1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
|
/*
* Copyright © 2014-2017 Broadcom
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice (including the next
* paragraph) shall be included in all copies or substantial portions of the
* Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
* IN THE SOFTWARE.
*/
#include "util/u_blitter.h"
#include "util/u_prim.h"
#include "util/u_format.h"
#include "util/u_pack_color.h"
#include "util/u_prim_restart.h"
#include "util/u_upload_mgr.h"
#include "indices/u_primconvert.h"
#include "v3d_context.h"
#include "v3d_resource.h"
#include "v3d_cl.h"
#include "broadcom/compiler/v3d_compiler.h"
#include "broadcom/common/v3d_macros.h"
#include "broadcom/cle/v3dx_pack.h"
/**
* Does the initial bining command list setup for drawing to a given FBO.
*/
static void
v3d_start_draw(struct v3d_context *v3d)
{
struct v3d_job *job = v3d->job;
if (job->needs_flush)
return;
/* Get space to emit our BCL state, using a branch to jump to a new BO
* if necessary.
*/
v3d_cl_ensure_space_with_branch(&job->bcl, 256 /* XXX */);
job->submit.bcl_start = job->bcl.bo->offset;
v3d_job_add_bo(job, job->bcl.bo);
/* The PTB will request the tile alloc initial size per tile at start
* of tile binning.
*/
uint32_t tile_alloc_size = (job->draw_tiles_x *
job->draw_tiles_y) * 64;
/* The PTB allocates in aligned 4k chunks after the initial setup. */
tile_alloc_size = align(tile_alloc_size, 4096);
/* Include the first two chunk allocations that the PTB does so that
* we definitely clear the OOM condition before triggering one (the HW
* won't trigger OOM during the first allocations).
*/
tile_alloc_size += 8192;
/* For performance, allocate some extra initial memory after the PTB's
* minimal allocations, so that we hopefully don't have to block the
* GPU on the kernel handling an OOM signal.
*/
tile_alloc_size += 512 * 1024;
job->tile_alloc = v3d_bo_alloc(v3d->screen, tile_alloc_size,
"tile_alloc");
uint32_t tsda_per_tile_size = v3d->screen->devinfo.ver >= 40 ? 256 : 64;
job->tile_state = v3d_bo_alloc(v3d->screen,
job->draw_tiles_y *
job->draw_tiles_x *
tsda_per_tile_size,
"TSDA");
#if V3D_VERSION >= 40
cl_emit(&job->bcl, TILE_BINNING_MODE_CFG, config) {
config.width_in_pixels = v3d->framebuffer.width;
config.height_in_pixels = v3d->framebuffer.height;
config.number_of_render_targets =
MAX2(v3d->framebuffer.nr_cbufs, 1);
config.multisample_mode_4x = job->msaa;
config.maximum_bpp_of_all_render_targets = job->internal_bpp;
}
#else /* V3D_VERSION < 40 */
/* "Binning mode lists start with a Tile Binning Mode Configuration
* item (120)"
*
* Part1 signals the end of binning config setup.
*/
cl_emit(&job->bcl, TILE_BINNING_MODE_CFG_PART2, config) {
config.tile_allocation_memory_address =
cl_address(job->tile_alloc, 0);
config.tile_allocation_memory_size = job->tile_alloc->size;
}
cl_emit(&job->bcl, TILE_BINNING_MODE_CFG_PART1, config) {
config.tile_state_data_array_base_address =
cl_address(job->tile_state, 0);
config.width_in_tiles = job->draw_tiles_x;
config.height_in_tiles = job->draw_tiles_y;
/* Must be >= 1 */
config.number_of_render_targets =
MAX2(v3d->framebuffer.nr_cbufs, 1);
config.multisample_mode_4x = job->msaa;
config.maximum_bpp_of_all_render_targets = job->internal_bpp;
}
#endif /* V3D_VERSION < 40 */
/* There's definitely nothing in the VCD cache we want. */
cl_emit(&job->bcl, FLUSH_VCD_CACHE, bin);
/* Disable any leftover OQ state from another job. */
cl_emit(&job->bcl, OCCLUSION_QUERY_COUNTER, counter);
/* "Binning mode lists must have a Start Tile Binning item (6) after
* any prefix state data before the binning list proper starts."
