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
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532
3533
3534
3535
3536
3537
3538
3539
3540
3541
3542
3543
3544
3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
3556
3557
3558
3559
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
3570
3571
3572
3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586
3587
3588
3589
3590
3591
3592
3593
3594
3595
3596
3597
3598
3599
3600
3601
3602
3603
3604
3605
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
3617
3618
3619
3620
3621
3622
3623
3624
3625
3626
3627
3628
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
3644
3645
3646
3647
3648
3649
3650
3651
3652
3653
3654
3655
3656
3657
3658
3659
3660
3661
3662
3663
3664
3665
3666
3667
3668
3669
3670
3671
3672
3673
3674
3675
3676
3677
3678
3679
3680
3681
3682
3683
3684
3685
3686
3687
3688
3689
3690
3691
3692
3693
3694
3695
3696
3697
3698
3699
3700
3701
3702
3703
3704
3705
3706
3707
3708
3709
3710
3711
3712
3713
3714
3715
3716
3717
3718
3719
3720
3721
3722
3723
3724
3725
3726
3727
3728
3729
3730
3731
3732
3733
3734
3735
3736
3737
3738
3739
3740
3741
3742
3743
3744
3745
3746
3747
3748
3749
3750
3751
3752
3753
3754
3755
3756
3757
3758
3759
3760
3761
3762
3763
3764
3765
3766
3767
3768
3769
3770
3771
3772
3773
3774
3775
3776
3777
3778
3779
3780
3781
3782
3783
3784
3785
3786
3787
3788
3789
3790
3791
3792
3793
3794
3795
3796
3797
3798
3799
3800
3801
3802
3803
3804
3805
3806
3807
3808
3809
3810
3811
3812
3813
3814
3815
3816
3817
3818
3819
3820
3821
3822
3823
3824
3825
3826
3827
3828
3829
3830
3831
3832
3833
3834
3835
3836
3837
3838
3839
3840
3841
3842
3843
3844
3845
3846
3847
3848
3849
3850
3851
3852
3853
3854
3855
3856
3857
3858
3859
3860
3861
3862
3863
3864
3865
3866
3867
3868
3869
3870
3871
3872
3873
3874
3875
3876
3877
3878
3879
3880
3881
3882
3883
3884
3885
3886
3887
3888
3889
3890
3891
3892
3893
3894
3895
3896
3897
3898
3899
3900
3901
3902
3903
3904
3905
3906
3907
3908
3909
3910
3911
3912
3913
3914
3915
3916
3917
3918
3919
3920
3921
3922
3923
3924
3925
3926
3927
3928
3929
3930
3931
3932
3933
3934
3935
3936
3937
3938
3939
3940
3941
3942
3943
3944
3945
3946
3947
3948
3949
3950
3951
3952
3953
3954
3955
3956
3957
3958
3959
3960
3961
3962
3963
3964
3965
3966
3967
3968
3969
3970
3971
3972
3973
3974
3975
3976
3977
3978
3979
3980
3981
3982
|
//===-- SIInstructions.td - SI Instruction Definitions --------------------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
// This file was originally auto-generated from a GPU register header file and
// all the instruction definitions were originally commented out. Instructions
// that are not yet supported remain commented out.
//===----------------------------------------------------------------------===//
class GCNPat<dag pattern, dag result> : Pat<pattern, result>, GCNPredicateControl {
}
class UniformSextInreg<ValueType VT> : PatFrag<
(ops node:$src),
(sext_inreg $src, VT),
[{ return !N->isDivergent(); }]>;
class DivergentSextInreg<ValueType VT> : PatFrag<
(ops node:$src),
(sext_inreg $src, VT),
[{ return N->isDivergent(); }]>;
include "SOPInstructions.td"
include "VOPInstructions.td"
include "SMInstructions.td"
include "FLATInstructions.td"
include "BUFInstructions.td"
include "EXPInstructions.td"
include "DSDIRInstructions.td"
include "VINTERPInstructions.td"
//===----------------------------------------------------------------------===//
// VINTRP Instructions
//===----------------------------------------------------------------------===//
// Used to inject printing of "_e32" suffix for VI (there are "_e64" variants for VI)
def VINTRPDst : VINTRPDstOperand <VGPR_32>;
let Uses = [MODE, M0, EXEC] in {
// FIXME: Specify SchedRW for VINTRP instructions.
multiclass V_INTERP_P1_F32_m : VINTRP_m <
0x00000000,
(outs VINTRPDst:$vdst),
(ins VGPR_32:$vsrc, InterpAttr:$attr, InterpAttrChan:$attrchan),
"v_interp_p1_f32$vdst, $vsrc, $attr$attrchan",
[(set f32:$vdst, (int_amdgcn_interp_p1 f32:$vsrc,
(i32 timm:$attrchan), (i32 timm:$attr), M0))]
>;
let OtherPredicates = [has32BankLDS, isNotGFX90APlus] in {
defm V_INTERP_P1_F32 : V_INTERP_P1_F32_m;
} // End OtherPredicates = [has32BankLDS, isNotGFX90APlus]
let OtherPredicates = [has16BankLDS, isNotGFX90APlus],
Constraints = "@earlyclobber $vdst", isAsmParserOnly=1 in {
defm V_INTERP_P1_F32_16bank : V_INTERP_P1_F32_m;
} // End OtherPredicates = [has32BankLDS, isNotGFX90APlus],
// Constraints = "@earlyclobber $vdst", isAsmParserOnly=1
let OtherPredicates = [isNotGFX90APlus] in {
let DisableEncoding = "$src0", Constraints = "$src0 = $vdst" in {
defm V_INTERP_P2_F32 : VINTRP_m <
0x00000001,
(outs VINTRPDst:$vdst),
(ins VGPR_32:$src0, VGPR_32:$vsrc, InterpAttr:$attr,
InterpAttrChan:$attrchan),
"v_interp_p2_f32$vdst, $vsrc, $attr$attrchan",
[(set f32:$vdst, (int_amdgcn_interp_p2 f32:$src0, f32:$vsrc,
(i32 timm:$attrchan), (i32 timm:$attr), M0))]>;
} // End DisableEncoding = "$src0", Constraints = "$src0 = $vdst"
defm V_INTERP_MOV_F32 : VINTRP_m <
0x00000002,
(outs VINTRPDst:$vdst),
(ins InterpSlot:$vsrc, InterpAttr:$attr, InterpAttrChan:$attrchan),
"v_interp_mov_f32$vdst, $vsrc, $attr$attrchan",
[(set f32:$vdst, (int_amdgcn_interp_mov (i32 timm:$vsrc),
(i32 timm:$attrchan), (i32 timm:$attr), M0))]>;
} // End OtherPredicates = [isNotGFX90APlus]
} // End Uses = [MODE, M0, EXEC]
//===----------------------------------------------------------------------===//
// Pseudo Instructions
//===----------------------------------------------------------------------===//
// Insert a branch to an endpgm block to use as a fallback trap.
def ENDPGM_TRAP : SPseudoInstSI<
(outs), (ins),
[(AMDGPUendpgm_trap)],
"ENDPGM_TRAP"> {
let hasSideEffects = 1;
let usesCustomInserter = 1;
}
def ATOMIC_FENCE : SPseudoInstSI<
(outs), (ins i32imm:$ordering, i32imm:$scope),
[(atomic_fence (i32 timm:$ordering), (i32 timm:$scope))],
"ATOMIC_FENCE $ordering, $scope"> {
let hasSideEffects = 1;
}
let hasSideEffects = 0, mayLoad = 0, mayStore = 0, Uses = [EXEC] in {
// For use in patterns
def V_CNDMASK_B64_PSEUDO : VOP3Common <(outs VReg_64:$vdst),
(ins VSrc_b64:$src0, VSrc_b64:$src1, SSrc_b64:$src2), "", []> {
let isPseudo = 1;
let isCodeGenOnly = 1;
let usesCustomInserter = 1;
}
// 64-bit vector move instruction. This is mainly used by the
// SIFoldOperands pass to enable folding of inline immediates.
def V_MOV_B64_PSEUDO : VPseudoInstSI <(outs VReg_64:$vdst),
(ins VSrc_b64:$src0)> {
let isReMaterializable = 1;
let isAsCheapAsAMove = 1;
let isMoveImm = 1;
let SchedRW = [Write64Bit];
let Size = 4;
let UseNamedOperandTable = 1;
}
// 64-bit vector move with dpp. Expanded post-RA.
def V_MOV_B64_DPP_PSEUDO : VOP_DPP_Pseudo <"v_mov_b64_dpp", VOP_I64_I64> {
let Size = 16; // Requires two 8-byte v_mov_b32_dpp to complete.
}
// 64-bit scalar move immediate instruction. This is used to avoid subregs
// initialization and allow rematerialization.
def S_MOV_B64_IMM_PSEUDO : SPseudoInstSI <(outs SReg_64:$sdst),
(ins i64imm:$src0)> {
let isReMaterializable = 1;
let isAsCheapAsAMove = 1;
let isMoveImm = 1;
let SchedRW = [WriteSALU, Write64Bit];
let Size = 4;
let Uses = [];
let UseNamedOperandTable = 1;
}
// Pseudoinstruction for @llvm.amdgcn.wqm. It is turned into a copy after the
// WQM pass processes it.
def WQM : PseudoInstSI <(outs unknown:$vdst), (ins unknown:$src0)>;
// Pseudoinstruction for @llvm.amdgcn.softwqm. Like @llvm.amdgcn.wqm it is
// turned into a copy by WQM pass, but does not seed WQM requirements.
def SOFT_WQM : PseudoInstSI <(outs unknown:$vdst), (ins unknown:$src0)>;
// Pseudoinstruction for @llvm.amdgcn.strict.wwm. It is turned into a copy post-RA, so
// that the @earlyclobber is respected. The @earlyclobber is to make sure that
// the instruction that defines $src0 (which is run in Whole Wave Mode) doesn't
// accidentally clobber inactive channels of $vdst.
let Constraints = "@earlyclobber $vdst" in {
def STRICT_WWM : PseudoInstSI <(outs unknown:$vdst), (ins unknown:$src0)>;
def STRICT_WQM : PseudoInstSI <(outs unknown:$vdst), (ins unknown:$src0)>;
}
} // End let hasSideEffects = 0, mayLoad = 0, mayStore = 0, Uses = [EXEC]
def WWM_COPY : SPseudoInstSI <
(outs unknown:$dst), (ins unknown:$src)> {
let hasSideEffects = 0;
let isAsCheapAsAMove = 1;
let isConvergent = 1;
}
def ENTER_STRICT_WWM : SPseudoInstSI <(outs SReg_1:$sdst), (ins i64imm:$src0)> {
let Uses = [EXEC];
let Defs = [EXEC, SCC];
let hasSideEffects = 0;
let mayLoad = 0;
let mayStore = 0;
}
def EXIT_STRICT_WWM : SPseudoInstSI <(outs SReg_1:$sdst), (ins SReg_1:$src0)> {
let hasSideEffects = 0;
let mayLoad = 0;
let mayStore = 0;
}
def ENTER_STRICT_WQM : SPseudoInstSI <(outs SReg_1:$sdst), (ins i64imm:$src0)> {
let Uses = [EXEC];
let Defs = [EXEC, SCC];
let hasSideEffects = 0;
let mayLoad = 0;
let mayStore = 0;
}
def EXIT_STRICT_WQM : SPseudoInstSI <(outs SReg_1:$sdst), (ins SReg_1:$src0)> {
let hasSideEffects = 0;
let mayLoad = 0;
let mayStore = 0;
}
let usesCustomInserter = 1 in {
def S_INVERSE_BALLOT_U32 : SPseudoInstSI <(outs SReg_32:$sdst), (ins SSrc_b32:$mask)>;
def S_INVERSE_BALLOT_U64 : SPseudoInstSI <(outs SReg_64:$sdst), (ins SSrc_b64:$mask)>;
} // End usesCustomInserter = 1
// PSEUDO_WM is treated like STRICT_WWM/STRICT_WQM without exec changes.
def ENTER_PSEUDO_WM : SPseudoInstSI <(outs), (ins)> {
let Uses = [EXEC];
let Defs = [EXEC];
let hasSideEffects = 0;
let mayLoad = 0;
let mayStore = 0;
}
def EXIT_PSEUDO_WM : SPseudoInstSI <(outs), (ins)> {
let hasSideEffects = 0;
let mayLoad = 0;
let mayStore = 0;
}
// Pseudo instructions used for @llvm.fptrunc.round upward
// and @llvm.fptrunc.round downward.
// These intrinsics will be legalized to G_FPTRUNC_ROUND_UPWARD
// and G_FPTRUNC_ROUND_DOWNWARD before being lowered to
// FPTRUNC_UPWARD_PSEUDO and FPTRUNC_DOWNWARD_PSEUDO.
// The final codegen is done in the ModeRegister pass.
let Uses = [MODE, EXEC] in {
def FPTRUNC_UPWARD_PSEUDO : VPseudoInstSI <(outs VGPR_32:$vdst),
(ins VGPR_32:$src0),
[(set f16:$vdst, (SIfptrunc_round_upward f32:$src0))]>;
def FPTRUNC_DOWNWARD_PSEUDO : VPseudoInstSI <(outs VGPR_32:$vdst),
(ins VGPR_32:$src0),
[(set f16:$vdst, (SIfptrunc_round_downward f32:$src0))]>;
} // End Uses = [MODE, EXEC]
// Invert the exec mask and overwrite the inactive lanes of dst with inactive,
// restoring it after we're done.
let Defs = [SCC], isConvergent = 1 in {
def V_SET_INACTIVE_B32 : VPseudoInstSI <(outs VGPR_32:$vdst),
(ins VSrc_b32: $src, VSrc_b32:$inactive),
[(set i32:$vdst, (int_amdgcn_set_inactive i32:$src, i32:$inactive))]> {
}
def V_SET_INACTIVE_B64 : VPseudoInstSI <(outs VReg_64:$vdst),
(ins VSrc_b64: $src, VSrc_b64:$inactive),
[(set i64:$vdst, (int_amdgcn_set_inactive i64:$src, i64:$inactive))]> {
}
} // End Defs = [SCC]
def : GCNPat<(i32 (int_amdgcn_set_inactive_chain_arg i32:$src, i32:$inactive)),
(V_SET_INACTIVE_B32 VGPR_32:$src, VGPR_32:$inactive)>;
def : GCNPat<(i64 (int_amdgcn_set_inactive_chain_arg i64:$src, i64:$inactive)),
(V_SET_INACTIVE_B64 VReg_64:$src, VReg_64:$inactive)>;
let usesCustomInserter = 1, hasSideEffects = 0, mayLoad = 0, mayStore = 0, Uses = [EXEC] in {
def WAVE_REDUCE_UMIN_PSEUDO_U32 : VPseudoInstSI <(outs SGPR_32:$sdst),
(ins VSrc_b32: $src, VSrc_b32:$strategy),
[(set i32:$sdst, (int_amdgcn_wave_reduce_umin i32:$src, i32:$strategy))]> {
}
def WAVE_REDUCE_UMAX_PSEUDO_U32 : VPseudoInstSI <(outs SGPR_32:$sdst),
(ins VSrc_b32: $src, VSrc_b32:$strategy),
[(set i32:$sdst, (int_amdgcn_wave_reduce_umax i32:$src, i32:$strategy))]> {
}
}
let usesCustomInserter = 1, Defs = [VCC] in {
def V_ADD_U64_PSEUDO : VPseudoInstSI <
(outs VReg_64:$vdst), (ins VSrc_b64:$src0, VSrc_b64:$src1),
[(set VReg_64:$vdst, (DivergentBinFrag<add> i64:$src0, i64:$src1))]
>;
def V_SUB_U64_PSEUDO : VPseudoInstSI <
(outs VReg_64:$vdst), (ins VSrc_b64:$src0, VSrc_b64:$src1),
[(set VReg_64:$vdst, (DivergentBinFrag<sub> i64:$src0, i64:$src1))]
>;
} // End usesCustomInserter = 1, Defs = [VCC]
let usesCustomInserter = 1, Defs = [SCC] in {
def S_ADD_U64_PSEUDO : SPseudoInstSI <
(outs SReg_64:$sdst), (ins SSrc_b64:$src0, SSrc_b64:$src1),
[(set SReg_64:$sdst, (UniformBinFrag<add> i64:$src0, i64:$src1))]
>;
def S_SUB_U64_PSEUDO : SPseudoInstSI <
(outs SReg_64:$sdst), (ins SSrc_b64:$src0, SSrc_b64:$src1),
[(set SReg_64:$sdst, (UniformBinFrag<sub> i64:$src0, i64:$src1))]
>;
def S_ADD_CO_PSEUDO : SPseudoInstSI <
(outs SReg_32:$sdst, SSrc_i1:$scc_out), (ins SSrc_b32:$src0, SSrc_b32:$src1, SSrc_i1:$scc_in)
>;
def S_SUB_CO_PSEUDO : SPseudoInstSI <
(outs SReg_32:$sdst, SSrc_i1:$scc_out), (ins SSrc_b32:$src0, SSrc_b32:$src1, SSrc_i1:$scc_in)
>;
def S_UADDO_PSEUDO : SPseudoInstSI <
(outs SReg_32:$sdst, SSrc_i1:$scc_out), (ins SSrc_b32:$src0, SSrc_b32:$src1)
>;
def S_USUBO_PSEUDO : SPseudoInstSI <
(outs SReg_32:$sdst, SSrc_i1:$scc_out), (ins SSrc_b32:$src0, SSrc_b32:$src1)
>;
let OtherPredicates = [HasShaderCyclesHiLoRegisters] in
def GET_SHADERCYCLESHILO : SPseudoInstSI<
(outs SReg_64:$sdst), (ins),
[(set SReg_64:$sdst, (i64 (readcyclecounter)))]
>;
} // End usesCustomInserter = 1, Defs = [SCC]
let usesCustomInserter = 1 in {
def GET_GROUPSTATICSIZE : SPseudoInstSI <(outs SReg_32:$sdst), (ins),
[(set SReg_32:$sdst, (int_amdgcn_groupstaticsize))]>;
} // End let usesCustomInserter = 1, SALU = 1
// Wrap an instruction by duplicating it, except for setting isTerminator.
class WrapTerminatorInst<SOP_Pseudo base_inst> : SPseudoInstSI<
base_inst.OutOperandList,
base_inst.InOperandList> {
let Uses = base_inst.Uses;
let Defs = base_inst.Defs;
let isTerminator = 1;
let isAsCheapAsAMove = base_inst.isAsCheapAsAMove;
let hasSideEffects = base_inst.hasSideEffects;
let UseNamedOperandTable = base_inst.UseNamedOperandTable;
let CodeSize = base_inst.CodeSize;
let SchedRW = base_inst.SchedRW;
}
let WaveSizePredicate = isWave64 in {
def S_MOV_B64_term : WrapTerminatorInst<S_MOV_B64>;
def S_XOR_B64_term : WrapTerminatorInst<S_XOR_B64>;
def S_OR_B64_term : WrapTerminatorInst<S_OR_B64>;
def S_ANDN2_B64_term : WrapTerminatorInst<S_ANDN2_B64>;
def S_AND_B64_term : WrapTerminatorInst<S_AND_B64>;
def S_AND_SAVEEXEC_B64_term : WrapTerminatorInst<S_AND_SAVEEXEC_B64>;
}
let WaveSizePredicate = isWave32 in {
def S_MOV_B32_term : WrapTerminatorInst<S_MOV_B32>;
def S_XOR_B32_term : WrapTerminatorInst<S_XOR_B32>;
def S_OR_B32_term : WrapTerminatorInst<S_OR_B32>;
def S_ANDN2_B32_term : WrapTerminatorInst<S_ANDN2_B32>;
def S_AND_B32_term : WrapTerminatorInst<S_AND_B32>;
def S_AND_SAVEEXEC_B32_term : WrapTerminatorInst<S_AND_SAVEEXEC_B32>;
}
def WAVE_BARRIER : SPseudoInstSI<(outs), (ins),
[(int_amdgcn_wave_barrier)]> {
let SchedRW = [];
let hasNoSchedulingInfo = 1;
let hasSideEffects = 1;
let mayLoad = 0;
let mayStore = 0;
let isConvergent = 1;
let FixedSize = 1;
let Size = 0;
let isMeta = 1;
}
def SCHED_BARRIER : SPseudoInstSI<(outs), (ins i32imm:$mask),
[(int_amdgcn_sched_barrier (i32 timm:$mask))]> {
let SchedRW = [];
let hasNoSchedulingInfo = 1;
let hasSideEffects = 1;
let mayLoad = 0;
let mayStore = 0;
let isConvergent = 1;
let FixedSize = 1;
let Size = 0;
let isMeta = 1;
}
def SCHED_GROUP_BARRIER : SPseudoInstSI<
(outs),
(ins i32imm:$mask, i32imm:$size, i32imm:$syncid),
[(int_amdgcn_sched_group_barrier (i32 timm:$mask), (i32 timm:$size), (i32 timm:$syncid))]> {
let SchedRW = [];
let hasNoSchedulingInfo = 1;
let hasSideEffects = 1;
let mayLoad = 0;
let mayStore = 0;
let isConvergent = 1;
let FixedSize = 1;
let Size = 0;
let isMeta = 1;
}
def IGLP_OPT : SPseudoInstSI<(outs), (ins i32imm:$mask),
[(int_amdgcn_iglp_opt (i32 timm:$mask))]> {
let SchedRW = [];
let hasNoSchedulingInfo = 1;
let hasSideEffects = 1;
let mayLoad = 0;
let mayStore = 0;
let isConvergent = 1;
let FixedSize = 1;
let Size = 0;
let isMeta = 1;
}
// SI pseudo instructions. These are used by the CFG structurizer pass
// and should be lowered to ISA instructions prior to codegen.
// As we have enhanced control flow intrinsics to work under unstructured CFG,
// duplicating such intrinsics can be actually treated as legal. On the contrary,
// by making them non-duplicable, we are observing better code generation result.
// So we choose to mark them non-duplicable in hope of getting better code
// generation as well as simplied CFG during Machine IR optimization stage.
let isTerminator = 1, isNotDuplicable = 1 in {
let OtherPredicates = [EnableLateCFGStructurize] in {
def SI_NON_UNIFORM_BRCOND_PSEUDO : CFPseudoInstSI <
(outs),
(ins SReg_1:$vcc, brtarget:$target),
[(brcond i1:$vcc, bb:$target)]> {
let Size = 12;
}
}
def SI_IF: CFPseudoInstSI <
(outs SReg_1:$dst), (ins SReg_1:$vcc, brtarget:$target),
[(set i1:$dst, (AMDGPUif i1:$vcc, bb:$target))], 1, 1> {
let Constraints = "";
let Size = 12;
let hasSideEffects = 1;
let IsNeverUniform = 1;
}
def SI_ELSE : CFPseudoInstSI <
(outs SReg_1:$dst),
(ins SReg_1:$src, brtarget:$target), [], 1, 1> {
let Size = 12;
let hasSideEffects = 1;
let IsNeverUniform = 1;
}
def SI_WATERFALL_LOOP : CFPseudoInstSI <
(outs),
(ins brtarget:$target), [], 1> {
let Size = 8;
let isBranch = 1;
let Defs = [];
}
def SI_LOOP : CFPseudoInstSI <
(outs), (ins SReg_1:$saved, brtarget:$target),
[(AMDGPUloop i1:$saved, bb:$target)], 1, 1> {
let Size = 8;
let isBranch = 1;
let hasSideEffects = 1;
let IsNeverUniform = 1;
}
} // End isTerminator = 1
def SI_END_CF : CFPseudoInstSI <
(outs), (ins SReg_1:$saved), [], 1, 1> {
let Size = 4;
let isAsCheapAsAMove = 1;
let isReMaterializable = 1;
let hasSideEffects = 1;
let isNotDuplicable = 1; // Not a hard requirement, see long comments above for details.
let mayLoad = 1; // FIXME: Should not need memory flags
let mayStore = 1;
}
def SI_IF_BREAK : CFPseudoInstSI <
(outs SReg_1:$dst), (ins SReg_1:$vcc, SReg_1:$src), []> {
let Size = 4;
let isNotDuplicable = 1; // Not a hard requirement, see long comments above for details.
let isAsCheapAsAMove = 1;
let isReMaterializable = 1;
}
// Branch to the early termination block of the shader if SCC is 0.
