summaryrefslogtreecommitdiffstats
path: root/src/intel/vulkan/genX_cmd_buffer.c
blob: ee47c2926e0579c00346a16d7eecc6b1b84c4081 (plain)
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
/*
 * 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 "anv_private.h"

#include "genxml/gen_macros.h"
#include "genxml/genX_pack.h"

void
genX(cmd_buffer_emit_state_base_address)(struct anv_cmd_buffer *cmd_buffer)
{
   struct anv_device *device = cmd_buffer->device;
   struct anv_bo *scratch_bo = NULL;

   cmd_buffer->state.scratch_size =
      anv_block_pool_size(&device->scratch_block_pool);
   if (cmd_buffer->state.scratch_size > 0)
      scratch_bo = &device->scratch_block_pool.bo;

/* XXX: Do we need this on more than just BDW? */
#if (GEN_GEN >= 8)
   /* Emit a render target cache flush.
    *
    * This isn't documented anywhere in the PRM.  However, it seems to be
    * necessary prior to changing the surface state base adress.  Without
    * this, we get GPU hangs when using multi-level command buffers which
    * clear depth, reset state base address, and then go render stuff.
    */
   anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), pc) {
      pc.RenderTargetCacheFlushEnable = true;
   }
#endif

   anv_batch_emit(&cmd_buffer->batch, GENX(STATE_BASE_ADDRESS), sba) {
      sba.GeneralStateBaseAddress = (struct anv_address) { scratch_bo, 0 };
      sba.GeneralStateMemoryObjectControlState = GENX(MOCS);
      sba.GeneralStateBaseAddressModifyEnable = true;

      sba.SurfaceStateBaseAddress =
         anv_cmd_buffer_surface_base_address(cmd_buffer);
      sba.SurfaceStateMemoryObjectControlState = GENX(MOCS);
      sba.SurfaceStateBaseAddressModifyEnable = true;

      sba.DynamicStateBaseAddress =
         (struct anv_address) { &device->dynamic_state_block_pool.bo, 0 };
      sba.DynamicStateMemoryObjectControlState = GENX(MOCS),
      sba.DynamicStateBaseAddressModifyEnable = true,

      sba.IndirectObjectBaseAddress = (struct anv_address) { NULL, 0 };
      sba.IndirectObjectMemoryObjectControlState = GENX(MOCS);
      sba.IndirectObjectBaseAddressModifyEnable = true;

      sba.InstructionBaseAddress =
         (struct anv_address) { &device->instruction_block_pool.bo, 0 };
      sba.InstructionMemoryObjectControlState = GENX(MOCS);
      sba.InstructionBaseAddressModifyEnable = true;

#  if (GEN_GEN >= 8)
      /* Broadwell requires that we specify a buffer size for a bunch of
       * these fields.  However, since we will be growing the BO's live, we
       * just set them all to the maximum.
       */
      sba.GeneralStateBufferSize                = 0xfffff;
      sba.GeneralStateBufferSizeModifyEnable    = true;
      sba.DynamicStateBufferSize                = 0xfffff;
      sba.DynamicStateBufferSizeModifyEnable    = true;
      sba.IndirectObjectBufferSize              = 0xfffff;
      sba.IndirectObjectBufferSizeModifyEnable  = true;
      sba.InstructionBufferSize                 = 0xfffff;
      sba.InstructionBuffersizeModifyEnable     = true;
#  endif
   }

   /* After re-setting the surface state base address, we have to do some
    * cache flusing so that the sampler engine will pick up the new
    * SURFACE_STATE objects and binding tables. From the Broadwell PRM,
    * Shared Function > 3D Sampler > State > State Caching (page 96):
    *
    *    Coherency with system memory in the state cache, like the texture
    *    cache is handled partially by software. It is expected that the
    *    command stream or shader will issue Cache Flush operation or
    *    Cache_Flush sampler message to ensure that the L1 cache remains
    *    coherent with system memory.
    *
    *    [...]
    *
    *    Whenever the value of the Dynamic_State_Base_Addr,
    *    Surface_State_Base_Addr are altered, the L1 state cache must be
    *    invalidated to ensure the new surface or sampler state is fetched
    *    from system memory.
    *
    * The PIPE_CONTROL command has a "State Cache Invalidation Enable" bit
    * which, according the PIPE_CONTROL instruction documentation in the
    * Broadwell PRM:
    *
    *    Setting this bit is independent of any other bit in this packet.
    *    This bit controls the invalidation of the L1 and L2 state caches
    *    at the top of the pipe i.e. at the parsing time.
    *
    * Unfortunately, experimentation seems to indicate that state cache
    * invalidation through a PIPE_CONTROL does nothing whatsoever in
    * regards to surface state and binding tables.  In stead, it seems that
    * invalidating the texture cache is what is actually needed.
    *
    * XXX:  As far as we have been able to determine through
    * experimentation, shows that flush the texture cache appears to be
    * sufficient.  The theory here is that all of the sampling/rendering
    * units cache the binding table in the texture cache.  However, we have
    * yet to be able to actually confirm this.
    */
   anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), pc) {
      pc.TextureCacheInvalidationEnable = true;
   }
}

void genX(CmdPipelineBarrier)(
    VkCommandBuffer                             commandBuffer,
    VkPipelineStageFlags                        srcStageMask,
    VkPipelineStageFlags                        destStageMask,
    VkBool32                                    byRegion,
    uint32_t                                    memoryBarrierCount,
    const VkMemoryBarrier*                      pMemoryBarriers,
    uint32_t                                    bufferMemoryBarrierCount,
    const VkBufferMemoryBarrier*                pBufferMemoryBarriers,
    uint32_t                                    imageMemoryBarrierCount,
    const VkImageMemoryBarrier*                 pImageMemoryBarriers)
{
   ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
   uint32_t b, *dw;

   /* XXX: Right now, we're really dumb and just flush whatever categories
    * the app asks for.  One of these days we may make this a bit better
    * but right now that's all the hardware allows for in most areas.
    */
   VkAccessFlags src_flags = 0;
   VkAccessFlags dst_flags = 0;

   for (uint32_t i = 0; i < memoryBarrierCount; i++) {
      src_flags |= pMemoryBarriers[i].srcAccessMask;
      dst_flags |= pMemoryBarriers[i].dstAccessMask;
   }

   for (uint32_t i = 0; i < bufferMemoryBarrierCount; i++) {
      src_flags |= pBufferMemoryBarriers[i].srcAccessMask;
      dst_flags |= pBufferMemoryBarriers[i].dstAccessMask;
   }

   for (uint32_t i = 0; i < imageMemoryBarrierCount; i++) {
      src_flags |= pImageMemoryBarriers[i].srcAccessMask;
      dst_flags |= pImageMemoryBarriers[i].dstAccessMask;
   }

   /* Mask out the Source access flags we care about */
   const uint32_t src_mask =
      VK_ACCESS_SHADER_WRITE_BIT |
      VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
      VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT |
      VK_ACCESS_TRANSFER_WRITE_BIT;

   src_flags = src_flags & src_mask;

