summaryrefslogtreecommitdiffstats
path: root/src/intel/compiler/brw_vec4.cpp
blob: fb4cf365cb455667517795e47d1636a78d11f18d (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
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
/*
 * Copyright © 2011 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.
 */

#include "brw_vec4.h"
#include "brw_fs.h"
#include "brw_cfg.h"
#include "brw_nir.h"
#include "brw_vec4_builder.h"
#include "brw_vec4_live_variables.h"
#include "brw_vec4_vs.h"
#include "brw_dead_control_flow.h"
#include "dev/gen_debug.h"
#include "program/prog_parameter.h"
#include "util/u_math.h"

#define MAX_INSTRUCTION (1 << 30)

using namespace brw;

namespace brw {

void
src_reg::init()
{
   memset((void*)this, 0, sizeof(*this));
   this->file = BAD_FILE;
   this->type = BRW_REGISTER_TYPE_UD;
}

src_reg::src_reg(enum brw_reg_file file, int nr, const glsl_type *type)
{
   init();

   this->file = file;
   this->nr = nr;
   if (type && (type->is_scalar() || type->is_vector() || type->is_matrix()))
      this->swizzle = brw_swizzle_for_size(type->vector_elements);
   else
      this->swizzle = BRW_SWIZZLE_XYZW;
   if (type)
      this->type = brw_type_for_base_type(type);
}

/** Generic unset register constructor. */
src_reg::src_reg()
{
   init();
}

src_reg::src_reg(struct ::brw_reg reg) :
   backend_reg(reg)
{
   this->offset = 0;
   this->reladdr = NULL;
}

src_reg::src_reg(const dst_reg &reg) :
   backend_reg(reg)
{
   this->reladdr = reg.reladdr;
   this->swizzle = brw_swizzle_for_mask(reg.writemask);
}

void
dst_reg::init()
{
   memset((void*)this, 0, sizeof(*this));
   this->file = BAD_FILE;
   this->type = BRW_REGISTER_TYPE_UD;
   this->writemask = WRITEMASK_XYZW;
}

dst_reg::dst_reg()
{
   init();
}

dst_reg::dst_reg(enum brw_reg_file file, int nr)
{
   init();

   this->file = file;
   this->nr = nr;
}

dst_reg::dst_reg(enum brw_reg_file file, int nr, const glsl_type *type,
                 unsigned writemask)
{
   init();

   this->file = file;
   this->nr = nr;
   this->type = brw_type_for_base_type(type);
   this->writemask = writemask;
}

dst_reg::dst_reg(enum brw_reg_file file, int nr, brw_reg_type type,
                 unsigned writemask)
{
   init();

   this->file = file;
   this->nr = nr;
   this->type = type;
   this->writemask = writemask;
}

dst_reg::dst_reg(struct ::brw_reg reg) :
   backend_reg(reg)
{
   this->offset = 0;
   this->reladdr = NULL;
}

dst_reg::dst_reg(const src_reg &reg) :
   backend_reg(reg)
{
   this->writemask = brw_mask_for_swizzle(reg.swizzle);
   this->reladdr = reg.reladdr;
}

bool
dst_reg::equals(const dst_reg &r) const
{
   return (this->backend_reg::equals(r) &&
           (reladdr == r.reladdr ||
            (reladdr && r.reladdr && reladdr->equals(*r.reladdr))));
}

bool
vec4_instruction::is_send_from_grf()
{
   switch (opcode) {
   case SHADER_OPCODE_SHADER_TIME_ADD:
   case VS_OPCODE_PULL_CONSTANT_LOAD_GEN7:
   case VEC4_OPCODE_UNTYPED_ATOMIC:
   case VEC4_OPCODE_UNTYPED_SURFACE_READ:
   case VEC4_OPCODE_UNTYPED_SURFACE_WRITE:
   case VEC4_OPCODE_URB_READ:
   case TCS_OPCODE_URB_WRITE:
   case TCS_OPCODE_RELEASE_INPUT:
   case SHADER_OPCODE_BARRIER:
      return true;
   default:
      return false;
   }
}

/**
 * Returns true if this instruction's sources and destinations cannot
 * safely be the same register.
 *
 * In most cases, a register can be written over safely by the same
 * instruction that is its last use.  For a single instruction, the
 * sources are dereferenced before writing of the destination starts
 * (naturally).
 *
 * However, there are a few cases where this can be problematic:
 *
 * - Virtual opcodes that translate to multiple instructions in the
 *   code generator: if src == dst and one instruction writes the
 *   destination before a later instruction reads the source, then
 *   src will have been clobbered.
 *
 * The register allocator uses this information to set up conflicts between
 * GRF sources and the destination.
 */
bool
vec4_instruction::has_source_and_destination_hazard() const
{
   switch (opcode) {
   case TCS_OPCODE_SET_INPUT_URB_OFFSETS:
   case TCS_OPCODE_SET_OUTPUT_URB_OFFSETS:
   case TES_OPCODE_ADD_INDIRECT_URB_OFFSET:
      return true;
   default:
      /* 8-wide compressed DF operations are executed as two 4-wide operations,
       * so we have a src/dst hazard if the first half of the instruction
       * overwrites the source of the second half. Prevent this by marking
       * compressed instructions as having src/dst hazards, so the register
       * allocator assigns safe register regions for dst and srcs.
       */
      return size_written > REG_SIZE;
   }
}

unsigned
vec4_instruction::size_read(unsigned arg) const
{
   switch (opcode) {
   case SHADER_OPCODE_SHADER_TIME_ADD:
   case VEC4_OPCODE_UNTYPED_ATOMIC:
   case VEC4_OPCODE_UNTYPED_SURFACE_READ:
   case VEC4_OPCODE_UNTYPED_SURFACE_WRITE:
   case TCS_OPCODE_URB_WRITE:
      if (arg == 0)
         return mlen * REG_SIZE;
      break;
   case VS_OPCODE_PULL_CONSTANT_LOAD_GEN7:
      if (arg == 1)
         return mlen * REG_SIZE;
      break;
   default:
      break;
   }

   switch (src[arg].file) {
   case BAD_FILE:
      return 0;
   case IMM:
   case UNIFORM:
      return 4 * type_sz(src[arg].type);
   default:
      /* XXX - Represent actual vertical stride. */
      return exec_size * type_sz(src[arg].type);
   }
}

bool
vec4_instruction::can_do_source_mods(const struct gen_device_info *devinfo)
{
   if (devinfo->gen == 6 && is_math())
      return false;

   if (is_send_from_grf())
      return false;

   if (!backend_instruction::can_do_source_mods())
      return false;

   return true;
}

bool
vec4_instruction::can_do_cmod()
{
   if (!backend_instruction::can_do_cmod())
      return false;

   /* The accumulator result appears to get used for the conditional modifier
    * generation.  When negating a UD value, there is a 33rd bit generated for
    * the sign in the accumulator value, so now you can't check, for example,
    * equality with a 32-bit value.  See piglit fs-op-neg-uvec4.
    */
   for (unsigned i = 0; i < 3; i++) {
      if (src[i].file != BAD_FILE &&
          type_is_unsigned_int(src[i].type) && src[i].negate)
         return false;
   }

   return true;
}

bool
vec4_instruction::can_do_writemask(const struct gen_device_info *devinfo)
{
   switch (opcode) {
   case SHADER_OPCODE_GEN4_SCRATCH_READ:
   case VEC4_OPCODE_DOUBLE_TO_F32:
   case VEC4_OPCODE_DOUBLE_TO_D32:
   case VEC4_OPCODE_DOUBLE_TO_U32:
   case VEC4_OPCODE_TO_DOUBLE:
   case VEC4_OPCODE_PICK_LOW_32BIT:
   case VEC4_OPCODE_PICK_HIGH_32BIT:
   case VEC4_OPCODE_SET_LOW_32BIT:
   case VEC4_OPCODE_SET_HIGH_32BIT:
   case VS_OPCODE_PULL_CONSTANT_LOAD:
   case VS_OPCODE_PULL_CONSTANT_LOAD_GEN7:
   case VS_OPCODE_SET_SIMD4X2_HEADER_GEN9:
   case TCS_OPCODE_SET_INPUT_URB_OFFSETS:
   case TCS_OPCODE_SET_OUTPUT_URB_OFFSETS:
   case TES_OPCODE_CREATE_INPUT_READ_HEADER:
   case TES_OPCODE_ADD_INDIRECT_URB_OFFSET:
   case VEC4_OPCODE_URB_READ:
   case SHADER_OPCODE_MOV_INDIRECT:
      return false;
   default:
      /* The MATH instruction on Gen6 only executes in align1 mode, which does
       * not support writemasking.
       */
      if (devinfo->gen == 6 && is_math())
         return false;

      if (is_tex())
         return false;

      return true;
   }
}

bool
vec4_instruction::can_change_types() const
{
   return dst.type == src[0].type &&
          !src[0].abs && !src[0].negate && !saturate &&
          (opcode == BRW_OPCODE_MOV ||
           (opcode == BRW_OPCODE_SEL &&
            dst.type == src[1].type &&
            predicate != BRW_PREDICATE_NONE &&
            !src[1].abs && !src[1].negate));
}

/**
 * Returns how many MRFs an opcode will write over.
 *
 * Note that this is not the 0 or 1 implied writes in an actual gen
 * instruction -- the generate_* functions generate additional MOVs
 * for setup.
 */
int
vec4_visitor::implied_mrf_writes(vec4_instruction *inst)
{
   if (inst->mlen == 0 || inst->is_send_from_grf())
      return 0;

   switch (inst->opcode) {
   case SHADER_OPCODE_RCP:
   case SHADER_OPCODE_RSQ:
   case SHADER_OPCODE_SQRT:
   case SHADER_OPCODE_EXP2:
   case SHADER_OPCODE_LOG2:
   case SHADER_OPCODE_SIN:
   case SHADER_OPCODE_COS:
      return 1;
   case SHADER_OPCODE_INT_QUOTIENT:
   case SHADER_OPCODE_INT_REMAINDER:
   case SHADER_OPCODE_POW:
   case TCS_OPCODE_THREAD_END:
      return 2;
   case VS_OPCODE_URB_WRITE:
      return 1;
   case VS_OPCODE_PULL_CONSTANT_LOAD:
      return 2;
   case SHADER_OPCODE_GEN4_SCRATCH_READ:
      return 2;
   case SHADER_OPCODE_GEN4_SCRATCH_WRITE:
      return 3;
   case GS_OPCODE_URB_WRITE:
   case GS_OPCODE_URB_WRITE_ALLOCATE:
   case GS_OPCODE_THREAD_END:
      return 0;
   case GS_OPCODE_FF_SYNC:
      return 1;
   case TCS_OPCODE_URB_WRITE:
      return 0;
   case SHADER_OPCODE_SHADER_TIME_ADD:
      return 0;
   case SHADER_OPCODE_TEX:
   case SHADER_OPCODE_TXL:
   case SHADER_OPCODE_TXD:
   case SHADER_OPCODE_TXF:
   case SHADER_OPCODE_TXF_CMS:
   case SHADER_OPCODE_TXF_CMS_W:
   case SHADER_OPCODE_TXF_MCS:
   case SHADER_OPCODE_TXS:
   case SHADER_OPCODE_TG4:
   case SHADER_OPCODE_TG4_OFFSET:
   case SHADER_OPCODE_SAMPLEINFO:
   case SHADER_OPCODE_GET_BUFFER_SIZE:
      return inst->header_size;
   default:
      unreachable("not reached");
   }
}

bool
src_reg::equals(const src_reg &r) const
{
   return (this->backend_reg::equals(r) &&
	   !reladdr && !r.reladdr);
}

bool
src_reg::negative_equals(const src_reg &r) const
{
   return this->backend_reg::negative_equals(r) &&
          !reladdr && !r.reladdr;
}

bool
vec4_visitor::opt_vector_float()
{
   bool progress = false;

   foreach_block(block, cfg) {
      unsigned last_reg = ~0u, last_offset = ~0u;
      enum brw_reg_file last_reg_file = BAD_FILE;

      uint8_t imm[4] = { 0 };
      int inst_count = 0;
      vec4_instruction *imm_inst[4];
      unsigned writemask = 0;
      enum brw_reg_type dest_type = BRW_REGISTER_TYPE_F;

      foreach_inst_in_block_safe(vec4_instruction, inst, block) {
         int vf = -1;
         enum brw_reg_type need_type = BRW_REGISTER_TYPE_LAST;

