summaryrefslogtreecommitdiff
path: root/lib/Target/NVPTX/NVPTXInstrInfo.td
blob: e94250b38da7e33fbfd4e694bbc85c77bd57f875 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
//===- NVPTXInstrInfo.td - NVPTX Instruction defs -------------*- tblgen-*-===//
//
//                     The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file describes the PTX instructions in TableGen format.
//
//===----------------------------------------------------------------------===//

include "NVPTXInstrFormats.td"

// A NOP instruction
def NOP : NVPTXInst<(outs), (ins), "", []>;

// List of vector specific properties
def isVecLD      : VecInstTypeEnum<1>;
def isVecST      : VecInstTypeEnum<2>;
def isVecBuild   : VecInstTypeEnum<3>;
def isVecShuffle : VecInstTypeEnum<4>;
def isVecExtract : VecInstTypeEnum<5>;
def isVecInsert  : VecInstTypeEnum<6>;
def isVecDest    : VecInstTypeEnum<7>;
def isVecOther   : VecInstTypeEnum<15>;

//===----------------------------------------------------------------------===//
// NVPTX Operand Definitions.
//===----------------------------------------------------------------------===//

def brtarget    : Operand<OtherVT>;

// CVT conversion modes
// These must match the enum in NVPTX.h
def CvtNONE : PatLeaf<(i32 0x0)>;
def CvtRNI  : PatLeaf<(i32 0x1)>;
def CvtRZI  : PatLeaf<(i32 0x2)>;
def CvtRMI  : PatLeaf<(i32 0x3)>;
def CvtRPI  : PatLeaf<(i32 0x4)>;
def CvtRN   : PatLeaf<(i32 0x5)>;
def CvtRZ   : PatLeaf<(i32 0x6)>;
def CvtRM   : PatLeaf<(i32 0x7)>;
def CvtRP   : PatLeaf<(i32 0x8)>;

def CvtNONE_FTZ : PatLeaf<(i32 0x10)>;
def CvtRNI_FTZ  : PatLeaf<(i32 0x11)>;
def CvtRZI_FTZ  : PatLeaf<(i32 0x12)>;
def CvtRMI_FTZ  : PatLeaf<(i32 0x13)>;
def CvtRPI_FTZ  : PatLeaf<(i32 0x14)>;
def CvtRN_FTZ   : PatLeaf<(i32 0x15)>;
def CvtRZ_FTZ   : PatLeaf<(i32 0x16)>;
def CvtRM_FTZ   : PatLeaf<(i32 0x17)>;
def CvtRP_FTZ   : PatLeaf<(i32 0x18)>;

def CvtSAT      : PatLeaf<(i32 0x20)>;
def CvtSAT_FTZ  : PatLeaf<(i32 0x30)>;

def CvtMode : Operand<i32> {
  let PrintMethod = "printCvtMode";
}

// Compare modes
// These must match the enum in NVPTX.h
def CmpEQ   : PatLeaf<(i32 0)>;
def CmpNE   : PatLeaf<(i32 1)>;
def CmpLT   : PatLeaf<(i32 2)>;
def CmpLE   : PatLeaf<(i32 3)>;
def CmpGT   : PatLeaf<(i32 4)>;
def CmpGE   : PatLeaf<(i32 5)>;
def CmpLO   : PatLeaf<(i32 6)>;
def CmpLS   : PatLeaf<(i32 7)>;
def CmpHI   : PatLeaf<(i32 8)>;
def CmpHS   : PatLeaf<(i32 9)>;
def CmpEQU  : PatLeaf<(i32 10)>;
def CmpNEU  : PatLeaf<(i32 11)>;
def CmpLTU  : PatLeaf<(i32 12)>;
def CmpLEU  : PatLeaf<(i32 13)>;
def CmpGTU  : PatLeaf<(i32 14)>;
def CmpGEU  : PatLeaf<(i32 15)>;
def CmpNUM  : PatLeaf<(i32 16)>;
def CmpNAN  : PatLeaf<(i32 17)>;

def CmpEQ_FTZ   : PatLeaf<(i32 0x100)>;
def CmpNE_FTZ   : PatLeaf<(i32 0x101)>;
def CmpLT_FTZ   : PatLeaf<(i32 0x102)>;
def CmpLE_FTZ   : PatLeaf<(i32 0x103)>;
def CmpGT_FTZ   : PatLeaf<(i32 0x104)>;
def CmpGE_FTZ   : PatLeaf<(i32 0x105)>;
def CmpLO_FTZ   : PatLeaf<(i32 0x106)>;
def CmpLS_FTZ   : PatLeaf<(i32 0x107)>;
def CmpHI_FTZ   : PatLeaf<(i32 0x108)>;
def CmpHS_FTZ   : PatLeaf<(i32 0x109)>;
def CmpEQU_FTZ  : PatLeaf<(i32 0x10A)>;
def CmpNEU_FTZ  : PatLeaf<(i32 0x10B)>;
def CmpLTU_FTZ  : PatLeaf<(i32 0x10C)>;
def CmpLEU_FTZ  : PatLeaf<(i32 0x10D)>;
def CmpGTU_FTZ  : PatLeaf<(i32 0x10E)>;
def CmpGEU_FTZ  : PatLeaf<(i32 0x10F)>;
def CmpNUM_FTZ  : PatLeaf<(i32 0x110)>;
def CmpNAN_FTZ  : PatLeaf<(i32 0x111)>;

def CmpMode : Operand<i32> {
  let PrintMethod = "printCmpMode";
}

def F32ConstZero : Operand<f32>, PatLeaf<(f32 fpimm)>, SDNodeXForm<fpimm, [{
    return CurDAG->getTargetConstantFP(0.0, MVT::f32);
  }]>;
def F32ConstOne : Operand<f32>, PatLeaf<(f32 fpimm)>, SDNodeXForm<fpimm, [{
    return CurDAG->getTargetConstantFP(1.0, MVT::f32);
  }]>;

//===----------------------------------------------------------------------===//
// NVPTX Instruction Predicate Definitions
//===----------------------------------------------------------------------===//


def hasAtomRedG32 : Predicate<"Subtarget.hasAtomRedG32()">;
def hasAtomRedS32 : Predicate<"Subtarget.hasAtomRedS32()">;
def hasAtomRedGen32 : Predicate<"Subtarget.hasAtomRedGen32()">;
def useAtomRedG32forGen32 :
  Predicate<"!Subtarget.hasAtomRedGen32() && Subtarget.hasAtomRedG32()">;
def hasBrkPt : Predicate<"Subtarget.hasBrkPt()">;
def hasAtomRedG64 : Predicate<"Subtarget.hasAtomRedG64()">;
def hasAtomRedS64 : Predicate<"Subtarget.hasAtomRedS64()">;
def hasAtomRedGen64 : Predicate<"Subtarget.hasAtomRedGen64()">;
def useAtomRedG64forGen64 :
  Predicate<"!Subtarget.hasAtomRedGen64() && Subtarget.hasAtomRedG64()">;
def hasAtomAddF32 : Predicate<"Subtarget.hasAtomAddF32()">;
def hasVote : Predicate<"Subtarget.hasVote()">;
def hasDouble : Predicate<"Subtarget.hasDouble()">;
def reqPTX20 : Predicate<"Subtarget.reqPTX20()">;
def hasLDG : Predicate<"Subtarget.hasLDG()">;
def hasLDU : Predicate<"Subtarget.hasLDU()">;
def hasGenericLdSt : Predicate<"Subtarget.hasGenericLdSt()">;

def doF32FTZ : Predicate<"useF32FTZ()">;
def doNoF32FTZ : Predicate<"!useF32FTZ()">;

def doFMAF32      : Predicate<"doFMAF32">;
def doFMAF32_ftz  : Predicate<"(doFMAF32 && useF32FTZ())">;
def doFMAF32AGG      : Predicate<"doFMAF32AGG">;
def doFMAF32AGG_ftz  : Predicate<"(doFMAF32AGG && useF32FTZ())">;
def doFMAF64      : Predicate<"doFMAF64">;
def doFMAF64AGG      : Predicate<"doFMAF64AGG">;

def doMulWide      : Predicate<"doMulWide">;

def allowFMA : Predicate<"allowFMA">;
def allowFMA_ftz : Predicate<"(allowFMA && useF32FTZ())">;

def do_DIVF32_APPROX : Predicate<"getDivF32Level()==0">;
def do_DIVF32_FULL : Predicate<"getDivF32Level()==1">;

def do_SQRTF32_APPROX : Predicate<"!usePrecSqrtF32()">;
def do_SQRTF32_RN : Predicate<"usePrecSqrtF32()">;

def hasHWROT32 : Predicate<"Subtarget.hasHWROT32()">;

def true : Predicate<"1">;

def hasPTX31 : Predicate<"Subtarget.getPTXVersion() >= 31">;


//===----------------------------------------------------------------------===//
// Some Common Instruction Class Templates
//===----------------------------------------------------------------------===//

multiclass I3<string OpcStr, SDNode OpNode> {
  def i64rr : NVPTXInst<(outs Int64Regs:$dst), (ins Int64Regs:$a, Int64Regs:$b),
                     !strconcat(OpcStr, "64 \t$dst, $a, $b;"),
                     [(set Int64Regs:$dst, (OpNode Int64Regs:$a,
                       Int64Regs:$b))]>;
  def i64ri : NVPTXInst<(outs Int64Regs:$dst), (ins Int64Regs:$a, i64imm:$b),
                     !strconcat(OpcStr, "64 \t$dst, $a, $b;"),
                     [(set Int64Regs:$dst, (OpNode Int64Regs:$a, imm:$b))]>;
  def i32rr : NVPTXInst<(outs Int32Regs:$dst), (ins Int32Regs:$a, Int32Regs:$b),
                     !strconcat(OpcStr, "32 \t$dst, $a, $b;"),
                     [(set Int32Regs:$dst, (OpNode Int32Regs:$a,
                       Int32Regs:$b))]>;
  def i32ri : NVPTXInst<(outs Int32Regs:$dst), (ins Int32Regs:$a, i32imm:$b),
                     !strconcat(OpcStr, "32 \t$dst, $a, $b;"),
                     [(set Int32Regs:$dst, (OpNode Int32Regs:$a, imm:$b))]>;
  def i16rr : NVPTXInst<(outs Int16Regs:$dst), (ins Int16Regs:$a, Int16Regs:$b),
                     !strconcat(OpcStr, "16 \t$dst, $a, $b;"),
                     [(set Int16Regs:$dst, (OpNode Int16Regs:$a,
                       Int16Regs:$b))]>;
  def i16ri : NVPTXInst<(outs Int16Regs:$dst), (ins Int16Regs:$a, i16imm:$b),
                     !strconcat(OpcStr, "16 \t$dst, $a, $b;"),
                     [(set Int16Regs:$dst, (OpNode Int16Regs:$a, (imm):$b))]>;
}

multiclass ADD_SUB_INT_32<string OpcStr, SDNode OpNode> {
   def i32rr : NVPTXInst<(outs Int32Regs:$dst), (ins Int32Regs:$a,
       Int32Regs:$b),
                      !strconcat(OpcStr, ".s32 \t$dst, $a, $b;"),
                      [(set Int32Regs:$dst, (OpNode Int32Regs:$a,
                        Int32Regs:$b))]>;
   def i32ri : NVPTXInst<(outs Int32Regs:$dst), (ins Int32Regs:$a, i32imm:$b),
                      !strconcat(OpcStr, ".s32 \t$dst, $a, $b;"),
                      [(set Int32Regs:$dst, (OpNode Int32Regs:$a, imm:$b))]>;
}

multiclass F3<string OpcStr, SDNode OpNode> {
   def f64rr : NVPTXInst<(outs Float64Regs:$dst),
                      (ins Float64Regs:$a, Float64Regs:$b),
                      !strconcat(OpcStr, ".f64 \t$dst, $a, $b;"),
                      [(set Float64Regs:$dst,
                        (OpNode Float64Regs:$a, Float64Regs:$b))]>,
                      Requires<[allowFMA]>;
   def f64ri : NVPTXInst<(outs Float64Regs:$dst),
                      (ins Float64Regs:$a, f64imm:$b),
                      !strconcat(OpcStr, ".f64 \t$dst, $a, $b;"),
                      [(set Float64Regs:$dst,
                        (OpNode Float64Regs:$a, fpimm:$b))]>,
                      Requires<[allowFMA]>;
   def f32rr_ftz : NVPTXInst<(outs Float32Regs:$dst),
                      (ins Float32Regs:$a, Float32Regs:$b),
                      !strconcat(OpcStr, ".ftz.f32 \t$dst, $a, $b;"),
                      [(set Float32Regs:$dst,
                        (OpNode Float32Regs:$a, Float32Regs:$b))]>,
                      Requires<[allowFMA_ftz]>;
   def f32ri_ftz : NVPTXInst<(outs Float32Regs:$dst),
                      (ins Float32Regs:$a, f32imm:$b),
                      !strconcat(OpcStr, ".ftz.f32 \t$dst, $a, $b;"),
                      [(set Float32Regs:$dst,
                        (OpNode Float32Regs:$a, fpimm:$b))]>,
                      Requires<[allowFMA_ftz]>;
   def f32rr : NVPTXInst<(outs Float32Regs:$dst),
                      (ins Float32Regs:$a, Float32Regs:$b),
                      !strconcat(OpcStr, ".f32 \t$dst, $a, $b;"),
                      [(set Float32Regs:$dst,
                        (OpNode Float32Regs:$a, Float32Regs:$b))]>,
                      Requires<[allowFMA]>;
   def f32ri : NVPTXInst<(outs Float32Regs:$dst),
                      (ins Float32Regs:$a, f32imm:$b),
                      !strconcat(OpcStr, ".f32 \t$dst, $a, $b;"),
                      [(set Float32Regs:$dst,
                        (OpNode Float32Regs:$a, fpimm:$b))]>,
                      Requires<[allowFMA]>;
}

multiclass F3_rn<string OpcStr, SDNode OpNode> {
   def f64rr : NVPTXInst<(outs Float64Regs:$dst),
                      (ins Float64Regs:$a, Float64Regs:$b),
                      !strconcat(OpcStr, ".rn.f64 \t$dst, $a, $b;"),
                      [(set Float64Regs:$dst,
                        (OpNode Float64Regs:$a, Float64Regs:$b))]>;
   def f64ri : NVPTXInst<(outs Float64Regs:$dst),
                      (ins Float64Regs:$a, f64imm:$b),
                      !strconcat(OpcStr, ".rn.f64 \t$dst, $a, $b;"),
                      [(set Float64Regs:$dst,
                        (OpNode Float64Regs:$a, fpimm:$b))]>;
   def f32rr_ftz : NVPTXInst<(outs Float32Regs:$dst),
                      (ins Float32Regs:$a, Float32Regs:$b),
                      !strconcat(OpcStr, ".rn.ftz.f32 \t$dst, $a, $b;"),
                      [(set Float32Regs:$dst,
                        (OpNode Float32Regs:$a, Float32Regs:$b))]>,
                      Requires<[doF32FTZ]>;
   def f32ri_ftz : NVPTXInst<(outs Float32Regs:$dst),
                      (ins Float32Regs:$a, f32imm:$b),
                      !strconcat(OpcStr, ".rn.ftz.f32 \t$dst, $a, $b;"),
                      [(set Float32Regs:$dst,
                        (OpNode Float32Regs:$a, fpimm:$b))]>,
                      Requires<[doF32FTZ]>;
   def f32rr : NVPTXInst<(outs Float32Regs:$dst),
                      (ins Float32Regs:$a, Float32Regs:$b),
                      !strconcat(OpcStr, ".rn.f32 \t$dst, $a, $b;"),
                      [(set Float32Regs:$dst,
                        (OpNode Float32Regs:$a, Float32Regs:$b))]>;
   def f32ri : NVPTXInst<(outs Float32Regs:$dst),
                      (ins Float32Regs:$a, f32imm:$b),
                      !strconcat(OpcStr, ".rn.f32 \t$dst, $a, $b;"),
                      [(set Float32Regs:$dst,
                        (OpNode Float32Regs:$a, fpimm:$b))]>;
}

multiclass F2<string OpcStr, SDNode OpNode> {
   def f64 : NVPTXInst<(outs Float64Regs:$dst), (ins Float64Regs:$a),
                      !strconcat(OpcStr, ".f64 \t$dst, $a;"),
                      [(set Float64Regs:$dst, (OpNode Float64Regs:$a))]>;
   def f32_ftz : NVPTXInst<(outs Float32Regs:$dst), (ins Float32Regs:$a),
                      !strconcat(OpcStr, ".ftz.f32 \t$dst, $a;"),
                      [(set Float32Regs:$dst, (OpNode Float32Regs:$a))]>,
                      Requires<[doF32FTZ]>;
   def f32 : NVPTXInst<(outs Float32Regs:$dst), (ins Float32Regs:$a),
                      !strconcat(OpcStr, ".f32 \t$dst, $a;"),
                      [(set Float32Regs:$dst, (OpNode Float32Regs:$a))]>;
}

//===----------------------------------------------------------------------===//
// NVPTX Instructions.
//===----------------------------------------------------------------------===//

//-----------------------------------
// General Type Conversion
//-----------------------------------

let neverHasSideEffects = 1 in {
// Generate a cvt to the given type from all possible types.
// Each instance takes a CvtMode immediate that defines the conversion mode to
// use.  It can be CvtNONE to omit a conversion mode.
multiclass CVT_FROM_ALL<string FromName, RegisterClass RC> {
  def _s16 : NVPTXInst<(outs RC:$dst),
                       (ins Int16Regs:$src, CvtMode:$mode),
                       !strconcat("cvt${mode:base}${mode:ftz}${mode:sat}.",
                       FromName, ".s16\t$dst, $src;"),
                       []>;
  def _u16 : NVPTXInst<(outs RC:$dst),
                       (ins Int16Regs:$src, CvtMode:$mode),
                       !strconcat("cvt${mode:base}${mode:ftz}${mode:sat}.",
                       FromName, ".u16\t$dst, $src;"),
                       []>;
  def _f16 : NVPTXInst<(outs RC:$dst),
                       (ins Int16Regs:$src, CvtMode:$mode),
                       !strconcat("cvt${mode:base}${mode:ftz}${mode:sat}.",
                       FromName, ".f16\t$dst, $src;"),
                       []>;
  def _s32 : NVPTXInst<(outs RC:$dst),
                       (ins Int32Regs:$src, CvtMode:$mode),
                       !strconcat("cvt${mode:base}${mode:ftz}${mode:sat}.",
                       FromName, ".s32\t$dst, $src;"),
                       []>;
  def _u32 : NVPTXInst<(outs RC:$dst),
                       (ins Int32Regs:$src, CvtMode:$mode),
                       !strconcat("cvt${mode:base}${mode:ftz}${mode:sat}.",
                       FromName, ".u32\t$dst, $src;"),
                       []>;
  def _s64 : NVPTXInst<(outs RC:$dst),
                       (ins Int64Regs:$src, CvtMode:$mode),
                       !strconcat("cvt${mode:base}${mode:ftz}${mode:sat}.",
                       FromName, ".s64\t$dst, $src;"),
                       []>;
  def _u64 : NVPTXInst<(outs RC:$dst),
                       (ins Int64Regs:$src, CvtMode:$mode),
                       !strconcat("cvt${mode:base}${mode:ftz}${mode:sat}.",
                       FromName, ".u64\t$dst, $src;"),
                       []>;
  def _f32 : NVPTXInst<(outs RC:$dst),
                       (ins Float32Regs:$src, CvtMode:$mode),
                       !strconcat("cvt${mode:base}${mode:ftz}${mode:sat}.",
                       FromName, ".f32\t$dst, $src;"),
                       []>;
  def _f64 : NVPTXInst<(outs RC:$dst),
                       (ins Float64Regs:$src, CvtMode:$mode),
                       !strconcat("cvt${mode:base}${mode:ftz}${mode:sat}.",
                       FromName, ".f64\t$dst, $src;"),
                       []>;
}