   /* Mask out the destination access flags we care about */
   const uint32_t dst_mask =
      VK_ACCESS_INDIRECT_COMMAND_READ_BIT |
      VK_ACCESS_INDEX_READ_BIT |
      VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT |
      VK_ACCESS_UNIFORM_READ_BIT |
      VK_ACCESS_SHADER_READ_BIT |
      VK_ACCESS_COLOR_ATTACHMENT_READ_BIT |
      VK_ACCESS_TRANSFER_READ_BIT;

   dst_flags = dst_flags & dst_mask;

   /* The src flags represent how things were used previously.  This is
    * what we use for doing flushes.
    */
   struct GENX(PIPE_CONTROL) flush_cmd = {
      GENX(PIPE_CONTROL_header),
      .PostSyncOperation = NoWrite,
   };

   for_each_bit(b, src_flags) {
      switch ((VkAccessFlagBits)(1 << b)) {
      case VK_ACCESS_SHADER_WRITE_BIT:
         flush_cmd.DCFlushEnable = true;
         break;
      case VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT:
         flush_cmd.RenderTargetCacheFlushEnable = true;
         break;
      case VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT:
         flush_cmd.DepthCacheFlushEnable = true;
         break;
      case VK_ACCESS_TRANSFER_WRITE_BIT:
         flush_cmd.RenderTargetCacheFlushEnable = true;
         flush_cmd.DepthCacheFlushEnable = true;
         break;
      default:
         unreachable("should've masked this out by now");
      }
   }

   /* If we end up doing two PIPE_CONTROLs, the first, flusing one also has to
    * stall and wait for the flushing to finish, so we don't re-dirty the
    * caches with in-flight rendering after the second PIPE_CONTROL
    * invalidates.
    */

   if (dst_flags)
      flush_cmd.CommandStreamerStallEnable = true;

   if (src_flags && dst_flags) {
      dw = anv_batch_emit_dwords(&cmd_buffer->batch, GENX(PIPE_CONTROL_length));
      GENX(PIPE_CONTROL_pack)(&cmd_buffer->batch, dw, &flush_cmd);
   }

   /* The dst flags represent how things will be used in the future.  This
    * is what we use for doing cache invalidations.
    */
   struct GENX(PIPE_CONTROL) invalidate_cmd = {
      GENX(PIPE_CONTROL_header),
      .PostSyncOperation = NoWrite,
   };

   for_each_bit(b, dst_flags) {
      switch ((VkAccessFlagBits)(1 << b)) {
      case VK_ACCESS_INDIRECT_COMMAND_READ_BIT:
      case VK_ACCESS_INDEX_READ_BIT:
      case VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT:
         invalidate_cmd.VFCacheInvalidationEnable = true;
         break;
      case VK_ACCESS_UNIFORM_READ_BIT:
         invalidate_cmd.ConstantCacheInvalidationEnable = true;
         /* fallthrough */
      case VK_ACCESS_SHADER_READ_BIT:
         invalidate_cmd.TextureCacheInvalidationEnable = true;
         break;
      case VK_ACCESS_COLOR_ATTACHMENT_READ_BIT:
         invalidate_cmd.TextureCacheInvalidationEnable = true;
         break;
      case VK_ACCESS_TRANSFER_READ_BIT:
         invalidate_cmd.TextureCacheInvalidationEnable = true;
         break;
      default:
         unreachable("should've masked this out by now");
      }
   }

   if (dst_flags) {
      dw = anv_batch_emit_dwords(&cmd_buffer->batch, GENX(PIPE_CONTROL_length));
      GENX(PIPE_CONTROL_pack)(&cmd_buffer->batch, dw, &invalidate_cmd);
   }
}

static uint32_t
cmd_buffer_flush_push_constants(struct anv_cmd_buffer *cmd_buffer)
{
   static const uint32_t push_constant_opcodes[] = {
      [MESA_SHADER_VERTEX]                      = 21,
      [MESA_SHADER_TESS_CTRL]                   = 25, /* HS */
      [MESA_SHADER_TESS_EVAL]                   = 26, /* DS */
      [MESA_SHADER_GEOMETRY]                    = 22,
      [MESA_SHADER_FRAGMENT]                    = 23,
      [MESA_SHADER_COMPUTE]                     = 0,
   };

   VkShaderStageFlags flushed = 0;

   anv_foreach_stage(stage, cmd_buffer->state.push_constants_dirty) {
      if (stage == MESA_SHADER_COMPUTE)
         continue;

      struct anv_state state = anv_cmd_buffer_push_constants(cmd_buffer, stage);

      if (state.offset == 0) {
         anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_CONSTANT_VS), c)
            c._3DCommandSubOpcode = push_constant_opcodes[stage];
      } else {
         anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_CONSTANT_VS), c) {
            c._3DCommandSubOpcode = push_constant_opcodes[stage],
            c.ConstantBody = (struct GENX(3DSTATE_CONSTANT_BODY)) {
#if GEN_GEN >= 9
               .PointerToConstantBuffer2 = { &cmd_buffer->device->dynamic_state_block_pool.bo, state.offset },
               .ConstantBuffer2ReadLength = DIV_ROUND_UP(state.alloc_size, 32),
#else
               .PointerToConstantBuffer0 = { .offset = state.offset },
               .ConstantBuffer0ReadLength = DIV_ROUND_UP(state.alloc_size, 32),
#endif
            };
         }
      }

      flushed |= mesa_to_vk_shader_stage(stage);
   }

   cmd_buffer->state.push_constants_dirty &= ~VK_SHADER_STAGE_ALL_GRAPHICS;

   return flushed;
}

void
genX(cmd_buffer_flush_state)(struct anv_cmd_buffer *cmd_buffer)
{
   struct anv_pipeline *pipeline = cmd_buffer->state.pipeline;
   uint32_t *p;

   uint32_t vb_emit = cmd_buffer->state.vb_dirty & pipeline->vb_used;

   assert((pipeline->active_stages & VK_SHADER_STAGE_COMPUTE_BIT) == 0);

   genX(cmd_buffer_config_l3)(cmd_buffer, pipeline);

   genX(flush_pipeline_select_3d)(cmd_buffer);

   if (vb_emit) {
      const uint32_t num_buffers = __builtin_popcount(vb_emit);
      const uint32_t num_dwords = 1 + num_buffers * 4;

      p = anv_batch_emitn(&cmd_buffer->batch, num_dwords,
                          GENX(3DSTATE_VERTEX_BUFFERS));
      uint32_t vb, i = 0;
      for_each_bit(vb, vb_emit) {
         struct anv_buffer *buffer = cmd_buffer->state.vertex_bindings[vb].buffer;
         uint32_t offset = cmd_buffer->state.vertex_bindings[vb].offset;

         struct GENX(VERTEX_BUFFER_STATE) state = {
            .VertexBufferIndex = vb,

#if GEN_GEN >= 8
            .MemoryObjectControlState = GENX(MOCS),
#else
            .BufferAccessType = pipeline->instancing_enable[vb] ? INSTANCEDATA : VERTEXDATA,
            .InstanceDataStepRate = 1,
            .VertexBufferMemoryObjectControlState = GENX(MOCS),
#endif

