Training courses

Kernel and Embedded Linux

Bootlin training courses

Embedded Linux, kernel,
Yocto Project, Buildroot, real-time,
graphics, boot time, debugging...

Bootlin logo

Elixir Cross Referencer

   1
   2
   3
   4
   5
   6
   7
   8
   9
  10
  11
  12
  13
  14
  15
  16
  17
  18
  19
  20
  21
  22
  23
  24
  25
  26
  27
  28
  29
  30
  31
  32
  33
  34
  35
  36
  37
  38
  39
  40
  41
  42
  43
  44
  45
  46
  47
  48
  49
  50
  51
  52
  53
  54
  55
  56
  57
  58
  59
  60
  61
  62
  63
  64
  65
  66
  67
  68
  69
  70
  71
  72
  73
  74
  75
  76
  77
  78
  79
  80
  81
  82
  83
  84
  85
  86
  87
  88
  89
  90
  91
  92
  93
  94
  95
  96
  97
  98
  99
 100
 101
 102
 103
 104
 105
 106
 107
 108
 109
 110
 111
 112
 113
 114
 115
 116
 117
 118
 119
 120
 121
 122
 123
 124
 125
 126
 127
 128
 129
 130
 131
 132
 133
 134
 135
 136
 137
 138
 139
 140
 141
 142
 143
 144
 145
 146
 147
 148
 149
 150
 151
 152
 153
 154
 155
 156
 157
 158
 159
 160
 161
 162
 163
 164
 165
 166
 167
 168
 169
 170
 171
 172
 173
 174
 175
 176
 177
 178
 179
 180
 181
 182
 183
 184
 185
 186
 187
 188
 189
 190
 191
 192
 193
 194
 195
 196
 197
 198
 199
 200
 201
 202
 203
 204
 205
 206
 207
 208
 209
 210
 211
 212
 213
 214
 215
 216
 217
 218
 219
 220
 221
 222
 223
 224
 225
 226
 227
 228
 229
 230
 231
 232
 233
 234
 235
 236
 237
 238
 239
 240
 241
 242
 243
 244
 245
 246
 247
 248
 249
 250
 251
 252
 253
 254
 255
 256
 257
 258
 259
 260
 261
 262
 263
 264
 265
 266
 267
 268
 269
 270
 271
 272
 273
 274
 275
 276
 277
 278
 279
 280
 281
 282
 283
 284
 285
 286
 287
 288
 289
 290
 291
 292
 293
 294
 295
 296
 297
 298
 299
 300
 301
 302
 303
 304
 305
 306
 307
 308
 309
 310
 311
 312
 313
 314
 315
 316
 317
 318
 319
 320
 321
 322
 323
 324
 325
 326
 327
 328
 329
 330
 331
 332
 333
 334
 335
 336
 337
 338
 339
 340
 341
 342
 343
 344
 345
 346
 347
 348
 349
 350
 351
 352
 353
 354
 355
 356
 357
 358
 359
 360
 361
 362
 363
 364
 365
 366
 367
 368
 369
 370
 371
 372
 373
 374
 375
 376
 377
 378
 379
 380
 381
 382
 383
 384
 385
 386
 387
 388
 389
 390
 391
 392
 393
 394
 395
 396
 397
 398
 399
 400
 401
 402
 403
 404
 405
 406
 407
 408
 409
 410
 411
 412
 413
 414
 415
 416
 417
 418
 419
 420
 421
 422
 423
 424
 425
 426
 427
 428
 429
 430
 431
 432
 433
 434
 435
 436
 437
 438
 439
 440
 441
 442
 443
 444
 445
 446
 447
 448
 449
 450
 451
 452
 453
 454
 455
 456
 457
 458
 459
 460
 461
 462
 463
 464
 465
 466
 467
 468
 469
 470
 471
 472
 473
 474
 475
 476
 477
 478
 479
 480
 481
 482
 483
 484
 485
 486
 487
 488
 489
 490
 491
 492
 493
 494
 495
 496
 497
 498
 499
 500
 501
 502
 503
 504
 505
 506
 507
 508
 509
 510
 511
 512
 513
 514
 515
 516
 517
 518
 519
 520
 521
 522
 523
 524
 525
 526
 527
 528
 529
 530
 531
 532
 533
 534
 535
 536
 537
 538
 539
 540
 541
 542
 543
 544
 545
 546
 547
 548
 549
 550
 551
 552
 553
 554
 555
 556
 557
 558
 559
 560
 561
 562
 563
 564
 565
 566
 567
 568
 569
 570
 571
 572
 573
 574
 575
 576
 577
 578
 579
 580
 581
 582
 583
 584
 585
 586
 587
 588
 589
 590
 591
 592
 593
 594
 595
 596
 597
 598
 599
 600
 601
 602
 603
 604
 605
 606
 607
 608
 609
 610
 611
 612
 613
 614
 615
 616
 617
 618
 619
 620
 621
 622
 623
 624
 625
 626
 627
 628
 629
 630
 631
 632
 633
 634
 635
 636
 637
 638
 639
 640
 641
 642
 643
 644
 645
 646
 647
 648
 649
 650
 651
 652
 653
 654
 655
 656
 657
 658
 659
 660
 661
 662
 663
 664
 665
 666
 667
 668
 669
 670
 671
 672
 673
 674
 675
 676
 677
 678
 679
 680
 681
 682
 683
 684
 685
 686
 687
 688
 689
 690
 691
 692
 693
 694
 695
 696
 697
 698
 699
 700
 701
 702
 703
 704
 705
 706
 707
 708
 709
 710
 711
 712
 713
 714
 715
 716
 717
 718
 719
 720
 721
 722
 723
 724
 725
 726
 727
 728
 729
 730
 731
 732
 733
 734
 735
 736
 737
 738
 739
 740
 741
 742
 743
 744
 745
 746
 747
 748
 749
 750
 751
 752
 753
 754
 755
 756
 757
 758
 759
 760
 761
 762
 763
 764
 765
 766
 767
 768
 769
 770
 771
 772
 773
 774
 775
 776
 777
 778
 779
 780
 781
 782
 783
 784
 785
 786
 787
 788
 789
 790
 791
 792
 793
 794
 795
 796
 797
 798
 799
 800
 801
 802
 803
 804
 805
 806
 807
 808
 809
 810
 811
 812
 813
 814
 815