// This uses SCC from a previous SALU operation, i.e. the update of
// a mask of live lanes after a kill/demote operation.
// Only valid in pixel shaders.
def SI_EARLY_TERMINATE_SCC0 : SPseudoInstSI <(outs), (ins)> {
let Uses = [EXEC,SCC];
}
let Uses = [EXEC] in {
multiclass PseudoInstKill <dag ins> {
// Even though this pseudo can usually be expanded without an SCC def, we
// conservatively assume that it has an SCC def, both because it is sometimes
// required in degenerate cases (when V_CMPX cannot be used due to constant
// bus limitations) and because it allows us to avoid having to track SCC
// liveness across basic blocks.
let Defs = [EXEC,SCC] in
def _PSEUDO : PseudoInstSI <(outs), ins> {
let isConvergent = 1;
let usesCustomInserter = 1;
}
let Defs = [EXEC,SCC] in
def _TERMINATOR : SPseudoInstSI <(outs), ins> {
let isTerminator = 1;
}
}
defm SI_KILL_I1 : PseudoInstKill <(ins SCSrc_i1:$src, i1imm:$killvalue)>;
let Defs = [VCC] in
defm SI_KILL_F32_COND_IMM : PseudoInstKill <(ins VSrc_b32:$src0, i32imm:$src1, i32imm:$cond)>;
let Defs = [EXEC,VCC] in
def SI_ILLEGAL_COPY : SPseudoInstSI <
(outs unknown:$dst), (ins unknown:$src),
[], " ; illegal copy $src to $dst">;
} // End Uses = [EXEC], Defs = [EXEC,VCC]
// Branch on undef scc. Used to avoid intermediate copy from
// IMPLICIT_DEF to SCC.
def SI_BR_UNDEF : SPseudoInstSI <(outs), (ins SOPPBrTarget:$simm16)> {
let isTerminator = 1;
let usesCustomInserter = 1;
let isBranch = 1;
}
def SI_PS_LIVE : PseudoInstSI <
(outs SReg_1:$dst), (ins),
[(set i1:$dst, (int_amdgcn_ps_live))]> {
let SALU = 1;
}
let Uses = [EXEC] in {
def SI_LIVE_MASK : PseudoInstSI <
(outs SReg_1:$dst), (ins),
[(set i1:$dst, (int_amdgcn_live_mask))]> {
let SALU = 1;
}
let Defs = [EXEC,SCC] in {
// Demote: Turn a pixel shader thread into a helper lane.
def SI_DEMOTE_I1 : SPseudoInstSI <(outs), (ins SCSrc_i1:$src, i1imm:$killvalue)>;
} // End Defs = [EXEC,SCC]
} // End Uses = [EXEC]
def SI_MASKED_UNREACHABLE : SPseudoInstSI <(outs), (ins),
[(int_amdgcn_unreachable)],
"; divergent unreachable"> {
let Size = 0;
let hasNoSchedulingInfo = 1;
let FixedSize = 1;
let isMeta = 1;
let maybeAtomic = 0;
}
// Used as an isel pseudo to directly emit initialization with an
// s_mov_b32 rather than a copy of another initialized
// register. MachineCSE skips copies, and we don't want to have to
// fold operands before it runs.
def SI_INIT_M0 : SPseudoInstSI <(outs), (ins SSrc_b32:$src)> {
let Defs = [M0];
let usesCustomInserter = 1;
let isAsCheapAsAMove = 1;
let isReMaterializable = 1;
}
def SI_INIT_EXEC : SPseudoInstSI <
(outs), (ins i64imm:$src),
[(int_amdgcn_init_exec (i64 timm:$src))]> {
let Defs = [EXEC];
let isAsCheapAsAMove = 1;
}
def SI_INIT_EXEC_FROM_INPUT : SPseudoInstSI <
(outs), (ins SSrc_b32:$input, i32imm:$shift),
[(int_amdgcn_init_exec_from_input i32:$input, (i32 timm:$shift))]> {
let Defs = [EXEC];
}
// Return for returning shaders to a shader variant epilog.
def SI_RETURN_TO_EPILOG : SPseudoInstSI <
(outs), (ins variable_ops), [(AMDGPUreturn_to_epilog)]> {
let isTerminator = 1;
let isBarrier = 1;
let isReturn = 1;
let hasNoSchedulingInfo = 1;
let DisableWQM = 1;
let FixedSize = 1;
// TODO: Should this be true?
let isMeta = 0;
}
// Return for returning function calls.
def SI_RETURN : SPseudoInstSI <
(outs), (ins), [(AMDGPUret_glue)],
"; return"> {
let isTerminator = 1;
let isBarrier = 1;
let isReturn = 1;
let SchedRW = [WriteBranch];
}
// Return for returning function calls without output register.
//
// This version is only needed so we can fill in the output register
// in the custom inserter.
def SI_CALL_ISEL : SPseudoInstSI <
(outs), (ins SSrc_b64:$src0, unknown:$callee),
[(AMDGPUcall i64:$src0, tglobaladdr:$callee)]> {
let Size = 4;
let isCall = 1;
let SchedRW = [WriteBranch];
let usesCustomInserter = 1;
// TODO: Should really base this on the call target
let isConvergent = 1;
}
def : GCNPat<
(AMDGPUcall i64:$src0, (i64 0)),
(SI_CALL_ISEL $src0, (i64 0))
>;
// Wrapper around s_swappc_b64 with extra $callee parameter to track
// the called function after regalloc.
def SI_CALL : SPseudoInstSI <
(outs SReg_64:$dst), (ins SSrc_b64:$src0, unknown:$callee)> {
let Size = 4;
let FixedSize = 1;
let isCall = 1;
let UseNamedOperandTable = 1;
let SchedRW = [WriteBranch];
// TODO: Should really base this on the call target
let isConvergent = 1;
}
class SI_TCRETURN_Pseudo<RegisterClass rc, SDNode sd> : SPseudoInstSI <(outs),
(ins rc:$src0, unknown:$callee, i32imm:$fpdiff),
[(sd i64:$src0, tglobaladdr:$callee, i32:$fpdiff)]> {
let Size = 4;
let FixedSize = 1;
let isCall = 1;
let isTerminator = 1;
let isReturn = 1;
let isBarrier = 1;
let UseNamedOperandTable = 1;
let SchedRW = [WriteBranch];
// TODO: Should really base this on the call target
let isConvergent = 1;
}
// Tail call handling pseudo
def SI_TCRETURN : SI_TCRETURN_Pseudo<CCR_SGPR_64, AMDGPUtc_return>;
def SI_TCRETURN_GFX : SI_TCRETURN_Pseudo<Gfx_CCR_SGPR_64, AMDGPUtc_return_gfx>;
// Handle selecting indirect tail calls
def : GCNPat<
(AMDGPUtc_return i64:$src0, (i64 0), (i32 timm:$fpdiff)),
(SI_TCRETURN CCR_SGPR_64:$src0, (i64 0), i32imm:$fpdiff)
>;
// Handle selecting indirect tail calls for AMDGPU_gfx
def : GCNPat<
(AMDGPUtc_return_gfx i64:$src0, (i64 0), (i32 timm:$fpdiff)),
(SI_TCRETURN_GFX Gfx_CCR_SGPR_64:$src0, (i64 0), i32imm:$fpdiff)
>;
// Pseudo for the llvm.amdgcn.cs.chain intrinsic.
// This is essentially a tail call, but it also takes a mask to put in EXEC
// right before jumping to the callee.
class SI_CS_CHAIN_TC<
ValueType execvt, Predicate wavesizepred,
RegisterOperand execrc = getSOPSrcForVT<execvt>.ret>
: SPseudoInstSI <(outs),
(ins CCR_SGPR_64:$src0, unknown:$callee, i32imm:$fpdiff, execrc:$exec)> {
let FixedSize = 0;
let isCall = 1;
let isTerminator = 1;
let isBarrier = 1;
let isReturn = 1;
let UseNamedOperandTable = 1;
let SchedRW = [WriteBranch];
let isConvergent = 1;
let WaveSizePredicate = wavesizepred;
}
def SI_CS_CHAIN_TC_W32 : SI_CS_CHAIN_TC<i32, isWave32>;
def SI_CS_CHAIN_TC_W64 : SI_CS_CHAIN_TC<i64, isWave64>;
// Handle selecting direct & indirect calls via SI_CS_CHAIN_TC_W32/64
multiclass si_cs_chain_tc_pattern<
dag callee, ValueType execvt, RegisterOperand execrc, Instruction tc> {
def : GCNPat<
(AMDGPUtc_return_chain i64:$src0, callee, (i32 timm:$fpdiff), execvt:$exec),
(tc CCR_SGPR_64:$src0, callee, i32imm:$fpdiff, execrc:$exec)
>;
}
multiclass si_cs_chain_tc_patterns<
ValueType execvt,
RegisterOperand execrc = getSOPSrcForVT<execvt>.ret,
Instruction tc = !if(!eq(execvt, i32), SI_CS_CHAIN_TC_W32, SI_CS_CHAIN_TC_W64)
> {
defm direct: si_cs_chain_tc_pattern<(tglobaladdr:$callee), execvt, execrc, tc>;
defm indirect: si_cs_chain_tc_pattern<(i64 0), execvt, execrc, tc>;
}
defm : si_cs_chain_tc_patterns<i32>;
defm : si_cs_chain_tc_patterns<i64>;
def ADJCALLSTACKUP : SPseudoInstSI<
(outs), (ins i32imm:$amt0, i32imm:$amt1),
[(callseq_start timm:$amt0, timm:$amt1)],
"; adjcallstackup $amt0 $amt1"> {
let Size = 8; // Worst case. (s_add_u32 + constant)
let FixedSize = 1;
let hasSideEffects = 1;
let usesCustomInserter = 1;
let SchedRW = [WriteSALU];
let Defs = [SCC];
}
def ADJCALLSTACKDOWN : SPseudoInstSI<
(outs), (ins i32imm:$amt1, i32imm:$amt2),
[(callseq_end timm:$amt1, timm:$amt2)],
"; adjcallstackdown $amt1"> {
let Size = 8; // Worst case. (s_add_u32 + constant)
let hasSideEffects = 1;
let usesCustomInserter = 1;
let SchedRW = [WriteSALU];
let Defs = [SCC];
}
let Defs = [M0, EXEC, SCC],
UseNamedOperandTable = 1 in {
// SI_INDIRECT_SRC/DST are only used by legacy SelectionDAG indirect
// addressing implementation.
class SI_INDIRECT_SRC<RegisterClass rc> : VPseudoInstSI <
(outs VGPR_32:$vdst),
(ins rc:$src, VS_32:$idx, i32imm:$offset)> {
let usesCustomInserter = 1;
}
class SI_INDIRECT_DST<RegisterClass rc> : VPseudoInstSI <
(outs rc:$vdst),
(ins rc:$src, VS_32:$idx, i32imm:$offset, VGPR_32:$val)> {
let Constraints = "$src = $vdst";
let usesCustomInserter = 1;
}
def SI_INDIRECT_SRC_V1 : SI_INDIRECT_SRC<VGPR_32>;
def SI_INDIRECT_SRC_V2 : SI_INDIRECT_SRC<VReg_64>;
def SI_INDIRECT_SRC_V4 : SI_INDIRECT_SRC<VReg_128>;
def SI_INDIRECT_SRC_V8 : SI_INDIRECT_SRC<VReg_256>;
def SI_INDIRECT_SRC_V9 : SI_INDIRECT_SRC<VReg_288>;
def SI_INDIRECT_SRC_V10 : SI_INDIRECT_SRC<VReg_320>;
def SI_INDIRECT_SRC_V11 : SI_INDIRECT_SRC<VReg_352>;
def SI_INDIRECT_SRC_V12 : SI_INDIRECT_SRC<VReg_384>;
def SI_INDIRECT_SRC_V16 : SI_INDIRECT_SRC<VReg_512>;
def SI_INDIRECT_SRC_V32 : SI_INDIRECT_SRC<VReg_1024>;
def SI_INDIRECT_DST_V1 : SI_INDIRECT_DST<VGPR_32>;
def SI_INDIRECT_DST_V2 : SI_INDIRECT_DST<VReg_64>;
def SI_INDIRECT_DST_V4 : SI_INDIRECT_DST<VReg_128>;
def SI_INDIRECT_DST_V8 : SI_INDIRECT_DST<VReg_256>;
def SI_INDIRECT_DST_V9 : SI_INDIRECT_DST<VReg_288>;
def SI_INDIRECT_DST_V10 : SI_INDIRECT_DST<VReg_320>;
def SI_INDIRECT_DST_V11 : SI_INDIRECT_DST<VReg_352>;
def SI_INDIRECT_DST_V12 : SI_INDIRECT_DST<VReg_384>;
def SI_INDIRECT_DST_V16 : SI_INDIRECT_DST<VReg_512>;
def SI_INDIRECT_DST_V32 : SI_INDIRECT_DST<VReg_1024>;
} // End Uses = [EXEC], Defs = [M0, EXEC]
// This is a pseudo variant of the v_movreld_b32 instruction in which the
// vector operand appears only twice, once as def and once as use. Using this
// pseudo avoids problems with the Two Address instructions pass.
class INDIRECT_REG_WRITE_MOVREL_pseudo<RegisterClass rc,
RegisterOperand val_ty> : PseudoInstSI <
(outs rc:$vdst), (ins rc:$vsrc, val_ty:$val, i32imm:$subreg)> {
let Constraints = "$vsrc = $vdst";
let Uses = [M0];
}
class V_INDIRECT_REG_WRITE_MOVREL_B32_pseudo<RegisterClass rc> :
INDIRECT_REG_WRITE_MOVREL_pseudo<rc, VSrc_b32> {
let VALU = 1;
let VOP1 = 1;
let Uses = [M0, EXEC];
}
class S_INDIRECT_REG_WRITE_MOVREL_pseudo<RegisterClass rc,
RegisterOperand val_ty> :
INDIRECT_REG_WRITE_MOVREL_pseudo<rc, val_ty> {
let SALU = 1;
let SOP1 = 1;
let Uses = [M0];
}
class S_INDIRECT_REG_WRITE_MOVREL_B32_pseudo<RegisterClass rc> :
S_INDIRECT_REG_WRITE_MOVREL_pseudo<rc, SSrc_b32>;
class S_INDIRECT_REG_WRITE_MOVREL_B64_pseudo<RegisterClass rc> :
S_INDIRECT_REG_WRITE_MOVREL_pseudo<rc, SSrc_b64>;
def V_INDIRECT_REG_WRITE_MOVREL_B32_V1 : V_INDIRECT_REG_WRITE_MOVREL_B32_pseudo<VGPR_32>;
def V_INDIRECT_REG_WRITE_MOVREL_B32_V2 : V_INDIRECT_REG_WRITE_MOVREL_B32_pseudo<VReg_64>;
def V_INDIRECT_REG_WRITE_MOVREL_B32_V3 : V_INDIRECT_REG_WRITE_MOVREL_B32_pseudo<VReg_96>;
def V_INDIRECT_REG_WRITE_MOVREL_B32_V4 : V_INDIRECT_REG_WRITE_MOVREL_B32_pseudo<VReg_128>;
def V_INDIRECT_REG_WRITE_MOVREL_B32_V5 : V_INDIRECT_REG_WRITE_MOVREL_B32_pseudo<VReg_160>;
def V_INDIRECT_REG_WRITE_MOVREL_B32_V8 : V_INDIRECT_REG_WRITE_MOVREL_B32_pseudo<VReg_256>;
def V_INDIRECT_REG_WRITE_MOVREL_B32_V9 : V_INDIRECT_REG_WRITE_MOVREL_B32_pseudo<VReg_288>;
def V_INDIRECT_REG_WRITE_MOVREL_B32_V10 : V_INDIRECT_REG_WRITE_MOVREL_B32_pseudo<VReg_320>;
def V_INDIRECT_REG_WRITE_MOVREL_B32_V11 : V_INDIRECT_REG_WRITE_MOVREL_B32_pseudo<VReg_352>;
def V_INDIRECT_REG_WRITE_MOVREL_B32_V12 : V_INDIRECT_REG_WRITE_MOVREL_B32_pseudo<VReg_384>;
def V_INDIRECT_REG_WRITE_MOVREL_B32_V16 : V_INDIRECT_REG_WRITE_MOVREL_B32_pseudo<VReg_512>;
def V_INDIRECT_REG_WRITE_MOVREL_B32_V32 : V_INDIRECT_REG_WRITE_MOVREL_B32_pseudo<VReg_1024>;
def S_INDIRECT_REG_WRITE_MOVREL_B32_V1 : S_INDIRECT_REG_WRITE_MOVREL_B32_pseudo<SReg_32>;
def S_INDIRECT_REG_WRITE_MOVREL_B32_V2 : S_INDIRECT_REG_WRITE_MOVREL_B32_pseudo<SReg_64>;
def S_INDIRECT_REG_WRITE_MOVREL_B32_V3 : S_INDIRECT_REG_WRITE_MOVREL_B32_pseudo<SReg_96>;
def S_INDIRECT_REG_WRITE_MOVREL_B32_V4 : S_INDIRECT_REG_WRITE_MOVREL_B32_pseudo<SReg_128>;
def S_INDIRECT_REG_WRITE_MOVREL_B32_V5 : S_INDIRECT_REG_WRITE_MOVREL_B32_pseudo<SReg_160>;
def S_INDIRECT_REG_WRITE_MOVREL_B32_V8 : S_INDIRECT_REG_WRITE_MOVREL_B32_pseudo<SReg_256>;
def S_INDIRECT_REG_WRITE_MOVREL_B32_V9 : S_INDIRECT_REG_WRITE_MOVREL_B32_pseudo<SReg_288>;
def S_INDIRECT_REG_WRITE_MOVREL_B32_V10 : S_INDIRECT_REG_WRITE_MOVREL_B32_pseudo<SReg_320>;
def S_INDIRECT_REG_WRITE_MOVREL_B32_V11 : S_INDIRECT_REG_WRITE_MOVREL_B32_pseudo<SReg_352>;
def S_INDIRECT_REG_WRITE_MOVREL_B32_V12 : S_INDIRECT_REG_WRITE_MOVREL_B32_pseudo<SReg_384>;
def S_INDIRECT_REG_WRITE_MOVREL_B32_V16 : S_INDIRECT_REG_WRITE_MOVREL_B32_pseudo<SReg_512>;
def S_INDIRECT_REG_WRITE_MOVREL_B32_V32 : S_INDIRECT_REG_WRITE_MOVREL_B32_pseudo<SReg_1024>;
def S_INDIRECT_REG_WRITE_MOVREL_B64_V1 : S_INDIRECT_REG_WRITE_MOVREL_B64_pseudo<SReg_64>;
def S_INDIRECT_REG_WRITE_MOVREL_B64_V2 : S_INDIRECT_REG_WRITE_MOVREL_B64_pseudo<SReg_128>;
def S_INDIRECT_REG_WRITE_MOVREL_B64_V4 : S_INDIRECT_REG_WRITE_MOVREL_B64_pseudo<SReg_256>;
def S_INDIRECT_REG_WRITE_MOVREL_B64_V8 : S_INDIRECT_REG_WRITE_MOVREL_B64_pseudo<SReg_512>;
def S_INDIRECT_REG_WRITE_MOVREL_B64_V16 : S_INDIRECT_REG_WRITE_MOVREL_B64_pseudo<SReg_1024>;
// These variants of V_INDIRECT_REG_READ/WRITE use VGPR indexing. By using these
// pseudos we avoid spills or copies being inserted within indirect sequences
// that switch the VGPR indexing mode. Spills to accvgprs could be effected by
// this mode switching.
class V_INDIRECT_REG_WRITE_GPR_IDX_pseudo<RegisterClass rc> : PseudoInstSI <
(outs rc:$vdst), (ins rc:$vsrc, VSrc_b32:$val, SSrc_b32:$idx, i32imm:$subreg)> {
let Constraints = "$vsrc = $vdst";
let VALU = 1;
let Uses = [M0, EXEC];
let Defs = [M0];
}
def V_INDIRECT_REG_WRITE_GPR_IDX_B32_V1 : V_INDIRECT_REG_WRITE_GPR_IDX_pseudo<VGPR_32>;
def V_INDIRECT_REG_WRITE_GPR_IDX_B32_V2 : V_INDIRECT_REG_WRITE_GPR_IDX_pseudo<VReg_64>;
def V_INDIRECT_REG_WRITE_GPR_IDX_B32_V3 : V_INDIRECT_REG_WRITE_GPR_IDX_pseudo<VReg_96>;
def V_INDIRECT_REG_WRITE_GPR_IDX_B32_V4 : V_INDIRECT_REG_WRITE_GPR_IDX_pseudo<VReg_128>;
def V_INDIRECT_REG_WRITE_GPR_IDX_B32_V5 : V_INDIRECT_REG_WRITE_GPR_IDX_pseudo<VReg_160>;
def V_INDIRECT_REG_WRITE_GPR_IDX_B32_V8 : V_INDIRECT_REG_WRITE_GPR_IDX_pseudo<VReg_256>;
def V_INDIRECT_REG_WRITE_GPR_IDX_B32_V9 : V_INDIRECT_REG_WRITE_GPR_IDX_pseudo<VReg_288>;
def V_INDIRECT_REG_WRITE_GPR_IDX_B32_V10 : V_INDIRECT_REG_WRITE_GPR_IDX_pseudo<VReg_320>;
def V_INDIRECT_REG_WRITE_GPR_IDX_B32_V11 : V_INDIRECT_REG_WRITE_GPR_IDX_pseudo<VReg_352>;
def V_INDIRECT_REG_WRITE_GPR_IDX_B32_V12 : V_INDIRECT_REG_WRITE_GPR_IDX_pseudo<VReg_384>;
def V_INDIRECT_REG_WRITE_GPR_IDX_B32_V16 : V_INDIRECT_REG_WRITE_GPR_IDX_pseudo<VReg_512>;
def V_INDIRECT_REG_WRITE_GPR_IDX_B32_V32 : V_INDIRECT_REG_WRITE_GPR_IDX_pseudo<VReg_1024>;
class V_INDIRECT_REG_READ_GPR_IDX_pseudo<RegisterClass rc> : PseudoInstSI <
(outs VGPR_32:$vdst), (ins rc:$vsrc, SSrc_b32:$idx, i32imm:$subreg)> {
let VALU = 1;
let Uses = [M0, EXEC];
let Defs = [M0];
}
def V_INDIRECT_REG_READ_GPR_IDX_B32_V1 : V_INDIRECT_REG_READ_GPR_IDX_pseudo<VGPR_32>;
def V_INDIRECT_REG_READ_GPR_IDX_B32_V2 : V_INDIRECT_REG_READ_GPR_IDX_pseudo<VReg_64>;
def V_INDIRECT_REG_READ_GPR_IDX_B32_V3 : V_INDIRECT_REG_READ_GPR_IDX_pseudo<VReg_96>;
def V_INDIRECT_REG_READ_GPR_IDX_B32_V4 : V_INDIRECT_REG_READ_GPR_IDX_pseudo<VReg_128>;
def V_INDIRECT_REG_READ_GPR_IDX_B32_V5 : V_INDIRECT_REG_READ_GPR_IDX_pseudo<VReg_160>;
def V_INDIRECT_REG_READ_GPR_IDX_B32_V8 : V_INDIRECT_REG_READ_GPR_IDX_pseudo<VReg_256>;
def V_INDIRECT_REG_READ_GPR_IDX_B32_V9 : V_INDIRECT_REG_READ_GPR_IDX_pseudo<VReg_288>;
def V_INDIRECT_REG_READ_GPR_IDX_B32_V10 : V_INDIRECT_REG_READ_GPR_IDX_pseudo<VReg_320>;
def V_INDIRECT_REG_READ_GPR_IDX_B32_V11 : V_INDIRECT_REG_READ_GPR_IDX_pseudo<VReg_352>;
def V_INDIRECT_REG_READ_GPR_IDX_B32_V12 : V_INDIRECT_REG_READ_GPR_IDX_pseudo<VReg_384>;
def V_INDIRECT_REG_READ_GPR_IDX_B32_V16 : V_INDIRECT_REG_READ_GPR_IDX_pseudo<VReg_512>;
def V_INDIRECT_REG_READ_GPR_IDX_B32_V32 : V_INDIRECT_REG_READ_GPR_IDX_pseudo<VReg_1024>;
multiclass SI_SPILL_SGPR <RegisterClass sgpr_class> {
let UseNamedOperandTable = 1, SGPRSpill = 1, Uses = [EXEC] in {
def _SAVE : PseudoInstSI <
(outs),
(ins sgpr_class:$data, i32imm:$addr)> {
let mayStore = 1;
let mayLoad = 0;
}
def _RESTORE : PseudoInstSI <
(outs sgpr_class:$data),
(ins i32imm:$addr)> {
let mayStore = 0;
let mayLoad = 1;
}
} // End UseNamedOperandTable = 1
}
// You cannot use M0 as the output of v_readlane_b32 instructions or
// use it in the sdata operand of SMEM instructions. We still need to
// be able to spill the physical register m0, so allow it for
// SI_SPILL_32_* instructions.