            .AddressModifyEnable = true,
            .BufferPitch = pipeline->binding_stride[vb],
            .BufferStartingAddress = { buffer->bo, buffer->offset + offset },

#if GEN_GEN >= 8
            .BufferSize = buffer->size - offset
#else
            .EndAddress = { buffer->bo, buffer->offset + buffer->size - 1},
#endif
         };

         GENX(VERTEX_BUFFER_STATE_pack)(&cmd_buffer->batch, &p[1 + i * 4], &state);
         i++;
      }
   }

   cmd_buffer->state.vb_dirty &= ~vb_emit;

   if (cmd_buffer->state.dirty & ANV_CMD_DIRTY_PIPELINE) {
      /* If somebody compiled a pipeline after starting a command buffer the
       * scratch bo may have grown since we started this cmd buffer (and
       * emitted STATE_BASE_ADDRESS).  If we're binding that pipeline now,
       * reemit STATE_BASE_ADDRESS so that we use the bigger scratch bo. */
      if (cmd_buffer->state.scratch_size < pipeline->total_scratch)
         anv_cmd_buffer_emit_state_base_address(cmd_buffer);

      anv_batch_emit_batch(&cmd_buffer->batch, &pipeline->batch);

      /* From the BDW PRM for 3DSTATE_PUSH_CONSTANT_ALLOC_VS:
       *
       *    "The 3DSTATE_CONSTANT_VS must be reprogrammed prior to
       *    the next 3DPRIMITIVE command after programming the
       *    3DSTATE_PUSH_CONSTANT_ALLOC_VS"
       *
       * Since 3DSTATE_PUSH_CONSTANT_ALLOC_VS is programmed as part of
       * pipeline setup, we need to dirty push constants.
       */
      cmd_buffer->state.push_constants_dirty |= VK_SHADER_STAGE_ALL_GRAPHICS;
   }

#if GEN_GEN <= 7
   if (cmd_buffer->state.descriptors_dirty & VK_SHADER_STAGE_VERTEX_BIT ||
       cmd_buffer->state.push_constants_dirty & VK_SHADER_STAGE_VERTEX_BIT) {
      /* From the IVB PRM Vol. 2, Part 1, Section 3.2.1:
       *
       *    "A PIPE_CONTROL with Post-Sync Operation set to 1h and a depth
       *    stall needs to be sent just prior to any 3DSTATE_VS,
       *    3DSTATE_URB_VS, 3DSTATE_CONSTANT_VS,
       *    3DSTATE_BINDING_TABLE_POINTER_VS,
       *    3DSTATE_SAMPLER_STATE_POINTER_VS command.  Only one
       *    PIPE_CONTROL needs to be sent before any combination of VS
       *    associated 3DSTATE."
       */
      anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), pc) {
         pc.DepthStallEnable  = true;
         pc.PostSyncOperation = WriteImmediateData;
         pc.Address           =
            (struct anv_address) { &cmd_buffer->device->workaround_bo, 0 };
      }
   }
#endif

   /* We emit the binding tables and sampler tables first, then emit push
    * constants and then finally emit binding table and sampler table
    * pointers.  It has to happen in this order, since emitting the binding
    * tables may change the push constants (in case of storage images). After
    * emitting push constants, on SKL+ we have to emit the corresponding
    * 3DSTATE_BINDING_TABLE_POINTER_* for the push constants to take effect.
    */
   uint32_t dirty = 0;
   if (cmd_buffer->state.descriptors_dirty)
      dirty = gen7_cmd_buffer_flush_descriptor_sets(cmd_buffer);

   if (cmd_buffer->state.push_constants_dirty) {
#if GEN_GEN >= 9
      /* On Sky Lake and later, the binding table pointers commands are
       * what actually flush the changes to push constant state so we need
       * to dirty them so they get re-emitted below.
       */
      dirty |= cmd_buffer_flush_push_constants(cmd_buffer);
#else
      cmd_buffer_flush_push_constants(cmd_buffer);
#endif
   }

   if (dirty)
      gen7_cmd_buffer_emit_descriptor_pointers(cmd_buffer, dirty);

   if (cmd_buffer->state.dirty & ANV_CMD_DIRTY_DYNAMIC_VIEWPORT)
      gen8_cmd_buffer_emit_viewport(cmd_buffer);

   if (cmd_buffer->state.dirty & ANV_CMD_DIRTY_DYNAMIC_SCISSOR)
      gen7_cmd_buffer_emit_scissor(cmd_buffer);

   genX(cmd_buffer_flush_dynamic_state)(cmd_buffer);
}

static void
emit_base_vertex_instance_bo(struct anv_cmd_buffer *cmd_buffer,
                             struct anv_bo *bo, uint32_t offset)
{
   uint32_t *p = anv_batch_emitn(&cmd_buffer->batch, 5,
                                 GENX(3DSTATE_VERTEX_BUFFERS));

   GENX(VERTEX_BUFFER_STATE_pack)(&cmd_buffer->batch, p + 1,
      &(struct GENX(VERTEX_BUFFER_STATE)) {
         .VertexBufferIndex = 32, /* Reserved for this */
         .AddressModifyEnable = true,
         .BufferPitch = 0,
#if (GEN_GEN >= 8)
         .MemoryObjectControlState = GENX(MOCS),
         .BufferStartingAddress = { bo, offset },
         .BufferSize = 8
#else
         .VertexBufferMemoryObjectControlState = GENX(MOCS),
         .BufferStartingAddress = { bo, offset },
         .EndAddress = { bo, offset + 8 },
#endif
      });
}

static void
emit_base_vertex_instance(struct anv_cmd_buffer *cmd_buffer,
                          uint32_t base_vertex, uint32_t base_instance)
{
   struct anv_state id_state =
      anv_cmd_buffer_alloc_dynamic_state(cmd_buffer, 8, 4);

