summaryrefslogtreecommitdiffstats
path: root/src/compiler/glsl/link_varyings.cpp
blob: 1fad31087ec29bb0a0880c6c18bf451e7d2f8026 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
/*
 * 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"


/**
 * Get the varying type stripped of the outermost array if we're processing
 * a stage whose varyings are arrays indexed by a vertex number (such as
 * geometry shader inputs).
 */
static const glsl_type *
get_varying_type(const ir_variable *var, gl_shader_stage stage)
{
   const glsl_type *type = var->type;

   if (!var->data.patch &&
       ((var->data.mode == ir_var_shader_out &&
         stage == MESA_SHADER_TESS_CTRL) ||
        (var->data.mode == ir_var_shader_in &&
         (stage == MESA_SHADER_TESS_CTRL || stage == MESA_SHADER_TESS_EVAL ||
          stage == MESA_SHADER_GEOMETRY)))) {
      assert(type->is_array());
      type = type->fields.array;
   }

   return type;
}

static void
create_xfb_varying_names(void *mem_ctx, const glsl_type *t, char **name,
                         size_t name_length, unsigned *count,
                         const char *ifc_member_name,
                         const glsl_type *ifc_member_t, char ***varying_names)
{
   if (t->is_interface()) {
      size_t new_length = name_length;

      assert(ifc_member_name && ifc_member_t);
      ralloc_asprintf_rewrite_tail(name, &new_length, ".%s", ifc_member_name);

      create_xfb_varying_names(mem_ctx, ifc_member_t, name, new_length, count,
                               NULL, NULL, varying_names);
   } else if (t->is_record()) {
      for (unsigned i = 0; i < t->length; i++) {
         const char *field = t->fields.structure[i].name;
         size_t new_length = name_length;

         ralloc_asprintf_rewrite_tail(name, &new_length, ".%s", field);

         create_xfb_varying_names(mem_ctx, t->fields.structure[i].type, name,
                                  new_length, count, NULL, NULL,
                                  varying_names);
      }
   } else if (t->without_array()->is_record() ||
              t->without_array()->is_interface() ||
              (t->is_array() && t->fields.array->is_array())) {
      for (unsigned i = 0; i < t->length; i++) {
         size_t new_length = name_length;

         /* Append the subscript to the current variable name */
         ralloc_asprintf_rewrite_tail(name, &new_length, "[%u]", i);

         create_xfb_varying_names(mem_ctx, t->fields.array, name, new_length,
                                  count, ifc_member_name, ifc_member_t,
                                  varying_names);
      }
   } else {
      (*varying_names)[(*count)++] = ralloc_strdup(mem_ctx, *name);
   }
}

bool
process_xfb_layout_qualifiers(void *mem_ctx, const gl_linked_shader *sh,
                              unsigned *num_tfeedback_decls,
                              char ***varying_names)
{
   bool has_xfb_qualifiers = false;

   /* We still need to enable transform feedback mode even if xfb_stride is
    * only applied to a global out. Also we don't bother to propagate
    * xfb_stride to interface block members so this will catch that case also.
    */
   for (unsigned j = 0; j < MAX_FEEDBACK_BUFFERS; j++) {
      if (sh->info.TransformFeedback.BufferStride[j]) {
         has_xfb_qualifiers = true;
      }
   }

   foreach_in_list(ir_instruction, node, sh->ir) {
      ir_variable *var = node->as_variable();
      if (!var || var->data.mode != ir_var_shader_out)
         continue;

      /* From the ARB_enhanced_layouts spec:
       *
       *    "Any shader making any static use (after preprocessing) of any of
       *     these *xfb_* qualifiers will cause the shader to be in a
       *     transform feedback capturing mode and hence responsible for
       *     describing the transform feedback setup.  This mode will capture
       *     any output selected by *xfb_offset*, directly or indirectly, to
       *     a transform feedback buffer."
       */
      if (var->data.explicit_xfb_buffer || var->data.explicit_xfb_stride) {
         has_xfb_qualifiers = true;
      }

      if (var->data.explicit_xfb_offset) {
         *num_tfeedback_decls += var->type->varying_count();
         has_xfb_qualifiers = true;
      }
   }

   if (*num_tfeedback_decls == 0)
      return has_xfb_qualifiers;

   unsigned i = 0;
   *varying_names = ralloc_array(mem_ctx, char *, *num_tfeedback_decls);
   foreach_in_list(ir_instruction, node, sh->ir) {
      ir_variable *var = node->as_variable();
      if (!var || var->data.mode != ir_var_shader_out)
         continue;

      if (var->data.explicit_xfb_offset) {
         char *name;
         const glsl_type *type, *member_type;

         if (var->data.from_named_ifc_block) {
            type = var->get_interface_type();
            /* Find the member type before it was altered by lowering */
            member_type =
               type->fields.structure[type->field_index(var->name)].type;
            name = ralloc_strdup(NULL, type->without_array()->name);
         } else {
            type = var->type;
            member_type = NULL;
            name = ralloc_strdup(NULL, var->name);
         }
         create_xfb_varying_names(mem_ctx, type, &name, strlen(name), &i,
                                  var->name, member_type, varying_names);
         ralloc_free(name);
      }
   }

   assert(i == *num_tfeedback_decls);
   return has_xfb_qualifiers;
}

static bool
anonymous_struct_type_matches(const glsl_type *output_type,
                              const glsl_type *to_match)
{
    while (output_type->is_array() && to_match->is_array()) {
        /* if the lengths at each level don't match fail. */
        if (output_type->length != to_match->length)
            return false;
        output_type = output_type->fields.array;
        to_match = to_match->fields.array;
    }

    if (output_type->is_array() || to_match->is_array())
        return false;
    return output_type->is_anonymous() &&
           to_match->is_anonymous() &&
           to_match->record_compare(output_type);
}

/**
 * Validate the types and qualifiers of an output from one stage against the
 * matching input to another stage.
 */
static void
cross_validate_types_and_qualifiers(struct gl_shader_program *prog,
                                    const ir_variable *input,
                                    const ir_variable *output,
                                    gl_shader_stage consumer_stage,
                                    gl_shader_stage producer_stage)
{
   /* Check that the types match between stages.
    */
   const glsl_type *type_to_match = input->type;

   /* VS -> GS, VS -> TCS, VS -> TES, TES -> GS */
   const bool extra_array_level = (producer_stage == MESA_SHADER_VERTEX &&
                                   consumer_stage != MESA_SHADER_FRAGMENT) ||
                                  consumer_stage == MESA_SHADER_GEOMETRY;
   if (extra_array_level) {
      assert(type_to_match->is_array());
      type_to_match = type_to_match->fields.array;
   }

   if (type_to_match != 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() || !is_gl_identifier(output->name)) {
         bool anon_matches = anonymous_struct_type_matches(output->type, type_to_match);

         if (!anon_matches) {
            linker_error(prog,
                         "%s shader output `%s' declared as type `%s', "
                         "but %s shader input declared as type `%s'\n",
                         _mesa_shader_stage_to_string(producer_stage),
                         output->name,
                         output->type->name,
                         _mesa_shader_stage_to_string(consumer_stage),
                         input->type->name);
            return;
         }
      }
   }

