aboutsummaryrefslogtreecommitdiffstats
path: root/src/mesa/drivers/dri/i965/brw_draw.c
blob: 8536c040109bd3053a340d12854c653d45e0485b (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
/*
 * Copyright 2003 VMware, Inc.
 * All Rights Reserved.
 *
 * 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 VMWARE AND/OR ITS SUPPLIERS 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 <sys/errno.h>

#include "main/arrayobj.h"
#include "main/blend.h"
#include "main/context.h"
#include "main/condrender.h"
#include "main/samplerobj.h"
#include "main/state.h"
#include "main/enums.h"
#include "main/macros.h"
#include "main/transformfeedback.h"
#include "main/framebuffer.h"
#include "main/varray.h"
#include "tnl/tnl.h"
#include "vbo/vbo.h"
#include "swrast/swrast.h"
#include "swrast_setup/swrast_setup.h"
#include "drivers/common/meta.h"
#include "util/bitscan.h"
#include "util/bitset.h"

#include "brw_blorp.h"
#include "brw_draw.h"
#include "brw_defines.h"
#include "compiler/brw_eu_defines.h"
#include "brw_context.h"
#include "brw_state.h"

#include "intel_batchbuffer.h"
#include "intel_buffers.h"
#include "intel_fbo.h"
#include "intel_mipmap_tree.h"
#include "intel_buffer_objects.h"

#define FILE_DEBUG_FLAG DEBUG_PRIMS


static const GLenum reduced_prim[GL_POLYGON+1] = {
   [GL_POINTS] = GL_POINTS,
   [GL_LINES] = GL_LINES,
   [GL_LINE_LOOP] = GL_LINES,
   [GL_LINE_STRIP] = GL_LINES,
   [GL_TRIANGLES] = GL_TRIANGLES,
   [GL_TRIANGLE_STRIP] = GL_TRIANGLES,
   [GL_TRIANGLE_FAN] = GL_TRIANGLES,
   [GL_QUADS] = GL_TRIANGLES,
   [GL_QUAD_STRIP] = GL_TRIANGLES,
   [GL_POLYGON] = GL_TRIANGLES
};

/* When the primitive changes, set a state bit and re-validate.  Not
 * the nicest and would rather deal with this by having all the
 * programs be immune to the active primitive (ie. cope with all
 * possibilities).  That may not be realistic however.
 */
static void
brw_set_prim(struct brw_context *brw, const struct _mesa_prim *prim)
{
   struct gl_context *ctx = &brw->ctx;
   uint32_t hw_prim = get_hw_prim_for_gl_prim(prim->mode);

   DBG("PRIM: %s\n", _mesa_enum_to_string(prim->mode));

   /* Slight optimization to avoid the GS program when not needed:
    */
   if (prim->mode == GL_QUAD_STRIP &&
       ctx->Light.ShadeModel != GL_FLAT &&
       ctx->Polygon.FrontMode == GL_FILL &&
       ctx->Polygon.BackMode == GL_FILL)
      hw_prim = _3DPRIM_TRISTRIP;

   if (prim->mode == GL_QUADS && prim->count == 4 &&
       ctx->Light.ShadeModel != GL_FLAT &&
       ctx->Polygon.FrontMode == GL_FILL &&
       ctx->Polygon.BackMode == GL_FILL) {
      hw_prim = _3DPRIM_TRIFAN;
   }

   if (hw_prim != brw->primitive) {
      brw->primitive = hw_prim;
      brw->ctx.NewDriverState |= BRW_NEW_PRIMITIVE;

      if (reduced_prim[prim->mode] != brw->reduced_primitive) {
         brw->reduced_primitive = reduced_prim[prim->mode];
         brw->ctx.NewDriverState |= BRW_NEW_REDUCED_PRIMITIVE;
      }
   }
}

static void
gen6_set_prim(struct brw_context *brw, const struct _mesa_prim *prim)
{
   const struct gl_context *ctx = &brw->ctx;
   uint32_t hw_prim;

   DBG("PRIM: %s\n", _mesa_enum_to_string(prim->mode));

   if (prim->mode == GL_PATCHES) {
      hw_prim = _3DPRIM_PATCHLIST(ctx->TessCtrlProgram.patch_vertices);
   } else {
      hw_prim = get_hw_prim_for_gl_prim(prim->mode);
   }

   if (hw_prim != brw->primitive) {
      brw->primitive = hw_prim;
      brw->ctx.NewDriverState |= BRW_NEW_PRIMITIVE;
      if (prim->mode == GL_PATCHES)
         brw->ctx.NewDriverState |= BRW_NEW_PATCH_PRIMITIVE;
   }
}


/**
 * The hardware is capable of removing dangling vertices on its own; however,
 * prior to Gen6, we sometimes convert quads into trifans (and quad strips
 * into tristrips), since pre-Gen6 hardware requires a GS to render quads.
 * This function manually trims dangling vertices from a draw call involving
 * quads so that those dangling vertices won't get drawn when we convert to
 * trifans/tristrips.
 */
static GLuint
trim(GLenum prim, GLuint length)
{
   if (prim == GL_QUAD_STRIP)
      return length > 3 ? (length - length % 2) : 0;
   else if (prim == GL_QUADS)
      return length - length % 4;
   else
      return length;
}


static void
brw_emit_prim(struct brw_context *brw,
              const struct _mesa_prim *prim,
              uint32_t hw_prim,
              struct brw_transform_feedback_object *xfb_obj,
              unsigned stream)
{
   const struct gen_device_info *devinfo = &brw->screen->devinfo;
   int verts_per_instance;
   int vertex_access_type;
   int indirect_flag;