   ((uint32_t *)id_state.map)[0] = base_vertex;
   ((uint32_t *)id_state.map)[1] = base_instance;

   if (!cmd_buffer->device->info.has_llc)
      anv_state_clflush(id_state);

   emit_base_vertex_instance_bo(cmd_buffer,
      &cmd_buffer->device->dynamic_state_block_pool.bo, id_state.offset);
}

void genX(CmdDraw)(
    VkCommandBuffer                             commandBuffer,
    uint32_t                                    vertexCount,
    uint32_t                                    instanceCount,
    uint32_t                                    firstVertex,
    uint32_t                                    firstInstance)
{
   ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
   struct anv_pipeline *pipeline = cmd_buffer->state.pipeline;
   const struct brw_vs_prog_data *vs_prog_data = get_vs_prog_data(pipeline);

   genX(cmd_buffer_flush_state)(cmd_buffer);

   if (vs_prog_data->uses_basevertex || vs_prog_data->uses_baseinstance)
      emit_base_vertex_instance(cmd_buffer, firstVertex, firstInstance);

   anv_batch_emit(&cmd_buffer->batch, GENX(3DPRIMITIVE), prim) {
      prim.VertexAccessType         = SEQUENTIAL;
      prim.PrimitiveTopologyType    = pipeline->topology;
      prim.VertexCountPerInstance   = vertexCount;
      prim.StartVertexLocation      = firstVertex;
      prim.InstanceCount            = instanceCount;
      prim.StartInstanceLocation    = firstInstance;
      prim.BaseVertexLocation       = 0;
   }
}

void genX(CmdDrawIndexed)(
    VkCommandBuffer                             commandBuffer,
    uint32_t                                    indexCount,
    uint32_t                                    instanceCount,
    uint32_t                                    firstIndex,
    int32_t                                     vertexOffset,
    uint32_t                                    firstInstance)
{
   ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
   struct anv_pipeline *pipeline = cmd_buffer->state.pipeline;
   const struct brw_vs_prog_data *vs_prog_data = get_vs_prog_data(pipeline);

   genX(cmd_buffer_flush_state)(cmd_buffer);

   if (vs_prog_data->uses_basevertex || vs_prog_data->uses_baseinstance)
      emit_base_vertex_instance(cmd_buffer, vertexOffset, firstInstance);

   anv_batch_emit(&cmd_buffer->batch, GENX(3DPRIMITIVE), prim) {
      prim.VertexAccessType         = RANDOM;
      prim.PrimitiveTopologyType    = pipeline->topology;
      prim.VertexCountPerInstance   = indexCount;
      prim.StartVertexLocation      = firstIndex;
      prim.InstanceCount            = instanceCount;
      prim.StartInstanceLocation    = firstInstance;
      prim.BaseVertexLocation       = vertexOffset;
   }
}

/* Auto-Draw / Indirect Registers */
#define GEN7_3DPRIM_END_OFFSET          0x2420
#define GEN7_3DPRIM_START_VERTEX        0x2430
#define GEN7_3DPRIM_VERTEX_COUNT        0x2434
#define GEN7_3DPRIM_INSTANCE_COUNT      0x2438
#define GEN7_3DPRIM_START_INSTANCE      0x243C
#define GEN7_3DPRIM_BASE_VERTEX         0x2440

static void
emit_lrm(struct anv_batch *batch,
         uint32_t reg, struct anv_bo *bo, uint32_t offset)
{
   anv_batch_emit(batch, GENX(MI_LOAD_REGISTER_MEM), lrm) {
      lrm.RegisterAddress  = reg;
      lrm.MemoryAddress    = (struct anv_address) { bo, offset };
   }
}

static void
emit_lri(struct anv_batch *batch, uint32_t reg, uint32_t imm)
{
   anv_batch_emit(batch, GENX(MI_LOAD_REGISTER_IMM), lri) {
      lri.RegisterOffset   = reg;
      lri.DataDWord        = imm;
   }
}

void genX(CmdDrawIndirect)(
    VkCommandBuffer                             commandBuffer,
    VkBuffer                                    _buffer,
    VkDeviceSize                                offset,
    uint32_t                                    drawCount,
    uint32_t                                    stride)
{
   ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
   ANV_FROM_HANDLE(anv_buffer, buffer, _buffer);
   struct anv_pipeline *pipeline = cmd_buffer->state.pipeline;
   const struct brw_vs_prog_data *vs_prog_data = get_vs_prog_data(pipeline);
   struct anv_bo *bo = buffer->bo;
   uint32_t bo_offset = buffer->offset + offset;

   genX(cmd_buffer_flush_state)(cmd_buffer);

   if (vs_prog_data->uses_basevertex || vs_prog_data->uses_baseinstance)
      emit_base_vertex_instance_bo(cmd_buffer, bo, bo_offset + 8);

   emit_lrm(&cmd_buffer->batch, GEN7_3DPRIM_VERTEX_COUNT, bo, bo_offset);
   emit_lrm(&cmd_buffer->batch, GEN7_3DPRIM_INSTANCE_COUNT, bo, bo_offset + 4);
   emit_lrm(&cmd_buffer->batch, GEN7_3DPRIM_START_VERTEX, bo, bo_offset + 8);
   emit_lrm(&cmd_buffer->batch, GEN7_3DPRIM_START_INSTANCE, bo, bo_offset + 12);
   emit_lri(&cmd_buffer->batch, GEN7_3DPRIM_BASE_VERTEX, 0);

   anv_batch_emit(&cmd_buffer->batch, GENX(3DPRIMITIVE), prim) {
      prim.IndirectParameterEnable  = true;
      prim.VertexAccessType         = SEQUENTIAL;
      prim.PrimitiveTopologyType    = pipeline->topology;
   }
}

void genX(CmdDrawIndexedIndirect)(
    VkCommandBuffer                             commandBuffer,
    VkBuffer                                    _buffer,
    VkDeviceSize                                offset,
    uint32_t                                    drawCount,
    uint32_t                                    stride)
{
   ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
   ANV_FROM_HANDLE(anv_buffer, buffer, _buffer);
   struct anv_pipeline *pipeline = cmd_buffer->state.pipeline;
   const struct brw_vs_prog_data *vs_prog_data = get_vs_prog_data(pipeline);
   struct anv_bo *bo = buffer->bo;
   uint32_t bo_offset = buffer->offset + offset;

   genX(cmd_buffer_flush_state)(cmd_buffer);

   /* TODO: We need to stomp base vertex to 0 somehow */
   if (vs_prog_data->uses_basevertex || vs_prog_data->uses_baseinstance)
      emit_base_vertex_instance_bo(cmd_buffer, bo, bo_offset + 12);

   emit_lrm(&cmd_buffer->batch, GEN7_3DPRIM_VERTEX_COUNT, bo, bo_offset);
   emit_lrm(&cmd_buffer->batch, GEN7_3DPRIM_INSTANCE_COUNT, bo, bo_offset + 4);
   emit_lrm(&cmd_buffer->batch, GEN7_3DPRIM_START_VERTEX, bo, bo_offset + 8);
   emit_lrm(&cmd_buffer->batch, GEN7_3DPRIM_BASE_VERTEX, bo, bo_offset + 12);
   emit_lrm(&cmd_buffer->batch, GEN7_3DPRIM_START_INSTANCE, bo, bo_offset + 16);

   anv_batch_emit(&cmd_buffer->batch, GENX(3DPRIMITIVE), prim) {
      prim.IndirectParameterEnable  = true;
      prim.VertexAccessType         = RANDOM;
      prim.PrimitiveTopologyType    = pipeline->topology;
   }
}

#if GEN_GEN == 7

static bool
verify_cmd_parser(const struct anv_device *device,
                  int required_version,
                  const char *function)
{
   if (device->instance->physicalDevice.cmd_parser_version < required_version) {
      vk_errorf(VK_ERROR_FEATURE_NOT_PRESENT,
                "cmd parser version %d is required for %s",
                required_version, function);
      return false;
   } else {
      return true;
   }
}