// Generate a cvt to all possible types.
defm CVT_s16 : CVT_FROM_ALL<"s16", Int16Regs>;
defm CVT_u16 : CVT_FROM_ALL<"u16", Int16Regs>;
defm CVT_f16 : CVT_FROM_ALL<"f16", Int16Regs>;
defm CVT_s32 : CVT_FROM_ALL<"s32", Int32Regs>;
defm CVT_u32 : CVT_FROM_ALL<"u32", Int32Regs>;
defm CVT_s64 : CVT_FROM_ALL<"s64", Int64Regs>;
defm CVT_u64 : CVT_FROM_ALL<"u64", Int64Regs>;
defm CVT_f32 : CVT_FROM_ALL<"f32", Float32Regs>;
defm CVT_f64 : CVT_FROM_ALL<"f64", Float64Regs>;

// This set of cvt is different from the above. The type of the source
// and target are the same.
//
def CVT_INREG_s16_s8 : NVPTXInst<(outs Int16Regs:$dst), (ins Int16Regs:$src),
                        "cvt.s16.s8 \t$dst, $src;", []>;
def CVT_INREG_s32_s8 : NVPTXInst<(outs Int32Regs:$dst), (ins Int32Regs:$src),
                        "cvt.s32.s8 \t$dst, $src;", []>;
def CVT_INREG_s32_s16 : NVPTXInst<(outs Int32Regs:$dst), (ins Int32Regs:$src),
                        "cvt.s32.s16 \t$dst, $src;", []>;
def CVT_INREG_s64_s8 : NVPTXInst<(outs Int64Regs:$dst), (ins Int64Regs:$src),
                        "cvt.s64.s8 \t$dst, $src;", []>;
def CVT_INREG_s64_s16 : NVPTXInst<(outs Int64Regs:$dst), (ins Int64Regs:$src),
                        "cvt.s64.s16 \t$dst, $src;", []>;
def CVT_INREG_s64_s32 : NVPTXInst<(outs Int64Regs:$dst), (ins Int64Regs:$src),
                        "cvt.s64.s32 \t$dst, $src;", []>;
}

//-----------------------------------
// Integer Arithmetic
//-----------------------------------

multiclass ADD_SUB_i1<SDNode OpNode> {
   def _rr: NVPTXInst<(outs Int1Regs:$dst), (ins Int1Regs:$a, Int1Regs:$b),
          "xor.pred \t$dst, $a, $b;",
      [(set Int1Regs:$dst, (OpNode Int1Regs:$a, Int1Regs:$b))]>;
   def _ri: NVPTXInst<(outs Int1Regs:$dst), (ins Int1Regs:$a, i1imm:$b),
          "xor.pred \t$dst, $a, $b;",
      [(set Int1Regs:$dst, (OpNode Int1Regs:$a, (imm):$b))]>;
}

defm ADD_i1 : ADD_SUB_i1<add>;
defm SUB_i1 : ADD_SUB_i1<sub>;


defm ADD : I3<"add.s", add>;
defm SUB : I3<"sub.s", sub>;

defm ADDCC : ADD_SUB_INT_32<"add.cc", addc>;
defm SUBCC : ADD_SUB_INT_32<"sub.cc", subc>;

defm ADDCCC : ADD_SUB_INT_32<"addc.cc", adde>;
defm SUBCCC : ADD_SUB_INT_32<"subc.cc", sube>;

//mul.wide PTX instruction
def SInt32Const : PatLeaf<(imm), [{
  const APInt &v = N->getAPIntValue();
  if (v.isSignedIntN(32))
    return true;
  return false;
}]>;

def UInt32Const : PatLeaf<(imm), [{
  const APInt &v = N->getAPIntValue();
  if (v.isIntN(32))
    return true;
  return false;
}]>;

def SInt16Const : PatLeaf<(imm), [{
  const APInt &v = N->getAPIntValue();
  if (v.isSignedIntN(16))
    return true;
  return false;
}]>;

def UInt16Const : PatLeaf<(imm), [{
  const APInt &v = N->getAPIntValue();
  if (v.isIntN(16))
    return true;
  return false;
}]>;

def Int5Const : PatLeaf<(imm), [{
  const APInt &v = N->getAPIntValue();
  // Check if 0 <= v < 32
  // Only then the result from (x << v) will be i32
  if (v.sge(0) && v.slt(32))
    return true;
  return false;
}]>;

def Int4Const : PatLeaf<(imm), [{
  const APInt &v = N->getAPIntValue();
  // Check if 0 <= v < 16
  // Only then the result from (x << v) will be i16
  if (v.sge(0) && v.slt(16))
    return true;
  return false;
}]>;

def SHL2MUL32 : SDNodeXForm<imm, [{
  const APInt &v = N->getAPIntValue();
  APInt temp(32, 1);
  return CurDAG->getTargetConstant(temp.shl(v), MVT::i32);
}]>;

def SHL2MUL16 : SDNodeXForm<imm, [{
  const APInt &v = N->getAPIntValue();
  APInt temp(16, 1);
  return CurDAG->getTargetConstant(temp.shl(v), MVT::i16);
}]>;

def MULWIDES64 : NVPTXInst<(outs Int64Regs:$dst),
                           (ins Int32Regs:$a, Int32Regs:$b),
                           "mul.wide.s32 \t$dst, $a, $b;", []>;
def MULWIDES64Imm : NVPTXInst<(outs Int64Regs:$dst),
                            (ins Int32Regs:$a, i64imm:$b),
                           "mul.wide.s32 \t$dst, $a, $b;", []>;

def MULWIDEU64 : NVPTXInst<(outs Int64Regs:$dst),
                           (ins Int32Regs:$a, Int32Regs:$b),
                           "mul.wide.u32 \t$dst, $a, $b;", []>;
def MULWIDEU64Imm : NVPTXInst<(outs Int64Regs:$dst),
                            (ins Int32Regs:$a, i64imm:$b),
                           "mul.wide.u32 \t$dst, $a, $b;", []>;

def MULWIDES32 : NVPTXInst<(outs Int32Regs:$dst),
                            (ins Int16Regs:$a, Int16Regs:$b),
                           "mul.wide.s16 \t$dst, $a, $b;", []>;
def MULWIDES32Imm : NVPTXInst<(outs Int32Regs:$dst),
                            (ins Int16Regs:$a, i32imm:$b),
                           "mul.wide.s16 \t$dst, $a, $b;", []>;

def MULWIDEU32 : NVPTXInst<(outs Int32Regs:$dst),
                            (ins Int16Regs:$a, Int16Regs:$b),
                           "mul.wide.u16 \t$dst, $a, $b;", []>;
def MULWIDEU32Imm : NVPTXInst<(outs Int32Regs:$dst),
                            (ins Int16Regs:$a, i32imm:$b),
                           "mul.wide.u16 \t$dst, $a, $b;", []>;

def : Pat<(shl (sext Int32Regs:$a), (i32 Int5Const:$b)),
          (MULWIDES64Imm Int32Regs:$a, (SHL2MUL32 node:$b))>,
          Requires<[doMulWide]>;
def : Pat<(shl (zext Int32Regs:$a), (i32 Int5Const:$b)),
          (MULWIDEU64Imm Int32Regs:$a, (SHL2MUL32 node:$b))>,
          Requires<[doMulWide]>;

def : Pat<(shl (sext Int16Regs:$a), (i16 Int4Const:$b)),
          (MULWIDES32Imm Int16Regs:$a, (SHL2MUL16 node:$b))>,
          Requires<[doMulWide]>;
def : Pat<(shl (zext Int16Regs:$a), (i16 Int4Const:$b)),
          (MULWIDEU32Imm Int16Regs:$a, (SHL2MUL16 node:$b))>,
          Requires<[doMulWide]>;

def : Pat<(mul (sext Int32Regs:$a), (sext Int32Regs:$b)),
          (MULWIDES64 Int32Regs:$a, Int32Regs:$b)>,
          Requires<[doMulWide]>;
def : Pat<(mul (sext Int32Regs:$a), (i64 SInt32Const:$b)),
          (MULWIDES64Imm Int32Regs:$a, (i64 SInt32Const:$b))>,
          Requires<[doMulWide]>;

def : Pat<(mul (zext Int32Regs:$a), (zext Int32Regs:$b)),
          (MULWIDEU64 Int32Regs:$a, Int32Regs:$b)>, Requires<[doMulWide]>;
def : Pat<(mul (zext Int32Regs:$a), (i64 UInt32Const:$b)),
          (MULWIDEU64Imm Int32Regs:$a, (i64 UInt32Const:$b))>,
          Requires<[doMulWide]>;

def : Pat<(mul (sext Int16Regs:$a), (sext Int16Regs:$b)),
          (MULWIDES32 Int16Regs:$a, Int16Regs:$b)>, Requires<[doMulWide]>;
def : Pat<(mul (sext Int16Regs:$a), (i32 SInt16Const:$b)),
          (MULWIDES32Imm Int16Regs:$a, (i32 SInt16Const:$b))>,
          Requires<[doMulWide]>;

def : Pat<(mul (zext Int16Regs:$a), (zext Int16Regs:$b)),
          (MULWIDEU32 Int16Regs:$a, Int16Regs:$b)>, Requires<[doMulWide]>;
def : Pat<(mul (zext Int16Regs:$a), (i32 UInt16Const:$b)),
          (MULWIDEU32Imm Int16Regs:$a, (i32 UInt16Const:$b))>,
          Requires<[doMulWide]>;

defm MULT : I3<"mul.lo.s", mul>;

defm MULTHS : I3<"mul.hi.s", mulhs>;
defm MULTHU : I3<"mul.hi.u", mulhu>;

defm SDIV : I3<"div.s", sdiv>;
defm UDIV : I3<"div.u", udiv>;

defm SREM : I3<"rem.s", srem>;
// The ri version will not be selected as DAGCombiner::visitSREM will lower it.
defm UREM : I3<"rem.u", urem>;
// The ri version will not be selected as DAGCombiner::visitUREM will lower it.

def MAD16rrr : NVPTXInst<(outs Int16Regs:$dst),
                      (ins Int16Regs:$a, Int16Regs:$b, Int16Regs:$c),
                      "mad.lo.s16 \t$dst, $a, $b, $c;",
                      [(set Int16Regs:$dst, (add
                        (mul Int16Regs:$a, Int16Regs:$b), Int16Regs:$c))]>;
def MAD16rri : NVPTXInst<(outs Int16Regs:$dst),
                      (ins Int16Regs:$a, Int16Regs:$b, i16imm:$c),
                      "mad.lo.s16 \t$dst, $a, $b, $c;",
                      [(set Int16Regs:$dst, (add
                        (mul Int16Regs:$a, Int16Regs:$b), imm:$c))]>;
def MAD16rir : NVPTXInst<(outs Int16Regs:$dst),
                      (ins Int16Regs:$a, i16imm:$b, Int16Regs:$c),
                      "mad.lo.s16 \t$dst, $a, $b, $c;",
                      [(set Int16Regs:$dst, (add
                        (mul Int16Regs:$a, imm:$b), Int16Regs:$c))]>;
def MAD16rii : NVPTXInst<(outs Int16Regs:$dst),
    (ins Int16Regs:$a, i16imm:$b, i16imm:$c),
                      "mad.lo.s16 \t$dst, $a, $b, $c;",
                      [(set Int16Regs:$dst, (add (mul Int16Regs:$a, imm:$b),
                        imm:$c))]>;

def MAD32rrr : NVPTXInst<(outs Int32Regs:$dst),
                      (ins Int32Regs:$a, Int32Regs:$b, Int32Regs:$c),
                      "mad.lo.s32 \t$dst, $a, $b, $c;",
                      [(set Int32Regs:$dst, (add
                        (mul Int32Regs:$a, Int32Regs:$b), Int32Regs:$c))]>;
def MAD32rri : NVPTXInst<(outs Int32Regs:$dst),
                      (ins Int32Regs:$a, Int32Regs:$b, i32imm:$c),
                      "mad.lo.s32 \t$dst, $a, $b, $c;",
                      [(set Int32Regs:$dst, (add
                        (mul Int32Regs:$a, Int32Regs:$b), imm:$c))]>;
def MAD32rir : NVPTXInst<(outs Int32Regs:$dst),
                      (ins Int32Regs:$a, i32imm:$b, Int32Regs:$c),
                      "mad.lo.s32 \t$dst, $a, $b, $c;",
                      [(set Int32Regs:$dst, (add
                        (mul Int32Regs:$a, imm:$b), Int32Regs:$c))]>;
def MAD32rii : NVPTXInst<(outs Int32Regs:$dst),
                      (ins Int32Regs:$a, i32imm:$b, i32imm:$c),
                      "mad.lo.s32 \t$dst, $a, $b, $c;",
                      [(set Int32Regs:$dst, (add
                        (mul Int32Regs:$a, imm:$b), imm:$c))]>;

def MAD64rrr : NVPTXInst<(outs Int64Regs:$dst),
                      (ins Int64Regs:$a, Int64Regs:$b, Int64Regs:$c),
                      "mad.lo.s64 \t$dst, $a, $b, $c;",
                      [(set Int64Regs:$dst, (add
                        (mul Int64Regs:$a, Int64Regs:$b), Int64Regs:$c))]>;
def MAD64rri : NVPTXInst<(outs Int64Regs:$dst),
                      (ins Int64Regs:$a, Int64Regs:$b, i64imm:$c),
                      "mad.lo.s64 \t$dst, $a, $b, $c;",
                      [(set Int64Regs:$dst, (add
                        (mul Int64Regs:$a, Int64Regs:$b), imm:$c))]>;
def MAD64rir : NVPTXInst<(outs Int64Regs:$dst),
                      (ins Int64Regs:$a, i64imm:$b, Int64Regs:$c),
                      "mad.lo.s64 \t$dst, $a, $b, $c;",
                      [(set Int64Regs:$dst, (add
                        (mul Int64Regs:$a, imm:$b), Int64Regs:$c))]>;
def MAD64rii : NVPTXInst<(outs Int64Regs:$dst),
                      (ins Int64Regs:$a, i64imm:$b, i64imm:$c),
                      "mad.lo.s64 \t$dst, $a, $b, $c;",
                      [(set Int64Regs:$dst, (add
                        (mul Int64Regs:$a, imm:$b), imm:$c))]>;


def INEG16 : NVPTXInst<(outs Int16Regs:$dst), (ins Int16Regs:$src),
                     "neg.s16 \t$dst, $src;",
         [(set Int16Regs:$dst, (ineg Int16Regs:$src))]>;
def INEG32 : NVPTXInst<(outs Int32Regs:$dst), (ins Int32Regs:$src),
                     "neg.s32 \t$dst, $src;",
         [(set Int32Regs:$dst, (ineg Int32Regs:$src))]>;
def INEG64 : NVPTXInst<(outs Int64Regs:$dst), (ins Int64Regs:$src),
                     "neg.s64 \t$dst, $src;",
         [(set Int64Regs:$dst, (ineg Int64Regs:$src))]>;

//-----------------------------------
// Floating Point Arithmetic
//-----------------------------------

// Constant 1.0f
def FloatConst1 : PatLeaf<(fpimm), [{
    if (&(N->getValueAPF().getSemantics()) != &llvm::APFloat::IEEEsingle)
      return false;
    float f = (float)N->getValueAPF().convertToFloat();
    return (f==1.0f);
}]>;
// Constand (double)1.0
def DoubleConst1 : PatLeaf<(fpimm), [{
    if (&(N->getValueAPF().getSemantics()) != &llvm::APFloat::IEEEdouble)
      return false;
    double d = (double)N->getValueAPF().convertToDouble();
    return (d==1.0);
}]>;

defm FADD : F3<"add", fadd>;
defm FSUB : F3<"sub", fsub>;
defm FMUL : F3<"mul", fmul>;

defm FADD_rn : F3_rn<"add", fadd>;
defm FSUB_rn : F3_rn<"sub", fsub>;
defm FMUL_rn : F3_rn<"mul", fmul>;

defm FABS : F2<"abs", fabs>;
defm FNEG : F2<"neg", fneg>;
defm FSQRT : F2<"sqrt.rn", fsqrt>;

//
// F64 division
//
def FDIV641r : NVPTXInst<(outs Float64Regs:$dst),
                      (ins f64imm:$a, Float64Regs:$b),
                      "rcp.rn.f64 \t$dst, $b;",
                      [(set Float64Regs:$dst,
                        (fdiv DoubleConst1:$a, Float64Regs:$b))]>;
def FDIV64rr : NVPTXInst<(outs Float64Regs:$dst),
                      (ins Float64Regs:$a, Float64Regs:$b),
                      "div.rn.f64 \t$dst, $a, $b;",
                      [(set Float64Regs:$dst,
                        (fdiv Float64Regs:$a, Float64Regs:$b))]>;
def FDIV64ri : NVPTXInst<(outs Float64Regs:$dst),
                      (ins Float64Regs:$a, f64imm:$b),
                      "div.rn.f64 \t$dst, $a, $b;",
                      [(set Float64Regs:$dst,
                        (fdiv Float64Regs:$a, fpimm:$b))]>;