   DBG("PRIM: %s %d %d\n", _mesa_enum_to_string(prim->mode),
       prim->start, prim->count);

   int start_vertex_location = prim->start;
   int base_vertex_location = prim->basevertex;

   if (prim->indexed) {
      vertex_access_type = devinfo->gen >= 7 ?
         GEN7_3DPRIM_VERTEXBUFFER_ACCESS_RANDOM :
         GEN4_3DPRIM_VERTEXBUFFER_ACCESS_RANDOM;
      start_vertex_location += brw->ib.start_vertex_offset;
      base_vertex_location += brw->vb.start_vertex_bias;
   } else {
      vertex_access_type = devinfo->gen >= 7 ?
         GEN7_3DPRIM_VERTEXBUFFER_ACCESS_SEQUENTIAL :
         GEN4_3DPRIM_VERTEXBUFFER_ACCESS_SEQUENTIAL;
      start_vertex_location += brw->vb.start_vertex_bias;
   }

   /* We only need to trim the primitive count on pre-Gen6. */
   if (devinfo->gen < 6)
      verts_per_instance = trim(prim->mode, prim->count);
   else
      verts_per_instance = prim->count;

   /* If nothing to emit, just return. */
   if (verts_per_instance == 0 && !prim->is_indirect && !xfb_obj)
      return;

   /* If we're set to always flush, do it before and after the primitive emit.
    * We want to catch both missed flushes that hurt instruction/state cache
    * and missed flushes of the render cache as it heads to other parts of
    * the besides the draw code.
    */
   if (brw->always_flush_cache)
      brw_emit_mi_flush(brw);

   /* If indirect, emit a bunch of loads from the indirect BO. */
   if (xfb_obj) {
      indirect_flag = GEN7_3DPRIM_INDIRECT_PARAMETER_ENABLE;

      brw_load_register_mem(brw, GEN7_3DPRIM_VERTEX_COUNT,
                            xfb_obj->prim_count_bo,
                            stream * sizeof(uint32_t));
      BEGIN_BATCH(9);
      OUT_BATCH(MI_LOAD_REGISTER_IMM | (9 - 2));
      OUT_BATCH(GEN7_3DPRIM_INSTANCE_COUNT);
      OUT_BATCH(prim->num_instances);
      OUT_BATCH(GEN7_3DPRIM_START_VERTEX);
      OUT_BATCH(0);
      OUT_BATCH(GEN7_3DPRIM_BASE_VERTEX);
      OUT_BATCH(0);
      OUT_BATCH(GEN7_3DPRIM_START_INSTANCE);
      OUT_BATCH(0);
      ADVANCE_BATCH();
   } else if (prim->is_indirect) {
      struct gl_buffer_object *indirect_buffer = brw->ctx.DrawIndirectBuffer;
      struct brw_bo *bo = intel_bufferobj_buffer(brw,
            intel_buffer_object(indirect_buffer),
            prim->indirect_offset, 5 * sizeof(GLuint), false);

      indirect_flag = GEN7_3DPRIM_INDIRECT_PARAMETER_ENABLE;

      brw_load_register_mem(brw, GEN7_3DPRIM_VERTEX_COUNT, bo,
                            prim->indirect_offset + 0);
      brw_load_register_mem(brw, GEN7_3DPRIM_INSTANCE_COUNT, bo,
                            prim->indirect_offset + 4);

      brw_load_register_mem(brw, GEN7_3DPRIM_START_VERTEX, bo,
                            prim->indirect_offset + 8);
      if (prim->indexed) {
         brw_load_register_mem(brw, GEN7_3DPRIM_BASE_VERTEX, bo,
                               prim->indirect_offset + 12);
         brw_load_register_mem(brw, GEN7_3DPRIM_START_INSTANCE, bo,
                               prim->indirect_offset + 16);
      } else {
         brw_load_register_mem(brw, GEN7_3DPRIM_START_INSTANCE, bo,
                               prim->indirect_offset + 12);
         brw_load_register_imm32(brw, GEN7_3DPRIM_BASE_VERTEX, 0);
      }
   } else {
      indirect_flag = 0;
   }

   BEGIN_BATCH(devinfo->gen >= 7 ? 7 : 6);

   if (devinfo->gen >= 7) {
      const int predicate_enable =
         (brw->predicate.state == BRW_PREDICATE_STATE_USE_BIT)
         ? GEN7_3DPRIM_PREDICATE_ENABLE : 0;

      OUT_BATCH(CMD_3D_PRIM << 16 | (7 - 2) | indirect_flag | predicate_enable);
      OUT_BATCH(hw_prim | vertex_access_type);
   } else {
      OUT_BATCH(CMD_3D_PRIM << 16 | (6 - 2) |
                hw_prim << GEN4_3DPRIM_TOPOLOGY_TYPE_SHIFT |
                vertex_access_type);
   }
   OUT_BATCH(verts_per_instance);
   OUT_BATCH(start_vertex_location);
   OUT_BATCH(prim->num_instances);
   OUT_BATCH(prim->base_instance);
   OUT_BATCH(base_vertex_location);
   ADVANCE_BATCH();

   if (brw->always_flush_cache)
      brw_emit_mi_flush(brw);
}


static void
brw_merge_inputs(struct brw_context *brw)
{
   const struct gen_device_info *devinfo = &brw->screen->devinfo;
   const struct gl_context *ctx = &brw->ctx;
   GLuint i;

   for (i = 0; i < brw->vb.nr_buffers; i++) {
      brw_bo_unreference(brw->vb.buffers[i].bo);
      brw->vb.buffers[i].bo = NULL;
   }
   brw->vb.nr_buffers = 0;