   /* Check that all of the qualifiers match between stages.
    */

   /* According to the OpenGL and OpenGLES GLSL specs, the centroid qualifier
    * should match until OpenGL 4.3 and OpenGLES 3.1. The OpenGLES 3.0
    * conformance test suite does not verify that the qualifiers must match.
    * The deqp test suite expects the opposite (OpenGLES 3.1) behavior for
    * OpenGLES 3.0 drivers, so we relax the checking in all cases.
    */
   if (false /* always skip the centroid check */ &&
       prog->Version < (prog->IsES ? 310 : 430) &&
       input->data.centroid != output->data.centroid) {
      linker_error(prog,
                   "%s shader output `%s' %s centroid qualifier, "
                   "but %s shader input %s centroid qualifier\n",
                   _mesa_shader_stage_to_string(producer_stage),
                   output->name,
                   (output->data.centroid) ? "has" : "lacks",
                   _mesa_shader_stage_to_string(consumer_stage),
                   (input->data.centroid) ? "has" : "lacks");
      return;
   }

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

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

   /* The GLSL 4.30 and GLSL ES 3.00 specifications say:
    *
    *    "As only outputs need be declared with invariant, an output from
    *     one shader stage will still match an input of a subsequent stage
    *     without the input being declared as invariant."
    *
    * while GLSL 4.20 says:
    *
    *    "For variables leaving one shader and coming into another shader,
    *     the invariant keyword has to be used in both shaders, or a link
    *     error will result."
    *
    * and GLSL ES 1.00 section 4.6.4 "Invariance and Linking" says:
    *
    *    "The invariance of varyings that are declared in both the vertex
    *     and fragment shaders must match."
    */
   if (input->data.invariant != output->data.invariant &&
       prog->Version < (prog->IsES ? 300 : 430)) {
      linker_error(prog,
                   "%s shader output `%s' %s invariant qualifier, "
                   "but %s shader input %s invariant qualifier\n",
                   _mesa_shader_stage_to_string(producer_stage),
                   output->name,
                   (output->data.invariant) ? "has" : "lacks",
                   _mesa_shader_stage_to_string(consumer_stage),
                   (input->data.invariant) ? "has" : "lacks");
      return;
   }

   /* GLSL >= 4.40 removes text requiring interpolation qualifiers
    * to match cross stage, they must only match within the same stage.
    *
    * From page 84 (page 90 of the PDF) of the GLSL 4.40 spec:
    *
    *     "It is a link-time error if, within the same stage, the interpolation
    *     qualifiers of variables of the same name do not match.
    *
    */
   if (input->data.interpolation != output->data.interpolation &&
       prog->Version < 440) {
      linker_error(prog,
                   "%s shader output `%s' specifies %s "
                   "interpolation qualifier, "
                   "but %s shader input specifies %s "
                   "interpolation qualifier\n",
                   _mesa_shader_stage_to_string(producer_stage),
                   output->name,
                   interpolation_string(output->data.interpolation),
                   _mesa_shader_stage_to_string(consumer_stage),
                   interpolation_string(input->data.interpolation));
      return;
   }
}

/**
 * Validate front and back color outputs against single color input
 */
static void
cross_validate_front_and_back_color(struct gl_shader_program *prog,
                                    const ir_variable *input,
                                    const ir_variable *front_color,
                                    const ir_variable *back_color,
                                    gl_shader_stage consumer_stage,
                                    gl_shader_stage producer_stage)
{
   if (front_color != NULL && front_color->data.assigned)
      cross_validate_types_and_qualifiers(prog, input, front_color,
                                          consumer_stage, producer_stage);

   if (back_color != NULL && back_color->data.assigned)
      cross_validate_types_and_qualifiers(prog, input, back_color,
                                          consumer_stage, producer_stage);
}

/**
 * Validate that outputs from one stage match inputs of another
 */
void
cross_validate_outputs_to_inputs(struct gl_shader_program *prog,
                                 gl_linked_shader *producer,
                                 gl_linked_shader *consumer)
{
   glsl_symbol_table parameters;
   ir_variable *explicit_locations[MAX_VARYINGS_INCL_PATCH][4] =
      { {NULL, NULL} };

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

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

      if (!var->data.explicit_location
          || var->data.location < VARYING_SLOT_VAR0)
         parameters.add_variable(var);
      else {
         /* User-defined varyings with explicit locations are handled
          * differently because they do not need to have matching names.
          */
         const glsl_type *type = get_varying_type(var, producer->Stage);
         unsigned num_elements = type->count_attribute_slots(false);
         unsigned idx = var->data.location - VARYING_SLOT_VAR0;
         unsigned slot_limit = idx + num_elements;
         unsigned last_comp;

         if (type->without_array()->is_record()) {
            /* The component qualifier can't be used on structs so just treat
             * all component slots as used.
             */
            last_comp = 4;
         } else {
            unsigned dmul = type->without_array()->is_64bit() ? 2 : 1;
            last_comp = var->data.location_frac +
               type->without_array()->vector_elements * dmul;
         }

         while (idx < slot_limit) {
            unsigned i = var->data.location_frac;
            while (i < last_comp) {
               if (explicit_locations[idx][i] != NULL) {
                  linker_error(prog,
                               "%s shader has multiple outputs explicitly "
                               "assigned to location %d and component %d\n",
                               _mesa_shader_stage_to_string(producer->Stage),
                               idx, var->data.location_frac);
                  return;
               }

               /* Make sure all component at this location have the same type.
                */
               for (unsigned j = 0; j < 4; j++) {
                  if (explicit_locations[idx][j] &&
                      (explicit_locations[idx][j]->type->without_array()
                       ->base_type != type->without_array()->base_type)) {
                     linker_error(prog,
                                  "Varyings sharing the same location must "
                                  "have the same underlying numerical type. "
                                  "Location %u component %u\n", idx,
                                  var->data.location_frac);
                     return;
                  }
               }

               explicit_locations[idx][i] = var;
               i++;

               /* We need to do some special handling for doubles as dvec3 and
                * dvec4 consume two consecutive locations. We don't need to
                * worry about components beginning at anything other than 0 as
                * the spec does not allow this for dvec3 and dvec4.
                */
               if (i == 4 && last_comp > 4) {
                  last_comp = last_comp - 4;
                  /* Bump location index and reset the component index */
                  idx++;
                  i = 0;
               }
            }
            idx++;
         }
      }
   }