defm SI_SPILL_S32 : SI_SPILL_SGPR <SReg_32>;
defm SI_SPILL_S64 : SI_SPILL_SGPR <SReg_64>;
defm SI_SPILL_S96 : SI_SPILL_SGPR <SReg_96>;
defm SI_SPILL_S128 : SI_SPILL_SGPR <SReg_128>;
defm SI_SPILL_S160 : SI_SPILL_SGPR <SReg_160>;
defm SI_SPILL_S192 : SI_SPILL_SGPR <SReg_192>;
defm SI_SPILL_S224 : SI_SPILL_SGPR <SReg_224>;
defm SI_SPILL_S256 : SI_SPILL_SGPR <SReg_256>;
defm SI_SPILL_S288 : SI_SPILL_SGPR <SReg_288>;
defm SI_SPILL_S320 : SI_SPILL_SGPR <SReg_320>;
defm SI_SPILL_S352 : SI_SPILL_SGPR <SReg_352>;
defm SI_SPILL_S384 : SI_SPILL_SGPR <SReg_384>;
defm SI_SPILL_S512 : SI_SPILL_SGPR <SReg_512>;
defm SI_SPILL_S1024 : SI_SPILL_SGPR <SReg_1024>;
let SGPRSpill = 1, VALU = 1, isConvergent = 1 in {
def SI_SPILL_S32_TO_VGPR : PseudoInstSI <(outs VGPR_32:$vdst),
(ins SReg_32:$src0, i32imm:$src1, VGPR_32:$vdst_in)> {
let Size = 4;
let FixedSize = 1;
let IsNeverUniform = 1;
let hasSideEffects = 0;
let mayLoad = 0;
let mayStore = 0;
let Constraints = "$vdst = $vdst_in";
}
def SI_RESTORE_S32_FROM_VGPR : PseudoInstSI <(outs SReg_32:$sdst),
(ins VGPR_32:$src0, i32imm:$src1)> {
let Size = 4;
let FixedSize = 1;
let hasSideEffects = 0;
let mayLoad = 0;
let mayStore = 0;
}
} // End SGPRSpill = 1, VALU = 1, isConvergent = 1
// VGPR or AGPR spill instructions. In case of AGPR spilling a temp register
// needs to be used and an extra instruction to move between VGPR and AGPR.
// UsesTmp adds to the total size of an expanded spill in this case.
multiclass SI_SPILL_VGPR <RegisterClass vgpr_class, bit UsesTmp = 0> {
let UseNamedOperandTable = 1, VGPRSpill = 1,
SchedRW = [WriteVMEM] in {
def _SAVE : VPseudoInstSI <
(outs),
(ins vgpr_class:$vdata, i32imm:$vaddr,
SReg_32:$soffset, i32imm:$offset)> {
let mayStore = 1;
let mayLoad = 0;
// (2 * 4) + (8 * num_subregs) bytes maximum
int MaxSize = !add(!shl(!srl(vgpr_class.Size, 5), !add(UsesTmp, 3)), 8);
// Size field is unsigned char and cannot fit more.
let Size = !if(!le(MaxSize, 256), MaxSize, 252);
}
def _RESTORE : VPseudoInstSI <
(outs vgpr_class:$vdata),
(ins i32imm:$vaddr,
SReg_32:$soffset, i32imm:$offset)> {
let mayStore = 0;
let mayLoad = 1;
// (2 * 4) + (8 * num_subregs) bytes maximum
int MaxSize = !add(!shl(!srl(vgpr_class.Size, 5), !add(UsesTmp, 3)), 8);
// Size field is unsigned char and cannot fit more.
let Size = !if(!le(MaxSize, 256), MaxSize, 252);
}
} // End UseNamedOperandTable = 1, VGPRSpill = 1, SchedRW = [WriteVMEM]
}
defm SI_SPILL_V32 : SI_SPILL_VGPR <VGPR_32>;
defm SI_SPILL_V64 : SI_SPILL_VGPR <VReg_64>;
defm SI_SPILL_V96 : SI_SPILL_VGPR <VReg_96>;
defm SI_SPILL_V128 : SI_SPILL_VGPR <VReg_128>;
defm SI_SPILL_V160 : SI_SPILL_VGPR <VReg_160>;
defm SI_SPILL_V192 : SI_SPILL_VGPR <VReg_192>;
defm SI_SPILL_V224 : SI_SPILL_VGPR <VReg_224>;
defm SI_SPILL_V256 : SI_SPILL_VGPR <VReg_256>;
defm SI_SPILL_V288 : SI_SPILL_VGPR <VReg_288>;
defm SI_SPILL_V320 : SI_SPILL_VGPR <VReg_320>;
defm SI_SPILL_V352 : SI_SPILL_VGPR <VReg_352>;
defm SI_SPILL_V384 : SI_SPILL_VGPR <VReg_384>;
defm SI_SPILL_V512 : SI_SPILL_VGPR <VReg_512>;
defm SI_SPILL_V1024 : SI_SPILL_VGPR <VReg_1024>;
defm SI_SPILL_A32 : SI_SPILL_VGPR <AGPR_32, 1>;
defm SI_SPILL_A64 : SI_SPILL_VGPR <AReg_64, 1>;
defm SI_SPILL_A96 : SI_SPILL_VGPR <AReg_96, 1>;
defm SI_SPILL_A128 : SI_SPILL_VGPR <AReg_128, 1>;
defm SI_SPILL_A160 : SI_SPILL_VGPR <AReg_160, 1>;
defm SI_SPILL_A192 : SI_SPILL_VGPR <AReg_192, 1>;
defm SI_SPILL_A224 : SI_SPILL_VGPR <AReg_224, 1>;
defm SI_SPILL_A256 : SI_SPILL_VGPR <AReg_256, 1>;
defm SI_SPILL_A288 : SI_SPILL_VGPR <AReg_288, 1>;
defm SI_SPILL_A320 : SI_SPILL_VGPR <AReg_320, 1>;
defm SI_SPILL_A352 : SI_SPILL_VGPR <AReg_352, 1>;
defm SI_SPILL_A384 : SI_SPILL_VGPR <AReg_384, 1>;
defm SI_SPILL_A512 : SI_SPILL_VGPR <AReg_512, 1>;
defm SI_SPILL_A1024 : SI_SPILL_VGPR <AReg_1024, 1>;
defm SI_SPILL_AV32 : SI_SPILL_VGPR <AV_32, 1>;
defm SI_SPILL_AV64 : SI_SPILL_VGPR <AV_64, 1>;
defm SI_SPILL_AV96 : SI_SPILL_VGPR <AV_96, 1>;
defm SI_SPILL_AV128 : SI_SPILL_VGPR <AV_128, 1>;
defm SI_SPILL_AV160 : SI_SPILL_VGPR <AV_160, 1>;
defm SI_SPILL_AV192 : SI_SPILL_VGPR <AV_192, 1>;
defm SI_SPILL_AV224 : SI_SPILL_VGPR <AV_224, 1>;
defm SI_SPILL_AV256 : SI_SPILL_VGPR <AV_256, 1>;
defm SI_SPILL_AV288 : SI_SPILL_VGPR <AV_288, 1>;
defm SI_SPILL_AV320 : SI_SPILL_VGPR <AV_320, 1>;
defm SI_SPILL_AV352 : SI_SPILL_VGPR <AV_352, 1>;
defm SI_SPILL_AV384 : SI_SPILL_VGPR <AV_384, 1>;
defm SI_SPILL_AV512 : SI_SPILL_VGPR <AV_512, 1>;
defm SI_SPILL_AV1024 : SI_SPILL_VGPR <AV_1024, 1>;
let isConvergent = 1 in {
defm SI_SPILL_WWM_V32 : SI_SPILL_VGPR <VGPR_32>;
defm SI_SPILL_WWM_AV32 : SI_SPILL_VGPR <AV_32, 1>;
}
def SI_PC_ADD_REL_OFFSET : SPseudoInstSI <
(outs SReg_64:$dst),
(ins si_ga:$ptr_lo, si_ga:$ptr_hi),
[(set SReg_64:$dst,
(i64 (SIpc_add_rel_offset tglobaladdr:$ptr_lo, tglobaladdr:$ptr_hi)))]> {
let Defs = [SCC];
}
def : GCNPat <
(SIpc_add_rel_offset tglobaladdr:$ptr_lo, 0),
(SI_PC_ADD_REL_OFFSET $ptr_lo, (i32 0))
>;
def : GCNPat<
(AMDGPUtrap timm:$trapid),
(S_TRAP $trapid)
>;
def : GCNPat<
(AMDGPUelse i1:$src, bb:$target),
(SI_ELSE $src, $target)
>;
def : Pat <
(int_amdgcn_kill i1:$src),
(SI_KILL_I1_PSEUDO SCSrc_i1:$src, 0)
>;
def : Pat <
(int_amdgcn_kill (i1 (not i1:$src))),
(SI_KILL_I1_PSEUDO SCSrc_i1:$src, -1)
>;
def : Pat <
(int_amdgcn_kill (i1 (setcc f32:$src, InlineImmFP32:$imm, cond:$cond))),
(SI_KILL_F32_COND_IMM_PSEUDO VSrc_b32:$src, (bitcast_fpimm_to_i32 $imm), (cond_as_i32imm $cond))
>;
def : Pat <
(int_amdgcn_wqm_demote i1:$src),
(SI_DEMOTE_I1 SCSrc_i1:$src, 0)
>;
def : Pat <
(int_amdgcn_wqm_demote (i1 (not i1:$src))),
(SI_DEMOTE_I1 SCSrc_i1:$src, -1)
>;
// TODO: we could add more variants for other types of conditionals
def : Pat <
(i64 (int_amdgcn_icmp i1:$src, (i1 0), (i32 33))),
(COPY $src) // Return the SGPRs representing i1 src
>;
def : Pat <
(i32 (int_amdgcn_icmp i1:$src, (i1 0), (i32 33))),
(COPY $src) // Return the SGPRs representing i1 src
>;
//===----------------------------------------------------------------------===//
// VOP1 Patterns
//===----------------------------------------------------------------------===//
multiclass f16_fp_Pats<Instruction cvt_f16_f32_inst_e64, Instruction cvt_f32_f16_inst_e64> {
// f16_to_fp patterns
def : GCNPat <
(f32 (any_f16_to_fp i32:$src0)),
(cvt_f32_f16_inst_e64 SRCMODS.NONE, $src0)
>;
def : GCNPat <
(f32 (f16_to_fp (and_oneuse i32:$src0, 0x7fff))),
(cvt_f32_f16_inst_e64 SRCMODS.ABS, $src0)
>;
def : GCNPat <
(f32 (f16_to_fp (i32 (srl_oneuse (and_oneuse i32:$src0, 0x7fff0000), (i32 16))))),
(cvt_f32_f16_inst_e64 SRCMODS.ABS, (i32 (V_LSHRREV_B32_e64 (i32 16), i32:$src0)))
>;
def : GCNPat <
(f32 (f16_to_fp (or_oneuse i32:$src0, 0x8000))),
(cvt_f32_f16_inst_e64 SRCMODS.NEG_ABS, $src0)
>;
def : GCNPat <
(f32 (f16_to_fp (xor_oneuse i32:$src0, 0x8000))),
(cvt_f32_f16_inst_e64 SRCMODS.NEG, $src0)
>;
def : GCNPat <
(f64 (any_fpextend f16:$src)),
(V_CVT_F64_F32_e32 (cvt_f32_f16_inst_e64 SRCMODS.NONE, $src))
>;
// fp_to_fp16 patterns
def : GCNPat <
(i32 (AMDGPUfp_to_f16 (f32 (VOP3Mods f32:$src0, i32:$src0_modifiers)))),
(cvt_f16_f32_inst_e64 $src0_modifiers, f32:$src0)
>;
def : GCNPat <
(i32 (fp_to_sint f16:$src)),
(V_CVT_I32_F32_e32 (cvt_f32_f16_inst_e64 SRCMODS.NONE, VSrc_b32:$src))
>;
def : GCNPat <
(i32 (fp_to_uint f16:$src)),
(V_CVT_U32_F32_e32 (cvt_f32_f16_inst_e64 SRCMODS.NONE, VSrc_b32:$src))
>;
def : GCNPat <
(f16 (sint_to_fp i32:$src)),
(cvt_f16_f32_inst_e64 SRCMODS.NONE, (V_CVT_F32_I32_e32 VSrc_b32:$src))
>;
def : GCNPat <
(f16 (uint_to_fp i32:$src)),
(cvt_f16_f32_inst_e64 SRCMODS.NONE, (V_CVT_F32_U32_e32 VSrc_b32:$src))
>;
// This is only used on targets without half support
// TODO: Introduce strict variant of AMDGPUfp_to_f16 and share custom lowering
def : GCNPat <
(i32 (strict_fp_to_f16 (f32 (VOP3Mods f32:$src0, i32:$src0_modifiers)))),
(cvt_f16_f32_inst_e64 $src0_modifiers, f32:$src0)
>;
}
let SubtargetPredicate = NotHasTrue16BitInsts in
defm : f16_fp_Pats<V_CVT_F16_F32_e64, V_CVT_F32_F16_e64>;
let SubtargetPredicate = HasTrue16BitInsts in
defm : f16_fp_Pats<V_CVT_F16_F32_t16_e64, V_CVT_F32_F16_t16_e64>;
//===----------------------------------------------------------------------===//
// VOP2 Patterns
//===----------------------------------------------------------------------===//
// NoMods pattern used for mac. If there are any source modifiers then it's
// better to select mad instead of mac.
class FMADPat <ValueType vt, Instruction inst>
: GCNPat <(vt (any_fmad (vt (VOP3NoMods vt:$src0)),
(vt (VOP3NoMods vt:$src1)),
(vt (VOP3NoMods vt:$src2)))),
(inst SRCMODS.NONE, $src0, SRCMODS.NONE, $src1,
SRCMODS.NONE, $src2, DSTCLAMP.NONE, DSTOMOD.NONE)
>;
// Prefer mac form when there are no modifiers.
let AddedComplexity = 9 in {
let OtherPredicates = [HasMadMacF32Insts] in
def : FMADPat <f32, V_MAC_F32_e64>;
// Don't allow source modifiers. If there are any source modifiers then it's
// better to select mad instead of mac.
let SubtargetPredicate = isGFX6GFX7GFX10,
OtherPredicates = [HasMadMacF32Insts, NoFP32Denormals] in
def : GCNPat <
(f32 (fadd (AMDGPUfmul_legacy (VOP3NoMods f32:$src0),
(VOP3NoMods f32:$src1)),
(VOP3NoMods f32:$src2))),
(V_MAC_LEGACY_F32_e64 SRCMODS.NONE, $src0, SRCMODS.NONE, $src1,
SRCMODS.NONE, $src2, DSTCLAMP.NONE, DSTOMOD.NONE)
>;
// Don't allow source modifiers. If there are any source modifiers then it's
// better to select fma instead of fmac.
let SubtargetPredicate = HasFmaLegacy32 in
def : GCNPat <
(f32 (int_amdgcn_fma_legacy (VOP3NoMods f32:$src0),
(VOP3NoMods f32:$src1),
(VOP3NoMods f32:$src2))),
(V_FMAC_LEGACY_F32_e64 SRCMODS.NONE, $src0, SRCMODS.NONE, $src1,
SRCMODS.NONE, $src2, DSTCLAMP.NONE, DSTOMOD.NONE)
>;
let SubtargetPredicate = Has16BitInsts in
def : FMADPat <f16, V_MAC_F16_e64>;
} // AddedComplexity = 9
let OtherPredicates = [HasMadMacF32Insts, NoFP32Denormals] in
def : GCNPat <
(f32 (fadd (AMDGPUfmul_legacy (VOP3Mods f32:$src0, i32:$src0_mod),
(VOP3Mods f32:$src1, i32:$src1_mod)),
(VOP3Mods f32:$src2, i32:$src2_mod))),
(V_MAD_LEGACY_F32_e64 $src0_mod, $src0, $src1_mod, $src1,
$src2_mod, $src2, DSTCLAMP.NONE, DSTOMOD.NONE)
>;
class VOPSelectModsPat <ValueType vt> : GCNPat <
(vt (select i1:$src0, (VOP3ModsNonCanonicalizing vt:$src1, i32:$src1_mods),
(VOP3ModsNonCanonicalizing vt:$src2, i32:$src2_mods))),
(V_CNDMASK_B32_e64 FP32InputMods:$src2_mods, VSrc_b32:$src2,
FP32InputMods:$src1_mods, VSrc_b32:$src1, SSrc_i1:$src0)
>;
class VOPSelectPat <ValueType vt> : GCNPat <
(vt (select i1:$src0, vt:$src1, vt:$src2)),
(V_CNDMASK_B32_e64 0, VSrc_b32:$src2, 0, VSrc_b32:$src1, SSrc_i1:$src0)
>;
def : VOPSelectModsPat <i32>;
def : VOPSelectModsPat <f32>;
def : VOPSelectPat <f16>;
def : VOPSelectPat <i16>;
let AddedComplexity = 1 in {
def : GCNPat <
(i32 (add (i32 (DivergentUnaryFrag<ctpop> i32:$popcnt)), i32:$val)),
(V_BCNT_U32_B32_e64 $popcnt, $val)
>;
}
def : GCNPat <
(i32 (DivergentUnaryFrag<ctpop> i32:$popcnt)),
(V_BCNT_U32_B32_e64 VSrc_b32:$popcnt, (i32 0))
>;
def : GCNPat <
(i16 (add (i16 (trunc (i32 (DivergentUnaryFrag<ctpop> i32:$popcnt)))), i16:$val)),
(V_BCNT_U32_B32_e64 $popcnt, $val)
>;
def : GCNPat <
(i64 (DivergentUnaryFrag<ctpop> i64:$src)),
(REG_SEQUENCE VReg_64,
(V_BCNT_U32_B32_e64 (i32 (EXTRACT_SUBREG i64:$src, sub1)),
(i32 (V_BCNT_U32_B32_e64 (i32 (EXTRACT_SUBREG i64:$src, sub0)), (i32 0)))), sub0,
(i32 (V_MOV_B32_e32 (i32 0))), sub1)
>;
/********** ============================================ **********/
/********** Extraction, Insertion, Building and Casting **********/
/********** ============================================ **********/
// Special case for 2 element vectors. REQ_SEQUENCE produces better code
// than an INSERT_SUBREG.
multiclass Insert_Element_V2<RegisterClass RC, ValueType elem_type, ValueType vec_type> {
def : GCNPat <
(insertelt vec_type:$vec, elem_type:$elem, 0),
(REG_SEQUENCE RC, $elem, sub0, (elem_type (EXTRACT_SUBREG $vec, sub1)), sub1)
>;
def : GCNPat <
(insertelt vec_type:$vec, elem_type:$elem, 1),
(REG_SEQUENCE RC, (elem_type (EXTRACT_SUBREG $vec, sub0)), sub0, $elem, sub1)
>;
}
foreach Index = 0-1 in {
def Extract_Element_v2i32_#Index : Extract_Element <
i32, v2i32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Extract_Element_v2f32_#Index : Extract_Element <
f32, v2f32, Index, !cast<SubRegIndex>(sub#Index)
>;
}
defm : Insert_Element_V2 <SReg_64, i32, v2i32>;
defm : Insert_Element_V2 <SReg_64, f32, v2f32>;
foreach Index = 0-2 in {
def Extract_Element_v3i32_#Index : Extract_Element <
i32, v3i32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Insert_Element_v3i32_#Index : Insert_Element <
i32, v3i32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Extract_Element_v3f32_#Index : Extract_Element <
f32, v3f32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Insert_Element_v3f32_#Index : Insert_Element <
f32, v3f32, Index, !cast<SubRegIndex>(sub#Index)
>;
}
foreach Index = 0-3 in {
def Extract_Element_v4i32_#Index : Extract_Element <
i32, v4i32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Insert_Element_v4i32_#Index : Insert_Element <
i32, v4i32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Extract_Element_v4f32_#Index : Extract_Element <
f32, v4f32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Insert_Element_v4f32_#Index : Insert_Element <
f32, v4f32, Index, !cast<SubRegIndex>(sub#Index)
>;
}
foreach Index = 0-4 in {
def Extract_Element_v5i32_#Index : Extract_Element <
i32, v5i32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Insert_Element_v5i32_#Index : Insert_Element <
i32, v5i32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Extract_Element_v5f32_#Index : Extract_Element <
f32, v5f32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Insert_Element_v5f32_#Index : Insert_Element <
f32, v5f32, Index, !cast<SubRegIndex>(sub#Index)
>;
}
foreach Index = 0-5 in {
def Extract_Element_v6i32_#Index : Extract_Element <
i32, v6i32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Insert_Element_v6i32_#Index : Insert_Element <
i32, v6i32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Extract_Element_v6f32_#Index : Extract_Element <
f32, v6f32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Insert_Element_v6f32_#Index : Insert_Element <
f32, v6f32, Index, !cast<SubRegIndex>(sub#Index)
>;
}
foreach Index = 0-6 in {
def Extract_Element_v7i32_#Index : Extract_Element <
i32, v7i32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Insert_Element_v7i32_#Index : Insert_Element <
i32, v7i32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Extract_Element_v7f32_#Index : Extract_Element <
f32, v7f32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Insert_Element_v7f32_#Index : Insert_Element <
f32, v7f32, Index, !cast<SubRegIndex>(sub#Index)
>;
}
foreach Index = 0-7 in {
def Extract_Element_v8i32_#Index : Extract_Element <
i32, v8i32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Insert_Element_v8i32_#Index : Insert_Element <
i32, v8i32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Extract_Element_v8f32_#Index : Extract_Element <
f32, v8f32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Insert_Element_v8f32_#Index : Insert_Element <
f32, v8f32, Index, !cast<SubRegIndex>(sub#Index)
>;
}
foreach Index = 0-8 in {
def Extract_Element_v9i32_#Index : Extract_Element <
i32, v9i32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Insert_Element_v9i32_#Index : Insert_Element <
i32, v9i32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Extract_Element_v9f32_#Index : Extract_Element <
f32, v9f32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Insert_Element_v9f32_#Index : Insert_Element <
f32, v9f32, Index, !cast<SubRegIndex>(sub#Index)
>;
}
foreach Index = 0-9 in {
def Extract_Element_v10i32_#Index : Extract_Element <
i32, v10i32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Insert_Element_v10i32_#Index : Insert_Element <
i32, v10i32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Extract_Element_v10f32_#Index : Extract_Element <
f32, v10f32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Insert_Element_v10f32_#Index : Insert_Element <
f32, v10f32, Index, !cast<SubRegIndex>(sub#Index)
>;
}
foreach Index = 0-10 in {
def Extract_Element_v11i32_#Index : Extract_Element <
i32, v11i32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Insert_Element_v11i32_#Index : Insert_Element <
i32, v11i32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Extract_Element_v11f32_#Index : Extract_Element <
f32, v11f32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Insert_Element_v11f32_#Index : Insert_Element <
f32, v11f32, Index, !cast<SubRegIndex>(sub#Index)
>;
}
foreach Index = 0-11 in {
def Extract_Element_v12i32_#Index : Extract_Element <
i32, v12i32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Insert_Element_v12i32_#Index : Insert_Element <
i32, v12i32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Extract_Element_v12f32_#Index : Extract_Element <
f32, v12f32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Insert_Element_v12f32_#Index : Insert_Element <
f32, v12f32, Index, !cast<SubRegIndex>(sub#Index)
>;
}
foreach Index = 0-15 in {
def Extract_Element_v16i32_#Index : Extract_Element <
i32, v16i32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Insert_Element_v16i32_#Index : Insert_Element <
i32, v16i32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Extract_Element_v16f32_#Index : Extract_Element <
f32, v16f32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Insert_Element_v16f32_#Index : Insert_Element <
f32, v16f32, Index, !cast<SubRegIndex>(sub#Index)
>;
}
foreach Index = 0-31 in {
def Extract_Element_v32i32_#Index : Extract_Element <
i32, v32i32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Insert_Element_v32i32_#Index : Insert_Element <
i32, v32i32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Extract_Element_v32f32_#Index : Extract_Element <
f32, v32f32, Index, !cast<SubRegIndex>(sub#Index)
>;
def Insert_Element_v32f32_#Index : Insert_Element <
f32, v32f32, Index, !cast<SubRegIndex>(sub#Index)
>;
}
// FIXME: Why do only some of these type combinations for SReg and
// VReg?