   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
      struct brw_vertex_element *input = &brw->vb.inputs[i];
      input->buffer = -1;
      _mesa_draw_attrib_and_binding(ctx, i,
                                    &input->glattrib, &input->glbinding);
   }

   if (devinfo->gen < 8 && !devinfo->is_haswell) {
      uint64_t mask = ctx->VertexProgram._Current->info.inputs_read;
      /* Prior to Haswell, the hardware can't natively support GL_FIXED or
       * 2_10_10_10_REV vertex formats.  Set appropriate workaround flags.
       */
      while (mask) {
         const struct gl_array_attributes *glattrib;
         uint8_t wa_flags = 0;

         i = u_bit_scan64(&mask);
         glattrib = brw->vb.inputs[i].glattrib;

         switch (glattrib->Type) {

         case GL_FIXED:
            wa_flags = glattrib->Size;
            break;

         case GL_INT_2_10_10_10_REV:
            wa_flags |= BRW_ATTRIB_WA_SIGN;
            /* fallthough */

         case GL_UNSIGNED_INT_2_10_10_10_REV:
            if (glattrib->Format == GL_BGRA)
               wa_flags |= BRW_ATTRIB_WA_BGRA;

            if (glattrib->Normalized)
               wa_flags |= BRW_ATTRIB_WA_NORMALIZE;
            else if (!glattrib->Integer)
               wa_flags |= BRW_ATTRIB_WA_SCALE;

            break;
         }

         if (brw->vb.attrib_wa_flags[i] != wa_flags) {
            brw->vb.attrib_wa_flags[i] = wa_flags;
            brw->ctx.NewDriverState |= BRW_NEW_VS_ATTRIB_WORKAROUNDS;
         }
      }
   }
}

/* Disable auxiliary buffers if a renderbuffer is also bound as a texture
 * or shader image.  This causes a self-dependency, where both rendering
 * and sampling may concurrently read or write the CCS buffer, causing
 * incorrect pixels.
 */
static bool
intel_disable_rb_aux_buffer(struct brw_context *brw,
                            bool *draw_aux_buffer_disabled,
                            struct intel_mipmap_tree *tex_mt,
                            unsigned min_level, unsigned num_levels,
                            const char *usage)
{
   const struct gl_framebuffer *fb = brw->ctx.DrawBuffer;
   bool found = false;

   /* We only need to worry about color compression and fast clears. */
   if (tex_mt->aux_usage != ISL_AUX_USAGE_CCS_D &&
       tex_mt->aux_usage != ISL_AUX_USAGE_CCS_E)
      return false;

   for (unsigned i = 0; i < fb->_NumColorDrawBuffers; i++) {
      const struct intel_renderbuffer *irb =
         intel_renderbuffer(fb->_ColorDrawBuffers[i]);

      if (irb && irb->mt->bo == tex_mt->bo &&
          irb->mt_level >= min_level &&
          irb->mt_level < min_level + num_levels) {
         found = draw_aux_buffer_disabled[i] = true;
      }
   }

   if (found) {
      perf_debug("Disabling CCS because a renderbuffer is also bound %s.\n",
                 usage);
   }

   return found;
}

/** Implement the ASTC 5x5 sampler workaround
 *
 * Gen9 sampling hardware has a bug where an ASTC 5x5 compressed surface
 * cannot live in the sampler cache at the same time as an aux compressed
 * surface.  In order to work around the bug we have to stall rendering with a
 * CS and pixel scoreboard stall (implicit in the CS stall) and invalidate the
 * texture cache whenever one of ASTC 5x5 or aux compressed may be in the
 * sampler cache and we're about to render with something which samples from
 * the other.
 *
 * In the case of a single shader which textures from both ASTC 5x5 and
 * a texture which is CCS or HiZ compressed, we have to resolve the aux
 * compressed texture prior to rendering.  This second part is handled in
 * brw_predraw_resolve_inputs() below.
 *
 * We have observed this issue to affect CCS and HiZ sampling but whether or
 * not it also affects MCS is unknown.  Because MCS has no concept of a
 * resolve (and doing one would be stupid expensive), we choose to simply
 * ignore the possibility and hope for the best.
 */
static void
gen9_apply_astc5x5_wa_flush(struct brw_context *brw,
                            enum gen9_astc5x5_wa_tex_type curr_mask)
{
   assert(brw->screen->devinfo.gen == 9);

   if (((brw->gen9_astc5x5_wa_tex_mask & GEN9_ASTC5X5_WA_TEX_TYPE_ASTC5x5) &&
        (curr_mask & GEN9_ASTC5X5_WA_TEX_TYPE_AUX)) ||
       ((brw->gen9_astc5x5_wa_tex_mask & GEN9_ASTC5X5_WA_TEX_TYPE_AUX) &&
        (curr_mask & GEN9_ASTC5X5_WA_TEX_TYPE_ASTC5x5))) {
      brw_emit_pipe_control_flush(brw, PIPE_CONTROL_CS_STALL);
      brw_emit_pipe_control_flush(brw, PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE);
   }

   brw->gen9_astc5x5_wa_tex_mask = curr_mask;
}

static enum gen9_astc5x5_wa_tex_type
gen9_astc5x5_wa_bits(mesa_format format, enum isl_aux_usage aux_usage)
{
   if (aux_usage != ISL_AUX_USAGE_NONE &&
       aux_usage != ISL_AUX_USAGE_MCS)
      return GEN9_ASTC5X5_WA_TEX_TYPE_AUX;

   if (format == MESA_FORMAT_RGBA_ASTC_5x5 ||
       format == MESA_FORMAT_SRGB8_ALPHA8_ASTC_5x5)
      return GEN9_ASTC5X5_WA_TEX_TYPE_ASTC5x5;

   return 0;
}

/* Helper for the gen9 ASTC 5x5 workaround.  This version exists for BLORP's
 * use-cases where only a single texture is bound.
 */
void
gen9_apply_single_tex_astc5x5_wa(struct brw_context *brw,
                                 mesa_format format,
                                 enum isl_aux_usage aux_usage)
{
   gen9_apply_astc5x5_wa_flush(brw, gen9_astc5x5_wa_bits(format, aux_usage));
}

static void
mark_textures_used_for_txf(BITSET_WORD *used_for_txf,
                           const struct gl_program *prog)
{
   if (!prog)
      return;

   unsigned mask = prog->SamplersUsed & prog->info.textures_used_by_txf;
   while (mask) {
      int s = u_bit_scan(&mask);
      BITSET_SET(used_for_txf, prog->SamplerUnits[s]);
   }
}