   /* 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.
    *
    * Exception: if the consumer is the geometry shader, then the inputs
    * should be arrays and the type of the array element should match the type
    * of the corresponding producer output.
    */
   foreach_in_list(ir_instruction, node, consumer->ir) {
      ir_variable *const input = node->as_variable();

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

      if (strcmp(input->name, "gl_Color") == 0 && input->data.used) {
         const ir_variable *const front_color =
            parameters.get_variable("gl_FrontColor");

         const ir_variable *const back_color =
            parameters.get_variable("gl_BackColor");

         cross_validate_front_and_back_color(prog, input,
                                             front_color, back_color,
                                             consumer->Stage, producer->Stage);
      } else if (strcmp(input->name, "gl_SecondaryColor") == 0 && input->data.used) {
         const ir_variable *const front_color =
            parameters.get_variable("gl_FrontSecondaryColor");

         const ir_variable *const back_color =
            parameters.get_variable("gl_BackSecondaryColor");

         cross_validate_front_and_back_color(prog, input,
                                             front_color, back_color,
                                             consumer->Stage, producer->Stage);
      } else {
         /* The rules for connecting inputs and outputs change in the presence
          * of explicit locations.  In this case, we no longer care about the
          * names of the variables.  Instead, we care only about the
          * explicitly assigned location.
          */
         ir_variable *output = NULL;
         if (input->data.explicit_location
             && input->data.location >= VARYING_SLOT_VAR0) {

            const glsl_type *type = get_varying_type(input, consumer->Stage);
            unsigned num_elements = type->count_attribute_slots(false);
            unsigned idx = input->data.location - VARYING_SLOT_VAR0;
            unsigned slot_limit = idx + num_elements;

            while (idx < slot_limit) {
               output = explicit_locations[idx][input->data.location_frac];

               if (output == NULL ||
                   input->data.location != output->data.location) {
                  linker_error(prog,
                               "%s shader input `%s' with explicit location "
                               "has no matching output\n",
                               _mesa_shader_stage_to_string(consumer->Stage),
                               input->name);
                  break;
               }
               idx++;
            }
         } else {
            output = parameters.get_variable(input->name);
         }

         if (output != NULL) {
            /* Interface blocks have their own validation elsewhere so don't
             * try validating them here.
             */
            if (!(input->get_interface_type() &&
                  output->get_interface_type()))
               cross_validate_types_and_qualifiers(prog, input, output,
                                                   consumer->Stage,
                                                   producer->Stage);
         } else {
            /* Check for input vars with unmatched output vars in prev stage
             * taking into account that interface blocks could have a matching
             * output but with different name, so we ignore them.
             */
            assert(!input->data.assigned);
            if (input->data.used && !input->get_interface_type() &&
                !input->data.explicit_location && !prog->SeparateShader)
               linker_error(prog,
                            "%s shader input `%s' "
                            "has no matching output in the previous stage\n",
                            _mesa_shader_stage_to_string(consumer->Stage),
                            input->name);
         }
      }
   }
}

/**
 * Demote shader inputs and outputs that are not used in other stages, and
 * remove them via dead code elimination.
 */
void
remove_unused_shader_inputs_and_outputs(bool is_separate_shader_object,
                                        gl_linked_shader *sh,
                                        enum ir_variable_mode mode)
{
   if (is_separate_shader_object)
      return;

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

      if (var == NULL || var->data.mode != int(mode))
         continue;

      /* A shader 'in' or 'out' variable is only really an input or output if
       * its value is used by other shader stages. This will cause the
       * variable to have a location assigned.
       */
      if (var->data.is_unmatched_generic_inout && !var->data.is_xfb_only) {
         assert(var->data.mode != ir_var_temporary);
         var->data.mode = ir_var_auto;
      }
   }

   /* Eliminate code that is now dead due to unused inputs/outputs being
    * demoted.
    */
   while (do_dead_code(sh->ir, false))
      ;

}

/**
 * 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, 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->lowered_builtin_array_variable = none;
   this->skip_components = 0;
   this->next_buffer_separator = false;
   this->matched_candidate = NULL;
   this->stream_id = 0;
   this->buffer = 0;
   this->offset = 0;

   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 (this->var_name == NULL) {
      _mesa_error_no_memory(__func__);
      return;
   }

   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->Const.ShaderCompilerOptions[MESA_SHADER_VERTEX].LowerCombinedClipCullDistance &&
       strcmp(this->var_name, "gl_ClipDistance") == 0) {
      this->lowered_builtin_array_variable = clip_distance;
   }
   if (ctx->Const.ShaderCompilerOptions[MESA_SHADER_VERTEX].LowerCombinedClipCullDistance &&
       strcmp(this->var_name, "gl_CullDistance") == 0) {
      this->lowered_builtin_array_variable = cull_distance;
   }

   if (ctx->Const.LowerTessLevel &&
       (strcmp(this->var_name, "gl_TessLevelOuter") == 0))
      this->lowered_builtin_array_variable = tess_level_outer;
   if (ctx->Const.LowerTessLevel &&
       (strcmp(this->var_name, "gl_TessLevelInner") == 0))
      this->lowered_builtin_array_variable = tess_level_inner;
}


/**
 * 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 and stream ID 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->data.location * 4
      + this->matched_candidate->toplevel_var->data.location_frac
      + this->matched_candidate->offset;
   const unsigned dmul =
      this->matched_candidate->type->without_array()->is_64bit() ? 2 : 1;

   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;
      switch (this->lowered_builtin_array_variable) {
      case clip_distance:
         actual_array_size = prog->LastClipDistanceArraySize;
         break;
      case cull_distance:
         actual_array_size = prog->LastCullDistanceArraySize;
         break;
      case tess_level_outer:
         actual_array_size = 4;
         break;
      case tess_level_inner:
         actual_array_size = 2;
         break;
      case none:
      default:
         actual_array_size = this->matched_candidate->type->array_size();
         break;
      }

      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->lowered_builtin_array_variable ?
            1 : vector_elements * matrix_cols * dmul;
         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->lowered_builtin_array_variable)
         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;
   }

   /* Only transform feedback varyings can be assigned to non-zero streams,
    * so assign the stream id here.
    */
   this->stream_id = this->matched_candidate->toplevel_var->data.stream;

   unsigned array_offset = this->array_subscript * 4 * dmul;
   unsigned struct_offset = this->matched_candidate->offset * 4 * dmul;
   this->buffer = this->matched_candidate->toplevel_var->data.xfb_buffer;
   this->offset = this->matched_candidate->toplevel_var->data.offset +
      array_offset + struct_offset;