         /* Look for unconditional MOVs from an immediate with a partial
          * writemask.  Skip type-conversion MOVs other than integer 0,
          * where the type doesn't matter.  See if the immediate can be
          * represented as a VF.
          */
         if (inst->opcode == BRW_OPCODE_MOV &&
             inst->src[0].file == IMM &&
             inst->predicate == BRW_PREDICATE_NONE &&
             inst->dst.writemask != WRITEMASK_XYZW &&
             type_sz(inst->src[0].type) < 8 &&
             (inst->src[0].type == inst->dst.type || inst->src[0].d == 0)) {

            vf = brw_float_to_vf(inst->src[0].d);
            need_type = BRW_REGISTER_TYPE_D;

            if (vf == -1) {
               vf = brw_float_to_vf(inst->src[0].f);
               need_type = BRW_REGISTER_TYPE_F;
            }
         } else {
            last_reg = ~0u;
         }

         /* If this wasn't a MOV, or the destination register doesn't match,
          * or we have to switch destination types, then this breaks our
          * sequence.  Combine anything we've accumulated so far.
          */
         if (last_reg != inst->dst.nr ||
             last_offset != inst->dst.offset ||
             last_reg_file != inst->dst.file ||
             (vf > 0 && dest_type != need_type)) {

            if (inst_count > 1) {
               unsigned vf;
               memcpy(&vf, imm, sizeof(vf));
               vec4_instruction *mov = MOV(imm_inst[0]->dst, brw_imm_vf(vf));
               mov->dst.type = dest_type;
               mov->dst.writemask = writemask;
               inst->insert_before(block, mov);

               for (int i = 0; i < inst_count; i++) {
                  imm_inst[i]->remove(block);
               }

               progress = true;
            }

            inst_count = 0;
            last_reg = ~0u;;
            writemask = 0;
            dest_type = BRW_REGISTER_TYPE_F;

            for (int i = 0; i < 4; i++) {
               imm[i] = 0;
            }
         }

         /* Record this instruction's value (if it was representable). */
         if (vf != -1) {
            if ((inst->dst.writemask & WRITEMASK_X) != 0)
               imm[0] = vf;
            if ((inst->dst.writemask & WRITEMASK_Y) != 0)
               imm[1] = vf;
            if ((inst->dst.writemask & WRITEMASK_Z) != 0)
               imm[2] = vf;
            if ((inst->dst.writemask & WRITEMASK_W) != 0)
               imm[3] = vf;

            writemask |= inst->dst.writemask;
            imm_inst[inst_count++] = inst;

            last_reg = inst->dst.nr;
            last_offset = inst->dst.offset;
            last_reg_file = inst->dst.file;
            if (vf > 0)
               dest_type = need_type;
         }
      }
   }

   if (progress)
      invalidate_live_intervals();

   return progress;
}

/* Replaces unused channels of a swizzle with channels that are used.
 *
 * For instance, this pass transforms
 *
 *    mov vgrf4.yz, vgrf5.wxzy
 *
 * into
 *
 *    mov vgrf4.yz, vgrf5.xxzx
 *
 * This eliminates false uses of some channels, letting dead code elimination
 * remove the instructions that wrote them.
 */
bool
vec4_visitor::opt_reduce_swizzle()
{
   bool progress = false;

   foreach_block_and_inst_safe(block, vec4_instruction, inst, cfg) {
      if (inst->dst.file == BAD_FILE ||
          inst->dst.file == ARF ||
          inst->dst.file == FIXED_GRF ||
          inst->is_send_from_grf())
         continue;

      unsigned swizzle;

      /* Determine which channels of the sources are read. */
      switch (inst->opcode) {
      case VEC4_OPCODE_PACK_BYTES:
      case BRW_OPCODE_DP4:
      case BRW_OPCODE_DPH: /* FINISHME: DPH reads only three channels of src0,
                            *           but all four of src1.
                            */
         swizzle = brw_swizzle_for_size(4);
         break;
      case BRW_OPCODE_DP3:
         swizzle = brw_swizzle_for_size(3);
         break;
      case BRW_OPCODE_DP2:
         swizzle = brw_swizzle_for_size(2);
         break;

      case VEC4_OPCODE_TO_DOUBLE:
      case VEC4_OPCODE_DOUBLE_TO_F32:
      case VEC4_OPCODE_DOUBLE_TO_D32:
      case VEC4_OPCODE_DOUBLE_TO_U32:
      case VEC4_OPCODE_PICK_LOW_32BIT:
      case VEC4_OPCODE_PICK_HIGH_32BIT:
      case VEC4_OPCODE_SET_LOW_32BIT:
      case VEC4_OPCODE_SET_HIGH_32BIT:
         swizzle = brw_swizzle_for_size(4);
         break;

      default:
         swizzle = brw_swizzle_for_mask(inst->dst.writemask);
         break;
      }

      /* Update sources' swizzles. */
      for (int i = 0; i < 3; i++) {
         if (inst->src[i].file != VGRF &&
             inst->src[i].file != ATTR &&
             inst->src[i].file != UNIFORM)
            continue;

         const unsigned new_swizzle =
            brw_compose_swizzle(swizzle, inst->src[i].swizzle);
         if (inst->src[i].swizzle != new_swizzle) {
            inst->src[i].swizzle = new_swizzle;
            progress = true;
         }
      }
   }

   if (progress)
      invalidate_live_intervals();

   return progress;
}

void
vec4_visitor::split_uniform_registers()
{
   /* Prior to this, uniforms have been in an array sized according to
    * the number of vector uniforms present, sparsely filled (so an
    * aggregate results in reg indices being skipped over).  Now we're
    * going to cut those aggregates up so each .nr index is one
    * vector.  The goal is to make elimination of unused uniform
    * components easier later.
    */
   foreach_block_and_inst(block, vec4_instruction, inst, cfg) {
      for (int i = 0 ; i < 3; i++) {
	 if (inst->src[i].file != UNIFORM)
	    continue;

	 assert(!inst->src[i].reladdr);

         inst->src[i].nr += inst->src[i].offset / 16;
	 inst->src[i].offset %= 16;
      }
   }
}

/* This function returns the register number where we placed the uniform */
static int
set_push_constant_loc(const int nr_uniforms, int *new_uniform_count,
                      const int src, const int size, const int channel_size,
                      int *new_loc, int *new_chan,
                      int *new_chans_used)
{
   int dst;
   /* Find the lowest place we can slot this uniform in. */
   for (dst = 0; dst < nr_uniforms; dst++) {
      if (ALIGN(new_chans_used[dst], channel_size) + size <= 4)
         break;
   }

   assert(dst < nr_uniforms);

   new_loc[src] = dst;
   new_chan[src] = ALIGN(new_chans_used[dst], channel_size);
   new_chans_used[dst] = ALIGN(new_chans_used[dst], channel_size) + size;

   *new_uniform_count = MAX2(*new_uniform_count, dst + 1);
   return dst;
}

void
vec4_visitor::pack_uniform_registers()
{
   if (!compiler->compact_params)
      return;

   uint8_t chans_used[this->uniforms];
   int new_loc[this->uniforms];
   int new_chan[this->uniforms];
   bool is_aligned_to_dvec4[this->uniforms];
   int new_chans_used[this->uniforms];
   int channel_sizes[this->uniforms];

   memset(chans_used, 0, sizeof(chans_used));
   memset(new_loc, 0, sizeof(new_loc));
   memset(new_chan, 0, sizeof(new_chan));
   memset(new_chans_used, 0, sizeof(new_chans_used));
   memset(is_aligned_to_dvec4, 0, sizeof(is_aligned_to_dvec4));
   memset(channel_sizes, 0, sizeof(channel_sizes));

   /* Find which uniform vectors are actually used by the program.  We
    * expect unused vector elements when we've moved array access out
    * to pull constants, and from some GLSL code generators like wine.
    */
   foreach_block_and_inst(block, vec4_instruction, inst, cfg) {
      unsigned readmask;
      switch (inst->opcode) {
      case VEC4_OPCODE_PACK_BYTES:
      case BRW_OPCODE_DP4:
      case BRW_OPCODE_DPH:
         readmask = 0xf;
         break;
      case BRW_OPCODE_DP3:
         readmask = 0x7;
         break;
      case BRW_OPCODE_DP2:
         readmask = 0x3;
         break;
      default:
         readmask = inst->dst.writemask;
         break;
      }

      for (int i = 0 ; i < 3; i++) {
         if (inst->src[i].file != UNIFORM)
            continue;

         assert(type_sz(inst->src[i].type) % 4 == 0);
         int channel_size = type_sz(inst->src[i].type) / 4;

         int reg = inst->src[i].nr;
         for (int c = 0; c < 4; c++) {
            if (!(readmask & (1 << c)))
               continue;

            unsigned channel = BRW_GET_SWZ(inst->src[i].swizzle, c) + 1;
            unsigned used = MAX2(chans_used[reg], channel * channel_size);
            if (used <= 4) {
               chans_used[reg] = used;
               channel_sizes[reg] = MAX2(channel_sizes[reg], channel_size);
            } else {
               is_aligned_to_dvec4[reg] = true;
               is_aligned_to_dvec4[reg + 1] = true;
               chans_used[reg + 1] = used - 4;
               channel_sizes[reg + 1] = MAX2(channel_sizes[reg + 1], channel_size);
            }
         }
      }

      if (inst->opcode == SHADER_OPCODE_MOV_INDIRECT &&
          inst->src[0].file == UNIFORM) {
         assert(inst->src[2].file == BRW_IMMEDIATE_VALUE);
         assert(inst->src[0].subnr == 0);

         unsigned bytes_read = inst->src[2].ud;
         assert(bytes_read % 4 == 0);
         unsigned vec4s_read = DIV_ROUND_UP(bytes_read, 16);

         /* We just mark every register touched by a MOV_INDIRECT as being
          * fully used.  This ensures that it doesn't broken up piecewise by
          * the next part of our packing algorithm.
          */
         int reg = inst->src[0].nr;
         int channel_size = type_sz(inst->src[0].type) / 4;
         for (unsigned i = 0; i < vec4s_read; i++) {
            chans_used[reg + i] = 4;
            channel_sizes[reg + i] = MAX2(channel_sizes[reg + i], channel_size);
         }
      }
   }

   int new_uniform_count = 0;

   /* As the uniforms are going to be reordered, take the data from a temporary
    * copy of the original param[].
    */
   uint32_t *param = ralloc_array(NULL, uint32_t, stage_prog_data->nr_params);
   memcpy(param, stage_prog_data->param,
          sizeof(uint32_t) * stage_prog_data->nr_params);

   /* Now, figure out a packing of the live uniform vectors into our
    * push constants. Start with dvec{3,4} because they are aligned to
    * dvec4 size (2 vec4).
    */
   for (int src = 0; src < uniforms; src++) {
      int size = chans_used[src];

      if (size == 0 || !is_aligned_to_dvec4[src])
         continue;

      /* dvec3 are aligned to dvec4 size, apply the alignment of the size
       * to 4 to avoid moving last component of a dvec3 to the available
       * location at the end of a previous dvec3. These available locations
       * could be filled by smaller variables in next loop.
       */
      size = ALIGN(size, 4);
      int dst = set_push_constant_loc(uniforms, &new_uniform_count,
                                      src, size, channel_sizes[src],
                                      new_loc, new_chan,
                                      new_chans_used);
      /* Move the references to the data */
      for (int j = 0; j < size; j++) {
         stage_prog_data->param[dst * 4 + new_chan[src] + j] =
            param[src * 4 + j];
      }
   }

   /* Continue with the rest of data, which is aligned to vec4. */
   for (int src = 0; src < uniforms; src++) {
      int size = chans_used[src];

      if (size == 0 || is_aligned_to_dvec4[src])
         continue;

      int dst = set_push_constant_loc(uniforms, &new_uniform_count,
                                      src, size, channel_sizes[src],
                                      new_loc, new_chan,
                                      new_chans_used);
      /* Move the references to the data */
      for (int j = 0; j < size; j++) {
         stage_prog_data->param[dst * 4 + new_chan[src] + j] =
            param[src * 4 + j];
      }
   }

   ralloc_free(param);
   this->uniforms = new_uniform_count;

   /* Now, update the instructions for our repacked uniforms. */
   foreach_block_and_inst(block, vec4_instruction, inst, cfg) {
      for (int i = 0 ; i < 3; i++) {
         int src = inst->src[i].nr;

         if (inst->src[i].file != UNIFORM)
            continue;

         int chan = new_chan[src] / channel_sizes[src];
         inst->src[i].nr = new_loc[src];
         inst->src[i].swizzle += BRW_SWIZZLE4(chan, chan, chan, chan);
      }
   }
}

