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
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
|
/*
* Copyright © 2015 Intel Corporation
*
* 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 <assert.h>
#include <stdbool.h>
#include <string.h>
#include <unistd.h>
#include <fcntl.h>
#include "util/mesa-sha1.h"
#include "common/gen_l3_config.h"
#include "anv_private.h"
#include "compiler/brw_nir.h"
#include "anv_nir.h"
#include "spirv/nir_spirv.h"
#include "vk_util.h"
/* Needed for SWIZZLE macros */
#include "program/prog_instruction.h"
// Shader functions
VkResult anv_CreateShaderModule(
VkDevice _device,
const VkShaderModuleCreateInfo* pCreateInfo,
const VkAllocationCallbacks* pAllocator,
VkShaderModule* pShaderModule)
{
ANV_FROM_HANDLE(anv_device, device, _device);
struct anv_shader_module *module;
assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO);
assert(pCreateInfo->flags == 0);
module = vk_alloc2(&device->alloc, pAllocator,
sizeof(*module) + pCreateInfo->codeSize, 8,
VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
if (module == NULL)
return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
module->size = pCreateInfo->codeSize;
memcpy(module->data, pCreateInfo->pCode, module->size);
_mesa_sha1_compute(module->data, module->size, module->sha1);
*pShaderModule = anv_shader_module_to_handle(module);
return VK_SUCCESS;
}
void anv_DestroyShaderModule(
VkDevice _device,
VkShaderModule _module,
const VkAllocationCallbacks* pAllocator)
{
ANV_FROM_HANDLE(anv_device, device, _device);
ANV_FROM_HANDLE(anv_shader_module, module, _module);
if (!module)
return;
vk_free2(&device->alloc, pAllocator, module);
}
#define SPIR_V_MAGIC_NUMBER 0x07230203
static const uint64_t stage_to_debug[] = {
[MESA_SHADER_VERTEX] = DEBUG_VS,
[MESA_SHADER_TESS_CTRL] = DEBUG_TCS,
[MESA_SHADER_TESS_EVAL] = DEBUG_TES,
[MESA_SHADER_GEOMETRY] = DEBUG_GS,
[MESA_SHADER_FRAGMENT] = DEBUG_WM,
[MESA_SHADER_COMPUTE] = DEBUG_CS,
};
/* Eventually, this will become part of anv_CreateShader. Unfortunately,
* we can't do that yet because we don't have the ability to copy nir.
*/
static nir_shader *
anv_shader_compile_to_nir(struct anv_pipeline *pipeline,
void *mem_ctx,
const struct anv_shader_module *module,
const char *entrypoint_name,
gl_shader_stage stage,
const VkSpecializationInfo *spec_info)
{
const struct anv_device *device = pipeline->device;
const struct brw_compiler *compiler =
device->instance->physicalDevice.compiler;
const nir_shader_compiler_options *nir_options =
compiler->glsl_compiler_options[stage].NirOptions;
uint32_t *spirv = (uint32_t *) module->data;
assert(spirv[0] == SPIR_V_MAGIC_NUMBER);
assert(module->size % 4 == 0);
uint32_t num_spec_entries = 0;
struct nir_spirv_specialization *spec_entries = NULL;
if (spec_info && spec_info->mapEntryCount > 0) {
num_spec_entries = spec_info->mapEntryCount;
spec_entries = malloc(num_spec_entries * sizeof(*spec_entries));
for (uint32_t i = 0; i < num_spec_entries; i++) {
VkSpecializationMapEntry entry = spec_info->pMapEntries[i];
const void *data = spec_info->pData + entry.offset;
assert(data + entry.size <= spec_info->pData + spec_info->dataSize);
spec_entries[i].id = spec_info->pMapEntries[i].constantID;
if (spec_info->dataSize == 8)
spec_entries[i].data64 = *(const uint64_t *)data;
else
spec_entries[i].data32 = *(const uint32_t *)data;
}
}
struct spirv_to_nir_options spirv_options = {
.lower_workgroup_access_to_offsets = true,
.caps = {
.float64 = device->instance->physicalDevice.info.gen >= 8,
.int64 = device->instance->physicalDevice.info.gen >= 8,
.tessellation = true,
.device_group = true,
.draw_parameters = true,
.image_write_without_format = true,
.multiview = true,
.variable_pointers = true,
.storage_16bit = device->instance->physicalDevice.info.gen >= 8,
.int16 = device->instance->physicalDevice.info.gen >= 8,
.shader_viewport_index_layer = true,
.subgroup_arithmetic = true,
.subgroup_basic = true,
.subgroup_ballot = true,
.subgroup_quad = true,
.subgroup_shuffle = true,
.subgroup_vote = true,
.stencil_export = device->instance->physicalDevice.info.gen >= 9,
.storage_8bit = device->instance->physicalDevice.info.gen >= 8,
.post_depth_coverage = device->instance->physicalDevice.info.gen >= 9,
},
};
nir_function *entry_point =
spirv_to_nir(spirv, module->size / 4,
spec_entries, num_spec_entries,
stage, entrypoint_name, &spirv_options, nir_options);
nir_shader *nir = entry_point->shader;
assert(nir->info.stage == stage);
nir_validate_shader(nir);
ralloc_steal(mem_ctx, nir);
free(spec_entries);
if (unlikely(INTEL_DEBUG & stage_to_debug[stage])) {
fprintf(stderr, "NIR (from SPIR-V) for %s shader:\n",
gl_shader_stage_name(stage));
nir_print_shader(nir, stderr);
}
/* We have to lower away local constant initializers right before we
* inline functions. That way they get properly initialized at the top
* of the function and not at the top of its caller.
*/
NIR_PASS_V(nir, nir_lower_constant_initializers, nir_var_local);
NIR_PASS_V(nir, nir_lower_returns);
NIR_PASS_V(nir, nir_inline_functions);
NIR_PASS_V(nir, nir_copy_prop);
/* Pick off the single entrypoint that we want */
foreach_list_typed_safe(nir_function, func, node, &nir->functions) {
if (func != entry_point)
exec_node_remove(&func->node);
}
assert(exec_list_length(&nir->functions) == 1);
entry_point->name = ralloc_strdup(entry_point, "main");
/* Now that we've deleted all but the main function, we can go ahead and
* lower the rest of the constant initializers. We do this here so that
* nir_remove_dead_variables and split_per_member_structs below see the
* corresponding stores.
*/
NIR_PASS_V(nir, nir_lower_constant_initializers, ~0);
/* Split member structs. We do this before lower_io_to_temporaries so that
* it doesn't lower system values to temporaries by accident.