   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, unsigned buffer_index,
                      const unsigned max_outputs, bool *explicit_stride,
                      bool has_xfb_qualifiers) const
{
   unsigned xfb_offset = 0;
   unsigned size = this->size;
   /* Handle gl_SkipComponents. */
   if (this->skip_components) {
      info->Buffers[buffer].Stride += this->skip_components;
      size = this->skip_components;
      goto store_varying;
   }

   if (this->next_buffer_separator) {
      size = 0;
      goto store_varying;
   }

   if (has_xfb_qualifiers) {
      xfb_offset = this->offset / 4;
   } else {
      xfb_offset = info->Buffers[buffer].Stride;
   }
   info->Varyings[info->NumVarying].Offset = xfb_offset * 4;

   {
      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 == 0 && max_outputs == 0) ||
                info->NumOutputs < max_outputs);

         /* From the ARB_enhanced_layouts spec:
          *
          *    "If such a block member or variable is not written during a shader
          *    invocation, the buffer contents at the assigned offset will be
          *    undefined.  Even if there are no static writes to a variable or
          *    member that is assigned a transform feedback offset, the space is
          *    still allocated in the buffer and still affects the stride."
          */
         if (this->is_varying_written()) {
            info->Outputs[info->NumOutputs].ComponentOffset = location_frac;
            info->Outputs[info->NumOutputs].OutputRegister = location;
            info->Outputs[info->NumOutputs].NumComponents = output_size;
            info->Outputs[info->NumOutputs].StreamId = stream_id;
            info->Outputs[info->NumOutputs].OutputBuffer = buffer;
            info->Outputs[info->NumOutputs].DstOffset = xfb_offset;
            ++info->NumOutputs;
         }
         info->Buffers[buffer].Stream = this->stream_id;
         xfb_offset += output_size;

         num_components -= output_size;
         location++;
         location_frac = 0;
      }
   }

   if (explicit_stride && explicit_stride[buffer]) {
      if (this->is_64bit() && info->Buffers[buffer].Stride % 2) {
         linker_error(prog, "invalid qualifier xfb_stride=%d must be a "
                      "multiple of 8 as its applied to a type that is or "
                      "contains a double.",
                      info->Buffers[buffer].Stride * 4);
         return false;
      }

      if ((this->offset / 4) / info->Buffers[buffer].Stride !=
          (xfb_offset - 1) / info->Buffers[buffer].Stride) {
         linker_error(prog, "xfb_offset (%d) overflows xfb_stride (%d) for "
                      "buffer (%d)", xfb_offset * 4,
                      info->Buffers[buffer].Stride * 4, buffer);
         return false;
      }
   } else {
      info->Buffers[buffer].Stride = xfb_offset;
   }

   /* 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.
    *
    * From GL_ARB_enhanced_layouts:
    *
    *   "The resulting stride (implicit or explicit) must be less than or
    *   equal to the implementation-dependent constant
    *   gl_MaxTransformFeedbackInterleavedComponents."
    */
   if ((prog->TransformFeedback.BufferMode == GL_INTERLEAVED_ATTRIBS ||
        has_xfb_qualifiers) &&
       info->Buffers[buffer].Stride >
       ctx->Const.MaxTransformFeedbackInterleavedComponents) {
      linker_error(prog, "The MAX_TRANSFORM_FEEDBACK_INTERLEAVED_COMPONENTS "
                   "limit has been exceeded.");
      return false;
   }

 store_varying:
   info->Varyings[info->NumVarying].Name = ralloc_strdup(prog,
                                                         this->orig_name);
   info->Varyings[info->NumVarying].Type = this->type;
   info->Varyings[info->NumVarying].Size = size;
   info->Varyings[info->NumVarying].BufferIndex = buffer_index;
   info->NumVarying++;
   info->Buffers[buffer].NumVaryings++;

   return true;
}


const tfeedback_candidate *
tfeedback_decl::find_candidate(gl_shader_program *prog,
                               hash_table *tfeedback_candidates)
{
   const char *name = this->var_name;
   switch (this->lowered_builtin_array_variable) {
   case none:
      name = this->var_name;
      break;
   case clip_distance:
      name = "gl_ClipDistanceMESA";
      break;
   case cull_distance:
      name = "gl_CullDistanceMESA";
      break;
   case tess_level_outer:
      name = "gl_TessLevelOuterMESA";
      break;
   case tess_level_inner:
      name = "gl_TessLevelInnerMESA";
      break;
   }
   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, 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;
}


static int
cmp_xfb_offset(const void * x_generic, const void * y_generic)
{
   tfeedback_decl *x = (tfeedback_decl *) x_generic;
   tfeedback_decl *y = (tfeedback_decl *) y_generic;

   if (x->get_buffer() != y->get_buffer())
      return x->get_buffer() - y->get_buffer();
   return x->get_offset() - y->get_offset();
}

/**
 * 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 has_xfb_qualifiers)
{
   /* Make sure MaxTransformFeedbackBuffers is less than 32 so the bitmask for
    * tracking the number of buffers doesn't overflow.
    */
   assert(ctx->Const.MaxTransformFeedbackBuffers < 32);

   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));

   /* The xfb_offset qualifier does not have to be used in increasing order
    * however some drivers expect to receive the list of transform feedback
    * declarations in order so sort it now for convenience.
    */
   if (has_xfb_qualifiers)
      qsort(tfeedback_decls, num_tfeedback_decls, sizeof(*tfeedback_decls),
            cmp_xfb_offset);

   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) {
      if (tfeedback_decls[i].is_varying_written())
         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;
   unsigned buffers = 0;

   if (!has_xfb_qualifiers && 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_buffers, num_outputs,
                                       NULL, has_xfb_qualifiers))
            return false;

         buffers |= 1 << num_buffers;
         num_buffers++;
      }
   }
   else {
      /* GL_INVERLEAVED_ATTRIBS */
      int buffer_stream_id = -1;
      unsigned buffer =
         num_tfeedback_decls ? tfeedback_decls[0].get_buffer() : 0;
      bool explicit_stride[MAX_FEEDBACK_BUFFERS] = { false };