/**
 * \brief Resolve buffers before drawing.
 *
 * Resolve the depth buffer's HiZ buffer, resolve the depth buffer of each
 * enabled depth texture, and flush the render cache for any dirty textures.
 */
void
brw_predraw_resolve_inputs(struct brw_context *brw, bool rendering,
                           bool *draw_aux_buffer_disabled)
{
   struct gl_context *ctx = &brw->ctx;
   struct intel_texture_object *tex_obj;

   BITSET_DECLARE(used_for_txf, MAX_COMBINED_TEXTURE_IMAGE_UNITS);
   memset(used_for_txf, 0, sizeof(used_for_txf));
   if (rendering) {
      mark_textures_used_for_txf(used_for_txf, ctx->VertexProgram._Current);
      mark_textures_used_for_txf(used_for_txf, ctx->TessCtrlProgram._Current);
      mark_textures_used_for_txf(used_for_txf, ctx->TessEvalProgram._Current);
      mark_textures_used_for_txf(used_for_txf, ctx->GeometryProgram._Current);
      mark_textures_used_for_txf(used_for_txf, ctx->FragmentProgram._Current);
   } else {
      mark_textures_used_for_txf(used_for_txf, ctx->ComputeProgram._Current);
   }

   int maxEnabledUnit = ctx->Texture._MaxEnabledTexImageUnit;

   enum gen9_astc5x5_wa_tex_type astc5x5_wa_bits = 0;
   if (brw->screen->devinfo.gen == 9) {
      /* In order to properly implement the ASTC 5x5 workaround for an
       * arbitrary draw or dispatch call, we have to walk the entire list of
       * textures looking for ASTC 5x5.  If there is any ASTC 5x5 in this draw
       * call, all aux compressed textures must be resolved and have aux
       * compression disabled while sampling.
       */
      for (int i = 0; i <= maxEnabledUnit; i++) {
         if (!ctx->Texture.Unit[i]._Current)
            continue;
         tex_obj = intel_texture_object(ctx->Texture.Unit[i]._Current);
         if (!tex_obj || !tex_obj->mt)
            continue;

         astc5x5_wa_bits |= gen9_astc5x5_wa_bits(tex_obj->_Format,
                                                 tex_obj->mt->aux_usage);
      }
      gen9_apply_astc5x5_wa_flush(brw, astc5x5_wa_bits);
   }

   /* Resolve depth buffer and render cache of each enabled texture. */
   for (int i = 0; i <= maxEnabledUnit; i++) {
      if (!ctx->Texture.Unit[i]._Current)
	 continue;
      tex_obj = intel_texture_object(ctx->Texture.Unit[i]._Current);
      if (!tex_obj || !tex_obj->mt)
	 continue;

      struct gl_sampler_object *sampler = _mesa_get_samplerobj(ctx, i);
      enum isl_format view_format =
         translate_tex_format(brw, tex_obj->_Format, sampler->sRGBDecode);

      unsigned min_level, min_layer, num_levels, num_layers;
      if (tex_obj->base.Immutable) {
         min_level  = tex_obj->base.MinLevel;
         num_levels = MIN2(tex_obj->base.NumLevels, tex_obj->_MaxLevel + 1);
         min_layer  = tex_obj->base.MinLayer;
         num_layers = tex_obj->base.Target != GL_TEXTURE_3D ?
                      tex_obj->base.NumLayers : INTEL_REMAINING_LAYERS;
      } else {
         min_level  = tex_obj->base.BaseLevel;
         num_levels = tex_obj->_MaxLevel - tex_obj->base.BaseLevel + 1;
         min_layer  = 0;
         num_layers = INTEL_REMAINING_LAYERS;
      }

      if (rendering) {
         intel_disable_rb_aux_buffer(brw, draw_aux_buffer_disabled,
                                     tex_obj->mt, min_level, num_levels,
                                     "for sampling");
      }

      intel_miptree_prepare_texture(brw, tex_obj->mt, view_format,
                                    min_level, num_levels,
                                    min_layer, num_layers,
                                    astc5x5_wa_bits);

      /* If any programs are using it with texelFetch, we may need to also do
       * a prepare with an sRGB format to ensure texelFetch works "properly".
       */
      if (BITSET_TEST(used_for_txf, i)) {
         enum isl_format txf_format =
            translate_tex_format(brw, tex_obj->_Format, GL_DECODE_EXT);
         if (txf_format != view_format) {
            intel_miptree_prepare_texture(brw, tex_obj->mt, txf_format,
                                          min_level, num_levels,
                                          min_layer, num_layers,
                                          astc5x5_wa_bits);
         }
      }

      brw_cache_flush_for_read(brw, tex_obj->mt->bo);

      if (tex_obj->base.StencilSampling ||
          tex_obj->mt->format == MESA_FORMAT_S_UINT8) {
         intel_update_r8stencil(brw, tex_obj->mt);
      }
   }

   /* Resolve color for each active shader image. */
   for (unsigned i = 0; i < MESA_SHADER_STAGES; i++) {
      const struct gl_program *prog = ctx->_Shader->CurrentProgram[i];

      if (unlikely(prog && prog->info.num_images)) {
         for (unsigned j = 0; j < prog->info.num_images; j++) {
            struct gl_image_unit *u =
               &ctx->ImageUnits[prog->sh.ImageUnits[j]];
            tex_obj = intel_texture_object(u->TexObj);

            if (tex_obj && tex_obj->mt) {
               if (rendering) {
                  intel_disable_rb_aux_buffer(brw, draw_aux_buffer_disabled,
                                              tex_obj->mt, 0, ~0,
                                              "as a shader image");
               }

               intel_miptree_prepare_image(brw, tex_obj->mt);

               brw_cache_flush_for_read(brw, tex_obj->mt->bo);
            }
         }
      }
   }
}

static void
brw_predraw_resolve_framebuffer(struct brw_context *brw,
                                bool *draw_aux_buffer_disabled)
{
   struct gl_context *ctx = &brw->ctx;
   struct intel_renderbuffer *depth_irb;