*/
cl_emit(&job->bcl, START_TILE_BINNING, bin);
job->needs_flush = true;
job->draw_width = v3d->framebuffer.width;
job->draw_height = v3d->framebuffer.height;
}
static void
v3d_predraw_check_stage_inputs(struct pipe_context *pctx,
enum pipe_shader_type s)
{
struct v3d_context *v3d = v3d_context(pctx);
/* Flush writes to textures we're sampling. */
for (int i = 0; i < v3d->tex[s].num_textures; i++) {
struct pipe_sampler_view *pview = v3d->tex[s].textures[i];
if (!pview)
continue;
struct v3d_sampler_view *view = v3d_sampler_view(pview);
if (view->texture != view->base.texture &&
view->base.format != PIPE_FORMAT_X32_S8X24_UINT)
v3d_update_shadow_texture(pctx, &view->base);
v3d_flush_jobs_writing_resource(v3d, view->texture, false);
}
/* Flush writes to UBOs. */
foreach_bit(i, v3d->constbuf[s].enabled_mask) {
struct pipe_constant_buffer *cb = &v3d->constbuf[s].cb[i];
if (cb->buffer)
v3d_flush_jobs_writing_resource(v3d, cb->buffer, false);
}
/* Flush writes to our image views */
foreach_bit(i, v3d->shaderimg[s].enabled_mask) {
struct v3d_image_view *view = &v3d->shaderimg[s].si[i];
v3d_flush_jobs_writing_resource(v3d, view->base.resource,
false);
}
/* Flush writes to our vertex buffers (i.e. from transform feedback) */
if (s == PIPE_SHADER_VERTEX) {
foreach_bit(i, v3d->vertexbuf.enabled_mask) {
struct pipe_vertex_buffer *vb = &v3d->vertexbuf.vb[i];
v3d_flush_jobs_writing_resource(v3d, vb->buffer.resource,
false);
}
}
}
static void
v3d_predraw_check_outputs(struct pipe_context *pctx)
{
struct v3d_context *v3d = v3d_context(pctx);
/* Flush jobs reading from TF buffers that we are about to write. */
if (v3d_transform_feedback_enabled(v3d)) {
struct v3d_streamout_stateobj *so = &v3d->streamout;
for (int i = 0; i < so->num_targets; i++) {
if (!so->targets[i])
continue;
const struct pipe_stream_output_target *target =
so->targets[i];
v3d_flush_jobs_reading_resource(v3d, target->buffer);
}
}
}
/**
* Checks if the state for the current draw reads a particular resource in
* in the given shader stage.
*/
static bool
v3d_state_reads_resource(struct v3d_context *v3d,
struct pipe_resource *prsc,
enum pipe_shader_type s)
{
struct v3d_resource *rsc = v3d_resource(prsc);
/* Vertex buffers */
if (s == PIPE_SHADER_VERTEX) {
foreach_bit(i, v3d->vertexbuf.enabled_mask) {
struct pipe_vertex_buffer *vb = &v3d->vertexbuf.vb[i];
if (!vb->buffer.resource)
continue;
struct v3d_resource *vb_rsc =
v3d_resource(vb->buffer.resource);
if (rsc->bo == vb_rsc->bo)
return true;
}
}
/* Constant buffers */
foreach_bit(i, v3d->constbuf[s].enabled_mask) {
struct pipe_constant_buffer *cb = &v3d->constbuf[s].cb[i];
if (!cb->buffer)
continue;
struct v3d_resource *cb_rsc = v3d_resource(cb->buffer);
if (rsc->bo == cb_rsc->bo)
return true;
}
/* Shader storage buffers */
foreach_bit(i, v3d->ssbo[s].enabled_mask) {
struct pipe_shader_buffer *sb = &v3d->ssbo[s].sb[i];
if (!sb->buffer)
continue;
struct v3d_resource *sb_rsc = v3d_resource(sb->buffer);
if (rsc->bo == sb_rsc->bo)
return true;
}
/* Textures */
for (int i = 0; i < v3d->tex[s].num_textures; i++) {
struct pipe_sampler_view *pview = v3d->tex[s].textures[i];
if (!pview)
continue;
struct v3d_sampler_view *view = v3d_sampler_view(pview);
struct v3d_resource *v_rsc = v3d_resource(view->texture);
if (rsc->bo == v_rsc->bo)
return true;
}
return false;
}
static void
v3d_emit_wait_for_tf(struct v3d_job *job)
{
/* XXX: we might be able to skip this in some cases, for now we
* always emit it.
*/
cl_emit(&job->bcl, FLUSH_TRANSFORM_FEEDBACK_DATA, flush);
cl_emit(&job->bcl, WAIT_FOR_TRANSFORM_FEEDBACK, wait) {
/* XXX: Wait for all outstanding writes... maybe we can do
* better in some cases.
*/
wait.block_count = 255;
}
/* We have just flushed all our outstanding TF work in this job so make
* sure we don't emit TF flushes again for any of it again.
*/
_mesa_set_clear(job->tf_write_prscs, NULL);
}
static void
v3d_emit_wait_for_tf_if_needed(struct v3d_context *v3d, struct v3d_job *job)
{
if (!job->tf_enabled)
return;
set_foreach(job->tf_write_prscs, entry) {
struct pipe_resource *prsc = (struct pipe_resource *)entry->key;
for (int s = 0; s < PIPE_SHADER_COMPUTE; s++) {
/* Fragment shaders can only start executing after all
* binning (and thus TF) is complete.
*
* XXX: For VS/GS/TES, if the binning shader does not
* read the resource then we could also avoid emitting
* the wait.