//
// F32 Approximate reciprocal
//
def FDIV321r_ftz : NVPTXInst<(outs Float32Regs:$dst),
                      (ins f32imm:$a, Float32Regs:$b),
                      "rcp.approx.ftz.f32 \t$dst, $b;",
                      [(set Float32Regs:$dst,
                        (fdiv FloatConst1:$a, Float32Regs:$b))]>,
                      Requires<[do_DIVF32_APPROX, doF32FTZ]>;
def FDIV321r : NVPTXInst<(outs Float32Regs:$dst),
                        (ins f32imm:$a, Float32Regs:$b),
                       "rcp.approx.f32 \t$dst, $b;",
                      [(set Float32Regs:$dst,
                        (fdiv FloatConst1:$a, Float32Regs:$b))]>,
                      Requires<[do_DIVF32_APPROX]>;
//
// F32 Approximate division
//
def FDIV32approxrr_ftz : NVPTXInst<(outs Float32Regs:$dst),
                      (ins Float32Regs:$a, Float32Regs:$b),
                      "div.approx.ftz.f32 \t$dst, $a, $b;",
                      [(set Float32Regs:$dst,
                        (fdiv Float32Regs:$a, Float32Regs:$b))]>,
                      Requires<[do_DIVF32_APPROX, doF32FTZ]>;
def FDIV32approxri_ftz : NVPTXInst<(outs Float32Regs:$dst),
                      (ins Float32Regs:$a, f32imm:$b),
                      "div.approx.ftz.f32 \t$dst, $a, $b;",
                      [(set Float32Regs:$dst,
                        (fdiv Float32Regs:$a, fpimm:$b))]>,
                      Requires<[do_DIVF32_APPROX, doF32FTZ]>;
def FDIV32approxrr     : NVPTXInst<(outs Float32Regs:$dst),
                      (ins Float32Regs:$a, Float32Regs:$b),
                      "div.approx.f32 \t$dst, $a, $b;",
                      [(set Float32Regs:$dst,
                        (fdiv Float32Regs:$a, Float32Regs:$b))]>,
                      Requires<[do_DIVF32_APPROX]>;
def FDIV32approxri : NVPTXInst<(outs Float32Regs:$dst),
                      (ins Float32Regs:$a, f32imm:$b),
                      "div.approx.f32 \t$dst, $a, $b;",
                      [(set Float32Regs:$dst,
                        (fdiv Float32Regs:$a, fpimm:$b))]>,
                      Requires<[do_DIVF32_APPROX]>;
//
// F32 Semi-accurate reciprocal
//
// rcp.approx gives the same result as div.full(1.0f, a) and is faster.
//
def FDIV321r_approx_ftz : NVPTXInst<(outs Float32Regs:$dst),
                      (ins f32imm:$a, Float32Regs:$b),
                      "rcp.approx.ftz.f32 \t$dst, $b;",
                      [(set Float32Regs:$dst,
                        (fdiv FloatConst1:$a, Float32Regs:$b))]>,
                      Requires<[do_DIVF32_FULL, doF32FTZ]>;
def FDIV321r_approx : NVPTXInst<(outs Float32Regs:$dst),
                      (ins f32imm:$a, Float32Regs:$b),
                      "rcp.approx.f32 \t$dst, $b;",
                      [(set Float32Regs:$dst,
                        (fdiv FloatConst1:$a, Float32Regs:$b))]>,
                      Requires<[do_DIVF32_FULL]>;
//
// F32 Semi-accurate division
//
def FDIV32rr_ftz : NVPTXInst<(outs Float32Regs:$dst),
                      (ins Float32Regs:$a, Float32Regs:$b),
                      "div.full.ftz.f32 \t$dst, $a, $b;",
                      [(set Float32Regs:$dst,
                        (fdiv Float32Regs:$a, Float32Regs:$b))]>,
                      Requires<[do_DIVF32_FULL, doF32FTZ]>;
def FDIV32ri_ftz : NVPTXInst<(outs Float32Regs:$dst),
                      (ins Float32Regs:$a, f32imm:$b),
                      "div.full.ftz.f32 \t$dst, $a, $b;",
                      [(set Float32Regs:$dst,
                        (fdiv Float32Regs:$a, fpimm:$b))]>,
                      Requires<[do_DIVF32_FULL, doF32FTZ]>;
def FDIV32rr : NVPTXInst<(outs Float32Regs:$dst),
                      (ins Float32Regs:$a, Float32Regs:$b),
                      "div.full.f32 \t$dst, $a, $b;",
                      [(set Float32Regs:$dst,
                        (fdiv Float32Regs:$a, Float32Regs:$b))]>,
                      Requires<[do_DIVF32_FULL]>;
def FDIV32ri : NVPTXInst<(outs Float32Regs:$dst),
                      (ins Float32Regs:$a, f32imm:$b),
                      "div.full.f32 \t$dst, $a, $b;",
                      [(set Float32Regs:$dst,
                        (fdiv Float32Regs:$a, fpimm:$b))]>,
                      Requires<[do_DIVF32_FULL]>;
//
// F32 Accurate reciprocal
//
def FDIV321r_prec_ftz : NVPTXInst<(outs Float32Regs:$dst),
                        (ins f32imm:$a, Float32Regs:$b),
                       "rcp.rn.ftz.f32 \t$dst, $b;",
                      [(set Float32Regs:$dst,
                        (fdiv FloatConst1:$a, Float32Regs:$b))]>,
                      Requires<[reqPTX20, doF32FTZ]>;
def FDIV321r_prec : NVPTXInst<(outs Float32Regs:$dst),
                      (ins f32imm:$a, Float32Regs:$b),
                       "rcp.rn.f32 \t$dst, $b;",
                      [(set Float32Regs:$dst,
                        (fdiv FloatConst1:$a, Float32Regs:$b))]>,
                      Requires<[reqPTX20]>;
//
// F32 Accurate division
//
def FDIV32rr_prec_ftz : NVPTXInst<(outs Float32Regs:$dst),
                      (ins Float32Regs:$a, Float32Regs:$b),
                      "div.rn.ftz.f32 \t$dst, $a, $b;",
                      [(set Float32Regs:$dst,
                        (fdiv Float32Regs:$a, Float32Regs:$b))]>,
                      Requires<[doF32FTZ, reqPTX20]>;
def FDIV32ri_prec_ftz : NVPTXInst<(outs Float32Regs:$dst),
                      (ins Float32Regs:$a, f32imm:$b),
                      "div.rn.ftz.f32 \t$dst, $a, $b;",
                      [(set Float32Regs:$dst,
                        (fdiv Float32Regs:$a, fpimm:$b))]>,
                      Requires<[doF32FTZ, reqPTX20]>;
def FDIV32rr_prec : NVPTXInst<(outs Float32Regs:$dst),
                      (ins Float32Regs:$a, Float32Regs:$b),
                      "div.rn.f32 \t$dst, $a, $b;",
                      [(set Float32Regs:$dst,
                        (fdiv Float32Regs:$a, Float32Regs:$b))]>,
                      Requires<[reqPTX20]>;
def FDIV32ri_prec : NVPTXInst<(outs Float32Regs:$dst),
                      (ins Float32Regs:$a, f32imm:$b),
                      "div.rn.f32 \t$dst, $a, $b;",
                      [(set Float32Regs:$dst,
                        (fdiv Float32Regs:$a, fpimm:$b))]>,
                      Requires<[reqPTX20]>;

//
// F32 rsqrt
//

def RSQRTF32approx1r : NVPTXInst<(outs Float32Regs:$dst), (ins Float32Regs:$b),
                       "rsqrt.approx.f32 \t$dst, $b;", []>;

def: Pat<(fdiv FloatConst1, (int_nvvm_sqrt_f Float32Regs:$b)),
         (RSQRTF32approx1r Float32Regs:$b)>,
         Requires<[do_DIVF32_FULL, do_SQRTF32_APPROX, doNoF32FTZ]>;

multiclass FPCONTRACT32<string OpcStr, Predicate Pred> {
   def rrr : NVPTXInst<(outs Float32Regs:$dst),
                      (ins Float32Regs:$a, Float32Regs:$b, Float32Regs:$c),
                      !strconcat(OpcStr, " \t$dst, $a, $b, $c;"),
                      [(set Float32Regs:$dst, (fadd
                        (fmul Float32Regs:$a, Float32Regs:$b),
                        Float32Regs:$c))]>, Requires<[Pred]>;
   // This is to WAR a weird bug in Tablegen that does not automatically
   // generate the following permutated rule rrr2 from the above rrr.
   // So we explicitly add it here. This happens to FMA32 only.
   // See the comments at FMAD32 and FMA32 for more information.
   def rrr2 : NVPTXInst<(outs Float32Regs:$dst),
                        (ins Float32Regs:$a, Float32Regs:$b, Float32Regs:$c),
                      !strconcat(OpcStr, " \t$dst, $a, $b, $c;"),
                      [(set Float32Regs:$dst, (fadd Float32Regs:$c,
                        (fmul Float32Regs:$a, Float32Regs:$b)))]>,
                      Requires<[Pred]>;
   def rri : NVPTXInst<(outs Float32Regs:$dst),
                      (ins Float32Regs:$a, Float32Regs:$b, f32imm:$c),
                      !strconcat(OpcStr, " \t$dst, $a, $b, $c;"),
                      [(set Float32Regs:$dst, (fadd
                        (fmul Float32Regs:$a, Float32Regs:$b), fpimm:$c))]>,
                      Requires<[Pred]>;
   def rir : NVPTXInst<(outs Float32Regs:$dst),
                      (ins Float32Regs:$a, f32imm:$b, Float32Regs:$c),
                      !strconcat(OpcStr, " \t$dst, $a, $b, $c;"),
                      [(set Float32Regs:$dst, (fadd
                        (fmul Float32Regs:$a, fpimm:$b), Float32Regs:$c))]>,
                      Requires<[Pred]>;
   def rii : NVPTXInst<(outs Float32Regs:$dst),
                      (ins Float32Regs:$a, f32imm:$b, f32imm:$c),
                      !strconcat(OpcStr, " \t$dst, $a, $b, $c;"),
                      [(set Float32Regs:$dst, (fadd
                        (fmul Float32Regs:$a, fpimm:$b), fpimm:$c))]>,
                      Requires<[Pred]>;
}

multiclass FPCONTRACT64<string OpcStr, Predicate Pred> {
   def rrr : NVPTXInst<(outs Float64Regs:$dst),
                      (ins Float64Regs:$a, Float64Regs:$b, Float64Regs:$c),
                      !strconcat(OpcStr, " \t$dst, $a, $b, $c;"),
                      [(set Float64Regs:$dst, (fadd
                        (fmul Float64Regs:$a, Float64Regs:$b),
                        Float64Regs:$c))]>, Requires<[Pred]>;
   def rri : NVPTXInst<(outs Float64Regs:$dst),
                      (ins Float64Regs:$a, Float64Regs:$b, f64imm:$c),
                      !strconcat(OpcStr, " \t$dst, $a, $b, $c;"),
                      [(set Float64Regs:$dst, (fadd (fmul Float64Regs:$a,
                        Float64Regs:$b), fpimm:$c))]>, Requires<[Pred]>;
   def rir : NVPTXInst<(outs Float64Regs:$dst),
                      (ins Float64Regs:$a, f64imm:$b, Float64Regs:$c),
                      !strconcat(OpcStr, " \t$dst, $a, $b, $c;"),
                      [(set Float64Regs:$dst, (fadd
                        (fmul Float64Regs:$a, fpimm:$b), Float64Regs:$c))]>,
                      Requires<[Pred]>;
   def rii : NVPTXInst<(outs Float64Regs:$dst),
                      (ins Float64Regs:$a, f64imm:$b, f64imm:$c),
                      !strconcat(OpcStr, " \t$dst, $a, $b, $c;"),
                      [(set Float64Regs:$dst, (fadd
                        (fmul Float64Regs:$a, fpimm:$b), fpimm:$c))]>,
                      Requires<[Pred]>;
}

// Due to a unknown reason (most likely a bug in tablegen), tablegen does not
// automatically generate the rrr2 rule from
// the rrr rule (see FPCONTRACT32) for FMA32, though it does for FMAD32.
// If we reverse the order of the following two lines, then rrr2 rule will be
// generated for FMA32, but not for rrr.
// Therefore, we manually write the rrr2 rule in FPCONTRACT32.
defm FMA32_ftz  : FPCONTRACT32<"fma.rn.ftz.f32", doFMAF32_ftz>;
defm FMA32  : FPCONTRACT32<"fma.rn.f32", doFMAF32>;
defm FMA64  : FPCONTRACT64<"fma.rn.f64", doFMAF64>;

// b*c-a => fmad(b, c, -a)
multiclass FPCONTRACT32_SUB_PAT_MAD<NVPTXInst Inst, Predicate Pred> {
  def : Pat<(fsub (fmul Float32Regs:$b, Float32Regs:$c), Float32Regs:$a),
          (Inst Float32Regs:$b, Float32Regs:$c, (FNEGf32 Float32Regs:$a))>,
          Requires<[Pred]>;
}

// a-b*c => fmad(-b,c, a)
// - legal because a-b*c <=> a+(-b*c) <=> a+(-b)*c
// b*c-a => fmad(b, c, -a)
// - legal because b*c-a <=> b*c+(-a)
multiclass FPCONTRACT32_SUB_PAT<NVPTXInst Inst, Predicate Pred> {
  def : Pat<(fsub Float32Regs:$a, (fmul Float32Regs:$b, Float32Regs:$c)),
          (Inst (FNEGf32 Float32Regs:$b), Float32Regs:$c, Float32Regs:$a)>,
          Requires<[Pred]>;
  def : Pat<(fsub (fmul Float32Regs:$b, Float32Regs:$c), Float32Regs:$a),
          (Inst Float32Regs:$b, Float32Regs:$c, (FNEGf32 Float32Regs:$a))>,
          Requires<[Pred]>;
}

// a-b*c => fmad(-b,c, a)
// b*c-a => fmad(b, c, -a)
multiclass FPCONTRACT64_SUB_PAT<NVPTXInst Inst, Predicate Pred> {
  def : Pat<(fsub Float64Regs:$a, (fmul Float64Regs:$b, Float64Regs:$c)),
          (Inst (FNEGf64 Float64Regs:$b), Float64Regs:$c, Float64Regs:$a)>,
          Requires<[Pred]>;

  def : Pat<(fsub (fmul Float64Regs:$b, Float64Regs:$c), Float64Regs:$a),
          (Inst Float64Regs:$b, Float64Regs:$c, (FNEGf64 Float64Regs:$a))>,
          Requires<[Pred]>;
}

defm FMAF32ext_ftz  : FPCONTRACT32_SUB_PAT<FMA32_ftzrrr, doFMAF32AGG_ftz>;
defm FMAF32ext  : FPCONTRACT32_SUB_PAT<FMA32rrr, doFMAF32AGG>;
defm FMAF64ext  : FPCONTRACT64_SUB_PAT<FMA64rrr, doFMAF64AGG>;

def SINF:  NVPTXInst<(outs Float32Regs:$dst), (ins Float32Regs:$src),
                      "sin.approx.f32 \t$dst, $src;",
                      [(set Float32Regs:$dst, (fsin Float32Regs:$src))]>;
def COSF:  NVPTXInst<(outs Float32Regs:$dst), (ins Float32Regs:$src),
                      "cos.approx.f32 \t$dst, $src;",
                      [(set Float32Regs:$dst, (fcos Float32Regs:$src))]>;

// Lower (frem x, y) into (sub x, (mul (floor (div x, y)) y))
// e.g. "poor man's fmod()"

// frem - f32 FTZ
def : Pat<(frem Float32Regs:$x, Float32Regs:$y),
          (FSUBf32rr_ftz Float32Regs:$x, (FMULf32rr_ftz (CVT_f32_f32
            (FDIV32rr_prec_ftz Float32Regs:$x, Float32Regs:$y), CvtRMI_FTZ),
             Float32Regs:$y))>,
          Requires<[doF32FTZ]>;
def : Pat<(frem Float32Regs:$x, fpimm:$y),
          (FSUBf32rr_ftz Float32Regs:$x, (FMULf32ri_ftz (CVT_f32_f32
            (FDIV32ri_prec_ftz Float32Regs:$x, fpimm:$y), CvtRMI_FTZ),
             fpimm:$y))>,
          Requires<[doF32FTZ]>;

// frem - f32
def : Pat<(frem Float32Regs:$x, Float32Regs:$y),
          (FSUBf32rr Float32Regs:$x, (FMULf32rr (CVT_f32_f32
            (FDIV32rr_prec Float32Regs:$x, Float32Regs:$y), CvtRMI),
             Float32Regs:$y))>;
def : Pat<(frem Float32Regs:$x, fpimm:$y),
          (FSUBf32rr Float32Regs:$x, (FMULf32ri (CVT_f32_f32
            (FDIV32ri_prec Float32Regs:$x, fpimm:$y), CvtRMI),
             fpimm:$y))>;

// frem - f64
def : Pat<(frem Float64Regs:$x, Float64Regs:$y),
          (FSUBf64rr Float64Regs:$x, (FMULf64rr (CVT_f64_f64
            (FDIV64rr Float64Regs:$x, Float64Regs:$y), CvtRMI),
             Float64Regs:$y))>;
def : Pat<(frem Float64Regs:$x, fpimm:$y),
          (FSUBf64rr Float64Regs:$x, (FMULf64ri (CVT_f64_f64
            (FDIV64ri Float64Regs:$x, fpimm:$y), CvtRMI),
             fpimm:$y))>;

//-----------------------------------
// Logical Arithmetic
//-----------------------------------

multiclass LOG_FORMAT<string OpcStr, SDNode OpNode> {
  def b1rr:  NVPTXInst<(outs Int1Regs:$dst), (ins Int1Regs:$a, Int1Regs:$b),
                      !strconcat(OpcStr, ".pred  \t$dst, $a, $b;"),
                      [(set Int1Regs:$dst, (OpNode Int1Regs:$a, Int1Regs:$b))]>;
  def b1ri:  NVPTXInst<(outs Int1Regs:$dst), (ins Int1Regs:$a, i1imm:$b),
                      !strconcat(OpcStr, ".pred  \t$dst, $a, $b;"),
                      [(set Int1Regs:$dst, (OpNode Int1Regs:$a, imm:$b))]>;
  def b16rr:  NVPTXInst<(outs Int16Regs:$dst), (ins Int16Regs:$a, Int16Regs:$b),
                      !strconcat(OpcStr, ".b16  \t$dst, $a, $b;"),
                      [(set Int16Regs:$dst, (OpNode Int16Regs:$a,
                        Int16Regs:$b))]>;
  def b16ri:  NVPTXInst<(outs Int16Regs:$dst), (ins Int16Regs:$a, i16imm:$b),
                      !strconcat(OpcStr, ".b16  \t$dst, $a, $b;"),
                      [(set Int16Regs:$dst, (OpNode Int16Regs:$a, imm:$b))]>;
  def b32rr:  NVPTXInst<(outs Int32Regs:$dst), (ins Int32Regs:$a, Int32Regs:$b),
                      !strconcat(OpcStr, ".b32  \t$dst, $a, $b;"),
                      [(set Int32Regs:$dst, (OpNode Int32Regs:$a,
                        Int32Regs:$b))]>;
  def b32ri:  NVPTXInst<(outs Int32Regs:$dst), (ins Int32Regs:$a, i32imm:$b),
                      !strconcat(OpcStr, ".b32  \t$dst, $a, $b;"),
                      [(set Int32Regs:$dst, (OpNode Int32Regs:$a, imm:$b))]>;
  def b64rr:  NVPTXInst<(outs Int64Regs:$dst), (ins Int64Regs:$a, Int64Regs:$b),
                      !strconcat(OpcStr, ".b64  \t$dst, $a, $b;"),
                      [(set Int64Regs:$dst, (OpNode Int64Regs:$a,
                        Int64Regs:$b))]>;
  def b64ri:  NVPTXInst<(outs Int64Regs:$dst), (ins Int64Regs:$a, i64imm:$b),
                      !strconcat(OpcStr, ".b64  \t$dst, $a, $b;"),
                      [(set Int64Regs:$dst, (OpNode Int64Regs:$a, imm:$b))]>;
}

defm OR  : LOG_FORMAT<"or", or>;
defm AND : LOG_FORMAT<"and", and>;
defm XOR : LOG_FORMAT<"xor", xor>;

def NOT1:  NVPTXInst<(outs Int1Regs:$dst), (ins Int1Regs:$src),
                      "not.pred \t$dst, $src;",
                      [(set Int1Regs:$dst, (not Int1Regs:$src))]>;
def NOT16:  NVPTXInst<(outs Int16Regs:$dst), (ins Int16Regs:$src),
                      "not.b16 \t$dst, $src;",
                      [(set Int16Regs:$dst, (not Int16Regs:$src))]>;
def NOT32:  NVPTXInst<(outs Int32Regs:$dst), (ins Int32Regs:$src),
                      "not.b32 \t$dst, $src;",
                      [(set Int32Regs:$dst, (not Int32Regs:$src))]>;
def NOT64:  NVPTXInst<(outs Int64Regs:$dst), (ins Int64Regs:$src),
                      "not.b64 \t$dst, $src;",
                      [(set Int64Regs:$dst, (not Int64Regs:$src))]>;

// For shifts, the second src operand must be 32-bit value
multiclass LSHIFT_FORMAT<string OpcStr, SDNode OpNode> {
   def i64rr : NVPTXInst<(outs Int64Regs:$dst), (ins Int64Regs:$a,
                      Int32Regs:$b),
                      !strconcat(OpcStr, "64 \t$dst, $a, $b;"),
                      [(set Int64Regs:$dst, (OpNode Int64Regs:$a,
                        Int32Regs:$b))]>;
   def i64ri : NVPTXInst<(outs Int64Regs:$dst), (ins Int64Regs:$a, i32imm:$b),
                      !strconcat(OpcStr, "64 \t$dst, $a, $b;"),
                      [(set Int64Regs:$dst, (OpNode Int64Regs:$a,
                        (i32 imm:$b)))]>;
   def i32rr : NVPTXInst<(outs Int32Regs:$dst), (ins Int32Regs:$a,
                      Int32Regs:$b),
                      !strconcat(OpcStr, "32 \t$dst, $a, $b;"),
                      [(set Int32Regs:$dst, (OpNode Int32Regs:$a,
                        Int32Regs:$b))]>;
   def i32ri : NVPTXInst<(outs Int32Regs:$dst), (ins Int32Regs:$a, i32imm:$b),
                      !strconcat(OpcStr, "32 \t$dst, $a, $b;"),
                      [(set Int32Regs:$dst, (OpNode Int32Regs:$a,
                        (i32 imm:$b)))]>;
   def i32ii : NVPTXInst<(outs Int32Regs:$dst), (ins i32imm:$a, i32imm:$b),
                      !strconcat(OpcStr, "32 \t$dst, $a, $b;"),
                      [(set Int32Regs:$dst, (OpNode (i32 imm:$a),
                        (i32 imm:$b)))]>;
   def i16rr : NVPTXInst<(outs Int16Regs:$dst), (ins Int16Regs:$a,
                      Int32Regs:$b),
                      !strconcat(OpcStr, "16 \t$dst, $a, $b;"),
                      [(set Int16Regs:$dst, (OpNode Int16Regs:$a,
                        Int32Regs:$b))]>;
   def i16ri : NVPTXInst<(outs Int16Regs:$dst), (ins Int16Regs:$a, i32imm:$b),
                      !strconcat(OpcStr, "16 \t$dst, $a, $b;"),
                      [(set Int16Regs:$dst, (OpNode Int16Regs:$a,
                        (i32 imm:$b)))]>;
}

defm SHL : LSHIFT_FORMAT<"shl.b", shl>;

