aboutsummaryrefslogtreecommitdiff
path: root/mesalib/src/glsl/link_varyings.cpp
blob: 2c7e4514e15bf77ec49dbd6d17b67a1419b8deff (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
/*
 * Copyright © 2012 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.
 */

/**
 * \file link_varyings.cpp
 *
 * Linker functions related specifically to linking varyings between shader
 * stages.
 */


#include "main/mtypes.h"
#include "glsl_symbol_table.h"
#include "glsl_parser_extras.h"
#include "ir_optimization.h"
#include "linker.h"
#include "link_varyings.h"
#include "main/macros.h"
#include "program/hash_table.h"
#include "program.h"


/**
 * Validate that outputs from one stage match inputs of another
 */
void
cross_validate_outputs_to_inputs(struct gl_shader_program *prog,
				 gl_shader *producer, gl_shader *consumer)
{
   glsl_symbol_table parameters;
   const char *const producer_stage =
      _mesa_glsl_shader_target_name(producer->Type);
   const char *const consumer_stage =
      _mesa_glsl_shader_target_name(consumer->Type);

   /* Find all shader outputs in the "producer" stage.
    */
   foreach_list(node, producer->ir) {
      ir_variable *const var = ((ir_instruction *) node)->as_variable();

      if ((var == NULL) || (var->mode != ir_var_shader_out))
	 continue;

      parameters.add_variable(var);
   }


   /* Find all shader inputs in the "consumer" stage.  Any variables that have
    * matching outputs already in the symbol table must have the same type and
    * qualifiers.
    */
   foreach_list(node, consumer->ir) {
      ir_variable *const input = ((ir_instruction *) node)->as_variable();

      if ((input == NULL) || (input->mode != ir_var_shader_in))
	 continue;

      ir_variable *const output = parameters.get_variable(input->name);
      if (output != NULL) {
	 /* Check that the types match between stages.
	  */
	 if (input->type != output->type) {
	    /* There is a bit of a special case for gl_TexCoord.  This
	     * built-in is unsized by default.  Applications that variable
	     * access it must redeclare it with a size.  There is some
	     * language in the GLSL spec that implies the fragment shader
	     * and vertex shader do not have to agree on this size.  Other
	     * driver behave this way, and one or two applications seem to
	     * rely on it.
	     *
	     * Neither declaration needs to be modified here because the array
	     * sizes are fixed later when update_array_sizes is called.
	     *
	     * From page 48 (page 54 of the PDF) of the GLSL 1.10 spec:
	     *
	     *     "Unlike user-defined varying variables, the built-in
	     *     varying variables don't have a strict one-to-one
	     *     correspondence between the vertex language and the
	     *     fragment language."
	     */
	    if (!output->type->is_array()
		|| (strncmp("gl_", output->name, 3) != 0)) {
	       linker_error(prog,
			    "%s shader output `%s' declared as type `%s', "
			    "but %s shader input declared as type `%s'\n",
			    producer_stage, output->name,
			    output->type->name,
			    consumer_stage, input->type->name);
	       return;
	    }
	 }

	 /* Check that all of the qualifiers match between stages.
	  */
	 if (input->centroid != output->centroid) {
	    linker_error(prog,
			 "%s shader output `%s' %s centroid qualifier, "
			 "but %s shader input %s centroid qualifier\n",
			 producer_stage,
			 output->name,
			 (output->centroid) ? "has" : "lacks",
			 consumer_stage,
			 (input->centroid) ? "has" : "lacks");
	    return;
	 }

	 if (input->invariant != output->invariant) {
	    linker_error(prog,
			 "%s shader output `%s' %s invariant qualifier, "
			 "but %s shader input %s invariant qualifier\n",
			 producer_stage,
			 output->name,
			 (output->invariant) ? "has" : "lacks",
			 consumer_stage,
			 (input->invariant) ? "has" : "lacks");
	    return;
	 }

	 if (input->interpolation != output->interpolation) {
	    linker_error(prog,
			 "%s shader output `%s' specifies %s "
			 "interpolation qualifier, "
			 "but %s shader input specifies %s "
			 "interpolation qualifier\n",
			 producer_stage,
			 output->name,
			 output->interpolation_string(),
			 consumer_stage,
			 input->interpolation_string());
	    return;
	 }
      }
   }
}