/**
 * Does algebraic optimizations (0 * a = 0, 1 * a = a, a + 0 = a).
 *
 * While GLSL IR also performs this optimization, we end up with it in
 * our instruction stream for a couple of reasons.  One is that we
 * sometimes generate silly instructions, for example in array access
 * where we'll generate "ADD offset, index, base" even if base is 0.
 * The other is that GLSL IR's constant propagation doesn't track the
 * components of aggregates, so some VS patterns (initialize matrix to
 * 0, accumulate in vertex blending factors) end up breaking down to
 * instructions involving 0.
 */
bool
vec4_visitor::opt_algebraic()
{
   bool progress = false;

   foreach_block_and_inst(block, vec4_instruction, inst, cfg) {
      switch (inst->opcode) {
      case BRW_OPCODE_MOV:
         if (inst->src[0].file != IMM)
            break;

         if (inst->saturate) {
            /* Full mixed-type saturates don't happen.  However, we can end up
             * with things like:
             *
             *    mov.sat(8) g21<1>DF       -1F
             *
             * Other mixed-size-but-same-base-type cases may also be possible.
             */
            if (inst->dst.type != inst->src[0].type &&
                inst->dst.type != BRW_REGISTER_TYPE_DF &&
                inst->src[0].type != BRW_REGISTER_TYPE_F)
               assert(!"unimplemented: saturate mixed types");

            if (brw_saturate_immediate(inst->src[0].type,
                                       &inst->src[0].as_brw_reg())) {
               inst->saturate = false;
               progress = true;
            }
         }
         break;

      case BRW_OPCODE_OR:
         if (inst->src[1].is_zero()) {
            inst->opcode = BRW_OPCODE_MOV;
            inst->src[1] = src_reg();
            progress = true;
         }
         break;

      case VEC4_OPCODE_UNPACK_UNIFORM:
         if (inst->src[0].file != UNIFORM) {
            inst->opcode = BRW_OPCODE_MOV;
            progress = true;
         }
         break;

      case BRW_OPCODE_ADD:
	 if (inst->src[1].is_zero()) {
	    inst->opcode = BRW_OPCODE_MOV;
	    inst->src[1] = src_reg();
	    progress = true;
	 }
	 break;

      case BRW_OPCODE_MUL:
	 if (inst->src[1].is_zero()) {
	    inst->opcode = BRW_OPCODE_MOV;
	    switch (inst->src[0].type) {
	    case BRW_REGISTER_TYPE_F:
	       inst->src[0] = brw_imm_f(0.0f);
	       break;
	    case BRW_REGISTER_TYPE_D:
	       inst->src[0] = brw_imm_d(0);
	       break;
	    case BRW_REGISTER_TYPE_UD:
	       inst->src[0] = brw_imm_ud(0u);
	       break;
	    default:
	       unreachable("not reached");
	    }
	    inst->src[1] = src_reg();
	    progress = true;
	 } else if (inst->src[1].is_one()) {
	    inst->opcode = BRW_OPCODE_MOV;
	    inst->src[1] = src_reg();
	    progress = true;
         } else if (inst->src[1].is_negative_one()) {
            inst->opcode = BRW_OPCODE_MOV;
            inst->src[0].negate = !inst->src[0].negate;
            inst->src[1] = src_reg();
            progress = true;
	 }
	 break;
      case SHADER_OPCODE_BROADCAST:
         if (is_uniform(inst->src[0]) ||
             inst->src[1].is_zero()) {
            inst->opcode = BRW_OPCODE_MOV;
            inst->src[1] = src_reg();
            inst->force_writemask_all = true;
            progress = true;
         }
         break;

      default:
	 break;
      }
   }

   if (progress)
      invalidate_live_intervals();

   return progress;
}

/**
 * Only a limited number of hardware registers may be used for push
 * constants, so this turns access to the overflowed constants into
 * pull constants.
 */
void
vec4_visitor::move_push_constants_to_pull_constants()
{
   int pull_constant_loc[this->uniforms];

   /* Only allow 32 registers (256 uniform components) as push constants,
    * which is the limit on gen6.
    *
    * If changing this value, note the limitation about total_regs in
    * brw_curbe.c.
    */
   int max_uniform_components = 32 * 8;
   if (this->uniforms * 4 <= max_uniform_components)
      return;

   /* Make some sort of choice as to which uniforms get sent to pull
    * constants.  We could potentially do something clever here like
    * look for the most infrequently used uniform vec4s, but leave
    * that for later.
    */
   for (int i = 0; i < this->uniforms * 4; i += 4) {
      pull_constant_loc[i / 4] = -1;

      if (i >= max_uniform_components) {
         uint32_t *values = &stage_prog_data->param[i];

         /* Try to find an existing copy of this uniform in the pull
          * constants if it was part of an array access already.
          */
         for (unsigned int j = 0; j < stage_prog_data->nr_pull_params; j += 4) {
            int matches;

            for (matches = 0; matches < 4; matches++) {
               if (stage_prog_data->pull_param[j + matches] != values[matches])
                  break;
            }

            if (matches == 4) {
               pull_constant_loc[i / 4] = j / 4;
               break;
            }
         }

         if (pull_constant_loc[i / 4] == -1) {
            assert(stage_prog_data->nr_pull_params % 4 == 0);
            pull_constant_loc[i / 4] = stage_prog_data->nr_pull_params / 4;

            for (int j = 0; j < 4; j++) {
               stage_prog_data->pull_param[stage_prog_data->nr_pull_params++] =
                  values[j];
            }
         }
      }
   }

   /* Now actually rewrite usage of the things we've moved to pull
    * constants.
    */
   foreach_block_and_inst_safe(block, vec4_instruction, inst, cfg) {
      for (int i = 0 ; i < 3; i++) {
         if (inst->src[i].file != UNIFORM ||
             pull_constant_loc[inst->src[i].nr] == -1)
            continue;

         int uniform = inst->src[i].nr;

         const glsl_type *temp_type = type_sz(inst->src[i].type) == 8 ?
            glsl_type::dvec4_type : glsl_type::vec4_type;
         dst_reg temp = dst_reg(this, temp_type);

         emit_pull_constant_load(block, inst, temp, inst->src[i],
                                 pull_constant_loc[uniform], src_reg());

         inst->src[i].file = temp.file;
         inst->src[i].nr = temp.nr;
         inst->src[i].offset %= 16;
         inst->src[i].reladdr = NULL;
      }
   }

   /* Repack push constants to remove the now-unused ones. */
   pack_uniform_registers();
}

/* Conditions for which we want to avoid setting the dependency control bits */
bool
vec4_visitor::is_dep_ctrl_unsafe(const vec4_instruction *inst)
{
#define IS_DWORD(reg) \
   (reg.type == BRW_REGISTER_TYPE_UD || \
    reg.type == BRW_REGISTER_TYPE_D)

#define IS_64BIT(reg) (reg.file != BAD_FILE && type_sz(reg.type) == 8)

   /* From the Cherryview and Broadwell PRMs:
    *
    * "When source or destination datatype is 64b or operation is integer DWord
    * multiply, DepCtrl must not be used."
    *
    * SKL PRMs don't include this restriction, however, gen7 seems to be
    * affected, at least by the 64b restriction, since DepCtrl with double
    * precision instructions seems to produce GPU hangs in some cases.
    */
   if (devinfo->gen == 8 || gen_device_info_is_9lp(devinfo)) {
      if (inst->opcode == BRW_OPCODE_MUL &&
         IS_DWORD(inst->src[0]) &&
         IS_DWORD(inst->src[1]))
         return true;
   }

   if (devinfo->gen >= 7 && devinfo->gen <= 8) {
      if (IS_64BIT(inst->dst) || IS_64BIT(inst->src[0]) ||
          IS_64BIT(inst->src[1]) || IS_64BIT(inst->src[2]))
      return true;
   }

#undef IS_64BIT
#undef IS_DWORD

   if (devinfo->gen >= 8) {
      if (inst->opcode == BRW_OPCODE_F32TO16)
         return true;
   }

   /*
    * mlen:
    * In the presence of send messages, totally interrupt dependency
    * control. They're long enough that the chance of dependency
    * control around them just doesn't matter.
    *
    * predicate:
    * From the Ivy Bridge PRM, volume 4 part 3.7, page 80:
    * When a sequence of NoDDChk and NoDDClr are used, the last instruction that
    * completes the scoreboard clear must have a non-zero execution mask. This
    * means, if any kind of predication can change the execution mask or channel
    * enable of the last instruction, the optimization must be avoided. This is
    * to avoid instructions being shot down the pipeline when no writes are
    * required.
    *
    * math:
    * Dependency control does not work well over math instructions.
    * NB: Discovered empirically
    */
   return (inst->mlen || inst->predicate || inst->is_math());
}

/**
 * Sets the dependency control fields on instructions after register
 * allocation and before the generator is run.
 *
 * When you have a sequence of instructions like:
 *
 * DP4 temp.x vertex uniform[0]
 * DP4 temp.y vertex uniform[0]
 * DP4 temp.z vertex uniform[0]
 * DP4 temp.w vertex uniform[0]
 *
 * The hardware doesn't know that it can actually run the later instructions
 * while the previous ones are in flight, producing stalls.  However, we have
 * manual fields we can set in the instructions that let it do so.
 */
void
vec4_visitor::opt_set_dependency_control()
{
   vec4_instruction *last_grf_write[BRW_MAX_GRF];
   uint8_t grf_channels_written[BRW_MAX_GRF];
   vec4_instruction *last_mrf_write[BRW_MAX_GRF];
   uint8_t mrf_channels_written[BRW_MAX_GRF];

   assert(prog_data->total_grf ||
          !"Must be called after register allocation");

   foreach_block (block, cfg) {
      memset(last_grf_write, 0, sizeof(last_grf_write));
      memset(last_mrf_write, 0, sizeof(last_mrf_write));

      foreach_inst_in_block (vec4_instruction, inst, block) {
         /* If we read from a register that we were doing dependency control
          * on, don't do dependency control across the read.
          */
         for (int i = 0; i < 3; i++) {
            int reg = inst->src[i].nr + inst->src[i].offset / REG_SIZE;
            if (inst->src[i].file == VGRF) {
               last_grf_write[reg] = NULL;
            } else if (inst->src[i].file == FIXED_GRF) {
               memset(last_grf_write, 0, sizeof(last_grf_write));
               break;
            }
            assert(inst->src[i].file != MRF);
         }

         if (is_dep_ctrl_unsafe(inst)) {
            memset(last_grf_write, 0, sizeof(last_grf_write));
            memset(last_mrf_write, 0, sizeof(last_mrf_write));
            continue;
         }

         /* Now, see if we can do dependency control for this instruction
          * against a previous one writing to its destination.
          */
         int reg = inst->dst.nr + inst->dst.offset / REG_SIZE;
         if (inst->dst.file == VGRF || inst->dst.file == FIXED_GRF) {
            if (last_grf_write[reg] &&
                last_grf_write[reg]->dst.offset == inst->dst.offset &&
                !(inst->dst.writemask & grf_channels_written[reg])) {
               last_grf_write[reg]->no_dd_clear = true;
               inst->no_dd_check = true;
            } else {
               grf_channels_written[reg] = 0;
            }

            last_grf_write[reg] = inst;
            grf_channels_written[reg] |= inst->dst.writemask;
         } else if (inst->dst.file == MRF) {
            if (last_mrf_write[reg] &&
                last_mrf_write[reg]->dst.offset == inst->dst.offset &&
                !(inst->dst.writemask & mrf_channels_written[reg])) {
               last_mrf_write[reg]->no_dd_clear = true;
               inst->no_dd_check = true;
            } else {
               mrf_channels_written[reg] = 0;
            }

            last_mrf_write[reg] = inst;
            mrf_channels_written[reg] |= inst->dst.writemask;
         }
      }
   }
}

bool
vec4_instruction::can_reswizzle(const struct gen_device_info *devinfo,
                                int dst_writemask,
                                int swizzle,
                                int swizzle_mask)
{
   /* Gen6 MATH instructions can not execute in align16 mode, so swizzles
    * are not allowed.
    */
   if (devinfo->gen == 6 && is_math() && swizzle != BRW_SWIZZLE_XYZW)
      return false;

   /* If we write to the flag register changing the swizzle would change
    * what channels are written to the flag register.
    */
   if (writes_flag())
      return false;