   /* Resolve the depth buffer's HiZ buffer. */
   depth_irb = intel_get_renderbuffer(ctx->DrawBuffer, BUFFER_DEPTH);
   if (depth_irb && depth_irb->mt) {
      intel_miptree_prepare_depth(brw, depth_irb->mt,
                                  depth_irb->mt_level,
                                  depth_irb->mt_layer,
                                  depth_irb->layer_count);
   }

   /* Resolve color buffers for non-coherent framebuffer fetch. */
   if (!ctx->Extensions.EXT_shader_framebuffer_fetch &&
       ctx->FragmentProgram._Current &&
       ctx->FragmentProgram._Current->info.outputs_read) {
      const struct gl_framebuffer *fb = ctx->DrawBuffer;

      /* This is only used for non-coherent framebuffer fetch, so we don't
       * need to worry about CCS_E and can simply pass 'false' below.
       */
      assert(brw->screen->devinfo.gen < 9);

      for (unsigned i = 0; i < fb->_NumColorDrawBuffers; i++) {
         const struct intel_renderbuffer *irb =
            intel_renderbuffer(fb->_ColorDrawBuffers[i]);

         if (irb) {
            intel_miptree_prepare_texture(brw, irb->mt, irb->mt->surf.format,
                                          irb->mt_level, 1,
                                          irb->mt_layer, irb->layer_count,
                                          brw->gen9_astc5x5_wa_tex_mask);
         }
      }
   }

   struct gl_framebuffer *fb = ctx->DrawBuffer;
   for (int i = 0; i < fb->_NumColorDrawBuffers; i++) {
      struct intel_renderbuffer *irb =
         intel_renderbuffer(fb->_ColorDrawBuffers[i]);

      if (irb == NULL || irb->mt == NULL)
         continue;

      mesa_format mesa_format =
         _mesa_get_render_format(ctx, intel_rb_format(irb));
      enum isl_format isl_format = brw_isl_format_for_mesa_format(mesa_format);
      bool blend_enabled = ctx->Color.BlendEnabled & (1 << i);
      enum isl_aux_usage aux_usage =
         intel_miptree_render_aux_usage(brw, irb->mt, isl_format,
                                        blend_enabled,
                                        draw_aux_buffer_disabled[i]);
      if (brw->draw_aux_usage[i] != aux_usage) {
         brw->ctx.NewDriverState |= BRW_NEW_AUX_STATE;
         brw->draw_aux_usage[i] = aux_usage;
      }

      intel_miptree_prepare_render(brw, irb->mt, irb->mt_level,
                                   irb->mt_layer, irb->layer_count,
                                   aux_usage);

      brw_cache_flush_for_render(brw, irb->mt->bo,
                                 isl_format, aux_usage);
   }
}

/**
 * \brief Call this after drawing to mark which buffers need resolving
 *
 * If the depth buffer was written to and if it has an accompanying HiZ
 * buffer, then mark that it needs a depth resolve.
 *
 * If the stencil buffer was written to then mark that it may need to be
 * copied to an R8 texture.
 *
 * If the color buffer is a multisample window system buffer, then
 * mark that it needs a downsample.
 *
 * Also mark any render targets which will be textured as needing a render
 * cache flush.
 */
static void
brw_postdraw_set_buffers_need_resolve(struct brw_context *brw)
{
   struct gl_context *ctx = &brw->ctx;
   struct gl_framebuffer *fb = ctx->DrawBuffer;

   struct intel_renderbuffer *front_irb = NULL;
   struct intel_renderbuffer *back_irb = intel_get_renderbuffer(fb, BUFFER_BACK_LEFT);
   struct intel_renderbuffer *depth_irb = intel_get_renderbuffer(fb, BUFFER_DEPTH);
   struct intel_renderbuffer *stencil_irb = intel_get_renderbuffer(fb, BUFFER_STENCIL);
   struct gl_renderbuffer_attachment *depth_att = &fb->Attachment[BUFFER_DEPTH];

   if (_mesa_is_front_buffer_drawing(fb))
      front_irb = intel_get_renderbuffer(fb, BUFFER_FRONT_LEFT);

   if (front_irb)
      front_irb->need_downsample = true;
   if (back_irb)
      back_irb->need_downsample = true;
   if (depth_irb) {
      bool depth_written = brw_depth_writes_enabled(brw);
      if (depth_att->Layered) {
         intel_miptree_finish_depth(brw, depth_irb->mt,
                                    depth_irb->mt_level,
                                    depth_irb->mt_layer,
                                    depth_irb->layer_count,
                                    depth_written);
      } else {
         intel_miptree_finish_depth(brw, depth_irb->mt,
                                    depth_irb->mt_level,
                                    depth_irb->mt_layer, 1,
                                    depth_written);
      }
      if (depth_written)
         brw_depth_cache_add_bo(brw, depth_irb->mt->bo);
   }

   if (stencil_irb && brw->stencil_write_enabled) {
      struct intel_mipmap_tree *stencil_mt =
         stencil_irb->mt->stencil_mt != NULL ?
         stencil_irb->mt->stencil_mt : stencil_irb->mt;
      brw_depth_cache_add_bo(brw, stencil_mt->bo);
      intel_miptree_finish_write(brw, stencil_mt, stencil_irb->mt_level,
                                 stencil_irb->mt_layer,
                                 stencil_irb->layer_count, ISL_AUX_USAGE_NONE);
   }

   for (unsigned i = 0; i < fb->_NumColorDrawBuffers; i++) {
      struct intel_renderbuffer *irb =
         intel_renderbuffer(fb->_ColorDrawBuffers[i]);

      if (!irb)
         continue;

      mesa_format mesa_format =
         _mesa_get_render_format(ctx, intel_rb_format(irb));
      enum isl_format isl_format = brw_isl_format_for_mesa_format(mesa_format);
      enum isl_aux_usage aux_usage = brw->draw_aux_usage[i];

      brw_render_cache_add_bo(brw, irb->mt->bo, isl_format, aux_usage);

      intel_miptree_finish_render(brw, irb->mt, irb->mt_level,
                                  irb->mt_layer, irb->layer_count,
                                  aux_usage);
   }
}

static void
intel_renderbuffer_move_temp_back(struct brw_context *brw,
                                  struct intel_renderbuffer *irb)
{
   if (irb->align_wa_mt == NULL)
      return;

   brw_cache_flush_for_read(brw, irb->align_wa_mt->bo);

   intel_miptree_copy_slice(brw, irb->align_wa_mt, 0, 0,
                            irb->mt,
                            irb->Base.Base.TexImage->Level, irb->mt_layer);

   intel_miptree_reference(&irb->align_wa_mt, NULL);