/**
 * Initialize this object based on a string that was passed to
 * glTransformFeedbackVaryings.
 *
 * If the input is mal-formed, this call still succeeds, but it sets
 * this->var_name to a mal-formed input, so tfeedback_decl::find_output_var()
 * will fail to find any matching variable.
 */
void
tfeedback_decl::init(struct gl_context *ctx, struct gl_shader_program *prog,
                     const void *mem_ctx, const char *input)
{
   /* We don't have to be pedantic about what is a valid GLSL variable name,
    * because any variable with an invalid name can't exist in the IR anyway.
    */

   this->location = -1;
   this->orig_name = input;
   this->is_clip_distance_mesa = false;
   this->skip_components = 0;
   this->next_buffer_separator = false;
   this->matched_candidate = NULL;

   if (ctx->Extensions.ARB_transform_feedback3) {
      /* Parse gl_NextBuffer. */
      if (strcmp(input, "gl_NextBuffer") == 0) {
         this->next_buffer_separator = true;
         return;
      }

      /* Parse gl_SkipComponents. */
      if (strcmp(input, "gl_SkipComponents1") == 0)
         this->skip_components = 1;
      else if (strcmp(input, "gl_SkipComponents2") == 0)
         this->skip_components = 2;
      else if (strcmp(input, "gl_SkipComponents3") == 0)
         this->skip_components = 3;
      else if (strcmp(input, "gl_SkipComponents4") == 0)
         this->skip_components = 4;

      if (this->skip_components)
         return;
   }

   /* Parse a declaration. */
   const char *base_name_end;
   long subscript = parse_program_resource_name(input, &base_name_end);
   this->var_name = ralloc_strndup(mem_ctx, input, base_name_end - input);
   if (subscript >= 0) {
      this->array_subscript = subscript;
      this->is_subscripted = true;
   } else {
      this->is_subscripted = false;
   }

   /* For drivers that lower gl_ClipDistance to gl_ClipDistanceMESA, this
    * class must behave specially to account for the fact that gl_ClipDistance
    * is converted from a float[8] to a vec4[2].
    */
   if (ctx->ShaderCompilerOptions[MESA_SHADER_VERTEX].LowerClipDistance &&
       strcmp(this->var_name, "gl_ClipDistance") == 0) {
      this->is_clip_distance_mesa = true;
   }
}


/**
 * Determine whether two tfeedback_decl objects refer to the same variable and
 * array index (if applicable).
 */
bool
tfeedback_decl::is_same(const tfeedback_decl &x, const tfeedback_decl &y)
{
   assert(x.is_varying() && y.is_varying());

   if (strcmp(x.var_name, y.var_name) != 0)
      return false;
   if (x.is_subscripted != y.is_subscripted)
      return false;
   if (x.is_subscripted && x.array_subscript != y.array_subscript)
      return false;
   return true;
}


/**
 * Assign a location for this tfeedback_decl object based on the transform
 * feedback candidate found by find_candidate.
 *
 * If an error occurs, the error is reported through linker_error() and false
 * is returned.
 */
bool
tfeedback_decl::assign_location(struct gl_context *ctx,
                                struct gl_shader_program *prog)
{
   assert(this->is_varying());

   unsigned fine_location
      = this->matched_candidate->toplevel_var->location * 4
      + this->matched_candidate->toplevel_var->location_frac
      + this->matched_candidate->offset;

   if (this->matched_candidate->type->is_array()) {
      /* Array variable */
      const unsigned matrix_cols =
         this->matched_candidate->type->fields.array->matrix_columns;
      const unsigned vector_elements =
         this->matched_candidate->type->fields.array->vector_elements;
      unsigned actual_array_size = this->is_clip_distance_mesa ?
         prog->Vert.ClipDistanceArraySize :
         this->matched_candidate->type->array_size();

      if (this->is_subscripted) {
         /* Check array bounds. */
         if (this->array_subscript >= actual_array_size) {
            linker_error(prog, "Transform feedback varying %s has index "
                         "%i, but the array size is %u.",
                         this->orig_name, this->array_subscript,
                         actual_array_size);
            return false;
         }
         unsigned array_elem_size = this->is_clip_distance_mesa ?
            1 : vector_elements * matrix_cols;
         fine_location += array_elem_size * this->array_subscript;
         this->size = 1;
      } else {
         this->size = actual_array_size;
      }
      this->vector_elements = vector_elements;
      this->matrix_columns = matrix_cols;
      if (this->is_clip_distance_mesa)
         this->type = GL_FLOAT;
      else
         this->type = this->matched_candidate->type->fields.array->gl_type;
   } else {
      /* Regular variable (scalar, vector, or matrix) */
      if (this->is_subscripted) {
         linker_error(prog, "Transform feedback varying %s requested, "
                      "but %s is not an array.",
                      this->orig_name, this->var_name);
         return false;
      }
      this->size = 1;
      this->vector_elements = this->matched_candidate->type->vector_elements;
      this->matrix_columns = this->matched_candidate->type->matrix_columns;
      this->type = this->matched_candidate->type->gl_type;
   }
   this->location = fine_location / 4;
   this->location_frac = fine_location % 4;