// 16-bit bitcast
def : BitConvert <i16, f16, VGPR_32>;
def : BitConvert <f16, i16, VGPR_32>;
def : BitConvert <f16, bf16, VGPR_32>;
def : BitConvert <bf16, f16, VGPR_32>;
def : BitConvert <i16, f16, SReg_32>;
def : BitConvert <f16, i16, SReg_32>;
def : BitConvert <f16, bf16, SReg_32>;
def : BitConvert <bf16, f16, SReg_32>;
def : BitConvert <i16, bf16, VGPR_32>;
def : BitConvert <bf16, i16, VGPR_32>;
def : BitConvert <i16, bf16, SReg_32>;
def : BitConvert <bf16, i16, SReg_32>;
// 32-bit bitcast
def : BitConvert <i32, f32, VGPR_32>;
def : BitConvert <f32, i32, VGPR_32>;
def : BitConvert <i32, f32, SReg_32>;
def : BitConvert <f32, i32, SReg_32>;
def : BitConvert <v2i16, i32, SReg_32>;
def : BitConvert <i32, v2i16, SReg_32>;
def : BitConvert <v2f16, i32, SReg_32>;
def : BitConvert <i32, v2f16, SReg_32>;
def : BitConvert <v2i16, v2f16, SReg_32>;
def : BitConvert <v2f16, v2i16, SReg_32>;
def : BitConvert <v2f16, f32, SReg_32>;
def : BitConvert <f32, v2f16, SReg_32>;
def : BitConvert <v2i16, f32, SReg_32>;
def : BitConvert <f32, v2i16, SReg_32>;
def : BitConvert <v2bf16, i32, SReg_32>;
def : BitConvert <i32, v2bf16, SReg_32>;
def : BitConvert <v2bf16, i32, VGPR_32>;
def : BitConvert <i32, v2bf16, VGPR_32>;
def : BitConvert <v2bf16, v2i16, SReg_32>;
def : BitConvert <v2i16, v2bf16, SReg_32>;
def : BitConvert <v2bf16, v2i16, VGPR_32>;
def : BitConvert <v2i16, v2bf16, VGPR_32>;
def : BitConvert <v2bf16, v2f16, SReg_32>;
def : BitConvert <v2f16, v2bf16, SReg_32>;
def : BitConvert <v2bf16, v2f16, VGPR_32>;
def : BitConvert <v2f16, v2bf16, VGPR_32>;
def : BitConvert <f32, v2bf16, VGPR_32>;
def : BitConvert <v2bf16, f32, VGPR_32>;
def : BitConvert <f32, v2bf16, SReg_32>;
def : BitConvert <v2bf16, f32, SReg_32>;
// 64-bit bitcast
def : BitConvert <i64, f64, VReg_64>;
def : BitConvert <f64, i64, VReg_64>;
def : BitConvert <v2i32, v2f32, VReg_64>;
def : BitConvert <v2f32, v2i32, VReg_64>;
def : BitConvert <i64, v2i32, VReg_64>;
def : BitConvert <v2i32, i64, VReg_64>;
def : BitConvert <i64, v2f32, VReg_64>;
def : BitConvert <v2f32, i64, VReg_64>;
def : BitConvert <f64, v2f32, VReg_64>;
def : BitConvert <v2f32, f64, VReg_64>;
def : BitConvert <f64, v2i32, VReg_64>;
def : BitConvert <v2i32, f64, VReg_64>;
def : BitConvert <v4i16, v4f16, VReg_64>;
def : BitConvert <v4f16, v4i16, VReg_64>;
def : BitConvert <v4bf16, v2i32, VReg_64>;
def : BitConvert <v2i32, v4bf16, VReg_64>;
def : BitConvert <v4bf16, i64, VReg_64>;
def : BitConvert <i64, v4bf16, VReg_64>;
def : BitConvert <v4bf16, v4i16, VReg_64>;
def : BitConvert <v4i16, v4bf16, VReg_64>;
def : BitConvert <v4bf16, v4f16, VReg_64>;
def : BitConvert <v4f16, v4bf16, VReg_64>;
def : BitConvert <v4bf16, v2f32, VReg_64>;
def : BitConvert <v2f32, v4bf16, VReg_64>;
def : BitConvert <v4bf16, f64, VReg_64>;
def : BitConvert <f64, v4bf16, VReg_64>;
// FIXME: Make SGPR
def : BitConvert <v2i32, v4f16, VReg_64>;
def : BitConvert <v4f16, v2i32, VReg_64>;
def : BitConvert <v2i32, v4f16, VReg_64>;
def : BitConvert <v2i32, v4i16, VReg_64>;
def : BitConvert <v4i16, v2i32, VReg_64>;
def : BitConvert <v2f32, v4f16, VReg_64>;
def : BitConvert <v4f16, v2f32, VReg_64>;
def : BitConvert <v2f32, v4i16, VReg_64>;
def : BitConvert <v4i16, v2f32, VReg_64>;
def : BitConvert <v4i16, f64, VReg_64>;
def : BitConvert <v4f16, f64, VReg_64>;
def : BitConvert <f64, v4i16, VReg_64>;
def : BitConvert <f64, v4f16, VReg_64>;
def : BitConvert <v4i16, i64, VReg_64>;
def : BitConvert <v4f16, i64, VReg_64>;
def : BitConvert <i64, v4i16, VReg_64>;
def : BitConvert <i64, v4f16, VReg_64>;
def : BitConvert <v4i32, v4f32, VReg_128>;
def : BitConvert <v4f32, v4i32, VReg_128>;
// 96-bit bitcast
def : BitConvert <v3i32, v3f32, SGPR_96>;
def : BitConvert <v3f32, v3i32, SGPR_96>;
// 128-bit bitcast
def : BitConvert <v2i64, v4i32, SReg_128>;
def : BitConvert <v4i32, v2i64, SReg_128>;
def : BitConvert <v2f64, v4f32, VReg_128>;
def : BitConvert <v2f64, v4i32, VReg_128>;
def : BitConvert <v4f32, v2f64, VReg_128>;
def : BitConvert <v4i32, v2f64, VReg_128>;
def : BitConvert <v2i64, v2f64, VReg_128>;
def : BitConvert <v2f64, v2i64, VReg_128>;
def : BitConvert <v4f32, v2i64, VReg_128>;
def : BitConvert <v2i64, v4f32, VReg_128>;
def : BitConvert <v8i16, v4i32, SReg_128>;
def : BitConvert <v4i32, v8i16, SReg_128>;
def : BitConvert <v8f16, v4f32, VReg_128>;
def : BitConvert <v8f16, v4i32, VReg_128>;
def : BitConvert <v4f32, v8f16, VReg_128>;
def : BitConvert <v4i32, v8f16, VReg_128>;
def : BitConvert <v8i16, v8f16, VReg_128>;
def : BitConvert <v8f16, v8i16, VReg_128>;
def : BitConvert <v4f32, v8i16, VReg_128>;
def : BitConvert <v8i16, v4f32, VReg_128>;
def : BitConvert <v8i16, v8f16, SReg_128>;
def : BitConvert <v8i16, v2i64, SReg_128>;
def : BitConvert <v8i16, v2f64, SReg_128>;
def : BitConvert <v8f16, v2i64, SReg_128>;
def : BitConvert <v8f16, v2f64, SReg_128>;
def : BitConvert <v8f16, v8i16, SReg_128>;
def : BitConvert <v2i64, v8i16, SReg_128>;
def : BitConvert <v2f64, v8i16, SReg_128>;
def : BitConvert <v2i64, v8f16, SReg_128>;
def : BitConvert <v2f64, v8f16, SReg_128>;
def : BitConvert <v4i32, v8bf16, SReg_128>;
def : BitConvert <v8bf16, v4i32, SReg_128>;
def : BitConvert <v4i32, v8bf16, VReg_128>;
def : BitConvert <v8bf16, v4i32, VReg_128>;
def : BitConvert <v4f32, v8bf16, SReg_128>;
def : BitConvert <v8bf16, v4f32, SReg_128>;
def : BitConvert <v4f32, v8bf16, VReg_128>;
def : BitConvert <v8bf16, v4f32, VReg_128>;
def : BitConvert <v8i16, v8bf16, SReg_128>;
def : BitConvert <v8bf16, v8i16, SReg_128>;
def : BitConvert <v8i16, v8bf16, VReg_128>;
def : BitConvert <v8bf16, v8i16, VReg_128>;
def : BitConvert <v8f16, v8bf16, SReg_128>;
def : BitConvert <v8bf16, v8f16, SReg_128>;
def : BitConvert <v8f16, v8bf16, VReg_128>;
def : BitConvert <v8bf16, v8f16, VReg_128>;
def : BitConvert <v2f64, v8bf16, SReg_128>;
def : BitConvert <v8bf16, v2f64, SReg_128>;
def : BitConvert <v2f64, v8bf16, VReg_128>;
def : BitConvert <v8bf16, v2f64, VReg_128>;
def : BitConvert <v2i64, v8bf16, SReg_128>;
def : BitConvert <v8bf16, v2i64, SReg_128>;
def : BitConvert <v2i64, v8bf16, VReg_128>;
def : BitConvert <v8bf16, v2i64, VReg_128>;
// 160-bit bitcast
def : BitConvert <v5i32, v5f32, SReg_160>;
def : BitConvert <v5f32, v5i32, SReg_160>;
def : BitConvert <v5i32, v5f32, VReg_160>;
def : BitConvert <v5f32, v5i32, VReg_160>;
// 192-bit bitcast
def : BitConvert <v6i32, v6f32, SReg_192>;
def : BitConvert <v6f32, v6i32, SReg_192>;
def : BitConvert <v6i32, v6f32, VReg_192>;
def : BitConvert <v6f32, v6i32, VReg_192>;
def : BitConvert <v3i64, v3f64, VReg_192>;
def : BitConvert <v3f64, v3i64, VReg_192>;
def : BitConvert <v3i64, v6i32, VReg_192>;
def : BitConvert <v3i64, v6f32, VReg_192>;
def : BitConvert <v3f64, v6i32, VReg_192>;
def : BitConvert <v3f64, v6f32, VReg_192>;
def : BitConvert <v6i32, v3i64, VReg_192>;
def : BitConvert <v6f32, v3i64, VReg_192>;
def : BitConvert <v6i32, v3f64, VReg_192>;
def : BitConvert <v6f32, v3f64, VReg_192>;
// 224-bit bitcast
def : BitConvert <v7i32, v7f32, SReg_224>;
def : BitConvert <v7f32, v7i32, SReg_224>;
def : BitConvert <v7i32, v7f32, VReg_224>;
def : BitConvert <v7f32, v7i32, VReg_224>;
// 256-bit bitcast
def : BitConvert <v8i32, v8f32, SReg_256>;
def : BitConvert <v8f32, v8i32, SReg_256>;
def : BitConvert <v8i32, v8f32, VReg_256>;
def : BitConvert <v8f32, v8i32, VReg_256>;
def : BitConvert <v4i64, v4f64, VReg_256>;
def : BitConvert <v4f64, v4i64, VReg_256>;
def : BitConvert <v4i64, v8i32, VReg_256>;
def : BitConvert <v4i64, v8f32, VReg_256>;
def : BitConvert <v4f64, v8i32, VReg_256>;
def : BitConvert <v4f64, v8f32, VReg_256>;
def : BitConvert <v8i32, v4i64, VReg_256>;
def : BitConvert <v8f32, v4i64, VReg_256>;
def : BitConvert <v8i32, v4f64, VReg_256>;
def : BitConvert <v8f32, v4f64, VReg_256>;
def : BitConvert <v16i16, v16f16, SReg_256>;
def : BitConvert <v16f16, v16i16, SReg_256>;
def : BitConvert <v16i16, v16f16, VReg_256>;
def : BitConvert <v16f16, v16i16, VReg_256>;
def : BitConvert <v16f16, v8i32, VReg_256>;
def : BitConvert <v16i16, v8i32, VReg_256>;
def : BitConvert <v16f16, v8f32, VReg_256>;
def : BitConvert <v16i16, v8f32, VReg_256>;
def : BitConvert <v8i32, v16f16, VReg_256>;
def : BitConvert <v8i32, v16i16, VReg_256>;
def : BitConvert <v8f32, v16f16, VReg_256>;
def : BitConvert <v8f32, v16i16, VReg_256>;
def : BitConvert <v16f16, v4i64, VReg_256>;
def : BitConvert <v16i16, v4i64, VReg_256>;
def : BitConvert <v16f16, v4f64, VReg_256>;
def : BitConvert <v16i16, v4f64, VReg_256>;
def : BitConvert <v4i64, v16f16, VReg_256>;
def : BitConvert <v4i64, v16i16, VReg_256>;
def : BitConvert <v4f64, v16f16, VReg_256>;
def : BitConvert <v4f64, v16i16, VReg_256>;
def : BitConvert <v8i32, v16bf16, VReg_256>;
def : BitConvert <v16bf16, v8i32, VReg_256>;
def : BitConvert <v8f32, v16bf16, VReg_256>;
def : BitConvert <v16bf16, v8f32, VReg_256>;
def : BitConvert <v4i64, v16bf16, VReg_256>;
def : BitConvert <v16bf16, v4i64, VReg_256>;
def : BitConvert <v4f64, v16bf16, VReg_256>;
def : BitConvert <v16bf16, v4f64, VReg_256>;
def : BitConvert <v16i16, v16bf16, SReg_256>;
def : BitConvert <v16bf16, v16i16, SReg_256>;
def : BitConvert <v16i16, v16bf16, VReg_256>;
def : BitConvert <v16bf16, v16i16, VReg_256>;
def : BitConvert <v16f16, v16bf16, SReg_256>;
def : BitConvert <v16bf16, v16f16, SReg_256>;
def : BitConvert <v16f16, v16bf16, VReg_256>;
def : BitConvert <v16bf16, v16f16, VReg_256>;
// 288-bit bitcast
def : BitConvert <v9i32, v9f32, SReg_288>;
def : BitConvert <v9f32, v9i32, SReg_288>;
def : BitConvert <v9i32, v9f32, VReg_288>;
def : BitConvert <v9f32, v9i32, VReg_288>;
// 320-bit bitcast
def : BitConvert <v10i32, v10f32, SReg_320>;
def : BitConvert <v10f32, v10i32, SReg_320>;
def : BitConvert <v10i32, v10f32, VReg_320>;
def : BitConvert <v10f32, v10i32, VReg_320>;
// 320-bit bitcast
def : BitConvert <v11i32, v11f32, SReg_352>;
def : BitConvert <v11f32, v11i32, SReg_352>;
def : BitConvert <v11i32, v11f32, VReg_352>;
def : BitConvert <v11f32, v11i32, VReg_352>;
// 384-bit bitcast
def : BitConvert <v12i32, v12f32, SReg_384>;
def : BitConvert <v12f32, v12i32, SReg_384>;
def : BitConvert <v12i32, v12f32, VReg_384>;
def : BitConvert <v12f32, v12i32, VReg_384>;
// 512-bit bitcast
def : BitConvert <v32f16, v32i16, VReg_512>;
def : BitConvert <v32i16, v32f16, VReg_512>;
def : BitConvert <v32f16, v16i32, VReg_512>;
def : BitConvert <v32f16, v16f32, VReg_512>;
def : BitConvert <v16f32, v32f16, VReg_512>;
def : BitConvert <v16i32, v32f16, VReg_512>;
def : BitConvert <v32i16, v16i32, VReg_512>;
def : BitConvert <v32i16, v16f32, VReg_512>;
def : BitConvert <v16f32, v32i16, VReg_512>;
def : BitConvert <v16i32, v32i16, VReg_512>;
def : BitConvert <v16i32, v16f32, VReg_512>;
def : BitConvert <v16f32, v16i32, VReg_512>;
def : BitConvert <v8i64, v8f64, VReg_512>;
def : BitConvert <v8f64, v8i64, VReg_512>;
def : BitConvert <v8i64, v16i32, VReg_512>;
def : BitConvert <v8f64, v16i32, VReg_512>;
def : BitConvert <v16i32, v8i64, VReg_512>;
def : BitConvert <v16i32, v8f64, VReg_512>;
def : BitConvert <v8i64, v16f32, VReg_512>;
def : BitConvert <v8f64, v16f32, VReg_512>;
def : BitConvert <v16f32, v8i64, VReg_512>;
def : BitConvert <v16f32, v8f64, VReg_512>;
def : BitConvert <v32bf16, v32i16, VReg_512>;
def : BitConvert <v32i16, v32bf16, VReg_512>;
def : BitConvert <v32bf16, v32i16, SReg_512>;
def : BitConvert <v32i16, v32bf16, SReg_512>;
def : BitConvert <v32bf16, v32f16, VReg_512>;
def : BitConvert <v32f16, v32bf16, VReg_512>;
def : BitConvert <v32bf16, v32f16, SReg_512>;
def : BitConvert <v32f16, v32bf16, SReg_512>;
def : BitConvert <v32bf16, v16i32, VReg_512>;
def : BitConvert <v16i32, v32bf16, VReg_512>;
def : BitConvert <v32bf16, v16i32, SReg_512>;
def : BitConvert <v16i32, v32bf16, SReg_512>;
def : BitConvert <v32bf16, v16f32, VReg_512>;
def : BitConvert <v16f32, v32bf16, VReg_512>;
def : BitConvert <v32bf16, v16f32, SReg_512>;
def : BitConvert <v16f32, v32bf16, SReg_512>;
def : BitConvert <v32bf16, v8f64, VReg_512>;
def : BitConvert <v8f64, v32bf16, VReg_512>;
def : BitConvert <v32bf16, v8f64, SReg_512>;
def : BitConvert <v8f64, v32bf16, SReg_512>;
def : BitConvert <v32bf16, v8i64, VReg_512>;
def : BitConvert <v8i64, v32bf16, VReg_512>;
def : BitConvert <v32bf16, v8i64, SReg_512>;
def : BitConvert <v8i64, v32bf16, SReg_512>;
// 1024-bit bitcast
def : BitConvert <v32i32, v32f32, VReg_1024>;
def : BitConvert <v32f32, v32i32, VReg_1024>;
def : BitConvert <v16i64, v16f64, VReg_1024>;
def : BitConvert <v16f64, v16i64, VReg_1024>;
def : BitConvert <v16i64, v32i32, VReg_1024>;
def : BitConvert <v32i32, v16i64, VReg_1024>;
def : BitConvert <v16f64, v32f32, VReg_1024>;
def : BitConvert <v32f32, v16f64, VReg_1024>;
def : BitConvert <v16i64, v32f32, VReg_1024>;
def : BitConvert <v32i32, v16f64, VReg_1024>;
def : BitConvert <v16f64, v32i32, VReg_1024>;
def : BitConvert <v32f32, v16i64, VReg_1024>;
/********** =================== **********/
/********** Src & Dst modifiers **********/
/********** =================== **********/
// If denormals are not enabled, it only impacts the compare of the
// inputs. The output result is not flushed.
class ClampPat<Instruction inst, ValueType vt> : GCNPat <
(vt (AMDGPUclamp (VOP3Mods vt:$src0, i32:$src0_modifiers))),
(inst i32:$src0_modifiers, vt:$src0,
i32:$src0_modifiers, vt:$src0, DSTCLAMP.ENABLE, DSTOMOD.NONE)
>;
def : ClampPat<V_MAX_F32_e64, f32>;
def : ClampPat<V_MAX_F64_e64, f64>;
let SubtargetPredicate = NotHasTrue16BitInsts in
def : ClampPat<V_MAX_F16_e64, f16>;
let SubtargetPredicate = UseRealTrue16Insts in
def : ClampPat<V_MAX_F16_t16_e64, f16>;
let SubtargetPredicate = UseFakeTrue16Insts in
def : ClampPat<V_MAX_F16_fake16_e64, f16>;
let SubtargetPredicate = HasVOP3PInsts in {
def : GCNPat <
(v2f16 (AMDGPUclamp (VOP3PMods v2f16:$src0, i32:$src0_modifiers))),
(V_PK_MAX_F16 $src0_modifiers, $src0,
$src0_modifiers, $src0, DSTCLAMP.ENABLE)
>;
}
/********** ================================ **********/
/********** Floating point absolute/negative **********/
/********** ================================ **********/
def : GCNPat <
(UniformUnaryFrag<fneg> (fabs (f32 SReg_32:$src))),
(S_OR_B32 SReg_32:$src, (S_MOV_B32 (i32 0x80000000))) // Set sign bit
>;
def : GCNPat <
(UniformUnaryFrag<fabs> (f32 SReg_32:$src)),
(S_AND_B32 SReg_32:$src, (S_MOV_B32 (i32 0x7fffffff)))
>;
def : GCNPat <
(UniformUnaryFrag<fneg> (f32 SReg_32:$src)),
(S_XOR_B32 SReg_32:$src, (S_MOV_B32 (i32 0x80000000)))
>;
def : GCNPat <
(UniformUnaryFrag<fneg> (f16 SReg_32:$src)),
(S_XOR_B32 SReg_32:$src, (S_MOV_B32 (i32 0x00008000)))
>;
def : GCNPat <
(UniformUnaryFrag<fabs> (f16 SReg_32:$src)),
(S_AND_B32 SReg_32:$src, (S_MOV_B32 (i32 0x00007fff)))
>;
def : GCNPat <
(UniformUnaryFrag<fneg> (fabs (f16 SReg_32:$src))),
(S_OR_B32 SReg_32:$src, (S_MOV_B32 (i32 0x00008000))) // Set sign bit
>;
def : GCNPat <
(UniformUnaryFrag<fneg> (v2f16 SReg_32:$src)),
(S_XOR_B32 SReg_32:$src, (S_MOV_B32 (i32 0x80008000)))
>;
def : GCNPat <
(UniformUnaryFrag<fabs> (v2f16 SReg_32:$src)),
(S_AND_B32 SReg_32:$src, (S_MOV_B32 (i32 0x7fff7fff)))
>;
// This is really (fneg (fabs v2f16:$src))
//
// fabs is not reported as free because there is modifier for it in
// VOP3P instructions, so it is turned into the bit op.