 816
 817
 818
 819
 820
 821
 822
 823
 824
 825
 826
 827
 828
 829
 830
 831
 832
 833
 834
 835
 836
 837
 838
 839
 840
 841
 842
 843
 844
 845
 846
 847
 848
 849
 850
 851
 852
 853
 854
 855
 856
 857
 858
 859
 860
 861
 862
 863
 864
 865
 866
 867
 868
 869
 870
 871
 872
 873
 874
 875
 876
 877
 878
 879
 880
 881
 882
 883
 884
 885
 886
 887
 888
 889
 890
 891
 892
 893
 894
 895
 896
 897
 898
 899
 900
 901
 902
 903
 904
 905
 906
 907
 908
 909
 910
 911
 912
 913
 914
 915
 916
 917
 918
 919
 920
 921
 922
 923
 924
 925
 926
 927
 928
 929
 930
 931
 932
 933
 934
 935
 936
 937
 938
 939
 940
 941
 942
 943
 944
 945
 946
 947
 948
 949
 950
 951
 952
 953
 954
 955
 956
 957
 958
 959
 960
 961
 962
 963
 964
 965
 966
 967
 968
 969
 970
 971
 972
 973
 974
 975
 976
 977
 978
 979
 980
 981
 982
 983
 984
 985
 986
 987
 988
 989
 990
 991
 992
 993
 994
 995
 996
 997
 998
 999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532
3533
3534
3535
3536
3537
3538
3539
3540
3541
3542
3543
3544
3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
3556
3557
3558
3559
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
3570
3571
3572
3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586
3587
3588
3589
3590
3591
3592
3593
3594
3595
3596
3597
3598
3599
3600
3601
3602
3603
3604
3605
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
3617
3618
3619
3620
3621
3622
3623
3624
3625
3626
3627
3628
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
3644
3645
3646
3647
3648
3649
3650
3651
3652
3653
3654
3655
3656
3657
3658
3659
3660
3661
3662
3663
3664
3665
3666
3667
3668
3669
3670
3671
3672
3673
3674
3675
3676
3677
3678
3679
3680
3681
3682
3683
3684
3685
3686
3687
3688
3689
3690
3691
3692
3693
3694
3695
3696
3697
3698
3699
3700
3701
3702
3703
3704
3705
3706
3707
3708
3709
3710
3711
3712
3713
3714
3715
3716
3717
3718
3719
3720
3721
3722
3723
3724
3725
3726
3727
3728
3729
3730
3731
3732
3733
3734
3735
3736
3737
3738
3739
3740
3741
3742
3743
3744
3745
3746
3747
3748
3749
3750
3751
3752
3753
3754
3755
3756
3757
3758
3759
3760
3761
3762
3763
3764
3765
3766
3767
3768
3769
3770
3771
3772
3773
3774
3775
3776
3777
3778
3779
3780
3781
3782
3783
3784
3785
3786
3787
3788
3789
3790
3791
3792
3793
3794
3795
3796
3797
3798
3799
3800
3801
3802
3803
3804
3805
3806
3807
3808
3809
3810
3811
3812
3813
3814
3815
3816
3817
3818
3819
3820
3821
3822
3823
3824
3825
3826
3827
3828
3829
3830
3831
3832
3833
3834
3835
3836
3837
3838
3839
3840
3841
3842
3843
3844
3845
3846
3847
3848
3849
3850
3851
3852
3853
3854
3855
3856
3857
3858
3859
3860
3861
3862
3863
3864
3865
3866
3867
3868
3869
3870
3871
3872
3873
3874
3875
3876
3877
3878
3879
3880
3881
3882
3883
3884
3885
3886
3887
3888
3889
3890
3891
3892
3893
3894
3895
3896
3897
3898
3899
3900
3901
3902
3903
3904
3905
3906
3907
3908
3909
3910
3911
3912
3913
3914
3915
3916
3917
3918
3919
3920
3921
3922
3923
3924
3925
3926
3927
3928
3929
3930
3931
3932
3933
3934
3935
3936
3937
3938
3939
3940
3941
3942
3943
3944
3945
3946
3947
3948
3949
3950
3951
3952
3953
3954
3955
3956
3957
3958
3959
3960
3961
3962
3963
3964
3965
3966
3967
3968
3969
3970
3971
3972
3973
3974
3975
3976
3977
3978
3979
3980
3981
3982
3983
3984
3985
3986
3987
3988
3989
3990
3991
3992
3993
3994
3995
3996
3997
3998
3999
4000
4001
4002
4003
4004
4005
4006
4007
4008
4009
4010
4011
4012
4013
4014
4015
4016
4017
4018
4019
4020
4021
4022
4023
4024
4025
4026
4027
4028
4029
4030
4031
4032
4033
4034
4035
4036
4037
4038
4039
4040
4041
4042
4043
4044
4045
4046
4047
4048
4049
4050
4051
4052
4053
4054
4055
4056
4057
4058
4059
4060
4061
4062
4063
4064
4065
4066
4067
4068
4069
4070
4071
4072
4073
4074
4075
4076
4077
4078
4079
4080
4081
4082
4083
4084
4085
4086
4087
4088
4089
4090
4091
4092
4093
4094
4095
4096
4097
4098
4099
4100
4101
4102
4103
4104
4105
4106
4107
4108
4109
4110
4111
4112
4113
4114
4115
4116
4117
4118
4119
4120
4121
4122
4123
4124
4125
4126
4127
4128
4129
4130
4131
4132
4133
4134
4135
4136
4137
4138
4139
4140
4141
4142
4143
4144
4145
4146
4147
4148
4149
4150
4151
4152
4153
4154
4155
4156
4157
4158
4159
4160
4161
4162
4163
4164
4165
4166
4167
4168
4169
4170
4171
4172
4173
4174
4175
4176
4177
4178
4179
4180
4181
4182
4183
4184
4185
4186
4187
4188
4189
4190
4191
4192
4193
4194
4195
4196
4197
4198