// For shifts, the second src operand must be 32-bit value
// Need to add cvt for the 8-bits.
multiclass RSHIFT_FORMAT<string OpcStr, SDNode OpNode> {
   def i64rr : NVPTXInst<(outs Int64Regs:$dst), (ins Int64Regs:$a,
                      Int32Regs:$b),
                      !strconcat(OpcStr, "64 \t$dst, $a, $b;"),
                      [(set Int64Regs:$dst, (OpNode Int64Regs:$a,
                        Int32Regs:$b))]>;
   def i64ri : NVPTXInst<(outs Int64Regs:$dst), (ins Int64Regs:$a, i32imm:$b),
                      !strconcat(OpcStr, "64 \t$dst, $a, $b;"),
                      [(set Int64Regs:$dst, (OpNode Int64Regs:$a,
                        (i32 imm:$b)))]>;
   def i32rr : NVPTXInst<(outs Int32Regs:$dst), (ins Int32Regs:$a,
                      Int32Regs:$b),
                      !strconcat(OpcStr, "32 \t$dst, $a, $b;"),
                      [(set Int32Regs:$dst, (OpNode Int32Regs:$a,
                        Int32Regs:$b))]>;
   def i32ri : NVPTXInst<(outs Int32Regs:$dst), (ins Int32Regs:$a, i32imm:$b),
                      !strconcat(OpcStr, "32 \t$dst, $a, $b;"),
                      [(set Int32Regs:$dst, (OpNode Int32Regs:$a,
                        (i32 imm:$b)))]>;
   def i32ii : NVPTXInst<(outs Int32Regs:$dst), (ins i32imm:$a, i32imm:$b),
                      !strconcat(OpcStr, "32 \t$dst, $a, $b;"),
                      [(set Int32Regs:$dst, (OpNode (i32 imm:$a),
                        (i32 imm:$b)))]>;
   def i16rr : NVPTXInst<(outs Int16Regs:$dst), (ins Int16Regs:$a,
                      Int32Regs:$b),
                      !strconcat(OpcStr, "16 \t$dst, $a, $b;"),
                      [(set Int16Regs:$dst, (OpNode Int16Regs:$a,
                        Int32Regs:$b))]>;
   def i16ri : NVPTXInst<(outs Int16Regs:$dst), (ins Int16Regs:$a, i32imm:$b),
                      !strconcat(OpcStr, "16 \t$dst, $a, $b;"),
                      [(set Int16Regs:$dst, (OpNode Int16Regs:$a,
                        (i32 imm:$b)))]>;
}

defm SRA : RSHIFT_FORMAT<"shr.s", sra>;
defm SRL : RSHIFT_FORMAT<"shr.u", srl>;

// 32bit
def ROT32imm_sw : NVPTXInst<(outs Int32Regs:$dst),
  (ins Int32Regs:$src, i32imm:$amt1, i32imm:$amt2),
    !strconcat("{{\n\t",
    !strconcat(".reg .b32 %lhs;\n\t",
    !strconcat(".reg .b32 %rhs;\n\t",
    !strconcat("shl.b32 \t%lhs, $src, $amt1;\n\t",
    !strconcat("shr.b32 \t%rhs, $src, $amt2;\n\t",
    !strconcat("add.u32 \t$dst, %lhs, %rhs;\n\t",
    !strconcat("}}", ""))))))),
    []>;

def SUB_FRM_32 : SDNodeXForm<imm, [{
    return CurDAG->getTargetConstant(32-N->getZExtValue(), MVT::i32);
}]>;

def : Pat<(rotl Int32Regs:$src, (i32 imm:$amt)),
          (ROT32imm_sw Int32Regs:$src, imm:$amt, (SUB_FRM_32 node:$amt))>;
def : Pat<(rotr Int32Regs:$src, (i32 imm:$amt)),
          (ROT32imm_sw Int32Regs:$src, (SUB_FRM_32 node:$amt), imm:$amt)>;

def ROTL32reg_sw : NVPTXInst<(outs Int32Regs:$dst), (ins Int32Regs:$src,
    Int32Regs:$amt),
    !strconcat("{{\n\t",
    !strconcat(".reg .b32 %lhs;\n\t",
    !strconcat(".reg .b32 %rhs;\n\t",
    !strconcat(".reg .b32 %amt2;\n\t",
    !strconcat("shl.b32 \t%lhs, $src, $amt;\n\t",
    !strconcat("sub.s32 \t%amt2, 32, $amt;\n\t",
    !strconcat("shr.b32 \t%rhs, $src, %amt2;\n\t",
    !strconcat("add.u32 \t$dst, %lhs, %rhs;\n\t",
    !strconcat("}}", ""))))))))),
    [(set Int32Regs:$dst, (rotl Int32Regs:$src, Int32Regs:$amt))]>;

def ROTR32reg_sw : NVPTXInst<(outs Int32Regs:$dst), (ins Int32Regs:$src,
    Int32Regs:$amt),
    !strconcat("{{\n\t",
    !strconcat(".reg .b32 %lhs;\n\t",
    !strconcat(".reg .b32 %rhs;\n\t",
    !strconcat(".reg .b32 %amt2;\n\t",
    !strconcat("shr.b32 \t%lhs, $src, $amt;\n\t",
    !strconcat("sub.s32 \t%amt2, 32, $amt;\n\t",
    !strconcat("shl.b32 \t%rhs, $src, %amt2;\n\t",
    !strconcat("add.u32 \t$dst, %lhs, %rhs;\n\t",
    !strconcat("}}", ""))))))))),
    [(set Int32Regs:$dst, (rotr Int32Regs:$src, Int32Regs:$amt))]>;

// 64bit
def ROT64imm_sw : NVPTXInst<(outs Int64Regs:$dst), (ins Int64Regs:$src,
    i32imm:$amt1, i32imm:$amt2),
    !strconcat("{{\n\t",
    !strconcat(".reg .b64 %lhs;\n\t",
    !strconcat(".reg .b64 %rhs;\n\t",
    !strconcat("shl.b64 \t%lhs, $src, $amt1;\n\t",
    !strconcat("shr.b64 \t%rhs, $src, $amt2;\n\t",
    !strconcat("add.u64 \t$dst, %lhs, %rhs;\n\t",
    !strconcat("}}", ""))))))),
    []>;

def SUB_FRM_64 : SDNodeXForm<imm, [{
    return CurDAG->getTargetConstant(64-N->getZExtValue(), MVT::i32);
}]>;

def : Pat<(rotl Int64Regs:$src, (i32 imm:$amt)),
          (ROT64imm_sw Int64Regs:$src, imm:$amt, (SUB_FRM_64 node:$amt))>;
def : Pat<(rotr Int64Regs:$src, (i32 imm:$amt)),
          (ROT64imm_sw Int64Regs:$src, (SUB_FRM_64 node:$amt), imm:$amt)>;

def ROTL64reg_sw : NVPTXInst<(outs Int64Regs:$dst), (ins Int64Regs:$src,
    Int32Regs:$amt),
    !strconcat("{{\n\t",
    !strconcat(".reg .b64 %lhs;\n\t",
    !strconcat(".reg .b64 %rhs;\n\t",
    !strconcat(".reg .u32 %amt2;\n\t",
    !strconcat("shl.b64 \t%lhs, $src, $amt;\n\t",
    !strconcat("sub.u32 \t%amt2, 64, $amt;\n\t",
    !strconcat("shr.b64 \t%rhs, $src, %amt2;\n\t",
    !strconcat("add.u64 \t$dst, %lhs, %rhs;\n\t",
    !strconcat("}}", ""))))))))),
    [(set Int64Regs:$dst, (rotl Int64Regs:$src, Int32Regs:$amt))]>;

def ROTR64reg_sw : NVPTXInst<(outs Int64Regs:$dst), (ins Int64Regs:$src,
    Int32Regs:$amt),
    !strconcat("{{\n\t",
    !strconcat(".reg .b64 %lhs;\n\t",
    !strconcat(".reg .b64 %rhs;\n\t",
    !strconcat(".reg .u32 %amt2;\n\t",
    !strconcat("shr.b64 \t%lhs, $src, $amt;\n\t",
    !strconcat("sub.u32 \t%amt2, 64, $amt;\n\t",
    !strconcat("shl.b64 \t%rhs, $src, %amt2;\n\t",
    !strconcat("add.u64 \t$dst, %lhs, %rhs;\n\t",
    !strconcat("}}", ""))))))))),
    [(set Int64Regs:$dst, (rotr Int64Regs:$src, Int32Regs:$amt))]>;

// BFE - bit-field extract

multiclass BFE<string TyStr, RegisterClass RC> {
  // BFE supports both 32-bit and 64-bit values, but the start and length
  // operands are always 32-bit
  def rrr
    : NVPTXInst<(outs RC:$d),
                (ins RC:$a, Int32Regs:$b, Int32Regs:$c),
                !strconcat("bfe.", TyStr, " \t$d, $a, $b, $c;"), []>;
  def rri
    : NVPTXInst<(outs RC:$d),
                (ins RC:$a, Int32Regs:$b, i32imm:$c),
                !strconcat("bfe.", TyStr, " \t$d, $a, $b, $c;"), []>;
  def rii
    : NVPTXInst<(outs RC:$d),
                (ins RC:$a, i32imm:$b, i32imm:$c),
                !strconcat("bfe.", TyStr, " \t$d, $a, $b, $c;"), []>;
}

defm BFE_S32 : BFE<"s32", Int32Regs>;
defm BFE_U32 : BFE<"u32", Int32Regs>;
defm BFE_S64 : BFE<"s64", Int64Regs>;
defm BFE_U64 : BFE<"u64", Int64Regs>;

//-----------------------------------
// General Comparison
//-----------------------------------

// General setp instructions
multiclass SETP<string TypeStr, RegisterClass RC, Operand ImmCls> {
  def rr : NVPTXInst<(outs Int1Regs:$dst),
                     (ins RC:$a, RC:$b, CmpMode:$cmp),
            !strconcat("setp${cmp:base}${cmp:ftz}.", TypeStr, "\t$dst, $a, $b;"),
                     []>;
  def ri : NVPTXInst<(outs Int1Regs:$dst),
                     (ins RC:$a, ImmCls:$b, CmpMode:$cmp),
            !strconcat("setp${cmp:base}${cmp:ftz}.", TypeStr, "\t$dst, $a, $b;"),
                     []>;
  def ir : NVPTXInst<(outs Int1Regs:$dst),
                     (ins ImmCls:$a, RC:$b, CmpMode:$cmp),
            !strconcat("setp${cmp:base}${cmp:ftz}.", TypeStr, "\t$dst, $a, $b;"),
                     []>;
}

defm SETP_b16 : SETP<"b16", Int16Regs, i16imm>;
defm SETP_s16 : SETP<"s16", Int16Regs, i16imm>;
defm SETP_u16 : SETP<"u16", Int16Regs, i16imm>;
defm SETP_b32 : SETP<"b32", Int32Regs, i32imm>;
defm SETP_s32 : SETP<"s32", Int32Regs, i32imm>;
defm SETP_u32 : SETP<"u32", Int32Regs, i32imm>;
defm SETP_b64 : SETP<"b64", Int64Regs, i64imm>;
defm SETP_s64 : SETP<"s64", Int64Regs, i64imm>;
defm SETP_u64 : SETP<"u64", Int64Regs, i64imm>;
defm SETP_f32 : SETP<"f32", Float32Regs, f32imm>;
defm SETP_f64 : SETP<"f64", Float64Regs, f64imm>;

// General set instructions
multiclass SET<string TypeStr, RegisterClass RC, Operand ImmCls> {
  def rr : NVPTXInst<(outs Int32Regs:$dst),
                     (ins RC:$a, RC:$b, CmpMode:$cmp),
                     !strconcat("set$cmp.", TypeStr, "\t$dst, $a, $b;"), []>;
  def ri : NVPTXInst<(outs Int32Regs:$dst),
                     (ins RC:$a, ImmCls:$b, CmpMode:$cmp),
                     !strconcat("set$cmp.", TypeStr, "\t$dst, $a, $b;"), []>;
  def ir : NVPTXInst<(outs Int32Regs:$dst),
                     (ins ImmCls:$a, RC:$b, CmpMode:$cmp),
                     !strconcat("set$cmp.", TypeStr, "\t$dst, $a, $b;"), []>;
}

defm SET_b16 : SET<"b16", Int16Regs, i16imm>;
defm SET_s16 : SET<"s16", Int16Regs, i16imm>;
defm SET_u16 : SET<"u16", Int16Regs, i16imm>;
defm SET_b32 : SET<"b32", Int32Regs, i32imm>;
defm SET_s32 : SET<"s32", Int32Regs, i32imm>;
defm SET_u32 : SET<"u32", Int32Regs, i32imm>;
defm SET_b64 : SET<"b64", Int64Regs, i64imm>;
defm SET_s64 : SET<"s64", Int64Regs, i64imm>;
defm SET_u64 : SET<"u64", Int64Regs, i64imm>;
defm SET_f32 : SET<"f32", Float32Regs, f32imm>;
defm SET_f64 : SET<"f64", Float64Regs, f64imm>;

//-----------------------------------
// General Selection
//-----------------------------------

// General selp instructions
multiclass SELP<string TypeStr, RegisterClass RC, Operand ImmCls> {
  def rr : NVPTXInst<(outs RC:$dst),
                     (ins RC:$a, RC:$b, Int1Regs:$p),
                     !strconcat("selp.", TypeStr, "\t$dst, $a, $b, $p;"), []>;
  def ri : NVPTXInst<(outs RC:$dst),
                     (ins RC:$a, ImmCls:$b, Int1Regs:$p),
                     !strconcat("selp.", TypeStr, "\t$dst, $a, $b, $p;"), []>;
  def ir : NVPTXInst<(outs RC:$dst),
                     (ins ImmCls:$a, RC:$b, Int1Regs:$p),
                     !strconcat("selp.", TypeStr, "\t$dst, $a, $b, $p;"), []>;
  def ii : NVPTXInst<(outs RC:$dst),
                     (ins ImmCls:$a, ImmCls:$b, Int1Regs:$p),
                     !strconcat("selp.", TypeStr, "\t$dst, $a, $b, $p;"), []>;
}

multiclass SELP_PATTERN<string TypeStr, RegisterClass RC, Operand ImmCls,
                        SDNode ImmNode> {
  def rr : NVPTXInst<(outs RC:$dst),
                     (ins RC:$a, RC:$b, Int1Regs:$p),
                     !strconcat("selp.", TypeStr, "\t$dst, $a, $b, $p;"),
                     [(set RC:$dst, (select Int1Regs:$p, RC:$a, RC:$b))]>;
  def ri : NVPTXInst<(outs RC:$dst),
                     (ins RC:$a, ImmCls:$b, Int1Regs:$p),
                     !strconcat("selp.", TypeStr, "\t$dst, $a, $b, $p;"),
                     [(set RC:$dst, (select Int1Regs:$p, RC:$a, ImmNode:$b))]>;
  def ir : NVPTXInst<(outs RC:$dst),
                     (ins ImmCls:$a, RC:$b, Int1Regs:$p),
                     !strconcat("selp.", TypeStr, "\t$dst, $a, $b, $p;"),
                     [(set RC:$dst, (select Int1Regs:$p, ImmNode:$a, RC:$b))]>;
  def ii : NVPTXInst<(outs RC:$dst),
                     (ins ImmCls:$a, ImmCls:$b, Int1Regs:$p),
                     !strconcat("selp.", TypeStr, "\t$dst, $a, $b, $p;"),
                 [(set RC:$dst, (select Int1Regs:$p, ImmNode:$a, ImmNode:$b))]>;
}

defm SELP_b16 : SELP_PATTERN<"b16", Int16Regs, i16imm, imm>;
defm SELP_s16 : SELP<"s16", Int16Regs, i16imm>;
defm SELP_u16 : SELP<"u16", Int16Regs, i16imm>;
defm SELP_b32 : SELP_PATTERN<"b32", Int32Regs, i32imm, imm>;
defm SELP_s32 : SELP<"s32", Int32Regs, i32imm>;
defm SELP_u32 : SELP<"u32", Int32Regs, i32imm>;
defm SELP_b64 : SELP_PATTERN<"b64", Int64Regs, i64imm, imm>;
defm SELP_s64 : SELP<"s64", Int64Regs, i64imm>;
defm SELP_u64 : SELP<"u64", Int64Regs, i64imm>;
defm SELP_f32 : SELP_PATTERN<"f32", Float32Regs, f32imm, fpimm>;
defm SELP_f64 : SELP_PATTERN<"f64", Float64Regs, f64imm, fpimm>;

// Special select for predicate operands
def : Pat<(i1 (select Int1Regs:$p, Int1Regs:$a, Int1Regs:$b)),
              (ORb1rr (ANDb1rr Int1Regs:$p, Int1Regs:$a),
              (ANDb1rr (NOT1 Int1Regs:$p), Int1Regs:$b))>;

//-----------------------------------
// Data Movement (Load / Store, Move)
//-----------------------------------

def ADDRri : ComplexPattern<i32, 2, "SelectADDRri", [frameindex],
  [SDNPWantRoot]>;
def ADDRri64 : ComplexPattern<i64, 2, "SelectADDRri64", [frameindex],
  [SDNPWantRoot]>;

def MEMri : Operand<i32> {
  let PrintMethod = "printMemOperand";
  let MIOperandInfo = (ops Int32Regs, i32imm);
}
def MEMri64 : Operand<i64> {
  let PrintMethod = "printMemOperand";
  let MIOperandInfo = (ops Int64Regs, i64imm);
}

def imem : Operand<iPTR> {
    let PrintMethod = "printOperand";
}

def imemAny : Operand<iPTRAny> {
    let PrintMethod = "printOperand";
}

def LdStCode : Operand<i32> {
    let PrintMethod = "printLdStCode";
}

def SDTWrapper : SDTypeProfile<1, 1, [SDTCisSameAs<0, 1>, SDTCisPtrTy<0>]>;
def Wrapper    : SDNode<"NVPTXISD::Wrapper", SDTWrapper>;

def MOV_ADDR : NVPTXInst<(outs Int32Regs:$dst), (ins imem:$a),
                     "mov.u32 \t$dst, $a;",
                     [(set Int32Regs:$dst, (Wrapper tglobaladdr:$a))]>;

def MOV_ADDR64 : NVPTXInst<(outs Int64Regs:$dst), (ins imem:$a),
                     "mov.u64 \t$dst, $a;",
                     [(set Int64Regs:$dst, (Wrapper tglobaladdr:$a))]>;

// Get pointer to local stack
def MOV_DEPOT_ADDR
  : NVPTXInst<(outs Int32Regs:$d), (ins i32imm:$num),
              "mov.u32 \t$d, __local_depot$num;", []>;
def MOV_DEPOT_ADDR_64
  : NVPTXInst<(outs Int64Regs:$d), (ins i32imm:$num),
              "mov.u64 \t$d, __local_depot$num;", []>;


