
,
137 &brw_vs_ubo_surfaces
,
138 &brw_wm_pull_constants
,
139 &brw_wm_ubo_surfaces
,
140 &gen6_renderbuffer_surfaces
,
141 &brw_texture_surfaces
,
143 &brw_vs_binding_table
,
144 &gen6_gs_binding_table
,
145 &brw_wm_binding_table
,
150 &gen6_multisample_state
,
160 &gen6_binding_table_pointers
,
164 &brw_polygon_stipple
,
165 &brw_polygon_stipple_offset
,
168 &brw_aa_line_parameters
,
177 static const struct brw_tracked_state
*gen7_atoms
[] =
183 /* Command packets: */
185 /* must do before binding table pointers, cc state ptrs */
186 &brw_state_base_address
,
189 &gen7_cc_viewport_state_pointer
, /* must do after brw_cc_vp */
190 &gen7_sf_clip_viewport
,
192 &gen7_push_constant_space
,
194 &gen6_blend_state
, /* must do before cc unit */
195 &gen6_color_calc_state
, /* must do before cc unit */
196 &gen6_depth_stencil_state
, /* must do before cc unit */
198 &gen6_vs_push_constants
, /* Before vs_state */
199 &gen7_gs_push_constants
, /* Before gs_state */
200 &gen6_wm_push_constants
, /* Before wm_surfaces and constant_buffer */
202 /* Surface state setup. Must come before the VS/WM unit. The binding
203 * table upload must be last.
205 &brw_vs_pull_constants
,
206 &brw_vs_ubo_surfaces
,
207 &brw_vs_abo_surfaces
,
208 &brw_gs_pull_constants
,
209 &brw_gs_ubo_surfaces
,
210 &brw_gs_abo_surfaces
,
211 &brw_wm_pull_constants
,
212 &brw_wm_ubo_surfaces
,
213 &brw_wm_abo_surfaces
,
214 &gen6_renderbuffer_surfaces
,
215 &brw_texture_surfaces
,
216 &brw_vs_binding_table
,
217 &brw_gs_binding_table
,
218 &brw_wm_binding_table
,
223 &gen6_multisample_state
,
225 &gen7_disable_stages
,
239 &brw_polygon_stipple
,
240 &brw_polygon_stipple_offset
,
243 &brw_aa_line_parameters
,
254 static const struct brw_tracked_state
*gen8_atoms
[] =
260 /* Command packets: */
261 &brw_state_base_address
,
264 &gen7_cc_viewport_state_pointer
, /* must do after brw_cc_vp */
265 &gen7_sf_clip_viewport
,
267 &gen7_push_constant_space
,
270 &gen6_color_calc_state
,
272 &gen6_vs_push_constants
, /* Before vs_state */
273 &gen7_gs_push_constants
, /* Before gs_state */
274 &gen6_wm_push_constants
, /* Before wm_surfaces and constant_buffer */
276 /* Surface state setup. Must come before the VS/WM unit. The binding
277 * table upload must be last.
279 &brw_vs_pull_constants
,
280 &brw_vs_ubo_surfaces
,
281 &brw_vs_abo_surfaces
,
282 &brw_gs_pull_constants
,
283 &brw_gs_ubo_surfaces
,
284 &brw_gs_abo_surfaces
,
285 &brw_wm_pull_constants
,
286 &brw_wm_ubo_surfaces
,
287 &brw_wm_abo_surfaces
,
288 &gen6_renderbuffer_surfaces
,
289 &brw_texture_surfaces
,
290 &brw_vs_binding_table
,
291 &brw_gs_binding_table
,
292 &brw_wm_binding_table
,
297 &gen6_multisample_state
,
299 &gen7_disable_stages
,
308 &gen8_wm_depth_stencil
,
316 &brw_polygon_stipple
,
317 &brw_polygon_stipple_offset
,
320 &brw_aa_line_parameters
,
332 brw_upload_initial_gpu_state(struct brw_context
*brw
)
334 /* On platforms with hardware contexts, we can set our initial GPU state
335 * right away rather than doing it via state atoms. This saves a small
336 * amount of overhead on every draw call.
