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 #include "brw_ff_gs.h"
44 #include "main/framebuffer.h"
46 static const struct brw_tracked_state
*gen4_atoms
[] =
48 &brw_interpolation_map
,
50 &brw_clip_prog
, /* must do before state base address */
51 &brw_sf_prog
, /* must do before state base address */
53 /* Once all the programs are done, we know how large urb entry
54 * sizes need to be and can decide if we need to change the urb
58 &brw_recalculate_urb_fence
,
63 /* Surface state setup. Must come before the VS/WM unit. The binding
64 * table upload must be last.
66 &brw_vs_pull_constants
,
67 &brw_wm_pull_constants
,
68 &brw_renderbuffer_surfaces
,
69 &brw_texture_surfaces
,
70 &brw_vs_binding_table
,
71 &brw_wm_binding_table
,
76 /* These set up state for brw_psp_urb_cbs */
80 &brw_vs_unit
, /* always required, enabled or not */
87 &brw_state_base_address
,
89 &brw_binding_table_pointers
,
90 &brw_blend_constant_color
,
95 &brw_polygon_stipple_offset
,
98 &brw_aa_line_parameters
,
103 &brw_indices
, /* must come before brw_vertices */
110 static const struct brw_tracked_state
*gen6_atoms
[] =
115 /* Command packets: */
117 /* must do before binding table pointers, cc state ptrs */
118 &brw_state_base_address
,
121 &gen6_viewport_state
, /* must do after *_vp stages */
124 &gen6_blend_state
, /* must do before cc unit */
125 &gen6_color_calc_state
, /* must do before cc unit */
126 &gen6_depth_stencil_state
, /* must do before cc unit */
128 &gen6_vs_push_constants
, /* Before vs_state */
129 &gen6_gs_push_constants
, /* Before gs_state */
130 &gen6_wm_push_constants
, /* Before wm_state */
132 /* Surface state setup. Must come before the VS/WM unit. The binding
133 * table upload must be last.
135 &brw_vs_pull_constants
,
136 &brw_vs_ubo_surfaces
,
137 &brw_gs_pull_constants
,
138 &brw_gs_ubo_surfaces
,
139 &brw_wm_pull_constants
,
140 &brw_wm_ubo_surfaces
,
141 &gen6_renderbuffer_surfaces
,
142 &brw_texture_surfaces
,
144 &brw_vs_binding_table
,
145 &gen6_gs_binding_table
,
146 &brw_wm_binding_table
,
152 &gen6_multisample_state
,
162 &gen6_binding_table_pointers
,
166 &brw_polygon_stipple
,
167 &brw_polygon_stipple_offset
,
170 &brw_aa_line_parameters
,
174 &brw_indices
, /* must come before brw_vertices */
179 static const struct brw_tracked_state
*gen7_render_atoms
[] =
181 /* Command packets: */
183 /* must do before binding table pointers, cc state ptrs */
184 &brw_state_base_address
,
187 &gen7_sf_clip_viewport
,
189 &gen7_push_constant_space
,
191 &gen6_blend_state
, /* must do before cc unit */
192 &gen6_color_calc_state
, /* must do before cc unit */
193 &gen6_depth_stencil_state
, /* must do before cc unit */
195 &gen7_hw_binding_tables
, /* Enable hw-generated binding tables for Haswell */
197 &brw_vs_image_surfaces
, /* Before vs push/pull constants and binding table */
198 &brw_gs_image_surfaces
, /* Before gs push/pull constants and binding table */
199 &brw_wm_image_surfaces
, /* Before wm push/pull constants and binding table */
201 &gen6_vs_push_constants
, /* Before vs_state */
202 &gen6_gs_push_constants
, /* Before gs_state */
203 &gen6_wm_push_constants
, /* Before wm_surfaces and constant_buffer */
205 /* Surface state setup. Must come before the VS/WM unit. The binding
206 * table upload must be last.