   /* Finally restore the x,y to correspond to full miptree. */
   intel_renderbuffer_set_draw_offset(irb);

   /* Make sure render surface state gets re-emitted with updated miptree. */
   brw->NewGLState |= _NEW_BUFFERS;
}

static void
brw_postdraw_reconcile_align_wa_slices(struct brw_context *brw)
{
   struct gl_context *ctx = &brw->ctx;
   struct gl_framebuffer *fb = ctx->DrawBuffer;

   struct intel_renderbuffer *depth_irb =
      intel_get_renderbuffer(fb, BUFFER_DEPTH);
   struct intel_renderbuffer *stencil_irb =
      intel_get_renderbuffer(fb, BUFFER_STENCIL);

   if (depth_irb && depth_irb->align_wa_mt)
      intel_renderbuffer_move_temp_back(brw, depth_irb);

   if (stencil_irb && stencil_irb->align_wa_mt)
      intel_renderbuffer_move_temp_back(brw, stencil_irb);

   for (unsigned i = 0; i < fb->_NumColorDrawBuffers; i++) {
      struct intel_renderbuffer *irb =
         intel_renderbuffer(fb->_ColorDrawBuffers[i]);

      if (!irb || irb->align_wa_mt == NULL)
         continue;

      intel_renderbuffer_move_temp_back(brw, irb);
   }
}

static void
brw_prepare_drawing(struct gl_context *ctx,
                    const struct _mesa_index_buffer *ib,
                    bool index_bounds_valid,
                    GLuint min_index,
                    GLuint max_index)
{
   struct brw_context *brw = brw_context(ctx);

   if (ctx->NewState)
      _mesa_update_state(ctx);

   /* We have to validate the textures *before* checking for fallbacks;
    * otherwise, the software fallback won't be able to rely on the
    * texture state, the firstLevel and lastLevel fields won't be
    * set in the intel texture object (they'll both be 0), and the
    * software fallback will segfault if it attempts to access any
    * texture level other than level 0.
    */
   brw_validate_textures(brw);

   /* Find the highest sampler unit used by each shader program.  A bit-count
    * won't work since ARB programs use the texture unit number as the sampler
    * index.
    */
   brw->wm.base.sampler_count =
      util_last_bit(ctx->FragmentProgram._Current->SamplersUsed);
   brw->gs.base.sampler_count = ctx->GeometryProgram._Current ?
      util_last_bit(ctx->GeometryProgram._Current->SamplersUsed) : 0;
   brw->tes.base.sampler_count = ctx->TessEvalProgram._Current ?
      util_last_bit(ctx->TessEvalProgram._Current->SamplersUsed) : 0;
   brw->tcs.base.sampler_count = ctx->TessCtrlProgram._Current ?
      util_last_bit(ctx->TessCtrlProgram._Current->SamplersUsed) : 0;
   brw->vs.base.sampler_count =
      util_last_bit(ctx->VertexProgram._Current->SamplersUsed);

   intel_prepare_render(brw);

   /* This workaround has to happen outside of brw_upload_render_state()
    * because it may flush the batchbuffer for a blit, affecting the state
    * flags.
    */
   brw_workaround_depthstencil_alignment(brw, 0);

   /* Resolves must occur after updating renderbuffers, updating context state,
    * and finalizing textures but before setting up any hardware state for
    * this draw call.
    */
   bool draw_aux_buffer_disabled[MAX_DRAW_BUFFERS] = { };
   brw_predraw_resolve_inputs(brw, true, draw_aux_buffer_disabled);
   brw_predraw_resolve_framebuffer(brw, draw_aux_buffer_disabled);

   /* Bind all inputs, derive varying and size information:
    */
   brw_merge_inputs(brw);

   brw->ib.ib = ib;
   brw->ctx.NewDriverState |= BRW_NEW_INDICES;

   brw->vb.index_bounds_valid = index_bounds_valid;
   brw->vb.min_index = min_index;
   brw->vb.max_index = max_index;
   brw->ctx.NewDriverState |= BRW_NEW_VERTICES;
}

static void
brw_finish_drawing(struct gl_context *ctx)
{
   struct brw_context *brw = brw_context(ctx);

   if (brw->always_flush_batch)
      intel_batchbuffer_flush(brw);

   brw_program_cache_check_size(brw);
   brw_postdraw_reconcile_align_wa_slices(brw);
   brw_postdraw_set_buffers_need_resolve(brw);

   if (brw->draw.draw_params_count_bo) {
      brw_bo_unreference(brw->draw.draw_params_count_bo);
      brw->draw.draw_params_count_bo = NULL;
   }
}

/* May fail if out of video memory for texture or vbo upload, or on
 * fallback conditions.
 */
static void
brw_draw_single_prim(struct gl_context *ctx,
                     const struct _mesa_prim *prim,
                     unsigned prim_id,
                     struct brw_transform_feedback_object *xfb_obj,
                     unsigned stream,
                     struct gl_buffer_object *indirect)
{
   struct brw_context *brw = brw_context(ctx);
   const struct gen_device_info *devinfo = &brw->screen->devinfo;
   bool fail_next = false;

   /* Flag BRW_NEW_DRAW_CALL on every draw.  This allows us to have
    * atoms that happen on every draw call.
    */
   brw->ctx.NewDriverState |= BRW_NEW_DRAW_CALL;

