summaryrefslogtreecommitdiffstats
path: root/src/glsl/linker.cpp
blob: ec3cc01d4046127909797275d43deeec1ebd1769 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
/*
 * Copyright © 2010 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 linker.cpp
 * GLSL linker implementation
 *
 * Given a set of shaders that are to be linked to generate a final program,
 * there are three distinct stages.
 *
 * In the first stage shaders are partitioned into groups based on the shader
 * type.  All shaders of a particular type (e.g., vertex shaders) are linked
 * together.
 *
 *   - Undefined references in each shader are resolve to definitions in
 *     another shader.
 *   - Types and qualifiers of uniforms, outputs, and global variables defined
 *     in multiple shaders with the same name are verified to be the same.
 *   - Initializers for uniforms and global variables defined
 *     in multiple shaders with the same name are verified to be the same.
 *
 * The result, in the terminology of the GLSL spec, is a set of shader
 * executables for each processing unit.
 *
 * After the first stage is complete, a series of semantic checks are performed
 * on each of the shader executables.
 *
 *   - Each shader executable must define a \c main function.
 *   - Each vertex shader executable must write to \c gl_Position.
 *   - Each fragment shader executable must write to either \c gl_FragData or
 *     \c gl_FragColor.
 *
 * In the final stage individual shader executables are linked to create a
 * complete exectuable.
 *
 *   - Types of uniforms defined in multiple shader stages with the same name
 *     are verified to be the same.
 *   - Initializers for uniforms defined in multiple shader stages with the
 *     same name are verified to be the same.
 *   - Types and qualifiers of outputs defined in one stage are verified to
 *     be the same as the types and qualifiers of inputs defined with the same
 *     name in a later stage.
 *
 * \author Ian Romanick <ian.d.romanick@intel.com>
 */
#include <cstdlib>
#include <cstdio>
#include <cstdarg>
#include <climits>

extern "C" {
#include <talloc.h>
}

#include "main/mtypes.h"
#include "main/macros.h"
#include "main/shaderobj.h"
#include "glsl_symbol_table.h"
#include "ir.h"
#include "program.h"
#include "hash_table.h"
#include "linker.h"
#include "ir_optimization.h"

/**
 * Visitor that determines whether or not a variable is ever written.
 */
class find_assignment_visitor : public ir_hierarchical_visitor {
public:
   find_assignment_visitor(const char *name)
      : name(name), found(false)
   {
      /* empty */
   }

   virtual ir_visitor_status visit_enter(ir_assignment *ir)
   {
      ir_variable *const var = ir->lhs->variable_referenced();

      if (strcmp(name, var->name) == 0) {
	 found = true;
	 return visit_stop;
      }

      return visit_continue_with_parent;
   }

   bool variable_found()
   {
      return found;
   }

private:
   const char *name;       /**< Find writes to a variable with this name. */
   bool found;             /**< Was a write to the variable found? */
};


void
linker_error_printf(gl_shader_program *prog, const char *fmt, ...)
{
   va_list ap;

   prog->InfoLog = talloc_strdup_append(prog->InfoLog, "error: ");
   va_start(ap, fmt);
   prog->InfoLog = talloc_vasprintf_append(prog->InfoLog, fmt, ap);
   va_end(ap);
}


void
invalidate_variable_locations(gl_shader *sh, enum ir_variable_mode mode,
			      int generic_base)
{
   foreach_list(node, sh->ir) {
      ir_variable *const var = ((ir_instruction *) node)->as_variable();

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

      /* Only assign locations for generic attributes / varyings / etc.
       */
      if (var->location >= generic_base)
	  var->location = -1;
   }
}


/**
 * Determine the number of attribute slots required for a particular type
 *
 * This code is here because it implements the language rules of a specific
 * GLSL version.  Since it's a property of the language and not a property of
 * types in general, it doesn't really belong in glsl_type.
 */
unsigned
count_attribute_slots(const glsl_type *t)
{
   /* From page 31 (page 37 of the PDF) of the GLSL 1.50 spec:
    *
    *     "A scalar input counts the same amount against this limit as a vec4,
    *     so applications may want to consider packing groups of four
    *     unrelated float inputs together into a vector to better utilize the
    *     capabilities of the underlying hardware. A matrix input will use up
    *     multiple locations.  The number of locations used will equal the
    *     number of columns in the matrix."
    *
    * The spec does not explicitly say how arrays are counted.  However, it
    * should be safe to assume the total number of slots consumed by an array
    * is the number of entries in the array multiplied by the number of slots
    * consumed by a single element of the array.
    */