208 &brw_vs_pull_constants
,
209 &brw_vs_ubo_surfaces
,
210 &brw_vs_abo_surfaces
,
211 &brw_gs_pull_constants
,
212 &brw_gs_ubo_surfaces
,
213 &brw_gs_abo_surfaces
,
214 &brw_wm_pull_constants
,
215 &brw_wm_ubo_surfaces
,
216 &brw_wm_abo_surfaces
,
217 &gen6_renderbuffer_surfaces
,
218 &brw_texture_surfaces
,
219 &brw_vs_binding_table
,
220 &brw_gs_binding_table
,
221 &brw_wm_binding_table
,
226 &gen6_multisample_state
,
244 &brw_polygon_stipple
,
245 &brw_polygon_stipple_offset
,
248 &brw_aa_line_parameters
,
252 &brw_indices
, /* must come before brw_vertices */
259 static const struct brw_tracked_state
*gen7_compute_atoms
[] =
261 &brw_state_base_address
,
262 &brw_cs_image_surfaces
,
263 &gen7_cs_push_constants
,
264 &brw_cs_pull_constants
,
265 &brw_cs_ubo_surfaces
,
266 &brw_cs_abo_surfaces
,
267 &brw_texture_surfaces
,
268 &brw_cs_work_groups_surface
,
272 static const struct brw_tracked_state
*gen8_render_atoms
[] =
274 /* Command packets: */
275 &gen8_state_base_address
,
278 &gen8_sf_clip_viewport
,
280 &gen7_push_constant_space
,
283 &gen6_color_calc_state
,
285 &gen7_hw_binding_tables
, /* Enable hw-generated binding tables for Broadwell */
287 &brw_vs_image_surfaces
, /* Before vs push/pull constants and binding table */
288 &brw_gs_image_surfaces
, /* Before gs push/pull constants and binding table */
289 &brw_wm_image_surfaces
, /* Before wm push/pull constants and binding table */
291 &gen6_vs_push_constants
, /* Before vs_state */
292 &gen6_gs_push_constants
, /* Before gs_state */
293 &gen6_wm_push_constants
, /* Before wm_surfaces and constant_buffer */
295 /* Surface state setup. Must come before the VS/WM unit. The binding
296 * table upload must be last.
298 &brw_vs_pull_constants
,
299 &brw_vs_ubo_surfaces
,
300 &brw_vs_abo_surfaces
,
301 &brw_gs_pull_constants
,
302 &brw_gs_ubo_surfaces
,
303 &brw_gs_abo_surfaces
,
304 &brw_wm_pull_constants
,
305 &brw_wm_ubo_surfaces
,
306 &brw_wm_abo_surfaces
,
307 &gen6_renderbuffer_surfaces
,
308 &brw_texture_surfaces
,
309 &brw_vs_binding_table
,
310 &brw_gs_binding_table
,
311 &brw_wm_binding_table
,
316 &gen8_multisample_state
,
318 &gen8_disable_stages
,
332 &gen8_wm_depth_stencil
,
339 &brw_polygon_stipple
,
340 &brw_polygon_stipple_offset
,
343 &brw_aa_line_parameters
,
357 static const struct brw_tracked_state
*gen8_compute_atoms
[] =
359 &gen8_state_base_address
,
360 &brw_cs_image_surfaces
,
361 &gen7_cs_push_constants
,
362 &brw_cs_pull_constants
,
363 &brw_cs_ubo_surfaces
,
364 &brw_cs_abo_surfaces
,
365 &brw_texture_surfaces
,
366 &brw_cs_work_groups_surface
,
371 brw_upload_initial_gpu_state(struct brw_context
*brw
)
373 /* On platforms with hardware contexts, we can set our initial GPU state
374 * right away rather than doing it via state atoms. This saves a small
375 * amount of overhead on every draw call.
381 brw_emit_post_sync_nonzero_flush(brw
);
383 brw_upload_invariant_state(brw
);
385 /* Recommended optimization for Victim Cache eviction in pixel backend. */
388 OUT_BATCH(MI_LOAD_REGISTER_IMM
| (3 - 2));
389 OUT_BATCH(GEN7_CACHE_MODE_1
);
390 OUT_BATCH((GEN9_PARTIAL_RESOLVE_DISABLE_IN_VC
<< 16) |
391 GEN9_PARTIAL_RESOLVE_DISABLE_IN_VC
);
396 gen8_emit_3dstate_sample_pattern(brw
);
400 static inline const struct brw_tracked_state
*
401 brw_get_pipeline_atoms(struct brw_context
*brw
,
402 enum brw_pipeline pipeline
)
405 case BRW_RENDER_PIPELINE
:
406 return brw
->render_atoms
;
407 case BRW_COMPUTE_PIPELINE
:
408 return brw
->compute_atoms
;
410 STATIC_ASSERT(BRW_NUM_PIPELINES
== 2);
411 unreachable("Unsupported pipeline");
417 brw_copy_pipeline_atoms(struct brw_context
*brw
,
418 enum brw_pipeline pipeline
,
419 const struct brw_tracked_state
**atoms
,
422 /* This is to work around brw_context::atoms being declared const. We want
423 * it to be const, but it needs to be initialized somehow!