*/
if (s == PIPE_SHADER_FRAGMENT)
continue;
if (v3d_state_reads_resource(v3d, prsc, s)) {
v3d_emit_wait_for_tf(job);
return;
}
}
}
}
static void
v3d_emit_gl_shader_state(struct v3d_context *v3d,
const struct pipe_draw_info *info)
{
struct v3d_job *job = v3d->job;
/* VC5_DIRTY_VTXSTATE */
struct v3d_vertex_stateobj *vtx = v3d->vtx;
/* VC5_DIRTY_VTXBUF */
struct v3d_vertexbuf_stateobj *vertexbuf = &v3d->vertexbuf;
/* Upload the uniforms to the indirect CL first */
struct v3d_cl_reloc fs_uniforms =
v3d_write_uniforms(v3d, v3d->prog.fs,
PIPE_SHADER_FRAGMENT);
struct v3d_cl_reloc vs_uniforms =
v3d_write_uniforms(v3d, v3d->prog.vs,
PIPE_SHADER_VERTEX);
struct v3d_cl_reloc cs_uniforms =
v3d_write_uniforms(v3d, v3d->prog.cs,
PIPE_SHADER_VERTEX);
/* See GFXH-930 workaround below */
uint32_t num_elements_to_emit = MAX2(vtx->num_elements, 1);
uint32_t shader_rec_offset =
v3d_cl_ensure_space(&job->indirect,
cl_packet_length(GL_SHADER_STATE_RECORD) +
num_elements_to_emit *
cl_packet_length(GL_SHADER_STATE_ATTRIBUTE_RECORD),
32);
/* XXX perf: We should move most of the SHADER_STATE_RECORD setup to
* compile time, so that we mostly just have to OR the VS and FS
* records together at draw time.
*/
cl_emit(&job->indirect, GL_SHADER_STATE_RECORD, shader) {
shader.enable_clipping = true;
/* VC5_DIRTY_PRIM_MODE | VC5_DIRTY_RASTERIZER */
shader.point_size_in_shaded_vertex_data =
(info->mode == PIPE_PRIM_POINTS &&
v3d->rasterizer->base.point_size_per_vertex);
/* Must be set if the shader modifies Z, discards, or modifies
* the sample mask. For any of these cases, the fragment
* shader needs to write the Z value (even just discards).
*/
shader.fragment_shader_does_z_writes =
v3d->prog.fs->prog_data.fs->writes_z;
/* Set if the EZ test must be disabled (due to shader side
* effects and the early_z flag not being present in the
* shader).
*/
shader.turn_off_early_z_test =
v3d->prog.fs->prog_data.fs->disable_ez;
shader.fragment_shader_uses_real_pixel_centre_w_in_addition_to_centroid_w2 =
v3d->prog.fs->prog_data.fs->uses_center_w;
#if V3D_VERSION >= 40
shader.do_scoreboard_wait_on_first_thread_switch =
v3d->prog.fs->prog_data.fs->lock_scoreboard_on_first_thrsw;
shader.disable_implicit_point_line_varyings =
!v3d->prog.fs->prog_data.fs->uses_implicit_point_line_varyings;
#endif
shader.number_of_varyings_in_fragment_shader =
v3d->prog.fs->prog_data.fs->num_inputs;
shader.coordinate_shader_propagate_nans = true;
shader.vertex_shader_propagate_nans = true;
shader.fragment_shader_propagate_nans = true;
shader.coordinate_shader_code_address =
cl_address(v3d_resource(v3d->prog.cs->resource)->bo,
v3d->prog.cs->offset);
shader.vertex_shader_code_address =
cl_address(v3d_resource(v3d->prog.vs->resource)->bo,
v3d->prog.vs->offset);
shader.fragment_shader_code_address =
cl_address(v3d_resource(v3d->prog.fs->resource)->bo,
v3d->prog.fs->offset);
/* XXX: Use combined input/output size flag in the common
* case.
*/
shader.coordinate_shader_has_separate_input_and_output_vpm_blocks =
v3d->prog.cs->prog_data.vs->separate_segments;
shader.vertex_shader_has_separate_input_and_output_vpm_blocks =
v3d->prog.vs->prog_data.vs->separate_segments;
shader.coordinate_shader_input_vpm_segment_size =
v3d->prog.cs->prog_data.vs->separate_segments ?
v3d->prog.cs->prog_data.vs->vpm_input_size : 1;
shader.vertex_shader_input_vpm_segment_size =
v3d->prog.vs->prog_data.vs->separate_segments ?