#endif

void genX(CmdDispatch)(
    VkCommandBuffer                             commandBuffer,
    uint32_t                                    x,
    uint32_t                                    y,
    uint32_t                                    z)
{
   ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
   struct anv_pipeline *pipeline = cmd_buffer->state.compute_pipeline;
   const struct brw_cs_prog_data *prog_data = get_cs_prog_data(pipeline);

   if (prog_data->uses_num_work_groups) {
      struct anv_state state =
         anv_cmd_buffer_alloc_dynamic_state(cmd_buffer, 12, 4);
      uint32_t *sizes = state.map;
      sizes[0] = x;
      sizes[1] = y;
      sizes[2] = z;
      if (!cmd_buffer->device->info.has_llc)
         anv_state_clflush(state);
      cmd_buffer->state.num_workgroups_offset = state.offset;
      cmd_buffer->state.num_workgroups_bo =
         &cmd_buffer->device->dynamic_state_block_pool.bo;
   }

   genX(cmd_buffer_flush_compute_state)(cmd_buffer);

   anv_batch_emit(&cmd_buffer->batch, GENX(GPGPU_WALKER), ggw) {
      ggw.SIMDSize                     = prog_data->simd_size / 16;
      ggw.ThreadDepthCounterMaximum    = 0;
      ggw.ThreadHeightCounterMaximum   = 0;
      ggw.ThreadWidthCounterMaximum    = pipeline->cs_thread_width_max - 1;
      ggw.ThreadGroupIDXDimension      = x;
      ggw.ThreadGroupIDYDimension      = y;
      ggw.ThreadGroupIDZDimension      = z;
      ggw.RightExecutionMask           = pipeline->cs_right_mask;
      ggw.BottomExecutionMask          = 0xffffffff;
   }

   anv_batch_emit(&cmd_buffer->batch, GENX(MEDIA_STATE_FLUSH), msf);
}

#define GPGPU_DISPATCHDIMX 0x2500
#define GPGPU_DISPATCHDIMY 0x2504
#define GPGPU_DISPATCHDIMZ 0x2508

#define MI_PREDICATE_SRC0  0x2400
#define MI_PREDICATE_SRC1  0x2408

void genX(CmdDispatchIndirect)(
    VkCommandBuffer                             commandBuffer,
    VkBuffer                                    _buffer,
    VkDeviceSize                                offset)
{
   ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
   ANV_FROM_HANDLE(anv_buffer, buffer, _buffer);
   struct anv_pipeline *pipeline = cmd_buffer->state.compute_pipeline;
   const struct brw_cs_prog_data *prog_data = get_cs_prog_data(pipeline);
   struct anv_bo *bo = buffer->bo;
   uint32_t bo_offset = buffer->offset + offset;
   struct anv_batch *batch = &cmd_buffer->batch;

#if GEN_GEN == 7
   /* Linux 4.4 added command parser version 5 which allows the GPGPU
    * indirect dispatch registers to be written.
    */
   if (!verify_cmd_parser(cmd_buffer->device, 5, "vkCmdDispatchIndirect"))
      return;
#endif

   if (prog_data->uses_num_work_groups) {
      cmd_buffer->state.num_workgroups_offset = bo_offset;
      cmd_buffer->state.num_workgroups_bo = bo;
   }

   genX(cmd_buffer_flush_compute_state)(cmd_buffer);

   emit_lrm(batch, GPGPU_DISPATCHDIMX, bo, bo_offset);
   emit_lrm(batch, GPGPU_DISPATCHDIMY, bo, bo_offset + 4);
   emit_lrm(batch, GPGPU_DISPATCHDIMZ, bo, bo_offset + 8);

#if GEN_GEN <= 7
   /* Clear upper 32-bits of SRC0 and all 64-bits of SRC1 */
   emit_lri(batch, MI_PREDICATE_SRC0 + 4, 0);
   emit_lri(batch, MI_PREDICATE_SRC1 + 0, 0);
   emit_lri(batch, MI_PREDICATE_SRC1 + 4, 0);

   /* Load compute_dispatch_indirect_x_size into SRC0 */
   emit_lrm(batch, MI_PREDICATE_SRC0, bo, bo_offset + 0);

   /* predicate = (compute_dispatch_indirect_x_size == 0); */
   anv_batch_emit(batch, GENX(MI_PREDICATE), mip) {
      mip.LoadOperation    = LOAD_LOAD;
      mip.CombineOperation = COMBINE_SET;
      mip.CompareOperation = COMPARE_SRCS_EQUAL;
   }

   /* Load compute_dispatch_indirect_y_size into SRC0 */
   emit_lrm(batch, MI_PREDICATE_SRC0, bo, bo_offset + 4);

   /* predicate |= (compute_dispatch_indirect_y_size == 0); */
   anv_batch_emit(batch, GENX(MI_PREDICATE), mip) {
      mip.LoadOperation    = LOAD_LOAD;
      mip.CombineOperation = COMBINE_OR;
      mip.CompareOperation = COMPARE_SRCS_EQUAL;
   }

   /* Load compute_dispatch_indirect_z_size into SRC0 */
   emit_lrm(batch, MI_PREDICATE_SRC0, bo, bo_offset + 8);

   /* predicate |= (compute_dispatch_indirect_z_size == 0); */
   anv_batch_emit(batch, GENX(MI_PREDICATE), mip) {
      mip.LoadOperation    = LOAD_LOAD;
      mip.CombineOperation = COMBINE_OR;
      mip.CompareOperation = COMPARE_SRCS_EQUAL;
   }

   /* predicate = !predicate; */
#define COMPARE_FALSE                           1
   anv_batch_emit(batch, GENX(MI_PREDICATE), mip) {
      mip.LoadOperation    = LOAD_LOADINV;
      mip.CombineOperation = COMBINE_OR;
      mip.CompareOperation = COMPARE_FALSE;
   }
#endif

   anv_batch_emit(batch, GENX(GPGPU_WALKER), ggw) {
      ggw.IndirectParameterEnable      = true;
      ggw.PredicateEnable              = GEN_GEN <= 7;
      ggw.SIMDSize                     = prog_data->simd_size / 16;
      ggw.ThreadDepthCounterMaximum    = 0;
      ggw.ThreadHeightCounterMaximum   = 0;
      ggw.ThreadWidthCounterMaximum    = pipeline->cs_thread_width_max - 1;
      ggw.RightExecutionMask           = pipeline->cs_right_mask;
      ggw.BottomExecutionMask          = 0xffffffff;
   }