   if (t->is_array())
      return t->array_size() * count_attribute_slots(t->element_type());

   if (t->is_matrix())
      return t->matrix_columns;

   return 1;
}


/**
 * Verify that a vertex shader executable meets all semantic requirements
 *
 * \param shader  Vertex shader executable to be verified
 */
bool
validate_vertex_shader_executable(struct gl_shader_program *prog,
				  struct gl_shader *shader)
{
   if (shader == NULL)
      return true;

   find_assignment_visitor find("gl_Position");
   find.run(shader->ir);
   if (!find.variable_found()) {
      linker_error_printf(prog,
			  "vertex shader does not write to `gl_Position'\n");
      return false;
   }

   return true;
}


/**
 * Verify that a fragment shader executable meets all semantic requirements
 *
 * \param shader  Fragment shader executable to be verified
 */
bool
validate_fragment_shader_executable(struct gl_shader_program *prog,
				    struct gl_shader *shader)
{
   if (shader == NULL)
      return true;

   find_assignment_visitor frag_color("gl_FragColor");
   find_assignment_visitor frag_data("gl_FragData");

   frag_color.run(shader->ir);
   frag_data.run(shader->ir);

   if (frag_color.variable_found() && frag_data.variable_found()) {
      linker_error_printf(prog,  "fragment shader writes to both "
			  "`gl_FragColor' and `gl_FragData'\n");
      return false;
   }

   return true;
}


/**
 * Generate a string describing the mode of a variable
 */
static const char *
mode_string(const ir_variable *var)
{
   switch (var->mode) {
   case ir_var_auto:
      return (var->read_only) ? "global constant" : "global variable";

   case ir_var_uniform: return "uniform";
   case ir_var_in:      return "shader input";
   case ir_var_out:     return "shader output";
   case ir_var_inout:   return "shader inout";

   case ir_var_temporary:
   default:
      assert(!"Should not get here.");
      return "invalid variable";
   }
}


/**
 * Perform validation of global variables used across multiple shaders
 */
bool
cross_validate_globals(struct gl_shader_program *prog,
		       struct gl_shader **shader_list,
		       unsigned num_shaders,
		       bool uniforms_only)
{
   /* Examine all of the uniforms in all of the shaders and cross validate
    * them.
    */
   glsl_symbol_table variables;
   for (unsigned i = 0; i < num_shaders; i++) {
      foreach_list(node, shader_list[i]->ir) {
	 ir_variable *const var = ((ir_instruction *) node)->as_variable();

	 if (var == NULL)
	    continue;

	 if (uniforms_only && (var->mode != ir_var_uniform))
	    continue;

	 /* Don't cross validate temporaries that are at global scope.  These
	  * will eventually get pulled into the shaders 'main'.
	  */
	 if (var->mode == ir_var_temporary)
	    continue;

	 /* If a global with this name has already been seen, verify that the
	  * new instance has the same type.  In addition, if the globals have
	  * initializers, the values of the initializers must be the same.
	  */
	 ir_variable *const existing = variables.get_variable(var->name);
	 if (existing != NULL) {
	    if (var->type != existing->type) {
	       linker_error_printf(prog, "%s `%s' declared as type "
				   "`%s' and type `%s'\n",
				   mode_string(var),
				   var->name, var->type->name,
				   existing->type->name);
	       return false;
	    }

	    /* FINISHME: Handle non-constant initializers.
	     */
	    if (var->constant_value != NULL) {
	       if (existing->constant_value != NULL) {
		  if (!var->constant_value->has_value(existing->constant_value)) {
		     linker_error_printf(prog, "initializers for %s "
					 "`%s' have differing values\n",
					 mode_string(var), var->name);
		     return false;
		  }
	       } else
		  /* If the first-seen instance of a particular uniform did not
		   * have an initializer but a later instance does, copy the
		   * initializer to the version stored in the symbol table.
		   */
		  /* FINISHME: This is wrong.  The constant_value field should
		   * FINISHME: not be modified!  Imagine a case where a shader
		   * FINISHME: without an initializer is linked in two different
		   * FINISHME: programs with shaders that have differing
		   * FINISHME: initializers.  Linking with the first will
		   * FINISHME: modify the shader, and linking with the second
		   * FINISHME: will fail.
		   */
		  existing->constant_value = var->constant_value->clone(NULL);
	    }
	 } else
	    variables.add_variable(var->name, var);
      }
   }

   return true;
}


/**
 * Perform validation of uniforms used across multiple shader stages
 */
bool
cross_validate_uniforms(struct gl_shader_program *prog)
{
   return cross_validate_globals(prog, prog->_LinkedShaders,
				 prog->_NumLinkedShaders, true);
}