v3d->prog.vs->prog_data.vs->vpm_input_size : 1;
shader.coordinate_shader_output_vpm_segment_size =
v3d->prog.cs->prog_data.vs->vpm_output_size;
shader.vertex_shader_output_vpm_segment_size =
v3d->prog.vs->prog_data.vs->vpm_output_size;
shader.coordinate_shader_uniforms_address = cs_uniforms;
shader.vertex_shader_uniforms_address = vs_uniforms;
shader.fragment_shader_uniforms_address = fs_uniforms;
#if V3D_VERSION >= 41
shader.min_coord_shader_input_segments_required_in_play = 1;
shader.min_vertex_shader_input_segments_required_in_play = 1;
shader.coordinate_shader_4_way_threadable =
v3d->prog.cs->prog_data.vs->base.threads == 4;
shader.vertex_shader_4_way_threadable =
v3d->prog.vs->prog_data.vs->base.threads == 4;
shader.fragment_shader_4_way_threadable =
v3d->prog.fs->prog_data.fs->base.threads == 4;
shader.coordinate_shader_start_in_final_thread_section =
v3d->prog.cs->prog_data.vs->base.single_seg;
shader.vertex_shader_start_in_final_thread_section =
v3d->prog.vs->prog_data.vs->base.single_seg;
shader.fragment_shader_start_in_final_thread_section =
v3d->prog.fs->prog_data.fs->base.single_seg;
#else
shader.coordinate_shader_4_way_threadable =
v3d->prog.cs->prog_data.vs->base.threads == 4;
shader.coordinate_shader_2_way_threadable =
v3d->prog.cs->prog_data.vs->base.threads == 2;
shader.vertex_shader_4_way_threadable =
v3d->prog.vs->prog_data.vs->base.threads == 4;
shader.vertex_shader_2_way_threadable =
v3d->prog.vs->prog_data.vs->base.threads == 2;
shader.fragment_shader_4_way_threadable =
v3d->prog.fs->prog_data.fs->base.threads == 4;
shader.fragment_shader_2_way_threadable =
v3d->prog.fs->prog_data.fs->base.threads == 2;
#endif
shader.vertex_id_read_by_coordinate_shader =
v3d->prog.cs->prog_data.vs->uses_vid;
shader.instance_id_read_by_coordinate_shader =
v3d->prog.cs->prog_data.vs->uses_iid;
shader.vertex_id_read_by_vertex_shader =
v3d->prog.vs->prog_data.vs->uses_vid;
shader.instance_id_read_by_vertex_shader =
v3d->prog.vs->prog_data.vs->uses_iid;
shader.address_of_default_attribute_values =
cl_address(v3d_resource(vtx->defaults)->bo,
vtx->defaults_offset);
}
bool cs_loaded_any = false;
for (int i = 0; i < vtx->num_elements; i++) {
struct pipe_vertex_element *elem = &vtx->pipe[i];
struct pipe_vertex_buffer *vb =
&vertexbuf->vb[elem->vertex_buffer_index];
struct v3d_resource *rsc = v3d_resource(vb->buffer.resource);
const uint32_t size =
cl_packet_length(GL_SHADER_STATE_ATTRIBUTE_RECORD);
cl_emit_with_prepacked(&job->indirect,
GL_SHADER_STATE_ATTRIBUTE_RECORD,
&vtx->attrs[i * size], attr) {
attr.stride = vb->stride;
attr.address = cl_address(rsc->bo,
vb->buffer_offset +
elem->src_offset);
attr.number_of_values_read_by_coordinate_shader =
v3d->prog.cs->prog_data.vs->vattr_sizes[i];
attr.number_of_values_read_by_vertex_shader =
v3d->prog.vs->prog_data.vs->vattr_sizes[i];
/* GFXH-930: At least one attribute must be enabled
* and read by CS and VS. If we have attributes being
* consumed by the VS but not the CS, then set up a
* dummy load of the last attribute into the CS's VPM
* inputs. (Since CS is just dead-code-elimination
* compared to VS, we can't have CS loading but not
* VS).
*/
if (v3d->prog.cs->prog_data.vs->vattr_sizes[i])
cs_loaded_any = true;
if (i == vtx->num_elements - 1 && !cs_loaded_any) {
attr.number_of_values_read_by_coordinate_shader = 1;
}
#if V3D_VERSION >= 41
attr.maximum_index = 0xffffff;
#endif
}
STATIC_ASSERT(sizeof(vtx->attrs) >= V3D_MAX_VS_INPUTS / 4 * size);
}
if (vtx->num_elements == 0) {
/* GFXH-930: At least one attribute must be enabled and read
* by CS and VS. If we have no attributes being consumed by
* the shader, set up a dummy to be loaded into the VPM.
*/
cl_emit(&job->indirect, GL_SHADER_STATE_ATTRIBUTE_RECORD, attr) {
/* Valid address of data whose value will be unused. */
attr.address = cl_address(job->indirect.bo, 0);
attr.type = ATTRIBUTE_FLOAT;
attr.stride = 0;
attr.vec_size = 1;
attr.number_of_values_read_by_coordinate_shader = 1;
attr.number_of_values_read_by_vertex_shader = 1;
}
}
cl_emit(&job->bcl, VCM_CACHE_SIZE, vcm) {
vcm.number_of_16_vertex_batches_for_binning =
v3d->prog.cs->prog_data.vs->vcm_cache_size;
vcm.number_of_16_vertex_batches_for_rendering =
v3d->prog.vs->prog_data.vs->vcm_cache_size;
}
cl_emit(&job->bcl, GL_SHADER_STATE, state) {
state.address = cl_address(job->indirect.bo, shader_rec_offset);
state.number_of_attribute_arrays = num_elements_to_emit;
}
v3d_bo_unreference(&cs_uniforms.bo);
v3d_bo_unreference(&vs_uniforms.bo);
v3d_bo_unreference(&fs_uniforms.bo);
job->shader_rec_count++;
}
/**
* Computes the various transform feedback statistics, since they can't be
* recorded by CL packets.
