summaryrefslogtreecommitdiffstats
path: root/src/gallium/drivers/ilo/ilo_shader.c
blob: 0798ccc8186612f5c6488c10b8325e28957b1449 (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
/*
 * Mesa 3-D graphics library
 *
 * Copyright (C) 2012-2013 LunarG, Inc.
 *
 * 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 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.
 *
 * Authors:
 *    Chia-I Wu <olv@lunarg.com>
 */

#include "genhw/genhw.h" /* for SBE setup */
#include "tgsi/tgsi_parse.h"
#include "intel_winsys.h"

#include "shader/ilo_shader_internal.h"
#include "ilo_builder.h"
#include "ilo_state.h"
#include "ilo_state_3d.h"
#include "ilo_shader.h"

struct ilo_shader_cache {
   struct list_head shaders;
   struct list_head changed;
};

/**
 * Create a shader cache.  A shader cache can manage shaders and upload them
 * to a bo as a whole.
 */
struct ilo_shader_cache *
ilo_shader_cache_create(void)
{
   struct ilo_shader_cache *shc;

   shc = CALLOC_STRUCT(ilo_shader_cache);
   if (!shc)
      return NULL;

   list_inithead(&shc->shaders);
   list_inithead(&shc->changed);

   return shc;
}

/**
 * Destroy a shader cache.
 */
void
ilo_shader_cache_destroy(struct ilo_shader_cache *shc)
{
   FREE(shc);
}

/**
 * Add a shader to the cache.
 */
void
ilo_shader_cache_add(struct ilo_shader_cache *shc,
                     struct ilo_shader_state *shader)
{
   struct ilo_shader *sh;

   shader->cache = shc;
   LIST_FOR_EACH_ENTRY(sh, &shader->variants, list)
      sh->uploaded = false;

   list_add(&shader->list, &shc->changed);
}

/**
 * Remove a shader from the cache.
 */
void
ilo_shader_cache_remove(struct ilo_shader_cache *shc,
                        struct ilo_shader_state *shader)
{
   list_del(&shader->list);
   shader->cache = NULL;
}

/**
 * Notify the cache that a managed shader has changed.
 */
static void
ilo_shader_cache_notify_change(struct ilo_shader_cache *shc,
                               struct ilo_shader_state *shader)
{
   if (shader->cache == shc) {
      list_del(&shader->list);
      list_add(&shader->list, &shc->changed);
   }
}

/**
 * Upload managed shaders to the bo.  Only shaders that are changed or added
 * after the last upload are uploaded.
 */
void
ilo_shader_cache_upload(struct ilo_shader_cache *shc,
                        struct ilo_builder *builder)
{
   struct ilo_shader_state *shader, *next;

   LIST_FOR_EACH_ENTRY_SAFE(shader, next, &shc->changed, list) {
      struct ilo_shader *sh;

      LIST_FOR_EACH_ENTRY(sh, &shader->variants, list) {
         if (sh->uploaded)
            continue;

         sh->cache_offset = ilo_builder_instruction_write(builder,
               sh->kernel_size, sh->kernel);

         sh->uploaded = true;
      }

      list_del(&shader->list);
      list_add(&shader->list, &shc->shaders);
   }
}

/**
 * Invalidate all shaders so that they get uploaded in next
 * ilo_shader_cache_upload().
 */
void
ilo_shader_cache_invalidate(struct ilo_shader_cache *shc)
{
   struct ilo_shader_state *shader, *next;

   LIST_FOR_EACH_ENTRY_SAFE(shader, next, &shc->shaders, list) {
      list_del(&shader->list);
      list_add(&shader->list, &shc->changed);
   }

   LIST_FOR_EACH_ENTRY(shader, &shc->changed, list) {
      struct ilo_shader *sh;

      LIST_FOR_EACH_ENTRY(sh, &shader->variants, list)
         sh->uploaded = false;
   }
}

/**
 * Initialize a shader variant.
 */
void
ilo_shader_variant_init(struct ilo_shader_variant *variant,
                        const struct ilo_shader_info *info,
                        const struct ilo_state_vector *vec)
{
   int num_views, i;

   memset(variant, 0, sizeof(*variant));