def : GCNPat <
(UniformUnaryFrag<fneg> (v2f16 (bitconvert (and_oneuse (i32 SReg_32:$src), 0x7fff7fff)))),
(S_OR_B32 SReg_32:$src, (S_MOV_B32 (i32 0x80008000))) // Set sign bit
>;
def : GCNPat <
(UniformUnaryFrag<fneg> (v2f16 (fabs SReg_32:$src))),
(S_OR_B32 SReg_32:$src, (S_MOV_B32 (i32 0x80008000))) // Set sign bit
>;
// COPY_TO_REGCLASS is needed to avoid using SCC from S_XOR_B32 instead
// of the real value.
def : GCNPat <
(UniformUnaryFrag<fneg> (v2f32 SReg_64:$src)),
(v2f32 (REG_SEQUENCE SReg_64,
(f32 (COPY_TO_REGCLASS (S_XOR_B32 (i32 (EXTRACT_SUBREG $src, sub0)),
(i32 (S_MOV_B32 (i32 0x80000000)))),
SReg_32)), sub0,
(f32 (COPY_TO_REGCLASS (S_XOR_B32 (i32 (EXTRACT_SUBREG $src, sub1)),
(i32 (S_MOV_B32 (i32 0x80000000)))),
SReg_32)), sub1))
>;
def : GCNPat <
(UniformUnaryFrag<fabs> (v2f32 SReg_64:$src)),
(v2f32 (REG_SEQUENCE SReg_64,
(f32 (COPY_TO_REGCLASS (S_AND_B32 (i32 (EXTRACT_SUBREG $src, sub0)),
(i32 (S_MOV_B32 (i32 0x7fffffff)))),
SReg_32)), sub0,
(f32 (COPY_TO_REGCLASS (S_AND_B32 (i32 (EXTRACT_SUBREG $src, sub1)),
(i32 (S_MOV_B32 (i32 0x7fffffff)))),
SReg_32)), sub1))
>;
def : GCNPat <
(UniformUnaryFrag<fneg> (fabs (v2f32 SReg_64:$src))),
(v2f32 (REG_SEQUENCE SReg_64,
(f32 (COPY_TO_REGCLASS (S_OR_B32 (i32 (EXTRACT_SUBREG $src, sub0)),
(i32 (S_MOV_B32 (i32 0x80000000)))),
SReg_32)), sub0,
(f32 (COPY_TO_REGCLASS (S_OR_B32 (i32 (EXTRACT_SUBREG $src, sub1)),
(i32 (S_MOV_B32 (i32 0x80000000)))),
SReg_32)), sub1))
>;
// FIXME: Use S_BITSET0_B32/B64?
def : GCNPat <
(UniformUnaryFrag<fabs> (f64 SReg_64:$src)),
(REG_SEQUENCE SReg_64,
(i32 (EXTRACT_SUBREG SReg_64:$src, sub0)),
sub0,
(i32 (COPY_TO_REGCLASS (S_AND_B32 (i32 (EXTRACT_SUBREG SReg_64:$src, sub1)),
(S_MOV_B32 (i32 0x7fffffff))), SReg_32)), // Set sign bit.
sub1)
>;
def : GCNPat <
(UniformUnaryFrag<fneg> (f64 SReg_64:$src)),
(REG_SEQUENCE SReg_64,
(i32 (EXTRACT_SUBREG SReg_64:$src, sub0)),
sub0,
(i32 (COPY_TO_REGCLASS (S_XOR_B32 (i32 (EXTRACT_SUBREG SReg_64:$src, sub1)),
(i32 (S_MOV_B32 (i32 0x80000000)))), SReg_32)),
sub1)
>;
def : GCNPat <
(UniformUnaryFrag<fneg> (fabs (f64 SReg_64:$src))),
(REG_SEQUENCE SReg_64,
(i32 (EXTRACT_SUBREG SReg_64:$src, sub0)),
sub0,
(i32 (COPY_TO_REGCLASS (S_OR_B32 (i32 (EXTRACT_SUBREG SReg_64:$src, sub1)),
(S_MOV_B32 (i32 0x80000000))), SReg_32)),// Set sign bit.
sub1)
>;
def : GCNPat <
(fneg (fabs (f32 VGPR_32:$src))),
(V_OR_B32_e64 (S_MOV_B32 (i32 0x80000000)), VGPR_32:$src) // Set sign bit
>;
def : GCNPat <
(fabs (f32 VGPR_32:$src)),
(V_AND_B32_e64 (S_MOV_B32 (i32 0x7fffffff)), VGPR_32:$src)
>;
def : GCNPat <
(fneg (f32 VGPR_32:$src)),
(V_XOR_B32_e64 (S_MOV_B32 (i32 0x80000000)), VGPR_32:$src)
>;
def : GCNPat <
(fabs (f16 VGPR_32:$src)),
(V_AND_B32_e64 (S_MOV_B32 (i32 0x00007fff)), VGPR_32:$src)
>;
def : GCNPat <
(fneg (f16 VGPR_32:$src)),
(V_XOR_B32_e64 (S_MOV_B32 (i32 0x00008000)), VGPR_32:$src)
>;
def : GCNPat <
(fneg (fabs (f16 VGPR_32:$src))),
(V_OR_B32_e64 (S_MOV_B32 (i32 0x00008000)), VGPR_32:$src) // Set sign bit
>;
def : GCNPat <
(fneg (v2f16 VGPR_32:$src)),
(V_XOR_B32_e64 (S_MOV_B32 (i32 0x80008000)), VGPR_32:$src)
>;
def : GCNPat <
(fabs (v2f16 VGPR_32:$src)),
(V_AND_B32_e64 (S_MOV_B32 (i32 0x7fff7fff)), VGPR_32:$src)
>;
def : GCNPat <
(fneg (v2f16 (fabs VGPR_32:$src))),
(V_OR_B32_e64 (S_MOV_B32 (i32 0x80008000)), VGPR_32:$src)
>;
def : GCNPat <
(fabs (f64 VReg_64:$src)),
(REG_SEQUENCE VReg_64,
(i32 (EXTRACT_SUBREG VReg_64:$src, sub0)),
sub0,
(V_AND_B32_e64 (i32 (S_MOV_B32 (i32 0x7fffffff))),
(i32 (EXTRACT_SUBREG VReg_64:$src, sub1))),
sub1)
>;
def : GCNPat <
(fneg (f64 VReg_64:$src)),
(REG_SEQUENCE VReg_64,
(i32 (EXTRACT_SUBREG VReg_64:$src, sub0)),
sub0,
(V_XOR_B32_e64 (i32 (S_MOV_B32 (i32 0x80000000))),
(i32 (EXTRACT_SUBREG VReg_64:$src, sub1))),
sub1)
>;
def : GCNPat <
(fneg (fabs (f64 VReg_64:$src))),
(REG_SEQUENCE VReg_64,
(i32 (EXTRACT_SUBREG VReg_64:$src, sub0)),
sub0,
(V_OR_B32_e64 (i32 (S_MOV_B32 (i32 0x80000000))),
(i32 (EXTRACT_SUBREG VReg_64:$src, sub1))),
sub1)
>;
def : GCNPat <
(DivergentUnaryFrag<fneg> (v2f32 VReg_64:$src)),
(V_PK_ADD_F32 11 /* OP_SEL_1 | NEG_LO | HEG_HI */, VReg_64:$src,
11 /* OP_SEL_1 | NEG_LO | HEG_HI */, 0,
0, 0, 0, 0, 0)
> {
let SubtargetPredicate = HasPackedFP32Ops;
}
foreach fp16vt = [f16, bf16] in {
def : GCNPat <
(fcopysign fp16vt:$src0, fp16vt:$src1),
(V_BFI_B32_e64 (S_MOV_B32 (i32 0x00007fff)), $src0, $src1)
>;
def : GCNPat <
(fcopysign f32:$src0, fp16vt:$src1),
(V_BFI_B32_e64 (S_MOV_B32 (i32 0x7fffffff)), $src0,
(V_LSHLREV_B32_e64 (i32 16), $src1))
>;
def : GCNPat <
(fcopysign f64:$src0, fp16vt:$src1),
(REG_SEQUENCE SReg_64,
(i32 (EXTRACT_SUBREG $src0, sub0)), sub0,
(V_BFI_B32_e64 (S_MOV_B32 (i32 0x7fffffff)), (i32 (EXTRACT_SUBREG $src0, sub1)),
(V_LSHLREV_B32_e64 (i32 16), $src1)), sub1)
>;
def : GCNPat <
(fcopysign fp16vt:$src0, f32:$src1),
(V_BFI_B32_e64 (S_MOV_B32 (i32 0x00007fff)), $src0,
(V_LSHRREV_B32_e64 (i32 16), $src1))
>;
def : GCNPat <
(fcopysign fp16vt:$src0, f64:$src1),
(V_BFI_B32_e64 (S_MOV_B32 (i32 0x00007fff)), $src0,
(V_LSHRREV_B32_e64 (i32 16), (EXTRACT_SUBREG $src1, sub1)))
>;
} // End foreach fp16vt = [f16, bf16]
/********** ================== **********/
/********** Immediate Patterns **********/
/********** ================== **********/
def : GCNPat <
(VGPRImm<(i32 imm)>:$imm),
(V_MOV_B32_e32 imm:$imm)
>;
def : GCNPat <
(VGPRImm<(f32 fpimm)>:$imm),
(V_MOV_B32_e32 (f32 (bitcast_fpimm_to_i32 $imm)))
>;
def : GCNPat <
(i32 imm:$imm),
(S_MOV_B32 imm:$imm)
>;
def : GCNPat <
(VGPRImm<(SIlds tglobaladdr:$ga)>),
(V_MOV_B32_e32 $ga)
>;
def : GCNPat <
(SIlds tglobaladdr:$ga),
(S_MOV_B32 $ga)
>;
// FIXME: Workaround for ordering issue with peephole optimizer where
// a register class copy interferes with immediate folding. Should
// use s_mov_b32, which can be shrunk to s_movk_i32
def : GCNPat <
(VGPRImm<(f16 fpimm)>:$imm),
(V_MOV_B32_e32 (f16 (bitcast_fpimm_to_i32 $imm)))
>;
def : GCNPat <
(VGPRImm<(bf16 fpimm)>:$imm),
(V_MOV_B32_e32 (bf16 (bitcast_fpimm_to_i32 $imm)))
>;
// V_MOV_B64_PSEUDO and S_MOV_B64_IMM_PSEUDO can be used with any 64-bit
// immediate and wil be expanded as needed, but we will only use these patterns
// for values which can be encoded.
def : GCNPat <
(VGPRImm<(i64 imm)>:$imm),
(V_MOV_B64_PSEUDO imm:$imm)
>;
def : GCNPat <
(VGPRImm<(f64 fpimm)>:$imm),
(V_MOV_B64_PSEUDO (f64 (bitcast_fpimm_to_i64 $imm)))
>;
def : GCNPat <
(i64 imm:$imm),
(S_MOV_B64_IMM_PSEUDO imm:$imm)
>;
def : GCNPat <
(f64 fpimm:$imm),
(S_MOV_B64_IMM_PSEUDO (i64 (bitcast_fpimm_to_i64 fpimm:$imm)))
>;
def : GCNPat <
(f32 fpimm:$imm),
(S_MOV_B32 (f32 (bitcast_fpimm_to_i32 $imm)))
>;
def : GCNPat <
(f16 fpimm:$imm),
(S_MOV_B32 (i32 (bitcast_fpimm_to_i32 $imm)))
>;
def : GCNPat <
(bf16 fpimm:$imm),
(S_MOV_B32 (i32 (bitcast_fpimm_to_i32 $imm)))
>;
def : GCNPat <
(p5 frameindex:$fi),
(V_MOV_B32_e32 (p5 (frameindex_to_targetframeindex $fi)))
>;
def : GCNPat <
(p5 frameindex:$fi),
(S_MOV_B32 (p5 (frameindex_to_targetframeindex $fi)))
>;
def : GCNPat <
(i64 InlineImm64:$imm),
(S_MOV_B64 InlineImm64:$imm)
>;
// XXX - Should this use a s_cmp to set SCC?
// Set to sign-extended 64-bit value (true = -1, false = 0)
def : GCNPat <
(i1 imm:$imm),
(S_MOV_B64 (i64 (as_i64imm $imm)))
> {
let WaveSizePredicate = isWave64;
}
def : GCNPat <
(i1 imm:$imm),
(S_MOV_B32 (i32 (as_i32imm $imm)))
> {
let WaveSizePredicate = isWave32;
}
def : GCNPat <
(f64 InlineImmFP64:$imm),
(S_MOV_B64 (f64 (bitcast_fpimm_to_i64 InlineImmFP64:$imm)))
>;
/********** ================== **********/
/********** Intrinsic Patterns **********/
/********** ================== **********/
def : GCNPat <
(f32 (fpow (VOP3Mods f32:$src0, i32:$src0_mods), (VOP3Mods f32:$src1, i32:$src1_mods))),
(V_EXP_F32_e64 SRCMODS.NONE, (V_MUL_LEGACY_F32_e64 $src1_mods, $src1, SRCMODS.NONE, (V_LOG_F32_e64 $src0_mods, $src0), 0, 0))
>;
def : GCNPat <
(i32 (sext i1:$src0)),
(V_CNDMASK_B32_e64 /*src0mod*/(i32 0), /*src0*/(i32 0),
/*src1mod*/(i32 0), /*src1*/(i32 -1), i1:$src0)
>;
class Ext32Pat <SDNode ext> : GCNPat <
(i32 (ext i1:$src0)),
(V_CNDMASK_B32_e64 /*src0mod*/(i32 0), /*src0*/(i32 0),
/*src1mod*/(i32 0), /*src1*/(i32 1), i1:$src0)
>;
def : Ext32Pat <zext>;
def : Ext32Pat <anyext>;
// The multiplication scales from [0,1) to the unsigned integer range,
// rounding down a bit to avoid unwanted overflow.
def : GCNPat <
(AMDGPUurecip i32:$src0),
(V_CVT_U32_F32_e32
(V_MUL_F32_e32 (i32 CONST.FP_4294966784),
(V_RCP_IFLAG_F32_e32 (V_CVT_F32_U32_e32 $src0))))
>;
//===----------------------------------------------------------------------===//
// VOP3 Patterns
//===----------------------------------------------------------------------===//
def : IMad24Pat<V_MAD_I32_I24_e64, 1>;
def : UMad24Pat<V_MAD_U32_U24_e64, 1>;
// BFI patterns
def BFIImm32 : PatFrag<
(ops node:$x, node:$y, node:$z),
(i32 (DivergentBinFrag<or> (and node:$y, node:$x), (and node:$z, imm))),
[{
auto *X = dyn_cast<ConstantSDNode>(N->getOperand(0)->getOperand(1));
auto *NotX = dyn_cast<ConstantSDNode>(N->getOperand(1)->getOperand(1));
return X && NotX &&
~(unsigned)X->getZExtValue() == (unsigned)NotX->getZExtValue();
}]
>;
// Definition from ISA doc:
// (y & x) | (z & ~x)
def : AMDGPUPatIgnoreCopies <
(DivergentBinFrag<or> (and i32:$y, i32:$x), (and i32:$z, (not i32:$x))),
(V_BFI_B32_e64 (COPY_TO_REGCLASS VSrc_b32:$x, VGPR_32),
(COPY_TO_REGCLASS VSrc_b32:$y, VGPR_32),
(COPY_TO_REGCLASS VSrc_b32:$z, VGPR_32))
>;
// (y & C) | (z & ~C)
def : AMDGPUPatIgnoreCopies <
(BFIImm32 i32:$x, i32:$y, i32:$z),
(V_BFI_B32_e64 VSrc_b32:$x, VSrc_b32:$y, VSrc_b32:$z)
>;
// 64-bit version
def : AMDGPUPatIgnoreCopies <
(DivergentBinFrag<or> (and i64:$y, i64:$x), (and i64:$z, (not i64:$x))),
(REG_SEQUENCE VReg_64,
(V_BFI_B32_e64 (i32 (EXTRACT_SUBREG VReg_64:$x, sub0)),
(i32 (EXTRACT_SUBREG VReg_64:$y, sub0)),
(i32 (EXTRACT_SUBREG VReg_64:$z, sub0))), sub0,
(V_BFI_B32_e64 (i32 (EXTRACT_SUBREG VReg_64:$x, sub1)),
(i32 (EXTRACT_SUBREG VReg_64:$y, sub1)),
(i32 (EXTRACT_SUBREG VReg_64:$z, sub1))), sub1)
>;
// SHA-256 Ch function
// z ^ (x & (y ^ z))
def : AMDGPUPatIgnoreCopies <
(DivergentBinFrag<xor> i32:$z, (and i32:$x, (xor i32:$y, i32:$z))),
(V_BFI_B32_e64 (COPY_TO_REGCLASS VSrc_b32:$x, VGPR_32),
(COPY_TO_REGCLASS VSrc_b32:$y, VGPR_32),
(COPY_TO_REGCLASS VSrc_b32:$z, VGPR_32))
>;
// 64-bit version
def : AMDGPUPatIgnoreCopies <
(DivergentBinFrag<xor> i64:$z, (and i64:$x, (xor i64:$y, i64:$z))),
(REG_SEQUENCE VReg_64,
(V_BFI_B32_e64 (i32 (EXTRACT_SUBREG VReg_64:$x, sub0)),
(i32 (EXTRACT_SUBREG VReg_64:$y, sub0)),
(i32 (EXTRACT_SUBREG VReg_64:$z, sub0))), sub0,
(V_BFI_B32_e64 (i32 (EXTRACT_SUBREG VReg_64:$x, sub1)),
(i32 (EXTRACT_SUBREG VReg_64:$y, sub1)),
(i32 (EXTRACT_SUBREG VReg_64:$z, sub1))), sub1)
>;
def : AMDGPUPat <
(fcopysign f32:$src0, f32:$src1),
(V_BFI_B32_e64 (S_MOV_B32 (i32 0x7fffffff)), $src0, $src1)
>;
def : AMDGPUPat <
(fcopysign f32:$src0, f64:$src1),
(V_BFI_B32_e64 (S_MOV_B32 (i32 0x7fffffff)), $src0,
(i32 (EXTRACT_SUBREG SReg_64:$src1, sub1)))
>;
def : AMDGPUPat <
(fcopysign f64:$src0, f64:$src1),
(REG_SEQUENCE SReg_64,
(i32 (EXTRACT_SUBREG $src0, sub0)), sub0,
(V_BFI_B32_e64 (S_MOV_B32 (i32 0x7fffffff)),
(i32 (EXTRACT_SUBREG SReg_64:$src0, sub1)),
(i32 (EXTRACT_SUBREG SReg_64:$src1, sub1))), sub1)
>;
def : AMDGPUPat <
(fcopysign f64:$src0, f32:$src1),
(REG_SEQUENCE SReg_64,
(i32 (EXTRACT_SUBREG $src0, sub0)), sub0,
(V_BFI_B32_e64 (S_MOV_B32 (i32 0x7fffffff)),
(i32 (EXTRACT_SUBREG SReg_64:$src0, sub1)),
$src1), sub1)
>;
def : ROTRPattern <V_ALIGNBIT_B32_e64>;
def : GCNPat<(i32 (trunc (srl i64:$src0, (and i32:$src1, (i32 31))))),
(V_ALIGNBIT_B32_e64 (i32 (EXTRACT_SUBREG (i64 $src0), sub1)),
(i32 (EXTRACT_SUBREG (i64 $src0), sub0)), $src1)>;
def : GCNPat<(i32 (trunc (srl i64:$src0, (i32 ShiftAmt32Imm:$src1)))),
(V_ALIGNBIT_B32_e64 (i32 (EXTRACT_SUBREG (i64 $src0), sub1)),
(i32 (EXTRACT_SUBREG (i64 $src0), sub0)), $src1)>;
/********** ====================== **********/
/********** Indirect addressing **********/
/********** ====================== **********/
multiclass SI_INDIRECT_Pattern <ValueType vt, ValueType eltvt, string VecSize> {
// Extract with offset
def : GCNPat<
(eltvt (extractelt vt:$src, (MOVRELOffset i32:$idx, (i32 imm:$offset)))),
(!cast<Instruction>("SI_INDIRECT_SRC_"#VecSize) $src, $idx, imm:$offset)
>;
// Insert with offset
def : GCNPat<
(insertelt vt:$src, eltvt:$val, (MOVRELOffset i32:$idx, (i32 imm:$offset))),
(!cast<Instruction>("SI_INDIRECT_DST_"#VecSize) $src, $idx, imm:$offset, $val)
>;
}
defm : SI_INDIRECT_Pattern <v2f32, f32, "V2">;
defm : SI_INDIRECT_Pattern <v4f32, f32, "V4">;
defm : SI_INDIRECT_Pattern <v8f32, f32, "V8">;
defm : SI_INDIRECT_Pattern <v9f32, f32, "V9">;
defm : SI_INDIRECT_Pattern <v10f32, f32, "V10">;
defm : SI_INDIRECT_Pattern <v11f32, f32, "V11">;
defm : SI_INDIRECT_Pattern <v12f32, f32, "V12">;
defm : SI_INDIRECT_Pattern <v16f32, f32, "V16">;
defm : SI_INDIRECT_Pattern <v32f32, f32, "V32">;
defm : SI_INDIRECT_Pattern <v2i32, i32, "V2">;
defm : SI_INDIRECT_Pattern <v4i32, i32, "V4">;
defm : SI_INDIRECT_Pattern <v8i32, i32, "V8">;
defm : SI_INDIRECT_Pattern <v9i32, i32, "V9">;
defm : SI_INDIRECT_Pattern <v10i32, i32, "V10">;
defm : SI_INDIRECT_Pattern <v11i32, i32, "V11">;
defm : SI_INDIRECT_Pattern <v12i32, i32, "V12">;
defm : SI_INDIRECT_Pattern <v16i32, i32, "V16">;
defm : SI_INDIRECT_Pattern <v32i32, i32, "V32">;
//===----------------------------------------------------------------------===//
// SAD Patterns
//===----------------------------------------------------------------------===//
def : GCNPat <
(add (sub_oneuse (umax i32:$src0, i32:$src1),
(umin i32:$src0, i32:$src1)),
i32:$src2),
(V_SAD_U32_e64 $src0, $src1, $src2, (i1 0))
>;
def : GCNPat <
(add (select_oneuse (i1 (setugt i32:$src0, i32:$src1)),
(sub i32:$src0, i32:$src1),
(sub i32:$src1, i32:$src0)),
i32:$src2),
(V_SAD_U32_e64 $src0, $src1, $src2, (i1 0))
>;
//===----------------------------------------------------------------------===//
// Conversion Patterns
//===----------------------------------------------------------------------===//
def : GCNPat<(i32 (UniformSextInreg<i1> i32:$src)),
(S_BFE_I32 i32:$src, (i32 65536))>; // 0 | 1 << 16
// Handle sext_inreg in i64
def : GCNPat <
(i64 (UniformSextInreg<i1> i64:$src)),
(S_BFE_I64 i64:$src, (i32 0x10000)) // 0 | 1 << 16
>;
def : GCNPat <
(i16 (UniformSextInreg<i1> i16:$src)),
(S_BFE_I32 $src, (i32 0x00010000)) // 0 | 1 << 16
>;
def : GCNPat <
(i16 (UniformSextInreg<i8> i16:$src)),
(S_BFE_I32 $src, (i32 0x80000)) // 0 | 8 << 16
>;
def : GCNPat <
(i64 (UniformSextInreg<i8> i64:$src)),
(S_BFE_I64 i64:$src, (i32 0x80000)) // 0 | 8 << 16
>;
def : GCNPat <
(i64 (UniformSextInreg<i16> i64:$src)),
(S_BFE_I64 i64:$src, (i32 0x100000)) // 0 | 16 << 16
>;
def : GCNPat <
(i64 (UniformSextInreg<i32> i64:$src)),
(S_BFE_I64 i64:$src, (i32 0x200000)) // 0 | 32 << 16
>;
def : GCNPat<
(i32 (DivergentSextInreg<i1> i32:$src)),
(V_BFE_I32_e64 i32:$src, (i32 0), (i32 1))>;
def : GCNPat <
(i16 (DivergentSextInreg<i1> i16:$src)),
(V_BFE_I32_e64 $src, (i32 0), (i32 1))
>;
def : GCNPat <
(i16 (DivergentSextInreg<i8> i16:$src)),
(V_BFE_I32_e64 $src, (i32 0), (i32 8))
>;
def : GCNPat<
(i32 (DivergentSextInreg<i8> i32:$src)),
(V_BFE_I32_e64 i32:$src, (i32 0), (i32 8))
>;
def : GCNPat <
(i32 (DivergentSextInreg<i16> i32:$src)),
(V_BFE_I32_e64 $src, (i32 0), (i32 16))
>;
def : GCNPat <
(i64 (DivergentSextInreg<i1> i64:$src)),
(REG_SEQUENCE VReg_64,
(V_BFE_I32_e64 (i32 (EXTRACT_SUBREG i64:$src, sub0)), (i32 0), (i32 1)), sub0,
(V_ASHRREV_I32_e32 (i32 31), (V_BFE_I32_e64 (i32 (EXTRACT_SUBREG i64:$src, sub0)), (i32 0), (i32 1))), sub1)
>;
def : GCNPat <
(i64 (DivergentSextInreg<i8> i64:$src)),
(REG_SEQUENCE VReg_64,
(V_BFE_I32_e64 (i32 (EXTRACT_SUBREG i64:$src, sub0)), (i32 0), (i32 8)), sub0,
(V_ASHRREV_I32_e32 (i32 31), (V_BFE_I32_e64 (i32 (EXTRACT_SUBREG i64:$src, sub0)), (i32 0), (i32 8))), sub1)
>;
def : GCNPat <
(i64 (DivergentSextInreg<i16> i64:$src)),
(REG_SEQUENCE VReg_64,
(V_BFE_I32_e64 (i32 (EXTRACT_SUBREG i64:$src, sub0)), (i32 0), (i32 16)), sub0,
(V_ASHRREV_I32_e32 (i32 31), (V_BFE_I32_e64 (i32 (EXTRACT_SUBREG i64:$src, sub0)), (i32 0), (i32 16))), sub1)
>;
def : GCNPat <
(i64 (DivergentSextInreg<i32> i64:$src)),
(REG_SEQUENCE VReg_64,
(i32 (EXTRACT_SUBREG i64:$src, sub0)), sub0,
(V_ASHRREV_I32_e32 (i32 31), (i32 (EXTRACT_SUBREG i64:$src, sub0))), sub1)
>;
def : GCNPat <
(i64 (zext i32:$src)),
(REG_SEQUENCE SReg_64, $src, sub0, (S_MOV_B32 (i32 0)), sub1)
>;
def : GCNPat <
(i64 (anyext i32:$src)),
(REG_SEQUENCE SReg_64, $src, sub0, (i32 (IMPLICIT_DEF)), sub1)
>;
class ZExt_i64_i1_Pat <SDNode ext> : GCNPat <
(i64 (ext i1:$src)),
(REG_SEQUENCE VReg_64,
(V_CNDMASK_B32_e64 /*src0mod*/(i32 0), /*src0*/(i32 0),
/*src1mod*/(i32 0), /*src1*/(i32 1), $src),
sub0, (S_MOV_B32 (i32 0)), sub1)
>;
def : ZExt_i64_i1_Pat<zext>;
def : ZExt_i64_i1_Pat<anyext>;
// FIXME: We need to use COPY_TO_REGCLASS to work-around the fact that
// REG_SEQUENCE patterns don't support instructions with multiple outputs.