/**
 * Validate that outputs from one stage match inputs of another
 */
bool
cross_validate_outputs_to_inputs(struct gl_shader_program *prog,
				 gl_shader *producer, gl_shader *consumer)
{
   glsl_symbol_table parameters;
   /* FINISHME: Figure these out dynamically. */
   const char *const producer_stage = "vertex";
   const char *const consumer_stage = "fragment";

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

      /* FINISHME: For geometry shaders, this should also look for inout
       * FINISHME: variables.
       */
      if ((var == NULL) || (var->mode != ir_var_out))
	 continue;

      parameters.add_variable(var->name, 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();

      /* FINISHME: For geometry shaders, this should also look for inout
       * FINISHME: variables.
       */
      if ((input == NULL) || (input->mode != ir_var_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) {
	    linker_error_printf(prog,
				"%s shader output `%s' delcared 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 false;
	 }

	 /* Check that all of the qualifiers match between stages.
	  */
	 if (input->centroid != output->centroid) {
	    linker_error_printf(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 false;
	 }

	 if (input->invariant != output->invariant) {
	    linker_error_printf(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 false;
	 }

	 if (input->interpolation != output->interpolation) {
	    linker_error_printf(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 false;
	 }
      }
   }

   return true;
}


/**
 * Populates a shaders symbol table with all global declarations
 */
static void
populate_symbol_table(gl_shader *sh)
{
   sh->symbols = new(sh) glsl_symbol_table;

   foreach_list(node, sh->ir) {
      ir_instruction *const inst = (ir_instruction *) node;
      ir_variable *var;
      ir_function *func;

      if ((func = inst->as_function()) != NULL) {
	 sh->symbols->add_function(func->name, func);
      } else if ((var = inst->as_variable()) != NULL) {
	 sh->symbols->add_variable(var->name, var);
      }
   }
}


/**
 * Remap variables referenced in an instruction tree
 *
 * This is used when instruction trees are cloned from one shader and placed in
 * another.  These trees will contain references to \c ir_variable nodes that
 * do not exist in the target shader.  This function finds these \c ir_variable
 * references and replaces the references with matching variables in the target
 * shader.
 *
 * If there is no matching variable in the target shader, a clone of the
 * \c ir_variable is made and added to the target shader.  The new variable is
 * added to \b both the instruction stream and the symbol table.
 *
 * \param inst         IR tree that is to be processed.
 * \param symbols      Symbol table containing global scope symbols in the
 *                     linked shader.
 * \param instructions Instruction stream where new variable declarations
 *                     should be added.
 */
void
remap_variables(ir_instruction *inst, glsl_symbol_table *symbols,
		exec_list *instructions, hash_table *temps)
{
   class remap_visitor : public ir_hierarchical_visitor {
   public:
      remap_visitor(glsl_symbol_table *symbols, exec_list *instructions,
		    hash_table *temps)
      {
	 this->symbols = symbols;
	 this->instructions = instructions;
	 this->temps = temps;
      }

      virtual ir_visitor_status visit(ir_dereference_variable *ir)
      {
	 if (ir->var->mode == ir_var_temporary) {
	    ir_variable *var = (ir_variable *) hash_table_find(temps, ir->var);

	    assert(var != NULL);
	    ir->var = var;
	    return visit_continue;
	 }

	 ir_variable *const existing =
	    this->symbols->get_variable(ir->var->name);
	 if (existing != NULL)
	    ir->var = existing;
	 else {
	    ir_variable *copy = ir->var->clone(NULL);

	    this->symbols->add_variable(copy->name, copy);
	    this->instructions->push_head(copy);
	    ir->var = copy;
	 }

	 return visit_continue;
      }

   private:
      glsl_symbol_table *symbols;
      exec_list *instructions;
      hash_table *temps;
   };

   remap_visitor v(symbols, instructions, temps);

   inst->accept(&v);
}


/**
 * Move non-declarations from one instruction stream to another
 *
 * The intended usage pattern of this function is to pass the pointer to the
 * head sentinal of a list (i.e., a pointer to the list cast to an \c exec_node
 * pointer) for \c last and \c false for \c make_copies on the first
 * call.  Successive calls pass the return value of the previous call for
 * \c last and \c true for \c make_copies.
 *
 * \param instructions Source instruction stream
 * \param last         Instruction after which new instructions should be
 *                     inserted in the target instruction stream
 * \param make_copies  Flag selecting whether instructions in \c instructions
 *                     should be copied (via \c ir_instruction::clone) into the
 *                     target list or moved.
 *
 * \return
 * The new "last" instruction in the target instruction stream.  This pointer
 * is suitable for use as the \c last parameter of a later call to this
 * function.
 */
exec_node *
move_non_declarations(exec_list *instructions, exec_node *last,
		      bool make_copies, gl_shader *target)
{
   hash_table *temps = NULL;