   /* From GL_EXT_transform_feedback:
    *   A program will fail to link if:
    *
    *   * the total number of components to capture in any varying
    *     variable in <varyings> is greater than the constant
    *     MAX_TRANSFORM_FEEDBACK_SEPARATE_COMPONENTS_EXT and the
    *     buffer mode is SEPARATE_ATTRIBS_EXT;
    */
   if (prog->TransformFeedback.BufferMode == GL_SEPARATE_ATTRIBS &&
       this->num_components() >
       ctx->Const.MaxTransformFeedbackSeparateComponents) {
      linker_error(prog, "Transform feedback varying %s exceeds "
                   "MAX_TRANSFORM_FEEDBACK_SEPARATE_COMPONENTS.",
                   this->orig_name);
      return false;
   }

   return true;
}


unsigned
tfeedback_decl::get_num_outputs() const
{
   if (!this->is_varying()) {
      return 0;
   }

   return (this->num_components() + this->location_frac + 3)/4;
}


/**
 * Update gl_transform_feedback_info to reflect this tfeedback_decl.
 *
 * If an error occurs, the error is reported through linker_error() and false
 * is returned.
 */
bool
tfeedback_decl::store(struct gl_context *ctx, struct gl_shader_program *prog,
                      struct gl_transform_feedback_info *info,
                      unsigned buffer, const unsigned max_outputs) const
{
   assert(!this->next_buffer_separator);

   /* Handle gl_SkipComponents. */
   if (this->skip_components) {
      info->BufferStride[buffer] += this->skip_components;
      return true;
   }

   /* From GL_EXT_transform_feedback:
    *   A program will fail to link if:
    *
    *     * the total number of components to capture is greater than
    *       the constant MAX_TRANSFORM_FEEDBACK_INTERLEAVED_COMPONENTS_EXT
    *       and the buffer mode is INTERLEAVED_ATTRIBS_EXT.
    */
   if (prog->TransformFeedback.BufferMode == GL_INTERLEAVED_ATTRIBS &&
       info->BufferStride[buffer] + this->num_components() >
       ctx->Const.MaxTransformFeedbackInterleavedComponents) {
      linker_error(prog, "The MAX_TRANSFORM_FEEDBACK_INTERLEAVED_COMPONENTS "
                   "limit has been exceeded.");
      return false;
   }

   unsigned location = this->location;
   unsigned location_frac = this->location_frac;
   unsigned num_components = this->num_components();
   while (num_components > 0) {
      unsigned output_size = MIN2(num_components, 4 - location_frac);
      assert(info->NumOutputs < max_outputs);
      info->Outputs[info->NumOutputs].ComponentOffset = location_frac;
      info->Outputs[info->NumOutputs].OutputRegister = location;
      info->Outputs[info->NumOutputs].NumComponents = output_size;
      info->Outputs[info->NumOutputs].OutputBuffer = buffer;
      info->Outputs[info->NumOutputs].DstOffset = info->BufferStride[buffer];
      ++info->NumOutputs;
      info->BufferStride[buffer] += output_size;
      num_components -= output_size;
      location++;
      location_frac = 0;
   }

   info->Varyings[info->NumVarying].Name = ralloc_strdup(prog, this->orig_name);
   info->Varyings[info->NumVarying].Type = this->type;
   info->Varyings[info->NumVarying].Size = this->size;
   info->NumVarying++;

   return true;
}


const tfeedback_candidate *
tfeedback_decl::find_candidate(gl_shader_program *prog,
                               hash_table *tfeedback_candidates)
{
   const char *name = this->is_clip_distance_mesa
      ? "gl_ClipDistanceMESA" : this->var_name;
   this->matched_candidate = (const tfeedback_candidate *)
      hash_table_find(tfeedback_candidates, name);
   if (!this->matched_candidate) {
      /* From GL_EXT_transform_feedback:
       *   A program will fail to link if:
       *
       *   * any variable name specified in the <varyings> array is not
       *     declared as an output in the geometry shader (if present) or
       *     the vertex shader (if no geometry shader is present);
       */
      linker_error(prog, "Transform feedback varying %s undeclared.",
                   this->orig_name);
   }
   return this->matched_candidate;
}


/**
 * Parse all the transform feedback declarations that were passed to
 * glTransformFeedbackVaryings() and store them in tfeedback_decl objects.
 *
 * If an error occurs, the error is reported through linker_error() and false
 * is returned.
 */
bool
parse_tfeedback_decls(struct gl_context *ctx, struct gl_shader_program *prog,
                      const void *mem_ctx, unsigned num_names,
                      char **varying_names, tfeedback_decl *decls)
{
   for (unsigned i = 0; i < num_names; ++i) {
      decls[i].init(ctx, prog, mem_ctx, varying_names[i]);

      if (!decls[i].is_varying())
         continue;