4199
4200
4201
4202
4203
4204
4205
4206
4207
4208
4209
4210
4211
4212
4213
4214
4215
4216
4217
4218
4219
4220
4221
4222
4223
4224
4225
4226
4227
4228
4229
4230
4231
4232
4233
4234
4235
4236
4237
4238
4239
4240
4241
4242
4243
4244
4245
4246
4247
4248
4249
4250
4251
4252
4253
4254
4255
4256
4257
4258
4259
4260
4261
4262
4263
4264
4265
4266
4267
4268
4269
4270
4271
4272
4273
4274
4275
4276
4277
4278
4279
4280
4281
4282
4283
4284
4285
4286
4287
4288
4289
4290
4291
4292
4293
4294
4295
4296
4297
4298
4299
4300
4301
4302
4303
4304
4305
4306
4307
4308
4309
4310
4311
4312
4313
4314
4315
4316
4317
4318
4319
4320
4321
4322
4323
4324
4325
4326
4327
4328
4329
4330
4331
4332
4333
4334
4335
4336
4337
4338
4339
4340
4341
4342
4343
4344
4345
4346
4347
4348
4349
4350
4351
4352
4353
4354
4355
4356
4357
4358
4359
4360
4361
4362
4363
4364
4365
4366
4367
4368
4369
4370
4371
4372
4373
4374
4375
4376
4377
4378
4379
4380
4381
4382
4383
4384
4385
4386
4387
4388
4389
4390
4391
4392
4393
4394
4395
4396
4397
4398
4399
4400
4401
4402
4403
4404
4405
4406
4407
4408
4409
4410
4411
4412
4413
4414
4415
4416
4417
4418
4419
4420
4421
4422
4423
4424
4425
4426
4427
4428
4429
4430
4431
4432
4433
4434
4435
4436
4437
4438
4439
4440
4441
4442
4443
4444
4445
4446
4447
4448
4449
4450
4451
4452
4453
4454
4455
4456
4457
4458
4459
4460
4461
4462
4463
4464
4465
4466
4467
4468
4469
4470
4471
4472
4473
4474
4475
4476
4477
4478
4479
4480
4481
4482
4483
4484
4485
4486
4487
4488
4489
4490
4491
4492
4493
4494
4495
4496
4497
4498
4499
4500
4501
4502
4503
4504
4505
4506
4507
4508
4509
4510
4511
4512
4513
4514
4515
4516
4517
4518
4519
4520
4521
4522
4523
4524
4525
4526
4527
4528
4529
4530
4531
4532
4533
4534
4535
4536
4537
4538
4539
4540
4541
4542
4543
4544
4545
4546
4547
4548
4549
4550
4551
4552
4553
4554
4555
4556
4557
4558
4559
4560
4561
4562
4563
4564
4565
4566
4567
4568
4569
4570
4571
4572
4573
4574
4575
4576
4577
4578
4579
4580
4581
4582
4583
4584
4585
4586
4587
4588
4589
4590
4591
4592
4593
4594
4595
4596
4597
4598
4599
4600
4601
4602
4603
4604
4605
4606
4607
4608
4609
4610
4611
4612
4613
4614
4615
4616
4617
4618
4619
4620
4621
4622
4623
4624
4625
4626
4627
4628
4629
4630
4631
4632
4633
4634
4635
4636
4637
4638
4639
4640
4641
4642
4643
4644
4645
4646
4647
4648
4649
4650
4651
4652
4653
4654
4655
4656
4657
4658
4659
4660
4661
4662
4663
4664
4665
4666
4667
4668
4669
4670
4671
4672
4673
4674
4675
4676
4677
4678
4679
4680
4681
4682
4683
4684
4685
4686
4687
4688
4689
4690
4691
4692
4693
4694
4695
4696
4697
4698
4699
4700
4701
4702
4703
4704
4705
4706
4707
4708
4709
4710
4711
4712
4713
4714
4715
4716
4717
4718
4719
4720
4721
4722
4723
4724
4725
4726
4727
4728
4729
4730
4731
4732
4733
4734
4735
4736
4737
4738
4739
4740
4741
4742
4743
4744
4745
4746
4747
4748
4749
4750
4751
4752
4753
4754
4755
4756
4757
4758
4759
4760
4761
4762
4763
4764
4765
4766
4767
4768
4769
4770
4771
4772
4773
4774
4775
4776
4777
4778
4779
4780
4781
4782
4783
4784
4785
4786
4787
4788
4789
4790
4791
4792
4793
4794
4795
4796
4797
4798
4799
4800
4801
4802
4803
4804
4805
4806
4807
4808
4809
4810
4811
4812
4813
4814
4815
4816
4817
4818
4819
4820
4821
4822
4823
4824
4825
4826
4827
4828
4829
4830
4831
4832
4833
4834
4835
4836
4837
4838
4839
4840
4841
4842
4843
4844
4845
4846
4847
4848
4849
4850
4851
4852
4853
4854
4855
4856
4857
4858
4859
4860
4861
4862
4863
4864
4865
4866
4867
4868
4869
4870
4871
4872
4873
4874
4875
4876
4877
4878
4879
4880
4881
4882
4883
4884
4885
4886
4887
4888
4889
4890
4891
4892
4893
4894
4895
4896
4897
4898
4899
4900
4901
4902
4903
4904
4905
4906
4907
4908
4909
4910
4911
4912
4913
4914
4915
4916
4917
4918
4919
4920
4921
4922
4923
4924
4925
4926
4927
4928
4929
4930
4931
4932
4933
4934
4935
4936
4937
4938
4939
4940
4941
4942
4943
4944
4945
4946
4947
4948
4949
4950
4951
4952
4953
4954
4955
4956
4957
4958
4959
4960
4961
4962
4963
4964
4965
4966
4967
4968
4969
4970
4971
4972
4973
4974
4975
4976
4977
4978
4979
4980
4981
4982
4983
4984
4985
4986
4987
4988
4989
4990
4991
4992
4993
4994
4995
4996
4997
4998
4999
5000
5001
5002
5003
5004
5005
5006
5007
5008
5009
5010
5011
5012
5013
5014
5015
5016
5017
5018
5019
5020
5021
5022
5023
5024
5025
5026
5027
5028
5029
5030
5031
5032
5033
5034
5035
5036
5037
5038
5039
5040
5041
5042
5043
5044
5045
5046
5047
5048
5049
5050
5051
5052
5053
5054