   /* We can't swizzle implicit accumulator access.  We'd have to
    * reswizzle the producer of the accumulator value in addition
    * to the consumer (i.e. both MUL and MACH).  Just skip this.
    */
   if (reads_accumulator_implicitly())
      return false;

   if (!can_do_writemask(devinfo) && dst_writemask != WRITEMASK_XYZW)
      return false;

   /* If this instruction sets anything not referenced by swizzle, then we'd
    * totally break it when we reswizzle.
    */
   if (dst.writemask & ~swizzle_mask)
      return false;

   if (mlen > 0)
      return false;

   for (int i = 0; i < 3; i++) {
      if (src[i].is_accumulator())
         return false;
   }

   return true;
}

/**
 * For any channels in the swizzle's source that were populated by this
 * instruction, rewrite the instruction to put the appropriate result directly
 * in those channels.
 *
 * e.g. for swizzle=yywx, MUL a.xy b c -> MUL a.yy_x b.yy z.yy_x
 */
void
vec4_instruction::reswizzle(int dst_writemask, int swizzle)
{
   /* Destination write mask doesn't correspond to source swizzle for the dot
    * product and pack_bytes instructions.
    */
   if (opcode != BRW_OPCODE_DP4 && opcode != BRW_OPCODE_DPH &&
       opcode != BRW_OPCODE_DP3 && opcode != BRW_OPCODE_DP2 &&
       opcode != VEC4_OPCODE_PACK_BYTES) {
      for (int i = 0; i < 3; i++) {
         if (src[i].file == BAD_FILE)
            continue;

         if (src[i].file == IMM) {
            assert(src[i].type != BRW_REGISTER_TYPE_V &&
                   src[i].type != BRW_REGISTER_TYPE_UV);

            /* Vector immediate types need to be reswizzled. */
            if (src[i].type == BRW_REGISTER_TYPE_VF) {
               const unsigned imm[] = {
                  (src[i].ud >>  0) & 0x0ff,
                  (src[i].ud >>  8) & 0x0ff,
                  (src[i].ud >> 16) & 0x0ff,
                  (src[i].ud >> 24) & 0x0ff,
               };

               src[i] = brw_imm_vf4(imm[BRW_GET_SWZ(swizzle, 0)],
                                    imm[BRW_GET_SWZ(swizzle, 1)],
                                    imm[BRW_GET_SWZ(swizzle, 2)],
                                    imm[BRW_GET_SWZ(swizzle, 3)]);
            }

            continue;
         }

         src[i].swizzle = brw_compose_swizzle(swizzle, src[i].swizzle);
      }
   }

   /* Apply the specified swizzle and writemask to the original mask of
    * written components.
    */
   dst.writemask = dst_writemask &
                   brw_apply_swizzle_to_mask(swizzle, dst.writemask);
}

/*
 * Tries to reduce extra MOV instructions by taking temporary GRFs that get
 * just written and then MOVed into another reg and making the original write
 * of the GRF write directly to the final destination instead.
 */
bool
vec4_visitor::opt_register_coalesce()
{
   bool progress = false;
   int next_ip = 0;

   calculate_live_intervals();

   foreach_block_and_inst_safe (block, vec4_instruction, inst, cfg) {
      int ip = next_ip;
      next_ip++;

      if (inst->opcode != BRW_OPCODE_MOV ||
          (inst->dst.file != VGRF && inst->dst.file != MRF) ||
	  inst->predicate ||
	  inst->src[0].file != VGRF ||
	  inst->dst.type != inst->src[0].type ||
	  inst->src[0].abs || inst->src[0].negate || inst->src[0].reladdr)
	 continue;

      /* Remove no-op MOVs */
      if (inst->dst.file == inst->src[0].file &&
          inst->dst.nr == inst->src[0].nr &&
          inst->dst.offset == inst->src[0].offset) {
         bool is_nop_mov = true;

         for (unsigned c = 0; c < 4; c++) {
            if ((inst->dst.writemask & (1 << c)) == 0)
               continue;

            if (BRW_GET_SWZ(inst->src[0].swizzle, c) != c) {
               is_nop_mov = false;
               break;
            }
         }

         if (is_nop_mov) {
            inst->remove(block);
            progress = true;
            continue;
         }
      }

      bool to_mrf = (inst->dst.file == MRF);

      /* Can't coalesce this GRF if someone else was going to
       * read it later.
       */
      if (var_range_end(var_from_reg(alloc, dst_reg(inst->src[0])), 8) > ip)
	 continue;

      /* We need to check interference with the final destination between this
       * instruction and the earliest instruction involved in writing the GRF
       * we're eliminating.  To do that, keep track of which of our source
       * channels we've seen initialized.
       */
      const unsigned chans_needed =
         brw_apply_inv_swizzle_to_mask(inst->src[0].swizzle,
                                       inst->dst.writemask);
      unsigned chans_remaining = chans_needed;

      /* Now walk up the instruction stream trying to see if we can rewrite
       * everything writing to the temporary to write into the destination
       * instead.
       */
      vec4_instruction *_scan_inst = (vec4_instruction *)inst->prev;
      foreach_inst_in_block_reverse_starting_from(vec4_instruction, scan_inst,
                                                  inst) {
         _scan_inst = scan_inst;

         if (regions_overlap(inst->src[0], inst->size_read(0),
                             scan_inst->dst, scan_inst->size_written)) {
            /* Found something writing to the reg we want to coalesce away. */
            if (to_mrf) {
               /* SEND instructions can't have MRF as a destination. */
               if (scan_inst->mlen)
                  break;

               if (devinfo->gen == 6) {
                  /* gen6 math instructions must have the destination be
                   * VGRF, so no compute-to-MRF for them.
                   */
                  if (scan_inst->is_math()) {
                     break;
                  }
               }
            }

            /* VS_OPCODE_UNPACK_FLAGS_SIMD4X2 generates a bunch of mov(1)
             * instructions, and this optimization pass is not capable of
             * handling that.  Bail on these instructions and hope that some
             * later optimization pass can do the right thing after they are
             * expanded.
             */
            if (scan_inst->opcode == VS_OPCODE_UNPACK_FLAGS_SIMD4X2)
               break;

            /* This doesn't handle saturation on the instruction we
             * want to coalesce away if the register types do not match.
             * But if scan_inst is a non type-converting 'mov', we can fix
             * the types later.
             */
            if (inst->saturate &&
                inst->dst.type != scan_inst->dst.type &&
                !(scan_inst->opcode == BRW_OPCODE_MOV &&
                  scan_inst->dst.type == scan_inst->src[0].type))
               break;

            /* Only allow coalescing between registers of the same type size.
             * Otherwise we would need to make the pass aware of the fact that
             * channel sizes are different for single and double precision.
             */
            if (type_sz(inst->src[0].type) != type_sz(scan_inst->src[0].type))
               break;

            /* Check that scan_inst writes the same amount of data as the
             * instruction, otherwise coalescing would lead to writing a
             * different (larger or smaller) region of the destination
             */
            if (scan_inst->size_written != inst->size_written)
               break;

            /* If we can't handle the swizzle, bail. */
            if (!scan_inst->can_reswizzle(devinfo, inst->dst.writemask,
                                          inst->src[0].swizzle,
                                          chans_needed)) {
               break;
            }

            /* This only handles coalescing writes of 8 channels (1 register
             * for single-precision and 2 registers for double-precision)
             * starting at the source offset of the copy instruction.
             */
            if (DIV_ROUND_UP(scan_inst->size_written,
                             type_sz(scan_inst->dst.type)) > 8 ||
                scan_inst->dst.offset != inst->src[0].offset)
               break;

	    /* Mark which channels we found unconditional writes for. */
	    if (!scan_inst->predicate)
               chans_remaining &= ~scan_inst->dst.writemask;

	    if (chans_remaining == 0)
	       break;
	 }

         /* You can't read from an MRF, so if someone else reads our MRF's
          * source GRF that we wanted to rewrite, that stops us.  If it's a
          * GRF we're trying to coalesce to, we don't actually handle
          * rewriting sources so bail in that case as well.
          */
	 bool interfered = false;
	 for (int i = 0; i < 3; i++) {
            if (regions_overlap(inst->src[0], inst->size_read(0),
                                scan_inst->src[i], scan_inst->size_read(i)))
	       interfered = true;
	 }
	 if (interfered)
	    break;

         /* If somebody else writes the same channels of our destination here,
          * we can't coalesce before that.
          */
         if (regions_overlap(inst->dst, inst->size_written,
                             scan_inst->dst, scan_inst->size_written) &&
             (inst->dst.writemask & scan_inst->dst.writemask) != 0) {
            break;
         }

         /* Check for reads of the register we're trying to coalesce into.  We
          * can't go rewriting instructions above that to put some other value
          * in the register instead.
          */
         if (to_mrf && scan_inst->mlen > 0) {
            unsigned start = scan_inst->base_mrf;
            unsigned end = scan_inst->base_mrf + scan_inst->mlen;

            if (inst->dst.nr >= start && inst->dst.nr < end) {
               break;
            }
         } else {
            for (int i = 0; i < 3; i++) {
               if (regions_overlap(inst->dst, inst->size_written,
                                   scan_inst->src[i], scan_inst->size_read(i)))
                  interfered = true;
            }
            if (interfered)
               break;
         }
      }

      if (chans_remaining == 0) {
	 /* If we've made it here, we have an MOV we want to coalesce out, and
	  * a scan_inst pointing to the earliest instruction involved in
	  * computing the value.  Now go rewrite the instruction stream
	  * between the two.
	  */
         vec4_instruction *scan_inst = _scan_inst;
	 while (scan_inst != inst) {
	    if (scan_inst->dst.file == VGRF &&
                scan_inst->dst.nr == inst->src[0].nr &&
		scan_inst->dst.offset == inst->src[0].offset) {
               scan_inst->reswizzle(inst->dst.writemask,
                                    inst->src[0].swizzle);
	       scan_inst->dst.file = inst->dst.file;
               scan_inst->dst.nr = inst->dst.nr;
	       scan_inst->dst.offset = inst->dst.offset;
               if (inst->saturate &&
                   inst->dst.type != scan_inst->dst.type) {
                  /* If we have reached this point, scan_inst is a non
                   * type-converting 'mov' and we can modify its register types
                   * to match the ones in inst. Otherwise, we could have an
                   * incorrect saturation result.
                   */
                  scan_inst->dst.type = inst->dst.type;
                  scan_inst->src[0].type = inst->src[0].type;
               }
	       scan_inst->saturate |= inst->saturate;
	    }
	    scan_inst = (vec4_instruction *)scan_inst->next;
	 }
	 inst->remove(block);
	 progress = true;
      }
   }

   if (progress)
      invalidate_live_intervals();

   return progress;
}

/**
 * Eliminate FIND_LIVE_CHANNEL instructions occurring outside any control
 * flow.  We could probably do better here with some form of divergence
 * analysis.
 */
bool
vec4_visitor::eliminate_find_live_channel()
{
   bool progress = false;
   unsigned depth = 0;

   if (!brw_stage_has_packed_dispatch(devinfo, stage, stage_prog_data)) {
      /* The optimization below assumes that channel zero is live on thread
       * dispatch, which may not be the case if the fixed function dispatches
       * threads sparsely.
       */
      return false;
   }

   foreach_block_and_inst_safe(block, vec4_instruction, inst, cfg) {
      switch (inst->opcode) {
      case BRW_OPCODE_IF:
      case BRW_OPCODE_DO:
         depth++;
         break;

      case BRW_OPCODE_ENDIF:
      case BRW_OPCODE_WHILE:
         depth--;
         break;

      case SHADER_OPCODE_FIND_LIVE_CHANNEL:
         if (depth == 0) {
            inst->opcode = BRW_OPCODE_MOV;
            inst->src[0] = brw_imm_d(0);
            inst->force_writemask_all = true;
            progress = true;
         }
         break;

      default:
         break;
      }
   }

   return progress;
}

/**
 * Splits virtual GRFs requesting more than one contiguous physical register.
 *
 * We initially create large virtual GRFs for temporary structures, arrays,
 * and matrices, so that the visitor functions can add offsets to work their
 * way down to the actual member being accessed.  But when it comes to
 * optimization, we'd like to treat each register as individual storage if
 * possible.
 *
 * So far, the only thing that might prevent splitting is a send message from
 * a GRF on IVB.
 */
void
vec4_visitor::split_virtual_grfs()
{
   int num_vars = this->alloc.count;
   int new_virtual_grf[num_vars];
   bool split_grf[num_vars];

   memset(new_virtual_grf, 0, sizeof(new_virtual_grf));