   anv_batch_emit(batch, GENX(MEDIA_STATE_FLUSH), msf);
}

static void
flush_pipeline_before_pipeline_select(struct anv_cmd_buffer *cmd_buffer,
                                      uint32_t pipeline)
{
#if GEN_GEN >= 8 && GEN_GEN < 10
   /* From the Broadwell PRM, Volume 2a: Instructions, PIPELINE_SELECT:
    *
    *   Software must clear the COLOR_CALC_STATE Valid field in
    *   3DSTATE_CC_STATE_POINTERS command prior to send a PIPELINE_SELECT
    *   with Pipeline Select set to GPGPU.
    *
    * The internal hardware docs recommend the same workaround for Gen9
    * hardware too.
    */
   if (pipeline == GPGPU)
      anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_CC_STATE_POINTERS), t);
#elif GEN_GEN <= 7
      /* From "BXML » GT » MI » vol1a GPU Overview » [Instruction]
       * PIPELINE_SELECT [DevBWR+]":
       *
       *   Project: DEVSNB+
       *
       *   Software must ensure all the write caches are flushed through a
       *   stalling PIPE_CONTROL command followed by another PIPE_CONTROL
       *   command to invalidate read only caches prior to programming
       *   MI_PIPELINE_SELECT command to change the Pipeline Select Mode.
       */
      anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), pc) {
         pc.RenderTargetCacheFlushEnable  = true;
         pc.DepthCacheFlushEnable         = true;
         pc.DCFlushEnable                 = true;
         pc.PostSyncOperation             = NoWrite;
         pc.CommandStreamerStallEnable    = true;
      }

      anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), pc) {
         pc.TextureCacheInvalidationEnable   = true;
         pc.ConstantCacheInvalidationEnable  = true;
         pc.StateCacheInvalidationEnable     = true;
         pc.InstructionCacheInvalidateEnable = true;
         pc.PostSyncOperation                = NoWrite;
      }
#endif
}

void
genX(flush_pipeline_select_3d)(struct anv_cmd_buffer *cmd_buffer)
{
   if (cmd_buffer->state.current_pipeline != _3D) {
      flush_pipeline_before_pipeline_select(cmd_buffer, _3D);

      anv_batch_emit(&cmd_buffer->batch, GENX(PIPELINE_SELECT), ps) {
#if GEN_GEN >= 9
         ps.MaskBits = 3;
#endif
         ps.PipelineSelection = _3D;
      }

      cmd_buffer->state.current_pipeline = _3D;
   }
}

void
genX(flush_pipeline_select_gpgpu)(struct anv_cmd_buffer *cmd_buffer)
{
   if (cmd_buffer->state.current_pipeline != GPGPU) {
      flush_pipeline_before_pipeline_select(cmd_buffer, GPGPU);

      anv_batch_emit(&cmd_buffer->batch, GENX(PIPELINE_SELECT), ps) {
#if GEN_GEN >= 9
         ps.MaskBits = 3;
#endif
         ps.PipelineSelection = GPGPU;
      }

      cmd_buffer->state.current_pipeline = GPGPU;
   }
}

struct anv_state
genX(cmd_buffer_alloc_null_surface_state)(struct anv_cmd_buffer *cmd_buffer,
                                          struct anv_framebuffer *fb)
{
   struct anv_state state =
      anv_state_stream_alloc(&cmd_buffer->surface_state_stream, 64, 64);

   struct GENX(RENDER_SURFACE_STATE) null_ss = {
      .SurfaceType = SURFTYPE_NULL,
      .SurfaceArray = fb->layers > 0,
      .SurfaceFormat = ISL_FORMAT_R8G8B8A8_UNORM,
#if GEN_GEN >= 8
      .TileMode = YMAJOR,
#else
      .TiledSurface = true,
#endif
      .Width = fb->width - 1,
      .Height = fb->height - 1,
      .Depth = fb->layers - 1,
      .RenderTargetViewExtent = fb->layers - 1,
   };

   GENX(RENDER_SURFACE_STATE_pack)(NULL, state.map, &null_ss);

   if (!cmd_buffer->device->info.has_llc)
      anv_state_clflush(state);

   return state;
}

static void
cmd_buffer_emit_depth_stencil(struct anv_cmd_buffer *cmd_buffer)
{
   struct anv_device *device = cmd_buffer->device;
   const struct anv_framebuffer *fb = cmd_buffer->state.framebuffer;
   const struct anv_image_view *iview =
      anv_cmd_buffer_get_depth_stencil_view(cmd_buffer);
   const struct anv_image *image = iview ? iview->image : NULL;
   const bool has_depth = image && (image->aspects & VK_IMAGE_ASPECT_DEPTH_BIT);
   const bool has_stencil =
      image && (image->aspects & VK_IMAGE_ASPECT_STENCIL_BIT);

   /* FIXME: Implement the PMA stall W/A */
   /* FIXME: Width and Height are wrong */

   /* Emit 3DSTATE_DEPTH_BUFFER */
   if (has_depth) {
      anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_DEPTH_BUFFER), db) {
         db.SurfaceType                   = SURFTYPE_2D;
         db.DepthWriteEnable              = true;
         db.StencilWriteEnable            = has_stencil;
         db.HierarchicalDepthBufferEnable = false;

         db.SurfaceFormat = isl_surf_get_depth_format(&device->isl_dev,
                                                      &image->depth_surface.isl);

         db.SurfaceBaseAddress = (struct anv_address) {
            .bo = image->bo,
            .offset = image->offset + image->depth_surface.offset,
         };
         db.DepthBufferObjectControlState = GENX(MOCS),

         db.SurfacePitch         = image->depth_surface.isl.row_pitch - 1;
         db.Height               = fb->height - 1;
         db.Width                = fb->width - 1;
         db.LOD                  = 0;
         db.Depth                = 1 - 1;
         db.MinimumArrayElement  = 0;

#if GEN_GEN >= 8
         db.SurfaceQPitch =
            isl_surf_get_array_pitch_el_rows(&image->depth_surface.isl) >> 2,
#endif
         db.RenderTargetViewExtent = 1 - 1;
      }
   } else {
      /* Even when no depth buffer is present, the hardware requires that
       * 3DSTATE_DEPTH_BUFFER be programmed correctly. The Broadwell PRM says:
       *
       *    If a null depth buffer is bound, the driver must instead bind depth as:
       *       3DSTATE_DEPTH.SurfaceType = SURFTYPE_2D
       *       3DSTATE_DEPTH.Width = 1
       *       3DSTATE_DEPTH.Height = 1
       *       3DSTATE_DEPTH.SuraceFormat = D16_UNORM
       *       3DSTATE_DEPTH.SurfaceBaseAddress = 0
       *       3DSTATE_DEPTH.HierarchicalDepthBufferEnable = 0
       *       3DSTATE_WM_DEPTH_STENCIL.DepthTestEnable = 0
       *       3DSTATE_WM_DEPTH_STENCIL.DepthBufferWriteEnable = 0
       *
       * The PRM is wrong, though. The width and height must be programmed to
       * actual framebuffer's width and height, even when neither depth buffer
       * nor stencil buffer is present.  Also, D16_UNORM is not allowed to
       * be combined with a stencil buffer so we use D32_FLOAT instead.
       */
      anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_DEPTH_BUFFER), db) {
         db.SurfaceType          = SURFTYPE_2D;
         db.SurfaceFormat        = D32_FLOAT;
         db.Width                = fb->width - 1;
         db.Height               = fb->height - 1;
         db.StencilWriteEnable   = has_stencil;
      }
   }