5055
5056
5057
5058
5059
5060
5061
5062
5063
5064
5065
5066
5067
5068
5069
5070
5071
5072
5073
5074
5075
5076
5077
5078
5079
5080
5081
5082
5083
5084
5085
5086
5087
5088
5089
5090
5091
5092
5093
5094
5095
5096
5097
5098
5099
5100
5101
5102
5103
5104
5105
5106
5107
5108
5109
5110
5111
5112
5113
5114
5115
5116
5117
5118
5119
5120
5121
5122
5123
5124
5125
5126
5127
5128
5129
5130
5131
5132
5133
5134
5135
5136
5137
5138
5139
5140
5141
5142
5143
5144
5145
5146
5147
5148
5149
5150
5151
5152
5153
5154
5155
5156
5157
5158
5159
5160
5161
5162
5163
5164
5165
5166
5167
5168
5169
5170
5171
5172
5173
5174
5175
5176
5177
5178
5179
5180
5181
5182
5183
5184
5185
5186
5187
5188
5189
5190
5191
5192
5193
5194
5195
5196
5197
5198
5199
5200
5201
5202
5203
5204
5205
5206
5207
5208
5209
5210
5211
5212
5213
5214
5215
5216
5217
5218
5219
5220
5221
5222
5223
5224
5225
5226
5227
5228
5229
5230
5231
5232
5233
5234
5235
5236
5237
5238
5239
5240
5241
5242
5243
5244
5245
5246
5247
5248
5249
5250
5251
5252
5253
5254
5255
5256
5257
5258
5259
5260
5261
5262
5263
5264
5265
5266
5267
5268
5269
5270
5271
5272
5273
5274
5275
5276
5277
5278
5279
5280
5281
5282
5283
5284
5285
5286
5287
5288
5289
5290
5291
5292
5293
5294
5295
5296
5297
5298
5299
5300
5301
5302
5303
5304
5305
5306
5307
5308
5309
5310
5311
5312
5313
5314
5315
5316
5317
5318
5319
5320
5321
5322
5323
5324
5325
5326
5327
5328
5329
5330
5331
5332
5333
5334
5335
5336
5337
5338
5339
5340
5341
5342
5343
5344
5345
5346
5347
5348
5349
5350
5351
5352
5353
5354
5355
5356
5357
5358
5359
5360
5361
5362
5363
5364
5365
5366
5367
5368
5369
5370
5371
5372
5373
5374
5375
5376
5377
5378
5379
5380
5381
5382
5383
5384
5385
5386
5387
5388
5389
5390
5391
5392
5393
5394
5395
5396
5397
5398
5399
5400
5401
5402
5403
5404
5405
5406
5407
5408
5409
5410
5411
5412
5413
5414
5415
5416
5417
5418
5419
5420
5421
5422
5423
5424
5425
5426
5427
5428
5429
5430
5431
5432
5433
5434
5435
5436
5437
5438
5439
5440
5441
5442
5443
5444
5445
5446
5447
5448
5449
5450
5451
5452
5453
5454
5455
5456
5457
5458
5459
5460
5461
5462
5463
5464
5465
5466
5467
5468
5469
5470
5471
5472
5473
5474
5475
5476
5477
5478
5479
5480
5481
5482
5483
5484
5485
5486
5487
5488
5489
5490
5491
5492
5493
5494
5495
5496
5497
5498
5499
5500
5501
5502
5503
5504
5505
5506
5507
5508
5509
5510
5511
5512
5513
5514
5515
5516
5517
5518
5519
5520
5521
5522
5523
5524
5525
5526
5527
5528
5529
5530
5531
5532
5533
5534
5535
5536
5537
5538
5539
5540
5541
5542
5543
5544
5545
5546
5547
5548
5549
5550
5551
5552
5553
5554
5555
5556
5557
5558
5559
5560
5561
5562
5563
5564
5565
5566
5567
5568
5569
5570
5571
5572
5573
5574
5575
5576
5577
5578
5579
5580
5581
5582
5583
5584
5585
5586
5587
5588
5589
5590
5591
5592
5593
5594
5595
5596
5597
5598
5599
5600
5601
5602
5603
5604
5605
5606
5607
5608
5609
5610
5611
5612
5613
5614
5615
5616
5617
5618
5619
5620
5621
5622
5623
5624
5625
5626
5627
5628
5629
5630
5631
5632
5633
5634
5635
5636
5637
5638
5639
5640
5641
5642
5643
5644
5645
5646
5647
5648
5649
5650
5651
5652
5653
5654
5655
5656
5657
5658
5659
5660
5661
5662
5663
5664
5665
5666
5667
5668
5669
5670
5671
5672
5673
5674
5675
5676
5677
5678
5679
5680
5681
5682
5683
5684
5685
5686
5687
5688
5689
5690
5691
5692
5693
5694
5695
5696
5697
5698
5699
5700
5701
5702
5703
5704
5705
5706
5707
5708
5709
5710
5711
5712
5713
5714
5715
5716
5717
5718
5719
5720
5721
5722
5723
5724
5725
5726
5727
5728
5729
5730
5731
5732
5733
5734
5735
5736
5737
5738
5739
5740
5741
5742
5743
5744
5745
5746
5747
5748
5749
5750
5751
5752
5753
5754
5755
5756
5757
5758
5759
5760
5761
5762
5763
5764
5765
5766
5767
5768
5769
5770
5771
5772
5773
5774
5775
5776
5777
5778
5779
5780
5781
5782
5783
5784
5785
5786
5787
5788
5789
5790
5791
5792
5793
5794
5795
5796
5797
5798
5799
5800
5801
5802
5803
5804
5805
5806
5807
5808
5809
5810
5811
5812
5813
5814
5815
5816
5817
5818
5819
5820
5821
5822
5823
5824
5825
5826
5827
5828
5829
5830
5831
5832
5833
5834
5835
5836
5837
5838
5839
5840
5841
5842
5843
5844
5845
5846
5847
5848
5849
5850
5851
5852
5853
5854
5855
5856
5857
5858
5859
5860
5861
5862
5863
5864
5865
5866
5867
5868
5869
5870
5871
5872
5873
5874
5875
5876
5877
5878
5879
5880
5881
5882
5883
5884
5885
5886
5887
5888
5889
5890
5891
5892
5893
5894
5895
5896
5897
5898
5899
5900
5901
5902
5903
5904
5905
5906
5907
5908
5909
5910