*/
NIR_PASS_V(nir, nir_split_var_copies);
NIR_PASS_V(nir, nir_split_per_member_structs);
NIR_PASS_V(nir, nir_remove_dead_variables,
nir_var_shader_in | nir_var_shader_out | nir_var_system_value);
if (stage == MESA_SHADER_FRAGMENT)
NIR_PASS_V(nir, nir_lower_wpos_center, pipeline->sample_shading_enable);
NIR_PASS_V(nir, nir_propagate_invariant);
NIR_PASS_V(nir, nir_lower_io_to_temporaries,
entry_point->impl, true, false);
/* Vulkan uses the separate-shader linking model */
nir->info.separate_shader = true;
nir = brw_preprocess_nir(compiler, nir);
if (stage == MESA_SHADER_FRAGMENT)
NIR_PASS_V(nir, anv_nir_lower_input_attachments);
return nir;
}
void anv_DestroyPipeline(
VkDevice _device,
VkPipeline _pipeline,
const VkAllocationCallbacks* pAllocator)
{
ANV_FROM_HANDLE(anv_device, device, _device);
ANV_FROM_HANDLE(anv_pipeline, pipeline, _pipeline);
if (!pipeline)
return;
anv_reloc_list_finish(&pipeline->batch_relocs,
pAllocator ? pAllocator : &device->alloc);
if (pipeline->blend_state.map)
anv_state_pool_free(&device->dynamic_state_pool, pipeline->blend_state);
for (unsigned s = 0; s < MESA_SHADER_STAGES; s++) {
if (pipeline->shaders[s])
anv_shader_bin_unref(device, pipeline->shaders[s]);
}
vk_free2(&device->alloc, pAllocator, pipeline);
}
static const uint32_t vk_to_gen_primitive_type[] = {
[VK_PRIMITIVE_TOPOLOGY_POINT_LIST] = _3DPRIM_POINTLIST,
[VK_PRIMITIVE_TOPOLOGY_LINE_LIST] = _3DPRIM_LINELIST,
[VK_PRIMITIVE_TOPOLOGY_LINE_STRIP] = _3DPRIM_LINESTRIP,
[VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST] = _3DPRIM_TRILIST,
[VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP] = _3DPRIM_TRISTRIP,
[VK_PRIMITIVE_TOPOLOGY_TRIANGLE_FAN] = _3DPRIM_TRIFAN,
[VK_PRIMITIVE_TOPOLOGY_LINE_LIST_WITH_ADJACENCY] = _3DPRIM_LINELIST_ADJ,
[VK_PRIMITIVE_TOPOLOGY_LINE_STRIP_WITH_ADJACENCY] = _3DPRIM_LINESTRIP_ADJ,
[VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST_WITH_ADJACENCY] = _3DPRIM_TRILIST_ADJ,
[VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP_WITH_ADJACENCY] = _3DPRIM_TRISTRIP_ADJ,
};
static void
populate_sampler_prog_key(const struct gen_device_info *devinfo,
struct brw_sampler_prog_key_data *key)
{
/* Almost all multisampled textures are compressed. The only time when we
* don't compress a multisampled texture is for 16x MSAA with a surface
* width greater than 8k which is a bit of an edge case. Since the sampler
* just ignores the MCS parameter to ld2ms when MCS is disabled, it's safe
* to tell the compiler to always assume compression.
*/
key->compressed_multisample_layout_mask = ~0;
/* SkyLake added support for 16x MSAA. With this came a new message for
* reading from a 16x MSAA surface with compression. The new message was
* needed because now the MCS data is 64 bits instead of 32 or lower as is
* the case for 8x, 4x, and 2x. The key->msaa_16 bit-field controls which
* message we use. Fortunately, the 16x message works for 8x, 4x, and 2x
* so we can just use it unconditionally. This may not be quite as
* efficient but it saves us from recompiling.
*/
if (devinfo->gen >= 9)
key->msaa_16 = ~0;
/* XXX: Handle texture swizzle on HSW- */
for (int i = 0; i < MAX_SAMPLERS; i++) {
/* Assume color sampler, no swizzling. (Works for BDW+) */
key->swizzles[i] = SWIZZLE_XYZW;
}
}
static void
populate_vs_prog_key(const struct gen_device_info *devinfo,
struct brw_vs_prog_key *key)
{
memset(key, 0, sizeof(*key));
populate_sampler_prog_key(devinfo, &key->tex);
/* XXX: Handle vertex input work-arounds */
/* XXX: Handle sampler_prog_key */
}
static void
populate_tcs_prog_key(const struct gen_device_info *devinfo,
unsigned input_vertices,
struct brw_tcs_prog_key *key)
{
memset(key, 0, sizeof(*key));
populate_sampler_prog_key(devinfo, &key->tex);
key->input_vertices = input_vertices;
}
static void
populate_tes_prog_key(const struct gen_device_info *devinfo,
struct brw_tes_prog_key *key)
{
memset(key, 0, sizeof(*key));
populate_sampler_prog_key(devinfo, &key->tex);
}
static void
populate_gs_prog_key(const struct gen_device_info *devinfo,
struct brw_gs_prog_key *key)
{
memset(key, 0, sizeof(*key));
populate_sampler_prog_key(devinfo, &key->tex);
}
static void
populate_wm_prog_key(const struct gen_device_info *devinfo,
const struct anv_subpass *subpass,
const VkPipelineMultisampleStateCreateInfo *ms_info,
struct brw_wm_prog_key *key)
{
memset(key, 0, sizeof(*key));
populate_sampler_prog_key(devinfo, &key->tex);
/* We set this to 0 here and set to the actual value before we call
* brw_compile_fs.
*/
key->input_slots_valid = 0;
/* Vulkan doesn't specify a default */
key->high_quality_derivatives = false;
/* XXX Vulkan doesn't appear to specify */
key->clamp_fragment_color = false;
assert(subpass->color_count <= MAX_RTS);
for (uint32_t i = 0; i < subpass->color_count; i++) {
if (subpass->color_attachments[i].attachment != VK_ATTACHMENT_UNUSED)
key->color_outputs_valid |= (1 << i);
}
key->nr_color_regions = _mesa_bitcount(key->color_outputs_valid);
key->replicate_alpha = key->nr_color_regions > 1 &&
ms_info && ms_info->alphaToCoverageEnable;
if (ms_info) {
/* We should probably pull this out of the shader, but it's fairly
* harmless to compute it and then let dead-code take care of it.