   switch (info->type) {
   case PIPE_SHADER_VERTEX:
      variant->u.vs.rasterizer_discard =
         vec->rasterizer->state.rasterizer_discard;
      variant->u.vs.num_ucps =
         util_last_bit(vec->rasterizer->state.clip_plane_enable);
      break;
   case PIPE_SHADER_GEOMETRY:
      variant->u.gs.rasterizer_discard =
         vec->rasterizer->state.rasterizer_discard;
      variant->u.gs.num_inputs = vec->vs->shader->out.count;
      for (i = 0; i < vec->vs->shader->out.count; i++) {
         variant->u.gs.semantic_names[i] =
            vec->vs->shader->out.semantic_names[i];
         variant->u.gs.semantic_indices[i] =
            vec->vs->shader->out.semantic_indices[i];
      }
      break;
   case PIPE_SHADER_FRAGMENT:
      variant->u.fs.flatshade =
         (info->has_color_interp && vec->rasterizer->state.flatshade);
      variant->u.fs.fb_height = (info->has_pos) ?
         vec->fb.state.height : 1;
      variant->u.fs.num_cbufs = vec->fb.state.nr_cbufs;
      break;
   default:
      assert(!"unknown shader type");
      break;
   }

   /* use PCB unless constant buffer 0 is not in user buffer  */
   if ((vec->cbuf[info->type].enabled_mask & 0x1) &&
       !vec->cbuf[info->type].cso[0].user_buffer)
      variant->use_pcb = false;
   else
      variant->use_pcb = true;

   num_views = vec->view[info->type].count;
   assert(info->num_samplers <= num_views);

   variant->num_sampler_views = info->num_samplers;
   for (i = 0; i < info->num_samplers; i++) {
      const struct pipe_sampler_view *view = vec->view[info->type].states[i];
      const struct ilo_sampler_cso *sampler = vec->sampler[info->type].cso[i];

      if (view) {
         variant->sampler_view_swizzles[i].r = view->swizzle_r;
         variant->sampler_view_swizzles[i].g = view->swizzle_g;
         variant->sampler_view_swizzles[i].b = view->swizzle_b;
         variant->sampler_view_swizzles[i].a = view->swizzle_a;
      }
      else if (info->shadow_samplers & (1 << i)) {
         variant->sampler_view_swizzles[i].r = PIPE_SWIZZLE_RED;
         variant->sampler_view_swizzles[i].g = PIPE_SWIZZLE_RED;
         variant->sampler_view_swizzles[i].b = PIPE_SWIZZLE_RED;
         variant->sampler_view_swizzles[i].a = PIPE_SWIZZLE_ONE;
      }
      else {
         variant->sampler_view_swizzles[i].r = PIPE_SWIZZLE_RED;
         variant->sampler_view_swizzles[i].g = PIPE_SWIZZLE_GREEN;
         variant->sampler_view_swizzles[i].b = PIPE_SWIZZLE_BLUE;
         variant->sampler_view_swizzles[i].a = PIPE_SWIZZLE_ALPHA;
      }

      /*
       * When non-nearest filter and PIPE_TEX_WRAP_CLAMP wrap mode is used,
       * the HW wrap mode is set to GEN6_TEXCOORDMODE_CLAMP_BORDER, and we
       * need to manually saturate the texture coordinates.
       */
      if (sampler) {
         variant->saturate_tex_coords[0] |= sampler->saturate_s << i;
         variant->saturate_tex_coords[1] |= sampler->saturate_t << i;
         variant->saturate_tex_coords[2] |= sampler->saturate_r << i;
      }
   }
}

/**
 * Guess the shader variant, knowing that the context may still change.
 */
static void
ilo_shader_variant_guess(struct ilo_shader_variant *variant,
                         const struct ilo_shader_info *info,
                         const struct ilo_state_vector *vec)
{
   int i;

   memset(variant, 0, sizeof(*variant));

   switch (info->type) {
   case PIPE_SHADER_VERTEX:
      break;
   case PIPE_SHADER_GEOMETRY:
      break;
   case PIPE_SHADER_FRAGMENT:
      variant->u.fs.flatshade = false;
      variant->u.fs.fb_height = (info->has_pos) ?
         vec->fb.state.height : 1;
      variant->u.fs.num_cbufs = 1;
      break;
   default:
      assert(!"unknown shader type");
      break;
   }

   variant->use_pcb = true;

   variant->num_sampler_views = info->num_samplers;
   for (i = 0; i < info->num_samplers; i++) {
      if (info->shadow_samplers & (1 << i)) {
         variant->sampler_view_swizzles[i].r = PIPE_SWIZZLE_RED;
         variant->sampler_view_swizzles[i].g = PIPE_SWIZZLE_RED;
         variant->sampler_view_swizzles[i].b = PIPE_SWIZZLE_RED;
         variant->sampler_view_swizzles[i].a = PIPE_SWIZZLE_ONE;
      }
      else {
         variant->sampler_view_swizzles[i].r = PIPE_SWIZZLE_RED;
         variant->sampler_view_swizzles[i].g = PIPE_SWIZZLE_GREEN;
         variant->sampler_view_swizzles[i].b = PIPE_SWIZZLE_BLUE;
         variant->sampler_view_swizzles[i].a = PIPE_SWIZZLE_ALPHA;
      }
   }
}