   if (make_copies)
      temps = hash_table_ctor(0, hash_table_pointer_hash,
			      hash_table_pointer_compare);

   foreach_list_safe(node, instructions) {
      ir_instruction *inst = (ir_instruction *) node;

      if (inst->as_function())
	 continue;

      ir_variable *var = inst->as_variable();
      if ((var != NULL) && (var->mode != ir_var_temporary))
	 continue;

      assert(inst->as_assignment()
	     || ((var != NULL) && (var->mode == ir_var_temporary)));

      if (make_copies) {
	 inst = inst->clone(NULL);

	 if (var != NULL)
	    hash_table_insert(temps, inst, var);
	 else
	    remap_variables(inst, target->symbols, target->ir, temps);
      } else {
	 inst->remove();
      }

      last->insert_after(inst);
      last = inst;
   }

   if (make_copies)
      hash_table_dtor(temps);

   return last;
}

/**
 * Get the function signature for main from a shader
 */
static ir_function_signature *
get_main_function_signature(gl_shader *sh)
{
   ir_function *const f = sh->symbols->get_function("main");
   if (f != NULL) {
      exec_list void_parameters;

      /* Look for the 'void main()' signature and ensure that it's defined.
       * This keeps the linker from accidentally pick a shader that just
       * contains a prototype for main.
       *
       * We don't have to check for multiple definitions of main (in multiple
       * shaders) because that would have already been caught above.
       */
      ir_function_signature *sig = f->matching_signature(&void_parameters);
      if ((sig != NULL) && sig->is_defined) {
	 return sig;
      }
   }

   return NULL;
}


/**
 * Combine a group of shaders for a single stage to generate a linked shader
 *
 * \note
 * If this function is supplied a single shader, it is cloned, and the new
 * shader is returned.
 */
static struct gl_shader *
link_intrastage_shaders(struct gl_shader_program *prog,
			struct gl_shader **shader_list,
			unsigned num_shaders)
{
   /* Check that global variables defined in multiple shaders are consistent.
    */
   if (!cross_validate_globals(prog, shader_list, num_shaders, false))
      return NULL;

   /* Check that there is only a single definition of each function signature
    * across all shaders.
    */
   for (unsigned i = 0; i < (num_shaders - 1); i++) {
      foreach_list(node, shader_list[i]->ir) {
	 ir_function *const f = ((ir_instruction *) node)->as_function();

	 if (f == NULL)
	    continue;

	 for (unsigned j = i + 1; j < num_shaders; j++) {
	    ir_function *const other =
	       shader_list[j]->symbols->get_function(f->name);

	    /* If the other shader has no function (and therefore no function
	     * signatures) with the same name, skip to the next shader.
	     */
	    if (other == NULL)
	       continue;

	    foreach_iter (exec_list_iterator, iter, *f) {
	       ir_function_signature *sig =
		  (ir_function_signature *) iter.get();

	       if (!sig->is_defined || sig->is_built_in)
		  continue;

	       ir_function_signature *other_sig =
		  other->exact_matching_signature(& sig->parameters);

	       if ((other_sig != NULL) && other_sig->is_defined
		   && !other_sig->is_built_in) {
		  linker_error_printf(prog,
				      "function `%s' is multiply defined",
				      f->name);
		  return NULL;
	       }
	    }
	 }
      }
   }

   /* Find the shader that defines main, and make a clone of it.
    *
    * Starting with the clone, search for undefined references.  If one is
    * found, find the shader that defines it.  Clone the reference and add
    * it to the shader.  Repeat until there are no undefined references or
    * until a reference cannot be resolved.
    */
   gl_shader *main = NULL;
   for (unsigned i = 0; i < num_shaders; i++) {
      if (get_main_function_signature(shader_list[i]) != NULL) {
	 main = shader_list[i];
	 break;
      }
   }

   if (main == NULL) {
      linker_error_printf(prog, "%s shader lacks `main'\n",
			  (shader_list[0]->Type == GL_VERTEX_SHADER)
			  ? "vertex" : "fragment");
      return NULL;
   }

   gl_shader *const linked = _mesa_new_shader(NULL, 0, main->Type);
   linked->ir = new(linked) exec_list;
   clone_ir_list(linked->ir, main->ir);

   populate_symbol_table(linked);