341 brw_upload_invariant_state(brw
);
344 void brw_init_state( struct brw_context
*brw
)
346 struct gl_context
*ctx
= &brw
->ctx
;
347 const struct brw_tracked_state
**atoms
;
350 brw_init_caches(brw
);
354 num_atoms
= ARRAY_SIZE(gen8_atoms
);
355 } else if (brw
->gen
== 7) {
357 num_atoms
= ARRAY_SIZE(gen7_atoms
);
358 } else if (brw
->gen
== 6) {
360 num_atoms
= ARRAY_SIZE(gen6_atoms
);
363 num_atoms
= ARRAY_SIZE(gen4_atoms
);
367 brw
->num_atoms
= num_atoms
;
369 while (num_atoms
--) {
370 assert((*atoms
)->dirty
.mesa
|
371 (*atoms
)->dirty
.brw
|
372 (*atoms
)->dirty
.cache
);
373 assert((*atoms
)->emit
);
377 brw_upload_initial_gpu_state(brw
);
379 brw
->state
.dirty
.mesa
= ~0;
380 brw
->state
.dirty
.brw
= ~0;
382 /* Make sure that brw->state.dirty.brw has enough bits to hold all possible
385 STATIC_ASSERT(BRW_NUM_STATE_BITS
<= 8 * sizeof(brw
->state
.dirty
.brw
));
387 ctx
->DriverFlags
.NewTransformFeedback
= BRW_NEW_TRANSFORM_FEEDBACK
;
388 ctx
->DriverFlags
.NewTransformFeedbackProg
= BRW_NEW_TRANSFORM_FEEDBACK
;
389 ctx
->DriverFlags
.NewRasterizerDiscard
= BRW_NEW_RASTERIZER_DISCARD
;
390 ctx
->DriverFlags
.NewUniformBuffer
= BRW_NEW_UNIFORM_BUFFER
;
391 ctx
->DriverFlags
.NewAtomicBuffer
= BRW_NEW_ATOMIC_BUFFER
;
395 void brw_destroy_state( struct brw_context
*brw
)
397 brw_destroy_caches(brw
);
400 /***********************************************************************
404 check_state(const struct brw_state_flags
*a
, const struct brw_state_flags
*b
)
406 return ((a
->mesa
& b
->mesa
) |
408 (a
->cache
& b
->cache
)) != 0;
411 static void accumulate_state( struct brw_state_flags
*a
,
412 const struct brw_state_flags
*b
)
416 a
->cache
|= b
->cache
;
420 static void xor_states( struct brw_state_flags
*result
,
421 const struct brw_state_flags
*a
,
422 const struct brw_state_flags
*b
)
424 result
->mesa
= a
->mesa
^ b
->mesa
;
425 result
->brw
= a
->brw
^ b
->brw
;
426 result
->cache
= a
->cache
^ b
->cache
;
429 struct dirty_bit_map
{
435 #define DEFINE_BIT(name) {name, #name, 0}
437 static struct dirty_bit_map mesa_bits
[] = {
438 DEFINE_BIT(_NEW_MODELVIEW
),
439 DEFINE_BIT(_NEW_PROJECTION
),
440 DEFINE_BIT(_NEW_TEXTURE_MATRIX
),
441 DEFINE_BIT(_NEW_COLOR
),
442 DEFINE_BIT(_NEW_DEPTH
),
443 DEFINE_BIT(_NEW_EVAL
),
444 DEFINE_BIT(_NEW_FOG
),
445 DEFINE_BIT(_NEW_HINT
),
446 DEFINE_BIT(_NEW_LIGHT
),
447 DEFINE_BIT(_NEW_LINE
),
448 DEFINE_BIT(_NEW_PIXEL
),
449 DEFINE_BIT(_NEW_POINT
),
450 DEFINE_BIT(_NEW_POLYGON
),
451 DEFINE_BIT(_NEW_POLYGONSTIPPLE
),
452 DEFINE_BIT(_NEW_SCISSOR
),
453 DEFINE_BIT(_NEW_STENCIL
),
454 DEFINE_BIT(_NEW_TEXTURE
),
455 DEFINE_BIT(_NEW_TRANSFORM
),
456 DEFINE_BIT(_NEW_VIEWPORT
),
457 DEFINE_BIT(_NEW_ARRAY
),
458 DEFINE_BIT(_NEW_RENDERMODE
),
459 DEFINE_BIT(_NEW_BUFFERS
),
460 DEFINE_BIT(_NEW_CURRENT_ATTRIB
),
461 DEFINE_BIT(_NEW_MULTISAMPLE
),
462 DEFINE_BIT(_NEW_TRACK_MATRIX
),
463 DEFINE_BIT(_NEW_PROGRAM