*/
if (ms_info->rasterizationSamples > 1) {
key->persample_interp =
(ms_info->minSampleShading * ms_info->rasterizationSamples) > 1;
key->multisample_fbo = true;
}
key->frag_coord_adds_sample_pos = ms_info->sampleShadingEnable;
}
}
static void
populate_cs_prog_key(const struct gen_device_info *devinfo,
struct brw_cs_prog_key *key)
{
memset(key, 0, sizeof(*key));
populate_sampler_prog_key(devinfo, &key->tex);
}
struct anv_pipeline_stage {
gl_shader_stage stage;
const struct anv_shader_module *module;
const char *entrypoint;
const VkSpecializationInfo *spec_info;
union brw_any_prog_key key;
struct {
gl_shader_stage stage;
unsigned char sha1[20];
} cache_key;
nir_shader *nir;
struct anv_pipeline_binding surface_to_descriptor[256];
struct anv_pipeline_binding sampler_to_descriptor[256];
struct anv_pipeline_bind_map bind_map;
union brw_any_prog_data prog_data;
};
static void
anv_pipeline_hash_shader(struct mesa_sha1 *ctx,
struct anv_pipeline_stage *stage)
{
_mesa_sha1_update(ctx, stage->module->sha1, sizeof(stage->module->sha1));
_mesa_sha1_update(ctx, stage->entrypoint, strlen(stage->entrypoint));
_mesa_sha1_update(ctx, &stage->stage, sizeof(stage->stage));
if (stage->spec_info) {
_mesa_sha1_update(ctx, stage->spec_info->pMapEntries,
stage->spec_info->mapEntryCount *
sizeof(*stage->spec_info->pMapEntries));
_mesa_sha1_update(ctx, stage->spec_info->pData,
stage->spec_info->dataSize);
}
_mesa_sha1_update(ctx, &stage->key, brw_prog_key_size(stage->stage));
}
static void
anv_pipeline_hash_graphics(struct anv_pipeline *pipeline,
struct anv_pipeline_layout *layout,
struct anv_pipeline_stage *stages,
unsigned char *sha1_out)
{
struct mesa_sha1 ctx;
_mesa_sha1_init(&ctx);
_mesa_sha1_update(&ctx, &pipeline->subpass->view_mask,
sizeof(pipeline->subpass->view_mask));
if (layout)
_mesa_sha1_update(&ctx, layout->sha1, sizeof(layout->sha1));
for (unsigned s = 0; s < MESA_SHADER_STAGES; s++) {
if (stages[s].entrypoint)
anv_pipeline_hash_shader(&ctx, &stages[s]);
}
_mesa_sha1_final(&ctx, sha1_out);
}
static void
anv_pipeline_hash_compute(struct anv_pipeline *pipeline,
struct anv_pipeline_layout *layout,
struct anv_pipeline_stage *stage,
unsigned char *sha1_out)
{
struct mesa_sha1 ctx;
_mesa_sha1_init(&ctx);
if (layout)
_mesa_sha1_update(&ctx, layout->sha1, sizeof(layout->sha1));
anv_pipeline_hash_shader(&ctx, stage);
_mesa_sha1_final(&ctx, sha1_out);
}
static nir_shader *
anv_pipeline_compile(struct anv_pipeline *pipeline,
void *mem_ctx,
struct anv_pipeline_layout *layout,
struct anv_pipeline_stage *stage,
struct brw_stage_prog_data *prog_data,
struct anv_pipeline_bind_map *map)
{
const struct brw_compiler *compiler =
pipeline->device->instance->physicalDevice.compiler;
nir_shader *nir = anv_shader_compile_to_nir(pipeline, mem_ctx,
stage->module,
stage->entrypoint,
stage->stage,
stage->spec_info);
if (nir == NULL)
return NULL;
NIR_PASS_V(nir, anv_nir_lower_ycbcr_textures, layout);
NIR_PASS_V(nir, anv_nir_lower_push_constants);
if (nir->info.stage != MESA_SHADER_COMPUTE)
NIR_PASS_V(nir, anv_nir_lower_multiview, pipeline->subpass->view_mask);
if (nir->info.stage == MESA_SHADER_COMPUTE)
prog_data->total_shared = nir->num_shared;
nir_shader_gather_info(nir, nir_shader_get_entrypoint(nir));
if (nir->num_uniforms > 0) {
assert(prog_data->nr_params == 0);
/* If the shader uses any push constants at all, we'll just give
* them the maximum possible number
*/
assert(nir->num_uniforms <= MAX_PUSH_CONSTANTS_SIZE);
nir->num_uniforms = MAX_PUSH_CONSTANTS_SIZE;
prog_data->nr_params += MAX_PUSH_CONSTANTS_SIZE / sizeof(float);
prog_data->param = ralloc_array(mem_ctx, uint32_t, prog_data->nr_params);
/* We now set the param values to be offsets into a
* anv_push_constant_data structure. Since the compiler doesn't
* actually dereference any of the gl_constant_value pointers in the
* params array, it doesn't really matter what we put here.
*/
struct anv_push_constants *null_data = NULL;
/* Fill out the push constants section of the param array */
for (unsigned i = 0; i < MAX_PUSH_CONSTANTS_SIZE / sizeof(float); i++) {
prog_data->param[i] = ANV_PARAM_PUSH(
(uintptr_t)&null_data->client_data[i * sizeof(float)]);
}
}
if (nir->info.num_ssbos > 0 || nir->info.num_images > 0)
pipeline->needs_data_cache = true;
/* Apply the actual pipeline layout to UBOs, SSBOs, and textures */
if (layout)
anv_nir_apply_pipeline_layout(pipeline, layout, nir, prog_data, map);
if (nir->info.stage != MESA_SHADER_COMPUTE)
brw_nir_analyze_ubo_ranges(compiler, nir, NULL, prog_data->ubo_ranges);
assert(nir->num_uniforms == prog_data->nr_params * 4);
return nir;
}
static void
anv_fill_binding_table(struct brw_stage_prog_data *prog_data, unsigned bias)
{
prog_data->binding_table.size_bytes = 0;
prog_data->binding_table.texture_start = bias;
prog_data->binding_table.gather_texture_start = bias;
prog_data->binding_table.ubo_start = bias;
prog_data->binding_table.ssbo_start = bias;
prog_data->binding_table.image_start = bias;
}
static void
anv_pipeline_link_vs(const struct brw_compiler *compiler,
struct anv_pipeline_stage *vs_stage,
struct anv_pipeline_stage *next_stage)
{
anv_fill_binding_table(&vs_stage->prog_data.vs.base.base, 0);
if (next_stage)
brw_nir_link_shaders(compiler, &vs_stage->nir, &next_stage->nir);
}
static const unsigned *
anv_pipeline_compile_vs(const struct brw_compiler *compiler,
void *mem_ctx,
struct anv_pipeline_stage *vs_stage)
{
brw_compute_vue_map(compiler->devinfo,
&vs_stage->prog_data.vs.base.vue_map,
vs_stage->nir->info.outputs_written,
vs_stage->nir->info.separate_shader);
return brw_compile_vs(compiler, NULL, mem_ctx, &vs_stage->key.vs,
&vs_stage->prog_data.vs, vs_stage->nir, -1, NULL);
}
static void
merge_tess_info(struct shader_info *tes_info,
const struct shader_info *tcs_info)
{
/* The Vulkan 1.0.38 spec, section 21.1 Tessellator says:
*
* "PointMode. Controls generation of points rather than triangles
* or lines. This functionality defaults to disabled, and is
* enabled if either shader stage includes the execution mode.