425 struct brw_tracked_state
*context_atoms
=
426 (struct brw_tracked_state
*) brw_get_pipeline_atoms(brw
, pipeline
);
428 for (int i
= 0; i
< num_atoms
; i
++) {
429 context_atoms
[i
] = *atoms
[i
];
430 assert(context_atoms
[i
].dirty
.mesa
| context_atoms
[i
].dirty
.brw
);
431 assert(context_atoms
[i
].emit
);
434 brw
->num_atoms
[pipeline
] = num_atoms
;
437 void brw_init_state( struct brw_context
*brw
)
439 struct gl_context
*ctx
= &brw
->ctx
;
441 /* Force the first brw_select_pipeline to emit pipeline select */
442 brw
->last_pipeline
= BRW_NUM_PIPELINES
;
444 STATIC_ASSERT(ARRAY_SIZE(gen4_atoms
) <= ARRAY_SIZE(brw
->render_atoms
));
445 STATIC_ASSERT(ARRAY_SIZE(gen6_atoms
) <= ARRAY_SIZE(brw
->render_atoms
));
446 STATIC_ASSERT(ARRAY_SIZE(gen7_render_atoms
) <=
447 ARRAY_SIZE(brw
->render_atoms
));
448 STATIC_ASSERT(ARRAY_SIZE(gen8_render_atoms
) <=
449 ARRAY_SIZE(brw
->render_atoms
));
450 STATIC_ASSERT(ARRAY_SIZE(gen7_compute_atoms
) <=
451 ARRAY_SIZE(brw
->compute_atoms
));
452 STATIC_ASSERT(ARRAY_SIZE(gen8_compute_atoms
) <=
453 ARRAY_SIZE(brw
->compute_atoms
));
455 brw_init_caches(brw
);
458 brw_copy_pipeline_atoms(brw
, BRW_RENDER_PIPELINE
,
460 ARRAY_SIZE(gen8_render_atoms
));
461 brw_copy_pipeline_atoms(brw
, BRW_COMPUTE_PIPELINE
,
463 ARRAY_SIZE(gen8_compute_atoms
));
464 } else if (brw
->gen
== 7) {
465 brw_copy_pipeline_atoms(brw
, BRW_RENDER_PIPELINE
,
467 ARRAY_SIZE(gen7_render_atoms
));
468 brw_copy_pipeline_atoms(brw
, BRW_COMPUTE_PIPELINE
,
470 ARRAY_SIZE(gen7_compute_atoms
));
471 } else if (brw
->gen
== 6) {
472 brw_copy_pipeline_atoms(brw
, BRW_RENDER_PIPELINE
,
473 gen6_atoms
, ARRAY_SIZE(gen6_atoms
));
475 brw_copy_pipeline_atoms(brw
, BRW_RENDER_PIPELINE
,
476 gen4_atoms
, ARRAY_SIZE(gen4_atoms
));
479 brw_upload_initial_gpu_state(brw
);
481 brw
->NewGLState
= ~0;
482 brw
->ctx
.NewDriverState
= ~0ull;
484 /* ~0 is a nonsensical value which won't match anything we program, so
485 * the programming will take effect on the first time around.