   /* The a pointer to the main function in the final linked shader (i.e., the
    * copy of the original shader that contained the main function).
    */
   ir_function_signature *const main_sig = get_main_function_signature(linked);

   /* Move any instructions other than variable declarations or function
    * declarations into main.
    */
   exec_node *insertion_point =
      move_non_declarations(linked->ir, (exec_node *) &main_sig->body, false,
			    linked);

   for (unsigned i = 0; i < num_shaders; i++) {
      if (shader_list[i] == main)
	 continue;

      insertion_point = move_non_declarations(shader_list[i]->ir,
					      insertion_point, true, linked);
   }

   /* Resolve initializers for global variables in the linked shader.
    */
   unsigned num_linking_shaders = num_shaders;
   for (unsigned i = 0; i < num_shaders; i++)
      num_linking_shaders += shader_list[i]->num_builtins_to_link;

   gl_shader **linking_shaders =
      (gl_shader **) calloc(num_linking_shaders, sizeof(gl_shader *));

   memcpy(linking_shaders, shader_list,
	  sizeof(linking_shaders[0]) * num_shaders);

   unsigned idx = num_shaders;
   for (unsigned i = 0; i < num_shaders; i++) {
      memcpy(&linking_shaders[idx], shader_list[i]->builtins_to_link,
	     sizeof(linking_shaders[0]) * shader_list[i]->num_builtins_to_link);
      idx += shader_list[i]->num_builtins_to_link;
   }

   assert(idx == num_linking_shaders);

   link_function_calls(prog, linked, linking_shaders, num_linking_shaders);

   free(linking_shaders);

   return linked;
}


struct uniform_node {
   exec_node link;
   struct gl_uniform *u;
   unsigned slots;
};

void
assign_uniform_locations(struct gl_shader_program *prog)
{
   /* */
   exec_list uniforms;
   unsigned total_uniforms = 0;
   hash_table *ht = hash_table_ctor(32, hash_table_string_hash,
				    hash_table_string_compare);

   for (unsigned i = 0; i < prog->_NumLinkedShaders; i++) {
      unsigned next_position = 0;

      foreach_list(node, prog->_LinkedShaders[i]->ir) {
	 ir_variable *const var = ((ir_instruction *) node)->as_variable();

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

	 const unsigned vec4_slots = (var->component_slots() + 3) / 4;
	 assert(vec4_slots != 0);

	 uniform_node *n = (uniform_node *) hash_table_find(ht, var->name);
	 if (n == NULL) {
	    n = (uniform_node *) calloc(1, sizeof(struct uniform_node));
	    n->u = (gl_uniform *) calloc(vec4_slots, sizeof(struct gl_uniform));
	    n->slots = vec4_slots;

	    n->u[0].Name = strdup(var->name);
	    for (unsigned j = 1; j < vec4_slots; j++)
	       n->u[j].Name = n->u[0].Name;

	    hash_table_insert(ht, n, n->u[0].Name);
	    uniforms.push_tail(& n->link);
	    total_uniforms += vec4_slots;
	 }

	 if (var->constant_value != NULL)
	    for (unsigned j = 0; j < vec4_slots; j++)
	       n->u[j].Initialized = true;

	 var->location = next_position;

	 for (unsigned j = 0; j < vec4_slots; j++) {
	    switch (prog->_LinkedShaders[i]->Type) {
	    case GL_VERTEX_SHADER:
	       n->u[j].VertPos = next_position;
	       break;
	    case GL_FRAGMENT_SHADER:
	       n->u[j].FragPos = next_position;
	       break;
	    case GL_GEOMETRY_SHADER:
	       /* FINISHME: Support geometry shaders. */
	       assert(prog->_LinkedShaders[i]->Type != GL_GEOMETRY_SHADER);
	       break;
	    }

	    next_position++;
	 }
      }
   }

   gl_uniform_list *ul = (gl_uniform_list *)
      calloc(1, sizeof(gl_uniform_list));

   ul->Size = total_uniforms;
   ul->NumUniforms = total_uniforms;
   ul->Uniforms = (gl_uniform *) calloc(total_uniforms, sizeof(gl_uniform));

   unsigned idx = 0;
   uniform_node *next;
   for (uniform_node *node = (uniform_node *) uniforms.head
	   ; node->link.next != NULL
	   ; node = next) {
      next = (uniform_node *) node->link.next;

      node->link.remove();
      memcpy(&ul->Uniforms[idx], node->u, sizeof(gl_uniform) * node->slots);
      idx += node->slots;

      free(node->u);
      free(node);
   }

   hash_table_dtor(ht);

   prog->Uniforms = ul;
}


/**
 * Find a contiguous set of available bits in a bitmask
 *
 * \param used_mask     Bits representing used (1) and unused (0) locations
 * \param needed_count  Number of contiguous bits needed.
 *
 * \return
 * Base location of the available bits on success or -1 on failure.
 */
int
find_available_slots(unsigned used_mask, unsigned needed_count)
{
   unsigned needed_mask = (1 << needed_count) - 1;
   const int max_bit_to_test = (8 * sizeof(used_mask)) - needed_count;