*
* and about Triangles, Quads, IsoLines, VertexOrderCw, VertexOrderCcw,
* PointMode, SpacingEqual, SpacingFractionalEven, SpacingFractionalOdd,
* and OutputVertices, it says:
*
* "One mode must be set in at least one of the tessellation
* shader stages."
*
* So, the fields can be set in either the TCS or TES, but they must
* agree if set in both. Our backend looks at TES, so bitwise-or in
* the values from the TCS.
*/
assert(tcs_info->tess.tcs_vertices_out == 0 ||
tes_info->tess.tcs_vertices_out == 0 ||
tcs_info->tess.tcs_vertices_out == tes_info->tess.tcs_vertices_out);
tes_info->tess.tcs_vertices_out |= tcs_info->tess.tcs_vertices_out;
assert(tcs_info->tess.spacing == TESS_SPACING_UNSPECIFIED ||
tes_info->tess.spacing == TESS_SPACING_UNSPECIFIED ||
tcs_info->tess.spacing == tes_info->tess.spacing);
tes_info->tess.spacing |= tcs_info->tess.spacing;
assert(tcs_info->tess.primitive_mode == 0 ||
tes_info->tess.primitive_mode == 0 ||
tcs_info->tess.primitive_mode == tes_info->tess.primitive_mode);
tes_info->tess.primitive_mode |= tcs_info->tess.primitive_mode;
tes_info->tess.ccw |= tcs_info->tess.ccw;
tes_info->tess.point_mode |= tcs_info->tess.point_mode;
}
static void
anv_pipeline_link_tcs(const struct brw_compiler *compiler,
struct anv_pipeline_stage *tcs_stage,
struct anv_pipeline_stage *tes_stage)
{
assert(tes_stage && tes_stage->stage == MESA_SHADER_TESS_EVAL);
anv_fill_binding_table(&tcs_stage->prog_data.tcs.base.base, 0);
brw_nir_link_shaders(compiler, &tcs_stage->nir, &tes_stage->nir);
nir_lower_patch_vertices(tes_stage->nir,
tcs_stage->nir->info.tess.tcs_vertices_out,
NULL);
/* Copy TCS info into the TES info */
merge_tess_info(&tes_stage->nir->info, &tcs_stage->nir->info);
anv_fill_binding_table(&tcs_stage->prog_data.tcs.base.base, 0);
anv_fill_binding_table(&tes_stage->prog_data.tes.base.base, 0);
/* Whacking the key after cache lookup is a bit sketchy, but all of
* this comes from the SPIR-V, which is part of the hash used for the
* pipeline cache. So it should be safe.
*/
tcs_stage->key.tcs.tes_primitive_mode =
tes_stage->nir->info.tess.primitive_mode;
tcs_stage->key.tcs.outputs_written =
tcs_stage->nir->info.outputs_written;
tcs_stage->key.tcs.patch_outputs_written =
tcs_stage->nir->info.patch_outputs_written;
tcs_stage->key.tcs.quads_workaround =
compiler->devinfo->gen < 9 &&
tes_stage->nir->info.tess.primitive_mode == 7 /* GL_QUADS */ &&
tes_stage->nir->info.tess.spacing == TESS_SPACING_EQUAL;
tes_stage->key.tes.inputs_read =
tcs_stage->nir->info.outputs_written;
tes_stage->key.tes.patch_inputs_read =
tcs_stage->nir->info.patch_outputs_written;
}
static const unsigned *
anv_pipeline_compile_tcs(const struct brw_compiler *compiler,
void *mem_ctx,
struct anv_pipeline_stage *tcs_stage,
struct anv_pipeline_stage *prev_stage)
{
return brw_compile_tcs(compiler, NULL, mem_ctx, &tcs_stage->key.tcs,
&tcs_stage->prog_data.tcs, tcs_stage->nir,
-1, NULL);
}
static void
anv_pipeline_link_tes(const struct brw_compiler *compiler,
struct anv_pipeline_stage *tes_stage,
struct anv_pipeline_stage *next_stage)
{
anv_fill_binding_table(&tes_stage->prog_data.tes.base.base, 0);
if (next_stage)
brw_nir_link_shaders(compiler, &tes_stage->nir, &next_stage->nir);
}
static const unsigned *
anv_pipeline_compile_tes(const struct brw_compiler *compiler,
void *mem_ctx,
struct anv_pipeline_stage *tes_stage,
struct anv_pipeline_stage *tcs_stage)
{
return brw_compile_tes(compiler, NULL, mem_ctx, &tes_stage->key.tes,
&tcs_stage->prog_data.tcs.base.vue_map,
&tes_stage->prog_data.tes, tes_stage->nir,
NULL, -1, NULL);
}
static void
anv_pipeline_link_gs(const struct brw_compiler *compiler,
struct anv_pipeline_stage *gs_stage,
struct anv_pipeline_stage *next_stage)
{
anv_fill_binding_table(&gs_stage->prog_data.gs.base.base, 0);
if (next_stage)
brw_nir_link_shaders(compiler, &gs_stage->nir, &next_stage->nir);
}
static const unsigned *
anv_pipeline_compile_gs(const struct brw_compiler *compiler,
void *mem_ctx,
struct anv_pipeline_stage *gs_stage,
struct anv_pipeline_stage *prev_stage)
{
brw_compute_vue_map(compiler->devinfo,
&gs_stage->prog_data.gs.base.vue_map,
gs_stage->nir->info.outputs_written,
gs_stage->nir->info.separate_shader);
return brw_compile_gs(compiler, NULL, mem_ctx, &gs_stage->key.gs,
&gs_stage->prog_data.gs, gs_stage->nir,
NULL, -1, NULL);
}
static void
anv_pipeline_link_fs(const struct brw_compiler *compiler,
struct anv_pipeline_stage *stage)
{
unsigned num_rts = 0;
const int max_rt = FRAG_RESULT_DATA7 - FRAG_RESULT_DATA0 + 1;
struct anv_pipeline_binding rt_bindings[max_rt];
nir_function_impl *impl = nir_shader_get_entrypoint(stage->nir);
int rt_to_bindings[max_rt];
memset(rt_to_bindings, -1, sizeof(rt_to_bindings));
bool rt_used[max_rt];
memset(rt_used, 0, sizeof(rt_used));
/* Flag used render targets */
nir_foreach_variable_safe(var, &stage->nir->outputs) {
if (var->data.location < FRAG_RESULT_DATA0)
continue;
const unsigned rt = var->data.location - FRAG_RESULT_DATA0;
/* Unused or out-of-bounds */
if (rt >= MAX_RTS || !(stage->key.wm.color_outputs_valid & (1 << rt)))
continue;
const unsigned array_len =
glsl_type_is_array(var->type) ? glsl_get_length(var->type) : 1;
assert(rt + array_len <= max_rt);
for (unsigned i = 0; i < array_len; i++)
rt_used[rt + i] = true;
}
/* Set new, compacted, location */
for (unsigned i = 0; i < max_rt; i++) {
if (!rt_used[i])