   /* Emit 3DSTATE_STENCIL_BUFFER */
   if (has_stencil) {
      anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_STENCIL_BUFFER), sb) {
#if GEN_GEN >= 8 || GEN_IS_HASWELL
         sb.StencilBufferEnable = true,
#endif
         sb.StencilBufferObjectControlState = GENX(MOCS),

         /* Stencil buffers have strange pitch. The PRM says:
          *
          *    The pitch must be set to 2x the value computed based on width,
          *    as the stencil buffer is stored with two rows interleaved.
          */
         sb.SurfacePitch = 2 * image->stencil_surface.isl.row_pitch - 1,

#if GEN_GEN >= 8
         sb.SurfaceQPitch = isl_surf_get_array_pitch_el_rows(&image->stencil_surface.isl) >> 2,
#endif
         sb.SurfaceBaseAddress = (struct anv_address) {
            .bo = image->bo,
            .offset = image->offset + image->stencil_surface.offset,
         };
      }
   } else {
      anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_STENCIL_BUFFER), sb);
   }

   /* Disable hierarchial depth buffers. */
   anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_HIER_DEPTH_BUFFER), hz);

   /* Clear the clear params. */
   anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_CLEAR_PARAMS), cp);
}

/**
 * @see anv_cmd_buffer_set_subpass()
 */
void
genX(cmd_buffer_set_subpass)(struct anv_cmd_buffer *cmd_buffer,
                             struct anv_subpass *subpass)
{
   cmd_buffer->state.subpass = subpass;

   cmd_buffer->state.descriptors_dirty |= VK_SHADER_STAGE_FRAGMENT_BIT;

   cmd_buffer_emit_depth_stencil(cmd_buffer);
}

void genX(CmdBeginRenderPass)(
    VkCommandBuffer                             commandBuffer,
    const VkRenderPassBeginInfo*                pRenderPassBegin,
    VkSubpassContents                           contents)
{
   ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
   ANV_FROM_HANDLE(anv_render_pass, pass, pRenderPassBegin->renderPass);
   ANV_FROM_HANDLE(anv_framebuffer, framebuffer, pRenderPassBegin->framebuffer);

   cmd_buffer->state.framebuffer = framebuffer;
   cmd_buffer->state.pass = pass;
   anv_cmd_state_setup_attachments(cmd_buffer, pRenderPassBegin);

   genX(flush_pipeline_select_3d)(cmd_buffer);

   const VkRect2D *render_area = &pRenderPassBegin->renderArea;

   anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_DRAWING_RECTANGLE), r) {
      r.ClippedDrawingRectangleYMin = MAX2(render_area->offset.y, 0);
      r.ClippedDrawingRectangleXMin = MAX2(render_area->offset.x, 0);
      r.ClippedDrawingRectangleYMax =
         render_area->offset.y + render_area->extent.height - 1;
      r.ClippedDrawingRectangleXMax =
         render_area->offset.x + render_area->extent.width - 1;
      r.DrawingRectangleOriginY     = 0;
      r.DrawingRectangleOriginX     = 0;
   }

   genX(cmd_buffer_set_subpass)(cmd_buffer, pass->subpasses);
   anv_cmd_buffer_clear_subpass(cmd_buffer);
}

void genX(CmdNextSubpass)(
    VkCommandBuffer                             commandBuffer,
    VkSubpassContents                           contents)
{
   ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);

   assert(cmd_buffer->level == VK_COMMAND_BUFFER_LEVEL_PRIMARY);

   anv_cmd_buffer_resolve_subpass(cmd_buffer);
   genX(cmd_buffer_set_subpass)(cmd_buffer, cmd_buffer->state.subpass + 1);
   anv_cmd_buffer_clear_subpass(cmd_buffer);
}

void genX(CmdEndRenderPass)(
    VkCommandBuffer                             commandBuffer)
{
   ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);

   anv_cmd_buffer_resolve_subpass(cmd_buffer);
}

static void
emit_ps_depth_count(struct anv_batch *batch,
                    struct anv_bo *bo, uint32_t offset)
{
   anv_batch_emit(batch, GENX(PIPE_CONTROL), pc) {
      pc.DestinationAddressType  = DAT_PPGTT;
      pc.PostSyncOperation       = WritePSDepthCount;
      pc.DepthStallEnable        = true;
      pc.Address                 = (struct anv_address) { bo, offset };
   }
}

static void
emit_query_availability(struct anv_batch *batch,
                        struct anv_bo *bo, uint32_t offset)
{
   anv_batch_emit(batch, GENX(PIPE_CONTROL), pc) {
      pc.DestinationAddressType  = DAT_PPGTT;
      pc.PostSyncOperation       = WriteImmediateData;
      pc.Address                 = (struct anv_address) { bo, offset };
      pc.ImmediateData           = 1;
   }
}

void genX(CmdBeginQuery)(
    VkCommandBuffer                             commandBuffer,
    VkQueryPool                                 queryPool,
    uint32_t                                    query,
    VkQueryControlFlags                         flags)
{
   ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
   ANV_FROM_HANDLE(anv_query_pool, pool, queryPool);

   /* Workaround: When meta uses the pipeline with the VS disabled, it seems
    * that the pipelining of the depth write breaks. What we see is that
    * samples from the render pass clear leaks into the first query
    * immediately after the clear. Doing a pipecontrol with a post-sync
    * operation and DepthStallEnable seems to work around the issue.
    */
   if (cmd_buffer->state.need_query_wa) {
      cmd_buffer->state.need_query_wa = false;
      anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), pc) {
         pc.DepthCacheFlushEnable   = true;
         pc.DepthStallEnable        = true;
      }
   }

   switch (pool->type) {
   case VK_QUERY_TYPE_OCCLUSION:
      emit_ps_depth_count(&cmd_buffer->batch, &pool->bo,
                          query * sizeof(struct anv_query_pool_slot));
      break;

   case VK_QUERY_TYPE_PIPELINE_STATISTICS:
   default:
      unreachable("");
   }
}

void genX(CmdEndQuery)(
    VkCommandBuffer                             commandBuffer,
    VkQueryPool                                 queryPool,
    uint32_t                                    query)
{
   ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
   ANV_FROM_HANDLE(anv_query_pool, pool, queryPool);

   switch (pool->type) {
   case VK_QUERY_TYPE_OCCLUSION:
      emit_ps_depth_count(&cmd_buffer->batch, &pool->bo,
                          query * sizeof(struct anv_query_pool_slot) + 8);

      emit_query_availability(&cmd_buffer->batch, &pool->bo,
                              query * sizeof(struct anv_query_pool_slot) + 16);
      break;

   case VK_QUERY_TYPE_PIPELINE_STATISTICS:
   default:
      unreachable("");
   }
}