5911
5912
5913
5914
5915
5916
5917
5918
5919
5920
5921
5922
5923
5924
5925
5926
5927
5928
5929
5930
5931
5932
5933
5934
5935
5936
5937
5938
5939
5940
5941
5942
5943
5944
5945
5946
5947
5948
5949
5950
5951
5952
5953
5954
5955
5956
5957
5958
5959
5960
5961
5962
5963
5964
5965
5966
5967
5968
5969
5970
5971
5972
5973
5974
5975
5976
5977
5978
5979
5980
5981
5982
5983
5984
5985
5986
5987
5988
5989
5990
5991
5992
5993
5994
5995
5996
5997
5998
5999
6000
6001
6002
6003
6004
6005
6006
6007
6008
6009
6010
6011
6012
6013
6014
6015
6016
6017
6018
6019
6020
6021
6022
6023
6024
6025
6026
6027
6028
6029
6030
6031
6032
6033
6034
6035
6036
6037
6038
6039
6040
6041
6042
6043
6044
6045
6046
6047
6048
6049
6050
6051
6052
6053
6054
6055
6056
6057
6058
6059
6060
6061
6062
6063
6064
6065
6066
6067
6068
6069
6070
6071
6072
6073
6074
6075
6076
6077
6078
6079
6080
6081
6082
6083
6084
6085
6086
6087
6088
6089
6090
6091
6092
6093
6094
6095
6096
6097
6098
6099
6100
6101
6102
6103
6104
6105
6106
6107
6108
6109
6110
6111
6112
6113
6114
6115
6116
6117
6118
6119
6120
6121
6122
6123
6124
6125
6126
6127
6128
6129
6130
6131
6132
6133
6134
6135
6136
6137
6138
6139
6140
6141
6142
6143
6144
6145
6146
6147
6148
6149
6150
6151
6152
6153
6154
6155
6156
6157
6158
6159
6160
6161
6162
6163
6164
6165
6166
6167
6168
6169
6170
6171
6172
6173
6174
6175
6176
6177
6178
6179
6180
6181
6182
6183
6184
6185
6186
6187
6188
6189
6190
6191
6192
6193
6194
6195
6196
6197
6198
6199
6200
6201
6202
6203
6204
6205
6206
6207
6208
6209
6210
6211
6212
6213
6214
6215
6216
6217
6218
6219
6220
6221
6222
6223
6224
6225
6226
6227
6228
6229
6230
6231
6232
6233
6234
6235
6236
6237
6238
6239
6240
6241
6242
6243
6244
6245
6246
6247
6248
6249
6250
6251
6252
6253
6254
6255
6256
6257
6258
6259
6260
6261
6262
6263
6264
6265
6266
6267
6268
6269
6270
6271
6272
6273
6274
6275
6276
6277
6278
6279
6280
6281
6282
6283
6284
6285
6286
6287
6288
6289
6290
6291
6292
6293
6294
6295
6296
6297
6298
6299
6300
6301
6302
6303
6304
6305
6306
6307
6308
6309
6310
6311
6312
6313
6314
6315
6316
6317
6318
6319
6320
6321
6322
6323
6324
6325
6326
6327
6328
6329
6330
6331
6332
6333
6334
6335
6336
6337
6338
6339
6340
6341
6342
6343
6344
6345
6346
6347
6348
6349
6350
6351
6352
6353
6354
6355
6356
6357
6358
6359
6360
6361
6362
6363
6364
6365
6366
6367
6368
6369
6370
6371
6372
6373
6374
6375
6376
6377
6378
6379
6380
6381
6382
6383
6384
6385
6386
6387
6388
6389
6390
6391
6392
6393
6394
6395
6396
6397
6398
6399
6400
6401
6402
6403
6404
6405
6406
6407
6408
6409
6410
6411
6412
6413
6414
6415
6416
6417
6418
6419
6420
6421
6422
6423
6424
6425
6426
6427
6428
6429
6430
6431
6432
6433
6434
6435
6436
6437
6438
6439
6440
6441
6442
6443
6444
6445
6446
6447
6448
6449
6450
6451
6452
6453
6454
6455
6456
6457
6458
6459
6460
6461
6462
6463
6464
6465
6466
6467
6468
6469
6470
6471
6472
6473
6474
6475
6476
6477
6478
6479
6480
6481
6482
6483
6484
6485
6486
6487
6488
6489
6490
6491
6492
6493
6494
6495
6496
6497
6498
6499
6500
6501
6502
6503
6504
6505
6506
6507
6508
6509
6510
6511
6512
6513
6514
6515
6516
6517
6518
6519
6520
6521
6522
6523
6524
6525
6526
6527
6528
6529
6530
6531
6532
6533
6534
6535
6536
6537
6538
6539
6540
6541
6542
6543
6544
6545
6546
6547
6548
6549
6550
6551
6552
6553
6554
6555
6556
6557
6558
6559
6560
6561
6562
6563
6564
6565
6566
6567
6568
6569
6570
6571
6572
6573
6574
6575
6576
6577
6578
6579
6580
6581
6582
6583
6584
6585
6586
6587
6588
6589
6590
6591
6592
6593
6594
6595
6596
6597
6598
6599
6600
6601
6602
6603
6604
6605
6606
6607
6608
6609
6610
6611
6612
6613
6614
6615
6616
6617
6618
6619
6620
6621
6622
6623
6624
6625
6626
6627
6628
6629
6630
6631
6632
6633
6634
6635
6636
6637
6638
6639
6640
6641
6642
6643
6644
6645
6646
6647
6648
6649
6650
6651
6652
6653
6654
6655
6656
6657
6658
6659
6660
6661
6662
6663
6664
6665
6666
6667
6668
6669
6670
6671
6672
6673
6674
6675
6676
6677
6678
6679
6680
6681
6682
6683
6684
6685
6686
6687
6688
6689
6690
6691
6692
6693
6694
6695
6696
6697
6698
6699
6700
6701
6702
6703
6704
6705
6706
6707
6708
6709
6710
6711
6712
6713
6714
6715
6716
6717
6718
6719
6720
6721
6722
6723
6724
6725
6726
6727
6728
6729
6730
6731
6732
6733
6734
6735
6736
6737
6738
6739
6740
6741
6742
6743
6744
6745
6746
6747
6748
6749
6750
6751
6752
6753
6754
6755
6756
6757
6758
6759
6760
6761
6762
6763
6764
6765
6766