continue;
rt_to_bindings[i] = num_rts;
rt_bindings[rt_to_bindings[i]] = (struct anv_pipeline_binding) {
.set = ANV_DESCRIPTOR_SET_COLOR_ATTACHMENTS,
.binding = 0,
.index = i,
};
num_rts++;
}
bool deleted_output = false;
nir_foreach_variable_safe(var, &stage->nir->outputs) {
if (var->data.location < FRAG_RESULT_DATA0)
continue;
const unsigned rt = var->data.location - FRAG_RESULT_DATA0;
if (rt >= MAX_RTS ||
!(stage->key.wm.color_outputs_valid & (1 << rt))) {
/* Unused or out-of-bounds, throw it away */
deleted_output = true;
var->data.mode = nir_var_local;
exec_node_remove(&var->node);
exec_list_push_tail(&impl->locals, &var->node);
continue;
}
/* Give it the new location */
assert(rt_to_bindings[rt] != -1);
var->data.location = rt_to_bindings[rt] + FRAG_RESULT_DATA0;
}
if (deleted_output)
nir_fixup_deref_modes(stage->nir);
if (num_rts == 0) {
/* If we have no render targets, we need a null render target */
rt_bindings[0] = (struct anv_pipeline_binding) {
.set = ANV_DESCRIPTOR_SET_COLOR_ATTACHMENTS,
.binding = 0,
.index = UINT32_MAX,
};
num_rts = 1;
}
/* Now that we've determined the actual number of render targets, adjust
* the key accordingly.
*/
stage->key.wm.nr_color_regions = num_rts;
stage->key.wm.color_outputs_valid = (1 << num_rts) - 1;
assert(num_rts <= max_rt);
assert(stage->bind_map.surface_count + num_rts <= 256);
memmove(stage->bind_map.surface_to_descriptor + num_rts,
stage->bind_map.surface_to_descriptor,
stage->bind_map.surface_count *
sizeof(*stage->bind_map.surface_to_descriptor));
typed_memcpy(stage->bind_map.surface_to_descriptor,
rt_bindings, num_rts);
stage->bind_map.surface_count += num_rts;
anv_fill_binding_table(&stage->prog_data.wm.base, num_rts);
}
static const unsigned *
anv_pipeline_compile_fs(const struct brw_compiler *compiler,
void *mem_ctx,
struct anv_pipeline_stage *fs_stage,
struct anv_pipeline_stage *prev_stage)
{
/* TODO: we could set this to 0 based on the information in nir_shader, but
* we need this before we call spirv_to_nir.
*/
assert(prev_stage);
fs_stage->key.wm.input_slots_valid =
prev_stage->prog_data.vue.vue_map.slots_valid;
return brw_compile_fs(compiler, NULL, mem_ctx, &fs_stage->key.wm,
&fs_stage->prog_data.wm, fs_stage->nir,
NULL, -1, -1, -1, true, false, NULL, NULL);
}
static VkResult
anv_pipeline_compile_graphics(struct anv_pipeline *pipeline,
struct anv_pipeline_cache *cache,
const VkGraphicsPipelineCreateInfo *info)
{
const struct brw_compiler *compiler =
pipeline->device->instance->physicalDevice.compiler;
struct anv_pipeline_stage stages[MESA_SHADER_STAGES] = {};
pipeline->active_stages = 0;
VkResult result;
for (uint32_t i = 0; i < info->stageCount; i++) {
const VkPipelineShaderStageCreateInfo *sinfo = &info->pStages[i];
gl_shader_stage stage = vk_to_mesa_shader_stage(sinfo->stage);
pipeline->active_stages |= sinfo->stage;
stages[stage].stage = stage;
stages[stage].module = anv_shader_module_from_handle(sinfo->module);
stages[stage].entrypoint = sinfo->pName;
stages[stage].spec_info = sinfo->pSpecializationInfo;
const struct gen_device_info *devinfo = &pipeline->device->info;
switch (stage) {
case MESA_SHADER_VERTEX:
populate_vs_prog_key(devinfo, &stages[stage].key.vs);
break;
case MESA_SHADER_TESS_CTRL:
populate_tcs_prog_key(devinfo,
info->pTessellationState->patchControlPoints,
&stages[stage].key.tcs);
break;
case MESA_SHADER_TESS_EVAL:
populate_tes_prog_key(devinfo, &stages[stage].key.tes);
break;
case MESA_SHADER_GEOMETRY:
populate_gs_prog_key(devinfo, &stages[stage].key.gs);
break;
case MESA_SHADER_FRAGMENT:
populate_wm_prog_key(devinfo, pipeline->subpass,
info->pMultisampleState,
&stages[stage].key.wm);
break;
default:
unreachable("Invalid graphics shader stage");
}
}
if (pipeline->active_stages & VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT)
pipeline->active_stages |= VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT;
assert(pipeline->active_stages & VK_SHADER_STAGE_VERTEX_BIT);
ANV_FROM_HANDLE(anv_pipeline_layout, layout, info->layout);
unsigned char sha1[20];
anv_pipeline_hash_graphics(pipeline, layout, stages, sha1);
unsigned found = 0;
for (unsigned s = 0; s < MESA_SHADER_STAGES; s++) {
if (!stages[s].entrypoint)
continue;
stages[s].cache_key.stage = s;
memcpy(stages[s].cache_key.sha1, sha1, sizeof(sha1));
struct anv_shader_bin *bin =
anv_device_search_for_kernel(pipeline->device, cache,
&stages[s].cache_key,
sizeof(stages[s].cache_key));
if (bin) {
found++;
pipeline->shaders[s] = bin;
}
}
if (found == __builtin_popcount(pipeline->active_stages)) {
/* We found all our shaders in the cache. We're done. */
return VK_SUCCESS;
} else if (found > 0) {
/* We found some but not all of our shaders. This shouldn't happen
* most of the time but it can if we have a partially populated
* pipeline cache.
*/
assert(found < __builtin_popcount(pipeline->active_stages));
vk_debug_report(&pipeline->device->instance->debug_report_callbacks,
VK_DEBUG_REPORT_WARNING_BIT_EXT |
VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_PIPELINE_CACHE_EXT,
(uint64_t)(uintptr_t)cache,
0, 0, "anv",
"Found a partial pipeline in the cache. This is "
"most likely caused by an incomplete pipeline cache "
"import or export");
/* We're going to have to recompile anyway, so just throw away our
* references to the shaders in the cache. We'll get them out of the
* cache again as part of the compilation process.