),
464 DEFINE_BIT(_NEW_PROGRAM_CONSTANTS
),
465 DEFINE_BIT(_NEW_BUFFER_OBJECT
),
466 DEFINE_BIT(_NEW_FRAG_CLAMP
),
467 DEFINE_BIT(_NEW_VARYING_VP_INPUTS
),
471 static struct dirty_bit_map brw_bits
[] = {
472 DEFINE_BIT(BRW_NEW_URB_FENCE
),
473 DEFINE_BIT(BRW_NEW_FRAGMENT_PROGRAM
),
474 DEFINE_BIT(BRW_NEW_GEOMETRY_PROGRAM
),
475 DEFINE_BIT(BRW_NEW_VERTEX_PROGRAM
),
476 DEFINE_BIT(BRW_NEW_CURBE_OFFSETS
),
477 DEFINE_BIT(BRW_NEW_REDUCED_PRIMITIVE
),
478 DEFINE_BIT(BRW_NEW_PRIMITIVE
),
479 DEFINE_BIT(BRW_NEW_CONTEXT
),
480 DEFINE_BIT(BRW_NEW_PSP
),
481 DEFINE_BIT(BRW_NEW_SURFACES
),
482 DEFINE_BIT(BRW_NEW_VS_BINDING_TABLE
),
483 DEFINE_BIT(BRW_NEW_GS_BINDING_TABLE
),
484 DEFINE_BIT(BRW_NEW_PS_BINDING_TABLE
),
485 DEFINE_BIT(BRW_NEW_INDICES
),
486 DEFINE_BIT(BRW_NEW_VERTICES
),
487 DEFINE_BIT(BRW_NEW_BATCH
),
488 DEFINE_BIT(BRW_NEW_INDEX_BUFFER
),
489 DEFINE_BIT(BRW_NEW_VS_CONSTBUF
),
490 DEFINE_BIT(BRW_NEW_GS_CONSTBUF
),
491 DEFINE_BIT(BRW_NEW_PROGRAM_CACHE
),
492 DEFINE_BIT(BRW_NEW_STATE_BASE_ADDRESS
),
493 DEFINE_BIT(BRW_NEW_VUE_MAP_VS
),
494 DEFINE_BIT(BRW_NEW_VUE_MAP_GEOM_OUT
),
495 DEFINE_BIT(BRW_NEW_TRANSFORM_FEEDBACK
),
496 DEFINE_BIT(BRW_NEW_RASTERIZER_DISCARD
),
497 DEFINE_BIT(BRW_NEW_STATS_WM
),
498 DEFINE_BIT(BRW_NEW_UNIFORM_BUFFER
),
499 DEFINE_BIT(BRW_NEW_ATOMIC_BUFFER
),
500 DEFINE_BIT(BRW_NEW_META_IN_PROGRESS
),
501 DEFINE_BIT(BRW_NEW_INTERPOLATION_MAP
),
502 DEFINE_BIT(BRW_NEW_PUSH_CONSTANT_ALLOCATION
),
506 static struct dirty_bit_map cache_bits
[] = {
507 DEFINE_BIT(CACHE_NEW_CC_VP
),
508 DEFINE_BIT(CACHE_NEW_CC_UNIT
),
509 DEFINE_BIT(CACHE_NEW_WM_PROG
),
510 DEFINE_BIT(CACHE_NEW_BLORP_BLIT_PROG
),
511 DEFINE_BIT(CACHE_NEW_BLORP_CONST_COLOR_PROG
),
512 DEFINE_BIT(CACHE_NEW_SAMPLER
),
513 DEFINE_BIT(CACHE_NEW_WM_UNIT
),
514 DEFINE_BIT(CACHE_NEW_SF_PROG
),
515 DEFINE_BIT(CACHE_NEW_SF_VP
),
516 DEFINE_BIT(CACHE_NEW_SF_UNIT
),
517 DEFINE_BIT(CACHE_NEW_VS_UNIT
),
518 DEFINE_BIT(CACHE_NEW_VS_PROG
),
519 DEFINE_BIT(CACHE_NEW_FF_GS_UNIT
),
520 DEFINE_BIT(CACHE_NEW_FF_GS_PROG
),
521 DEFINE_BIT(CACHE_NEW_GS_PROG
),
522 DEFINE_BIT(CACHE_NEW_CLIP_VP
),
523 DEFINE_BIT(CACHE_NEW_CLIP_UNIT
),
524 DEFINE_BIT(CACHE_NEW_CLIP_PROG
),
530 brw_update_dirty_count(struct dirty_bit_map
*bit_map
, int32_t bits
)
534 for (i
= 0; i
< 32; i
++) {
535 if (bit_map
[i
].bit
== 0)
538 if (bit_map
[i
].bit
& bits
)
544 brw_print_dirty_count(struct dirty_bit_map
*bit_map
)
548 for (i
= 0; i
< 32; i
++) {
549 if (bit_map
[i
].bit
== 0)
552 fprintf(stderr
, "0x%08x: %12d (%s)\n",
553 bit_map
[i
].bit
, bit_map
[i
].count
, bit_map
[i
].name
);
557 /***********************************************************************
560 void brw_upload_state(struct brw_context
*brw
)
562 struct gl_context
*ctx
= &brw
->ctx
;
563 struct brw_state_flags
*state
= &brw
->state
.dirty
;
565 static int dirty_count
= 0;
567 state
->mesa
|= brw
->NewGLState
;
570 state
->brw
|= ctx