   /* Try to split anything > 0 sized. */
   for (int i = 0; i < num_vars; i++) {
      split_grf[i] = this->alloc.sizes[i] != 1;
   }

   /* Check that the instructions are compatible with the registers we're trying
    * to split.
    */
   foreach_block_and_inst(block, vec4_instruction, inst, cfg) {
      if (inst->dst.file == VGRF && regs_written(inst) > 1)
         split_grf[inst->dst.nr] = false;

      for (int i = 0; i < 3; i++) {
         if (inst->src[i].file == VGRF && regs_read(inst, i) > 1)
            split_grf[inst->src[i].nr] = false;
      }
   }

   /* Allocate new space for split regs.  Note that the virtual
    * numbers will be contiguous.
    */
   for (int i = 0; i < num_vars; i++) {
      if (!split_grf[i])
         continue;

      new_virtual_grf[i] = alloc.allocate(1);
      for (unsigned j = 2; j < this->alloc.sizes[i]; j++) {
         unsigned reg = alloc.allocate(1);
         assert(reg == new_virtual_grf[i] + j - 1);
         (void) reg;
      }
      this->alloc.sizes[i] = 1;
   }

   foreach_block_and_inst(block, vec4_instruction, inst, cfg) {
      if (inst->dst.file == VGRF && split_grf[inst->dst.nr] &&
          inst->dst.offset / REG_SIZE != 0) {
         inst->dst.nr = (new_virtual_grf[inst->dst.nr] +
                         inst->dst.offset / REG_SIZE - 1);
         inst->dst.offset %= REG_SIZE;
      }
      for (int i = 0; i < 3; i++) {
         if (inst->src[i].file == VGRF && split_grf[inst->src[i].nr] &&
             inst->src[i].offset / REG_SIZE != 0) {
            inst->src[i].nr = (new_virtual_grf[inst->src[i].nr] +
                                inst->src[i].offset / REG_SIZE - 1);
            inst->src[i].offset %= REG_SIZE;
         }
      }
   }
   invalidate_live_intervals();
}

void
vec4_visitor::dump_instruction(backend_instruction *be_inst)
{
   dump_instruction(be_inst, stderr);
}

void
vec4_visitor::dump_instruction(backend_instruction *be_inst, FILE *file)
{
   vec4_instruction *inst = (vec4_instruction *)be_inst;

   if (inst->predicate) {
      fprintf(file, "(%cf%d.%d%s) ",
              inst->predicate_inverse ? '-' : '+',
              inst->flag_subreg / 2,
              inst->flag_subreg % 2,
              pred_ctrl_align16[inst->predicate]);
   }

   fprintf(file, "%s(%d)", brw_instruction_name(devinfo, inst->opcode),
           inst->exec_size);
   if (inst->saturate)
      fprintf(file, ".sat");
   if (inst->conditional_mod) {
      fprintf(file, "%s", conditional_modifier[inst->conditional_mod]);
      if (!inst->predicate &&
          (devinfo->gen < 5 || (inst->opcode != BRW_OPCODE_SEL &&
                                inst->opcode != BRW_OPCODE_CSEL &&
                                inst->opcode != BRW_OPCODE_IF &&
                                inst->opcode != BRW_OPCODE_WHILE))) {
         fprintf(file, ".f%d.%d", inst->flag_subreg / 2, inst->flag_subreg % 2);
      }
   }
   fprintf(file, " ");

   switch (inst->dst.file) {
   case VGRF:
      fprintf(file, "vgrf%d", inst->dst.nr);
      break;
   case FIXED_GRF:
      fprintf(file, "g%d", inst->dst.nr);
      break;
   case MRF:
      fprintf(file, "m%d", inst->dst.nr);
      break;
   case ARF:
      switch (inst->dst.nr) {
      case BRW_ARF_NULL:
         fprintf(file, "null");
         break;
      case BRW_ARF_ADDRESS:
         fprintf(file, "a0.%d", inst->dst.subnr);
         break;
      case BRW_ARF_ACCUMULATOR:
         fprintf(file, "acc%d", inst->dst.subnr);
         break;
      case BRW_ARF_FLAG:
         fprintf(file, "f%d.%d", inst->dst.nr & 0xf, inst->dst.subnr);
         break;
      default:
         fprintf(file, "arf%d.%d", inst->dst.nr & 0xf, inst->dst.subnr);
         break;
      }
      break;
   case BAD_FILE:
      fprintf(file, "(null)");
      break;
   case IMM:
   case ATTR:
   case UNIFORM:
      unreachable("not reached");
   }
   if (inst->dst.offset ||
       (inst->dst.file == VGRF &&
        alloc.sizes[inst->dst.nr] * REG_SIZE != inst->size_written)) {
      const unsigned reg_size = (inst->dst.file == UNIFORM ? 16 : REG_SIZE);
      fprintf(file, "+%d.%d", inst->dst.offset / reg_size,
              inst->dst.offset % reg_size);
   }
   if (inst->dst.writemask != WRITEMASK_XYZW) {
      fprintf(file, ".");
      if (inst->dst.writemask & 1)
         fprintf(file, "x");
      if (inst->dst.writemask & 2)
         fprintf(file, "y");
      if (inst->dst.writemask & 4)
         fprintf(file, "z");
      if (inst->dst.writemask & 8)
         fprintf(file, "w");
   }
   fprintf(file, ":%s", brw_reg_type_to_letters(inst->dst.type));

   if (inst->src[0].file != BAD_FILE)
      fprintf(file, ", ");

   for (int i = 0; i < 3 && inst->src[i].file != BAD_FILE; i++) {
      if (inst->src[i].negate)
         fprintf(file, "-");
      if (inst->src[i].abs)
         fprintf(file, "|");
      switch (inst->src[i].file) {
      case VGRF:
         fprintf(file, "vgrf%d", inst->src[i].nr);
         break;
      case FIXED_GRF:
         fprintf(file, "g%d.%d", inst->src[i].nr, inst->src[i].subnr);
         break;
      case ATTR:
         fprintf(file, "attr%d", inst->src[i].nr);
         break;
      case UNIFORM:
         fprintf(file, "u%d", inst->src[i].nr);
         break;
      case IMM:
         switch (inst->src[i].type) {
         case BRW_REGISTER_TYPE_F:
            fprintf(file, "%fF", inst->src[i].f);
            break;
         case BRW_REGISTER_TYPE_DF:
            fprintf(file, "%fDF", inst->src[i].df);
            break;
         case BRW_REGISTER_TYPE_D:
            fprintf(file, "%dD", inst->src[i].d);
            break;
         case BRW_REGISTER_TYPE_UD:
            fprintf(file, "%uU", inst->src[i].ud);
            break;
         case BRW_REGISTER_TYPE_VF:
            fprintf(file, "[%-gF, %-gF, %-gF, %-gF]",
                    brw_vf_to_float((inst->src[i].ud >>  0) & 0xff),
                    brw_vf_to_float((inst->src[i].ud >>  8) & 0xff),
                    brw_vf_to_float((inst->src[i].ud >> 16) & 0xff),
                    brw_vf_to_float((inst->src[i].ud >> 24) & 0xff));
            break;
         default:
            fprintf(file, "???");
            break;
         }
         break;
      case ARF:
         switch (inst->src[i].nr) {
         case BRW_ARF_NULL:
            fprintf(file, "null");
            break;
         case BRW_ARF_ADDRESS:
            fprintf(file, "a0.%d", inst->src[i].subnr);
            break;
         case BRW_ARF_ACCUMULATOR:
            fprintf(file, "acc%d", inst->src[i].subnr);
            break;
         case BRW_ARF_FLAG:
            fprintf(file, "f%d.%d", inst->src[i].nr & 0xf, inst->src[i].subnr);
            break;
         default:
            fprintf(file, "arf%d.%d", inst->src[i].nr & 0xf, inst->src[i].subnr);
            break;
         }
         break;
      case BAD_FILE:
         fprintf(file, "(null)");
         break;
      case MRF:
         unreachable("not reached");
      }

      if (inst->src[i].offset ||
          (inst->src[i].file == VGRF &&
           alloc.sizes[inst->src[i].nr] * REG_SIZE != inst->size_read(i))) {
         const unsigned reg_size = (inst->src[i].file == UNIFORM ? 16 : REG_SIZE);
         fprintf(file, "+%d.%d", inst->src[i].offset / reg_size,
                 inst->src[i].offset % reg_size);
      }

      if (inst->src[i].file != IMM) {
         static const char *chans[4] = {"x", "y", "z", "w"};
         fprintf(file, ".");
         for (int c = 0; c < 4; c++) {
            fprintf(file, "%s", chans[BRW_GET_SWZ(inst->src[i].swizzle, c)]);
         }
      }

      if (inst->src[i].abs)
         fprintf(file, "|");

      if (inst->src[i].file != IMM) {
         fprintf(file, ":%s", brw_reg_type_to_letters(inst->src[i].type));
      }

      if (i < 2 && inst->src[i + 1].file != BAD_FILE)
         fprintf(file, ", ");
   }

   if (inst->force_writemask_all)
      fprintf(file, " NoMask");

   if (inst->exec_size != 8)
      fprintf(file, " group%d", inst->group);

   fprintf(file, "\n");
}


int
vec4_vs_visitor::setup_attributes(int payload_reg)
{
   foreach_block_and_inst(block, vec4_instruction, inst, cfg) {
      for (int i = 0; i < 3; i++) {
         if (inst->src[i].file == ATTR) {
            assert(inst->src[i].offset % REG_SIZE == 0);
            int grf = payload_reg + inst->src[i].nr +
                      inst->src[i].offset / REG_SIZE;

            struct brw_reg reg = brw_vec8_grf(grf, 0);
            reg.swizzle = inst->src[i].swizzle;
            reg.type = inst->src[i].type;
            reg.abs = inst->src[i].abs;
            reg.negate = inst->src[i].negate;
            inst->src[i] = reg;
         }
      }
   }

   return payload_reg + vs_prog_data->nr_attribute_slots;
}

int
vec4_visitor::setup_uniforms(int reg)
{
   prog_data->base.dispatch_grf_start_reg = reg;

   /* The pre-gen6 VS requires that some push constants get loaded no
    * matter what, or the GPU would hang.
    */
   if (devinfo->gen < 6 && this->uniforms == 0) {
      brw_stage_prog_data_add_params(stage_prog_data, 4);
      for (unsigned int i = 0; i < 4; i++) {
	 unsigned int slot = this->uniforms * 4 + i;
	 stage_prog_data->param[slot] = BRW_PARAM_BUILTIN_ZERO;
      }

      this->uniforms++;
      reg++;
   } else {
      reg += ALIGN(uniforms, 2) / 2;
   }

   for (int i = 0; i < 4; i++)
      reg += stage_prog_data->ubo_ranges[i].length;

   stage_prog_data->nr_params = this->uniforms * 4;

   prog_data->base.curb_read_length =
      reg - prog_data->base.dispatch_grf_start_reg;

   return reg;
}

void
vec4_vs_visitor::setup_payload(void)
{
   int reg = 0;

   /* The payload always contains important data in g0, which contains
    * the URB handles that are passed on to the URB write at the end
    * of the thread.  So, we always start push constants at g1.
    */
   reg++;

   reg = setup_uniforms(reg);

   reg = setup_attributes(reg);

   this->first_non_payload_grf = reg;
}

bool
vec4_visitor::lower_minmax()
{
   assert(devinfo->gen < 6);

   bool progress = false;

   foreach_block_and_inst_safe(block, vec4_instruction, inst, cfg) {
      const vec4_builder ibld(this, block, inst);

      if (inst->opcode == BRW_OPCODE_SEL &&
          inst->predicate == BRW_PREDICATE_NONE) {
         /* FIXME: Using CMP doesn't preserve the NaN propagation semantics of
          *        the original SEL.L/GE instruction
          */
         ibld.CMP(ibld.null_reg_d(), inst->src[0], inst->src[1],
                  inst->conditional_mod);
         inst->predicate = BRW_PREDICATE_NORMAL;
         inst->conditional_mod = BRW_CONDITIONAL_NONE;

         progress = true;
      }
   }

   if (progress)
      invalidate_live_intervals();

   return progress;
}

src_reg
vec4_visitor::get_timestamp()
{
   assert(devinfo->gen >= 7);

   src_reg ts = src_reg(brw_reg(BRW_ARCHITECTURE_REGISTER_FILE,
                                BRW_ARF_TIMESTAMP,
                                0,
                                0,
                                0,
                                BRW_REGISTER_TYPE_UD,
                                BRW_VERTICAL_STRIDE_0,
                                BRW_WIDTH_4,
                                BRW_HORIZONTAL_STRIDE_4,
                                BRW_SWIZZLE_XYZW,
                                WRITEMASK_XYZW));

   dst_reg dst = dst_reg(this, glsl_type::uvec4_type);

   vec4_instruction *mov = emit(MOV(dst, ts));
   /* We want to read the 3 fields we care about (mostly field 0, but also 2)
    * even if it's not enabled in the dispatch.
    */
   mov->force_writemask_all = true;

   return src_reg(dst);
}

void
vec4_visitor::emit_shader_time_begin()
{
   current_annotation = "shader time start";
   shader_start_time = get_timestamp();
}

void
vec4_visitor::emit_shader_time_end()
{
   current_annotation = "shader time end";
   src_reg shader_end_time = get_timestamp();