*/
for (unsigned s = 0; s < MESA_SHADER_STAGES; s++) {
if (pipeline->shaders[s]) {
anv_shader_bin_unref(pipeline->device, pipeline->shaders[s]);
pipeline->shaders[s] = NULL;
}
}
}
void *pipeline_ctx = ralloc_context(NULL);
for (unsigned s = 0; s < MESA_SHADER_STAGES; s++) {
if (!stages[s].entrypoint)
continue;
assert(stages[s].stage == s);
assert(pipeline->shaders[s] == NULL);
stages[s].bind_map = (struct anv_pipeline_bind_map) {
.surface_to_descriptor = stages[s].surface_to_descriptor,
.sampler_to_descriptor = stages[s].sampler_to_descriptor
};
stages[s].nir = anv_pipeline_compile(pipeline, pipeline_ctx, layout,
&stages[s],
&stages[s].prog_data.base,
&stages[s].bind_map);
if (stages[s].nir == NULL)
goto fail;
}
/* Walk backwards to link */
struct anv_pipeline_stage *next_stage = NULL;
for (int s = MESA_SHADER_STAGES - 1; s >= 0; s--) {
if (!stages[s].entrypoint)
continue;
switch (s) {
case MESA_SHADER_VERTEX:
anv_pipeline_link_vs(compiler, &stages[s], next_stage);
break;
case MESA_SHADER_TESS_CTRL:
anv_pipeline_link_tcs(compiler, &stages[s], next_stage);
break;
case MESA_SHADER_TESS_EVAL:
anv_pipeline_link_tes(compiler, &stages[s], next_stage);
break;
case MESA_SHADER_GEOMETRY:
anv_pipeline_link_gs(compiler, &stages[s], next_stage);
break;
case MESA_SHADER_FRAGMENT:
anv_pipeline_link_fs(compiler, &stages[s]);
break;
default:
unreachable("Invalid graphics shader stage");
}
next_stage = &stages[s];
}
struct anv_pipeline_stage *prev_stage = NULL;
for (unsigned s = 0; s < MESA_SHADER_STAGES; s++) {
if (!stages[s].entrypoint)
continue;
void *stage_ctx = ralloc_context(NULL);
const unsigned *code;
switch (s) {
case MESA_SHADER_VERTEX:
code = anv_pipeline_compile_vs(compiler, stage_ctx, &stages[s]);
break;
case MESA_SHADER_TESS_CTRL:
code = anv_pipeline_compile_tcs(compiler, stage_ctx,
&stages[s], prev_stage);
break;
case MESA_SHADER_TESS_EVAL:
code = anv_pipeline_compile_tes(compiler, stage_ctx,
&stages[s], prev_stage);
break;
case MESA_SHADER_GEOMETRY:
code = anv_pipeline_compile_gs(compiler, stage_ctx,
&stages[s], prev_stage);
break;
case MESA_SHADER_FRAGMENT:
code = anv_pipeline_compile_fs(compiler, stage_ctx,
&stages[s], prev_stage);
break;
default:
unreachable("Invalid graphics shader stage");
}
if (code == NULL) {
ralloc_free(stage_ctx);
result = vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
goto fail;
}
struct anv_shader_bin *bin =
anv_device_upload_kernel(pipeline->device, cache,
&stages[s].cache_key,
sizeof(stages[s].cache_key),
code, stages[s].prog_data.base.program_size,
stages[s].nir->constant_data,
stages[s].nir->constant_data_size,
&stages[s].prog_data.base,
brw_prog_data_size(s),
&stages[s].bind_map);
if (!bin) {
ralloc_free(stage_ctx);
result = vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
goto fail;
}
pipeline->shaders[s] = bin;
ralloc_free(stage_ctx);
prev_stage = &stages[s];
}
ralloc_free(pipeline_ctx);
return VK_SUCCESS;
fail:
ralloc_free(pipeline_ctx);
for (unsigned s = 0; s < MESA_SHADER_STAGES; s++) {
if (pipeline->shaders[s])
anv_shader_bin_unref(pipeline->device, pipeline->shaders[s]);
}
return result;
}
VkResult
anv_pipeline_compile_cs(struct anv_pipeline *pipeline,
struct anv_pipeline_cache *cache,
const VkComputePipelineCreateInfo *info,
const struct anv_shader_module *module,
const char *entrypoint,
const VkSpecializationInfo *spec_info)
{
const struct brw_compiler *compiler =
pipeline->device->instance->physicalDevice.compiler;
struct anv_pipeline_stage stage = {
.stage = MESA_SHADER_COMPUTE,
.module = module,
.entrypoint = entrypoint,
.spec_info = spec_info,
};
struct anv_shader_bin *bin = NULL;
populate_cs_prog_key(&pipeline->device->info, &stage.key.cs);
ANV_FROM_HANDLE(anv_pipeline_layout, layout, info->layout);
unsigned char sha1[20];
anv_pipeline_hash_compute(pipeline, layout, &stage, sha1);
bin = anv_device_search_for_kernel(pipeline->device, cache, sha1, 20);
if (bin == NULL) {
struct brw_cs_prog_data prog_data = {};
stage.bind_map = (struct anv_pipeline_bind_map) {
.surface_to_descriptor = stage.surface_to_descriptor,
.sampler_to_descriptor = stage.sampler_to_descriptor
};
void *mem_ctx = ralloc_context(NULL);
nir_shader *nir = anv_pipeline_compile(pipeline, mem_ctx, layout, &stage,
&prog_data.base, &stage.bind_map);
if (nir == NULL) {
ralloc_free(mem_ctx);
return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
}
NIR_PASS_V(nir, anv_nir_add_base_work_group_id, &prog_data);
anv_fill_binding_table(&prog_data.base, 1);
const unsigned *shader_code =
brw_compile_cs(compiler, NULL, mem_ctx, &stage.key.cs,
&prog_data, nir, -1, NULL);
if (shader_code == NULL) {
ralloc_free(mem_ctx);
return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
}
const unsigned code_size = prog_data.base.program_size;
bin = anv_device_upload_kernel(pipeline->device, cache, sha1, 20,
shader_code, code_size,
nir->constant_data,
nir->constant_data_size,
&prog_data.base, sizeof(prog_data),
&stage.bind_map);
if (!bin) {
ralloc_free(mem_ctx);
return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
}
ralloc_free(mem_ctx);
}
pipeline->active_stages = VK_SHADER_STAGE_COMPUTE_BIT;
pipeline->shaders[MESA_SHADER_COMPUTE] = bin;
return VK_SUCCESS;
}
/**
* Copy pipeline state not marked as dynamic.