/**
 * Parse a TGSI instruction for the shader info.
 */
static void
ilo_shader_info_parse_inst(struct ilo_shader_info *info,
                           const struct tgsi_full_instruction *inst)
{
   int i;

   /* look for edgeflag passthrough */
   if (info->edgeflag_out >= 0 &&
       inst->Instruction.Opcode == TGSI_OPCODE_MOV &&
       inst->Dst[0].Register.File == TGSI_FILE_OUTPUT &&
       inst->Dst[0].Register.Index == info->edgeflag_out) {

      assert(inst->Src[0].Register.File == TGSI_FILE_INPUT);
      info->edgeflag_in = inst->Src[0].Register.Index;
   }

   if (inst->Instruction.Texture) {
      bool shadow;

      switch (inst->Texture.Texture) {
      case TGSI_TEXTURE_SHADOW1D:
      case TGSI_TEXTURE_SHADOW2D:
      case TGSI_TEXTURE_SHADOWRECT:
      case TGSI_TEXTURE_SHADOW1D_ARRAY:
      case TGSI_TEXTURE_SHADOW2D_ARRAY:
      case TGSI_TEXTURE_SHADOWCUBE:
      case TGSI_TEXTURE_SHADOWCUBE_ARRAY:
         shadow = true;
         break;
      default:
         shadow = false;
         break;
      }

      for (i = 0; i < inst->Instruction.NumSrcRegs; i++) {
         const struct tgsi_full_src_register *src = &inst->Src[i];

         if (src->Register.File == TGSI_FILE_SAMPLER) {
            const int idx = src->Register.Index;

            if (idx >= info->num_samplers)
               info->num_samplers = idx + 1;

            if (shadow)
               info->shadow_samplers |= 1 << idx;
         }
      }
   }
}

/**
 * Parse a TGSI property for the shader info.
 */
static void
ilo_shader_info_parse_prop(struct ilo_shader_info *info,
                           const struct tgsi_full_property *prop)
{
   switch (prop->Property.PropertyName) {
   case TGSI_PROPERTY_FS_COLOR0_WRITES_ALL_CBUFS:
      info->fs_color0_writes_all_cbufs = prop->u[0].Data;
      break;
   default:
      break;
   }
}

/**
 * Parse a TGSI declaration for the shader info.
 */
static void
ilo_shader_info_parse_decl(struct ilo_shader_info *info,
                           const struct tgsi_full_declaration *decl)
{
   switch (decl->Declaration.File) {
   case TGSI_FILE_INPUT:
      if (decl->Declaration.Interpolate &&
          decl->Interp.Interpolate == TGSI_INTERPOLATE_COLOR)
         info->has_color_interp = true;
      if (decl->Declaration.Semantic &&
          decl->Semantic.Name == TGSI_SEMANTIC_POSITION)
         info->has_pos = true;
      break;
   case TGSI_FILE_OUTPUT:
      if (decl->Declaration.Semantic &&
          decl->Semantic.Name == TGSI_SEMANTIC_EDGEFLAG)
         info->edgeflag_out = decl->Range.First;
      break;
   case TGSI_FILE_CONSTANT:
      {
         const int idx = (decl->Declaration.Dimension) ?
            decl->Dim.Index2D : 0;
         if (info->constant_buffer_count <= idx)
            info->constant_buffer_count = idx + 1;
      }
      break;
   case TGSI_FILE_SYSTEM_VALUE:
      if (decl->Declaration.Semantic &&
          decl->Semantic.Name == TGSI_SEMANTIC_INSTANCEID)
         info->has_instanceid = true;
      if (decl->Declaration.Semantic &&
          decl->Semantic.Name == TGSI_SEMANTIC_VERTEXID)
         info->has_vertexid = true;
      break;
   default:
      break;
   }
}

static void
ilo_shader_info_parse_tokens(struct ilo_shader_info *info)
{
   struct tgsi_parse_context parse;

   info->edgeflag_in = -1;
   info->edgeflag_out = -1;

   tgsi_parse_init(&parse, info->tokens);
   while (!tgsi_parse_end_of_tokens(&parse)) {
      const union tgsi_full_token *token;

      tgsi_parse_token(&parse);
      token = &parse.FullToken;

      switch (token->Token.Type) {
      case TGSI_TOKEN_TYPE_DECLARATION:
         ilo_shader_info_parse_decl(info, &token->FullDeclaration);
         break;
      case TGSI_TOKEN_TYPE_INSTRUCTION:
         ilo_shader_info_parse_inst(info, &token->FullInstruction);
         break;
      case TGSI_TOKEN_TYPE_PROPERTY:
         ilo_shader_info_parse_prop(info, &token->FullProperty);
         break;
      default:
         break;
      }
   }
   tgsi_parse_free(&parse);
}