*/
static void
v3d_tf_statistics_record(struct v3d_context *v3d,
const struct pipe_draw_info *info)
{
if (!v3d->active_queries)
return;
uint32_t prims = u_prims_for_vertices(info->mode, info->count);
v3d->prims_generated += prims;
if (v3d->streamout.num_targets <= 0)
return;
/* XXX: Only count if we didn't overflow. */
v3d->tf_prims_generated += prims;
}
static void
v3d_update_job_ez(struct v3d_context *v3d, struct v3d_job *job)
{
switch (v3d->zsa->ez_state) {
case VC5_EZ_UNDECIDED:
/* If the Z/S state didn't pick a direction but didn't
* disable, then go along with the current EZ state. This
* allows EZ optimization for Z func == EQUAL or NEVER.
*/
break;
case VC5_EZ_LT_LE:
case VC5_EZ_GT_GE:
/* If the Z/S state picked a direction, then it needs to match
* the current direction if we've decided on one.
*/
if (job->ez_state == VC5_EZ_UNDECIDED)
job->ez_state = v3d->zsa->ez_state;
else if (job->ez_state != v3d->zsa->ez_state)
job->ez_state = VC5_EZ_DISABLED;
break;
case VC5_EZ_DISABLED:
/* If the current Z/S state disables EZ because of a bad Z
* func or stencil operation, then we can't do any more EZ in
* this frame.
*/
job->ez_state = VC5_EZ_DISABLED;
break;
}
/* If the FS affects the Z of the pixels, then it may update against
* the chosen EZ direction (though we could use
* ARB_conservative_depth's hints to avoid this)
*/
if (v3d->prog.fs->prog_data.fs->writes_z) {
job->ez_state = VC5_EZ_DISABLED;
}
if (job->first_ez_state == VC5_EZ_UNDECIDED &&
(job->ez_state != VC5_EZ_DISABLED || job->draw_calls_queued == 0))
job->first_ez_state = job->ez_state;
}
static void
v3d_draw_vbo(struct pipe_context *pctx, const struct pipe_draw_info *info)
{
struct v3d_context *v3d = v3d_context(pctx);
if (!info->count_from_stream_output && !info->indirect &&
!info->primitive_restart &&
!u_trim_pipe_prim(info->mode, (unsigned*)&info->count))
return;
/* Fall back for weird desktop GL primitive restart values. */
if (info->primitive_restart &&
info->index_size) {
uint32_t mask = ~0;
switch (info->index_size) {
case 2:
mask = 0xffff;
break;
case 1:
mask = 0xff;
break;
}
if (info->restart_index != mask) {
util_draw_vbo_without_prim_restart(pctx, info);
return;
}
}
if (info->mode >= PIPE_PRIM_QUADS) {
util_primconvert_save_rasterizer_state(v3d->primconvert, &v3d->rasterizer->base);
util_primconvert_draw_vbo(v3d->primconvert, info);
perf_debug("Fallback conversion for %d %s vertices\n",
info->count, u_prim_name(info->mode));
return;
}
/* Before setting up the draw, flush anything writing to the resources
* that we read from or reading from resources we write to.
*/
for (int s = 0; s < PIPE_SHADER_COMPUTE; s++)
v3d_predraw_check_stage_inputs(pctx, s);
if (info->indirect) {
v3d_flush_jobs_writing_resource(v3d, info->indirect->buffer,
false);
}
v3d_predraw_check_outputs(pctx);
struct v3d_job *job = v3d_get_job_for_fbo(v3d);
/* If vertex texturing depends on the output of rendering, we need to
* ensure that that rendering is complete before we run a coordinate
* shader that depends on it.
*
* Given that doing that is unusual, for now we just block the binner
* on the last submitted render, rather than tracking the last
* rendering to each texture's BO.
*/
if (v3d->tex[PIPE_SHADER_VERTEX].num_textures || info->indirect) {
perf_debug("Blocking binner on last render "
"due to vertex texturing or indirect drawing.\n");
job->submit.in_sync_bcl = v3d->out_sync;
}
/* Mark SSBOs as being written. We don't actually know which ones are
* read vs written, so just assume the worst
*/
for (int s = 0; s < PIPE_SHADER_COMPUTE; s++) {
foreach_bit(i, v3d->ssbo[s].enabled_mask) {
v3d_job_add_write_resource(job,
v3d->ssbo[s].sb[i].buffer);
job->tmu_dirty_rcl = true;
}
foreach_bit(i, v3d->shaderimg[s].enabled_mask) {
v3d_job_add_write_resource(job,
v3d->shaderimg[s].si[i].base.resource);
job->tmu_dirty_rcl = true;
}
}
/* Get space to emit our draw call into the BCL, using a branch to
* jump to a new BO if necessary.