// copyPhysreg is hard-coded in NVPTXInstrInfo.cpp
let IsSimpleMove=1 in {
def IMOV1rr: NVPTXInst<(outs Int1Regs:$dst), (ins Int1Regs:$sss),
                   "mov.pred \t$dst, $sss;", []>;
def IMOV16rr: NVPTXInst<(outs Int16Regs:$dst), (ins Int16Regs:$sss),
                    "mov.u16 \t$dst, $sss;", []>;
def IMOV32rr: NVPTXInst<(outs Int32Regs:$dst), (ins Int32Regs:$sss),
                    "mov.u32 \t$dst, $sss;", []>;
def IMOV64rr: NVPTXInst<(outs Int64Regs:$dst), (ins Int64Regs:$sss),
                    "mov.u64 \t$dst, $sss;", []>;

def FMOV32rr: NVPTXInst<(outs Float32Regs:$dst), (ins Float32Regs:$src),
                    "mov.f32 \t$dst, $src;", []>;
def FMOV64rr: NVPTXInst<(outs Float64Regs:$dst), (ins Float64Regs:$src),
                    "mov.f64 \t$dst, $src;", []>;
}
def IMOV1ri: NVPTXInst<(outs Int1Regs:$dst), (ins i1imm:$src),
                    "mov.pred \t$dst, $src;",
          [(set Int1Regs:$dst, imm:$src)]>;
def IMOV16ri: NVPTXInst<(outs Int16Regs:$dst), (ins i16imm:$src),
                    "mov.u16 \t$dst, $src;",
          [(set Int16Regs:$dst, imm:$src)]>;
def IMOV32ri: NVPTXInst<(outs Int32Regs:$dst), (ins i32imm:$src),
                    "mov.u32 \t$dst, $src;",
          [(set Int32Regs:$dst, imm:$src)]>;
def IMOV64i: NVPTXInst<(outs Int64Regs:$dst), (ins i64imm:$src),
                    "mov.u64 \t$dst, $src;",
          [(set Int64Regs:$dst, imm:$src)]>;

def FMOV32ri: NVPTXInst<(outs Float32Regs:$dst), (ins f32imm:$src),
                    "mov.f32 \t$dst, $src;",
          [(set Float32Regs:$dst, fpimm:$src)]>;
def FMOV64ri: NVPTXInst<(outs Float64Regs:$dst), (ins f64imm:$src),
                    "mov.f64 \t$dst, $src;",
          [(set Float64Regs:$dst, fpimm:$src)]>;

def : Pat<(i32 (Wrapper texternalsym:$dst)), (IMOV32ri texternalsym:$dst)>;

//---- Copy Frame Index ----
def LEA_ADDRi : NVPTXInst<(outs Int32Regs:$dst), (ins MEMri:$addr),
                        "add.u32 \t$dst, ${addr:add};",
                        [(set Int32Regs:$dst, ADDRri:$addr)]>;
def LEA_ADDRi64 : NVPTXInst<(outs Int64Regs:$dst), (ins MEMri64:$addr),
                        "add.u64 \t$dst, ${addr:add};",
                        [(set Int64Regs:$dst, ADDRri64:$addr)]>;

//-----------------------------------
// Comparison and Selection
//-----------------------------------

multiclass ISET_FORMAT<PatFrag OpNode, PatLeaf Mode,
                       Instruction setp_16rr,
                       Instruction setp_16ri,
                       Instruction setp_16ir,
                       Instruction setp_32rr,
                       Instruction setp_32ri,
                       Instruction setp_32ir,
                       Instruction setp_64rr,
                       Instruction setp_64ri,
                       Instruction setp_64ir,
                       Instruction set_16rr,
                       Instruction set_16ri,
                       Instruction set_16ir,
                       Instruction set_32rr,
                       Instruction set_32ri,
                       Instruction set_32ir,
                       Instruction set_64rr,
                       Instruction set_64ri,
                       Instruction set_64ir> {
  // i16 -> pred
  def : Pat<(i1 (OpNode Int16Regs:$a, Int16Regs:$b)),
            (setp_16rr Int16Regs:$a, Int16Regs:$b, Mode)>;
  def : Pat<(i1 (OpNode Int16Regs:$a, imm:$b)),
            (setp_16ri Int16Regs:$a, imm:$b, Mode)>;
  def : Pat<(i1 (OpNode imm:$a, Int16Regs:$b)),
            (setp_16ir imm:$a, Int16Regs:$b, Mode)>;
  // i32 -> pred
  def : Pat<(i1 (OpNode Int32Regs:$a, Int32Regs:$b)),
            (setp_32rr Int32Regs:$a, Int32Regs:$b, Mode)>;
  def : Pat<(i1 (OpNode Int32Regs:$a, imm:$b)),
            (setp_32ri Int32Regs:$a, imm:$b, Mode)>;
  def : Pat<(i1 (OpNode imm:$a, Int32Regs:$b)),
            (setp_32ir imm:$a, Int32Regs:$b, Mode)>;
  // i64 -> pred
  def : Pat<(i1 (OpNode Int64Regs:$a, Int64Regs:$b)),
            (setp_64rr Int64Regs:$a, Int64Regs:$b, Mode)>;
  def : Pat<(i1 (OpNode Int64Regs:$a, imm:$b)),
            (setp_64ri Int64Regs:$a, imm:$b, Mode)>;
  def : Pat<(i1 (OpNode imm:$a, Int64Regs:$b)),
            (setp_64ir imm:$a, Int64Regs:$b, Mode)>;

  // i16 -> i32
  def : Pat<(i32 (OpNode Int16Regs:$a, Int16Regs:$b)),
            (set_16rr Int16Regs:$a, Int16Regs:$b, Mode)>;
  def : Pat<(i32 (OpNode Int16Regs:$a, imm:$b)),
            (set_16ri Int16Regs:$a, imm:$b, Mode)>;
  def : Pat<(i32 (OpNode imm:$a, Int16Regs:$b)),
            (set_16ir imm:$a, Int16Regs:$b, Mode)>;
  // i32 -> i32
  def : Pat<(i32 (OpNode Int32Regs:$a, Int32Regs:$b)),
            (set_32rr Int32Regs:$a, Int32Regs:$b, Mode)>;
  def : Pat<(i32 (OpNode Int32Regs:$a, imm:$b)),
            (set_32ri Int32Regs:$a, imm:$b, Mode)>;
  def : Pat<(i32 (OpNode imm:$a, Int32Regs:$b)),
            (set_32ir imm:$a, Int32Regs:$b, Mode)>;
  // i64 -> i32
  def : Pat<(i32 (OpNode Int64Regs:$a, Int64Regs:$b)),
            (set_64rr Int64Regs:$a, Int64Regs:$b, Mode)>;
  def : Pat<(i32 (OpNode Int64Regs:$a, imm:$b)),
            (set_64ri Int64Regs:$a, imm:$b, Mode)>;
  def : Pat<(i32 (OpNode imm:$a, Int64Regs:$b)),
            (set_64ir imm:$a, Int64Regs:$b, Mode)>;
}

multiclass ISET_FORMAT_SIGNED<PatFrag OpNode, PatLeaf Mode>
  : ISET_FORMAT<OpNode, Mode,
                SETP_s16rr, SETP_s16ri, SETP_s16ir,
                SETP_s32rr, SETP_s32ri, SETP_s32ir,
                SETP_s64rr, SETP_s64ri, SETP_s64ir,
                SET_s16rr, SET_s16ri, SET_s16ir,
                SET_s32rr, SET_s32ri, SET_s32ir,
                SET_s64rr, SET_s64ri, SET_s64ir> {
  // TableGen doesn't like empty multiclasses
  def : PatLeaf<(i32 0)>;
}

multiclass ISET_FORMAT_UNSIGNED<PatFrag OpNode, PatLeaf Mode>
  : ISET_FORMAT<OpNode, Mode,
                SETP_u16rr, SETP_u16ri, SETP_u16ir,
                SETP_u32rr, SETP_u32ri, SETP_u32ir,
                SETP_u64rr, SETP_u64ri, SETP_u64ir,
                SET_u16rr, SET_u16ri, SET_u16ir,
                SET_u32rr, SET_u32ri, SET_u32ir,
                SET_u64rr, SET_u64ri, SET_u64ir> {
  // TableGen doesn't like empty multiclasses
  def : PatLeaf<(i32 0)>;
}

defm : ISET_FORMAT_SIGNED<setgt, CmpGT>;
defm : ISET_FORMAT_UNSIGNED<setugt, CmpGT>;
defm : ISET_FORMAT_SIGNED<setlt, CmpLT>;
defm : ISET_FORMAT_UNSIGNED<setult, CmpLT>;
defm : ISET_FORMAT_SIGNED<setge, CmpGE>;
defm : ISET_FORMAT_UNSIGNED<setuge, CmpGE>;
defm : ISET_FORMAT_SIGNED<setle, CmpLE>;
defm : ISET_FORMAT_UNSIGNED<setule, CmpLE>;
defm : ISET_FORMAT_SIGNED<seteq, CmpEQ>;
defm : ISET_FORMAT_UNSIGNED<setueq, CmpEQ>;
defm : ISET_FORMAT_SIGNED<setne, CmpNE>;
defm : ISET_FORMAT_UNSIGNED<setune, CmpNE>;

// i1 compares
def : Pat<(setne Int1Regs:$a, Int1Regs:$b),
          (XORb1rr Int1Regs:$a, Int1Regs:$b)>;
def : Pat<(setune Int1Regs:$a, Int1Regs:$b),
          (XORb1rr Int1Regs:$a, Int1Regs:$b)>;

def : Pat<(seteq Int1Regs:$a, Int1Regs:$b),
          (NOT1 (XORb1rr Int1Regs:$a, Int1Regs:$b))>;
def : Pat<(setueq Int1Regs:$a, Int1Regs:$b),
          (NOT1 (XORb1rr Int1Regs:$a, Int1Regs:$b))>;

// i1 compare -> i32
def : Pat<(i32 (setne Int1Regs:$a, Int1Regs:$b)),
          (SELP_u32ii -1, 0, (XORb1rr Int1Regs:$a, Int1Regs:$b))>;
def : Pat<(i32 (setne Int1Regs:$a, Int1Regs:$b)),
          (SELP_u32ii 0, -1, (XORb1rr Int1Regs:$a, Int1Regs:$b))>;



multiclass FSET_FORMAT<PatFrag OpNode, PatLeaf Mode, PatLeaf ModeFTZ> {
  // f32 -> pred
  def : Pat<(i1 (OpNode Float32Regs:$a, Float32Regs:$b)),
            (SETP_f32rr Float32Regs:$a, Float32Regs:$b, ModeFTZ)>,
        Requires<[doF32FTZ]>;
  def : Pat<(i1 (OpNode Float32Regs:$a, Float32Regs:$b)),
            (SETP_f32rr Float32Regs:$a, Float32Regs:$b, Mode)>;
  def : Pat<(i1 (OpNode Float32Regs:$a, fpimm:$b)),
            (SETP_f32ri Float32Regs:$a, fpimm:$b, ModeFTZ)>,
        Requires<[doF32FTZ]>;
  def : Pat<(i1 (OpNode Float32Regs:$a, fpimm:$b)),
            (SETP_f32ri Float32Regs:$a, fpimm:$b, Mode)>;
  def : Pat<(i1 (OpNode fpimm:$a, Float32Regs:$b)),
            (SETP_f32ir fpimm:$a, Float32Regs:$b, ModeFTZ)>,
        Requires<[doF32FTZ]>;
  def : Pat<(i1 (OpNode fpimm:$a, Float32Regs:$b)),
            (SETP_f32ir fpimm:$a, Float32Regs:$b, Mode)>;

  // f64 -> pred
  def : Pat<(i1 (OpNode Float64Regs:$a, Float64Regs:$b)),
            (SETP_f64rr Float64Regs:$a, Float64Regs:$b, Mode)>;
  def : Pat<(i1 (OpNode Float64Regs:$a, fpimm:$b)),
            (SETP_f64ri Float64Regs:$a, fpimm:$b, Mode)>;
  def : Pat<(i1 (OpNode fpimm:$a, Float64Regs:$b)),
            (SETP_f64ir fpimm:$a, Float64Regs:$b, Mode)>;

  // f32 -> i32
  def : Pat<(i32 (OpNode Float32Regs:$a, Float32Regs:$b)),
            (SET_f32rr Float32Regs:$a, Float32Regs:$b, ModeFTZ)>,
        Requires<[doF32FTZ]>;
  def : Pat<(i32 (OpNode Float32Regs:$a, Float32Regs:$b)),
            (SET_f32rr Float32Regs:$a, Float32Regs:$b, Mode)>;
  def : Pat<(i32 (OpNode Float32Regs:$a, fpimm:$b)),
            (SET_f32ri Float32Regs:$a, fpimm:$b, ModeFTZ)>,
        Requires<[doF32FTZ]>;
  def : Pat<(i32 (OpNode Float32Regs:$a, fpimm:$b)),
            (SET_f32ri Float32Regs:$a, fpimm:$b, Mode)>;
  def : Pat<(i32 (OpNode fpimm:$a, Float32Regs:$b)),
            (SET_f32ir fpimm:$a, Float32Regs:$b, ModeFTZ)>,
        Requires<[doF32FTZ]>;
  def : Pat<(i32 (OpNode fpimm:$a, Float32Regs:$b)),
            (SET_f32ir fpimm:$a, Float32Regs:$b, Mode)>;

  // f64 -> i32
  def : Pat<(i32 (OpNode Float64Regs:$a, Float64Regs:$b)),
            (SET_f64rr Float64Regs:$a, Float64Regs:$b, Mode)>;
  def : Pat<(i32 (OpNode Float64Regs:$a, fpimm:$b)),
            (SET_f64ri Float64Regs:$a, fpimm:$b, Mode)>;
  def : Pat<(i32 (OpNode fpimm:$a, Float64Regs:$b)),
            (SET_f64ir fpimm:$a, Float64Regs:$b, Mode)>;
}

defm FSetGT : FSET_FORMAT<setogt, CmpGT, CmpGT_FTZ>;
defm FSetLT : FSET_FORMAT<setolt, CmpLT, CmpLT_FTZ>;
defm FSetGE : FSET_FORMAT<setoge, CmpGE, CmpGE_FTZ>;
defm FSetLE : FSET_FORMAT<setole, CmpLE, CmpLE_FTZ>;
defm FSetEQ : FSET_FORMAT<setoeq, CmpEQ, CmpEQ_FTZ>;
defm FSetNE : FSET_FORMAT<setone, CmpNE, CmpNE_FTZ>;

defm FSetUGT : FSET_FORMAT<setugt, CmpGTU, CmpGTU_FTZ>;
defm FSetULT : FSET_FORMAT<setult, CmpLTU, CmpLTU_FTZ>;
defm FSetUGE : FSET_FORMAT<setuge, CmpGEU, CmpGEU_FTZ>;
defm FSetULE : FSET_FORMAT<setule, CmpLEU, CmpLEU_FTZ>;
defm FSetUEQ : FSET_FORMAT<setueq, CmpEQU, CmpEQU_FTZ>;
defm FSetUNE : FSET_FORMAT<setune, CmpNEU, CmpNEU_FTZ>;

defm FSetNUM : FSET_FORMAT<seto, CmpNUM, CmpNUM_FTZ>;
defm FSetNAN : FSET_FORMAT<setuo, CmpNAN, CmpNAN_FTZ>;

//def ld_param         : SDNode<"NVPTXISD::LOAD_PARAM", SDTLoad,
//                        [SDNPHasChain, SDNPMayLoad, SDNPMemOperand]>;

def SDTDeclareParamProfile : SDTypeProfile<0, 3, [SDTCisInt<0>, SDTCisInt<1>,
  SDTCisInt<2>]>;
def SDTDeclareScalarParamProfile : SDTypeProfile<0, 3, [SDTCisInt<0>,
  SDTCisInt<1>, SDTCisInt<2>]>;
def SDTLoadParamProfile : SDTypeProfile<1, 2, [SDTCisInt<1>, SDTCisInt<2>]>;
def SDTLoadParamV2Profile : SDTypeProfile<2, 2, [SDTCisSameAs<0, 1>, SDTCisInt<2>, SDTCisInt<3>]>;
def SDTLoadParamV4Profile : SDTypeProfile<4, 2, [SDTCisInt<4>, SDTCisInt<5>]>;
def SDTPrintCallProfile : SDTypeProfile<0, 1, [SDTCisInt<0>]>;
def SDTPrintCallUniProfile : SDTypeProfile<0, 1, [SDTCisInt<0>]>;
def SDTStoreParamProfile : SDTypeProfile<0, 3, [SDTCisInt<0>, SDTCisInt<1>]>;
def SDTStoreParamV2Profile : SDTypeProfile<0, 4, [SDTCisInt<0>, SDTCisInt<1>]>;
def SDTStoreParamV4Profile : SDTypeProfile<0, 6, [SDTCisInt<0>, SDTCisInt<1>]>;
def SDTStoreParam32Profile : SDTypeProfile<0, 3, [SDTCisInt<0>, SDTCisInt<1>]>;
def SDTCallArgProfile : SDTypeProfile<0, 2, [SDTCisInt<0>]>;
def SDTCallArgMarkProfile : SDTypeProfile<0, 0, []>;
def SDTCallVoidProfile : SDTypeProfile<0, 1, []>;
def SDTCallValProfile : SDTypeProfile<1, 0, []>;
def SDTMoveParamProfile : SDTypeProfile<1, 1, []>;
def SDTStoreRetvalProfile : SDTypeProfile<0, 2, [SDTCisInt<0>]>;
def SDTStoreRetvalV2Profile : SDTypeProfile<0, 3, [SDTCisInt<0>]>;
def SDTStoreRetvalV4Profile : SDTypeProfile<0, 5, [SDTCisInt<0>]>;
def SDTPseudoUseParamProfile : SDTypeProfile<0, 1, []>;