   /* The comparison to 32 is redundant, but without it GCC emits "warning:
    * cannot optimize possibly infinite loops" for the loop below.
    */
   if ((needed_count == 0) || (max_bit_to_test < 0) || (max_bit_to_test > 32))
      return -1;

   for (int i = 0; i <= max_bit_to_test; i++) {
      if ((needed_mask & ~used_mask) == needed_mask)
	 return i;

      needed_mask <<= 1;
   }

   return -1;
}


bool
assign_attribute_locations(gl_shader_program *prog, unsigned max_attribute_index)
{
   /* Mark invalid attribute locations as being used.
    */
   unsigned used_locations = (max_attribute_index >= 32)
      ? ~0 : ~((1 << max_attribute_index) - 1);

   gl_shader *const sh = prog->_LinkedShaders[0];
   assert(sh->Type == GL_VERTEX_SHADER);

   /* Operate in a total of four passes.
    *
    * 1. Invalidate the location assignments for all vertex shader inputs.
    *
    * 2. Assign locations for inputs that have user-defined (via
    *    glBindVertexAttribLocation) locatoins.
    *
    * 3. Sort the attributes without assigned locations by number of slots
    *    required in decreasing order.  Fragmentation caused by attribute
    *    locations assigned by the application may prevent large attributes
    *    from having enough contiguous space.
    *
    * 4. Assign locations to any inputs without assigned locations.
    */

   invalidate_variable_locations(sh, ir_var_in, VERT_ATTRIB_GENERIC0);

   if (prog->Attributes != NULL) {
      for (unsigned i = 0; i < prog->Attributes->NumParameters; i++) {
	 ir_variable *const var =
	    sh->symbols->get_variable(prog->Attributes->Parameters[i].Name);

	 /* Note: attributes that occupy multiple slots, such as arrays or
	  * matrices, may appear in the attrib array multiple times.
	  */
	 if ((var == NULL) || (var->location != -1))
	    continue;

	 /* From page 61 of the OpenGL 4.0 spec:
	  *
	  *     "LinkProgram will fail if the attribute bindings assigned by
	  *     BindAttribLocation do not leave not enough space to assign a
	  *     location for an active matrix attribute or an active attribute
	  *     array, both of which require multiple contiguous generic
	  *     attributes."
	  *
	  * Previous versions of the spec contain similar language but omit the
	  * bit about attribute arrays.
	  *
	  * Page 61 of the OpenGL 4.0 spec also says:
	  *
	  *     "It is possible for an application to bind more than one
	  *     attribute name to the same location. This is referred to as
	  *     aliasing. This will only work if only one of the aliased
	  *     attributes is active in the executable program, or if no path
	  *     through the shader consumes more than one attribute of a set
	  *     of attributes aliased to the same location. A link error can
	  *     occur if the linker determines that every path through the
	  *     shader consumes multiple aliased attributes, but
	  *     implementations are not required to generate an error in this
	  *     case."
	  *
	  * These two paragraphs are either somewhat contradictory, or I don't
	  * fully understand one or both of them.
	  */
	 /* FINISHME: The code as currently written does not support attribute
	  * FINISHME: location aliasing (see comment above).
	  */
	 const int attr = prog->Attributes->Parameters[i].StateIndexes[0];
	 const unsigned slots = count_attribute_slots(var->type);

	 /* Mask representing the contiguous slots that will be used by this
	  * attribute.
	  */
	 const unsigned use_mask = (1 << slots) - 1;

	 /* Generate a link error if the set of bits requested for this
	  * attribute overlaps any previously allocated bits.
	  */
	 if ((~(use_mask << attr) & used_locations) != used_locations) {
	    linker_error_printf(prog,
				"insufficient contiguous attribute locations "
				"available for vertex shader input `%s'",
				var->name);
	    return false;
	 }

	 var->location = VERT_ATTRIB_GENERIC0 + attr;
	 used_locations |= (use_mask << attr);
      }
   }

   /* Temporary storage for the set of attributes that need locations assigned.
    */
   struct temp_attr {
      unsigned slots;
      ir_variable *var;