   /* Flush the batch if the batch/state buffers are nearly full.  We can
    * grow them if needed, but this is not free, so we'd like to avoid it.
    */
   intel_batchbuffer_require_space(brw, 1500);
   brw_require_statebuffer_space(brw, 2400);
   intel_batchbuffer_save_state(brw);

   if (brw->num_instances != prim->num_instances ||
       brw->basevertex != prim->basevertex ||
       brw->baseinstance != prim->base_instance) {
      brw->num_instances = prim->num_instances;
      brw->basevertex = prim->basevertex;
      brw->baseinstance = prim->base_instance;
      if (prim_id > 0) { /* For i == 0 we just did this before the loop */
         brw->ctx.NewDriverState |= BRW_NEW_VERTICES;
         brw_merge_inputs(brw);
      }
   }

   /* Determine if we need to flag BRW_NEW_VERTICES for updating the
    * gl_BaseVertexARB or gl_BaseInstanceARB values. For indirect draw, we
    * always flag if the shader uses one of the values. For direct draws,
    * we only flag if the values change.
    */
   const int new_firstvertex =
      prim->indexed ? prim->basevertex : prim->start;
   const int new_baseinstance = prim->base_instance;
   const struct brw_vs_prog_data *vs_prog_data =
      brw_vs_prog_data(brw->vs.base.prog_data);
   if (prim_id > 0) {
      const bool uses_draw_parameters =
         vs_prog_data->uses_firstvertex ||
         vs_prog_data->uses_baseinstance;

      if ((uses_draw_parameters && prim->is_indirect) ||
          (vs_prog_data->uses_firstvertex &&
           brw->draw.params.firstvertex != new_firstvertex) ||
          (vs_prog_data->uses_baseinstance &&
           brw->draw.params.gl_baseinstance != new_baseinstance))
         brw->ctx.NewDriverState |= BRW_NEW_VERTICES;
   }

   brw->draw.params.firstvertex = new_firstvertex;
   brw->draw.params.gl_baseinstance = new_baseinstance;
   brw_bo_unreference(brw->draw.draw_params_bo);

   if (prim->is_indirect) {
      /* Point draw_params_bo at the indirect buffer. */
      brw->draw.draw_params_bo =
         intel_buffer_object(ctx->DrawIndirectBuffer)->buffer;
      brw_bo_reference(brw->draw.draw_params_bo);
      brw->draw.draw_params_offset =
         prim->indirect_offset + (prim->indexed ? 12 : 8);
   } else {
      /* Set draw_params_bo to NULL so brw_prepare_vertices knows it
       * has to upload gl_BaseVertex and such if they're needed.
       */
      brw->draw.draw_params_bo = NULL;
      brw->draw.draw_params_offset = 0;
   }

   /* gl_DrawID always needs its own vertex buffer since it's not part of
    * the indirect parameter buffer. Same for is_indexed_draw, which shares
    * the buffer with gl_DrawID. If the program uses gl_DrawID, we need to
    * flag BRW_NEW_VERTICES. For the first iteration, we don't have valid
    * vs_prog_data, but we always flag BRW_NEW_VERTICES before the loop.
    */
   if (prim_id > 0 && vs_prog_data->uses_drawid)
      brw->ctx.NewDriverState |= BRW_NEW_VERTICES;

   brw->draw.derived_params.gl_drawid = prim->draw_id;
   brw->draw.derived_params.is_indexed_draw = prim->indexed ? ~0 : 0;

   brw_bo_unreference(brw->draw.derived_draw_params_bo);
   brw->draw.derived_draw_params_bo = NULL;
   brw->draw.derived_draw_params_offset = 0;

   if (devinfo->gen < 6)
      brw_set_prim(brw, prim);
   else
      gen6_set_prim(brw, prim);

retry:

   /* Note that before the loop, brw->ctx.NewDriverState was set to != 0, and
    * that the state updated in the loop outside of this block is that in
    * *_set_prim or intel_batchbuffer_flush(), which only impacts
    * brw->ctx.NewDriverState.
    */
   if (brw->ctx.NewDriverState) {
      brw->batch.no_wrap = true;
      brw_upload_render_state(brw);
   }

   brw_emit_prim(brw, prim, brw->primitive, xfb_obj, stream);

   brw->batch.no_wrap = false;

   if (!brw_batch_has_aperture_space(brw, 0)) {
      if (!fail_next) {
         intel_batchbuffer_reset_to_saved(brw);
         intel_batchbuffer_flush(brw);
         fail_next = true;
         goto retry;
      } else {
         int ret = intel_batchbuffer_flush(brw);
         WARN_ONCE(ret == -ENOSPC,
                   "i965: Single primitive emit exceeded "
                   "available aperture space\n");
      }
   }

   /* Now that we know we haven't run out of aperture space, we can safely
    * reset the dirty bits.
    */
   if (brw->ctx.NewDriverState)
      brw_render_state_finished(brw);

   return;
}



void
brw_draw_prims(struct gl_context *ctx,
               const struct _mesa_prim *prims,
               GLuint nr_prims,
               const struct _mesa_index_buffer *ib,
               GLboolean index_bounds_valid,
               GLuint min_index,
               GLuint max_index,
               struct gl_transform_feedback_object *gl_xfb_obj,
               unsigned stream,
               struct gl_buffer_object *indirect)
{
   unsigned i;
   struct brw_context *brw = brw_context(ctx);
   int predicate_state = brw->predicate.state;
   struct brw_transform_feedback_object *xfb_obj =
      (struct brw_transform_feedback_object *) gl_xfb_obj;

   if (!brw_check_conditional_render(brw))
      return;

   /* Handle primitive restart if needed */
   if (brw_handle_primitive_restart(ctx, prims, nr_prims, ib, indirect)) {
      /* The draw was handled, so we can exit now */
      return;
   }