def DeclareParam : SDNode<"NVPTXISD::DeclareParam", SDTDeclareParamProfile,
                       [SDNPHasChain, SDNPOutGlue, SDNPInGlue, SDNPSideEffect]>;
def DeclareScalarParam : SDNode<"NVPTXISD::DeclareScalarParam",
  SDTDeclareScalarParamProfile,
                       [SDNPHasChain, SDNPOutGlue, SDNPInGlue, SDNPSideEffect]>;
def DeclareRetParam : SDNode<"NVPTXISD::DeclareRetParam",
  SDTDeclareParamProfile,
                       [SDNPHasChain, SDNPOutGlue, SDNPInGlue, SDNPSideEffect]>;
def DeclareRet   : SDNode<"NVPTXISD::DeclareRet", SDTDeclareScalarParamProfile,
                       [SDNPHasChain, SDNPOutGlue, SDNPInGlue, SDNPSideEffect]>;
def LoadParam    : SDNode<"NVPTXISD::LoadParam", SDTLoadParamProfile,
                         [SDNPHasChain, SDNPMayLoad, SDNPOutGlue, SDNPInGlue]>;
def LoadParamV2  : SDNode<"NVPTXISD::LoadParamV2", SDTLoadParamV2Profile,
                         [SDNPHasChain, SDNPMayLoad, SDNPOutGlue, SDNPInGlue]>;
def LoadParamV4  : SDNode<"NVPTXISD::LoadParamV4", SDTLoadParamV4Profile,
                         [SDNPHasChain, SDNPMayLoad, SDNPOutGlue, SDNPInGlue]>;
def PrintCall    : SDNode<"NVPTXISD::PrintCall", SDTPrintCallProfile,
                       [SDNPHasChain, SDNPOutGlue, SDNPInGlue, SDNPSideEffect]>;
def PrintCallUni : SDNode<"NVPTXISD::PrintCallUni", SDTPrintCallUniProfile,
                       [SDNPHasChain, SDNPOutGlue, SDNPInGlue, SDNPSideEffect]>;
def StoreParam   : SDNode<"NVPTXISD::StoreParam", SDTStoreParamProfile,
                       [SDNPHasChain, SDNPOutGlue, SDNPInGlue, SDNPSideEffect]>;
def StoreParamV2 : SDNode<"NVPTXISD::StoreParamV2", SDTStoreParamV2Profile,
                         [SDNPHasChain, SDNPOutGlue, SDNPInGlue, SDNPSideEffect]>;
def StoreParamV4 : SDNode<"NVPTXISD::StoreParamV4", SDTStoreParamV4Profile,
                         [SDNPHasChain, SDNPOutGlue, SDNPInGlue, SDNPSideEffect]>;
def StoreParamU32 : SDNode<"NVPTXISD::StoreParamU32", SDTStoreParam32Profile,
                       [SDNPHasChain, SDNPOutGlue, SDNPInGlue, SDNPSideEffect]>;
def StoreParamS32 : SDNode<"NVPTXISD::StoreParamS32", SDTStoreParam32Profile,
                       [SDNPHasChain, SDNPOutGlue, SDNPInGlue, SDNPSideEffect]>;
def CallArgBegin : SDNode<"NVPTXISD::CallArgBegin", SDTCallArgMarkProfile,
                       [SDNPHasChain, SDNPOutGlue, SDNPInGlue, SDNPSideEffect]>;
def CallArg      : SDNode<"NVPTXISD::CallArg", SDTCallArgProfile,
                       [SDNPHasChain, SDNPOutGlue, SDNPInGlue, SDNPSideEffect]>;
def LastCallArg  : SDNode<"NVPTXISD::LastCallArg", SDTCallArgProfile,
                       [SDNPHasChain, SDNPOutGlue, SDNPInGlue, SDNPSideEffect]>;
def CallArgEnd   : SDNode<"NVPTXISD::CallArgEnd", SDTCallVoidProfile,
                       [SDNPHasChain, SDNPOutGlue, SDNPInGlue, SDNPSideEffect]>;
def CallVoid     : SDNode<"NVPTXISD::CallVoid", SDTCallVoidProfile,
                       [SDNPHasChain, SDNPOutGlue, SDNPInGlue, SDNPSideEffect]>;
def Prototype    : SDNode<"NVPTXISD::Prototype", SDTCallVoidProfile,
                       [SDNPHasChain, SDNPOutGlue, SDNPInGlue, SDNPSideEffect]>;
def CallVal      : SDNode<"NVPTXISD::CallVal", SDTCallValProfile,
                       [SDNPHasChain, SDNPOutGlue, SDNPInGlue, SDNPSideEffect]>;
def MoveParam    : SDNode<"NVPTXISD::MoveParam", SDTMoveParamProfile,
                         []>;
def StoreRetval  : SDNode<"NVPTXISD::StoreRetval", SDTStoreRetvalProfile,
                         [SDNPHasChain, SDNPSideEffect]>;
def StoreRetvalV2  : SDNode<"NVPTXISD::StoreRetvalV2", SDTStoreRetvalV2Profile,
                           [SDNPHasChain, SDNPSideEffect]>;
def StoreRetvalV4  : SDNode<"NVPTXISD::StoreRetvalV4", SDTStoreRetvalV4Profile,
                           [SDNPHasChain, SDNPSideEffect]>;
def PseudoUseParam : SDNode<"NVPTXISD::PseudoUseParam",
  SDTPseudoUseParamProfile,
                       [SDNPHasChain, SDNPOutGlue, SDNPInGlue, SDNPSideEffect]>;
def RETURNNode   : SDNode<"NVPTXISD::RETURN", SDTCallArgMarkProfile,
                         [SDNPHasChain, SDNPSideEffect]>;

class LoadParamMemInst<NVPTXRegClass regclass, string opstr> :
      NVPTXInst<(outs regclass:$dst), (ins i32imm:$b),
                !strconcat(!strconcat("ld.param", opstr),
                "\t$dst, [retval0+$b];"),
                []>;

class LoadParamRegInst<NVPTXRegClass regclass, string opstr> :
      NVPTXInst<(outs regclass:$dst), (ins i32imm:$b),
                !strconcat(!strconcat("mov", opstr),
                "\t$dst, retval$b;"),
                [(set regclass:$dst, (LoadParam (i32 0), (i32 imm:$b)))]>;

class LoadParamV2MemInst<NVPTXRegClass regclass, string opstr> :
      NVPTXInst<(outs regclass:$dst, regclass:$dst2), (ins i32imm:$b),
                !strconcat(!strconcat("ld.param.v2", opstr),
                "\t{{$dst, $dst2}}, [retval0+$b];"), []>;

class LoadParamV4MemInst<NVPTXRegClass regclass, string opstr> :
      NVPTXInst<(outs regclass:$dst, regclass:$dst2, regclass:$dst3,
                      regclass:$dst4),
                (ins i32imm:$b),
                !strconcat(!strconcat("ld.param.v4", opstr),
                "\t{{$dst, $dst2, $dst3, $dst4}}, [retval0+$b];"), []>;

class StoreParamInst<NVPTXRegClass regclass, string opstr> :
      NVPTXInst<(outs), (ins regclass:$val, i32imm:$a, i32imm:$b),
                !strconcat(!strconcat("st.param", opstr),
                "\t[param$a+$b], $val;"),
                []>;

class StoreParamV2Inst<NVPTXRegClass regclass, string opstr> :
      NVPTXInst<(outs), (ins regclass:$val, regclass:$val2,
                             i32imm:$a, i32imm:$b),
                !strconcat(!strconcat("st.param.v2", opstr),
                "\t[param$a+$b], {{$val, $val2}};"),
                []>;

class StoreParamV4Inst<NVPTXRegClass regclass, string opstr> :
      NVPTXInst<(outs), (ins regclass:$val, regclass:$val1, regclass:$val2,
                             regclass:$val3, i32imm:$a, i32imm:$b),
                !strconcat(!strconcat("st.param.v4", opstr),
                "\t[param$a+$b], {{$val, $val2, $val3, $val4}};"),
                []>;

class StoreRetvalInst<NVPTXRegClass regclass, string opstr> :
      NVPTXInst<(outs), (ins regclass:$val, i32imm:$a),
                !strconcat(!strconcat("st.param", opstr),
                "\t[func_retval0+$a], $val;"),
                []>;

class StoreRetvalV2Inst<NVPTXRegClass regclass, string opstr> :
      NVPTXInst<(outs), (ins regclass:$val, regclass:$val2, i32imm:$a),
                !strconcat(!strconcat("st.param.v2", opstr),
                "\t[func_retval0+$a], {{$val, $val2}};"),
                []>;

class StoreRetvalV4Inst<NVPTXRegClass regclass, string opstr> :
      NVPTXInst<(outs),
                (ins regclass:$val, regclass:$val2, regclass:$val3,
                     regclass:$val4, i32imm:$a),
                !strconcat(!strconcat("st.param.v4", opstr),
                "\t[func_retval0+$a], {{$val, $val2, $val3, $val4}};"),
                []>;

def PrintCallRetInst1 : NVPTXInst<(outs), (ins),
"call (retval0), ",
                                [(PrintCall (i32 1))]>;
def PrintCallRetInst2 : NVPTXInst<(outs), (ins),
"call (retval0, retval1), ",
                                [(PrintCall (i32 2))]>;
def PrintCallRetInst3 : NVPTXInst<(outs), (ins),
"call (retval0, retval1, retval2), ",
                                [(PrintCall (i32 3))]>;
def PrintCallRetInst4 : NVPTXInst<(outs), (ins),
"call (retval0, retval1, retval2, retval3), ",
                                [(PrintCall (i32 4))]>;
def PrintCallRetInst5 : NVPTXInst<(outs), (ins),
"call (retval0, retval1, retval2, retval3, retval4), ",
                                [(PrintCall (i32 5))]>;
def PrintCallRetInst6 : NVPTXInst<(outs), (ins),
"call (retval0, retval1, retval2, retval3, retval4, retval5), ",
                                [(PrintCall (i32 6))]>;
def PrintCallRetInst7 : NVPTXInst<(outs), (ins),
"call (retval0, retval1, retval2, retval3, retval4, retval5, retval6), ",
                                [(PrintCall (i32 7))]>;
def PrintCallRetInst8 : NVPTXInst<(outs), (ins),
!strconcat("call (retval0, retval1, retval2, retval3, retval4",
           ", retval5, retval6, retval7), "),
                                [(PrintCall (i32 8))]>;

def PrintCallNoRetInst : NVPTXInst<(outs), (ins), "call ",
                                [(PrintCall (i32 0))]>;

def PrintCallUniRetInst1 : NVPTXInst<(outs), (ins),
"call.uni (retval0), ",
                                [(PrintCallUni (i32 1))]>;
def PrintCallUniRetInst2 : NVPTXInst<(outs), (ins),
"call.uni (retval0, retval1), ",
                                [(PrintCallUni (i32 2))]>;
def PrintCallUniRetInst3 : NVPTXInst<(outs), (ins),
"call.uni (retval0, retval1, retval2), ",
                                [(PrintCallUni (i32 3))]>;
def PrintCallUniRetInst4 : NVPTXInst<(outs), (ins),
"call.uni (retval0, retval1, retval2, retval3), ",
                                [(PrintCallUni (i32 4))]>;
def PrintCallUniRetInst5 : NVPTXInst<(outs), (ins),
"call.uni (retval0, retval1, retval2, retval3, retval4), ",
                                [(PrintCallUni (i32 5))]>;
def PrintCallUniRetInst6 : NVPTXInst<(outs), (ins),
"call.uni (retval0, retval1, retval2, retval3, retval4, retval5), ",
                                [(PrintCallUni (i32 6))]>;
def PrintCallUniRetInst7 : NVPTXInst<(outs), (ins),
"call.uni (retval0, retval1, retval2, retval3, retval4, retval5, retval6), ",
                                [(PrintCallUni (i32 7))]>;
def PrintCallUniRetInst8 : NVPTXInst<(outs), (ins),
!strconcat("call.uni (retval0, retval1, retval2, retval3, retval4",
           ", retval5, retval6, retval7), "),
                                [(PrintCallUni (i32 8))]>;

def PrintCallUniNoRetInst : NVPTXInst<(outs), (ins), "call.uni ",
                                [(PrintCallUni (i32 0))]>;

def LoadParamMemI64    : LoadParamMemInst<Int64Regs, ".b64">;
def LoadParamMemI32    : LoadParamMemInst<Int32Regs, ".b32">;
def LoadParamMemI16    : LoadParamMemInst<Int16Regs, ".b16">;
def LoadParamMemI8     : LoadParamMemInst<Int16Regs, ".b8">;
def LoadParamMemV2I64  : LoadParamV2MemInst<Int64Regs, ".b64">;
def LoadParamMemV2I32  : LoadParamV2MemInst<Int32Regs, ".b32">;
def LoadParamMemV2I16  : LoadParamV2MemInst<Int16Regs, ".b16">;
def LoadParamMemV2I8   : LoadParamV2MemInst<Int16Regs, ".b8">;
def LoadParamMemV4I32  : LoadParamV4MemInst<Int32Regs, ".b32">;
def LoadParamMemV4I16  : LoadParamV4MemInst<Int16Regs, ".b16">;
def LoadParamMemV4I8   : LoadParamV4MemInst<Int16Regs, ".b8">;
def LoadParamMemF32    : LoadParamMemInst<Float32Regs, ".f32">;
def LoadParamMemF64    : LoadParamMemInst<Float64Regs, ".f64">;
def LoadParamMemV2F32  : LoadParamV2MemInst<Float32Regs, ".f32">;
def LoadParamMemV2F64  : LoadParamV2MemInst<Float64Regs, ".f64">;
def LoadParamMemV4F32  : LoadParamV4MemInst<Float32Regs, ".f32">;

def StoreParamI64    : StoreParamInst<Int64Regs, ".b64">;
def StoreParamI32    : StoreParamInst<Int32Regs, ".b32">;

def StoreParamI16    : StoreParamInst<Int16Regs, ".b16">;
def StoreParamI8     : StoreParamInst<Int16Regs, ".b8">;
def StoreParamV2I64  : StoreParamV2Inst<Int64Regs, ".b64">;
def StoreParamV2I32  : StoreParamV2Inst<Int32Regs, ".b32">;
def StoreParamV2I16  : StoreParamV2Inst<Int16Regs, ".b16">;
def StoreParamV2I8   : StoreParamV2Inst<Int16Regs, ".b8">;

// FIXME: StoreParamV4Inst crashes llvm-tblgen :(
//def StoreParamV4I32    : StoreParamV4Inst<Int32Regs, ".b32">;
def StoreParamV4I32    : NVPTXInst<(outs), (ins Int32Regs:$val, Int32Regs:$val2,
                                               Int32Regs:$val3, Int32Regs:$val4,
                                                i32imm:$a, i32imm:$b),
                   "st.param.b32\t[param$a+$b], {{$val, $val2, $val3, $val4}};",
                         []>;

def StoreParamV4I16    : NVPTXInst<(outs), (ins Int16Regs:$val, Int16Regs:$val2,
                                               Int16Regs:$val3, Int16Regs:$val4,
                                                i32imm:$a, i32imm:$b),
                "st.param.v4.b16\t[param$a+$b], {{$val, $val2, $val3, $val4}};",
                         []>;

def StoreParamV4I8     : NVPTXInst<(outs), (ins Int16Regs:$val, Int16Regs:$val2,
                                                Int16Regs:$val3, Int16Regs:$val4,
                                                i32imm:$a, i32imm:$b),
                 "st.param.v4.b8\t[param$a+$b], {{$val, $val2, $val3, $val4}};",
                         []>;

def StoreParamF32    : StoreParamInst<Float32Regs, ".f32">;
def StoreParamF64    : StoreParamInst<Float64Regs, ".f64">;
def StoreParamV2F32    : StoreParamV2Inst<Float32Regs, ".f32">;
def StoreParamV2F64    : StoreParamV2Inst<Float64Regs, ".f64">;
// FIXME: StoreParamV4Inst crashes llvm-tblgen :(
//def StoreParamV4F32    : StoreParamV4Inst<Float32Regs, ".f32">;
def StoreParamV4F32    : NVPTXInst<(outs),
                                   (ins Float32Regs:$val, Float32Regs:$val2,
                                        Float32Regs:$val3, Float32Regs:$val4,
                                        i32imm:$a, i32imm:$b),
                "st.param.v4.f32\t[param$a+$b], {{$val, $val2, $val3, $val4}};",
                        []>;


def StoreRetvalI64    : StoreRetvalInst<Int64Regs, ".b64">;
def StoreRetvalI32    : StoreRetvalInst<Int32Regs, ".b32">;
def StoreRetvalI16    : StoreRetvalInst<Int16Regs, ".b16">;
def StoreRetvalI8     : StoreRetvalInst<Int16Regs, ".b8">;
def StoreRetvalV2I64  : StoreRetvalV2Inst<Int64Regs, ".b64">;
def StoreRetvalV2I32  : StoreRetvalV2Inst<Int32Regs, ".b32">;
def StoreRetvalV2I16  : StoreRetvalV2Inst<Int16Regs, ".b16">;
def StoreRetvalV2I8   : StoreRetvalV2Inst<Int16Regs, ".b8">;
def StoreRetvalV4I32  : StoreRetvalV4Inst<Int32Regs, ".b32">;
def StoreRetvalV4I16  : StoreRetvalV4Inst<Int16Regs, ".b16">;
def StoreRetvalV4I8   : StoreRetvalV4Inst<Int16Regs, ".b8">;

def StoreRetvalF64    : StoreRetvalInst<Float64Regs, ".f64">;
def StoreRetvalF32    : StoreRetvalInst<Float32Regs, ".f32">;
def StoreRetvalV2F64  : StoreRetvalV2Inst<Float64Regs, ".f64">;
def StoreRetvalV2F32  : StoreRetvalV2Inst<Float32Regs, ".f32">;
def StoreRetvalV4F32  : StoreRetvalV4Inst<Float32Regs, ".f32">;

def CallArgBeginInst : NVPTXInst<(outs), (ins), "(", [(CallArgBegin)]>;
def CallArgEndInst1  : NVPTXInst<(outs), (ins), ");", [(CallArgEnd (i32 1))]>;
def CallArgEndInst0  : NVPTXInst<(outs), (ins), ")", [(CallArgEnd (i32 0))]>;
def RETURNInst       : NVPTXInst<(outs), (ins), "ret;", [(RETURNNode)]>;

class CallArgInst<NVPTXRegClass regclass> :
      NVPTXInst<(outs), (ins regclass:$a), "$a, ",
                [(CallArg (i32 0), regclass:$a)]>;

class LastCallArgInst<NVPTXRegClass regclass> :
      NVPTXInst<(outs), (ins regclass:$a), "$a",
                [(LastCallArg (i32 0), regclass:$a)]>;

def CallArgI64     : CallArgInst<Int64Regs>;
def CallArgI32     : CallArgInst<Int32Regs>;
def CallArgI16     : CallArgInst<Int16Regs>;

def CallArgF64     : CallArgInst<Float64Regs>;
def CallArgF32     : CallArgInst<Float32Regs>;

def LastCallArgI64 : LastCallArgInst<Int64Regs>;
def LastCallArgI32 : LastCallArgInst<Int32Regs>;
def LastCallArgI16 : LastCallArgInst<Int16Regs>;

def LastCallArgF64 : LastCallArgInst<Float64Regs>;
def LastCallArgF32 : LastCallArgInst<Float32Regs>;

def CallArgI32imm : NVPTXInst<(outs), (ins i32imm:$a), "$a, ",
                              [(CallArg (i32 0), (i32 imm:$a))]>;
def LastCallArgI32imm : NVPTXInst<(outs), (ins i32imm:$a), "$a",
                              [(LastCallArg (i32 0), (i32 imm:$a))]>;

def CallArgParam : NVPTXInst<(outs), (ins i32imm:$a), "param$a, ",
                             [(CallArg (i32 1), (i32 imm:$a))]>;
def LastCallArgParam : NVPTXInst<(outs), (ins i32imm:$a), "param$a",
                             [(LastCallArg (i32 1), (i32 imm:$a))]>;

def CallVoidInst : NVPTXInst<(outs), (ins imem:$addr),
                             "$addr, ",
                             [(CallVoid (Wrapper tglobaladdr:$addr))]>;
def CallVoidInstReg : NVPTXInst<(outs), (ins Int32Regs:$addr),
                             "$addr, ",
                             [(CallVoid Int32Regs:$addr)]>;
def CallVoidInstReg64 : NVPTXInst<(outs), (ins Int64Regs:$addr),
                             "$addr, ",
                             [(CallVoid Int64Regs:$addr)]>;
def PrototypeInst : NVPTXInst<(outs), (ins i32imm:$val),
                             ", prototype_$val;",
                             [(Prototype (i32 imm:$val))]>;

def DeclareRetMemInst : NVPTXInst<(outs),
  (ins i32imm:$align, i32imm:$size, i32imm:$num),
         ".param .align $align .b8 retval$num[$size];",
         [(DeclareRetParam (i32 imm:$align), (i32 imm:$size), (i32 imm:$num))]>;
def DeclareRetScalarInst : NVPTXInst<(outs), (ins i32imm:$size, i32imm:$num),
         ".param .b$size retval$num;",
         [(DeclareRet (i32 1), (i32 imm:$size), (i32 imm:$num))]>;
def DeclareRetRegInst : NVPTXInst<(outs), (ins i32imm:$size, i32imm:$num),
         ".reg .b$size retval$num;",
         [(DeclareRet (i32 2), (i32 imm:$size), (i32 imm:$num))]>;

def DeclareParamInst : NVPTXInst<(outs),
  (ins i32imm:$align, i32imm:$a, i32imm:$size),
         ".param .align $align .b8 param$a[$size];",
         [(DeclareParam (i32 imm:$align), (i32 imm:$a), (i32 imm:$size))]>;
def DeclareScalarParamInst : NVPTXInst<(outs), (ins i32imm:$a, i32imm:$size),
         ".param .b$size param$a;",
         [(DeclareScalarParam (i32 imm:$a), (i32 imm:$size), (i32 0))]>;
def DeclareScalarRegInst : NVPTXInst<(outs), (ins i32imm:$a, i32imm:$size),
         ".reg .b$size param$a;",
         [(DeclareScalarParam (i32 imm:$a), (i32 imm:$size), (i32 1))]>;

class MoveParamInst<NVPTXRegClass regclass, string asmstr> :
      NVPTXInst<(outs regclass:$dst), (ins regclass:$src),
                !strconcat(!strconcat("mov", asmstr), "\t$dst, $src;"),
                [(set regclass:$dst, (MoveParam regclass:$src))]>;

def MoveParamI64 : MoveParamInst<Int64Regs, ".b64">;
def MoveParamI32 : MoveParamInst<Int32Regs, ".b32">;
def MoveParamI16 : NVPTXInst<(outs Int16Regs:$dst), (ins Int16Regs:$src),
                   "cvt.u16.u32\t$dst, $src;",
                   [(set Int16Regs:$dst, (MoveParam Int16Regs:$src))]>;
def MoveParamF64 : MoveParamInst<Float64Regs, ".f64">;
def MoveParamF32 : MoveParamInst<Float32Regs, ".f32">;

class PseudoUseParamInst<NVPTXRegClass regclass> :
      NVPTXInst<(outs), (ins regclass:$src),
      "// Pseudo use of $src",
      [(PseudoUseParam regclass:$src)]>;

def PseudoUseParamI64 : PseudoUseParamInst<Int64Regs>;
def PseudoUseParamI32 : PseudoUseParamInst<Int32Regs>;
def PseudoUseParamI16 : PseudoUseParamInst<Int16Regs>;
def PseudoUseParamF64 : PseudoUseParamInst<Float64Regs>;
def PseudoUseParamF32 : PseudoUseParamInst<Float32Regs>;