487 brw
->pma_stall_bits
= ~0;
489 /* Make sure that brw->ctx.NewDriverState has enough bits to hold all possible
492 STATIC_ASSERT(BRW_NUM_STATE_BITS
<= 8 * sizeof(brw
->ctx
.NewDriverState
));
494 ctx
->DriverFlags
.NewTransformFeedback
= BRW_NEW_TRANSFORM_FEEDBACK
;
495 ctx
->DriverFlags
.NewTransformFeedbackProg
= BRW_NEW_TRANSFORM_FEEDBACK
;
496 ctx
->DriverFlags
.NewRasterizerDiscard
= BRW_NEW_RASTERIZER_DISCARD
;
497 ctx
->DriverFlags
.NewUniformBuffer
= BRW_NEW_UNIFORM_BUFFER
;
498 ctx
->DriverFlags
.NewShaderStorageBuffer
= BRW_NEW_UNIFORM_BUFFER
;
499 ctx
->DriverFlags
.NewTextureBuffer
= BRW_NEW_TEXTURE_BUFFER
;
500 ctx
->DriverFlags
.NewAtomicBuffer
= BRW_NEW_ATOMIC_BUFFER
;
501 ctx
->DriverFlags
.NewImageUnits
= BRW_NEW_IMAGE_UNITS
;
505 void brw_destroy_state( struct brw_context
*brw
)
507 brw_destroy_caches(brw
);
510 /***********************************************************************
514 check_state(const struct brw_state_flags
*a
, const struct brw_state_flags
*b
)
516 return ((a
->mesa
& b
->mesa
) | (a
->brw
& b
->brw
)) != 0;
519 static void accumulate_state( struct brw_state_flags
*a
,
520 const struct brw_state_flags
*b
)
527 static void xor_states( struct brw_state_flags
*result
,
528 const struct brw_state_flags
*a
,
529 const struct brw_state_flags
*b
)
531 result
->mesa
= a
->mesa
^ b
->mesa
;
532 result
->brw
= a
->brw
^ b
->brw
;
535 struct dirty_bit_map
{
541 #define DEFINE_BIT(name) {name, #name, 0}
543 static struct dirty_bit_map mesa_bits
[] = {
544 DEFINE_BIT(_NEW_MODELVIEW
),
545 DEFINE_BIT(_NEW_PROJECTION
),
546 DEFINE_BIT(_NEW_TEXTURE_MATRIX
),
547 DEFINE_BIT(_NEW_COLOR
),
548 DEFINE_BIT(_NEW_DEPTH
),
549 DEFINE_BIT(_NEW_EVAL
),
550 DEFINE_BIT(_NEW_FOG
),
551 DEFINE_BIT(_NEW_HINT
),
552 DEFINE_BIT(_NEW_LIGHT
),
553 DEFINE_BIT(_NEW_LINE
),
554 DEFINE_BIT(_NEW_PIXEL
),
555 DEFINE_BIT(_NEW_POINT
),
556 DEFINE_BIT(_NEW_POLYGON
),
557 DEFINE_BIT(_NEW_POLYGONSTIPPLE
),
558 DEFINE_BIT(_NEW_SCISSOR
),
559 DEFINE_BIT(_NEW_STENCIL
),
560 DEFINE_BIT(_NEW_TEXTURE
),
561 DEFINE_BIT(_NEW_TRANSFORM
),
562 DEFINE_BIT(_NEW_VIEWPORT
),
563 DEFINE_BIT(_NEW_ARRAY
),
564 DEFINE_BIT(_NEW_RENDERMODE
),
565 DEFINE_BIT(_NEW_BUFFERS
),
566 DEFINE_BIT(_NEW_CURRENT_ATTRIB
),
567 DEFINE_BIT(_NEW_MULTISAMPLE
),
568 DEFINE_BIT(_NEW_TRACK_MATRIX
),
569 DEFINE_BIT(_NEW_PROGRAM
),
570 DEFINE_BIT(_NEW_PROGRAM_CONSTANTS
),
571 DEFINE_BIT(_NEW_BUFFER_OBJECT
),
572 DEFINE_BIT(_NEW_FRAG_CLAMP
),
573 /* Avoid sign extension problems. */
574 {(unsigned) _NEW_VARYING_VP_INPUTS
, "_NEW_VARYING_VP_INPUTS", 0},
578 static struct dirty_bit_map brw_bits
[] = {
579 DEFINE_BIT(BRW_NEW_FS_PROG_DATA
),
580 DEFINE_BIT(BRW_NEW_BLORP_BLIT_PROG_DATA
),