      /* Apply any xfb_stride global qualifiers */
      if (has_xfb_qualifiers) {
         for (unsigned j = 0; j < MAX_FEEDBACK_BUFFERS; j++) {
            if (prog->TransformFeedback.BufferStride[j]) {
               buffers |= 1 << j;
               explicit_stride[j] = true;
               prog->LinkedTransformFeedback.Buffers[j].Stride =
                  prog->TransformFeedback.BufferStride[j] / 4;
            }
         }
      }

      for (unsigned i = 0; i < num_tfeedback_decls; ++i) {
         if (has_xfb_qualifiers &&
             buffer != tfeedback_decls[i].get_buffer()) {
            /* we have moved to the next buffer so reset stream id */
            buffer_stream_id = -1;
            num_buffers++;
         }

         if (tfeedback_decls[i].is_next_buffer_separator()) {
            if (!tfeedback_decls[i].store(ctx, prog,
                                          &prog->LinkedTransformFeedback,
                                          buffer, num_buffers, num_outputs,
                                          explicit_stride, has_xfb_qualifiers))
               return false;
            num_buffers++;
            buffer_stream_id = -1;
            continue;
         } else if (tfeedback_decls[i].is_varying()) {
            if (buffer_stream_id == -1)  {
               /* First varying writing to this buffer: remember its stream */
               buffer_stream_id = (int) tfeedback_decls[i].get_stream_id();
            } else if (buffer_stream_id !=
                       (int) tfeedback_decls[i].get_stream_id()) {
               /* Varying writes to the same buffer from a different stream */
               linker_error(prog,
                            "Transform feedback can't capture varyings belonging "
                            "to different vertex streams in a single buffer. "
                            "Varying %s writes to buffer from stream %u, other "
                            "varyings in the same buffer write from stream %u.",
                            tfeedback_decls[i].name(),
                            tfeedback_decls[i].get_stream_id(),
                            buffer_stream_id);
               return false;
            }
         }

         if (has_xfb_qualifiers) {
            buffer = tfeedback_decls[i].get_buffer();
         } else {
            buffer = num_buffers;
         }
         buffers |= 1 << buffer;

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

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

   prog->LinkedTransformFeedback.ActiveBuffers = buffers;
   return true;
}

namespace {

/**
 * 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 xfb_enabled,
                   gl_shader_stage producer_stage,
                   gl_shader_stage consumer_stage);
   ~varying_matches();
   void record(ir_variable *producer_var, ir_variable *consumer_var);
   unsigned assign_locations(struct gl_shader_program *prog,
                             uint64_t reserved_slots);
   void store_locations() const;

private:
   bool is_varying_packing_safe(const glsl_type *type,
                                const ir_variable *var);

   /**
    * 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.
    *
    * Packing may also be disabled because our current packing method is not
    * safe in SSO or versions of OpenGL where interpolation qualifiers are not
    * guaranteed to match across stages.
    */
   const bool disable_varying_packing;

   /**
    * If true, this driver has transform feedback enabled. The transform
    * feedback code requires at least some packing be done even when varying
    * packing is disabled, fortunately where transform feedback requires
    * packing it's safe to override the disabled setting. See
    * is_varying_packing_safe().
    */
   const bool xfb_enabled;

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

   gl_shader_stage producer_stage;
   gl_shader_stage consumer_stage;
};

} /* anonymous namespace */

varying_matches::varying_matches(bool disable_varying_packing,
                                 bool xfb_enabled,
                                 gl_shader_stage producer_stage,
                                 gl_shader_stage consumer_stage)
   : disable_varying_packing(disable_varying_packing),
     xfb_enabled(xfb_enabled),
     producer_stage(producer_stage),
     consumer_stage(consumer_stage)
{
   /* 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);
}


/**
 * Packing is always safe on individual arrays, structures, and matrices. It
 * is also safe if the varying is only used for transform feedback.
 */
bool
varying_matches::is_varying_packing_safe(const glsl_type *type,
                                         const ir_variable *var)
{
   if (consumer_stage == MESA_SHADER_TESS_EVAL ||
       consumer_stage == MESA_SHADER_TESS_CTRL ||
       producer_stage == MESA_SHADER_TESS_CTRL)
      return false;

   return xfb_enabled && (type->is_array() || type->is_record() ||
                          type->is_matrix() || var->data.is_xfb_only);
}


/**
 * 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)
{
   assert(producer_var != NULL || consumer_var != NULL);

   if ((producer_var && (!producer_var->data.is_unmatched_generic_inout ||
       producer_var->data.explicit_location)) ||
       (consumer_var && (!consumer_var->data.is_unmatched_generic_inout ||
       consumer_var->data.explicit_location))) {
      /* 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;
   }

   bool needs_flat_qualifier = consumer_var == NULL &&
      (producer_var->type->contains_integer() ||
       producer_var->type->contains_double());

   if (!disable_varying_packing &&
       (needs_flat_qualifier ||
        (consumer_stage != -1 && consumer_stage != MESA_SHADER_FRAGMENT))) {
      /* 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
       * or a double.
       * If the consumer stage is unknown, don't modify the interpolation
       * type as it could affect rendering later with separate shaders.
       *
       * 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.
       */
      if (producer_var) {
         producer_var->data.centroid = false;
         producer_var->data.sample = false;
         producer_var->data.interpolation = INTERP_MODE_FLAT;
      }

      if (consumer_var) {
         consumer_var->data.centroid = false;
         consumer_var->data.sample = false;
         consumer_var->data.interpolation = INTERP_MODE_FLAT;
      }
   }

   if (this->num_matches == this->matches_capacity) {
      this->matches_capacity *= 2;
      this->matches = (match *)
         realloc(this->matches,
                 sizeof(*this->matches) * this->matches_capacity);
   }

   /* We must use the consumer to compute the packing class because in GL4.4+
    * there is no guarantee interpolation qualifiers will match across stages.
    *
    * From Section 4.5 (Interpolation Qualifiers) of the GLSL 4.30 spec:
    *
    *    "The type and presence of interpolation qualifiers of variables with
    *    the same name declared in all linked shaders for the same cross-stage
    *    interface must match, otherwise the link command will fail.
    *
    *    When comparing an output from one stage to an input of a subsequent
    *    stage, the input and output don't match if their interpolation
    *    qualifiers (or lack thereof) are not the same."
    *
    * This text was also in at least revison 7 of the 4.40 spec but is no
    * longer in revision 9 and not in the 4.50 spec.
    */
   const ir_variable *const var = (consumer_var != NULL)
      ? consumer_var : producer_var;
   const gl_shader_stage stage = (consumer_var != NULL)
      ? consumer_stage : producer_stage;
   const glsl_type *type = get_varying_type(var, stage);

   this->matches[this->num_matches].packing_class
      = this->compute_packing_class(var);
   this->matches[this->num_matches].packing_order
      = this->compute_packing_order(var);
   if (this->disable_varying_packing && !is_varying_packing_safe(type, var)) {
      unsigned slots = type->count_attribute_slots(false);
      this->matches[this->num_matches].num_components = slots * 4;
   } else {
      this->matches[this->num_matches].num_components
         = type->component_slots();
   }
   this->matches[this->num_matches].producer_var = producer_var;
   this->matches[this->num_matches].consumer_var = consumer_var;
   this->num_matches++;
   if (producer_var)
      producer_var->data.is_unmatched_generic_inout = 0;
   if (consumer_var)
      consumer_var->data.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(struct gl_shader_program *prog,
                                  uint64_t reserved_slots)
{
   /* If packing has been disabled then we cannot safely sort the varyings by
    * class as it may mean we are using a version of OpenGL where
    * interpolation qualifiers are not guaranteed to be matching across
    * shaders, sorting in this case could result in mismatching shader
    * interfaces.
    * When packing is disabled the sort orders varyings used by transform
    * feedback first, but also depends on *undefined behaviour* of qsort to
    * reverse the order of the varyings. See: xfb_comparator().
    */
   if (!this->disable_varying_packing) {
      /* 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);
   } else {
      /* Only sort varyings that are only used by transform feedback. */
      qsort(this->matches, this->num_matches, sizeof(*this->matches),
            &varying_matches::xfb_comparator);
   }