/**
 * Create a shader state.
 */
static struct ilo_shader_state *
ilo_shader_state_create(const struct ilo_dev_info *dev,
                        const struct ilo_state_vector *vec,
                        int type, const void *templ)
{
   struct ilo_shader_state *state;
   struct ilo_shader_variant variant;

   state = CALLOC_STRUCT(ilo_shader_state);
   if (!state)
      return NULL;

   state->info.dev = dev;
   state->info.type = type;

   if (type == PIPE_SHADER_COMPUTE) {
      const struct pipe_compute_state *c =
         (const struct pipe_compute_state *) templ;

      state->info.tokens = tgsi_dup_tokens(c->prog);
      state->info.compute.req_local_mem = c->req_local_mem;
      state->info.compute.req_private_mem = c->req_private_mem;
      state->info.compute.req_input_mem = c->req_input_mem;
   }
   else {
      const struct pipe_shader_state *s =
         (const struct pipe_shader_state *) templ;

      state->info.tokens = tgsi_dup_tokens(s->tokens);
      state->info.stream_output = s->stream_output;
   }

   list_inithead(&state->variants);

   ilo_shader_info_parse_tokens(&state->info);

   /* guess and compile now */
   ilo_shader_variant_guess(&variant, &state->info, vec);
   if (!ilo_shader_state_use_variant(state, &variant)) {
      ilo_shader_destroy(state);
      return NULL;
   }

   return state;
}

/**
 * Add a compiled shader to the shader state.
 */
static void
ilo_shader_state_add_shader(struct ilo_shader_state *state,
                            struct ilo_shader *sh)
{
   list_add(&sh->list, &state->variants);
   state->num_variants++;
   state->total_size += sh->kernel_size;

   if (state->cache)
      ilo_shader_cache_notify_change(state->cache, state);
}

/**
 * Remove a compiled shader from the shader state.
 */
static void
ilo_shader_state_remove_shader(struct ilo_shader_state *state,
                               struct ilo_shader *sh)
{
   list_del(&sh->list);
   state->num_variants--;
   state->total_size -= sh->kernel_size;
}

/**
 * Garbage collect shader variants in the shader state.
 */
static void
ilo_shader_state_gc(struct ilo_shader_state *state)
{
   /* activate when the variants take up more than 4KiB of space */
   const int limit = 4 * 1024;
   struct ilo_shader *sh, *next;

   if (state->total_size < limit)
      return;

   /* remove from the tail as the most recently ones are at the head */
   LIST_FOR_EACH_ENTRY_SAFE_REV(sh, next, &state->variants, list) {
      ilo_shader_state_remove_shader(state, sh);
      ilo_shader_destroy_kernel(sh);

      if (state->total_size <= limit / 2)
         break;
   }
}

/**
 * Search for a shader variant.
 */
static struct ilo_shader *
ilo_shader_state_search_variant(struct ilo_shader_state *state,
                                const struct ilo_shader_variant *variant)
{
   struct ilo_shader *sh = NULL, *tmp;

   LIST_FOR_EACH_ENTRY(tmp, &state->variants, list) {
      if (memcmp(&tmp->variant, variant, sizeof(*variant)) == 0) {
         sh = tmp;
         break;
      }
   }

   return sh;
}

static void
copy_so_info(struct ilo_shader *sh,
             const struct pipe_stream_output_info *so_info)
{
   unsigned i, attr;

   if (!so_info->num_outputs)
      return;

   sh->so_info = *so_info;

   for (i = 0; i < so_info->num_outputs; i++) {
      /* figure out which attribute is sourced */
      for (attr = 0; attr < sh->out.count; attr++) {
         const int reg_idx = sh->out.register_indices[attr];
         if (reg_idx == so_info->output[i].register_index)
            break;
      }

      if (attr < sh->out.count) {
         sh->so_info.output[i].register_index = attr;
      }
      else {
         assert(!"stream output an undefined register");
         sh->so_info.output[i].register_index = 0;
      }

      /* PSIZE is at W channel */
      if (sh->out.semantic_names[attr] == TGSI_SEMANTIC_PSIZE) {
         assert(so_info->output[i].start_component == 0);
         assert(so_info->output[i].num_components == 1);
         sh->so_info.output[i].start_component = 3;
      }
   }
}