      /* From GL_EXT_transform_feedback:
       *   A program will fail to link if:
       *
       *   * any two entries in the <varyings> array specify the same varying
       *     variable;
       *
       * We interpret this to mean "any two entries in the <varyings> array
       * specify the same varying variable and array index", since transform
       * feedback of arrays would be useless otherwise.
       */
      for (unsigned j = 0; j < i; ++j) {
         if (!decls[j].is_varying())
            continue;

         if (tfeedback_decl::is_same(decls[i], decls[j])) {
            linker_error(prog, "Transform feedback varying %s specified "
                         "more than once.", varying_names[i]);
            return false;
         }
      }
   }
   return true;
}


/**
 * Store transform feedback location assignments into
 * prog->LinkedTransformFeedback based on the data stored in tfeedback_decls.
 *
 * If an error occurs, the error is reported through linker_error() and false
 * is returned.
 */
bool
store_tfeedback_info(struct gl_context *ctx, struct gl_shader_program *prog,
                     unsigned num_tfeedback_decls,
                     tfeedback_decl *tfeedback_decls)
{
   bool separate_attribs_mode =
      prog->TransformFeedback.BufferMode == GL_SEPARATE_ATTRIBS;

   ralloc_free(prog->LinkedTransformFeedback.Varyings);
   ralloc_free(prog->LinkedTransformFeedback.Outputs);

   memset(&prog->LinkedTransformFeedback, 0,
          sizeof(prog->LinkedTransformFeedback));

   prog->LinkedTransformFeedback.Varyings =
      rzalloc_array(prog,
		    struct gl_transform_feedback_varying_info,
		    num_tfeedback_decls);

   unsigned num_outputs = 0;
   for (unsigned i = 0; i < num_tfeedback_decls; ++i)
      num_outputs += tfeedback_decls[i].get_num_outputs();

   prog->LinkedTransformFeedback.Outputs =
      rzalloc_array(prog,
                    struct gl_transform_feedback_output,
                    num_outputs);

   unsigned num_buffers = 0;

   if (separate_attribs_mode) {
      /* GL_SEPARATE_ATTRIBS */
      for (unsigned i = 0; i < num_tfeedback_decls; ++i) {
         if (!tfeedback_decls[i].store(ctx, prog, &prog->LinkedTransformFeedback,
                                       num_buffers, num_outputs))
            return false;

         num_buffers++;
      }
   }
   else {
      /* GL_INVERLEAVED_ATTRIBS */
      for (unsigned i = 0; i < num_tfeedback_decls; ++i) {
         if (tfeedback_decls[i].is_next_buffer_separator()) {
            num_buffers++;
            continue;
         }

         if (!tfeedback_decls[i].store(ctx, prog,
                                       &prog->LinkedTransformFeedback,
                                       num_buffers, num_outputs))
            return false;
      }
      num_buffers++;
   }

   assert(prog->LinkedTransformFeedback.NumOutputs == num_outputs);

   prog->LinkedTransformFeedback.NumBuffers = num_buffers;
   return true;
}


/**
 * Data structure recording the relationship between outputs of one shader
 * stage (the "producer") and inputs of another (the "consumer").
 */
class varying_matches
{
public:
   varying_matches(bool disable_varying_packing, bool consumer_is_fs);
   ~varying_matches();
   void record(ir_variable *producer_var, ir_variable *consumer_var);
   unsigned assign_locations();
   void store_locations(unsigned producer_base, unsigned consumer_base) const;

private:
   /**
    * If true, this driver disables varying packing, so all varyings need to
    * be aligned on slot boundaries, and take up a number of slots equal to
    * their number of matrix columns times their array size.
    */
   const bool disable_varying_packing;

   /**
    * Enum representing the order in which varyings are packed within a
    * packing class.
    *
    * Currently we pack vec4's first, then vec2's, then scalar values, then
    * vec3's.  This order ensures that the only vectors that are at risk of
    * having to be "double parked" (split between two adjacent varying slots)
    * are the vec3's.
    */
   enum packing_order_enum {
      PACKING_ORDER_VEC4,
      PACKING_ORDER_VEC2,
      PACKING_ORDER_SCALAR,
      PACKING_ORDER_VEC3,
   };

   static unsigned compute_packing_class(ir_variable *var);
   static packing_order_enum compute_packing_order(ir_variable *var);
   static int match_comparator(const void *x_generic, const void *y_generic);

   /**
    * Structure recording the relationship between a single producer output
    * and a single consumer input.
    */
   struct match {
      /**
       * Packing class for this varying, computed by compute_packing_class().
       */
      unsigned packing_class;

      /**
       * Packing order for this varying, computed by compute_packing_order().
       */
      packing_order_enum packing_order;
      unsigned num_components;