   unsigned generic_location = 0;
   unsigned generic_patch_location = MAX_VARYING*4;
   bool previous_var_xfb_only = false;

   for (unsigned i = 0; i < this->num_matches; i++) {
      unsigned *location = &generic_location;

      const ir_variable *var;
      const glsl_type *type;
      bool is_vertex_input = false;
      if (matches[i].consumer_var) {
         var = matches[i].consumer_var;
         type = get_varying_type(var, consumer_stage);
         if (consumer_stage == MESA_SHADER_VERTEX)
            is_vertex_input = true;
      } else {
         var = matches[i].producer_var;
         type = get_varying_type(var, producer_stage);
      }

      if (var->data.patch)
         location = &generic_patch_location;

      /* 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.
       *
       * Also advance to the next slot if packing is disabled. This makes sure
       * we don't assign varyings the same locations which is possible
       * because we still pack individual arrays, records and matrices even
       * when packing is disabled. Note we don't advance to the next slot if
       * we can pack varyings together that are only used for transform
       * feedback.
       */
      if ((this->disable_varying_packing &&
           !(previous_var_xfb_only && var->data.is_xfb_only)) ||
          (i > 0 && this->matches[i - 1].packing_class
          != this->matches[i].packing_class )) {
         *location = ALIGN(*location, 4);
      }

      previous_var_xfb_only = var->data.is_xfb_only;

      unsigned num_elements =  type->count_attribute_slots(is_vertex_input);
      unsigned slot_end;
      if (this->disable_varying_packing &&
          !is_varying_packing_safe(type, var))
         slot_end = 4;
      else
         slot_end = type->without_array()->vector_elements;
      slot_end += *location - 1;

      /* FIXME: We could be smarter in the below code and loop back over
       * trying to fill any locations that we skipped because we couldn't pack
       * the varying between an explicit location. For now just let the user
       * hit the linking error if we run out of room and suggest they use
       * explicit locations.
       */
      for (unsigned j = 0; j < num_elements; j++) {
         while ((slot_end < MAX_VARYING * 4u) &&
                ((reserved_slots & (UINT64_C(1) << *location / 4u) ||
                 (reserved_slots & (UINT64_C(1) << slot_end / 4u))))) {

            *location = ALIGN(*location + 1, 4);
            slot_end = *location;

            /* reset the counter and try again */
            j = 0;
         }

         /* Increase the slot to make sure there is enough room for next
          * array element.
          */
         if (this->disable_varying_packing &&
             !is_varying_packing_safe(type, var))
            slot_end += 4;
         else
            slot_end += type->without_array()->vector_elements;
      }

      if (!var->data.patch && *location >= MAX_VARYING * 4u) {
         linker_error(prog, "insufficient contiguous locations available for "
                      "%s it is possible an array or struct could not be "
                      "packed between varyings with explicit locations. Try "
                      "using an explicit location for arrays and structs.",
                      var->name);
      }

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

      *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() 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;

      if (producer_var) {
         producer_var->data.location = VARYING_SLOT_VAR0 + slot;
         producer_var->data.location_frac = offset;
      }

      if (consumer_var) {
         assert(consumer_var->data.location == -1);
         consumer_var->data.location = VARYING_SLOT_VAR0 + slot;
         consumer_var->data.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(const 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->data.centroid | (var->data.sample << 1) |
                            (var->data.patch << 2);
   packing_class *= 4;
   packing_class += var->is_interpolation_flat()
      ? unsigned(INTERP_MODE_FLAT) : var->data.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(const 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;
}


/**
 * Comparison function passed to qsort() to sort varyings used only by
 * transform feedback when packing of other varyings is disabled.
 */
int
varying_matches::xfb_comparator(const void *x_generic, const void *y_generic)
{
   const match *x = (const match *) x_generic;

   if (x->producer_var != NULL && x->producer_var->data.is_xfb_only)
         return match_comparator(x_generic, y_generic);

   /* FIXME: When the comparator returns 0 it means the elements being
    * compared are equivalent. However the qsort documentation says:
    *
    *    "The order of equivalent elements is undefined."
    *
    * In practice the sort ends up reversing the order of the varyings which
    * means locations are also assigned in this reversed order and happens to
    * be what we want. This is also whats happening in
    * varying_matches::match_comparator().
    */
   return 0;
}


/**
 * 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
var_counts_against_varying_limit(gl_shader_stage stage, const ir_variable *var)
{
   /* Only fragment shaders will take a varying variable as an input */
   if (stage == MESA_SHADER_FRAGMENT &&
       var->data.mode == ir_var_shader_in) {
      switch (var->data.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)
   {
      /* All named varying interface blocks should be flattened by now */
      assert(!var->is_interface_instance());

      this->toplevel_var = var;
      this->varying_floats = 0;
      program_resource_visitor::process(var);
   }

private:
   virtual void visit_field(const glsl_type *type, const char *name,
                            bool row_major)
   {
      assert(!type->without_array()->is_record());
      assert(!type->without_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;
};


namespace linker {

void
populate_consumer_input_sets(void *mem_ctx, exec_list *ir,
                             hash_table *consumer_inputs,
                             hash_table *consumer_interface_inputs,
                             ir_variable *consumer_inputs_with_locations[VARYING_SLOT_TESS_MAX])
{
   memset(consumer_inputs_with_locations,
          0,
          sizeof(consumer_inputs_with_locations[0]) * VARYING_SLOT_TESS_MAX);

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

      if (input_var != NULL && input_var->data.mode == ir_var_shader_in) {
         /* All interface blocks should have been lowered by this point */
         assert(!input_var->type->is_interface());

         if (input_var->data.explicit_location) {
            /* assign_varying_locations only cares about finding the
             * ir_variable at the start of a contiguous location block.
             *
             *     - For !producer, consumer_inputs_with_locations isn't used.
             *
             *     - For !consumer, consumer_inputs_with_locations is empty.
             *
             * For consumer && producer, if you were trying to set some
             * ir_variable to the middle of a location block on the other side
             * of producer/consumer, cross_validate_outputs_to_inputs() should
             * be link-erroring due to either type mismatch or location
             * overlaps.  If the variables do match up, then they've got a
             * matching data.location and you only looked at
             * consumer_inputs_with_locations[var->data.location], not any
             * following entries for the array/structure.
             */
            consumer_inputs_with_locations[input_var->data.location] =
               input_var;
         } else if (input_var->get_interface_type() != NULL) {
            char *const iface_field_name =
               ralloc_asprintf(mem_ctx, "%s.%s",
                  input_var->get_interface_type()->without_array()->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));
         }
      }
   }
}