/**
 * Add a shader variant to the shader state.
 */
static struct ilo_shader *
ilo_shader_state_add_variant(struct ilo_shader_state *state,
                             const struct ilo_shader_variant *variant)
{
   struct ilo_shader *sh;

   switch (state->info.type) {
   case PIPE_SHADER_VERTEX:
      sh = ilo_shader_compile_vs(state, variant);
      break;
   case PIPE_SHADER_FRAGMENT:
      sh = ilo_shader_compile_fs(state, variant);
      break;
   case PIPE_SHADER_GEOMETRY:
      sh = ilo_shader_compile_gs(state, variant);
      break;
   case PIPE_SHADER_COMPUTE:
      sh = ilo_shader_compile_cs(state, variant);
      break;
   default:
      sh = NULL;
      break;
   }
   if (!sh) {
      assert(!"failed to compile shader");
      return NULL;
   }

   sh->variant = *variant;

   copy_so_info(sh, &state->info.stream_output);

   ilo_shader_state_add_shader(state, sh);

   return sh;
}

/**
 * Update state->shader to point to a variant.  If the variant does not exist,
 * it will be added first.
 */
bool
ilo_shader_state_use_variant(struct ilo_shader_state *state,
                             const struct ilo_shader_variant *variant)
{
   struct ilo_shader *sh;
   bool construct_cso = false;

   sh = ilo_shader_state_search_variant(state, variant);
   if (!sh) {
      ilo_shader_state_gc(state);

      sh = ilo_shader_state_add_variant(state, variant);
      if (!sh)
         return false;

      construct_cso = true;
   }

   /* move to head */
   if (state->variants.next != &sh->list) {
      list_del(&sh->list);
      list_add(&sh->list, &state->variants);
   }

   state->shader = sh;

   if (construct_cso) {
      switch (state->info.type) {
      case PIPE_SHADER_VERTEX:
         ilo_gpe_init_vs_cso(state->info.dev, state, &sh->cso);
         break;
      case PIPE_SHADER_GEOMETRY:
         ilo_gpe_init_gs_cso(state->info.dev, state, &sh->cso);
         break;
      case PIPE_SHADER_FRAGMENT:
         ilo_gpe_init_fs_cso(state->info.dev, state, &sh->cso);
         break;
      default:
         break;
      }
   }

   return true;
}

struct ilo_shader_state *
ilo_shader_create_vs(const struct ilo_dev_info *dev,
                     const struct pipe_shader_state *state,
                     const struct ilo_state_vector *precompile)
{
   struct ilo_shader_state *shader;

   shader = ilo_shader_state_create(dev, precompile,
         PIPE_SHADER_VERTEX, state);

   /* states used in ilo_shader_variant_init() */
   shader->info.non_orthogonal_states = ILO_DIRTY_VIEW_VS |
                                        ILO_DIRTY_RASTERIZER |
                                        ILO_DIRTY_CBUF;

   return shader;
}

struct ilo_shader_state *
ilo_shader_create_gs(const struct ilo_dev_info *dev,
                     const struct pipe_shader_state *state,
                     const struct ilo_state_vector *precompile)
{
   struct ilo_shader_state *shader;

   shader = ilo_shader_state_create(dev, precompile,
         PIPE_SHADER_GEOMETRY, state);

   /* states used in ilo_shader_variant_init() */
   shader->info.non_orthogonal_states = ILO_DIRTY_VIEW_GS |
                                        ILO_DIRTY_VS |
                                        ILO_DIRTY_RASTERIZER |
                                        ILO_DIRTY_CBUF;

   return shader;
}

struct ilo_shader_state *
ilo_shader_create_fs(const struct ilo_dev_info *dev,
                     const struct pipe_shader_state *state,
                     const struct ilo_state_vector *precompile)
{
   struct ilo_shader_state *shader;

   shader = ilo_shader_state_create(dev, precompile,
         PIPE_SHADER_FRAGMENT, state);

   /* states used in ilo_shader_variant_init() */
   shader->info.non_orthogonal_states = ILO_DIRTY_VIEW_FS |
                                        ILO_DIRTY_RASTERIZER |
                                        ILO_DIRTY_FB |
                                        ILO_DIRTY_CBUF;

   return shader;
}

struct ilo_shader_state *
ilo_shader_create_cs(const struct ilo_dev_info *dev,
                     const struct pipe_compute_state *state,
                     const struct ilo_state_vector *precompile)
{
   struct ilo_shader_state *shader;

   shader = ilo_shader_state_create(dev, precompile,
         PIPE_SHADER_COMPUTE, state);

   shader->info.non_orthogonal_states = 0;

   return shader;
}

/**
 * Destroy a shader state.
 */
void
ilo_shader_destroy(struct ilo_shader_state *shader)
{
   struct ilo_shader *sh, *next;

   LIST_FOR_EACH_ENTRY_SAFE(sh, next, &shader->variants, list)
      ilo_shader_destroy_kernel(sh);