//
// Load / Store Handling
//
multiclass LD<NVPTXRegClass regclass> {
  def _avar : NVPTXInst<(outs regclass:$dst),
    (ins LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec, LdStCode:$Sign,
      i32imm:$fromWidth, imem:$addr),
!strconcat("ld${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
           "$fromWidth \t$dst, [$addr];"), []>;
  def _areg : NVPTXInst<(outs regclass:$dst),
    (ins LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec, LdStCode:$Sign,
      i32imm:$fromWidth, Int32Regs:$addr),
!strconcat("ld${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
           "$fromWidth \t$dst, [$addr];"), []>;
  def _areg_64 : NVPTXInst<(outs regclass:$dst),
    (ins LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec, LdStCode:$Sign,
     i32imm:$fromWidth, Int64Regs:$addr),
     !strconcat("ld${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}$fromWidth",
                " \t$dst, [$addr];"), []>;
  def _ari : NVPTXInst<(outs regclass:$dst),
    (ins LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec, LdStCode:$Sign,
      i32imm:$fromWidth, Int32Regs:$addr, i32imm:$offset),
!strconcat("ld${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
           "$fromWidth \t$dst, [$addr+$offset];"), []>;
  def _ari_64 : NVPTXInst<(outs regclass:$dst),
    (ins LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec, LdStCode:$Sign,
     i32imm:$fromWidth, Int64Regs:$addr, i32imm:$offset),
    !strconcat("ld${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}$fromWidth",
               " \t$dst, [$addr+$offset];"), []>;
  def _asi : NVPTXInst<(outs regclass:$dst),
    (ins LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec, LdStCode:$Sign,
      i32imm:$fromWidth, imem:$addr, i32imm:$offset),
!strconcat("ld${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
           "$fromWidth \t$dst, [$addr+$offset];"), []>;
}

let mayLoad=1, neverHasSideEffects=1 in {
defm LD_i8  : LD<Int16Regs>;
defm LD_i16 : LD<Int16Regs>;
defm LD_i32 : LD<Int32Regs>;
defm LD_i64 : LD<Int64Regs>;
defm LD_f32 : LD<Float32Regs>;
defm LD_f64 : LD<Float64Regs>;
}

multiclass ST<NVPTXRegClass regclass> {
  def _avar : NVPTXInst<(outs),
    (ins regclass:$src, LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec,
      LdStCode:$Sign, i32imm:$toWidth, imem:$addr),
!strconcat("st${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}$toWidth",
           " \t[$addr], $src;"), []>;
  def _areg : NVPTXInst<(outs),
    (ins regclass:$src, LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec,
      LdStCode:$Sign, i32imm:$toWidth, Int32Regs:$addr),
!strconcat("st${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}$toWidth",
           " \t[$addr], $src;"), []>;
  def _areg_64 : NVPTXInst<(outs),
    (ins regclass:$src, LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec,
     LdStCode:$Sign, i32imm:$toWidth, Int64Regs:$addr),
  !strconcat("st${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}$toWidth ",
               "\t[$addr], $src;"), []>;
  def _ari : NVPTXInst<(outs),
    (ins regclass:$src, LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec,
      LdStCode:$Sign, i32imm:$toWidth, Int32Regs:$addr, i32imm:$offset),
!strconcat("st${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}$toWidth",
           " \t[$addr+$offset], $src;"), []>;
  def _ari_64 : NVPTXInst<(outs),
    (ins regclass:$src, LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec,
     LdStCode:$Sign, i32imm:$toWidth, Int64Regs:$addr, i32imm:$offset),
  !strconcat("st${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}$toWidth ",
               "\t[$addr+$offset], $src;"), []>;
  def _asi : NVPTXInst<(outs),
    (ins regclass:$src, LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec,
      LdStCode:$Sign, i32imm:$toWidth, imem:$addr, i32imm:$offset),
!strconcat("st${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}$toWidth",
           " \t[$addr+$offset], $src;"), []>;
}

let mayStore=1, neverHasSideEffects=1 in {
defm ST_i8  : ST<Int16Regs>;
defm ST_i16 : ST<Int16Regs>;
defm ST_i32 : ST<Int32Regs>;
defm ST_i64 : ST<Int64Regs>;
defm ST_f32 : ST<Float32Regs>;
defm ST_f64 : ST<Float64Regs>;
}

// The following is used only in and after vector elementizations.
// Vector elementization happens at the machine instruction level, so the
// following instruction
// never appears in the DAG.
multiclass LD_VEC<NVPTXRegClass regclass> {
  def _v2_avar : NVPTXInst<(outs regclass:$dst1, regclass:$dst2),
    (ins LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec, LdStCode:$Sign,
      i32imm:$fromWidth, imem:$addr),
    !strconcat("ld${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
               "$fromWidth \t{{$dst1, $dst2}}, [$addr];"), []>;
  def _v2_areg : NVPTXInst<(outs regclass:$dst1, regclass:$dst2),
    (ins LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec, LdStCode:$Sign,
      i32imm:$fromWidth, Int32Regs:$addr),
    !strconcat("ld${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
               "$fromWidth \t{{$dst1, $dst2}}, [$addr];"), []>;
  def _v2_areg_64 : NVPTXInst<(outs regclass:$dst1, regclass:$dst2),
    (ins LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec, LdStCode:$Sign,
     i32imm:$fromWidth, Int64Regs:$addr),
    !strconcat("ld${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
               "$fromWidth \t{{$dst1, $dst2}}, [$addr];"), []>;
  def _v2_ari : NVPTXInst<(outs regclass:$dst1, regclass:$dst2),
    (ins LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec, LdStCode:$Sign,
      i32imm:$fromWidth, Int32Regs:$addr, i32imm:$offset),
    !strconcat("ld${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
               "$fromWidth \t{{$dst1, $dst2}}, [$addr+$offset];"), []>;
  def _v2_ari_64 : NVPTXInst<(outs regclass:$dst1, regclass:$dst2),
    (ins LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec, LdStCode:$Sign,
     i32imm:$fromWidth, Int64Regs:$addr, i32imm:$offset),
    !strconcat("ld${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
               "$fromWidth \t{{$dst1, $dst2}}, [$addr+$offset];"), []>;
  def _v2_asi : NVPTXInst<(outs regclass:$dst1, regclass:$dst2),
    (ins LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec, LdStCode:$Sign,
      i32imm:$fromWidth, imem:$addr, i32imm:$offset),
    !strconcat("ld${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
               "$fromWidth \t{{$dst1, $dst2}}, [$addr+$offset];"), []>;
  def _v4_avar : NVPTXInst<(outs regclass:$dst1, regclass:$dst2,
      regclass:$dst3, regclass:$dst4),
    (ins LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec, LdStCode:$Sign,
      i32imm:$fromWidth, imem:$addr),
    !strconcat("ld${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
               "$fromWidth \t{{$dst1, $dst2, $dst3, $dst4}}, [$addr];"), []>;
  def _v4_areg : NVPTXInst<(outs regclass:$dst1, regclass:$dst2, regclass:$dst3,
      regclass:$dst4),
    (ins LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec, LdStCode:$Sign,
      i32imm:$fromWidth, Int32Regs:$addr),
    !strconcat("ld${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
               "$fromWidth \t{{$dst1, $dst2, $dst3, $dst4}}, [$addr];"), []>;
  def _v4_areg_64 : NVPTXInst<(outs regclass:$dst1, regclass:$dst2,
                               regclass:$dst3, regclass:$dst4),
    (ins LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec, LdStCode:$Sign,
     i32imm:$fromWidth, Int64Regs:$addr),
    !strconcat("ld${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
               "$fromWidth \t{{$dst1, $dst2, $dst3, $dst4}}, [$addr];"), []>;
  def _v4_ari : NVPTXInst<(outs regclass:$dst1, regclass:$dst2, regclass:$dst3,
      regclass:$dst4),
    (ins LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec, LdStCode:$Sign,
      i32imm:$fromWidth, Int32Regs:$addr, i32imm:$offset),
    !strconcat("ld${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
               "$fromWidth \t{{$dst1, $dst2, $dst3, $dst4}}, [$addr+$offset];"),
                []>;
  def _v4_ari_64 : NVPTXInst<(outs regclass:$dst1, regclass:$dst2,
                              regclass:$dst3, regclass:$dst4),
    (ins LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec, LdStCode:$Sign,
     i32imm:$fromWidth, Int64Regs:$addr, i32imm:$offset),
    !strconcat("ld${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
               "$fromWidth \t{{$dst1, $dst2, $dst3, $dst4}}, [$addr+$offset];"),
    []>;
  def _v4_asi : NVPTXInst<(outs regclass:$dst1, regclass:$dst2, regclass:$dst3,
      regclass:$dst4),
    (ins LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec, LdStCode:$Sign,
      i32imm:$fromWidth, imem:$addr, i32imm:$offset),
    !strconcat("ld${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
               "$fromWidth \t{{$dst1, $dst2, $dst3, $dst4}}, [$addr+$offset];"),
                []>;
}
let mayLoad=1, neverHasSideEffects=1 in {
defm LDV_i8  : LD_VEC<Int16Regs>;
defm LDV_i16 : LD_VEC<Int16Regs>;
defm LDV_i32 : LD_VEC<Int32Regs>;
defm LDV_i64 : LD_VEC<Int64Regs>;
defm LDV_f32 : LD_VEC<Float32Regs>;
defm LDV_f64 : LD_VEC<Float64Regs>;
}

multiclass ST_VEC<NVPTXRegClass regclass> {
  def _v2_avar : NVPTXInst<(outs),
    (ins regclass:$src1, regclass:$src2, LdStCode:$isVol, LdStCode:$addsp,
      LdStCode:$Vec, LdStCode:$Sign, i32imm:$fromWidth, imem:$addr),
    !strconcat("st${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
               "$fromWidth \t[$addr], {{$src1, $src2}};"), []>;
  def _v2_areg : NVPTXInst<(outs),
    (ins regclass:$src1, regclass:$src2, LdStCode:$isVol, LdStCode:$addsp,
      LdStCode:$Vec, LdStCode:$Sign, i32imm:$fromWidth, Int32Regs:$addr),
    !strconcat("st${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
               "$fromWidth \t[$addr], {{$src1, $src2}};"), []>;
  def _v2_areg_64 : NVPTXInst<(outs),
    (ins regclass:$src1, regclass:$src2, LdStCode:$isVol, LdStCode:$addsp,
     LdStCode:$Vec, LdStCode:$Sign, i32imm:$fromWidth, Int64Regs:$addr),
    !strconcat("st${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
               "$fromWidth \t[$addr], {{$src1, $src2}};"), []>;
  def _v2_ari : NVPTXInst<(outs),
    (ins regclass:$src1, regclass:$src2, LdStCode:$isVol, LdStCode:$addsp,
      LdStCode:$Vec, LdStCode:$Sign, i32imm:$fromWidth, Int32Regs:$addr,
      i32imm:$offset),
    !strconcat("st${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
               "$fromWidth \t[$addr+$offset], {{$src1, $src2}};"), []>;
  def _v2_ari_64 : NVPTXInst<(outs),
    (ins regclass:$src1, regclass:$src2, LdStCode:$isVol, LdStCode:$addsp,
     LdStCode:$Vec, LdStCode:$Sign, i32imm:$fromWidth, Int64Regs:$addr,
     i32imm:$offset),
    !strconcat("st${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
               "$fromWidth \t[$addr+$offset], {{$src1, $src2}};"), []>;
  def _v2_asi : NVPTXInst<(outs),
    (ins regclass:$src1, regclass:$src2, LdStCode:$isVol, LdStCode:$addsp,
      LdStCode:$Vec, LdStCode:$Sign, i32imm:$fromWidth, imem:$addr,
      i32imm:$offset),
    !strconcat("st${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
               "$fromWidth \t[$addr+$offset], {{$src1, $src2}};"), []>;
  def _v4_avar : NVPTXInst<(outs),
    (ins regclass:$src1, regclass:$src2, regclass:$src3, regclass:$src4,
      LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec, LdStCode:$Sign,
      i32imm:$fromWidth, imem:$addr),
    !strconcat("st${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
               "$fromWidth \t[$addr], {{$src1, $src2, $src3, $src4}};"), []>;
  def _v4_areg : NVPTXInst<(outs),
    (ins regclass:$src1, regclass:$src2, regclass:$src3, regclass:$src4,
      LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec, LdStCode:$Sign,
      i32imm:$fromWidth, Int32Regs:$addr),
    !strconcat("st${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
               "$fromWidth \t[$addr], {{$src1, $src2, $src3, $src4}};"), []>;
  def _v4_areg_64 : NVPTXInst<(outs),
    (ins regclass:$src1, regclass:$src2, regclass:$src3, regclass:$src4,
     LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec, LdStCode:$Sign,
     i32imm:$fromWidth, Int64Regs:$addr),
    !strconcat("st${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
               "$fromWidth \t[$addr], {{$src1, $src2, $src3, $src4}};"), []>;
  def _v4_ari : NVPTXInst<(outs),
    (ins regclass:$src1, regclass:$src2, regclass:$src3, regclass:$src4,
      LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec, LdStCode:$Sign,
      i32imm:$fromWidth, Int32Regs:$addr, i32imm:$offset),
    !strconcat("st${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
               "$fromWidth \t[$addr+$offset], {{$src1, $src2, $src3, $src4}};"),
    []>;
  def _v4_ari_64 : NVPTXInst<(outs),
    (ins regclass:$src1, regclass:$src2, regclass:$src3, regclass:$src4,
     LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec, LdStCode:$Sign,
     i32imm:$fromWidth, Int64Regs:$addr, i32imm:$offset),
    !strconcat("st${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
               "$fromWidth \t[$addr+$offset], {{$src1, $src2, $src3, $src4}};"),
     []>;
  def _v4_asi : NVPTXInst<(outs),
    (ins regclass:$src1, regclass:$src2, regclass:$src3, regclass:$src4,
      LdStCode:$isVol, LdStCode:$addsp, LdStCode:$Vec, LdStCode:$Sign,
      i32imm:$fromWidth, imem:$addr, i32imm:$offset),
    !strconcat("st${isVol:volatile}${addsp:addsp}${Vec:vec}.${Sign:sign}",
               "$fromWidth \t[$addr+$offset], {{$src1, $src2, $src3, $src4}};"),
    []>;
}
let mayStore=1, neverHasSideEffects=1 in {
defm STV_i8  : ST_VEC<Int16Regs>;
defm STV_i16 : ST_VEC<Int16Regs>;
defm STV_i32 : ST_VEC<Int32Regs>;
defm STV_i64 : ST_VEC<Int64Regs>;
defm STV_f32 : ST_VEC<Float32Regs>;
defm STV_f64 : ST_VEC<Float64Regs>;
}


//---- Conversion ----

class F_BITCONVERT<string SzStr, NVPTXRegClass regclassIn,
  NVPTXRegClass regclassOut> :
           NVPTXInst<(outs regclassOut:$d), (ins regclassIn:$a),
           !strconcat("mov.b", !strconcat(SzStr, " \t $d, $a;")),
     [(set regclassOut:$d, (bitconvert regclassIn:$a))]>;

def BITCONVERT_32_I2F : F_BITCONVERT<"32", Int32Regs, Float32Regs>;
def BITCONVERT_32_F2I : F_BITCONVERT<"32", Float32Regs, Int32Regs>;
def BITCONVERT_64_I2F : F_BITCONVERT<"64", Int64Regs, Float64Regs>;
def BITCONVERT_64_F2I : F_BITCONVERT<"64", Float64Regs, Int64Regs>;

// NOTE: pred->fp are currently sub-optimal due to an issue in TableGen where
// we cannot specify floating-point literals in isel patterns.  Therefore, we
// use an integer selp to select either 1 or 0 and then cvt to floating-point.