      /**
       * The output variable in the producer stage.
       */
      ir_variable *producer_var;

      /**
       * The input variable in the consumer stage.
       */
      ir_variable *consumer_var;

      /**
       * The location which has been assigned for this varying.  This is
       * expressed in multiples of a float, with the first generic varying
       * (i.e. the one referred to by VARYING_SLOT_VAR0) represented by the
       * value 0.
       */
      unsigned generic_location;
   } *matches;

   /**
    * The number of elements in the \c matches array that are currently in
    * use.
    */
   unsigned num_matches;

   /**
    * The number of elements that were set aside for the \c matches array when
    * it was allocated.
    */
   unsigned matches_capacity;

   const bool consumer_is_fs;
};


varying_matches::varying_matches(bool disable_varying_packing,
                                 bool consumer_is_fs)
   : disable_varying_packing(disable_varying_packing),
     consumer_is_fs(consumer_is_fs)
{
   /* Note: this initial capacity is rather arbitrarily chosen to be large
    * enough for many cases without wasting an unreasonable amount of space.
    * varying_matches::record() will resize the array if there are more than
    * this number of varyings.
    */
   this->matches_capacity = 8;
   this->matches = (match *)
      malloc(sizeof(*this->matches) * this->matches_capacity);
   this->num_matches = 0;
}


varying_matches::~varying_matches()
{
   free(this->matches);
}


/**
 * Record the given producer/consumer variable pair in the list of variables
 * that should later be assigned locations.
 *
 * It is permissible for \c consumer_var to be NULL (this happens if a
 * variable is output by the producer and consumed by transform feedback, but
 * not consumed by the consumer).
 *
 * If \c producer_var has already been paired up with a consumer_var, or
 * producer_var is part of fixed pipeline functionality (and hence already has
 * a location assigned), this function has no effect.
 *
 * Note: as a side effect this function may change the interpolation type of
 * \c producer_var, but only when the change couldn't possibly affect
 * rendering.
 */
void
varying_matches::record(ir_variable *producer_var, ir_variable *consumer_var)
{
   if (!producer_var->is_unmatched_generic_inout) {
      /* Either a location already exists for this variable (since it is part
       * of fixed functionality), or it has already been recorded as part of a
       * previous match.
       */
      return;
   }

   if ((consumer_var == NULL && producer_var->type->contains_integer()) ||
       !consumer_is_fs) {
      /* Since this varying is not being consumed by the fragment shader, its
       * interpolation type varying cannot possibly affect rendering.  Also,
       * this variable is non-flat and is (or contains) an integer.
       *
       * lower_packed_varyings requires all integer varyings to flat,
       * regardless of where they appear.  We can trivially satisfy that
       * requirement by changing the interpolation type to flat here.
       */
      producer_var->centroid = false;
      producer_var->interpolation = INTERP_QUALIFIER_FLAT;

      if (consumer_var) {
         consumer_var->centroid = false;
         consumer_var->interpolation = INTERP_QUALIFIER_FLAT;
      }
   }

   if (this->num_matches == this->matches_capacity) {
      this->matches_capacity *= 2;
      this->matches = (match *)
         realloc(this->matches,
                 sizeof(*this->matches) * this->matches_capacity);
   }
   this->matches[this->num_matches].packing_class
      = this->compute_packing_class(producer_var);
   this->matches[this->num_matches].packing_order
      = this->compute_packing_order(producer_var);
   if (this->disable_varying_packing) {
      unsigned slots = producer_var->type->is_array()
         ? (producer_var->type->length
            * producer_var->type->fields.array->matrix_columns)
         : producer_var->type->matrix_columns;
      this->matches[this->num_matches].num_components = 4 * slots;
   } else {
      this->matches[this->num_matches].num_components
         = producer_var->type->component_slots();
   }
   this->matches[this->num_matches].producer_var = producer_var;
   this->matches[this->num_matches].consumer_var = consumer_var;
   this->num_matches++;
   producer_var->is_unmatched_generic_inout = 0;
   if (consumer_var)
      consumer_var->is_unmatched_generic_inout = 0;
}