6767
6768
6769
6770
6771
6772
6773
6774
6775
6776
6777
6778
6779
6780
6781
6782
6783
6784
6785
6786
6787
6788
6789
6790
6791
6792
6793
6794
6795
6796
6797
6798
6799
6800
6801
6802
6803
6804
6805
6806
6807
6808
6809
6810
6811
6812
6813
6814
6815
6816
6817
6818
6819
6820
6821
6822
6823
6824
6825
6826
6827
6828
6829
6830
6831
6832
6833
6834
6835
6836
6837
6838
6839
6840
6841
6842
6843
6844
6845
6846
6847
6848
6849
6850
6851
6852
6853
6854
6855
6856
6857
6858
6859
6860
6861
6862
6863
6864
6865
6866
6867
6868
6869
6870
6871
6872
6873
6874
6875
6876
6877
6878
6879
6880
6881
6882
6883
6884
6885
6886
6887
6888
6889
6890
6891
6892
6893
6894
6895
6896
6897
6898
6899
6900
6901
6902
6903
6904
6905
6906
6907
6908
6909
6910
6911
6912
6913
6914
6915
6916
6917
6918
6919
6920
6921
6922
6923
6924
6925
6926
6927
6928
6929
6930
6931
6932
6933
6934
6935
6936
6937
6938
6939
6940
6941
6942
6943
6944
6945
6946
6947
6948
6949
6950
6951
6952
6953
6954
6955
6956
6957
6958
6959
6960
6961
6962
6963
6964
6965
6966
6967
6968
6969
6970
6971
6972
6973
6974
6975
6976
6977
6978
6979
6980
6981
6982
6983
6984
6985
6986
6987
6988
6989
6990
6991
6992
6993
6994
6995
6996
6997
6998
6999
7000
7001
7002
7003
7004
7005
7006
7007
7008
7009
7010
7011
7012
7013
7014
7015
7016
7017
7018
7019
7020
7021
7022
7023
7024
7025
7026
7027
7028
7029
7030
7031
7032
7033
7034
7035
7036
7037
7038
7039
7040
7041
7042
7043
7044
7045
7046
7047
7048
7049
7050
7051
7052
7053
7054
7055
7056
7057
7058
7059
7060
7061
7062
7063
7064
7065
7066
7067
7068
7069
7070
7071
7072
7073
7074
7075
7076
7077
7078
7079
7080
7081
7082
7083
7084
7085
7086
7087
7088
7089
7090
7091
7092
7093
7094
7095
7096
7097
7098
7099
7100
7101
7102
7103
7104
7105
7106
7107
7108
7109
7110
7111
7112
7113
7114
7115
7116
7117
7118
7119
7120
7121
7122
7123
7124
7125
7126
7127
7128
7129
7130
7131
7132
7133
7134
7135
7136
7137
7138
7139
7140
7141
7142
7143
7144
7145
7146
7147
7148
7149
7150
7151
7152
7153
7154
7155
7156
7157
7158
7159
7160
7161
7162
7163
7164
7165
7166
7167
7168
7169
7170
7171
7172
7173
7174
7175
7176
7177
7178
7179
7180
7181
7182
7183
7184
7185
7186
7187
7188
7189
7190
7191
7192
7193
7194
7195
7196
7197
7198
7199
7200
7201
7202
7203
7204
7205
7206
7207
7208
7209
7210
7211
7212
7213
7214
7215
7216
7217
7218
7219
7220
7221
7222
7223
7224
7225
7226
7227
7228
7229
7230
7231
7232
7233
7234
7235
7236
7237
7238
7239
7240
7241
7242
7243
7244
7245
7246
7247
7248
7249
7250
7251
7252
7253
7254
7255
7256
7257
7258
7259
7260
7261
7262
7263
7264
7265
7266
7267
7268
7269
7270
7271
7272
7273
7274
7275
7276
7277
7278
7279
7280
7281
7282
7283
7284
7285
7286
7287
7288
7289
7290
7291
7292
7293
7294
7295
7296
7297
7298
7299
7300
7301
7302
7303
7304
7305
7306
7307
7308
7309
7310
7311
7312
7313
7314
7315
7316
7317
7318
7319
7320
7321
7322
7323
7324
7325
7326
7327
7328
7329
7330
7331
7332
7333
7334
7335
7336
7337
7338
7339
7340
7341
7342
7343
7344
7345
7346
7347
7348
7349
7350
7351
7352
7353
7354
7355
7356
7357
7358
7359
7360
7361
7362
7363
7364
7365
7366
7367
7368
7369
7370
7371
7372
7373
7374
7375
7376
7377
7378
7379
7380
7381
7382
7383
7384
7385
7386
7387
7388
7389
7390
7391
7392
7393
7394
7395
7396
7397
7398
7399
7400
7401
7402
7403
7404
7405
7406
7407
7408
7409
7410
7411
7412
7413
7414
7415
7416
7417
7418
7419
7420
7421
7422
7423
7424
7425
7426
7427
7428
7429
7430
7431
7432
7433
7434
7435
7436
7437
7438
7439
7440
7441
7442
7443
7444
7445
7446
7447
7448
7449
7450
7451
7452
7453
7454
7455
7456
7457
7458
7459
7460
7461
7462
7463
7464
7465
7466
7467
7468
7469
7470
7471
7472
7473
7474
7475
7476
7477
7478
7479
7480
7481
7482
7483
7484
7485
7486
7487
7488
7489
7490
7491
7492
7493
7494
7495
7496
7497
7498
7499
7500
7501
7502
7503
7504
7505
7506
7507
7508
7509
7510
7511
7512
7513
7514
7515
7516
7517
7518
7519
7520
7521
7522
7523
7524
7525
7526
7527
7528
7529
7530
7531
7532
7533
7534
7535
7536
7537
7538
7539
7540
7541
7542
7543
7544
7545
7546
7547
7548
7549
7550
7551
7552
7553
7554
7555
7556
7557
7558
7559
7560
7561
7562
7563
7564
7565
7566
7567
7568
7569
7570
7571
7572
7573
7574
7575
7576
7577
7578
7579
7580
7581
7582
7583
7584
7585
7586
7587
7588
7589
7590
7591
7592
7593
7594
7595
7596
7597
7598
7599
7600
7601
7602
7603
7604
7605
7606
7607
7608
7609
7610
7611
7612
7613
7614
7615
7616
7617
7618
7619
7620
7621
7622