   FREE((struct tgsi_token *) shader->info.tokens);
   FREE(shader);
}

/**
 * Return the type (PIPE_SHADER_x) of the shader.
 */
int
ilo_shader_get_type(const struct ilo_shader_state *shader)
{
   return shader->info.type;
}

/**
 * Select a kernel for the given context.  This will compile a new kernel if
 * none of the existing kernels work with the context.
 *
 * \param ilo the context
 * \param dirty states of the context that are considered changed
 * \return true if a different kernel is selected
 */
bool
ilo_shader_select_kernel(struct ilo_shader_state *shader,
                         const struct ilo_state_vector *vec,
                         uint32_t dirty)
{
   const struct ilo_shader * const cur = shader->shader;
   struct ilo_shader_variant variant;

   if (!(shader->info.non_orthogonal_states & dirty))
      return false;

   ilo_shader_variant_init(&variant, &shader->info, vec);
   ilo_shader_state_use_variant(shader, &variant);

   return (shader->shader != cur);
}

static int
route_attr(const int *semantics, const int *indices, int len,
           int semantic, int index)
{
   int i;

   for (i = 0; i < len; i++) {
      if (semantics[i] == semantic && indices[i] == index)
         return i;
   }

   /* failed to match for COLOR, try BCOLOR */
   if (semantic == TGSI_SEMANTIC_COLOR) {
      for (i = 0; i < len; i++) {
         if (semantics[i] == TGSI_SEMANTIC_BCOLOR && indices[i] == index)
            return i;
      }
   }

   return -1;
}

/**
 * Select a routing for the given source shader and rasterizer state.
 *
 * \return true if a different routing is selected
 */
bool
ilo_shader_select_kernel_routing(struct ilo_shader_state *shader,
                                 const struct ilo_shader_state *source,
                                 const struct ilo_rasterizer_state *rasterizer)
{
   const uint32_t sprite_coord_enable = rasterizer->state.sprite_coord_enable;
   const bool light_twoside = rasterizer->state.light_twoside;
   struct ilo_shader *kernel = shader->shader;
   struct ilo_kernel_routing *routing = &kernel->routing;
   const int *src_semantics, *src_indices;
   int src_len, max_src_slot;
   int dst_len, dst_slot;

   /* we are constructing 3DSTATE_SBE here */
   ILO_DEV_ASSERT(shader->info.dev, 6, 8);

   assert(kernel);

   if (source) {
      assert(source->shader);
      src_semantics = source->shader->out.semantic_names;
      src_indices = source->shader->out.semantic_indices;
      src_len = source->shader->out.count;
   }
   else {
      src_semantics = kernel->in.semantic_names;
      src_indices = kernel->in.semantic_indices;
      src_len = kernel->in.count;
   }

   /* no change */
   if (kernel->routing_initialized &&
       routing->source_skip + routing->source_len <= src_len &&
       kernel->routing_sprite_coord_enable == sprite_coord_enable &&
       !memcmp(kernel->routing_src_semantics,
          &src_semantics[routing->source_skip],
          sizeof(kernel->routing_src_semantics[0]) * routing->source_len) &&
       !memcmp(kernel->routing_src_indices,
          &src_indices[routing->source_skip],
          sizeof(kernel->routing_src_indices[0]) * routing->source_len))
      return false;

   if (source) {
      /* skip PSIZE and POSITION (how about the optional CLIPDISTs?) */
      assert(src_semantics[0] == TGSI_SEMANTIC_PSIZE);
      assert(src_semantics[1] == TGSI_SEMANTIC_POSITION);
      routing->source_skip = 2;

      routing->source_len = src_len - routing->source_skip;
      src_semantics += routing->source_skip;
      src_indices += routing->source_skip;
   }
   else {
      routing->source_skip = 0;
      routing->source_len = src_len;
   }

   routing->const_interp_enable = kernel->in.const_interp_enable;
   routing->point_sprite_enable = 0;
   routing->swizzle_enable = false;

   assert(kernel->in.count <= Elements(routing->swizzles));
   dst_len = MIN2(kernel->in.count, Elements(routing->swizzles));
   max_src_slot = -1;

   for (dst_slot = 0; dst_slot < dst_len; dst_slot++) {
      const int semantic = kernel->in.semantic_names[dst_slot];
      const int index = kernel->in.semantic_indices[dst_slot];
      int src_slot;

      if (semantic == TGSI_SEMANTIC_GENERIC &&
          (sprite_coord_enable & (1 << index)))
         routing->point_sprite_enable |= 1 << dst_slot;

      if (source) {
         src_slot = route_attr(src_semantics, src_indices,
               routing->source_len, semantic, index);