def : GCNPat <
(i64 (UniformUnaryFrag<sext> i32:$src)),
(REG_SEQUENCE SReg_64, $src, sub0,
(i32 (COPY_TO_REGCLASS (S_ASHR_I32 $src, (i32 31)), SReg_32_XM0)), sub1)
>;
def : GCNPat <
(i64 (DivergentUnaryFrag<sext> i32:$src)),
(REG_SEQUENCE VReg_64, $src, sub0,
(i32 (COPY_TO_REGCLASS (V_ASHRREV_I32_e64 (i32 31), $src), VGPR_32)), sub1)
>;
def : GCNPat <
(i64 (sext i1:$src)),
(REG_SEQUENCE VReg_64,
(V_CNDMASK_B32_e64 /*src0mod*/(i32 0), /*src0*/(i32 0),
/*src1mod*/(i32 0), /*src1*/(i32 -1), $src), sub0,
(V_CNDMASK_B32_e64 /*src0mod*/(i32 0), /*src0*/(i32 0),
/*src1mod*/(i32 0), /*src1*/(i32 -1), $src), sub1)
>;
class FPToI1Pat<Instruction Inst, int KOne, ValueType kone_type, ValueType vt, SDPatternOperator fp_to_int> : GCNPat <
(i1 (fp_to_int (vt (VOP3Mods vt:$src0, i32:$src0_modifiers)))),
(i1 (Inst 0, (kone_type KOne), $src0_modifiers, $src0, DSTCLAMP.NONE))
>;
let OtherPredicates = [NotHasTrue16BitInsts] in {
def : FPToI1Pat<V_CMP_EQ_F16_e64, CONST.FP16_ONE, i16, f16, fp_to_uint>;
def : FPToI1Pat<V_CMP_EQ_F16_e64, CONST.FP16_NEG_ONE, i16, f16, fp_to_sint>;
} // end OtherPredicates = [NotHasTrue16BitInsts]
let OtherPredicates = [HasTrue16BitInsts] in {
def : FPToI1Pat<V_CMP_EQ_F16_t16_e64, CONST.FP16_ONE, i16, f16, fp_to_uint>;
def : FPToI1Pat<V_CMP_EQ_F16_t16_e64, CONST.FP16_NEG_ONE, i16, f16, fp_to_sint>;
} // end OtherPredicates = [HasTrue16BitInsts]
def : FPToI1Pat<V_CMP_EQ_F32_e64, CONST.FP32_ONE, i32, f32, fp_to_uint>;
def : FPToI1Pat<V_CMP_EQ_F32_e64, CONST.FP32_NEG_ONE, i32, f32, fp_to_sint>;
def : FPToI1Pat<V_CMP_EQ_F64_e64, CONST.FP64_ONE, i64, f64, fp_to_uint>;
def : FPToI1Pat<V_CMP_EQ_F64_e64, CONST.FP64_NEG_ONE, i64, f64, fp_to_sint>;
// If we need to perform a logical operation on i1 values, we need to
// use vector comparisons since there is only one SCC register. Vector
// comparisons may write to a pair of SGPRs or a single SGPR, so treat
// these as 32 or 64-bit comparisons. When legalizing SGPR copies,
// instructions resulting in the copies from SCC to these instructions
// will be moved to the VALU.
let WaveSizePredicate = isWave64 in {
def : GCNPat <
(i1 (and i1:$src0, i1:$src1)),
(S_AND_B64 $src0, $src1)
>;
def : GCNPat <
(i1 (or i1:$src0, i1:$src1)),
(S_OR_B64 $src0, $src1)
>;
def : GCNPat <
(i1 (xor i1:$src0, i1:$src1)),
(S_XOR_B64 $src0, $src1)
>;
def : GCNPat <
(i1 (add i1:$src0, i1:$src1)),
(S_XOR_B64 $src0, $src1)
>;
def : GCNPat <
(i1 (sub i1:$src0, i1:$src1)),
(S_XOR_B64 $src0, $src1)
>;
let AddedComplexity = 1 in {
def : GCNPat <
(i1 (add i1:$src0, (i1 -1))),
(S_NOT_B64 $src0)
>;
def : GCNPat <
(i1 (sub i1:$src0, (i1 -1))),
(S_NOT_B64 $src0)
>;
}
} // end isWave64
let WaveSizePredicate = isWave32 in {
def : GCNPat <
(i1 (and i1:$src0, i1:$src1)),
(S_AND_B32 $src0, $src1)
>;
def : GCNPat <
(i1 (or i1:$src0, i1:$src1)),
(S_OR_B32 $src0, $src1)
>;
def : GCNPat <
(i1 (xor i1:$src0, i1:$src1)),
(S_XOR_B32 $src0, $src1)
>;
def : GCNPat <
(i1 (add i1:$src0, i1:$src1)),
(S_XOR_B32 $src0, $src1)
>;
def : GCNPat <
(i1 (sub i1:$src0, i1:$src1)),
(S_XOR_B32 $src0, $src1)
>;
let AddedComplexity = 1 in {
def : GCNPat <
(i1 (add i1:$src0, (i1 -1))),
(S_NOT_B32 $src0)
>;
def : GCNPat <
(i1 (sub i1:$src0, (i1 -1))),
(S_NOT_B32 $src0)
>;
}
} // end isWave32
def : GCNPat <
(i32 (DivergentBinFrag<xor> i32:$src0, (i32 -1))),
(V_NOT_B32_e32 $src0)
>;
def : GCNPat <
(i64 (DivergentBinFrag<xor> i64:$src0, (i64 -1))),
(REG_SEQUENCE VReg_64,
(V_NOT_B32_e32 (i32 (EXTRACT_SUBREG i64:$src0, sub0))), sub0,
(V_NOT_B32_e32 (i32 (EXTRACT_SUBREG i64:$src0, sub1))), sub1
)
>;
let SubtargetPredicate = NotHasTrue16BitInsts in
def : GCNPat <
(f16 (sint_to_fp i1:$src)),
(V_CVT_F16_F32_e32 (
V_CNDMASK_B32_e64 /*src0mod*/(i32 0), /*src0*/(i32 0),
/*src1mod*/(i32 0), /*src1*/(i32 CONST.FP32_NEG_ONE),
SSrc_i1:$src))
>;
let SubtargetPredicate = HasTrue16BitInsts in
def : GCNPat <
(f16 (sint_to_fp i1:$src)),
(V_CVT_F16_F32_t16_e32 (
V_CNDMASK_B32_e64 /*src0mod*/(i32 0), /*src0*/(i32 0),
/*src1mod*/(i32 0), /*src1*/(i32 CONST.FP32_NEG_ONE),
SSrc_i1:$src))
>;
let SubtargetPredicate = NotHasTrue16BitInsts in
def : GCNPat <
(f16 (uint_to_fp i1:$src)),
(V_CVT_F16_F32_e32 (
V_CNDMASK_B32_e64 /*src0mod*/(i32 0), /*src0*/(i32 0),
/*src1mod*/(i32 0), /*src1*/(i32 CONST.FP32_ONE),
SSrc_i1:$src))
>;
let SubtargetPredicate = HasTrue16BitInsts in
def : GCNPat <
(f16 (uint_to_fp i1:$src)),
(V_CVT_F16_F32_t16_e32 (
V_CNDMASK_B32_e64 /*src0mod*/(i32 0), /*src0*/(i32 0),
/*src1mod*/(i32 0), /*src1*/(i32 CONST.FP32_ONE),
SSrc_i1:$src))
>;
def : GCNPat <
(f32 (sint_to_fp i1:$src)),
(V_CNDMASK_B32_e64 /*src0mod*/(i32 0), /*src0*/(i32 0),
/*src1mod*/(i32 0), /*src1*/(i32 CONST.FP32_NEG_ONE),
SSrc_i1:$src)
>;
def : GCNPat <
(f32 (uint_to_fp i1:$src)),
(V_CNDMASK_B32_e64 /*src0mod*/(i32 0), /*src0*/(i32 0),
/*src1mod*/(i32 0), /*src1*/(i32 CONST.FP32_ONE),
SSrc_i1:$src)
>;
def : GCNPat <
(f64 (sint_to_fp i1:$src)),
(V_CVT_F64_I32_e32 (V_CNDMASK_B32_e64 /*src0mod*/(i32 0), /*src0*/(i32 0),
/*src1mod*/(i32 0), /*src1*/(i32 -1),
SSrc_i1:$src))
>;
def : GCNPat <
(f64 (uint_to_fp i1:$src)),
(V_CVT_F64_U32_e32 (V_CNDMASK_B32_e64 /*src0mod*/(i32 0), /*src0*/(i32 0),
/*src1mod*/(i32 0), /*src1*/(i32 1),
SSrc_i1:$src))
>;
//===----------------------------------------------------------------------===//
// Miscellaneous Patterns
//===----------------------------------------------------------------------===//
// Eliminate a zero extension from an fp16 operation if it already
// zeros the high bits of the 32-bit register.
//
// This is complicated on gfx9+. Some instructions maintain the legacy
// zeroing behavior, but others preserve the high bits. Some have a
// control bit to change the behavior. We can't simply say with
// certainty what the source behavior is without more context on how
// the src is lowered. e.g. fptrunc + fma may be lowered to a
// v_fma_mix* instruction which does not zero, or may not.
def : GCNPat<
(i32 (DivergentUnaryFrag<abs> i32:$src)),
(V_MAX_I32_e64 (V_SUB_CO_U32_e32 (i32 0), $src), $src)>;
let AddedComplexity = 1 in {
def : GCNPat<
(i32 (DivergentUnaryFrag<abs> i32:$src)),
(V_MAX_I32_e64 (V_SUB_U32_e32 (i32 0), $src), $src)>{
let SubtargetPredicate = HasAddNoCarryInsts;
}
} // AddedComplexity = 1
def : GCNPat<
(i32 (DivergentUnaryFrag<zext> i16:$src)),
(V_AND_B32_e64 (S_MOV_B32 (i32 0xffff)), $src)
>;
def : GCNPat<
(i64 (DivergentUnaryFrag<zext> i16:$src)),
(REG_SEQUENCE VReg_64,
(V_AND_B32_e64 (S_MOV_B32 (i32 0xffff)), $src), sub0,
(S_MOV_B32 (i32 0)), sub1)
>;
def : GCNPat<
(i32 (zext (i16 (bitconvert fp16_zeros_high_16bits:$src)))),
(COPY VSrc_b16:$src)>;
def : GCNPat <
(i32 (trunc i64:$a)),
(EXTRACT_SUBREG $a, sub0)
>;
def : GCNPat <
(i1 (UniformUnaryFrag<trunc> i32:$a)),
(S_CMP_EQ_U32 (S_AND_B32 (i32 1), $a), (i32 1))
>;
def : GCNPat <
(i1 (UniformUnaryFrag<trunc> i16:$a)),
(S_CMP_EQ_U32 (S_AND_B32 (i32 1), $a), (i32 1))
>;
def : GCNPat <
(i1 (UniformUnaryFrag<trunc> i64:$a)),
(S_CMP_EQ_U32 (S_AND_B32 (i32 1),
(i32 (EXTRACT_SUBREG $a, sub0))), (i32 1))
>;
def : GCNPat <
(i1 (DivergentUnaryFrag<trunc> i32:$a)),
(V_CMP_EQ_U32_e64 (V_AND_B32_e64 (i32 1), $a), (i32 1))
>;
def : GCNPat <
(i1 (DivergentUnaryFrag<trunc> i16:$a)),
(V_CMP_EQ_U32_e64 (V_AND_B32_e64 (i32 1), $a), (i32 1))
>;
def IMMBitSelConst : SDNodeXForm<imm, [{
return CurDAG->getTargetConstant(1ULL << N->getZExtValue(), SDLoc(N),
MVT::i32);
}]>;
// Matching separate SRL and TRUNC instructions
// with dependent operands (SRL dest is source of TRUNC)
// generates three instructions. However, by using bit shifts,
// the V_LSHRREV_B32_e64 result can be directly used in the
// operand of the V_AND_B32_e64 instruction:
// (trunc i32 (srl i32 $a, i32 $b)) ->
// v_and_b32_e64 $a, (1 << $b), $a
// v_cmp_ne_u32_e64 $a, 0, $a
// Handle the VALU case.
def : GCNPat <
(i1 (DivergentUnaryFrag<trunc> (i32 (srl i32:$a, (i32 imm:$b))))),
(V_CMP_NE_U32_e64 (V_AND_B32_e64 (i32 (IMMBitSelConst $b)), $a),
(i32 0))
>;
// Handle the scalar case.
def : GCNPat <
(i1 (UniformUnaryFrag<trunc> (i32 (srl i32:$a, (i32 imm:$b))))),
(S_CMP_LG_U32 (S_AND_B32 (i32 (IMMBitSelConst $b)), $a),
(i32 0))
>;
def : GCNPat <
(i1 (DivergentUnaryFrag<trunc> i64:$a)),
(V_CMP_EQ_U32_e64 (V_AND_B32_e64 (i32 1),
(i32 (EXTRACT_SUBREG $a, sub0))), (i32 1))
>;
def : GCNPat <
(i32 (bswap i32:$a)),
(V_BFI_B32_e64 (S_MOV_B32 (i32 0x00ff00ff)),
(V_ALIGNBIT_B32_e64 VSrc_b32:$a, VSrc_b32:$a, (i32 24)),
(V_ALIGNBIT_B32_e64 VSrc_b32:$a, VSrc_b32:$a, (i32 8)))
>;
// FIXME: This should have been narrowed to i32 during legalization.
// This pattern should also be skipped for GlobalISel
def : GCNPat <
(i64 (bswap i64:$a)),
(REG_SEQUENCE VReg_64,
(V_BFI_B32_e64 (S_MOV_B32 (i32 0x00ff00ff)),
(V_ALIGNBIT_B32_e64 (i32 (EXTRACT_SUBREG VReg_64:$a, sub1)),
(i32 (EXTRACT_SUBREG VReg_64:$a, sub1)),
(i32 24)),
(V_ALIGNBIT_B32_e64 (i32 (EXTRACT_SUBREG VReg_64:$a, sub1)),
(i32 (EXTRACT_SUBREG VReg_64:$a, sub1)),
(i32 8))),
sub0,
(V_BFI_B32_e64 (S_MOV_B32 (i32 0x00ff00ff)),
(V_ALIGNBIT_B32_e64 (i32 (EXTRACT_SUBREG VReg_64:$a, sub0)),
(i32 (EXTRACT_SUBREG VReg_64:$a, sub0)),
(i32 24)),
(V_ALIGNBIT_B32_e64 (i32 (EXTRACT_SUBREG VReg_64:$a, sub0)),
(i32 (EXTRACT_SUBREG VReg_64:$a, sub0)),
(i32 8))),
sub1)
>;
// FIXME: The AddedComplexity should not be needed, but in GlobalISel
// the BFI pattern ends up taking precedence without it.
let SubtargetPredicate = isGFX8Plus, AddedComplexity = 1 in {
// Magic number: 3 | (2 << 8) | (1 << 16) | (0 << 24)
//
// My reading of the manual suggests we should be using src0 for the
// register value, but this is what seems to work.
def : GCNPat <
(i32 (bswap i32:$a)),
(V_PERM_B32_e64 (i32 0), VSrc_b32:$a, (S_MOV_B32 (i32 0x00010203)))
>;
// FIXME: This should have been narrowed to i32 during legalization.
// This pattern should also be skipped for GlobalISel
def : GCNPat <
(i64 (bswap i64:$a)),
(REG_SEQUENCE VReg_64,
(V_PERM_B32_e64 (i32 0), (EXTRACT_SUBREG VReg_64:$a, sub1),
(S_MOV_B32 (i32 0x00010203))),
sub0,
(V_PERM_B32_e64 (i32 0), (EXTRACT_SUBREG VReg_64:$a, sub0),
(S_MOV_B32 (i32 0x00010203))),
sub1)
>;
// Magic number: 1 | (0 << 8) | (12 << 16) | (12 << 24)
// The 12s emit 0s.
def : GCNPat <
(i16 (bswap i16:$a)),
(V_PERM_B32_e64 (i32 0), VSrc_b32:$a, (S_MOV_B32 (i32 0x0c0c0001)))
>;
def : GCNPat <
(i32 (zext (bswap i16:$a))),
(V_PERM_B32_e64 (i32 0), VSrc_b32:$a, (S_MOV_B32 (i32 0x0c0c0001)))
>;
// Magic number: 1 | (0 << 8) | (3 << 16) | (2 << 24)
def : GCNPat <
(v2i16 (bswap v2i16:$a)),
(V_PERM_B32_e64 (i32 0), VSrc_b32:$a, (S_MOV_B32 (i32 0x02030001)))
>;
}
def : GCNPat<
(i64 (DivergentUnaryFrag<bitreverse> i64:$a)),
(REG_SEQUENCE VReg_64,
(V_BFREV_B32_e64 (i32 (EXTRACT_SUBREG VReg_64:$a, sub1))), sub0,
(V_BFREV_B32_e64 (i32 (EXTRACT_SUBREG VReg_64:$a, sub0))), sub1)>;
// Prefer selecting to max when legal, but using mul is always valid.
let AddedComplexity = -5 in {
let OtherPredicates = [NotHasTrue16BitInsts] in {
def : GCNPat<
(fcanonicalize (f16 (VOP3Mods f16:$src, i32:$src_mods))),
(V_MUL_F16_e64 0, (i32 CONST.FP16_ONE), $src_mods, $src)
>;
def : GCNPat<
(fcanonicalize (f16 (fneg (VOP3Mods f16:$src, i32:$src_mods)))),
(V_MUL_F16_e64 0, (i32 CONST.FP16_NEG_ONE), $src_mods, $src)
>;
} // End OtherPredicates
let OtherPredicates = [HasTrue16BitInsts] in {
def : GCNPat<
(fcanonicalize (f16 (VOP3Mods f16:$src, i32:$src_mods))),
(V_MUL_F16_fake16_e64 0, (i32 CONST.FP16_ONE), $src_mods, $src)
>;
def : GCNPat<
(fcanonicalize (f16 (fneg (VOP3Mods f16:$src, i32:$src_mods)))),
(V_MUL_F16_fake16_e64 0, (i32 CONST.FP16_NEG_ONE), $src_mods, $src)
>;
} // End OtherPredicates
def : GCNPat<
(fcanonicalize (v2f16 (VOP3PMods v2f16:$src, i32:$src_mods))),
(V_PK_MUL_F16 0, (i32 CONST.FP16_ONE), $src_mods, $src, DSTCLAMP.NONE)
>;
def : GCNPat<
(fcanonicalize (f32 (VOP3Mods f32:$src, i32:$src_mods))),
(V_MUL_F32_e64 0, (i32 CONST.FP32_ONE), $src_mods, $src)
>;
def : GCNPat<
(fcanonicalize (f32 (fneg (VOP3Mods f32:$src, i32:$src_mods)))),
(V_MUL_F32_e64 0, (i32 CONST.FP32_NEG_ONE), $src_mods, $src)
>;
let SubtargetPredicate = HasPackedFP32Ops in {
def : GCNPat<
(fcanonicalize (v2f32 (VOP3PMods v2f32:$src, i32:$src_mods))),
(V_PK_MUL_F32 0, CONST.FP32_ONE, $src_mods, $src)
>;
}
// TODO: Handle fneg like other types.
def : GCNPat<
(fcanonicalize (f64 (VOP3Mods f64:$src, i32:$src_mods))),
(V_MUL_F64_e64 0, CONST.FP64_ONE, $src_mods, $src)
>;
} // End AddedComplexity = -5
multiclass SelectCanonicalizeAsMax<
list<Predicate> f32_preds = [],
list<Predicate> f64_preds = [],
list<Predicate> f16_preds = []> {
def : GCNPat<
(fcanonicalize (f32 (VOP3Mods f32:$src, i32:$src_mods))),
(V_MAX_F32_e64 $src_mods, $src, $src_mods, $src)> {
let OtherPredicates = f32_preds;
}
def : GCNPat<
(fcanonicalize (f64 (VOP3Mods f64:$src, i32:$src_mods))),
(V_MAX_F64_e64 $src_mods, $src, $src_mods, $src)> {
let OtherPredicates = f64_preds;
}
def : GCNPat<
(fcanonicalize (f16 (VOP3Mods f16:$src, i32:$src_mods))),
(V_MAX_F16_e64 $src_mods, $src, $src_mods, $src, 0, 0)> {
let OtherPredicates = !listconcat(f16_preds, [Has16BitInsts, NotHasTrue16BitInsts]);
}
def : GCNPat<
(fcanonicalize (f16 (VOP3Mods f16:$src, i32:$src_mods))),
(V_MAX_F16_fake16_e64 $src_mods, $src, $src_mods, $src, 0, 0)> {
let OtherPredicates = !listconcat(f16_preds, [Has16BitInsts, HasTrue16BitInsts]);
}
def : GCNPat<
(fcanonicalize (v2f16 (VOP3PMods v2f16:$src, i32:$src_mods))),
(V_PK_MAX_F16 $src_mods, $src, $src_mods, $src, DSTCLAMP.NONE)> {
// FIXME: Should have VOP3P subtarget predicate
let OtherPredicates = f16_preds;
}
}
// On pre-gfx9 targets, v_max_*/v_min_* did not respect the denormal
// mode, and would never flush. For f64, it's faster to do implement
// this with a max. For f16/f32 it's a wash, but prefer max when
// valid.