7623
7624
7625
7626
7627
7628
7629
7630
7631
7632
7633
7634
7635
7636
7637
7638
7639
7640
7641
7642
7643
7644
7645
7646
7647
7648
7649
7650
7651
7652
7653
7654
7655
7656
7657
7658
7659
7660
7661
7662
7663
7664
7665
7666
7667
7668
7669
7670
7671
7672
7673
7674
7675
7676
7677
7678
7679
7680
7681
7682
7683
7684
7685
7686
7687
7688
7689
7690
7691
7692
7693
7694
7695
7696
7697
7698
7699
7700
7701
7702
7703
7704
7705
7706
7707
7708
7709
7710
7711
7712
7713
7714
7715
7716
7717
7718
7719
7720
7721
7722
7723
7724
7725
7726
7727
7728
7729
7730
7731
7732
7733
7734
7735
7736
7737
7738
7739
7740
7741
7742
7743
7744
7745
7746
7747
7748
7749
7750
7751
7752
7753
7754
7755
7756
7757
7758
7759
7760
7761
7762
7763
7764
7765
7766
7767
7768
7769
7770
7771
7772
7773
7774
7775
7776
7777
7778
7779
7780
7781
7782
7783
7784
7785
7786
7787
7788
7789
7790
7791
7792
7793
7794
7795
7796
7797
7798
7799
7800
7801
7802
7803
7804
7805
7806
7807
7808
7809
7810
7811
7812
7813
7814
7815
7816
7817
7818
7819
7820
7821
7822
7823
7824
7825
7826
7827
7828
7829
7830
7831
7832
7833
7834
7835
7836
7837
7838
7839
7840
7841
7842
7843
7844
7845
7846
7847
7848
7849
7850
7851
7852
7853
7854
7855
7856
7857
7858
7859
7860
7861
7862
7863
7864
7865
7866
7867
7868
7869
7870
7871
7872
7873
7874
7875
7876
7877
7878
7879
7880
7881
7882
7883
7884
7885
7886
7887
7888
7889
7890
7891
7892
7893
7894
7895
7896
7897
7898
7899
7900
7901
7902
7903
7904
7905
7906
7907
7908
7909
7910
7911
7912
7913
7914
7915
7916
7917
7918
7919
7920
7921
7922
7923
7924
7925
7926
7927
7928
7929
7930
7931
7932
7933
7934
7935
7936
7937
7938
7939
7940
7941
7942
7943
7944
7945
7946
7947
7948
7949
7950
7951
7952
7953
7954
7955
7956
7957
7958
7959
7960
7961
7962
7963
7964
7965
7966
7967
7968
7969
7970
7971
7972
7973
7974
7975
7976
7977
7978
7979
7980
7981
7982
7983
7984
7985
7986
7987
7988
7989
7990
7991
7992
7993
7994
7995
7996
7997
7998
7999
8000
8001
8002
8003
8004
8005
8006
8007
8008
8009
8010
8011
8012
8013
8014
8015
8016
8017
8018
8019
8020
8021
8022
8023
8024
8025
8026
8027
8028
8029
8030
8031
8032
8033
8034
8035
8036
8037
8038
8039
8040
8041
8042
8043
8044
8045
8046
8047
8048
8049
8050
8051
8052
8053
8054
8055
8056
8057
8058
8059
8060
8061
8062
8063
8064
8065
8066
8067
8068
8069
8070
8071
8072
8073
8074
8075
8076
8077
8078
8079
8080
8081
8082
8083
8084
8085
8086
8087
8088
8089
8090
8091
8092
8093
8094
8095
8096
8097
8098
8099
8100
8101
8102
8103
8104
8105
8106
8107
8108
8109
8110
8111
8112
8113
8114
8115
8116
8117
8118
8119
8120
8121
8122
8123
8124
8125
8126
8127
8128
8129
8130
8131
8132
8133
8134
8135
8136
8137
8138
8139
8140
8141
8142
8143
8144
8145
8146
8147
8148
8149
8150
8151
8152
8153
8154
8155
8156
8157
8158
8159
8160
8161
8162
8163
8164
8165
8166
8167
8168
8169
8170
8171
8172
8173
8174
8175
8176
8177
8178
8179
8180
8181
8182
8183
8184
8185
8186
8187
8188
8189
8190
8191
8192
8193
8194
8195
8196
8197
8198
8199
8200
8201
8202
8203
8204
8205
8206
8207
8208
8209
8210
8211
8212
8213
8214
8215
8216
8217
8218
8219
8220
8221
8222
8223
8224
8225
8226
8227
8228
8229
8230
8231
8232
8233
8234
8235
8236
8237
8238
8239
8240
8241
8242
8243
8244
8245
8246
8247
8248
8249
8250
8251
8252
8253
8254
8255
8256
8257
8258
8259
8260
8261
8262
8263
8264
8265
8266
8267
8268
8269
8270
8271
8272
8273
8274
8275
8276
8277
8278
8279
8280
8281
8282
8283
8284
8285
8286
8287
8288
8289
8290
8291
8292
8293
8294
8295
8296
8297
8298
8299
8300
8301
8302
8303
8304
8305
8306
8307
8308
8309
8310
8311
8312
8313
8314
8315
8316
8317
8318
8319
8320
8321
8322
8323
8324
8325
8326
8327
8328
8329
8330
8331
8332
8333
8334
8335
8336
8337
8338
8339
8340
8341
8342
8343
8344
8345
8346
8347
8348
8349
8350
8351
8352
8353
8354
8355
8356
8357
8358
8359
8360
8361
8362
8363
8364
8365
8366
8367
8368
8369
8370
8371
8372
8373
8374
8375
8376
8377
8378
8379
8380
8381
8382
8383
8384
8385
8386
8387
8388
8389
8390
8391
8392
8393
8394
8395
8396
8397
8398
8399
8400
8401
8402
8403
8404
8405
8406
8407
8408
8409
8410
8411
8412
8413
8414
8415
8416
8417
8418
8419
8420
8421
8422
//===-- SystemZISelLowering.cpp - SystemZ DAG lowering implementation -----===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
//
// This file implements the SystemZTargetLowering class.
//
//===----------------------------------------------------------------------===//