*/
v3d_cl_ensure_space_with_branch(&job->bcl, 256 /* XXX */);
if (v3d->prim_mode != info->mode) {
v3d->prim_mode = info->mode;
v3d->dirty |= VC5_DIRTY_PRIM_MODE;
}
v3d_start_draw(v3d);
v3d_update_compiled_shaders(v3d, info->mode);
v3d_update_job_ez(v3d, job);
/* If this job was writing to transform feedback buffers before this
* draw and we are reading from them here, then we need to wait for TF
* to complete before we emit this draw.
*
* Notice this check needs to happen before we emit state for the
* current draw call, where we update job->tf_enabled, so we can ensure
* that we only check TF writes for prior draws.
*/
v3d_emit_wait_for_tf_if_needed(v3d, job);
#if V3D_VERSION >= 41
v3d41_emit_state(pctx);
#else
v3d33_emit_state(pctx);
#endif
if (v3d->dirty & (VC5_DIRTY_VTXBUF |
VC5_DIRTY_VTXSTATE |
VC5_DIRTY_PRIM_MODE |
VC5_DIRTY_RASTERIZER |
VC5_DIRTY_COMPILED_CS |
VC5_DIRTY_COMPILED_VS |
VC5_DIRTY_COMPILED_FS |
v3d->prog.cs->uniform_dirty_bits |
v3d->prog.vs->uniform_dirty_bits |
v3d->prog.fs->uniform_dirty_bits)) {
v3d_emit_gl_shader_state(v3d, info);
}
v3d->dirty = 0;
/* The Base Vertex/Base Instance packet sets those values to nonzero
* for the next draw call only.
*/
if (info->index_bias || info->start_instance) {
cl_emit(&job->bcl, BASE_VERTEX_BASE_INSTANCE, base) {
base.base_instance = info->start_instance;
base.base_vertex = info->index_bias;
}
}
uint32_t prim_tf_enable = 0;
#if V3D_VERSION < 40
/* V3D 3.x: The HW only processes transform feedback on primitives
* with the flag set.
*/
if (v3d->streamout.num_targets)
prim_tf_enable = (V3D_PRIM_POINTS_TF - V3D_PRIM_POINTS);
#endif
v3d_tf_statistics_record(v3d, info);
/* Note that the primitive type fields match with OpenGL/gallium
* definitions, up to but not including QUADS.
*/
if (info->index_size) {
uint32_t index_size = info->index_size;
uint32_t offset = info->start * index_size;
struct pipe_resource *prsc;
if (info->has_user_indices) {
prsc = NULL;
u_upload_data(v3d->uploader, 0,
info->count * info->index_size, 4,
info->index.user,
&offset, &prsc);
} else {
prsc = info->index.resource;
}
struct v3d_resource *rsc = v3d_resource(prsc);
#if V3D_VERSION >= 40
cl_emit(&job->bcl, INDEX_BUFFER_SETUP, ib) {
ib.address = cl_address(rsc->bo, 0);
ib.size = rsc->bo->size;
}
#endif
if (info->indirect) {
cl_emit(&job->bcl, INDIRECT_INDEXED_INSTANCED_PRIM_LIST, prim) {
prim.index_type = ffs(info->index_size) - 1;
#if V3D_VERSION < 40
prim.address_of_indices_list =
cl_address(rsc->bo, offset);
#endif /* V3D_VERSION < 40 */
prim.mode = info->mode | prim_tf_enable;
prim.enable_primitive_restarts = info->primitive_restart;
prim.number_of_draw_indirect_indexed_records = info->indirect->draw_count;
prim.stride_in_multiples_of_4_bytes = info->indirect->stride >> 2;
prim.address = cl_address(v3d_resource(info->indirect->buffer)->bo,
info->indirect->offset);
}
} else if (info->instance_count > 1) {
cl_emit(&job->bcl, INDEXED_INSTANCED_PRIM_LIST, prim) {
prim.index_type = ffs(info->index_size) - 1;
#if V3D_VERSION >= 40
prim.index_offset = offset;
#else /* V3D_VERSION < 40 */
prim.maximum_index = (1u << 31) - 1; /* XXX */
prim.address_of_indices_list =
cl_address(rsc->bo, offset);
#endif /* V3D_VERSION < 40 */
prim.mode = info->mode | prim_tf_enable;
prim.enable_primitive_restarts = info->primitive_restart;
prim.number_of_instances = info->instance_count;
prim.instance_length = info->count;
}
} else {
cl_emit(&job->bcl, INDEXED_PRIM_LIST, prim) {
prim.index_type = ffs(info->index_size) - 1;
prim.length = info->count;
#if V3D_VERSION >= 40
prim.index_offset = offset;
#else /* V3D_VERSION < 40 */
prim.maximum_index = (1u << 31) - 1; /* XXX */
prim.address_of_indices_list =
cl_address(rsc->bo, offset);
#endif /* V3D_VERSION < 40 */
prim.mode = info->mode | prim_tf_enable;
prim.enable_primitive_restarts = info->primitive_restart;
}
}
job->draw_calls_queued++;
if (info->has_user_indices)
pipe_resource_reference(&prsc, NULL);
} else {
if (info->indirect) {
cl_emit(&job->bcl, INDIRECT_VERTEX_ARRAY_INSTANCED_PRIMS, prim) {
prim.mode = info->mode | prim_tf_enable;
prim.number_of_draw_indirect_array_records = info->indirect->draw_count;
prim.stride_in_multiples_of_4_bytes = info->indirect->stride >> 2;
prim.address = cl_address(v3d_resource(info->indirect->buffer)->bo,
info->indirect->offset);
}
} else if (info->instance_count > 1) {
cl_emit(&job->bcl, VERTEX_ARRAY_INSTANCED_PRIMS, prim) {
prim.mode = info->mode | prim_tf_enable;
prim.index_of_first_vertex = info->start;
prim.number_of_instances = info->instance_count;
prim.instance_length = info->count;
}
} else {
cl_emit(&job->bcl, VERTEX_ARRAY_PRIMS, prim) {
prim.mode = info->mode | prim_tf_enable;
prim.length = info->count;
prim.index_of_first_vertex = info->start;
}
}
}
/* A flush is required in between a TF draw and any following TF specs
* packet, or the GPU may hang. Just flush each time for now.