         /*
          * The source shader stage does not output this attribute.  The value
          * is supposed to be undefined, unless the attribute goes through
          * point sprite replacement or the attribute is
          * TGSI_SEMANTIC_POSITION.  In all cases, we do not care which source
          * attribute is picked.
          *
          * We should update the kernel code and omit the output of
          * TGSI_SEMANTIC_POSITION here.
          */
         if (src_slot < 0)
            src_slot = 0;
      }
      else {
         src_slot = dst_slot;
      }

      routing->swizzles[dst_slot] = src_slot;

      /* use the following slot for two-sided lighting */
      if (semantic == TGSI_SEMANTIC_COLOR && light_twoside &&
          src_slot + 1 < routing->source_len &&
          src_semantics[src_slot + 1] == TGSI_SEMANTIC_BCOLOR &&
          src_indices[src_slot + 1] == index) {
         routing->swizzles[dst_slot] |= GEN8_SBE_SWIZ_INPUTATTR_FACING;
         src_slot++;
      }

      if (routing->swizzles[dst_slot] != dst_slot)
         routing->swizzle_enable = true;

      if (max_src_slot < src_slot)
         max_src_slot = src_slot;
   }

   memset(&routing->swizzles[dst_slot], 0, sizeof(routing->swizzles) -
         sizeof(routing->swizzles[0]) * dst_slot);

   /*
    * From the Sandy Bridge PRM, volume 2 part 1, page 248:
    *
    *     "It is UNDEFINED to set this field (Vertex URB Entry Read Length) to
    *      0 indicating no Vertex URB data to be read.
    *
    *      This field should be set to the minimum length required to read the
    *      maximum source attribute. The maximum source attribute is indicated
    *      by the maximum value of the enabled Attribute # Source Attribute if
    *      Attribute Swizzle Enable is set, Number of Output Attributes-1 if
    *      enable is not set.
    *
    *        read_length = ceiling((max_source_attr+1)/2)
    *
    *      [errata] Corruption/Hang possible if length programmed larger than
    *      recommended"
    */
   routing->source_len = max_src_slot + 1;

   /* remember the states of the source */
   kernel->routing_initialized = true;
   kernel->routing_sprite_coord_enable = sprite_coord_enable;
   memcpy(kernel->routing_src_semantics, src_semantics,
         sizeof(kernel->routing_src_semantics[0]) * routing->source_len);
   memcpy(kernel->routing_src_indices, src_indices,
         sizeof(kernel->routing_src_indices[0]) * routing->source_len);

   return true;
}

/**
 * Return the cache offset of the selected kernel.  This must be called after
 * ilo_shader_select_kernel() and ilo_shader_cache_upload().
 */
uint32_t
ilo_shader_get_kernel_offset(const struct ilo_shader_state *shader)
{
   const struct ilo_shader *kernel = shader->shader;

   assert(kernel && kernel->uploaded);

   return kernel->cache_offset;
}

/**
 * Query a kernel parameter for the selected kernel.
 */
int
ilo_shader_get_kernel_param(const struct ilo_shader_state *shader,
                            enum ilo_kernel_param param)
{
   const struct ilo_shader *kernel = shader->shader;
   int val;

   assert(kernel);

   switch (param) {
   case ILO_KERNEL_INPUT_COUNT:
      val = kernel->in.count;
      break;
   case ILO_KERNEL_OUTPUT_COUNT:
      val = kernel->out.count;
      break;
   case ILO_KERNEL_SAMPLER_COUNT:
      val = shader->info.num_samplers;
      break;
   case ILO_KERNEL_URB_DATA_START_REG:
      val = kernel->in.start_grf;
      break;
   case ILO_KERNEL_SKIP_CBUF0_UPLOAD:
      val = kernel->skip_cbuf0_upload;
      break;
   case ILO_KERNEL_PCB_CBUF0_SIZE:
      val = kernel->pcb.cbuf0_size;
      break;

   case ILO_KERNEL_SURFACE_TOTAL_COUNT:
      val = kernel->bt.total_count;
      break;
   case ILO_KERNEL_SURFACE_TEX_BASE:
      val = kernel->bt.tex_base;
      break;
   case ILO_KERNEL_SURFACE_TEX_COUNT:
      val = kernel->bt.tex_count;
      break;
   case ILO_KERNEL_SURFACE_CONST_BASE:
      val = kernel->bt.const_base;
      break;
   case ILO_KERNEL_SURFACE_CONST_COUNT:
      val = kernel->bt.const_count;
      break;
   case ILO_KERNEL_SURFACE_RES_BASE:
      val = kernel->bt.res_base;
      break;
   case ILO_KERNEL_SURFACE_RES_COUNT:
      val = kernel->bt.res_count;
      break;