// sint -> f32
def : Pat<(f32 (sint_to_fp Int1Regs:$a)),
          (CVT_f32_s32 (SELP_u32ii 1, 0, Int1Regs:$a), CvtRN)>;
def : Pat<(f32 (sint_to_fp Int16Regs:$a)),
          (CVT_f32_s16 Int16Regs:$a, CvtRN)>;
def : Pat<(f32 (sint_to_fp Int32Regs:$a)),
          (CVT_f32_s32 Int32Regs:$a, CvtRN)>;
def : Pat<(f32 (sint_to_fp Int64Regs:$a)),
          (CVT_f32_s64 Int64Regs:$a, CvtRN)>;

// uint -> f32
def : Pat<(f32 (uint_to_fp Int1Regs:$a)),
          (CVT_f32_u32 (SELP_u32ii 1, 0, Int1Regs:$a), CvtRN)>;
def : Pat<(f32 (uint_to_fp Int16Regs:$a)),
          (CVT_f32_u16 Int16Regs:$a, CvtRN)>;
def : Pat<(f32 (uint_to_fp Int32Regs:$a)),
          (CVT_f32_u32 Int32Regs:$a, CvtRN)>;
def : Pat<(f32 (uint_to_fp Int64Regs:$a)),
          (CVT_f32_u64 Int64Regs:$a, CvtRN)>;

// sint -> f64
def : Pat<(f64 (sint_to_fp Int1Regs:$a)),
          (CVT_f64_s32 (SELP_u32ii 1, 0, Int1Regs:$a), CvtRN)>;
def : Pat<(f64 (sint_to_fp Int16Regs:$a)),
          (CVT_f64_s16 Int16Regs:$a, CvtRN)>;
def : Pat<(f64 (sint_to_fp Int32Regs:$a)),
          (CVT_f64_s32 Int32Regs:$a, CvtRN)>;
def : Pat<(f64 (sint_to_fp Int64Regs:$a)),
          (CVT_f64_s64 Int64Regs:$a, CvtRN)>;

// uint -> f64
def : Pat<(f64 (uint_to_fp Int1Regs:$a)),
          (CVT_f64_u32 (SELP_u32ii 1, 0, Int1Regs:$a), CvtRN)>;
def : Pat<(f64 (uint_to_fp Int16Regs:$a)),
          (CVT_f64_u16 Int16Regs:$a, CvtRN)>;
def : Pat<(f64 (uint_to_fp Int32Regs:$a)),
          (CVT_f64_u32 Int32Regs:$a, CvtRN)>;
def : Pat<(f64 (uint_to_fp Int64Regs:$a)),
          (CVT_f64_u64 Int64Regs:$a, CvtRN)>;


// f32 -> sint
def : Pat<(i1 (fp_to_sint Float32Regs:$a)),
          (SETP_b32ri (BITCONVERT_32_F2I Float32Regs:$a), 0, CmpEQ)>;
def : Pat<(i16 (fp_to_sint Float32Regs:$a)),
          (CVT_s16_f32 Float32Regs:$a, CvtRZI_FTZ)>, Requires<[doF32FTZ]>;
def : Pat<(i16 (fp_to_sint Float32Regs:$a)),
          (CVT_s16_f32 Float32Regs:$a, CvtRZI)>;
def : Pat<(i32 (fp_to_sint Float32Regs:$a)),
          (CVT_s32_f32 Float32Regs:$a, CvtRZI_FTZ)>, Requires<[doF32FTZ]>;
def : Pat<(i32 (fp_to_sint Float32Regs:$a)),
          (CVT_s32_f32 Float32Regs:$a, CvtRZI)>;
def : Pat<(i64 (fp_to_sint Float32Regs:$a)),
          (CVT_s64_f32 Float32Regs:$a, CvtRZI_FTZ)>, Requires<[doF32FTZ]>;
def : Pat<(i64 (fp_to_sint Float32Regs:$a)),
          (CVT_s64_f32 Float32Regs:$a, CvtRZI)>;

// f32 -> uint
def : Pat<(i1 (fp_to_uint Float32Regs:$a)),
          (SETP_b32ri (BITCONVERT_32_F2I Float32Regs:$a), 0, CmpEQ)>;
def : Pat<(i16 (fp_to_uint Float32Regs:$a)),
          (CVT_u16_f32 Float32Regs:$a, CvtRZI_FTZ)>, Requires<[doF32FTZ]>;
def : Pat<(i16 (fp_to_uint Float32Regs:$a)),
          (CVT_u16_f32 Float32Regs:$a, CvtRZI)>;
def : Pat<(i32 (fp_to_uint Float32Regs:$a)),
          (CVT_u32_f32 Float32Regs:$a, CvtRZI_FTZ)>, Requires<[doF32FTZ]>;
def : Pat<(i32 (fp_to_uint Float32Regs:$a)),
          (CVT_u32_f32 Float32Regs:$a, CvtRZI)>;
def : Pat<(i64 (fp_to_uint Float32Regs:$a)),
          (CVT_u64_f32 Float32Regs:$a, CvtRZI_FTZ)>, Requires<[doF32FTZ]>;
def : Pat<(i64 (fp_to_uint Float32Regs:$a)),
          (CVT_u64_f32 Float32Regs:$a, CvtRZI)>;

// f64 -> sint
def : Pat<(i1 (fp_to_sint Float64Regs:$a)),
          (SETP_b64ri (BITCONVERT_64_F2I Float64Regs:$a), 0, CmpEQ)>;
def : Pat<(i16 (fp_to_sint Float64Regs:$a)),
          (CVT_s16_f64 Float64Regs:$a, CvtRZI)>;
def : Pat<(i32 (fp_to_sint Float64Regs:$a)),
          (CVT_s32_f64 Float64Regs:$a, CvtRZI)>;
def : Pat<(i64 (fp_to_sint Float64Regs:$a)),
          (CVT_s64_f64 Float64Regs:$a, CvtRZI)>;

// f64 -> uint
def : Pat<(i1 (fp_to_uint Float64Regs:$a)),
          (SETP_b64ri (BITCONVERT_64_F2I Float64Regs:$a), 0, CmpEQ)>;
def : Pat<(i16 (fp_to_uint Float64Regs:$a)),
          (CVT_u16_f64 Float64Regs:$a, CvtRZI)>;
def : Pat<(i32 (fp_to_uint Float64Regs:$a)),
          (CVT_u32_f64 Float64Regs:$a, CvtRZI)>;
def : Pat<(i64 (fp_to_uint Float64Regs:$a)),
          (CVT_u64_f64 Float64Regs:$a, CvtRZI)>;

// sext i1
def : Pat<(i16 (sext Int1Regs:$a)),
          (SELP_s16ii -1, 0, Int1Regs:$a)>;
def : Pat<(i32 (sext Int1Regs:$a)),
          (SELP_s32ii -1, 0, Int1Regs:$a)>;
def : Pat<(i64 (sext Int1Regs:$a)),
          (SELP_s64ii -1, 0, Int1Regs:$a)>;

// zext i1
def : Pat<(i16 (zext Int1Regs:$a)),
          (SELP_u16ii 1, 0, Int1Regs:$a)>;
def : Pat<(i32 (zext Int1Regs:$a)),
          (SELP_u32ii 1, 0, Int1Regs:$a)>;
def : Pat<(i64 (zext Int1Regs:$a)),
          (SELP_u64ii 1, 0, Int1Regs:$a)>;

// anyext i1
def : Pat<(i16 (anyext Int1Regs:$a)),
          (SELP_u16ii -1, 0, Int1Regs:$a)>;
def : Pat<(i32 (anyext Int1Regs:$a)),
          (SELP_u32ii -1, 0, Int1Regs:$a)>;
def : Pat<(i64 (anyext Int1Regs:$a)),
          (SELP_u64ii -1, 0, Int1Regs:$a)>;

// sext i16
def : Pat<(i32 (sext Int16Regs:$a)),
          (CVT_s32_s16 Int16Regs:$a, CvtNONE)>;
def : Pat<(i64 (sext Int16Regs:$a)),
          (CVT_s64_s16 Int16Regs:$a, CvtNONE)>;

// zext i16
def : Pat<(i32 (zext Int16Regs:$a)),
          (CVT_u32_u16 Int16Regs:$a, CvtNONE)>;
def : Pat<(i64 (zext Int16Regs:$a)),
          (CVT_u64_u16 Int16Regs:$a, CvtNONE)>;

// anyext i16
def : Pat<(i32 (anyext Int16Regs:$a)),
          (CVT_u32_u16 Int16Regs:$a, CvtNONE)>;
def : Pat<(i64 (anyext Int16Regs:$a)),
          (CVT_u64_u16 Int16Regs:$a, CvtNONE)>;

// sext i32
def : Pat<(i64 (sext Int32Regs:$a)),
          (CVT_s64_s32 Int32Regs:$a, CvtNONE)>;

// zext i32
def : Pat<(i64 (zext Int32Regs:$a)),
          (CVT_u64_u32 Int32Regs:$a, CvtNONE)>;

// anyext i32
def : Pat<(i64 (anyext Int32Regs:$a)),
          (CVT_u64_u32 Int32Regs:$a, CvtNONE)>;


// truncate i64
def : Pat<(i32 (trunc Int64Regs:$a)),
          (CVT_u32_u64 Int64Regs:$a, CvtNONE)>;
def : Pat<(i16 (trunc Int64Regs:$a)),
          (CVT_u16_u64 Int64Regs:$a, CvtNONE)>;
def : Pat<(i1 (trunc Int64Regs:$a)),
          (SETP_b64ri (ANDb64ri Int64Regs:$a, 1), 1, CmpEQ)>;

// truncate i32
def : Pat<(i16 (trunc Int32Regs:$a)),
          (CVT_u16_u32 Int32Regs:$a, CvtNONE)>;
def : Pat<(i1 (trunc Int32Regs:$a)),
          (SETP_b32ri (ANDb32ri Int32Regs:$a, 1), 1, CmpEQ)>;

// truncate i16
def : Pat<(i1 (trunc Int16Regs:$a)),
          (SETP_b16ri (ANDb16ri Int16Regs:$a, 1), 1, CmpEQ)>;

// sext_inreg
def : Pat<(sext_inreg Int16Regs:$a, i8), (CVT_INREG_s16_s8 Int16Regs:$a)>;
def : Pat<(sext_inreg Int32Regs:$a, i8), (CVT_INREG_s32_s8 Int32Regs:$a)>;
def : Pat<(sext_inreg Int32Regs:$a, i16), (CVT_INREG_s32_s16 Int32Regs:$a)>;
def : Pat<(sext_inreg Int64Regs:$a, i8), (CVT_INREG_s64_s8 Int64Regs:$a)>;
def : Pat<(sext_inreg Int64Regs:$a, i16), (CVT_INREG_s64_s16 Int64Regs:$a)>;
def : Pat<(sext_inreg Int64Regs:$a, i32), (CVT_INREG_s64_s32 Int64Regs:$a)>;


// Select instructions with 32-bit predicates
def : Pat<(select Int32Regs:$pred, Int16Regs:$a, Int16Regs:$b),
          (SELP_b16rr Int16Regs:$a, Int16Regs:$b,
          (SETP_b32ri (ANDb32ri Int32Regs:$pred, 1), 1, CmpEQ))>;
def : Pat<(select Int32Regs:$pred, Int32Regs:$a, Int32Regs:$b),
          (SELP_b32rr Int32Regs:$a, Int32Regs:$b,
          (SETP_b32ri (ANDb32ri Int32Regs:$pred, 1), 1, CmpEQ))>;
def : Pat<(select Int32Regs:$pred, Int64Regs:$a, Int64Regs:$b),
          (SELP_b64rr Int64Regs:$a, Int64Regs:$b,
          (SETP_b32ri (ANDb32ri Int32Regs:$pred, 1), 1, CmpEQ))>;
def : Pat<(select Int32Regs:$pred, Float32Regs:$a, Float32Regs:$b),
          (SELP_f32rr Float32Regs:$a, Float32Regs:$b,
          (SETP_b32ri (ANDb32ri Int32Regs:$pred, 1), 1, CmpEQ))>;
def : Pat<(select Int32Regs:$pred, Float64Regs:$a, Float64Regs:$b),
          (SELP_f64rr Float64Regs:$a, Float64Regs:$b,
          (SETP_b32ri (ANDb32ri Int32Regs:$pred, 1), 1, CmpEQ))>;


// pack a set of smaller int registers to a larger int register
def V4I16toI64 : NVPTXInst<(outs Int64Regs:$d),
                          (ins Int16Regs:$s1, Int16Regs:$s2,
                               Int16Regs:$s3, Int16Regs:$s4),
                          "mov.b64\t$d, {{$s1, $s2, $s3, $s4}};",
                          []>;
def V2I16toI32 : NVPTXInst<(outs Int32Regs:$d),
                          (ins Int16Regs:$s1, Int16Regs:$s2),
                          "mov.b32\t$d, {{$s1, $s2}};",
                          []>;
def V2I32toI64 : NVPTXInst<(outs Int64Regs:$d),
                          (ins Int32Regs:$s1, Int32Regs:$s2),
                          "mov.b64\t$d, {{$s1, $s2}};",
                          []>;
def V2F32toF64 : NVPTXInst<(outs Float64Regs:$d),
                          (ins Float32Regs:$s1, Float32Regs:$s2),
                          "mov.b64\t$d, {{$s1, $s2}};",
                          []>;

// unpack a larger int register to a set of smaller int registers
def I64toV4I16 : NVPTXInst<(outs Int16Regs:$d1, Int16Regs:$d2,
                                 Int16Regs:$d3, Int16Regs:$d4),
                           (ins Int64Regs:$s),
                           "mov.b64\t{{$d1, $d2, $d3, $d4}}, $s;",
                          []>;
def I32toV2I16 : NVPTXInst<(outs Int16Regs:$d1, Int16Regs:$d2),
                           (ins Int32Regs:$s),
                           "mov.b32\t{{$d1, $d2}}, $s;",
                          []>;
def I64toV2I32 : NVPTXInst<(outs Int32Regs:$d1, Int32Regs:$d2),
                           (ins Int64Regs:$s),
                           "mov.b64\t{{$d1, $d2}}, $s;",
                          []>;
def F64toV2F32 : NVPTXInst<(outs Float32Regs:$d1, Float32Regs:$d2),
                           (ins Float64Regs:$s),
                           "mov.b64\t{{$d1, $d2}}, $s;",
                          []>;

// Count leading zeros
def CLZr32 : NVPTXInst<(outs Int32Regs:$d), (ins Int32Regs:$a),
                       "clz.b32\t$d, $a;",
                       []>;
def CLZr64 : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
                       "clz.b64\t$d, $a;",
                       []>;

// 32-bit has a direct PTX instruction
def : Pat<(ctlz Int32Regs:$a),
          (CLZr32 Int32Regs:$a)>;
def : Pat<(ctlz_zero_undef Int32Regs:$a),
          (CLZr32 Int32Regs:$a)>;

// For 64-bit, the result in PTX is actually 32-bit so we zero-extend
// to 64-bit to match the LLVM semantics
def : Pat<(ctlz Int64Regs:$a),
          (CVT_u64_u32 (CLZr64 Int64Regs:$a), CvtNONE)>;
def : Pat<(ctlz_zero_undef Int64Regs:$a),
          (CVT_u64_u32 (CLZr64 Int64Regs:$a), CvtNONE)>;

// For 16-bit, we zero-extend to 32-bit, then trunc the result back
// to 16-bits (ctlz of a 16-bit value is guaranteed to require less
// than 16 bits to store). We also need to subtract 16 because the
// high-order 16 zeros were counted.
def : Pat<(ctlz Int16Regs:$a),
          (SUBi16ri (CVT_u16_u32 (CLZr32
            (CVT_u32_u16 Int16Regs:$a, CvtNONE)),
           CvtNONE), 16)>;
def : Pat<(ctlz_zero_undef Int16Regs:$a),
          (SUBi16ri (CVT_u16_u32 (CLZr32
            (CVT_u32_u16 Int16Regs:$a, CvtNONE)),
           CvtNONE), 16)>;

// Population count
def POPCr32 : NVPTXInst<(outs Int32Regs:$d), (ins Int32Regs:$a),
                        "popc.b32\t$d, $a;",
                        []>;
def POPCr64 : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
                        "popc.b64\t$d, $a;",
                        []>;

// 32-bit has a direct PTX instruction
def : Pat<(ctpop Int32Regs:$a),
          (POPCr32 Int32Regs:$a)>;

// For 64-bit, the result in PTX is actually 32-bit so we zero-extend
// to 64-bit to match the LLVM semantics
def : Pat<(ctpop Int64Regs:$a),
          (CVT_u64_u32 (POPCr64 Int64Regs:$a), CvtNONE)>;

// For 16-bit, we zero-extend to 32-bit, then trunc the result back
// to 16-bits (ctpop of a 16-bit value is guaranteed to require less
// than 16 bits to store)
def : Pat<(ctpop Int16Regs:$a),
          (CVT_u16_u32 (POPCr32 (CVT_u32_u16 Int16Regs:$a, CvtNONE)),
           CvtNONE)>;

// fround f64 -> f32
def : Pat<(f32 (fround Float64Regs:$a)),
          (CVT_f32_f64 Float64Regs:$a, CvtRN_FTZ)>, Requires<[doF32FTZ]>;
def : Pat<(f32 (fround Float64Regs:$a)),
          (CVT_f32_f64 Float64Regs:$a, CvtRN)>;

// fextend f32 -> f64
def : Pat<(f64 (fextend Float32Regs:$a)),
          (CVT_f64_f32 Float32Regs:$a, CvtNONE_FTZ)>, Requires<[doF32FTZ]>;
def : Pat<(f64 (fextend Float32Regs:$a)),
          (CVT_f64_f32 Float32Regs:$a, CvtNONE)>;

def retflag       : SDNode<"NVPTXISD::RET_FLAG", SDTNone,
                           [SDNPHasChain, SDNPOptInGlue]>;

//-----------------------------------
// Control-flow
//-----------------------------------

let isTerminator=1 in {
   let isReturn=1, isBarrier=1 in
      def Return : NVPTXInst<(outs), (ins), "ret;", [(retflag)]>;

   let isBranch=1 in
      def CBranch : NVPTXInst<(outs), (ins Int1Regs:$a, brtarget:$target),
                          "@$a bra \t$target;",
                           [(brcond Int1Regs:$a, bb:$target)]>;
   let isBranch=1 in
      def CBranchOther : NVPTXInst<(outs), (ins Int1Regs:$a, brtarget:$target),
                          "@!$a bra \t$target;",
                           []>;

   let isBranch=1, isBarrier=1 in
      def GOTO : NVPTXInst<(outs), (ins brtarget:$target),
                        "bra.uni \t$target;",
                  [(br bb:$target)]>;
}

def : Pat<(brcond Int32Regs:$a, bb:$target),
          (CBranch (SETP_u32ri Int32Regs:$a, 0, CmpNE), bb:$target)>;

// SelectionDAGBuilder::visitSWitchCase() will invert the condition of a
// conditional branch if
// the target block is the next block so that the code can fall through to the
// target block.
// The invertion is done by 'xor condition, 1', which will be translated to
// (setne condition, -1).
// Since ptx supports '@!pred bra target', we should use it.
def : Pat<(brcond (i1 (setne Int1Regs:$a, -1)), bb:$target),
  (CBranchOther Int1Regs:$a, bb:$target)>;

// Call
def SDT_NVPTXCallSeqStart : SDCallSeqStart<[ SDTCisVT<0, i32> ]>;
def SDT_NVPTXCallSeqEnd   : SDCallSeqEnd<[ SDTCisVT<0, i32>,
                                        SDTCisVT<1, i32> ]>;

def callseq_start : SDNode<"ISD::CALLSEQ_START", SDT_NVPTXCallSeqStart,
                           [SDNPHasChain, SDNPOutGlue, SDNPSideEffect]>;
def callseq_end   : SDNode<"ISD::CALLSEQ_END",   SDT_NVPTXCallSeqEnd,
                           [SDNPHasChain, SDNPOptInGlue, SDNPOutGlue,
                           SDNPSideEffect]>;

def SDT_NVPTXCall : SDTypeProfile<0, 1, [SDTCisVT<0, i32>]>;
def call          : SDNode<"NVPTXISD::CALL", SDT_NVPTXCall,
                           [SDNPHasChain, SDNPOptInGlue, SDNPOutGlue]>;
def calltarget : Operand<i32>;
let isCall=1 in {
   def CALL : NVPTXInst<(outs), (ins calltarget:$dst),
                  "call \t$dst, (1);", []>;
}

def : Pat<(call tglobaladdr:$dst),
          (CALL tglobaladdr:$dst)>;
def : Pat<(call texternalsym:$dst),
          (CALL texternalsym:$dst)>;

// Pseudo instructions.
class Pseudo<dag outs, dag ins, string asmstr, list<dag> pattern>
   : NVPTXInst<outs, ins, asmstr, pattern>;

// @TODO: We use some tricks here to emit curly braces.  Can we clean this up
// a bit without TableGen modifications?
def Callseq_Start : NVPTXInst<(outs), (ins i32imm:$amt),
  "// Callseq Start $amt\n\t{{\n\t.reg .b32 temp_param_reg;\n\t// <end>}}",
                               [(callseq_start timm:$amt)]>;
def Callseq_End : NVPTXInst<(outs), (ins i32imm:$amt1, i32imm:$amt2),
  "\n\t//{{\n\t}}// Callseq End $amt1",
                            [(callseq_end timm:$amt1, timm:$amt2)]>;

// trap instruction

def trapinst : NVPTXInst<(outs), (ins),
                         "trap;",
                         [(trap)]>;

// Call prototype wrapper
def SDTCallPrototype : SDTypeProfile<0, 1, [SDTCisInt<0>]>;
def CallPrototype
  : SDNode<"NVPTXISD::CallPrototype", SDTCallPrototype,
           [SDNPHasChain, SDNPOutGlue, SDNPInGlue, SDNPSideEffect]>;
def ProtoIdent : Operand<i32> {
  let PrintMethod = "printProtoIdent";
}
def CALL_PROTOTYPE
  : NVPTXInst<(outs), (ins ProtoIdent:$ident),
              "$ident", [(CallPrototype (i32 texternalsym:$ident))]>;



include "NVPTXIntrinsics.td"


//-----------------------------------
// Notes
//-----------------------------------
// BSWAP is currently expanded. The following is a more efficient
// - for < sm_20, use vector scalar mov, as tesla support native 16-bit register
// - for sm_20, use pmpt (use vector scalar mov to get the pack and
//   unpack). sm_20 supports native 32-bit register, but not native 16-bit
// register.