      /* Used below in the call to qsort. */
      static int compare(const void *a, const void *b)
      {
	 const temp_attr *const l = (const temp_attr *) a;
	 const temp_attr *const r = (const temp_attr *) b;

	 /* Reversed because we want a descending order sort below. */
	 return r->slots - l->slots;
      }
   } to_assign[16];

   unsigned num_attr = 0;

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

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

      /* The location was explicitly assigned, nothing to do here.
       */
      if (var->location != -1)
	 continue;

      to_assign[num_attr].slots = count_attribute_slots(var->type);
      to_assign[num_attr].var = var;
      num_attr++;
   }

   /* If all of the attributes were assigned locations by the application (or
    * are built-in attributes with fixed locations), return early.  This should
    * be the common case.
    */
   if (num_attr == 0)
      return true;

   qsort(to_assign, num_attr, sizeof(to_assign[0]), temp_attr::compare);

   /* VERT_ATTRIB_GENERIC0 is a psdueo-alias for VERT_ATTRIB_POS.  It can only
    * be explicitly assigned by via glBindAttribLocation.  Mark it as reserved
    * to prevent it from being automatically allocated below.
    */
   used_locations |= (1 << 0);

   for (unsigned i = 0; i < num_attr; i++) {
      /* Mask representing the contiguous slots that will be used by this
       * attribute.
       */
      const unsigned use_mask = (1 << to_assign[i].slots) - 1;

      int location = find_available_slots(used_locations, to_assign[i].slots);

      if (location < 0) {
	 linker_error_printf(prog,
			     "insufficient contiguous attribute locations "
			     "available for vertex shader input `%s'",
			     to_assign[i].var->name);
	 return false;
      }

      to_assign[i].var->location = VERT_ATTRIB_GENERIC0 + location;
      used_locations |= (use_mask << location);
   }

   return true;
}


void
assign_varying_locations(gl_shader *producer, gl_shader *consumer)
{
   /* FINISHME: Set dynamically when geometry shader support is added. */
   unsigned output_index = VERT_RESULT_VAR0;
   unsigned input_index = FRAG_ATTRIB_VAR0;

   /* 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.
    */

   invalidate_variable_locations(producer, ir_var_out, VERT_RESULT_VAR0);
   invalidate_variable_locations(consumer, ir_var_in, FRAG_ATTRIB_VAR0);

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

      if ((output_var == NULL) || (output_var->mode != ir_var_out)
	  || (output_var->location != -1))
	 continue;

      ir_variable *const input_var =
	 consumer->symbols->get_variable(output_var->name);

      if ((input_var == NULL) || (input_var->mode != ir_var_in))
	 continue;

      assert(input_var->location == -1);

      /* FINISHME: Location assignment will need some changes when arrays,
       * FINISHME: matrices, and structures are allowed as shader inputs /
       * FINISHME: outputs.
       */
      output_var->location = output_index;
      input_var->location = input_index;

      output_index++;
      input_index++;
   }

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

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

      /* An 'out' variable is only really a shader output if its value is read
       * by the following stage.
       */
      if (var->location == -1) {
	 var->shader_out = false;
	 var->mode = ir_var_auto;
      }
   }

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

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

      /* An 'in' variable is only really a shader input if its value is written
       * by the previous stage.
       */
      var->shader_in = (var->location != -1);
      var->mode = ir_var_auto;
   }
}


void
link_shaders(struct gl_shader_program *prog)
{
   prog->LinkStatus = false;
   prog->Validated = false;
   prog->_Used = false;

   if (prog->InfoLog != NULL)
      talloc_free(prog->InfoLog);

   prog->InfoLog = talloc_strdup(NULL, "");

   /* Separate the shaders into groups based on their type.
    */
   struct gl_shader **vert_shader_list;
   unsigned num_vert_shaders = 0;
   struct gl_shader **frag_shader_list;
   unsigned num_frag_shaders = 0;

   vert_shader_list = (struct gl_shader **)
      calloc(2 * prog->NumShaders, sizeof(struct gl_shader *));
   frag_shader_list =  &vert_shader_list[prog->NumShaders];

   unsigned min_version = UINT_MAX;
   unsigned max_version = 0;
   for (unsigned i = 0; i < prog->NumShaders; i++) {
      min_version = MIN2(min_version, prog->Shaders[i]->Version);
      max_version = MAX2(max_version, prog->Shaders[i]->Version);

      switch (prog->Shaders[i]->Type) {
      case GL_VERTEX_SHADER:
	 vert_shader_list[num_vert_shaders] = prog->Shaders[i];
	 num_vert_shaders++;
	 break;
      case GL_FRAGMENT_SHADER:
	 frag_shader_list[num_frag_shaders] = prog->Shaders[i];
	 num_frag_shaders++;
	 break;
      case GL_GEOMETRY_SHADER:
	 /* FINISHME: Support geometry shaders. */
	 assert(prog->Shaders[i]->Type != GL_GEOMETRY_SHADER);
	 break;
      }
   }