//
// FIXME: Lowering f32/f16 with max is worse since we can use a
// smaller encoding if the input is fneg'd. It also adds an extra
// register use.
let SubtargetPredicate = HasMinMaxDenormModes in {
defm : SelectCanonicalizeAsMax<[], [], []>;
} // End SubtargetPredicate = HasMinMaxDenormModes
let SubtargetPredicate = NotHasMinMaxDenormModes in {
// Use the max lowering if we don't need to flush.
// FIXME: We don't do use this for f32 as a workaround for the
// library being compiled with the default ieee mode, but
// potentially being called from flushing kernels. Really we should
// not be mixing code expecting different default FP modes, but mul
// works in any FP environment.
defm : SelectCanonicalizeAsMax<[FalsePredicate], [FP64Denormals], [FP16Denormals]>;
} // End SubtargetPredicate = NotHasMinMaxDenormModes
let OtherPredicates = [HasDLInsts] in {
// Don't allow source modifiers. If there are any source modifiers then it's
// better to select fma instead of fmac.
def : GCNPat <
(fma (f32 (VOP3NoMods f32:$src0)),
(f32 (VOP3NoMods f32:$src1)),
(f32 (VOP3NoMods f32:$src2))),
(V_FMAC_F32_e64 SRCMODS.NONE, $src0, SRCMODS.NONE, $src1,
SRCMODS.NONE, $src2)
>;
} // End OtherPredicates = [HasDLInsts]
let SubtargetPredicate = isGFX10Plus in {
// Don't allow source modifiers. If there are any source modifiers then it's
// better to select fma instead of fmac.
let OtherPredicates = [NotHasTrue16BitInsts] in
def : GCNPat <
(fma (f16 (VOP3NoMods f32:$src0)),
(f16 (VOP3NoMods f32:$src1)),
(f16 (VOP3NoMods f32:$src2))),
(V_FMAC_F16_e64 SRCMODS.NONE, $src0, SRCMODS.NONE, $src1,
SRCMODS.NONE, $src2)
>;
let OtherPredicates = [HasTrue16BitInsts] in
def : GCNPat <
(fma (f16 (VOP3NoMods f32:$src0)),
(f16 (VOP3NoMods f32:$src1)),
(f16 (VOP3NoMods f32:$src2))),
(V_FMAC_F16_t16_e64 SRCMODS.NONE, $src0, SRCMODS.NONE, $src1,
SRCMODS.NONE, $src2)
>;
}
let OtherPredicates = [HasFmacF64Inst] in
// Don't allow source modifiers. If there are any source modifiers then it's
// better to select fma instead of fmac.
def : GCNPat <
(fma (f64 (VOP3NoMods f64:$src0)),
(f64 (VOP3NoMods f64:$src1)),
(f64 (VOP3NoMods f64:$src2))),
(V_FMAC_F64_e64 SRCMODS.NONE, $src0, SRCMODS.NONE, $src1,
SRCMODS.NONE, $src2)
>;
// COPY is workaround tablegen bug from multiple outputs
// from S_LSHL_B32's multiple outputs from implicit scc def.
let AddedComplexity = 1 in {
def : GCNPat <
(v2i16 (UniformBinFrag<build_vector> (i16 0), (i16 SReg_32:$src1))),
(S_LSHL_B32 SReg_32:$src1, (i16 16))
>;
def : GCNPat <
(v2i16 (DivergentBinFrag<build_vector> (i16 0), (i16 VGPR_32:$src1))),
(v2i16 (V_LSHLREV_B32_e64 (i16 16), VGPR_32:$src1))
>;
def : GCNPat <
(v2i16 (UniformBinFrag<build_vector> (i16 SReg_32:$src1), (i16 0))),
(S_AND_B32 (S_MOV_B32 (i32 0xffff)), SReg_32:$src1)
>;
def : GCNPat <
(v2i16 (DivergentBinFrag<build_vector> (i16 VGPR_32:$src1), (i16 0))),
(v2i16 (V_AND_B32_e64 (i32 (V_MOV_B32_e32 (i32 0xffff))), VGPR_32:$src1))
>;
def : GCNPat <
(v2f16 (UniformBinFrag<build_vector> (f16 SReg_32:$src1), (f16 FP_ZERO))),
(S_AND_B32 (S_MOV_B32 (i32 0xffff)), SReg_32:$src1)
>;
def : GCNPat <
(v2f16 (DivergentBinFrag<build_vector> (f16 VGPR_32:$src1), (f16 FP_ZERO))),
(v2f16 (V_AND_B32_e64 (i32 (V_MOV_B32_e32 (i32 0xffff))), VGPR_32:$src1))
>;
def : GCNPat <
(v2i16 (UniformBinFrag<build_vector> (i16 SReg_32:$src0), (i16 undef))),
(COPY_TO_REGCLASS SReg_32:$src0, SReg_32)
>;
def : GCNPat <
(v2i16 (DivergentBinFrag<build_vector> (i16 VGPR_32:$src0), (i16 undef))),
(COPY_TO_REGCLASS VGPR_32:$src0, VGPR_32)
>;
def : GCNPat <
(v2f16 (build_vector f16:$src0, (f16 undef))),
(COPY $src0)
>;
def : GCNPat <
(v2i16 (UniformBinFrag<build_vector> (i16 undef), (i16 SReg_32:$src1))),
(S_LSHL_B32 SReg_32:$src1, (i32 16))
>;
def : GCNPat <
(v2i16 (DivergentBinFrag<build_vector> (i16 undef), (i16 VGPR_32:$src1))),
(v2i16 (V_LSHLREV_B32_e64 (i32 16), VGPR_32:$src1))
>;
def : GCNPat <
(v2f16 (UniformBinFrag<build_vector> (f16 undef), (f16 SReg_32:$src1))),
(S_LSHL_B32 SReg_32:$src1, (i32 16))
>;
def : GCNPat <
(v2f16 (DivergentBinFrag<build_vector> (f16 undef), (f16 VGPR_32:$src1))),
(v2f16 (V_LSHLREV_B32_e64 (i32 16), VGPR_32:$src1))
>;
}
let SubtargetPredicate = HasVOP3PInsts in {
def : GCNPat <
(v2i16 (UniformBinFrag<build_vector> (i16 SReg_32:$src0), (i16 SReg_32:$src1))),
(S_PACK_LL_B32_B16 SReg_32:$src0, SReg_32:$src1)
>;
def : GCNPat <
(v2i16 (DivergentBinFrag<build_vector> (i16 VGPR_32:$src0), (i16 VGPR_32:$src1))),
(v2i16 (V_LSHL_OR_B32_e64 $src1, (i32 16), (i32 (V_AND_B32_e64 (i32 (V_MOV_B32_e32 (i32 0xffff))), $src0))))
>;
// With multiple uses of the shift, this will duplicate the shift and
// increase register pressure.
def : GCNPat <
(v2i16 (UniformBinFrag<build_vector> (i16 SReg_32:$src0), (i16 (trunc (srl_oneuse SReg_32:$src1, (i32 16)))))),
(v2i16 (S_PACK_LH_B32_B16 SReg_32:$src0, SReg_32:$src1))
>;
def : GCNPat <
(v2i16 (UniformBinFrag<build_vector> (i16 (trunc (srl_oneuse SReg_32:$src0, (i32 16)))),
(i16 (trunc (srl_oneuse SReg_32:$src1, (i32 16)))))),
(S_PACK_HH_B32_B16 SReg_32:$src0, SReg_32:$src1)
>;
def : GCNPat <
(v2f16 (UniformBinFrag<build_vector> (f16 SReg_32:$src0), (f16 SReg_32:$src1))),
(S_PACK_LL_B32_B16 SReg_32:$src0, SReg_32:$src1)
>;
foreach Ty = [i16, f16] in {
defvar vecTy = !if(!eq(Ty, i16), v2i16, v2f16);
defvar immzeroTy = !if(!eq(Ty, i16), immzero, fpimmzero);
// Take the lower 16 bits from each VGPR_32 and concat them
def : GCNPat <
(vecTy (DivergentBinFrag<build_vector> (Ty VGPR_32:$a), (Ty VGPR_32:$b))),
(V_PERM_B32_e64 VGPR_32:$b, VGPR_32:$a, (S_MOV_B32 (i32 0x05040100)))
>;
// Take the lower 16 bits from V[0] and the upper 16 bits from V[1]
// Special case, can use V_BFI (0xffff literal likely more reusable than 0x70601000)
def : GCNPat <
(vecTy (DivergentBinFrag<build_vector> (Ty (immzeroTy)),
(Ty !if(!eq(Ty, i16),
(Ty (trunc (srl VGPR_32:$b, (i32 16)))),
(Ty (bitconvert (i16 (trunc (srl VGPR_32:$b, (i32 16)))))))))),
(V_AND_B32_e64 (S_MOV_B32 (i32 0xffff0000)), VGPR_32:$b)
>;
// Take the lower 16 bits from V[0] and the upper 16 bits from V[1]
// Special case, can use V_BFI (0xffff literal likely more reusable than 0x70601000)
def : GCNPat <
(vecTy (DivergentBinFrag<build_vector> (Ty VGPR_32:$a),
(Ty !if(!eq(Ty, i16),
(Ty (trunc (srl VGPR_32:$b, (i32 16)))),
(Ty (bitconvert (i16 (trunc (srl VGPR_32:$b, (i32 16)))))))))),
(V_BFI_B32_e64 (S_MOV_B32 (i32 0x0000ffff)), VGPR_32:$a, VGPR_32:$b)
>;
// Take the upper 16 bits from V[0] and the lower 16 bits from V[1]
// Special case, can use V_ALIGNBIT (always uses encoded literal)
def : GCNPat <
(vecTy (DivergentBinFrag<build_vector>
(Ty !if(!eq(Ty, i16),
(Ty (trunc (srl VGPR_32:$a, (i32 16)))),
(Ty (bitconvert (i16 (trunc (srl VGPR_32:$a, (i32 16)))))))),
(Ty VGPR_32:$b))),
(V_ALIGNBIT_B32_e64 VGPR_32:$b, VGPR_32:$a, (i32 16))
>;
// Take the upper 16 bits from each VGPR_32 and concat them
def : GCNPat <
(vecTy (DivergentBinFrag<build_vector>
(Ty !if(!eq(Ty, i16),
(Ty (trunc (srl VGPR_32:$a, (i32 16)))),
(Ty (bitconvert (i16 (trunc (srl VGPR_32:$a, (i32 16)))))))),
(Ty !if(!eq(Ty, i16),
(Ty (trunc (srl VGPR_32:$b, (i32 16)))),
(Ty (bitconvert (i16 (trunc (srl VGPR_32:$b, (i32 16)))))))))),
(V_PERM_B32_e64 VGPR_32:$b, VGPR_32:$a, (S_MOV_B32 (i32 0x07060302)))
>;
} // end foreach Ty
let AddedComplexity = 5 in {
def : GCNPat <
(v2f16 (is_canonicalized<build_vector> (f16 (VOP3Mods (f16 VGPR_32:$src0), i32:$src0_mods)),
(f16 (VOP3Mods (f16 VGPR_32:$src1), i32:$src1_mods)))),
(V_PACK_B32_F16_e64 $src0_mods, VGPR_32:$src0, $src1_mods, VGPR_32:$src1)
>;
}
} // End SubtargetPredicate = HasVOP3PInsts
// With multiple uses of the shift, this will duplicate the shift and
// increase register pressure.
let SubtargetPredicate = isGFX11Plus in
def : GCNPat <
(v2i16 (build_vector (i16 (trunc (srl_oneuse SReg_32:$src0, (i32 16)))), (i16 SReg_32:$src1))),
(v2i16 (S_PACK_HL_B32_B16 SReg_32:$src0, SReg_32:$src1))
>;
def : GCNPat <
(v2f16 (scalar_to_vector f16:$src0)),
(COPY $src0)
>;
def : GCNPat <
(v2i16 (scalar_to_vector i16:$src0)),
(COPY $src0)
>;
def : GCNPat <
(v4i16 (scalar_to_vector i16:$src0)),
(INSERT_SUBREG (IMPLICIT_DEF), $src0, sub0)
>;
def : GCNPat <
(v4f16 (scalar_to_vector f16:$src0)),
(INSERT_SUBREG (IMPLICIT_DEF), $src0, sub0)
>;
def : GCNPat <
(i64 (int_amdgcn_mov_dpp i64:$src, timm:$dpp_ctrl, timm:$row_mask,
timm:$bank_mask, timm:$bound_ctrl)),
(V_MOV_B64_DPP_PSEUDO VReg_64_Align2:$src, VReg_64_Align2:$src,
(as_i32timm $dpp_ctrl), (as_i32timm $row_mask),
(as_i32timm $bank_mask),
(as_i1timm $bound_ctrl))
>;
def : GCNPat <
(i64 (int_amdgcn_update_dpp i64:$old, i64:$src, timm:$dpp_ctrl, timm:$row_mask,
timm:$bank_mask, timm:$bound_ctrl)),
(V_MOV_B64_DPP_PSEUDO VReg_64_Align2:$old, VReg_64_Align2:$src, (as_i32timm $dpp_ctrl),
(as_i32timm $row_mask), (as_i32timm $bank_mask),
(as_i1timm $bound_ctrl))
>;
//===----------------------------------------------------------------------===//
// Fract Patterns
//===----------------------------------------------------------------------===//
let SubtargetPredicate = isGFX6 in {
// V_FRACT is buggy on SI, so the F32 version is never used and (x-floor(x)) is
// used instead. However, SI doesn't have V_FLOOR_F64, so the most efficient
// way to implement it is using V_FRACT_F64.
// The workaround for the V_FRACT bug is:
// fract(x) = isnan(x) ? x : min(V_FRACT(x), 0.99999999999999999)
// Convert floor(x) to (x - fract(x))
// Don't bother handling this for GlobalISel, it's handled during
// lowering.
//
// FIXME: DAG should also custom lower this.
def : GCNPat <
(f64 (ffloor (f64 (VOP3Mods f64:$x, i32:$mods)))),
(V_ADD_F64_e64
$mods,
$x,
SRCMODS.NEG,
(V_CNDMASK_B64_PSEUDO
(V_MIN_F64_e64
SRCMODS.NONE,
(V_FRACT_F64_e64 $mods, $x),
SRCMODS.NONE,
(V_MOV_B64_PSEUDO 0x3fefffffffffffff)),
$x,
(V_CMP_CLASS_F64_e64 SRCMODS.NONE, $x, (i32 3 /*NaN*/))))
>;
} // End SubtargetPredicates = isGFX6
//============================================================================//
// Miscellaneous Optimization Patterns
//============================================================================//
// Undo sub x, c -> add x, -c canonicalization since c is more likely
// an inline immediate than -c.
// TODO: Also do for 64-bit.
def : GCNPat<
(UniformBinFrag<add> i32:$src0, (i32 NegSubInlineConst32:$src1)),
(S_SUB_I32 SReg_32:$src0, NegSubInlineConst32:$src1)
>;
def : GCNPat<
(DivergentBinFrag<add> i32:$src0, (i32 NegSubInlineConst32:$src1)),
(V_SUB_U32_e64 VS_32:$src0, NegSubInlineConst32:$src1)> {
let SubtargetPredicate = HasAddNoCarryInsts;
}
def : GCNPat<
(DivergentBinFrag<add> i32:$src0, (i32 NegSubInlineConst32:$src1)),
(V_SUB_CO_U32_e64 VS_32:$src0, NegSubInlineConst32:$src1)> {
let SubtargetPredicate = NotHasAddNoCarryInsts;
}
// Avoid pointlessly materializing a constant in VGPR.
// FIXME: Should also do this for readlane, but tablegen crashes on
// the ignored src1.