*/
if (v3d->streamout.num_targets)
cl_emit(&job->bcl, TRANSFORM_FEEDBACK_FLUSH_AND_COUNT, flush);
job->draw_calls_queued++;
/* Increment the TF offsets by how many verts we wrote. XXX: This
* needs some clamping to the buffer size.
*/
for (int i = 0; i < v3d->streamout.num_targets; i++)
v3d->streamout.offsets[i] += info->count;
if (v3d->zsa && job->zsbuf && v3d->zsa->base.depth.enabled) {
struct v3d_resource *rsc = v3d_resource(job->zsbuf->texture);
v3d_job_add_bo(job, rsc->bo);
job->load |= PIPE_CLEAR_DEPTH & ~job->clear;
if (v3d->zsa->base.depth.writemask)
job->store |= PIPE_CLEAR_DEPTH;
rsc->initialized_buffers = PIPE_CLEAR_DEPTH;
}
if (v3d->zsa && job->zsbuf && v3d->zsa->base.stencil[0].enabled) {
struct v3d_resource *rsc = v3d_resource(job->zsbuf->texture);
if (rsc->separate_stencil)
rsc = rsc->separate_stencil;
v3d_job_add_bo(job, rsc->bo);
job->load |= PIPE_CLEAR_STENCIL & ~job->clear;
if (v3d->zsa->base.stencil[0].writemask ||
v3d->zsa->base.stencil[1].writemask) {
job->store |= PIPE_CLEAR_STENCIL;
}
rsc->initialized_buffers |= PIPE_CLEAR_STENCIL;
}
for (int i = 0; i < V3D_MAX_DRAW_BUFFERS; i++) {
uint32_t bit = PIPE_CLEAR_COLOR0 << i;
int blend_rt = v3d->blend->base.independent_blend_enable ? i : 0;
if (job->store & bit || !job->cbufs[i])
continue;
struct v3d_resource *rsc = v3d_resource(job->cbufs[i]->texture);
job->load |= bit & ~job->clear;
if (v3d->blend->base.rt[blend_rt].colormask)
job->store |= bit;
v3d_job_add_bo(job, rsc->bo);
}
if (job->referenced_size > 768 * 1024 * 1024) {
perf_debug("Flushing job with %dkb to try to free up memory\n",
job->referenced_size / 1024);
v3d_flush(pctx);
}
if (V3D_DEBUG & V3D_DEBUG_ALWAYS_FLUSH)
v3d_flush(pctx);
}
/**
* Implements gallium's clear() hook (glClear()) by drawing a pair of triangles.
*/
static void
v3d_draw_clear(struct v3d_context *v3d,
unsigned buffers,
const union pipe_color_union *color,
double depth, unsigned stencil)
{
static const union pipe_color_union dummy_color = {};
/* The blitter util dereferences the color regardless, even though the
* gallium clear API may not pass one in when only Z/S are cleared.
*/
if (!color)
color = &dummy_color;
v3d_blitter_save(v3d);
util_blitter_clear(v3d->blitter,
v3d->framebuffer.width,
v3d->framebuffer.height,
util_framebuffer_get_num_layers(&v3d->framebuffer),
buffers, color, depth, stencil,
util_framebuffer_get_num_samples(&v3d->framebuffer) > 1);
}
/**
* Attempts to perform the GL clear by using the TLB's fast clear at the start
* of the frame.
*/
static unsigned
v3d_tlb_clear(struct v3d_job *job, unsigned buffers,
const union pipe_color_union *color,
double depth, unsigned stencil)
{
struct v3d_context *v3d = job->v3d;
if (job->draw_calls_queued) {
/* If anything in the CL has drawn using the buffer, then the
* TLB clear we're trying to add now would happen before that
* drawing.