#define TIMESTAMP 0x2358

void genX(CmdWriteTimestamp)(
    VkCommandBuffer                             commandBuffer,
    VkPipelineStageFlagBits                     pipelineStage,
    VkQueryPool                                 queryPool,
    uint32_t                                    query)
{
   ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
   ANV_FROM_HANDLE(anv_query_pool, pool, queryPool);
   uint32_t offset = query * sizeof(struct anv_query_pool_slot);

   assert(pool->type == VK_QUERY_TYPE_TIMESTAMP);

   switch (pipelineStage) {
   case VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT:
      anv_batch_emit(&cmd_buffer->batch, GENX(MI_STORE_REGISTER_MEM), srm) {
         srm.RegisterAddress  = TIMESTAMP;
         srm.MemoryAddress    = (struct anv_address) { &pool->bo, offset };
      }
      anv_batch_emit(&cmd_buffer->batch, GENX(MI_STORE_REGISTER_MEM), srm) {
         srm.RegisterAddress  = TIMESTAMP + 4;
         srm.MemoryAddress    = (struct anv_address) { &pool->bo, offset + 4 };
      }
      break;

   default:
      /* Everything else is bottom-of-pipe */
      anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), pc) {
         pc.DestinationAddressType  = DAT_PPGTT,
         pc.PostSyncOperation       = WriteTimestamp,
         pc.Address = (struct anv_address) { &pool->bo, offset };
      }
      break;
   }

   emit_query_availability(&cmd_buffer->batch, &pool->bo, query + 16);
}

#if GEN_GEN > 7 || GEN_IS_HASWELL

#define alu_opcode(v)   __gen_uint((v),  20, 31)
#define alu_operand1(v) __gen_uint((v),  10, 19)
#define alu_operand2(v) __gen_uint((v),   0,  9)
#define alu(opcode, operand1, operand2) \
   alu_opcode(opcode) | alu_operand1(operand1) | alu_operand2(operand2)

#define OPCODE_NOOP      0x000
#define OPCODE_LOAD      0x080
#define OPCODE_LOADINV   0x480
#define OPCODE_LOAD0     0x081
#define OPCODE_LOAD1     0x481
#define OPCODE_ADD       0x100
#define OPCODE_SUB       0x101
#define OPCODE_AND       0x102
#define OPCODE_OR        0x103
#define OPCODE_XOR       0x104
#define OPCODE_STORE     0x180
#define OPCODE_STOREINV  0x580

#define OPERAND_R0   0x00
#define OPERAND_R1   0x01
#define OPERAND_R2   0x02
#define OPERAND_R3   0x03
#define OPERAND_R4   0x04
#define OPERAND_SRCA 0x20
#define OPERAND_SRCB 0x21
#define OPERAND_ACCU 0x31
#define OPERAND_ZF   0x32
#define OPERAND_CF   0x33

#define CS_GPR(n) (0x2600 + (n) * 8)

static void
emit_load_alu_reg_u64(struct anv_batch *batch, uint32_t reg,
                      struct anv_bo *bo, uint32_t offset)
{
   anv_batch_emit(batch, GENX(MI_LOAD_REGISTER_MEM), lrm) {
      lrm.RegisterAddress  = reg,
      lrm.MemoryAddress    = (struct anv_address) { bo, offset };
   }
   anv_batch_emit(batch, GENX(MI_LOAD_REGISTER_MEM), lrm) {
      lrm.RegisterAddress  = reg + 4;
      lrm.MemoryAddress    = (struct anv_address) { bo, offset + 4 };
   }
}

static void
store_query_result(struct anv_batch *batch, uint32_t reg,
                   struct anv_bo *bo, uint32_t offset, VkQueryResultFlags flags)
{
   anv_batch_emit(batch, GENX(MI_STORE_REGISTER_MEM), srm) {
      srm.RegisterAddress  = reg;
      srm.MemoryAddress    = (struct anv_address) { bo, offset };
   }

   if (flags & VK_QUERY_RESULT_64_BIT) {
      anv_batch_emit(batch, GENX(MI_STORE_REGISTER_MEM), srm) {
         srm.RegisterAddress  = reg + 4;
         srm.MemoryAddress    = (struct anv_address) { bo, offset + 4 };
      }
   }
}

void genX(CmdCopyQueryPoolResults)(
    VkCommandBuffer                             commandBuffer,
    VkQueryPool                                 queryPool,
    uint32_t                                    firstQuery,
    uint32_t                                    queryCount,
    VkBuffer                                    destBuffer,
    VkDeviceSize                                destOffset,
    VkDeviceSize                                destStride,
    VkQueryResultFlags                          flags)
{
   ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
   ANV_FROM_HANDLE(anv_query_pool, pool, queryPool);
   ANV_FROM_HANDLE(anv_buffer, buffer, destBuffer);
   uint32_t slot_offset, dst_offset;

   if (flags & VK_QUERY_RESULT_WAIT_BIT) {
      anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), pc) {
         pc.CommandStreamerStallEnable = true;
         pc.StallAtPixelScoreboard     = true;
      }
   }

   dst_offset = buffer->offset + destOffset;
   for (uint32_t i = 0; i < queryCount; i++) {

      slot_offset = (firstQuery + i) * sizeof(struct anv_query_pool_slot);
      switch (pool->type) {
      case VK_QUERY_TYPE_OCCLUSION:
         emit_load_alu_reg_u64(&cmd_buffer->batch,
                               CS_GPR(0), &pool->bo, slot_offset);
         emit_load_alu_reg_u64(&cmd_buffer->batch,
                               CS_GPR(1), &pool->bo, slot_offset + 8);

         /* FIXME: We need to clamp the result for 32 bit. */

         uint32_t *dw = anv_batch_emitn(&cmd_buffer->batch, 5, GENX(MI_MATH));
         dw[1] = alu(OPCODE_LOAD, OPERAND_SRCA, OPERAND_R1);
         dw[2] = alu(OPCODE_LOAD, OPERAND_SRCB, OPERAND_R0);
         dw[3] = alu(OPCODE_SUB, 0, 0);
         dw[4] = alu(OPCODE_STORE, OPERAND_R2, OPERAND_ACCU);
         break;

      case VK_QUERY_TYPE_TIMESTAMP:
         emit_load_alu_reg_u64(&cmd_buffer->batch,
                               CS_GPR(2), &pool->bo, slot_offset);
         break;

      default:
         unreachable("unhandled query type");
      }

      store_query_result(&cmd_buffer->batch,
                         CS_GPR(2), buffer->bo, dst_offset, flags);

      if (flags & VK_QUERY_RESULT_WITH_AVAILABILITY_BIT) {
         emit_load_alu_reg_u64(&cmd_buffer->batch, CS_GPR(0),
                               &pool->bo, slot_offset + 16);
         if (flags & VK_QUERY_RESULT_64_BIT)
            store_query_result(&cmd_buffer->batch,
                               CS_GPR(0), buffer->bo, dst_offset + 8, flags);
         else
            store_query_result(&cmd_buffer->batch,
                               CS_GPR(0), buffer->bo, dst_offset + 4, flags);
      }

      dst_offset += destStride;
   }
}

#endif