   /* Previous to GLSL version 1.30, different compilation units could mix and
    * match shading language versions.  With GLSL 1.30 and later, the versions
    * of all shaders must match.
    */
   assert(min_version >= 110);
   assert(max_version <= 130);
   if ((max_version >= 130) && (min_version != max_version)) {
      linker_error_printf(prog, "all shaders must use same shading "
			  "language version\n");
      goto done;
   }

   prog->Version = max_version;

   /* Link all shaders for a particular stage and validate the result.
    */
   prog->_NumLinkedShaders = 0;
   if (num_vert_shaders > 0) {
      gl_shader *const sh =
	 link_intrastage_shaders(prog, vert_shader_list, num_vert_shaders);

      if (sh == NULL)
	 goto done;

      if (!validate_vertex_shader_executable(prog, sh))
	  goto done;

      prog->_LinkedShaders[prog->_NumLinkedShaders] = sh;
      prog->_NumLinkedShaders++;
   }

   if (num_frag_shaders > 0) {
      gl_shader *const sh =
	 link_intrastage_shaders(prog, frag_shader_list, num_frag_shaders);

      if (sh == NULL)
	 goto done;

      if (!validate_fragment_shader_executable(prog, sh))
	  goto done;

      prog->_LinkedShaders[prog->_NumLinkedShaders] = sh;
      prog->_NumLinkedShaders++;
   }

   /* Here begins the inter-stage linking phase.  Some initial validation is
    * performed, then locations are assigned for uniforms, attributes, and
    * varyings.
    */
   if (cross_validate_uniforms(prog)) {
      /* Validate the inputs of each stage with the output of the preceeding
       * stage.
       */
      for (unsigned i = 1; i < prog->_NumLinkedShaders; i++) {
	 if (!cross_validate_outputs_to_inputs(prog,
					       prog->_LinkedShaders[i - 1],
					       prog->_LinkedShaders[i]))
	    goto done;
      }

      prog->LinkStatus = true;
   }

   /* FINISHME: Perform whole-program optimization here. */
   for (unsigned i = 0; i < prog->_NumLinkedShaders; i++) {
      /* Optimization passes */
      bool progress;
      exec_list *ir = prog->_LinkedShaders[i]->ir;

      /* Lowering */
      do_mat_op_to_vec(ir);
      do_mod_to_fract(ir);
      do_div_to_mul_rcp(ir);

      do {
	 progress = false;

	 progress = do_function_inlining(ir) || progress;
	 progress = do_if_simplification(ir) || progress;
	 progress = do_copy_propagation(ir) || progress;
	 progress = do_dead_code_local(ir) || progress;
	 progress = do_dead_code(ir) || progress;
	 progress = do_constant_variable_unlinked(ir) || progress;
	 progress = do_constant_folding(ir) || progress;
	 progress = do_if_return(ir) || progress;
#if 0
	 if (ctx->Shader.EmitNoIfs)
	    progress = do_if_to_cond_assign(ir) || progress;
#endif

	 progress = do_vec_index_to_swizzle(ir) || progress;
	 /* Do this one after the previous to let the easier pass handle
	  * constant vector indexing.
	  */
	 progress = do_vec_index_to_cond_assign(ir) || progress;

	 progress = do_swizzle_swizzle(ir) || progress;
      } while (progress);
   }

   assign_uniform_locations(prog);

   if (prog->_LinkedShaders[0]->Type == GL_VERTEX_SHADER)
      /* FINISHME: The value of the max_attribute_index parameter is
       * FINISHME: implementation dependent based on the value of
       * FINISHME: GL_MAX_VERTEX_ATTRIBS.  GL_MAX_VERTEX_ATTRIBS must be
       * FINISHME: at least 16, so hardcode 16 for now.
       */
      if (!assign_attribute_locations(prog, 16))
	 goto done;

   for (unsigned i = 1; i < prog->_NumLinkedShaders; i++)
      assign_varying_locations(prog->_LinkedShaders[i - 1],
			       prog->_LinkedShaders[i]);

   /* FINISHME: Assign fragment shader output locations. */

done:
   free(vert_shader_list);
}