/**
 * Choose locations for all of the variable matches that were previously
 * passed to varying_matches::record().
 */
unsigned
varying_matches::assign_locations()
{
   /* Sort varying matches into an order that makes them easy to pack. */
   qsort(this->matches, this->num_matches, sizeof(*this->matches),
         &varying_matches::match_comparator);

   unsigned generic_location = 0;

   for (unsigned i = 0; i < this->num_matches; i++) {
      /* Advance to the next slot if this varying has a different packing
       * class than the previous one, and we're not already on a slot
       * boundary.
       */
      if (i > 0 &&
          this->matches[i - 1].packing_class
          != this->matches[i].packing_class) {
         generic_location = ALIGN(generic_location, 4);
      }

      this->matches[i].generic_location = generic_location;

      generic_location += this->matches[i].num_components;
   }

   return (generic_location + 3) / 4;
}


/**
 * Update the producer and consumer shaders to reflect the locations
 * assignments that were made by varying_matches::assign_locations().
 */
void
varying_matches::store_locations(unsigned producer_base,
                                 unsigned consumer_base) const
{
   for (unsigned i = 0; i < this->num_matches; i++) {
      ir_variable *producer_var = this->matches[i].producer_var;
      ir_variable *consumer_var = this->matches[i].consumer_var;
      unsigned generic_location = this->matches[i].generic_location;
      unsigned slot = generic_location / 4;
      unsigned offset = generic_location % 4;

      producer_var->location = producer_base + slot;
      producer_var->location_frac = offset;
      if (consumer_var) {
         assert(consumer_var->location == -1);
         consumer_var->location = consumer_base + slot;
         consumer_var->location_frac = offset;
      }
   }
}


/**
 * Compute the "packing class" of the given varying.  This is an unsigned
 * integer with the property that two variables in the same packing class can
 * be safely backed into the same vec4.
 */
unsigned
varying_matches::compute_packing_class(ir_variable *var)
{
   /* Without help from the back-end, there is no way to pack together
    * variables with different interpolation types, because
    * lower_packed_varyings must choose exactly one interpolation type for
    * each packed varying it creates.
    *
    * However, we can safely pack together floats, ints, and uints, because:
    *
    * - varyings of base type "int" and "uint" must use the "flat"
    *   interpolation type, which can only occur in GLSL 1.30 and above.
    *
    * - On platforms that support GLSL 1.30 and above, lower_packed_varyings
    *   can store flat floats as ints without losing any information (using
    *   the ir_unop_bitcast_* opcodes).
    *
    * Therefore, the packing class depends only on the interpolation type.
    */
   unsigned packing_class = var->centroid ? 1 : 0;
   packing_class *= 4;
   packing_class += var->interpolation;
   return packing_class;
}


/**
 * Compute the "packing order" of the given varying.  This is a sort key we
 * use to determine when to attempt to pack the given varying relative to
 * other varyings in the same packing class.
 */
varying_matches::packing_order_enum
varying_matches::compute_packing_order(ir_variable *var)
{
   const glsl_type *element_type = var->type;

   while (element_type->base_type == GLSL_TYPE_ARRAY) {
      element_type = element_type->fields.array;
   }

   switch (element_type->component_slots() % 4) {
   case 1: return PACKING_ORDER_SCALAR;
   case 2: return PACKING_ORDER_VEC2;
   case 3: return PACKING_ORDER_VEC3;
   case 0: return PACKING_ORDER_VEC4;
   default:
      assert(!"Unexpected value of vector_elements");
      return PACKING_ORDER_VEC4;
   }
}


/**
 * Comparison function passed to qsort() to sort varyings by packing_class and
 * then by packing_order.
 */
int
varying_matches::match_comparator(const void *x_generic, const void *y_generic)
{
   const match *x = (const match *) x_generic;
   const match *y = (const match *) y_generic;

   if (x->packing_class != y->packing_class)
      return x->packing_class - y->packing_class;
   return x->packing_order - y->packing_order;
}


/**
 * Is the given variable a varying variable to be counted against the
 * limit in ctx->Const.MaxVarying?
 * This includes variables such as texcoords, colors and generic
 * varyings, but excludes variables such as gl_FrontFacing and gl_FragCoord.
 */
static bool
is_varying_var(GLenum shaderType, const ir_variable *var)
{
   /* Only fragment shaders will take a varying variable as an input */
   if (shaderType == GL_FRAGMENT_SHADER &&
       var->mode == ir_var_shader_in) {
      switch (var->location) {
      case VARYING_SLOT_POS:
      case VARYING_SLOT_FACE:
      case VARYING_SLOT_PNTC:
         return false;
      default:
         return true;
      }
   }
   return false;
}


/**
 * Visitor class that generates tfeedback_candidate structs describing all
 * possible targets of transform feedback.
 *
 * tfeedback_candidate structs are stored in the hash table
 * tfeedback_candidates, which is passed to the constructor.  This hash table
 * maps varying names to instances of the tfeedback_candidate struct.
 */
class tfeedback_candidate_generator : public program_resource_visitor
{
public:
   tfeedback_candidate_generator(void *mem_ctx,
                                 hash_table *tfeedback_candidates)
      : mem_ctx(mem_ctx),
        tfeedback_candidates(tfeedback_candidates),
        toplevel_var(NULL),
        varying_floats(0)
   {
   }

   void process(ir_variable *var)
   {
      this->toplevel_var = var;
      this->varying_floats = 0;
      if (var->is_interface_instance())
         program_resource_visitor::process(var->interface_type,
                                           var->interface_type->name);
      else
         program_resource_visitor::process(var);
   }

private:
   virtual void visit_field(const glsl_type *type, const char *name,
                            bool row_major)
   {
      assert(!type->is_record());
      assert(!(type->is_array() && type->fields.array->is_record()));
      assert(!type->is_interface());
      assert(!(type->is_array() && type->fields.array->is_interface()));