/**
 * Find a variable from the consumer that "matches" the specified variable
 *
 * This function only finds inputs with names that match.  There is no
 * validation (here) that the types, etc. are compatible.
 */
ir_variable *
get_matching_input(void *mem_ctx,
                   const ir_variable *output_var,
                   hash_table *consumer_inputs,
                   hash_table *consumer_interface_inputs,
                   ir_variable *consumer_inputs_with_locations[VARYING_SLOT_TESS_MAX])
{
   ir_variable *input_var;

   if (output_var->data.explicit_location) {
      input_var = consumer_inputs_with_locations[output_var->data.location];
   } else if (output_var->get_interface_type() != NULL) {
      char *const iface_field_name =
         ralloc_asprintf(mem_ctx, "%s.%s",
            output_var->get_interface_type()->without_array()->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);
   }

   return (input_var == NULL || input_var->data.mode != ir_var_shader_in)
      ? NULL : input_var;
}

}

static int
io_variable_cmp(const void *_a, const void *_b)
{
   const ir_variable *const a = *(const ir_variable **) _a;
   const ir_variable *const b = *(const ir_variable **) _b;

   if (a->data.explicit_location && b->data.explicit_location)
      return b->data.location - a->data.location;

   if (a->data.explicit_location && !b->data.explicit_location)
      return 1;

   if (!a->data.explicit_location && b->data.explicit_location)
      return -1;

   return -strcmp(a->name, b->name);
}

/**
 * Sort the shader IO variables into canonical order
 */
static void
canonicalize_shader_io(exec_list *ir, enum ir_variable_mode io_mode)
{
   ir_variable *var_table[MAX_PROGRAM_OUTPUTS * 4];
   unsigned num_variables = 0;

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

      if (var == NULL || var->data.mode != io_mode)
         continue;

      /* If we have already encountered more I/O variables that could
       * successfully link, bail.
       */
      if (num_variables == ARRAY_SIZE(var_table))
         return;

      var_table[num_variables++] = var;
   }

   if (num_variables == 0)
      return;

   /* Sort the list in reverse order (io_variable_cmp handles this).  Later
    * we're going to push the variables on to the IR list as a stack, so we
    * want the last variable (in canonical order) to be first in the list.
    */
   qsort(var_table, num_variables, sizeof(var_table[0]), io_variable_cmp);

   /* Remove the variable from it's current location in the IR, and put it at
    * the front.
    */
   for (unsigned i = 0; i < num_variables; i++) {
      var_table[i]->remove();
      ir->push_head(var_table[i]);
   }
}

/**
 * Generate a bitfield map of the explicit locations for shader varyings.
 *
 * Note: For Tessellation shaders we are sitting right on the limits of the
 * 64 bit map. Per-vertex and per-patch both have separate location domains
 * with a max of MAX_VARYING.
 */
uint64_t
reserved_varying_slot(struct gl_linked_shader *stage,
                      ir_variable_mode io_mode)
{
   assert(io_mode == ir_var_shader_in || io_mode == ir_var_shader_out);
   /* Avoid an overflow of the returned value */
   assert(MAX_VARYINGS_INCL_PATCH <= 64);

   uint64_t slots = 0;
   int var_slot;

   if (!stage)
      return slots;

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

      if (var == NULL || var->data.mode != io_mode ||
          !var->data.explicit_location ||
          var->data.location < VARYING_SLOT_VAR0)
         continue;

      var_slot = var->data.location - VARYING_SLOT_VAR0;

      unsigned num_elements = get_varying_type(var, stage->Stage)
         ->count_attribute_slots(stage->Stage == MESA_SHADER_VERTEX);
      for (unsigned i = 0; i < num_elements; i++) {
         if (var_slot >= 0 && var_slot < MAX_VARYINGS_INCL_PATCH)
            slots |= UINT64_C(1) << var_slot;
         var_slot += 1;
      }
   }

   return slots;
}


/**
 * 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_linked_shader *producer,
                         gl_linked_shader *consumer,
                         unsigned num_tfeedback_decls,
                         tfeedback_decl *tfeedback_decls,
                         const uint64_t reserved_slots)
{
   /* Tessellation shaders treat inputs and outputs as shared memory and can
    * access inputs and outputs of other invocations.
    * Therefore, they can't be lowered to temps easily (and definitely not
    * efficiently).
    */
   bool unpackable_tess =
      (consumer && consumer->Stage == MESA_SHADER_TESS_EVAL) ||
      (consumer && consumer->Stage == MESA_SHADER_TESS_CTRL) ||
      (producer && producer->Stage == MESA_SHADER_TESS_CTRL);

   /* Transform feedback code assumes varying arrays are packed, so if the
    * driver has disabled varying packing, make sure to at least enable
    * packing required by transform feedback.
    */
   bool xfb_enabled =
      ctx->Extensions.EXT_transform_feedback && !unpackable_tess;

   /* Disable packing on outward facing interfaces for SSO because in ES we
    * need to retain the unpacked varying information for draw time
    * validation.
    *
    * Packing is still enabled on individual arrays, structs, and matrices as
    * these are required by the transform feedback code and it is still safe
    * to do so. We also enable packing when a varying is only used for
    * transform feedback and its not a SSO.
    */
   bool disable_varying_packing =
      ctx->Const.DisableVaryingPacking || unpackable_tess;
   if (prog->SeparateShader && (producer == NULL || consumer == NULL))
      disable_varying_packing = true;

   varying_matches matches(disable_varying_packing, xfb_enabled,
                           producer ? producer->Stage : (gl_shader_stage)-1,
                           consumer ? consumer->Stage : (gl_shader_stage)-1);
   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);
   ir_variable *consumer_inputs_with_locations[VARYING_SLOT_TESS_MAX] = {
      NULL,
   };

   unsigned consumer_vertices = 0;
   if (consumer && consumer->Stage == MESA_SHADER_GEOMETRY)
      consumer_vertices = prog->Geom.VerticesIn;

   /* Operate in a total of four passes.
    *
    * 1. Sort inputs / outputs into a canonical order.  This is necessary so
    *    that inputs / outputs of separable shaders will be assigned
    *    predictable locations regardless of the order in which declarations
    *    appeared in the shader source.
    *
    * 2. Assign locations for any matching inputs and outputs.
    *
    * 3. Mark output variables in the producer that do not have locations as
    *    not being outputs.  This lets the optimizer eliminate them.
    *
    * 4. Mark input variables in the consumer that do not have locations as
    *    not being inputs.  This lets the optimizer eliminate them.
    */
   if (consumer)
      canonicalize_shader_io(consumer->ir, ir_var_shader_in);

   if (producer)
      canonicalize_shader_io(producer->ir, ir_var_shader_out);

   if (consumer)
      linker::populate_consumer_input_sets(mem_ctx, consumer->ir,
                                           consumer_inputs,
                                           consumer_interface_inputs,
                                           consumer_inputs_with_locations);