def : GCNPat<
(int_amdgcn_readfirstlane (i32 imm:$src)),
(S_MOV_B32 SReg_32:$src)
>;
multiclass BFMPatterns <ValueType vt, PatFrag SHL, PatFrag ADD, InstSI BFM> {
def : GCNPat <
(vt (SHL (vt (add (vt (shl 1, vt:$a)), -1)), vt:$b)),
(BFM $a, $b)
>;
def : GCNPat <
(vt (ADD (vt (shl 1, vt:$a)), -1)),
(BFM $a, (i32 0))
>;
}
defm : BFMPatterns <i32, UniformBinFrag<shl>, UniformBinFrag<add>, S_BFM_B32>;
// FIXME: defm : BFMPatterns <i64, UniformBinFrag<shl>, UniformBinFrag<add>, S_BFM_B64>;
defm : BFMPatterns <i32, DivergentBinFrag<shl>, DivergentBinFrag<add>, V_BFM_B32_e64>;
// Bitfield extract patterns
def IMMZeroBasedBitfieldMask : ImmLeaf <i32, [{
return isMask_32(Imm);
}]>;
def IMMPopCount : SDNodeXForm<imm, [{
return CurDAG->getTargetConstant(llvm::popcount(N->getZExtValue()), SDLoc(N),
MVT::i32);
}]>;
def : AMDGPUPat <
(DivergentBinFrag<and> (i32 (srl i32:$src, i32:$rshift)),
IMMZeroBasedBitfieldMask:$mask),
(V_BFE_U32_e64 $src, $rshift, (i32 (IMMPopCount $mask)))
>;
// x & ((1 << y) - 1)
def : AMDGPUPat <
(DivergentBinFrag<and> i32:$src, (add_oneuse (shl_oneuse 1, i32:$width), -1)),
(V_BFE_U32_e64 $src, (i32 0), $width)
>;
// x & ~(-1 << y)
def : AMDGPUPat <
(DivergentBinFrag<and> i32:$src,
(xor_oneuse (shl_oneuse -1, i32:$width), -1)),
(V_BFE_U32_e64 $src, (i32 0), $width)
>;
// x & (-1 >> (bitwidth - y))
def : AMDGPUPat <
(DivergentBinFrag<and> i32:$src, (srl_oneuse -1, (sub 32, i32:$width))),
(V_BFE_U32_e64 $src, (i32 0), $width)
>;
// x << (bitwidth - y) >> (bitwidth - y)
def : AMDGPUPat <
(DivergentBinFrag<srl> (shl_oneuse i32:$src, (sub 32, i32:$width)),
(sub 32, i32:$width)),
(V_BFE_U32_e64 $src, (i32 0), $width)
>;
def : AMDGPUPat <
(DivergentBinFrag<sra> (shl_oneuse i32:$src, (sub 32, i32:$width)),
(sub 32, i32:$width)),
(V_BFE_I32_e64 $src, (i32 0), $width)
>;
// SHA-256 Ma patterns
// ((x & z) | (y & (x | z))) -> BFI (XOR x, y), z, y
def : AMDGPUPatIgnoreCopies <
(DivergentBinFrag<or> (and i32:$x, i32:$z),
(and i32:$y, (or i32:$x, i32:$z))),
(V_BFI_B32_e64 (V_XOR_B32_e64 (COPY_TO_REGCLASS VSrc_b32:$x, VGPR_32),
(COPY_TO_REGCLASS VSrc_b32:$y, VGPR_32)),
(COPY_TO_REGCLASS VSrc_b32:$z, VGPR_32),
(COPY_TO_REGCLASS VSrc_b32:$y, VGPR_32))
>;
def : AMDGPUPatIgnoreCopies <
(DivergentBinFrag<or> (and i64:$x, i64:$z),
(and i64:$y, (or i64:$x, i64:$z))),
(REG_SEQUENCE VReg_64,
(V_BFI_B32_e64 (V_XOR_B32_e64 (i32 (EXTRACT_SUBREG VReg_64:$x, sub0)),
(i32 (EXTRACT_SUBREG VReg_64:$y, sub0))),
(i32 (EXTRACT_SUBREG VReg_64:$z, sub0)),
(i32 (EXTRACT_SUBREG VReg_64:$y, sub0))), sub0,
(V_BFI_B32_e64 (V_XOR_B32_e64 (i32 (EXTRACT_SUBREG VReg_64:$x, sub1)),
(i32 (EXTRACT_SUBREG VReg_64:$y, sub1))),
(i32 (EXTRACT_SUBREG VReg_64:$z, sub1)),
(i32 (EXTRACT_SUBREG VReg_64:$y, sub1))), sub1)
>;
multiclass IntMed3Pat<Instruction med3Inst,
SDPatternOperator min,
SDPatternOperator max> {
// This matches 16 permutations of
// min(max(a, b), max(min(a, b), c))
def : AMDGPUPat <
(min (max i32:$src0, i32:$src1),
(max (min i32:$src0, i32:$src1), i32:$src2)),
(med3Inst VSrc_b32:$src0, VSrc_b32:$src1, VSrc_b32:$src2)
>;
// This matches 16 permutations of
// max(min(x, y), min(max(x, y), z))
def : AMDGPUPat <
(max (min i32:$src0, i32:$src1),
(min (max i32:$src0, i32:$src1), i32:$src2)),
(med3Inst VSrc_b32:$src0, VSrc_b32:$src1, VSrc_b32:$src2)
>;
}
defm : IntMed3Pat<V_MED3_I32_e64, smin, smax>;
defm : IntMed3Pat<V_MED3_U32_e64, umin, umax>;
multiclass FPMed3Pat<ValueType vt,
Instruction med3Inst> {
// This matches 16 permutations of max(min(x, y), min(max(x, y), z))
def : GCNPat<
(fmaxnum_like_nnan
(fminnum_like (VOP3Mods vt:$src0, i32:$src0_mods),
(VOP3Mods vt:$src1, i32:$src1_mods)),
(fminnum_like (fmaxnum_like (VOP3Mods vt:$src0, i32:$src0_mods),
(VOP3Mods vt:$src1, i32:$src1_mods)),
(vt (VOP3Mods vt:$src2, i32:$src2_mods)))),
(med3Inst $src0_mods, $src0, $src1_mods, $src1, $src2_mods, $src2,
DSTCLAMP.NONE, DSTOMOD.NONE)>;
// This matches 16 permutations of min(max(x, y), max(min(x, y), z))
def : GCNPat<
(fminnum_like_nnan
(fmaxnum_like (VOP3Mods vt:$src0, i32:$src0_mods),
(VOP3Mods vt:$src1, i32:$src1_mods)),
(fmaxnum_like (fminnum_like (VOP3Mods vt:$src0, i32:$src0_mods),
(VOP3Mods vt:$src1, i32:$src1_mods)),
(vt (VOP3Mods vt:$src2, i32:$src2_mods)))),
(med3Inst $src0_mods, $src0, $src1_mods, $src1, $src2_mods, $src2,
DSTCLAMP.NONE, DSTOMOD.NONE)>;
}
class FP16Med3Pat<ValueType vt,
Instruction med3Inst> : GCNPat<
(fmaxnum_like_nnan (fminnum_like (VOP3Mods vt:$src0, i32:$src0_mods),
(VOP3Mods vt:$src1, i32:$src1_mods)),
(fminnum_like (fmaxnum_like (VOP3Mods vt:$src0, i32:$src0_mods),
(VOP3Mods vt:$src1, i32:$src1_mods)),
(vt (VOP3Mods vt:$src2, i32:$src2_mods)))),
(med3Inst $src0_mods, $src0, $src1_mods, $src1, $src2_mods, $src2, DSTCLAMP.NONE)
>;
multiclass Int16Med3Pat<Instruction med3Inst,
SDPatternOperator min,
SDPatternOperator max> {
// This matches 16 permutations of
// max(min(x, y), min(max(x, y), z))
def : GCNPat <
(max (min i16:$src0, i16:$src1),
(min (max i16:$src0, i16:$src1), i16:$src2)),
(med3Inst SRCMODS.NONE, VSrc_b16:$src0, SRCMODS.NONE, VSrc_b16:$src1, SRCMODS.NONE, VSrc_b16:$src2, DSTCLAMP.NONE)
>;
// This matches 16 permutations of
// min(max(a, b), max(min(a, b), c))
def : GCNPat <
(min (max i16:$src0, i16:$src1),
(max (min i16:$src0, i16:$src1), i16:$src2)),
(med3Inst SRCMODS.NONE, VSrc_b16:$src0, SRCMODS.NONE, VSrc_b16:$src1, SRCMODS.NONE, VSrc_b16:$src2, DSTCLAMP.NONE)
>;
}
defm : FPMed3Pat<f32, V_MED3_F32_e64>;
class
IntMinMaxPat<Instruction minmaxInst, SDPatternOperator min_or_max,
SDPatternOperator max_or_min_oneuse> : AMDGPUPat <
(DivergentBinFrag<min_or_max> (max_or_min_oneuse i32:$src0, i32:$src1),
i32:$src2),
(minmaxInst VSrc_b32:$src0, VSrc_b32:$src1, VSrc_b32:$src2)
>;
class
FPMinMaxPat<Instruction minmaxInst, ValueType vt, SDPatternOperator min_or_max,
SDPatternOperator max_or_min_oneuse> : GCNPat <
(min_or_max (max_or_min_oneuse (VOP3Mods vt:$src0, i32:$src0_mods),
(VOP3Mods vt:$src1, i32:$src1_mods)),
(vt (VOP3Mods vt:$src2, i32:$src2_mods))),
(minmaxInst $src0_mods, $src0, $src1_mods, $src1, $src2_mods, $src2,
DSTCLAMP.NONE, DSTOMOD.NONE)
>;
let OtherPredicates = [isGFX11Plus] in {
def : IntMinMaxPat<V_MAXMIN_I32_e64, smin, smax_oneuse>;
def : IntMinMaxPat<V_MINMAX_I32_e64, smax, smin_oneuse>;
def : IntMinMaxPat<V_MAXMIN_U32_e64, umin, umax_oneuse>;
def : IntMinMaxPat<V_MINMAX_U32_e64, umax, umin_oneuse>;
def : FPMinMaxPat<V_MINMAX_F32_e64, f32, fmaxnum_like, fminnum_like_oneuse>;
def : FPMinMaxPat<V_MAXMIN_F32_e64, f32, fminnum_like, fmaxnum_like_oneuse>;
def : FPMinMaxPat<V_MINMAX_F16_e64, f16, fmaxnum_like, fminnum_like_oneuse>;
def : FPMinMaxPat<V_MAXMIN_F16_e64, f16, fminnum_like, fmaxnum_like_oneuse>;
}
let OtherPredicates = [isGFX9Plus] in {
def : FP16Med3Pat<f16, V_MED3_F16_e64>;
defm : Int16Med3Pat<V_MED3_I16_e64, smin, smax>;
defm : Int16Med3Pat<V_MED3_U16_e64, umin, umax>;
} // End Predicates = [isGFX9Plus]
let OtherPredicates = [isGFX12Plus] in {
def : FPMinMaxPat<V_MINIMUMMAXIMUM_F32_e64, f32, DivergentBinFrag<fmaximum>, fminimum_oneuse>;
def : FPMinMaxPat<V_MAXIMUMMINIMUM_F32_e64, f32, DivergentBinFrag<fminimum>, fmaximum_oneuse>;
def : FPMinMaxPat<V_MINIMUMMAXIMUM_F16_e64, f16, DivergentBinFrag<fmaximum>, fminimum_oneuse>;
def : FPMinMaxPat<V_MAXIMUMMINIMUM_F16_e64, f16, DivergentBinFrag<fminimum>, fmaximum_oneuse>;
}
// Convert a floating-point power of 2 to the integer exponent.
def FPPow2ToExponentXForm : SDNodeXForm<fpimm, [{
const auto &APF = N->getValueAPF();
int Log2 = APF.getExactLog2Abs();
assert(Log2 != INT_MIN);
return CurDAG->getTargetConstant(Log2, SDLoc(N), MVT::i32);
}]>;
// Check if a floating point value is a power of 2 floating-point
// immediate where it's preferable to emit a multiply by as an
// ldexp. We skip over 0.5 to 4.0 as those are inline immediates
// anyway.
def fpimm_pos_pow2_prefer_ldexp_f64 : FPImmLeaf<f64, [{
if (Imm.isNegative())
return false;
int Exp = Imm.getExactLog2Abs();
// Prefer leaving the FP inline immediates as they are.
// 0.5, 1.0, 2.0, 4.0
// For f64 ldexp is always better than materializing a 64-bit
// constant.
return Exp != INT_MIN && (Exp < -1 || Exp > 2);
}], FPPow2ToExponentXForm
>;
def fpimm_neg_pow2_prefer_ldexp_f64 : FPImmLeaf<f64, [{
if (!Imm.isNegative())
return false;
int Exp = Imm.getExactLog2Abs();
// Prefer leaving the FP inline immediates as they are.
// 0.5, 1.0, 2.0, 4.0
// For f64 ldexp is always better than materializing a 64-bit
// constant.
return Exp != INT_MIN && (Exp < -1 || Exp > 2);
}], FPPow2ToExponentXForm
>;
// f64 is different because we also want to handle cases that may
// require materialization of the exponent.
// TODO: If we know f64 ops are fast, prefer add (ldexp x, N), y over fma
// TODO: For f32/f16, it's not a clear win on code size to use ldexp
// in place of mul since we have to use the vop3 form. Are there power
// savings or some other reason to prefer ldexp over mul?
def : GCNPat<
(any_fmul (f64 (VOP3Mods f64:$src0, i32:$src0_mods)),
fpimm_pos_pow2_prefer_ldexp_f64:$src1),
(V_LDEXP_F64_e64 i32:$src0_mods, VSrc_b64:$src0,
0, (S_MOV_B32 (i32 (FPPow2ToExponentXForm $src1))))
>;
def : GCNPat<
(any_fmul f64:$src0, fpimm_neg_pow2_prefer_ldexp_f64:$src1),
(V_LDEXP_F64_e64 SRCMODS.NEG, VSrc_b64:$src0,
0, (S_MOV_B32 (i32 (FPPow2ToExponentXForm $src1))))
>;
// We want to avoid using VOP3Mods which could pull in another fneg
// which we would need to be re-negated (which should never happen in
// practice). I don't see a way to apply an SDNodeXForm that accounts
// for a second operand.
def : GCNPat<
(any_fmul (fabs f64:$src0), fpimm_neg_pow2_prefer_ldexp_f64:$src1),
(V_LDEXP_F64_e64 SRCMODS.NEG_ABS, VSrc_b64:$src0,
0, (S_MOV_B32 (i32 (FPPow2ToExponentXForm $src1))))
>;
class AMDGPUGenericInstruction : GenericInstruction {
let Namespace = "AMDGPU";
}
// Convert a wave address to a swizzled vector address (i.e. this is
// for copying the stack pointer to a vector address appropriate to
// use in the offset field of mubuf instructions).
def G_AMDGPU_WAVE_ADDRESS : AMDGPUGenericInstruction {
let OutOperandList = (outs type0:$dst);
let InOperandList = (ins type0:$src);
let hasSideEffects = 0;
}
// Returns -1 if the input is zero.
def G_AMDGPU_FFBH_U32 : AMDGPUGenericInstruction {
let OutOperandList = (outs type0:$dst);
let InOperandList = (ins type1:$src);
let hasSideEffects = 0;
}
// Returns -1 if the input is zero.
def G_AMDGPU_FFBL_B32 : AMDGPUGenericInstruction {
let OutOperandList = (outs type0:$dst);
let InOperandList = (ins type1:$src);
let hasSideEffects = 0;
}
def G_AMDGPU_RCP_IFLAG : AMDGPUGenericInstruction {
let OutOperandList = (outs type0:$dst);
let InOperandList = (ins type1:$src);
let hasSideEffects = 0;
}
class BufferLoadGenericInstruction : AMDGPUGenericInstruction {
let OutOperandList = (outs type0:$dst);
let InOperandList = (ins type1:$rsrc, type2:$vindex, type2:$voffset,
type2:$soffset, untyped_imm_0:$offset,
untyped_imm_0:$cachepolicy, untyped_imm_0:$idxen);
let hasSideEffects = 0;
let mayLoad = 1;
}
class TBufferLoadGenericInstruction : AMDGPUGenericInstruction {
let OutOperandList = (outs type0:$dst);
let InOperandList = (ins type1:$rsrc, type2:$vindex, type2:$voffset,
type2:$soffset, untyped_imm_0:$offset, untyped_imm_0:$format,
untyped_imm_0:$cachepolicy, untyped_imm_0:$idxen);
let hasSideEffects = 0;
let mayLoad = 1;
}
def G_AMDGPU_BUFFER_LOAD_UBYTE : BufferLoadGenericInstruction;
def G_AMDGPU_BUFFER_LOAD_SBYTE : BufferLoadGenericInstruction;
def G_AMDGPU_BUFFER_LOAD_USHORT : BufferLoadGenericInstruction;
def G_AMDGPU_BUFFER_LOAD_SSHORT : BufferLoadGenericInstruction;
def G_AMDGPU_BUFFER_LOAD : BufferLoadGenericInstruction;
def G_AMDGPU_BUFFER_LOAD_FORMAT : BufferLoadGenericInstruction;
def G_AMDGPU_BUFFER_LOAD_FORMAT_TFE : BufferLoadGenericInstruction;
def G_AMDGPU_BUFFER_LOAD_FORMAT_D16 : BufferLoadGenericInstruction;
def G_AMDGPU_TBUFFER_LOAD_FORMAT : TBufferLoadGenericInstruction;
def G_AMDGPU_TBUFFER_LOAD_FORMAT_D16 : TBufferLoadGenericInstruction;
class BufferStoreGenericInstruction : AMDGPUGenericInstruction {
let OutOperandList = (outs);
let InOperandList = (ins type0:$vdata, type1:$rsrc, type2:$vindex, type2:$voffset,
type2:$soffset, untyped_imm_0:$offset,
untyped_imm_0:$cachepolicy, untyped_imm_0:$idxen);
let hasSideEffects = 0;
let mayStore = 1;
}
class TBufferStoreGenericInstruction : AMDGPUGenericInstruction {
let OutOperandList = (outs);
let InOperandList = (ins type0:$vdata, type1:$rsrc, type2:$vindex, type2:$voffset,
type2:$soffset, untyped_imm_0:$offset,
untyped_imm_0:$format,
untyped_imm_0:$cachepolicy, untyped_imm_0:$idxen);
let hasSideEffects = 0;
let mayStore = 1;
}
def G_AMDGPU_BUFFER_STORE : BufferStoreGenericInstruction;
def G_AMDGPU_BUFFER_STORE_BYTE : BufferStoreGenericInstruction;
def G_AMDGPU_BUFFER_STORE_SHORT : BufferStoreGenericInstruction;
def G_AMDGPU_BUFFER_STORE_FORMAT : BufferStoreGenericInstruction;
def G_AMDGPU_BUFFER_STORE_FORMAT_D16 : BufferStoreGenericInstruction;
def G_AMDGPU_TBUFFER_STORE_FORMAT : TBufferStoreGenericInstruction;
def G_AMDGPU_TBUFFER_STORE_FORMAT_D16 : TBufferStoreGenericInstruction;
def G_AMDGPU_FMIN_LEGACY : AMDGPUGenericInstruction {
let OutOperandList = (outs type0:$dst);
let InOperandList = (ins type0:$src0, type0:$src1);
let hasSideEffects = 0;
}
def G_AMDGPU_FMAX_LEGACY : AMDGPUGenericInstruction {
let OutOperandList = (outs type0:$dst);
let InOperandList = (ins type0:$src0, type0:$src1);
let hasSideEffects = 0;
}
foreach N = 0-3 in {
def G_AMDGPU_CVT_F32_UBYTE#N : AMDGPUGenericInstruction {
let OutOperandList = (outs type0:$dst);
let InOperandList = (ins type0:$src0);
let hasSideEffects = 0;
}
}
def G_AMDGPU_CVT_PK_I16_I32 : AMDGPUGenericInstruction {
let OutOperandList = (outs type0:$dst);
let InOperandList = (ins type0:$src0, type0:$src1);
let hasSideEffects = 0;
}
def G_AMDGPU_SMED3 : AMDGPUGenericInstruction {
let OutOperandList = (outs type0:$dst);
let InOperandList = (ins type0:$src0, type0:$src1, type0:$src2);
let hasSideEffects = 0;
}
def G_AMDGPU_UMED3 : AMDGPUGenericInstruction {
let OutOperandList = (outs type0:$dst);
let InOperandList = (ins type0:$src0, type0:$src1, type0:$src2);
let hasSideEffects = 0;
}
def G_AMDGPU_FMED3 : AMDGPUGenericInstruction {
let OutOperandList = (outs type0:$dst);
let InOperandList = (ins type0:$src0, type0:$src1, type0:$src2);
let hasSideEffects = 0;
}
def G_AMDGPU_CLAMP : AMDGPUGenericInstruction {
let OutOperandList = (outs type0:$dst);
let InOperandList = (ins type0:$src);
let hasSideEffects = 0;
}
// Integer multiply-add: arg0 * arg1 + arg2.
//
// arg0 and arg1 are 32-bit integers (interpreted as signed or unsigned),
// arg2 is a 64-bit integer. Result is a 64-bit integer and a 1-bit carry-out.
class G_AMDGPU_MAD_64_32 : AMDGPUGenericInstruction {
let OutOperandList = (outs type0:$dst, type1:$carry_out);
let InOperandList = (ins type2:$arg0, type2:$arg1, type0:$arg2);
let hasSideEffects = 0;
}
def G_AMDGPU_MAD_U64_U32 : G_AMDGPU_MAD_64_32;
def G_AMDGPU_MAD_I64_I32 : G_AMDGPU_MAD_64_32;
// Atomic cmpxchg. $cmpval ad $newval are packed in a single vector
// operand Expects a MachineMemOperand in addition to explicit
// operands.
def G_AMDGPU_ATOMIC_CMPXCHG : AMDGPUGenericInstruction {
let OutOperandList = (outs type0:$oldval);
let InOperandList = (ins ptype1:$addr, type0:$cmpval_newval);
let hasSideEffects = 0;
let mayLoad = 1;
let mayStore = 1;
}
let Namespace = "AMDGPU" in {
def G_AMDGPU_ATOMIC_FMIN : G_ATOMICRMW_OP;
def G_AMDGPU_ATOMIC_FMAX : G_ATOMICRMW_OP;
}
class BufferAtomicGenericInstruction : AMDGPUGenericInstruction {
let OutOperandList = (outs type0:$dst);
let InOperandList = (ins type0:$vdata, type1:$rsrc, type2:$vindex, type2:$voffset,
type2:$soffset, untyped_imm_0:$offset,
untyped_imm_0:$cachepolicy, untyped_imm_0:$idxen);
let hasSideEffects = 0;
let mayLoad = 1;
let mayStore = 1;
}
def G_AMDGPU_BUFFER_ATOMIC_SWAP : BufferAtomicGenericInstruction;
def G_AMDGPU_BUFFER_ATOMIC_ADD : BufferAtomicGenericInstruction;
def G_AMDGPU_BUFFER_ATOMIC_SUB : BufferAtomicGenericInstruction;
def G_AMDGPU_BUFFER_ATOMIC_SMIN : BufferAtomicGenericInstruction;
def G_AMDGPU_BUFFER_ATOMIC_UMIN : BufferAtomicGenericInstruction;
def G_AMDGPU_BUFFER_ATOMIC_SMAX : BufferAtomicGenericInstruction;
def G_AMDGPU_BUFFER_ATOMIC_UMAX : BufferAtomicGenericInstruction;
def G_AMDGPU_BUFFER_ATOMIC_AND : BufferAtomicGenericInstruction;
def G_AMDGPU_BUFFER_ATOMIC_OR : BufferAtomicGenericInstruction;
def G_AMDGPU_BUFFER_ATOMIC_XOR : BufferAtomicGenericInstruction;
def G_AMDGPU_BUFFER_ATOMIC_INC : BufferAtomicGenericInstruction;
def G_AMDGPU_BUFFER_ATOMIC_DEC : BufferAtomicGenericInstruction;
def G_AMDGPU_BUFFER_ATOMIC_FADD : BufferAtomicGenericInstruction;
def G_AMDGPU_BUFFER_ATOMIC_FMIN : BufferAtomicGenericInstruction;
def G_AMDGPU_BUFFER_ATOMIC_FMAX : BufferAtomicGenericInstruction;
def G_AMDGPU_BUFFER_ATOMIC_CMPSWAP : AMDGPUGenericInstruction {
let OutOperandList = (outs type0:$dst);
let InOperandList = (ins type0:$vdata, type0:$cmp, type1:$rsrc, type2:$vindex,
type2:$voffset, type2:$soffset, untyped_imm_0:$offset,
untyped_imm_0:$cachepolicy, untyped_imm_0:$idxen);
let hasSideEffects = 0;
let mayLoad = 1;
let mayStore = 1;
}
// Wrapper around llvm.amdgcn.s.buffer.load. This is mostly needed as
// a workaround for the intrinsic being defined as readnone, but
// really needs a memory operand.
def G_AMDGPU_S_BUFFER_LOAD : AMDGPUGenericInstruction {
let OutOperandList = (outs type0:$dst);
let InOperandList = (ins type1:$rsrc, type2:$offset, untyped_imm_0:$cachepolicy);
let hasSideEffects = 0;
let mayLoad = 1;
let mayStore = 0;
}
def G_AMDGPU_S_MUL_U64_U32 : AMDGPUGenericInstruction {
let OutOperandList = (outs type0:$dst);
let InOperandList = (ins type0:$src0, type0:$src1);
let hasSideEffects = 0;
}
def G_AMDGPU_S_MUL_I64_I32 : AMDGPUGenericInstruction {
let OutOperandList = (outs type0:$dst);
let InOperandList = (ins type0:$src0, type0:$src1);
let hasSideEffects = 0;
}
// This is equivalent to the G_INTRINSIC*, but the operands may have
// been legalized depending on the subtarget requirements.
def G_AMDGPU_INTRIN_IMAGE_LOAD : AMDGPUGenericInstruction {
let OutOperandList = (outs type0:$dst);
let InOperandList = (ins unknown:$intrin, variable_ops);
let hasSideEffects = 0;
let mayLoad = 1;
// FIXME: Use separate opcode for atomics.
let mayStore = 1;
}
def G_AMDGPU_INTRIN_IMAGE_LOAD_D16 : AMDGPUGenericInstruction {
let OutOperandList = (outs type0:$dst);
let InOperandList = (ins unknown:$intrin, variable_ops);
let hasSideEffects = 0;
let mayLoad = 1;
// FIXME: Use separate opcode for atomics.
let mayStore = 1;
}
// This is equivalent to the G_INTRINSIC*, but the operands may have
// been legalized depending on the subtarget requirements.
def G_AMDGPU_INTRIN_IMAGE_STORE : AMDGPUGenericInstruction {
let OutOperandList = (outs);
let InOperandList = (ins unknown:$intrin, variable_ops);
let hasSideEffects = 0;
let mayStore = 1;
}
def G_AMDGPU_INTRIN_IMAGE_STORE_D16 : AMDGPUGenericInstruction {
let OutOperandList = (outs);
let InOperandList = (ins unknown:$intrin, variable_ops);
let hasSideEffects = 0;
let mayStore = 1;
}
def G_AMDGPU_INTRIN_BVH_INTERSECT_RAY : AMDGPUGenericInstruction {
let OutOperandList = (outs type0:$dst);
let InOperandList = (ins unknown:$intrin, variable_ops);
let hasSideEffects = 0;
let mayLoad = 1;
let mayStore = 0;
}
// Generic instruction for SI_CALL, so we can select the register bank and insert a waterfall loop
// if necessary.
def G_SI_CALL : AMDGPUGenericInstruction {
let OutOperandList = (outs SReg_64:$dst);
let InOperandList = (ins type0:$src0, unknown:$callee);
let Size = 4;
let isCall = 1;
let UseNamedOperandTable = 1;
let SchedRW = [WriteBranch];
// TODO: Should really base this on the call target
let isConvergent = 1;
}
def G_FPTRUNC_ROUND_UPWARD : AMDGPUGenericInstruction {
let OutOperandList = (outs type0:$vdst);
let InOperandList = (ins type1:$src0);
let hasSideEffects = 0;
}
def G_FPTRUNC_ROUND_DOWNWARD : AMDGPUGenericInstruction {
let OutOperandList = (outs type0:$vdst);
let InOperandList = (ins type1:$src0);
let hasSideEffects = 0;
}
//============================================================================//
// Dummy Instructions
//============================================================================//
def V_ILLEGAL : Enc32, InstSI<(outs), (ins), "v_illegal"> {
let Inst{31-0} = 0x00000000;
let FixedSize = 1;
let Size = 4;
let Uses = [EXEC];
let hasSideEffects = 1;
let SubtargetPredicate = isGFX10Plus;
}
|