   /* Check that there weren't any timestamp reset events (assuming these
    * were the only two timestamp reads that happened).
    */
   src_reg reset_end = shader_end_time;
   reset_end.swizzle = BRW_SWIZZLE_ZZZZ;
   vec4_instruction *test = emit(AND(dst_null_ud(), reset_end, brw_imm_ud(1u)));
   test->conditional_mod = BRW_CONDITIONAL_Z;

   emit(IF(BRW_PREDICATE_NORMAL));

   /* Take the current timestamp and get the delta. */
   shader_start_time.negate = true;
   dst_reg diff = dst_reg(this, glsl_type::uint_type);
   emit(ADD(diff, shader_start_time, shader_end_time));

   /* If there were no instructions between the two timestamp gets, the diff
    * is 2 cycles.  Remove that overhead, so I can forget about that when
    * trying to determine the time taken for single instructions.
    */
   emit(ADD(diff, src_reg(diff), brw_imm_ud(-2u)));

   emit_shader_time_write(0, src_reg(diff));
   emit_shader_time_write(1, brw_imm_ud(1u));
   emit(BRW_OPCODE_ELSE);
   emit_shader_time_write(2, brw_imm_ud(1u));
   emit(BRW_OPCODE_ENDIF);
}

void
vec4_visitor::emit_shader_time_write(int shader_time_subindex, src_reg value)
{
   dst_reg dst =
      dst_reg(this, glsl_type::get_array_instance(glsl_type::vec4_type, 2));

   dst_reg offset = dst;
   dst_reg time = dst;
   time.offset += REG_SIZE;

   offset.type = BRW_REGISTER_TYPE_UD;
   int index = shader_time_index * 3 + shader_time_subindex;
   emit(MOV(offset, brw_imm_d(index * BRW_SHADER_TIME_STRIDE)));

   time.type = BRW_REGISTER_TYPE_UD;
   emit(MOV(time, value));

   vec4_instruction *inst =
      emit(SHADER_OPCODE_SHADER_TIME_ADD, dst_reg(), src_reg(dst));
   inst->mlen = 2;
}

static bool
is_align1_df(vec4_instruction *inst)
{
   switch (inst->opcode) {
   case VEC4_OPCODE_DOUBLE_TO_F32:
   case VEC4_OPCODE_DOUBLE_TO_D32:
   case VEC4_OPCODE_DOUBLE_TO_U32:
   case VEC4_OPCODE_TO_DOUBLE:
   case VEC4_OPCODE_PICK_LOW_32BIT:
   case VEC4_OPCODE_PICK_HIGH_32BIT:
   case VEC4_OPCODE_SET_LOW_32BIT:
   case VEC4_OPCODE_SET_HIGH_32BIT:
      return true;
   default:
      return false;
   }
}

/**
 * Three source instruction must have a GRF/MRF destination register.
 * ARF NULL is not allowed.  Fix that up by allocating a temporary GRF.
 */
void
vec4_visitor::fixup_3src_null_dest()
{
   bool progress = false;

   foreach_block_and_inst_safe (block, vec4_instruction, inst, cfg) {
      if (inst->is_3src(devinfo) && inst->dst.is_null()) {
         const unsigned size_written = type_sz(inst->dst.type);
         const unsigned num_regs = DIV_ROUND_UP(size_written, REG_SIZE);

         inst->dst = retype(dst_reg(VGRF, alloc.allocate(num_regs)),
                            inst->dst.type);
         progress = true;
      }
   }

   if (progress)
      invalidate_live_intervals();
}

void
vec4_visitor::convert_to_hw_regs()
{
   foreach_block_and_inst(block, vec4_instruction, inst, cfg) {
      for (int i = 0; i < 3; i++) {
         class src_reg &src = inst->src[i];
         struct brw_reg reg;
         switch (src.file) {
         case VGRF: {
            reg = byte_offset(brw_vecn_grf(4, src.nr, 0), src.offset);
            reg.type = src.type;
            reg.abs = src.abs;
            reg.negate = src.negate;
            break;
         }

         case UNIFORM: {
            reg = stride(byte_offset(brw_vec4_grf(
                                        prog_data->base.dispatch_grf_start_reg +
                                        src.nr / 2, src.nr % 2 * 4),
                                     src.offset),
                         0, 4, 1);
            reg.type = src.type;
            reg.abs = src.abs;
            reg.negate = src.negate;

            /* This should have been moved to pull constants. */
            assert(!src.reladdr);
            break;
         }

         case FIXED_GRF:
            if (type_sz(src.type) == 8) {
               reg = src.as_brw_reg();
               break;
            }
            /* fallthrough */
         case ARF:
         case IMM:
            continue;

         case BAD_FILE:
            /* Probably unused. */
            reg = brw_null_reg();
            reg = retype(reg, src.type);
            break;

         case MRF:
         case ATTR:
            unreachable("not reached");
         }

         apply_logical_swizzle(&reg, inst, i);
         src = reg;

         /* From IVB PRM, vol4, part3, "General Restrictions on Regioning
          * Parameters":
          *
          *   "If ExecSize = Width and HorzStride ≠ 0, VertStride must be set
          *    to Width * HorzStride."
          *
          * We can break this rule with DF sources on DF align1
          * instructions, because the exec_size would be 4 and width is 4.
          * As we know we are not accessing to next GRF, it is safe to
          * set vstride to the formula given by the rule itself.
          */
         if (is_align1_df(inst) && (cvt(inst->exec_size) - 1) == src.width)
            src.vstride = src.width + src.hstride;
      }

      if (inst->is_3src(devinfo)) {
         /* 3-src instructions with scalar sources support arbitrary subnr,
          * but don't actually use swizzles.  Convert swizzle into subnr.
          * Skip this for double-precision instructions: RepCtrl=1 is not
          * allowed for them and needs special handling.
          */
         for (int i = 0; i < 3; i++) {
            if (inst->src[i].vstride == BRW_VERTICAL_STRIDE_0 &&
                type_sz(inst->src[i].type) < 8) {
               assert(brw_is_single_value_swizzle(inst->src[i].swizzle));
               inst->src[i].subnr += 4 * BRW_GET_SWZ(inst->src[i].swizzle, 0);
            }
         }
      }

      dst_reg &dst = inst->dst;
      struct brw_reg reg;

      switch (inst->dst.file) {
      case VGRF:
         reg = byte_offset(brw_vec8_grf(dst.nr, 0), dst.offset);
         reg.type = dst.type;
         reg.writemask = dst.writemask;
         break;

      case MRF:
         reg = byte_offset(brw_message_reg(dst.nr), dst.offset);
         assert((reg.nr & ~BRW_MRF_COMPR4) < BRW_MAX_MRF(devinfo->gen));
         reg.type = dst.type;
         reg.writemask = dst.writemask;
         break;

      case ARF:
      case FIXED_GRF:
         reg = dst.as_brw_reg();
         break;

      case BAD_FILE:
         reg = brw_null_reg();
         reg = retype(reg, dst.type);
         break;

      case IMM:
      case ATTR:
      case UNIFORM:
         unreachable("not reached");
      }

      dst = reg;
   }
}

static bool
stage_uses_interleaved_attributes(unsigned stage,
                                  enum shader_dispatch_mode dispatch_mode)
{
   switch (stage) {
   case MESA_SHADER_TESS_EVAL:
      return true;
   case MESA_SHADER_GEOMETRY:
      return dispatch_mode != DISPATCH_MODE_4X2_DUAL_OBJECT;
   default:
      return false;
   }
}

/**
 * Get the closest native SIMD width supported by the hardware for instruction
 * \p inst.  The instruction will be left untouched by
 * vec4_visitor::lower_simd_width() if the returned value matches the
 * instruction's original execution size.
 */
static unsigned
get_lowered_simd_width(const struct gen_device_info *devinfo,
                       enum shader_dispatch_mode dispatch_mode,
                       unsigned stage, const vec4_instruction *inst)
{
   /* Do not split some instructions that require special handling */
   switch (inst->opcode) {
   case SHADER_OPCODE_GEN4_SCRATCH_READ:
   case SHADER_OPCODE_GEN4_SCRATCH_WRITE:
      return inst->exec_size;
   default:
      break;
   }

   unsigned lowered_width = MIN2(16, inst->exec_size);

   /* We need to split some cases of double-precision instructions that write
    * 2 registers. We only need to care about this in gen7 because that is the
    * only hardware that implements fp64 in Align16.
    */
   if (devinfo->gen == 7 && inst->size_written > REG_SIZE) {
      /* Align16 8-wide double-precision SEL does not work well. Verified
       * empirically.
       */
      if (inst->opcode == BRW_OPCODE_SEL && type_sz(inst->dst.type) == 8)
         lowered_width = MIN2(lowered_width, 4);

      /* HSW PRM, 3D Media GPGPU Engine, Region Alignment Rules for Direct
       * Register Addressing:
       *
       *    "When destination spans two registers, the source MUST span two
       *     registers."
       */
      for (unsigned i = 0; i < 3; i++) {
         if (inst->src[i].file == BAD_FILE)
            continue;
         if (inst->size_read(i) <= REG_SIZE)
            lowered_width = MIN2(lowered_width, 4);

         /* Interleaved attribute setups use a vertical stride of 0, which
          * makes them hit the associated instruction decompression bug in gen7.
          * Split them to prevent this.
          */
         if (inst->src[i].file == ATTR &&
             stage_uses_interleaved_attributes(stage, dispatch_mode))
            lowered_width = MIN2(lowered_width, 4);
      }
   }

   /* IvyBridge can manage a maximum of 4 DFs per SIMD4x2 instruction, since
    * it doesn't support compression in Align16 mode, no matter if it has
    * force_writemask_all enabled or disabled (the latter is affected by the
    * compressed instruction bug in gen7, which is another reason to enforce
    * this limit).
    */
   if (devinfo->gen == 7 && !devinfo->is_haswell &&
       (get_exec_type_size(inst) == 8 || type_sz(inst->dst.type) == 8))
      lowered_width = MIN2(lowered_width, 4);

   return lowered_width;
}

static bool
dst_src_regions_overlap(vec4_instruction *inst)
{
   if (inst->size_written == 0)
      return false;

   unsigned dst_start = inst->dst.offset;
   unsigned dst_end = dst_start + inst->size_written - 1;
   for (int i = 0; i < 3; i++) {
      if (inst->src[i].file == BAD_FILE)
         continue;

      if (inst->dst.file != inst->src[i].file ||
          inst->dst.nr != inst->src[i].nr)
         continue;

      unsigned src_start = inst->src[i].offset;
      unsigned src_end = src_start + inst->size_read(i) - 1;

      if ((dst_start >= src_start && dst_start <= src_end) ||
          (dst_end >= src_start && dst_end <= src_end) ||
          (dst_start <= src_start && dst_end >= src_end)) {
         return true;
      }
   }

   return false;
}

bool
vec4_visitor::lower_simd_width()
{
   bool progress = false;

   foreach_block_and_inst_safe(block, vec4_instruction, inst, cfg) {
      const unsigned lowered_width =
         get_lowered_simd_width(devinfo, prog_data->dispatch_mode, stage, inst);
      assert(lowered_width <= inst->exec_size);
      if (lowered_width == inst->exec_size)
         continue;

      /* We need to deal with source / destination overlaps when splitting.
       * The hardware supports reading from and writing to the same register
       * in the same instruction, but we need to be careful that each split
       * instruction we produce does not corrupt the source of the next.
       *
       * The easiest way to handle this is to make the split instructions write
       * to temporaries if there is an src/dst overlap and then move from the
       * temporaries to the original destination. We also need to consider
       * instructions that do partial writes via align1 opcodes, in which case
       * we need to make sure that the we initialize the temporary with the
       * value of the instruction's dst.
       */
      bool needs_temp = dst_src_regions_overlap(inst);
      for (unsigned n = 0; n < inst->exec_size / lowered_width; n++)  {
         unsigned channel_offset = lowered_width * n;

         unsigned size_written = lowered_width * type_sz(inst->dst.type);