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

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

         /* Only geometry shaders can use non-zero streams */
         assert(output_var->data.stream == 0 ||
                (output_var->data.stream < MAX_VERTEX_STREAMS &&
                 producer->Stage == MESA_SHADER_GEOMETRY));

         if (num_tfeedback_decls > 0) {
            tfeedback_candidate_generator g(mem_ctx, tfeedback_candidates);
            g.process(output_var);
         }

         ir_variable *const input_var =
            linker::get_matching_input(mem_ctx, output_var, consumer_inputs,
                                       consumer_interface_inputs,
                                       consumer_inputs_with_locations);

         /* If a matching input variable was found, add this output (and the
          * input) to the set.  If this is a separable program and there is no
          * consumer stage, add the output.
          *
          * Always add TCS outputs. They are shared by all invocations
          * within a patch and can be used as shared memory.
          */
         if (input_var || (prog->SeparateShader && consumer == NULL) ||
             producer->Stage == MESA_SHADER_TESS_CTRL) {
            matches.record(output_var, input_var);
         }

         /* Only stream 0 outputs can be consumed in the next stage */
         if (input_var && output_var->data.stream != 0) {
            linker_error(prog, "output %s is assigned to stream=%d but "
                         "is linked to an input, which requires stream=0",
                         output_var->name, output_var->data.stream);
            return false;
         }
      }
   } else {
      /* If there's no producer stage, then this must be a separable program.
       * For example, we may have a program that has just a fragment shader.
       * Later this program will be used with some arbitrary vertex (or
       * geometry) shader program.  This means that locations must be assigned
       * for all the inputs.
       */
      foreach_in_list(ir_instruction, node, consumer->ir) {
         ir_variable *const input_var = node->as_variable();

         if (input_var == NULL || input_var->data.mode != ir_var_shader_in)
            continue;

         matches.record(NULL, input_var);
      }
   }

   hash_table_dtor(consumer_inputs);
   hash_table_dtor(consumer_interface_inputs);

   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);
         return false;
      }

      if (matched_candidate->toplevel_var->data.is_unmatched_generic_inout) {
         matched_candidate->toplevel_var->data.is_xfb_only = 1;
         matches.record(matched_candidate->toplevel_var, NULL);
      }
   }

   const unsigned slots_used = matches.assign_locations(prog, reserved_slots);
   matches.store_locations();

   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);
         return false;
      }
   }
   hash_table_dtor(tfeedback_candidates);

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

         if (var && var->data.mode == ir_var_shader_in &&
             var->data.is_unmatched_generic_inout) {
            if (!prog->IsES && 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_shader_stage_to_string(consumer->Stage),
                            var->name,
                            _mesa_shader_stage_to_string(producer->Stage));
            } else {
               linker_warning(prog, "%s shader varying %s not written "
                              "by %s shader\n.",
                              _mesa_shader_stage_to_string(consumer->Stage),
                              var->name,
                              _mesa_shader_stage_to_string(producer->Stage));
            }
         }
      }

      /* Now that validation is done its safe to remove unused varyings. As
       * we have both a producer and consumer its safe to remove unused
       * varyings even if the program is a SSO because the stages are being
       * linked together i.e. we have a multi-stage SSO.
       */
      remove_unused_shader_inputs_and_outputs(false, producer,
                                              ir_var_shader_out);
      remove_unused_shader_inputs_and_outputs(false, consumer,
                                              ir_var_shader_in);
   }

   if (producer) {
      lower_packed_varyings(mem_ctx, slots_used, ir_var_shader_out,
                            0, producer, disable_varying_packing,
                            xfb_enabled);
   }

   if (consumer) {
      lower_packed_varyings(mem_ctx, slots_used, ir_var_shader_in,
                            consumer_vertices, consumer,
                            disable_varying_packing, xfb_enabled);
   }

   return true;
}

bool
check_against_output_limit(struct gl_context *ctx,
                           struct gl_shader_program *prog,
                           gl_linked_shader *producer,
                           unsigned num_explicit_locations)
{
   unsigned output_vectors = num_explicit_locations;

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

      if (var && !var->data.explicit_location &&
          var->data.mode == ir_var_shader_out &&
          var_counts_against_varying_limit(producer->Stage, var)) {
         /* outputs for fragment shader can't be doubles */
         output_vectors += var->type->count_attribute_slots(false);
      }
   }

   assert(producer->Stage != MESA_SHADER_FRAGMENT);
   unsigned max_output_components =
      ctx->Const.Program[producer->Stage].MaxOutputComponents;

   const unsigned output_components = output_vectors * 4;
   if (output_components > max_output_components) {
      if (ctx->API == API_OPENGLES2 || prog->IsES)
         linker_error(prog, "%s shader uses too many output vectors "
                      "(%u > %u)\n",
                      _mesa_shader_stage_to_string(producer->Stage),
                      output_vectors,
                      max_output_components / 4);
      else
         linker_error(prog, "%s shader uses too many output components "
                      "(%u > %u)\n",
                      _mesa_shader_stage_to_string(producer->Stage),
                      output_components,
                      max_output_components);

      return false;
   }

   return true;
}

bool
check_against_input_limit(struct gl_context *ctx,
                          struct gl_shader_program *prog,
                          gl_linked_shader *consumer,
                          unsigned num_explicit_locations)
{
   unsigned input_vectors = num_explicit_locations;

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

      if (var && !var->data.explicit_location &&
          var->data.mode == ir_var_shader_in &&
          var_counts_against_varying_limit(consumer->Stage, var)) {
         /* vertex inputs aren't varying counted */
         input_vectors += var->type->count_attribute_slots(false);
      }
   }

   assert(consumer->Stage != MESA_SHADER_VERTEX);
   unsigned max_input_components =
      ctx->Const.Program[consumer->Stage].MaxInputComponents;

   const unsigned input_components = input_vectors * 4;
   if (input_components > max_input_components) {
      if (ctx->API == API_OPENGLES2 || prog->IsES)
         linker_error(prog, "%s shader uses too many input vectors "
                      "(%u > %u)\n",
                      _mesa_shader_stage_to_string(consumer->Stage),
                      input_vectors,
                      max_input_components / 4);
      else
         linker_error(prog, "%s shader uses too many input components "
                      "(%u > %u)\n",
                      _mesa_shader_stage_to_string(consumer->Stage),
                      input_components,
                      max_input_components);

      return false;
   }

   return true;
}