->NewDriverState
;
571 ctx
->NewDriverState
= 0;
574 /* Always re-emit all state. */
580 if (brw
->fragment_program
!= ctx
->FragmentProgram
._Current
) {
581 brw
->fragment_program
= ctx
->FragmentProgram
._Current
;
582 brw
->state
.dirty
.brw
|= BRW_NEW_FRAGMENT_PROGRAM
;
585 if (brw
->geometry_program
!= ctx
->GeometryProgram
._Current
) {
586 brw
->geometry_program
= ctx
->GeometryProgram
._Current
;
587 brw
->state
.dirty
.brw
|= BRW_NEW_GEOMETRY_PROGRAM
;
590 if (brw
->vertex_program
!= ctx
->VertexProgram
._Current
) {
591 brw
->vertex_program
= ctx
->VertexProgram
._Current
;
592 brw
->state
.dirty
.brw
|= BRW_NEW_VERTEX_PROGRAM
;
595 if (brw
->meta_in_progress
!= _mesa_meta_in_progress(ctx
)) {
596 brw
->meta_in_progress
= _mesa_meta_in_progress(ctx
);
597 brw
->state
.dirty
.brw
|= BRW_NEW_META_IN_PROGRESS
;
600 if ((state
->mesa
| state
->cache
| state
->brw
) == 0)
603 intel_check_front_buffer_rendering(brw
);
605 if (unlikely(INTEL_DEBUG
)) {
606 /* Debug version which enforces various sanity checks on the
607 * state flags which are generated and checked to help ensure
608 * state atoms are ordered correctly in the list.
610 struct brw_state_flags examined
, prev
;
611 memset(&examined
, 0, sizeof(examined
));
614 for (i
= 0; i
< brw
->num_atoms
; i
++) {
615 const struct brw_tracked_state
*atom
= brw
->atoms
[i
];
616 struct brw_state_flags generated
;
618 if (check_state(state
, &atom
->dirty
)) {
622 accumulate_state(&examined
, &atom
->dirty
);
624 /* generated = (prev ^ state)
625 * if (examined & generated)
628 xor_states(&generated
, &prev
, state
);
629 assert(!check_state(&examined
, &generated
));
634 for (i
= 0; i
< brw
->num_atoms
; i
++) {
635 const struct brw_tracked_state
*atom
= brw
->atoms
[i
];
637 if (check_state(state
, &atom
->dirty
)) {
643 if (unlikely(INTEL_DEBUG
& DEBUG_STATE
)) {
644 STATIC_ASSERT(ARRAY_SIZE(brw_bits
) == BRW_NUM_STATE_BITS
+ 1);
645 STATIC_ASSERT(ARRAY_SIZE(cache_bits
) == BRW_MAX_CACHE
+ 1);
647 brw_update_dirty_count(mesa_bits
, state
->mesa
);
648 brw_update_dirty_count(brw_bits
, state
->brw
);
649 brw_update_dirty_count(cache_bits
, state
->cache
);
650 if (dirty_count
++ % 1000 == 0) {
651 brw_print_dirty_count(mesa_bits
);
652 brw_print_dirty_count(brw_bits
);
653 brw_print_dirty_count(cache_bits
);
654 fprintf(stderr
, "\n");
661 * Clear dirty bits to account for the fact that the state emitted by
662 * brw_upload_state() has been committed to the hardware. This is a separate
663 * call from brw_upload_state() because it's possible that after the call to
664 * brw_upload_state(), we will discover that we've run out of aperture space,
665 * and need to rewind the batch buffer to the state it had before the
666 * brw_upload_state() call.
669 brw_clear_dirty_bits(struct brw_context
*brw
)
671 struct brw_state_flags
*state
= &brw
->state
.dirty
;
672 memset(state
, 0, sizeof(*state
));