   case ILO_KERNEL_VS_INPUT_INSTANCEID:
      val = shader->info.has_instanceid;
      break;
   case ILO_KERNEL_VS_INPUT_VERTEXID:
      val = shader->info.has_vertexid;
      break;
   case ILO_KERNEL_VS_INPUT_EDGEFLAG:
      if (shader->info.edgeflag_in >= 0) {
         /* we rely on the state tracker here */
         assert(shader->info.edgeflag_in == kernel->in.count - 1);
         val = true;
      }
      else {
         val = false;
      }
      break;
   case ILO_KERNEL_VS_PCB_UCP_SIZE:
      val = kernel->pcb.clip_state_size;
      break;
   case ILO_KERNEL_VS_GEN6_SO:
      val = kernel->stream_output;
      break;
   case ILO_KERNEL_VS_GEN6_SO_START_REG:
      val = kernel->gs_start_grf;
      break;
   case ILO_KERNEL_VS_GEN6_SO_POINT_OFFSET:
      val = kernel->gs_offsets[0];
      break;
   case ILO_KERNEL_VS_GEN6_SO_LINE_OFFSET:
      val = kernel->gs_offsets[1];
      break;
   case ILO_KERNEL_VS_GEN6_SO_TRI_OFFSET:
      val = kernel->gs_offsets[2];
      break;
   case ILO_KERNEL_VS_GEN6_SO_SURFACE_COUNT:
      val = kernel->gs_bt_so_count;
      break;

   case ILO_KERNEL_GS_DISCARD_ADJACENCY:
      val = kernel->in.discard_adj;
      break;
   case ILO_KERNEL_GS_GEN6_SVBI_POST_INC:
      val = kernel->svbi_post_inc;
      break;
   case ILO_KERNEL_GS_GEN6_SURFACE_SO_BASE:
      val = kernel->bt.gen6_so_base;
      break;
   case ILO_KERNEL_GS_GEN6_SURFACE_SO_COUNT:
      val = kernel->bt.gen6_so_count;
      break;

   case ILO_KERNEL_FS_INPUT_Z:
   case ILO_KERNEL_FS_INPUT_W:
      val = kernel->in.has_pos;
      break;
   case ILO_KERNEL_FS_OUTPUT_Z:
      val = kernel->out.has_pos;
      break;
   case ILO_KERNEL_FS_USE_KILL:
      val = kernel->has_kill;
      break;
   case ILO_KERNEL_FS_BARYCENTRIC_INTERPOLATIONS:
      val = kernel->in.barycentric_interpolation_mode;
      break;
   case ILO_KERNEL_FS_DISPATCH_16_OFFSET:
      val = 0;
      break;
   case ILO_KERNEL_FS_SURFACE_RT_BASE:
      val = kernel->bt.rt_base;
      break;
   case ILO_KERNEL_FS_SURFACE_RT_COUNT:
      val = kernel->bt.rt_count;
      break;

   case ILO_KERNEL_CS_LOCAL_SIZE:
      val = shader->info.compute.req_local_mem;
      break;
   case ILO_KERNEL_CS_PRIVATE_SIZE:
      val = shader->info.compute.req_private_mem;
      break;
   case ILO_KERNEL_CS_INPUT_SIZE:
      val = shader->info.compute.req_input_mem;
      break;
   case ILO_KERNEL_CS_SIMD_SIZE:
      val = 16;
      break;
   case ILO_KERNEL_CS_SURFACE_GLOBAL_BASE:
      val = kernel->bt.global_base;
      break;
   case ILO_KERNEL_CS_SURFACE_GLOBAL_COUNT:
      val = kernel->bt.global_count;
      break;

   default:
      assert(!"unknown kernel parameter");
      val = 0;
      break;
   }

   return val;
}

/**
 * Return the CSO of the selected kernel.
 */
const struct ilo_shader_cso *
ilo_shader_get_kernel_cso(const struct ilo_shader_state *shader)
{
   const struct ilo_shader *kernel = shader->shader;

   assert(kernel);

   return &kernel->cso;
}

/**
 * Return the SO info of the selected kernel.
 */
const struct pipe_stream_output_info *
ilo_shader_get_kernel_so_info(const struct ilo_shader_state *shader)
{
   const struct ilo_shader *kernel = shader->shader;

   assert(kernel);

   return &kernel->so_info;
}

/**
 * Return the routing info of the selected kernel.
 */
const struct ilo_kernel_routing *
ilo_shader_get_kernel_routing(const struct ilo_shader_state *shader)
{
   const struct ilo_shader *kernel = shader->shader;

   assert(kernel);

   return &kernel->routing;
}