581 DEFINE_BIT(BRW_NEW_SF_PROG_DATA
),
582 DEFINE_BIT(BRW_NEW_VS_PROG_DATA
),
583 DEFINE_BIT(BRW_NEW_FF_GS_PROG_DATA
),
584 DEFINE_BIT(BRW_NEW_GS_PROG_DATA
),
585 DEFINE_BIT(BRW_NEW_TCS_PROG_DATA
),
586 DEFINE_BIT(BRW_NEW_TES_PROG_DATA
),
587 DEFINE_BIT(BRW_NEW_CLIP_PROG_DATA
),
588 DEFINE_BIT(BRW_NEW_CS_PROG_DATA
),
589 DEFINE_BIT(BRW_NEW_URB_FENCE
),
590 DEFINE_BIT(BRW_NEW_FRAGMENT_PROGRAM
),
591 DEFINE_BIT(BRW_NEW_GEOMETRY_PROGRAM
),
592 DEFINE_BIT(BRW_NEW_TESS_EVAL_PROGRAM
),
593 DEFINE_BIT(BRW_NEW_TESS_CTRL_PROGRAM
),
594 DEFINE_BIT(BRW_NEW_VERTEX_PROGRAM
),
595 DEFINE_BIT(BRW_NEW_CURBE_OFFSETS
),
596 DEFINE_BIT(BRW_NEW_REDUCED_PRIMITIVE
),
597 DEFINE_BIT(BRW_NEW_PATCH_PRIMITIVE
),
598 DEFINE_BIT(BRW_NEW_PRIMITIVE
),
599 DEFINE_BIT(BRW_NEW_CONTEXT
),
600 DEFINE_BIT(BRW_NEW_PSP
),
601 DEFINE_BIT(BRW_NEW_SURFACES
),
602 DEFINE_BIT(BRW_NEW_BINDING_TABLE_POINTERS
),
603 DEFINE_BIT(BRW_NEW_INDICES
),
604 DEFINE_BIT(BRW_NEW_VERTICES
),
605 DEFINE_BIT(BRW_NEW_BATCH
),
606 DEFINE_BIT(BRW_NEW_INDEX_BUFFER
),
607 DEFINE_BIT(BRW_NEW_VS_CONSTBUF
),
608 DEFINE_BIT(BRW_NEW_GS_CONSTBUF
),
609 DEFINE_BIT(BRW_NEW_PROGRAM_CACHE
),
610 DEFINE_BIT(BRW_NEW_STATE_BASE_ADDRESS
),
611 DEFINE_BIT(BRW_NEW_VUE_MAP_GEOM_OUT
),
612 DEFINE_BIT(BRW_NEW_TRANSFORM_FEEDBACK
),
613 DEFINE_BIT(BRW_NEW_RASTERIZER_DISCARD
),
614 DEFINE_BIT(BRW_NEW_STATS_WM
),
615 DEFINE_BIT(BRW_NEW_UNIFORM_BUFFER
),
616 DEFINE_BIT(BRW_NEW_ATOMIC_BUFFER
),
617 DEFINE_BIT(BRW_NEW_IMAGE_UNITS
),
618 DEFINE_BIT(BRW_NEW_META_IN_PROGRESS
),
619 DEFINE_BIT(BRW_NEW_INTERPOLATION_MAP
),
620 DEFINE_BIT(BRW_NEW_PUSH_CONSTANT_ALLOCATION
),
621 DEFINE_BIT(BRW_NEW_NUM_SAMPLES
),
622 DEFINE_BIT(BRW_NEW_TEXTURE_BUFFER
),
623 DEFINE_BIT(BRW_NEW_GEN4_UNIT_STATE
),
624 DEFINE_BIT(BRW_NEW_CC_VP
),
625 DEFINE_BIT(BRW_NEW_SF_VP
),
626 DEFINE_BIT(BRW_NEW_CLIP_VP
),
627 DEFINE_BIT(BRW_NEW_SAMPLER_STATE_TABLE
),
628 DEFINE_BIT(BRW_NEW_VS_ATTRIB_WORKAROUNDS
),
629 DEFINE_BIT(BRW_NEW_COMPUTE_PROGRAM
),
630 DEFINE_BIT(BRW_NEW_CS_WORK_GROUPS
),
635 brw_update_dirty_count(struct dirty_bit_map
*bit_map
, uint64_t bits
)
637 for (int i
= 0; bit_map
[i
].bit
!= 0; i
++) {
638 if (bit_map
[i
].bit
& bits
)
644 brw_print_dirty_count(struct dirty_bit_map
*bit_map
)
646 for (int i
= 0; bit_map
[i
].bit
!= 0; i
++) {
647 if (bit_map
[i
].count
> 1) {
648 fprintf(stderr
, "0x%016lx: %12d (%s)\n",
649 bit_map
[i
].bit
, bit_map
[i
].count
, bit_map
[i
].name
);
655 brw_upload_programs(struct brw_context
*brw
,
656 enum brw_pipeline pipeline
)
658 if (pipeline
== BRW_RENDER_PIPELINE
) {
659 brw_upload_vs_prog(brw
);
662 brw_upload_ff_gs_prog(brw
);
664 brw_upload_gs_prog(brw
);
666 /* Update the VUE map for data exiting the GS stage of the pipeline.