   /* Do GL_SELECT and GL_FEEDBACK rendering using swrast, even though it
    * won't support all the extensions we support.
    */
   if (ctx->RenderMode != GL_RENDER) {
      perf_debug("%s render mode not supported in hardware\n",
                 _mesa_enum_to_string(ctx->RenderMode));
      _swsetup_Wakeup(ctx);
      _tnl_wakeup(ctx);
      _tnl_draw(ctx, prims, nr_prims, ib,
                index_bounds_valid, min_index, max_index, NULL, 0, NULL);
      return;
   }

   /* If we're going to have to upload any of the user's vertex arrays, then
    * get the minimum and maximum of their index buffer so we know what range
    * to upload.
    */
   if (!index_bounds_valid && _mesa_draw_user_array_bits(ctx) != 0) {
      perf_debug("Scanning index buffer to compute index buffer bounds.  "
                 "Use glDrawRangeElements() to avoid this.\n");
      vbo_get_minmax_indices(ctx, prims, ib, &min_index, &max_index, nr_prims);
      index_bounds_valid = true;
   }

   brw_prepare_drawing(ctx, ib, index_bounds_valid, min_index, max_index);
   /* Try drawing with the hardware, but don't do anything else if we can't
    * manage it.  swrast doesn't support our featureset, so we can't fall back
    * to it.
    */

   for (i = 0; i < nr_prims; i++) {
      /* Implementation of ARB_indirect_parameters via predicates */
      if (brw->draw.draw_params_count_bo) {
         brw_emit_pipe_control_flush(brw, PIPE_CONTROL_FLUSH_ENABLE);

         /* Upload the current draw count from the draw parameters buffer to
          * MI_PREDICATE_SRC0.
          */
         brw_load_register_mem(brw, MI_PREDICATE_SRC0,
                               brw->draw.draw_params_count_bo,
                               brw->draw.draw_params_count_offset);
         /* Zero the top 32-bits of MI_PREDICATE_SRC0 */
         brw_load_register_imm32(brw, MI_PREDICATE_SRC0 + 4, 0);
         /* Upload the id of the current primitive to MI_PREDICATE_SRC1. */
         brw_load_register_imm64(brw, MI_PREDICATE_SRC1, prims[i].draw_id);

         BEGIN_BATCH(1);
         if (i == 0 && brw->predicate.state != BRW_PREDICATE_STATE_USE_BIT) {
            OUT_BATCH(GEN7_MI_PREDICATE | MI_PREDICATE_LOADOP_LOADINV |
                      MI_PREDICATE_COMBINEOP_SET |
                      MI_PREDICATE_COMPAREOP_SRCS_EQUAL);
         } else {
            OUT_BATCH(GEN7_MI_PREDICATE |
                      MI_PREDICATE_LOADOP_LOAD | MI_PREDICATE_COMBINEOP_XOR |
                      MI_PREDICATE_COMPAREOP_SRCS_EQUAL);
         }
         ADVANCE_BATCH();

         brw->predicate.state = BRW_PREDICATE_STATE_USE_BIT;
      }

      brw_draw_single_prim(ctx, &prims[i], i, xfb_obj, stream, indirect);
   }

   brw_finish_drawing(ctx);
   brw->predicate.state = predicate_state;
}

void
brw_draw_indirect_prims(struct gl_context *ctx,
                        GLuint mode,
                        struct gl_buffer_object *indirect_data,
                        GLsizeiptr indirect_offset,
                        unsigned draw_count,
                        unsigned stride,
                        struct gl_buffer_object *indirect_params,
                        GLsizeiptr indirect_params_offset,
                        const struct _mesa_index_buffer *ib)
{
   struct brw_context *brw = brw_context(ctx);
   struct _mesa_prim *prim;
   GLsizei i;

   prim = calloc(draw_count, sizeof(*prim));
   if (prim == NULL) {
      _mesa_error(ctx, GL_OUT_OF_MEMORY, "gl%sDraw%sIndirect%s",
                  (draw_count > 1) ? "Multi" : "",
                  ib ? "Elements" : "Arrays",
                  indirect_params ? "CountARB" : "");
      return;
   }

   prim[0].begin = 1;
   prim[draw_count - 1].end = 1;
   for (i = 0; i < draw_count; ++i, indirect_offset += stride) {
      prim[i].mode = mode;
      prim[i].indexed = ib != NULL;
      prim[i].indirect_offset = indirect_offset;
      prim[i].is_indirect = 1;
      prim[i].draw_id = i;
   }

   if (indirect_params) {
      brw->draw.draw_params_count_bo =
         intel_buffer_object(indirect_params)->buffer;
      brw_bo_reference(brw->draw.draw_params_count_bo);
      brw->draw.draw_params_count_offset = indirect_params_offset;
   }

   brw_draw_prims(ctx, prim, draw_count,
                  ib, false, 0, ~0,
                  NULL, 0,
                  indirect_data);

   free(prim);
}

void
brw_init_draw_functions(struct dd_function_table *functions)
{
   /* Register our drawing function:
    */
   functions->Draw = brw_draw_prims;
   functions->DrawIndirect = brw_draw_indirect_prims;
}

void
brw_draw_init(struct brw_context *brw)
{
   for (int i = 0; i < VERT_ATTRIB_MAX; i++)
      brw->vb.inputs[i].buffer = -1;
   brw->vb.nr_buffers = 0;
   brw->vb.nr_enabled = 0;
}

void
brw_draw_destroy(struct brw_context *brw)
{
   unsigned i;

   for (i = 0; i < brw->vb.nr_buffers; i++) {
      brw_bo_unreference(brw->vb.buffers[i].bo);
      brw->vb.buffers[i].bo = NULL;
   }
   brw->vb.nr_buffers = 0;

   for (i = 0; i < brw->vb.nr_enabled; i++) {
      brw->vb.enabled[i]->buffer = -1;
   }
   brw->vb.nr_enabled = 0;

   brw_bo_unreference(brw->ib.bo);
   brw->ib.bo = NULL;
}