* Dynamic state is pipeline state which hasn't been provided at pipeline
* creation time, but is dynamically provided afterwards using various
* vkCmdSet* functions.
*
* The set of state considered "non_dynamic" is determined by the pieces of
* state that have their corresponding VkDynamicState enums omitted from
* VkPipelineDynamicStateCreateInfo::pDynamicStates.
*
* @param[out] pipeline Destination non_dynamic state.
* @param[in] pCreateInfo Source of non_dynamic state to be copied.
*/
static void
copy_non_dynamic_state(struct anv_pipeline *pipeline,
const VkGraphicsPipelineCreateInfo *pCreateInfo)
{
anv_cmd_dirty_mask_t states = ANV_CMD_DIRTY_DYNAMIC_ALL;
struct anv_subpass *subpass = pipeline->subpass;
pipeline->dynamic_state = default_dynamic_state;
if (pCreateInfo->pDynamicState) {
/* Remove all of the states that are marked as dynamic */
uint32_t count = pCreateInfo->pDynamicState->dynamicStateCount;
for (uint32_t s = 0; s < count; s++)
states &= ~(1 << pCreateInfo->pDynamicState->pDynamicStates[s]);
}
struct anv_dynamic_state *dynamic = &pipeline->dynamic_state;
/* Section 9.2 of the Vulkan 1.0.15 spec says:
*
* pViewportState is [...] NULL if the pipeline
* has rasterization disabled.
*/
if (!pCreateInfo->pRasterizationState->rasterizerDiscardEnable) {
assert(pCreateInfo->pViewportState);
dynamic->viewport.count = pCreateInfo->pViewportState->viewportCount;
if (states & (1 << VK_DYNAMIC_STATE_VIEWPORT)) {
typed_memcpy(dynamic->viewport.viewports,
pCreateInfo->pViewportState->pViewports,
pCreateInfo->pViewportState->viewportCount);
}
dynamic->scissor.count = pCreateInfo->pViewportState->scissorCount;
if (states & (1 << VK_DYNAMIC_STATE_SCISSOR)) {
typed_memcpy(dynamic->scissor.scissors,
pCreateInfo->pViewportState->pScissors,
pCreateInfo->pViewportState->scissorCount);
}
}
if (states & (1 << VK_DYNAMIC_STATE_LINE_WIDTH)) {
assert(pCreateInfo->pRasterizationState);
dynamic->line_width = pCreateInfo->pRasterizationState->lineWidth;
}
if (states & (1 << VK_DYNAMIC_STATE_DEPTH_BIAS)) {
assert(pCreateInfo->pRasterizationState);
dynamic->depth_bias.bias =
pCreateInfo->pRasterizationState->depthBiasConstantFactor;
dynamic->depth_bias.clamp =
pCreateInfo->pRasterizationState->depthBiasClamp;
dynamic->depth_bias.slope =
pCreateInfo->pRasterizationState->depthBiasSlopeFactor;
}
/* Section 9.2 of the Vulkan 1.0.15 spec says:
*
* pColorBlendState is [...] NULL if the pipeline has rasterization
* disabled or if the subpass of the render pass the pipeline is
* created against does not use any color attachments.
*/
bool uses_color_att = false;
for (unsigned i = 0; i < subpass->color_count; ++i) {
if (subpass->color_attachments[i].attachment != VK_ATTACHMENT_UNUSED) {
uses_color_att = true;
break;
}
}
if (uses_color_att &&
!pCreateInfo->pRasterizationState->rasterizerDiscardEnable) {
assert(pCreateInfo->pColorBlendState);
if (states & (1 << VK_DYNAMIC_STATE_BLEND_CONSTANTS))
typed_memcpy(dynamic->blend_constants,
pCreateInfo->pColorBlendState->blendConstants, 4);
}
/* If there is no depthstencil attachment, then don't read
* pDepthStencilState. The Vulkan spec states that pDepthStencilState may
* be NULL in this case. Even if pDepthStencilState is non-NULL, there is
* no need to override the depthstencil defaults in
* anv_pipeline::dynamic_state when there is no depthstencil attachment.
*
* Section 9.2 of the Vulkan 1.0.15 spec says:
*
* pDepthStencilState is [...] NULL if the pipeline has rasterization
* disabled or if the subpass of the render pass the pipeline is created
* against does not use a depth/stencil attachment.
*/
if (!pCreateInfo->pRasterizationState->rasterizerDiscardEnable &&
subpass->depth_stencil_attachment) {
assert(pCreateInfo->pDepthStencilState);
if (states & (1 << VK_DYNAMIC_STATE_DEPTH_BOUNDS)) {
dynamic->depth_bounds.min =
pCreateInfo->pDepthStencilState->minDepthBounds;
dynamic->depth_bounds.max =
pCreateInfo->pDepthStencilState->maxDepthBounds;
}
if (states & (1 << VK_DYNAMIC_STATE_STENCIL_COMPARE_MASK)) {
dynamic->stencil_compare_mask.front =
pCreateInfo->pDepthStencilState->front.compareMask;
dynamic->stencil_compare_mask.back =
pCreateInfo->pDepthStencilState->back.compareMask;
}
if (states & (1 << VK_DYNAMIC_STATE_STENCIL_WRITE_MASK)) {
dynamic->stencil_write_mask.front =
pCreateInfo->pDepthStencilState->front.writeMask;
dynamic->stencil_write_mask.back =
pCreateInfo->pDepthStencilState->back.writeMask;
}
if (states & (1 << VK_DYNAMIC_STATE_STENCIL_REFERENCE)) {
dynamic->stencil_reference.front =
pCreateInfo->pDepthStencilState->front.reference;
dynamic->stencil_reference.back =
pCreateInfo->pDepthStencilState->back.reference;
}
}
pipeline->dynamic_state_mask = states;
}
static void
anv_pipeline_validate_create_info(const VkGraphicsPipelineCreateInfo *info)
{
#ifdef DEBUG
struct anv_render_pass *renderpass = NULL;
struct anv_subpass *subpass = NULL;
/* Assert that all required members of VkGraphicsPipelineCreateInfo are
* present. See the Vulkan 1.0.28 spec, Section 9.2 Graphics Pipelines.