      (void) row_major;

      tfeedback_candidate *candidate
         = rzalloc(this->mem_ctx, tfeedback_candidate);
      candidate->toplevel_var = this->toplevel_var;
      candidate->type = type;
      candidate->offset = this->varying_floats;
      hash_table_insert(this->tfeedback_candidates, candidate,
                        ralloc_strdup(this->mem_ctx, name));
      this->varying_floats += type->component_slots();
   }

   /**
    * Memory context used to allocate hash table keys and values.
    */
   void * const mem_ctx;

   /**
    * Hash table in which tfeedback_candidate objects should be stored.
    */
   hash_table * const tfeedback_candidates;

   /**
    * Pointer to the toplevel variable that is being traversed.
    */
   ir_variable *toplevel_var;

   /**
    * Total number of varying floats that have been visited so far.  This is
    * used to determine the offset to each varying within the toplevel
    * variable.
    */
   unsigned varying_floats;
};


/**
 * Assign locations for all variables that are produced in one pipeline stage
 * (the "producer") and consumed in the next stage (the "consumer").
 *
 * Variables produced by the producer may also be consumed by transform
 * feedback.
 *
 * \param num_tfeedback_decls is the number of declarations indicating
 *        variables that may be consumed by transform feedback.
 *
 * \param tfeedback_decls is a pointer to an array of tfeedback_decl objects
 *        representing the result of parsing the strings passed to
 *        glTransformFeedbackVaryings().  assign_location() will be called for
 *        each of these objects that matches one of the outputs of the
 *        producer.
 *
 * When num_tfeedback_decls is nonzero, it is permissible for the consumer to
 * be NULL.  In this case, varying locations are assigned solely based on the
 * requirements of transform feedback.
 */
bool
assign_varying_locations(struct gl_context *ctx,
			 void *mem_ctx,
			 struct gl_shader_program *prog,
			 gl_shader *producer, gl_shader *consumer,
                         unsigned num_tfeedback_decls,
                         tfeedback_decl *tfeedback_decls)
{
   const unsigned producer_base = VARYING_SLOT_VAR0;
   const unsigned consumer_base = VARYING_SLOT_VAR0;
   varying_matches matches(ctx->Const.DisableVaryingPacking,
                           consumer && consumer->Type == GL_FRAGMENT_SHADER);
   hash_table *tfeedback_candidates
      = hash_table_ctor(0, hash_table_string_hash, hash_table_string_compare);
   hash_table *consumer_inputs
      = hash_table_ctor(0, hash_table_string_hash, hash_table_string_compare);
   hash_table *consumer_interface_inputs
      = hash_table_ctor(0, hash_table_string_hash, hash_table_string_compare);

   /* Operate in a total of three passes.
    *
    * 1. Assign locations for any matching inputs and outputs.
    *
    * 2. Mark output variables in the producer that do not have locations as
    *    not being outputs.  This lets the optimizer eliminate them.
    *
    * 3. Mark input variables in the consumer that do not have locations as
    *    not being inputs.  This lets the optimizer eliminate them.
    */

   if (consumer) {
      foreach_list(node, consumer->ir) {
         ir_variable *const input_var =
            ((ir_instruction *) node)->as_variable();

         if ((input_var != NULL) && (input_var->mode == ir_var_shader_in)) {
            if (input_var->interface_type != NULL) {
               char *const iface_field_name =
                  ralloc_asprintf(mem_ctx, "%s.%s",
                                  input_var->interface_type->name,
                                  input_var->name);
               hash_table_insert(consumer_interface_inputs, input_var,
                                 iface_field_name);
            } else {
               hash_table_insert(consumer_inputs, input_var,
                                 ralloc_strdup(mem_ctx, input_var->name));
            }
         }
      }
   }

   foreach_list(node, producer->ir) {
      ir_variable *const output_var = ((ir_instruction *) node)->as_variable();

      if ((output_var == NULL) || (output_var->mode != ir_var_shader_out))
	 continue;

      tfeedback_candidate_generator g(mem_ctx, tfeedback_candidates);
      g.process(output_var);