667 * This comes from the last enabled shader stage.
669 GLbitfield64 old_slots
= brw
->vue_map_geom_out
.slots_valid
;
670 bool old_separate
= brw
->vue_map_geom_out
.separate
;
671 if (brw
->geometry_program
)
672 brw
->vue_map_geom_out
= brw
->gs
.prog_data
->base
.vue_map
;
674 brw
->vue_map_geom_out
= brw
->vs
.prog_data
->base
.vue_map
;
676 /* If the layout has changed, signal BRW_NEW_VUE_MAP_GEOM_OUT. */
677 if (old_slots
!= brw
->vue_map_geom_out
.slots_valid
||
678 old_separate
!= brw
->vue_map_geom_out
.separate
)
679 brw
->ctx
.NewDriverState
|= BRW_NEW_VUE_MAP_GEOM_OUT
;
681 brw_upload_wm_prog(brw
);
682 } else if (pipeline
== BRW_COMPUTE_PIPELINE
) {
683 brw_upload_cs_prog(brw
);
688 merge_ctx_state(struct brw_context
*brw
,
689 struct brw_state_flags
*state
)
691 state
->mesa
|= brw
->NewGLState
;
692 state
->brw
|= brw
->ctx
.NewDriverState
;
696 check_and_emit_atom(struct brw_context
*brw
,
697 struct brw_state_flags
*state
,
698 const struct brw_tracked_state
*atom
)
700 if (check_state(state
, &atom
->dirty
)) {
702 merge_ctx_state(brw
, state
);
707 brw_upload_pipeline_state(struct brw_context
*brw
,
708 enum brw_pipeline pipeline
)
710 struct gl_context
*ctx
= &brw
->ctx
;
712 static int dirty_count
= 0;
713 struct brw_state_flags state
= brw
->state
.pipelines
[pipeline
];
714 unsigned int fb_samples
= _mesa_geometric_samples(ctx
->DrawBuffer
);
716 brw_select_pipeline(brw
, pipeline
);
719 /* Always re-emit all state. */
720 brw
->NewGLState
= ~0;
721 ctx
->NewDriverState
= ~0ull;
724 if (pipeline
== BRW_RENDER_PIPELINE
) {
725 if (brw
->fragment_program
!= ctx
->FragmentProgram
._Current
) {
726 brw
->fragment_program
= ctx
->FragmentProgram
._Current
;
727 brw
->ctx
.NewDriverState
|= BRW_NEW_FRAGMENT_PROGRAM
;
730 if (brw
->tess_eval_program
!= ctx
->TessEvalProgram
._Current
) {
731 brw
->tess_eval_program
= ctx
->TessEvalProgram
._Current
;
732 brw
->ctx
.NewDriverState
|= BRW_NEW_TESS_EVAL_PROGRAM
;
735 if (brw
->tess_ctrl_program
!= ctx
->TessCtrlProgram
._Current
) {
736 brw
->tess_ctrl_program
= ctx
->TessCtrlProgram
._Current
;
737 brw
->ctx
.NewDriverState
|= BRW_NEW_TESS_CTRL_PROGRAM
;
740 if (brw
->geometry_program
!= ctx
->GeometryProgram
._Current
) {
741 brw
->geometry_program
= ctx
->GeometryProgram
._Current
;
742 brw
->ctx
.NewDriverState
|= BRW_NEW_GEOMETRY_PROGRAM
;
745 if (brw
->vertex_program
!= ctx
->VertexProgram
._Current
) {
746 brw
->vertex_program
= ctx
->VertexProgram
._Current
;
747 brw
->ctx
.NewDriverState
|= BRW_NEW_VERTEX_PROGRAM
;
751 if (brw
->compute_program
!= ctx
->ComputeProgram
._Current
) {
752 brw
->compute_program
= ctx
->ComputeProgram
._Current
;
753 brw
->ctx
.NewDriverState
|= BRW_NEW_COMPUTE_PROGRAM
;
756 if (brw
->meta_in_progress
!= _mesa_meta_in_progress(ctx
)) {
757 brw
->meta_in_progress
= _mesa_meta_in_progress(ctx
);
758 brw
->ctx
.NewDriverState
|= BRW_NEW_META_IN_PROGRESS
;
761 if (brw
->num_samples
!= fb_samples
) {
762 brw
->num_samples
= fb_samples
;
763 brw
->ctx
.NewDriverState
|= BRW_NEW_NUM_SAMPLES
;
766 /* Exit early if no state is flagged as dirty */