         /* Create the split instruction from the original so that we copy all
          * relevant instruction fields, then set the width and calculate the
          * new dst/src regions.
          */
         vec4_instruction *linst = new(mem_ctx) vec4_instruction(*inst);
         linst->exec_size = lowered_width;
         linst->group = channel_offset;
         linst->size_written = size_written;

         /* Compute split dst region */
         dst_reg dst;
         if (needs_temp) {
            unsigned num_regs = DIV_ROUND_UP(size_written, REG_SIZE);
            dst = retype(dst_reg(VGRF, alloc.allocate(num_regs)),
                         inst->dst.type);
            if (inst->is_align1_partial_write()) {
               vec4_instruction *copy = MOV(dst, src_reg(inst->dst));
               copy->exec_size = lowered_width;
               copy->group = channel_offset;
               copy->size_written = size_written;
               inst->insert_before(block, copy);
            }
         } else {
            dst = horiz_offset(inst->dst, channel_offset);
         }
         linst->dst = dst;

         /* Compute split source regions */
         for (int i = 0; i < 3; i++) {
            if (linst->src[i].file == BAD_FILE)
               continue;

            bool is_interleaved_attr =
               linst->src[i].file == ATTR &&
               stage_uses_interleaved_attributes(stage,
                                                 prog_data->dispatch_mode);

            if (!is_uniform(linst->src[i]) && !is_interleaved_attr)
               linst->src[i] = horiz_offset(linst->src[i], channel_offset);
         }

         inst->insert_before(block, linst);

         /* If we used a temporary to store the result of the split
          * instruction, copy the result to the original destination
          */
         if (needs_temp) {
            vec4_instruction *mov =
               MOV(offset(inst->dst, lowered_width, n), src_reg(dst));
            mov->exec_size = lowered_width;
            mov->group = channel_offset;
            mov->size_written = size_written;
            mov->predicate = inst->predicate;
            inst->insert_before(block, mov);
         }
      }

      inst->remove(block);
      progress = true;
   }

   if (progress)
      invalidate_live_intervals();

   return progress;
}

static brw_predicate
scalarize_predicate(brw_predicate predicate, unsigned writemask)
{
   if (predicate != BRW_PREDICATE_NORMAL)
      return predicate;

   switch (writemask) {
   case WRITEMASK_X:
      return BRW_PREDICATE_ALIGN16_REPLICATE_X;
   case WRITEMASK_Y:
      return BRW_PREDICATE_ALIGN16_REPLICATE_Y;
   case WRITEMASK_Z:
      return BRW_PREDICATE_ALIGN16_REPLICATE_Z;
   case WRITEMASK_W:
      return BRW_PREDICATE_ALIGN16_REPLICATE_W;
   default:
      unreachable("invalid writemask");
   }
}

/* Gen7 has a hardware decompression bug that we can exploit to represent
 * handful of additional swizzles natively.
 */
static bool
is_gen7_supported_64bit_swizzle(vec4_instruction *inst, unsigned arg)
{
   switch (inst->src[arg].swizzle) {
   case BRW_SWIZZLE_XXXX:
   case BRW_SWIZZLE_YYYY:
   case BRW_SWIZZLE_ZZZZ:
   case BRW_SWIZZLE_WWWW:
   case BRW_SWIZZLE_XYXY:
   case BRW_SWIZZLE_YXYX:
   case BRW_SWIZZLE_ZWZW:
   case BRW_SWIZZLE_WZWZ:
      return true;
   default:
      return false;
   }
}

/* 64-bit sources use regions with a width of 2. These 2 elements in each row
 * can be addressed using 32-bit swizzles (which is what the hardware supports)
 * but it also means that the swizzle we apply on the first two components of a
 * dvec4 is coupled with the swizzle we use for the last 2. In other words,
 * only some specific swizzle combinations can be natively supported.
 *
 * FIXME: we can go an step further and implement even more swizzle
 *        variations using only partial scalarization.
 *
 * For more details see:
 * https://bugs.freedesktop.org/show_bug.cgi?id=92760#c82
 */
bool
vec4_visitor::is_supported_64bit_region(vec4_instruction *inst, unsigned arg)
{
   const src_reg &src = inst->src[arg];
   assert(type_sz(src.type) == 8);

   /* Uniform regions have a vstride=0. Because we use 2-wide rows with
    * 64-bit regions it means that we cannot access components Z/W, so
    * return false for any such case. Interleaved attributes will also be
    * mapped to GRF registers with a vstride of 0, so apply the same
    * treatment.
    */
   if ((is_uniform(src) ||
        (stage_uses_interleaved_attributes(stage, prog_data->dispatch_mode) &&
         src.file == ATTR)) &&
       (brw_mask_for_swizzle(src.swizzle) & 12))
      return false;

   switch (src.swizzle) {
   case BRW_SWIZZLE_XYZW:
   case BRW_SWIZZLE_XXZZ:
   case BRW_SWIZZLE_YYWW:
   case BRW_SWIZZLE_YXWZ:
      return true;
   default:
      return devinfo->gen == 7 && is_gen7_supported_64bit_swizzle(inst, arg);
   }
}

bool
vec4_visitor::scalarize_df()
{
   bool progress = false;

   foreach_block_and_inst_safe(block, vec4_instruction, inst, cfg) {
      /* Skip DF instructions that operate in Align1 mode */
      if (is_align1_df(inst))
         continue;

      /* Check if this is a double-precision instruction */
      bool is_double = type_sz(inst->dst.type) == 8;
      for (int arg = 0; !is_double && arg < 3; arg++) {
         is_double = inst->src[arg].file != BAD_FILE &&
                     type_sz(inst->src[arg].type) == 8;
      }

      if (!is_double)
         continue;

      /* Skip the lowering for specific regioning scenarios that we can
       * support natively.
       */
      bool skip_lowering = true;

      /* XY and ZW writemasks operate in 32-bit, which means that they don't
       * have a native 64-bit representation and they should always be split.
       */
      if (inst->dst.writemask == WRITEMASK_XY ||
          inst->dst.writemask == WRITEMASK_ZW) {
         skip_lowering = false;
      } else {
         for (unsigned i = 0; i < 3; i++) {
            if (inst->src[i].file == BAD_FILE || type_sz(inst->src[i].type) < 8)
               continue;
            skip_lowering = skip_lowering && is_supported_64bit_region(inst, i);
         }
      }

      if (skip_lowering)
         continue;

      /* Generate scalar instructions for each enabled channel */
      for (unsigned chan = 0; chan < 4; chan++) {
         unsigned chan_mask = 1 << chan;
         if (!(inst->dst.writemask & chan_mask))
            continue;

         vec4_instruction *scalar_inst = new(mem_ctx) vec4_instruction(*inst);

         for (unsigned i = 0; i < 3; i++) {
            unsigned swz = BRW_GET_SWZ(inst->src[i].swizzle, chan);
            scalar_inst->src[i].swizzle = BRW_SWIZZLE4(swz, swz, swz, swz);
         }

         scalar_inst->dst.writemask = chan_mask;

         if (inst->predicate != BRW_PREDICATE_NONE) {
            scalar_inst->predicate =
               scalarize_predicate(inst->predicate, chan_mask);
         }

         inst->insert_before(block, scalar_inst);
      }

      inst->remove(block);
      progress = true;
   }

   if (progress)
      invalidate_live_intervals();

   return progress;
}

bool
vec4_visitor::lower_64bit_mad_to_mul_add()
{
   bool progress = false;

   foreach_block_and_inst_safe(block, vec4_instruction, inst, cfg) {
      if (inst->opcode != BRW_OPCODE_MAD)
         continue;

      if (type_sz(inst->dst.type) != 8)
         continue;

      dst_reg mul_dst = dst_reg(this, glsl_type::dvec4_type);

      /* Use the copy constructor so we copy all relevant instruction fields
       * from the original mad into the add and mul instructions
       */
      vec4_instruction *mul = new(mem_ctx) vec4_instruction(*inst);
      mul->opcode = BRW_OPCODE_MUL;
      mul->dst = mul_dst;
      mul->src[0] = inst->src[1];
      mul->src[1] = inst->src[2];
      mul->src[2].file = BAD_FILE;

      vec4_instruction *add = new(mem_ctx) vec4_instruction(*inst);
      add->opcode = BRW_OPCODE_ADD;
      add->src[0] = src_reg(mul_dst);
      add->src[1] = inst->src[0];
      add->src[2].file = BAD_FILE;

      inst->insert_before(block, mul);
      inst->insert_before(block, add);
      inst->remove(block);

      progress = true;
   }

   if (progress)
      invalidate_live_intervals();

   return progress;
}

/* The align16 hardware can only do 32-bit swizzle channels, so we need to
 * translate the logical 64-bit swizzle channels that we use in the Vec4 IR
 * to 32-bit swizzle channels in hardware registers.
 *
 * @inst and @arg identify the original vec4 IR source operand we need to
 * translate the swizzle for and @hw_reg is the hardware register where we
 * will write the hardware swizzle to use.
 *
 * This pass assumes that Align16/DF instructions have been fully scalarized
 * previously so there is just one 64-bit swizzle channel to deal with for any
 * given Vec4 IR source.
 */
void
vec4_visitor::apply_logical_swizzle(struct brw_reg *hw_reg,
                                    vec4_instruction *inst, int arg)
{
   src_reg reg = inst->src[arg];

   if (reg.file == BAD_FILE || reg.file == BRW_IMMEDIATE_VALUE)
      return;

   /* If this is not a 64-bit operand or this is a scalar instruction we don't
    * need to do anything about the swizzles.
    */
   if(type_sz(reg.type) < 8 || is_align1_df(inst)) {
      hw_reg->swizzle = reg.swizzle;
      return;
   }

   /* Take the 64-bit logical swizzle channel and translate it to 32-bit */
   assert(brw_is_single_value_swizzle(reg.swizzle) ||
          is_supported_64bit_region(inst, arg));

   /* Apply the region <2, 2, 1> for GRF or <0, 2, 1> for uniforms, as align16
    * HW can only do 32-bit swizzle channels.
    */
   hw_reg->width = BRW_WIDTH_2;

   if (is_supported_64bit_region(inst, arg) &&
       !is_gen7_supported_64bit_swizzle(inst, arg)) {
      /* Supported 64-bit swizzles are those such that their first two
       * components, when expanded to 32-bit swizzles, match the semantics
       * of the original 64-bit swizzle with 2-wide row regioning.
       */
      unsigned swizzle0 = BRW_GET_SWZ(reg.swizzle, 0);
      unsigned swizzle1 = BRW_GET_SWZ(reg.swizzle, 1);
      hw_reg->swizzle = BRW_SWIZZLE4(swizzle0 * 2, swizzle0 * 2 + 1,
                                     swizzle1 * 2, swizzle1 * 2 + 1);
   } else {
      /* If we got here then we have one of the following:
       *
       * 1. An unsupported swizzle, which should be single-value thanks to the
       *    scalarization pass.
       *
       * 2. A gen7 supported swizzle. These can be single-value or double-value
       *    swizzles. If the latter, they are never cross-dvec2 channels. For
       *    these we always need to activate the gen7 vstride=0 exploit.
       */
      unsigned swizzle0 = BRW_GET_SWZ(reg.swizzle, 0);
      unsigned swizzle1 = BRW_GET_SWZ(reg.swizzle, 1);
      assert((swizzle0 < 2) == (swizzle1 < 2));

      /* To gain access to Z/W components we need to select the second half
       * of the register and then use a X/Y swizzle to select Z/W respectively.
       */
      if (swizzle0 >= 2) {
         *hw_reg = suboffset(*hw_reg, 2);
         swizzle0 -= 2;
         swizzle1 -= 2;
      }

      /* All gen7-specific supported swizzles require the vstride=0 exploit */
      if (devinfo->gen == 7 && is_gen7_supported_64bit_swizzle(inst, arg))
         hw_reg->vstride = BRW_VERTICAL_STRIDE_0;