*/
assert(info->sType == VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO);
renderpass = anv_render_pass_from_handle(info->renderPass);
assert(renderpass);
assert(info->subpass < renderpass->subpass_count);
subpass = &renderpass->subpasses[info->subpass];
assert(info->stageCount >= 1);
assert(info->pVertexInputState);
assert(info->pInputAssemblyState);
assert(info->pRasterizationState);
if (!info->pRasterizationState->rasterizerDiscardEnable) {
assert(info->pViewportState);
assert(info->pMultisampleState);
if (subpass && subpass->depth_stencil_attachment)
assert(info->pDepthStencilState);
if (subpass && subpass->color_count > 0) {
bool all_color_unused = true;
for (int i = 0; i < subpass->color_count; i++) {
if (subpass->color_attachments[i].attachment != VK_ATTACHMENT_UNUSED)
all_color_unused = false;
}
/* pColorBlendState is ignored if the pipeline has rasterization
* disabled or if the subpass of the render pass the pipeline is
* created against does not use any color attachments.
*/
assert(info->pColorBlendState || all_color_unused);
}
}
for (uint32_t i = 0; i < info->stageCount; ++i) {
switch (info->pStages[i].stage) {
case VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT:
case VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT:
assert(info->pTessellationState);
break;
default:
break;
}
}
#endif
}
/**
* Calculate the desired L3 partitioning based on the current state of the
* pipeline. For now this simply returns the conservative defaults calculated
* by get_default_l3_weights(), but we could probably do better by gathering
* more statistics from the pipeline state (e.g. guess of expected URB usage
* and bound surfaces), or by using feed-back from performance counters.
*/
void
anv_pipeline_setup_l3_config(struct anv_pipeline *pipeline, bool needs_slm)
{
const struct gen_device_info *devinfo = &pipeline->device->info;
const struct gen_l3_weights w =
gen_get_default_l3_weights(devinfo, pipeline->needs_data_cache, needs_slm);
pipeline->urb.l3_config = gen_get_l3_config(devinfo, w);
pipeline->urb.total_size =
gen_get_l3_config_urb_size(devinfo, pipeline->urb.l3_config);
}
VkResult
anv_pipeline_init(struct anv_pipeline *pipeline,
struct anv_device *device,
struct anv_pipeline_cache *cache,
const VkGraphicsPipelineCreateInfo *pCreateInfo,
const VkAllocationCallbacks *alloc)
{
VkResult result;
anv_pipeline_validate_create_info(pCreateInfo);
if (alloc == NULL)
alloc = &device->alloc;
pipeline->device = device;
ANV_FROM_HANDLE(anv_render_pass, render_pass, pCreateInfo->renderPass);
assert(pCreateInfo->subpass < render_pass->subpass_count);
pipeline->subpass = &render_pass->subpasses[pCreateInfo->subpass];
result = anv_reloc_list_init(&pipeline->batch_relocs, alloc);
if (result != VK_SUCCESS)
return result;
pipeline->batch.alloc = alloc;
pipeline->batch.next = pipeline->batch.start = pipeline->batch_data;
pipeline->batch.end = pipeline->batch.start + sizeof(pipeline->batch_data);
pipeline->batch.relocs = &pipeline->batch_relocs;
pipeline->batch.status = VK_SUCCESS;
copy_non_dynamic_state(pipeline, pCreateInfo);
pipeline->depth_clamp_enable = pCreateInfo->pRasterizationState &&
pCreateInfo->pRasterizationState->depthClampEnable;
pipeline->sample_shading_enable = pCreateInfo->pMultisampleState &&
pCreateInfo->pMultisampleState->sampleShadingEnable;
pipeline->needs_data_cache = false;
/* When we free the pipeline, we detect stages based on the NULL status
* of various prog_data pointers. Make them NULL by default.
*/
memset(pipeline->shaders, 0, sizeof(pipeline->shaders));
result = anv_pipeline_compile_graphics(pipeline, cache, pCreateInfo);
if (result != VK_SUCCESS) {
anv_reloc_list_finish(&pipeline->batch_relocs, alloc);
return result;
}
assert(pipeline->shaders[MESA_SHADER_VERTEX]);
anv_pipeline_setup_l3_config(pipeline, false);
const VkPipelineVertexInputStateCreateInfo *vi_info =
pCreateInfo->pVertexInputState;
const uint64_t inputs_read = get_vs_prog_data(pipeline)->inputs_read;
pipeline->vb_used = 0;
for (uint32_t i = 0; i < vi_info->vertexAttributeDescriptionCount; i++) {
const VkVertexInputAttributeDescription *desc =
&vi_info->pVertexAttributeDescriptions[i];
if (inputs_read & (1ull << (VERT_ATTRIB_GENERIC0 + desc->location)))
pipeline->vb_used |= 1 << desc->binding;
}
for (uint32_t i = 0; i < vi_info->vertexBindingDescriptionCount; i++) {
const VkVertexInputBindingDescription *desc =
&vi_info->pVertexBindingDescriptions[i];
pipeline->vb[desc->binding].stride = desc->stride;
/* Step rate is programmed per vertex element (attribute), not
* binding. Set up a map of which bindings step per instance, for
* reference by vertex element setup. */
switch (desc->inputRate) {
default:
case VK_VERTEX_INPUT_RATE_VERTEX:
pipeline->vb[desc->binding].instanced = false;
break;
case VK_VERTEX_INPUT_RATE_INSTANCE:
pipeline->vb[desc->binding].instanced = true;
break;
}
pipeline->vb[desc->binding].instance_divisor = 1;
}
const VkPipelineVertexInputDivisorStateCreateInfoEXT *vi_div_state =
vk_find_struct_const(vi_info->pNext,
PIPELINE_VERTEX_INPUT_DIVISOR_STATE_CREATE_INFO_EXT);
if (vi_div_state) {
for (uint32_t i = 0; i < vi_div_state->vertexBindingDivisorCount; i++) {
const VkVertexInputBindingDivisorDescriptionEXT *desc =
&vi_div_state->pVertexBindingDivisors[i];
pipeline->vb[desc->binding].instance_divisor = desc->divisor;
}
}
/* Our implementation of VK_KHR_multiview uses instancing to draw the
* different views. If the client asks for instancing, we need to multiply
* the instance divisor by the number of views ensure that we repeat the
* client's per-instance data once for each view.
*/
if (pipeline->subpass->view_mask) {
const uint32_t view_count = anv_subpass_view_count(pipeline->subpass);
for (uint32_t vb = 0; vb < MAX_VBS; vb++) {
if (pipeline->vb[vb].instanced)
pipeline->vb[vb].instance_divisor *= view_count;
}
}
const VkPipelineInputAssemblyStateCreateInfo *ia_info =
pCreateInfo->pInputAssemblyState;
const VkPipelineTessellationStateCreateInfo *tess_info =
pCreateInfo->pTessellationState;
pipeline->primitive_restart = ia_info->primitiveRestartEnable;
if (anv_pipeline_has_stage(pipeline, MESA_SHADER_TESS_EVAL))
pipeline->topology = _3DPRIM_PATCHLIST(tess_info->patchControlPoints);
else
pipeline->topology = vk_to_gen_primitive_type[ia_info->topology];
return VK_SUCCESS;
}
|