767 merge_ctx_state(brw
, &state
);
768 if ((state
.mesa
| state
.brw
) == 0)
771 /* Emit Sandybridge workaround flushes on every primitive, for safety. */
773 brw_emit_post_sync_nonzero_flush(brw
);
775 brw_upload_programs(brw
, pipeline
);
776 merge_ctx_state(brw
, &state
);
778 const struct brw_tracked_state
*atoms
=
779 brw_get_pipeline_atoms(brw
, pipeline
);
780 const int num_atoms
= brw
->num_atoms
[pipeline
];
782 if (unlikely(INTEL_DEBUG
)) {
783 /* Debug version which enforces various sanity checks on the
784 * state flags which are generated and checked to help ensure
785 * state atoms are ordered correctly in the list.
787 struct brw_state_flags examined
, prev
;
788 memset(&examined
, 0, sizeof(examined
));
791 for (i
= 0; i
< num_atoms
; i
++) {
792 const struct brw_tracked_state
*atom
= &atoms
[i
];
793 struct brw_state_flags generated
;
795 check_and_emit_atom(brw
, &state
, atom
);
797 accumulate_state(&examined
, &atom
->dirty
);
799 /* generated = (prev ^ state)
800 * if (examined & generated)
803 xor_states(&generated
, &prev
, &state
);
804 assert(!check_state(&examined
, &generated
));
809 for (i
= 0; i
< num_atoms
; i
++) {
810 const struct brw_tracked_state
*atom
= &atoms
[i
];
812 check_and_emit_atom(brw
, &state
, atom
);
816 if (unlikely(INTEL_DEBUG
& DEBUG_STATE
)) {
817 STATIC_ASSERT(ARRAY_SIZE(brw_bits
) == BRW_NUM_STATE_BITS
+ 1);
819 brw_update_dirty_count(mesa_bits
, state
.mesa
);
820 brw_update_dirty_count(brw_bits
, state
.brw
);
821 if (dirty_count
++ % 1000 == 0) {
822 brw_print_dirty_count(mesa_bits
);
823 brw_print_dirty_count(brw_bits
);
824 fprintf(stderr
, "\n");
829 /***********************************************************************
832 void brw_upload_render_state(struct brw_context
*brw
)
834 brw_upload_pipeline_state(brw
, BRW_RENDER_PIPELINE
);
838 brw_pipeline_state_finished(struct brw_context
*brw
,
839 enum brw_pipeline pipeline
)
841 /* Save all dirty state into the other pipelines */
842 for (unsigned i
= 0; i
< BRW_NUM_PIPELINES
; i
++) {
844 brw
->state
.pipelines
[i
].mesa
|= brw
->NewGLState
;
845 brw
->state
.pipelines
[i
].brw
|= brw
->ctx
.NewDriverState
;
847 memset(&brw
->state
.pipelines
[i
], 0, sizeof(struct brw_state_flags
));
852 brw
->ctx
.NewDriverState
= 0ull;
856 * Clear dirty bits to account for the fact that the state emitted by
857 * brw_upload_render_state() has been committed to the hardware. This is a
858 * separate call from brw_upload_render_state() because it's possible that
859 * after the call to brw_upload_render_state(), we will discover that we've
860 * run out of aperture space, and need to rewind the batch buffer to the state
861 * it had before the brw_upload_render_state() call.
864 brw_render_state_finished(struct brw_context
*brw
)
866 brw_pipeline_state_finished(brw
, BRW_RENDER_PIPELINE
);
870 brw_upload_compute_state(struct brw_context
*brw
)
872 brw_upload_pipeline_state(brw
, BRW_COMPUTE_PIPELINE
);
876 brw_compute_state_finished(struct brw_context
*brw
)
878 brw_pipeline_state_finished(brw
, BRW_COMPUTE_PIPELINE
);