      /* Any 64-bit source with an offset at 16B is intended to address the
       * second half of a register and needs a vertical stride of 0 so we:
       *
       * 1. Don't violate register region restrictions.
       * 2. Activate the gen7 instruction decompresion bug exploit when
       *    execsize > 4
       */
      if (hw_reg->subnr % REG_SIZE == 16) {
         assert(devinfo->gen == 7);
         hw_reg->vstride = BRW_VERTICAL_STRIDE_0;
      }

      hw_reg->swizzle = BRW_SWIZZLE4(swizzle0 * 2, swizzle0 * 2 + 1,
                                     swizzle1 * 2, swizzle1 * 2 + 1);
   }
}

bool
vec4_visitor::run()
{
   if (shader_time_index >= 0)
      emit_shader_time_begin();

   emit_prolog();

   emit_nir_code();
   if (failed)
      return false;
   base_ir = NULL;

   emit_thread_end();

   calculate_cfg();

   /* Before any optimization, push array accesses out to scratch
    * space where we need them to be.  This pass may allocate new
    * virtual GRFs, so we want to do it early.  It also makes sure
    * that we have reladdr computations available for CSE, since we'll
    * often do repeated subexpressions for those.
    */
   move_grf_array_access_to_scratch();
   move_uniform_array_access_to_pull_constants();

   pack_uniform_registers();
   move_push_constants_to_pull_constants();
   split_virtual_grfs();

#define OPT(pass, args...) ({                                          \
      pass_num++;                                                      \
      bool this_progress = pass(args);                                 \
                                                                       \
      if (unlikely(INTEL_DEBUG & DEBUG_OPTIMIZER) && this_progress) {  \
         char filename[64];                                            \
         snprintf(filename, 64, "%s-%s-%02d-%02d-" #pass,              \
                  stage_abbrev, nir->info.name, iteration, pass_num); \
                                                                       \
         backend_shader::dump_instructions(filename);                  \
      }                                                                \
                                                                       \
      progress = progress || this_progress;                            \
      this_progress;                                                   \
   })


   if (unlikely(INTEL_DEBUG & DEBUG_OPTIMIZER)) {
      char filename[64];
      snprintf(filename, 64, "%s-%s-00-00-start",
               stage_abbrev, nir->info.name);

      backend_shader::dump_instructions(filename);
   }

   bool progress;
   int iteration = 0;
   int pass_num = 0;
   do {
      progress = false;
      pass_num = 0;
      iteration++;

      OPT(opt_predicated_break, this);
      OPT(opt_reduce_swizzle);
      OPT(dead_code_eliminate);
      OPT(dead_control_flow_eliminate, this);
      OPT(opt_copy_propagation);
      OPT(opt_cmod_propagation);
      OPT(opt_cse);
      OPT(opt_algebraic);
      OPT(opt_register_coalesce);
      OPT(eliminate_find_live_channel);
   } while (progress);

   pass_num = 0;

   if (OPT(opt_vector_float)) {
      OPT(opt_cse);
      OPT(opt_copy_propagation, false);
      OPT(opt_copy_propagation, true);
      OPT(dead_code_eliminate);
   }

   if (devinfo->gen <= 5 && OPT(lower_minmax)) {
      OPT(opt_cmod_propagation);
      OPT(opt_cse);
      OPT(opt_copy_propagation);
      OPT(dead_code_eliminate);
   }

   if (OPT(lower_simd_width)) {
      OPT(opt_copy_propagation);
      OPT(dead_code_eliminate);
   }

   if (failed)
      return false;

   OPT(lower_64bit_mad_to_mul_add);

   /* Run this before payload setup because tesselation shaders
    * rely on it to prevent cross dvec2 regioning on DF attributes
    * that are setup so that XY are on the second half of register and
    * ZW are in the first half of the next.
    */
   OPT(scalarize_df);

   setup_payload();

   if (unlikely(INTEL_DEBUG & DEBUG_SPILL_VEC4)) {
      /* Debug of register spilling: Go spill everything. */
      const int grf_count = alloc.count;
      float spill_costs[alloc.count];
      bool no_spill[alloc.count];
      evaluate_spill_costs(spill_costs, no_spill);
      for (int i = 0; i < grf_count; i++) {
         if (no_spill[i])
            continue;
         spill_reg(i);
      }

      /* We want to run this after spilling because 64-bit (un)spills need to
       * emit code to shuffle 64-bit data for the 32-bit scratch read/write
       * messages that can produce unsupported 64-bit swizzle regions.
       */
      OPT(scalarize_df);
   }

   fixup_3src_null_dest();

   bool allocated_without_spills = reg_allocate();

   if (!allocated_without_spills) {
      compiler->shader_perf_log(log_data,
                                "%s shader triggered register spilling.  "
                                "Try reducing the number of live vec4 values "
                                "to improve performance.\n",
                                stage_name);

      while (!reg_allocate()) {
         if (failed)
            return false;
      }

      /* We want to run this after spilling because 64-bit (un)spills need to
       * emit code to shuffle 64-bit data for the 32-bit scratch read/write
       * messages that can produce unsupported 64-bit swizzle regions.
       */
      OPT(scalarize_df);
   }

   opt_schedule_instructions();

   opt_set_dependency_control();

   convert_to_hw_regs();

   if (last_scratch > 0) {
      prog_data->base.total_scratch =
         brw_get_scratch_size(last_scratch * REG_SIZE);
   }

   return !failed;
}

} /* namespace brw */

extern "C" {

/**
 * Compile a vertex shader.
 *
 * Returns the final assembly and the program's size.
 */
const unsigned *
brw_compile_vs(const struct brw_compiler *compiler, void *log_data,
               void *mem_ctx,
               const struct brw_vs_prog_key *key,
               struct brw_vs_prog_data *prog_data,
               nir_shader *shader,
               int shader_time_index,
               struct brw_compile_stats *stats,
               char **error_str)
{
   const bool is_scalar = compiler->scalar_stage[MESA_SHADER_VERTEX];
   brw_nir_apply_key(shader, compiler, &key->base, 8, is_scalar);

   const unsigned *assembly = NULL;

   if (prog_data->base.vue_map.varying_to_slot[VARYING_SLOT_EDGE] != -1) {
      /* If the output VUE map contains VARYING_SLOT_EDGE then we need to copy
       * the edge flag from VERT_ATTRIB_EDGEFLAG.  This will be done
       * automatically by brw_vec4_visitor::emit_urb_slot but we need to
       * ensure that prog_data->inputs_read is accurate.
       *
       * In order to make late NIR passes aware of the change, we actually
       * whack shader->info.inputs_read instead.  This is safe because we just
       * made a copy of the shader.
       */
      assert(!is_scalar);
      assert(key->copy_edgeflag);
      shader->info.inputs_read |= VERT_BIT_EDGEFLAG;
   }

   prog_data->inputs_read = shader->info.inputs_read;
   prog_data->double_inputs_read = shader->info.vs.double_inputs;

   brw_nir_lower_vs_inputs(shader, key->gl_attrib_wa_flags);
   brw_nir_lower_vue_outputs(shader);
   brw_postprocess_nir(shader, compiler, is_scalar);

   prog_data->base.clip_distance_mask =
      ((1 << shader->info.clip_distance_array_size) - 1);
   prog_data->base.cull_distance_mask =
      ((1 << shader->info.cull_distance_array_size) - 1) <<
      shader->info.clip_distance_array_size;

   unsigned nr_attribute_slots = util_bitcount64(prog_data->inputs_read);

   /* gl_VertexID and gl_InstanceID are system values, but arrive via an
    * incoming vertex attribute.  So, add an extra slot.
    */
   if (shader->info.system_values_read &
       (BITFIELD64_BIT(SYSTEM_VALUE_FIRST_VERTEX) |
        BITFIELD64_BIT(SYSTEM_VALUE_BASE_INSTANCE) |
        BITFIELD64_BIT(SYSTEM_VALUE_VERTEX_ID_ZERO_BASE) |
        BITFIELD64_BIT(SYSTEM_VALUE_INSTANCE_ID))) {
      nr_attribute_slots++;
   }

   /* gl_DrawID and IsIndexedDraw share its very own vec4 */
   if (shader->info.system_values_read &
       (BITFIELD64_BIT(SYSTEM_VALUE_DRAW_ID) |
        BITFIELD64_BIT(SYSTEM_VALUE_IS_INDEXED_DRAW))) {
      nr_attribute_slots++;
   }

   if (shader->info.system_values_read &
       BITFIELD64_BIT(SYSTEM_VALUE_IS_INDEXED_DRAW))
      prog_data->uses_is_indexed_draw = true;

   if (shader->info.system_values_read &
       BITFIELD64_BIT(SYSTEM_VALUE_FIRST_VERTEX))
      prog_data->uses_firstvertex = true;

   if (shader->info.system_values_read &
       BITFIELD64_BIT(SYSTEM_VALUE_BASE_INSTANCE))
      prog_data->uses_baseinstance = true;

   if (shader->info.system_values_read &
       BITFIELD64_BIT(SYSTEM_VALUE_VERTEX_ID_ZERO_BASE))
      prog_data->uses_vertexid = true;

   if (shader->info.system_values_read &
       BITFIELD64_BIT(SYSTEM_VALUE_INSTANCE_ID))
      prog_data->uses_instanceid = true;

   if (shader->info.system_values_read &
       BITFIELD64_BIT(SYSTEM_VALUE_DRAW_ID))
          prog_data->uses_drawid = true;

   /* The 3DSTATE_VS documentation lists the lower bound on "Vertex URB Entry
    * Read Length" as 1 in vec4 mode, and 0 in SIMD8 mode.  Empirically, in
    * vec4 mode, the hardware appears to wedge unless we read something.
    */
   if (is_scalar)
      prog_data->base.urb_read_length =
         DIV_ROUND_UP(nr_attribute_slots, 2);
   else
      prog_data->base.urb_read_length =
         DIV_ROUND_UP(MAX2(nr_attribute_slots, 1), 2);

   prog_data->nr_attribute_slots = nr_attribute_slots;

   /* Since vertex shaders reuse the same VUE entry for inputs and outputs
    * (overwriting the original contents), we need to make sure the size is
    * the larger of the two.
    */
   const unsigned vue_entries =
      MAX2(nr_attribute_slots, (unsigned)prog_data->base.vue_map.num_slots);

   if (compiler->devinfo->gen == 6) {
      prog_data->base.urb_entry_size = DIV_ROUND_UP(vue_entries, 8);
   } else {
      prog_data->base.urb_entry_size = DIV_ROUND_UP(vue_entries, 4);
      /* On Cannonlake software shall not program an allocation size that
       * specifies a size that is a multiple of 3 64B (512-bit) cachelines.
       */
      if (compiler->devinfo->gen == 10 &&
          prog_data->base.urb_entry_size % 3 == 0)
         prog_data->base.urb_entry_size++;
   }

   if (INTEL_DEBUG & DEBUG_VS) {
      fprintf(stderr, "VS Output ");
      brw_print_vue_map(stderr, &prog_data->base.vue_map);
   }

   if (is_scalar) {
      prog_data->base.dispatch_mode = DISPATCH_MODE_SIMD8;

      fs_visitor v(compiler, log_data, mem_ctx, &key->base,
                   &prog_data->base.base,
                   shader, 8, shader_time_index);
      if (!v.run_vs()) {
         if (error_str)
            *error_str = ralloc_strdup(mem_ctx, v.fail_msg);

         return NULL;
      }

      prog_data->base.base.dispatch_grf_start_reg = v.payload.num_regs;

      fs_generator g(compiler, log_data, mem_ctx,
                     &prog_data->base.base, v.shader_stats,
                     v.runtime_check_aads_emit, MESA_SHADER_VERTEX);
      if (INTEL_DEBUG & DEBUG_VS) {
         const char *debug_name =
            ralloc_asprintf(mem_ctx, "%s vertex shader %s",
                            shader->info.label ? shader->info.label :
                               "unnamed",
                            shader->info.name);

         g.enable_debug(debug_name);
      }
      g.generate_code(v.cfg, 8, stats);
      assembly = g.get_assembly();
   }

   if (!assembly) {
      prog_data->base.dispatch_mode = DISPATCH_MODE_4X2_DUAL_OBJECT;

      vec4_vs_visitor v(compiler, log_data, key, prog_data,
                        shader, mem_ctx, shader_time_index);
      if (!v.run()) {
         if (error_str)
            *error_str = ralloc_strdup(mem_ctx, v.fail_msg);

         return NULL;
      }

      assembly = brw_vec4_generate_assembly(compiler, log_data, mem_ctx,
                                            shader, &prog_data->base,
                                            v.cfg, stats);
   }

   return assembly;
}

} /* extern "C" */