      ir_variable *input_var;
      if (output_var->interface_type != NULL) {
         char *const iface_field_name =
            ralloc_asprintf(mem_ctx, "%s.%s",
                            output_var->interface_type->name,
                            output_var->name);
         input_var =
            (ir_variable *) hash_table_find(consumer_interface_inputs,
                                            iface_field_name);
      } else {
         input_var =
            (ir_variable *) hash_table_find(consumer_inputs, output_var->name);
      }

      if (input_var && input_var->mode != ir_var_shader_in)
         input_var = NULL;

      if (input_var) {
         matches.record(output_var, input_var);
      }
   }

   for (unsigned i = 0; i < num_tfeedback_decls; ++i) {
      if (!tfeedback_decls[i].is_varying())
         continue;

      const tfeedback_candidate *matched_candidate
         = tfeedback_decls[i].find_candidate(prog, tfeedback_candidates);

      if (matched_candidate == NULL) {
         hash_table_dtor(tfeedback_candidates);
         hash_table_dtor(consumer_inputs);
         hash_table_dtor(consumer_interface_inputs);
         return false;
      }

      if (matched_candidate->toplevel_var->is_unmatched_generic_inout)
         matches.record(matched_candidate->toplevel_var, NULL);
   }

   const unsigned slots_used = matches.assign_locations();
   matches.store_locations(producer_base, consumer_base);

   for (unsigned i = 0; i < num_tfeedback_decls; ++i) {
      if (!tfeedback_decls[i].is_varying())
         continue;

      if (!tfeedback_decls[i].assign_location(ctx, prog)) {
         hash_table_dtor(tfeedback_candidates);
         hash_table_dtor(consumer_inputs);
         hash_table_dtor(consumer_interface_inputs);
         return false;
      }
   }

   hash_table_dtor(tfeedback_candidates);
   hash_table_dtor(consumer_inputs);
   hash_table_dtor(consumer_interface_inputs);

   if (ctx->Const.DisableVaryingPacking) {
      /* Transform feedback code assumes varyings are packed, so if the driver
       * has disabled varying packing, make sure it does not support transform
       * feedback.
       */
      assert(!ctx->Extensions.EXT_transform_feedback);
   } else {
      lower_packed_varyings(mem_ctx, producer_base, slots_used,
                            ir_var_shader_out, producer);
      if (consumer) {
         lower_packed_varyings(mem_ctx, consumer_base, slots_used,
                               ir_var_shader_in, consumer);
      }
   }

   if (consumer) {
      foreach_list(node, consumer->ir) {
         ir_variable *const var = ((ir_instruction *) node)->as_variable();

         if (var && var->mode == ir_var_shader_in &&
             var->is_unmatched_generic_inout) {
            if (prog->Version <= 120) {
               /* On page 25 (page 31 of the PDF) of the GLSL 1.20 spec:
                *
                *     Only those varying variables used (i.e. read) in
                *     the fragment shader executable must be written to
                *     by the vertex shader executable; declaring
                *     superfluous varying variables in a vertex shader is
                *     permissible.
                *
                * We interpret this text as meaning that the VS must
                * write the variable for the FS to read it.  See
                * "glsl1-varying read but not written" in piglit.
                */

               linker_error(prog, "%s shader varying %s not written "
                            "by %s shader\n.",
                            _mesa_glsl_shader_target_name(consumer->Type),
			    var->name,
                            _mesa_glsl_shader_target_name(producer->Type));
            }

            /* An 'in' variable is only really a shader input if its
             * value is written by the previous stage.
             */
            var->mode = ir_var_auto;
         }
      }
   }

   return true;
}

bool
check_against_varying_limit(struct gl_context *ctx,
                            struct gl_shader_program *prog,
                            gl_shader *consumer)
{
   unsigned varying_vectors = 0;

   foreach_list(node, consumer->ir) {
      ir_variable *const var = ((ir_instruction *) node)->as_variable();

      if (var && var->mode == ir_var_shader_in &&
          is_varying_var(consumer->Type, var)) {
         /* The packing rules used for vertex shader inputs are also
          * used for fragment shader inputs.
          */
         varying_vectors += count_attribute_slots(var->type);
      }
   }

   if (ctx->API == API_OPENGLES2 || prog->IsES) {
      if (varying_vectors > ctx->Const.MaxVarying) {
         linker_error(prog, "shader uses too many varying vectors "
                      "(%u > %u)\n",
                      varying_vectors, ctx->Const.MaxVarying);
         return false;
      }
   } else {
      const unsigned float_components = varying_vectors * 4;
      if (float_components > ctx->Const.MaxVarying * 4) {
         linker_error(prog, "shader uses too many varying components "
                      "(%u > %u)\n",
                      float_components, ctx->Const.MaxVarying * 4);
         return false;
      }
   }

   return true;
}