*/
buffers &= ~(job->load | job->store);
}
/* GFXH-1461: If we were to emit a load of just depth or just stencil,
* then the clear for the other may get lost. We need to decide now
* if it would be possible to need to emit a load of just one after
* we've set up our TLB clears.
*/
if (buffers & PIPE_CLEAR_DEPTHSTENCIL &&
(buffers & PIPE_CLEAR_DEPTHSTENCIL) != PIPE_CLEAR_DEPTHSTENCIL &&
job->zsbuf &&
util_format_is_depth_and_stencil(job->zsbuf->texture->format)) {
buffers &= ~PIPE_CLEAR_DEPTHSTENCIL;
}
for (int i = 0; i < V3D_MAX_DRAW_BUFFERS; i++) {
uint32_t bit = PIPE_CLEAR_COLOR0 << i;
if (!(buffers & bit))
continue;
struct pipe_surface *psurf = v3d->framebuffer.cbufs[i];
struct v3d_surface *surf = v3d_surface(psurf);
struct v3d_resource *rsc = v3d_resource(psurf->texture);
union util_color uc;
uint32_t internal_size = 4 << surf->internal_bpp;
static union pipe_color_union swapped_color;
if (v3d->swap_color_rb & (1 << i)) {
swapped_color.f[0] = color->f[2];
swapped_color.f[1] = color->f[1];
swapped_color.f[2] = color->f[0];
swapped_color.f[3] = color->f[3];
color = &swapped_color;
}
switch (surf->internal_type) {
case V3D_INTERNAL_TYPE_8:
util_pack_color(color->f, PIPE_FORMAT_R8G8B8A8_UNORM,
&uc);
memcpy(job->clear_color[i], uc.ui, internal_size);
break;
case V3D_INTERNAL_TYPE_8I:
case V3D_INTERNAL_TYPE_8UI:
job->clear_color[i][0] = ((color->ui[0] & 0xff) |
(color->ui[1] & 0xff) << 8 |
(color->ui[2] & 0xff) << 16 |
(color->ui[3] & 0xff) << 24);
break;
case V3D_INTERNAL_TYPE_16F:
util_pack_color(color->f, PIPE_FORMAT_R16G16B16A16_FLOAT,
&uc);
memcpy(job->clear_color[i], uc.ui, internal_size);
break;
case V3D_INTERNAL_TYPE_16I:
case V3D_INTERNAL_TYPE_16UI:
job->clear_color[i][0] = ((color->ui[0] & 0xffff) |
color->ui[1] << 16);
job->clear_color[i][1] = ((color->ui[2] & 0xffff) |
color->ui[3] << 16);
break;
case V3D_INTERNAL_TYPE_32F:
case V3D_INTERNAL_TYPE_32I:
case V3D_INTERNAL_TYPE_32UI:
memcpy(job->clear_color[i], color->ui, internal_size);
break;
}
rsc->initialized_buffers |= bit;
}
unsigned zsclear = buffers & PIPE_CLEAR_DEPTHSTENCIL;
if (zsclear) {
struct v3d_resource *rsc =
v3d_resource(v3d->framebuffer.zsbuf->texture);
if (zsclear & PIPE_CLEAR_DEPTH)
job->clear_z = depth;
if (zsclear & PIPE_CLEAR_STENCIL)
job->clear_s = stencil;
rsc->initialized_buffers |= zsclear;
}
job->draw_min_x = 0;
job->draw_min_y = 0;
job->draw_max_x = v3d->framebuffer.width;
job->draw_max_y = v3d->framebuffer.height;
job->clear |= buffers;
job->store |= buffers;
v3d_start_draw(v3d);
return buffers;
}
static void
v3d_clear(struct pipe_context *pctx, unsigned buffers,
const union pipe_color_union *color, double depth, unsigned stencil)
{
struct v3d_context *v3d = v3d_context(pctx);
struct v3d_job *job = v3d_get_job_for_fbo(v3d);
buffers &= ~v3d_tlb_clear(job, buffers, color, depth, stencil);
if (buffers)
v3d_draw_clear(v3d, buffers, color, depth, stencil);
}
static void
v3d_clear_render_target(struct pipe_context *pctx, struct pipe_surface *ps,
const union pipe_color_union *color,
unsigned x, unsigned y, unsigned w, unsigned h,
bool render_condition_enabled)
{
fprintf(stderr, "unimpl: clear RT\n");
}
static void
v3d_clear_depth_stencil(struct pipe_context *pctx, struct pipe_surface *ps,
unsigned buffers, double depth, unsigned stencil,
unsigned x, unsigned y, unsigned w, unsigned h,
bool render_condition_enabled)
{
fprintf(stderr, "unimpl: clear DS\n");
}
void
v3dX(draw_init)(struct pipe_context *pctx)
{
pctx->draw_vbo = v3d_draw_vbo;
pctx->clear = v3d_clear;
pctx->clear_render_target = v3d_clear_render_target;
pctx->clear_depth_stencil = v3d_clear_depth_stencil;
}
|