#include "SystemZISelLowering.h"
#include "SystemZCallingConv.h"
#include "SystemZConstantPoolValue.h"
#include "SystemZMachineFunctionInfo.h"
#include "SystemZTargetMachine.h"
#include "llvm/CodeGen/CallingConvLower.h"
#include "llvm/CodeGen/MachineInstrBuilder.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
#include "llvm/CodeGen/TargetLoweringObjectFileImpl.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/IR/Intrinsics.h"
#include "llvm/IR/IntrinsicsS390.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/KnownBits.h"
#include <cctype>

using namespace llvm;

#define DEBUG_TYPE "systemz-lower"

namespace {
// Represents information about a comparison.
struct Comparison {
  Comparison(SDValue Op0In, SDValue Op1In, SDValue ChainIn)
    : Op0(Op0In), Op1(Op1In), Chain(ChainIn),
      Opcode(0), ICmpType(0), CCValid(0), CCMask(0) {}

  // The operands to the comparison.
  SDValue Op0, Op1;

  // Chain if this is a strict floating-point comparison.
  SDValue Chain;

  // The opcode that should be used to compare Op0 and Op1.
  unsigned Opcode;

  // A SystemZICMP value.  Only used for integer comparisons.
  unsigned ICmpType;

  // The mask of CC values that Opcode can produce.
  unsigned CCValid;

  // The mask of CC values for which the original condition is true.
  unsigned CCMask;
};
} // end anonymous namespace

// Classify VT as either 32 or 64 bit.
static bool is32Bit(EVT VT) {
  switch (VT.getSimpleVT().SimpleTy) {
  case MVT::i32:
    return true;
  case MVT::i64:
    return false;
  default:
    llvm_unreachable("Unsupported type");
  }
}

// Return a version of MachineOperand that can be safely used before the
// final use.
static MachineOperand earlyUseOperand(MachineOperand Op) {
  if (Op.isReg())
    Op.setIsKill(false);
  return Op;
}

SystemZTargetLowering::SystemZTargetLowering(const TargetMachine &TM,
                                             const SystemZSubtarget &STI)
    : TargetLowering(TM), Subtarget(STI) {
  MVT PtrVT = MVT::getIntegerVT(8 * TM.getPointerSize(0));

  // Set up the register classes.
  if (Subtarget.hasHighWord())
    addRegisterClass(MVT::i32, &SystemZ::GRX32BitRegClass);
  else
    addRegisterClass(MVT::i32, &SystemZ::GR32BitRegClass);
  addRegisterClass(MVT::i64, &SystemZ::GR64BitRegClass);
  if (!useSoftFloat()) {
    if (Subtarget.hasVector()) {
      addRegisterClass(MVT::f32, &SystemZ::VR32BitRegClass);
      addRegisterClass(MVT::f64, &SystemZ::VR64BitRegClass);
    } else {
      addRegisterClass(MVT::f32, &SystemZ::FP32BitRegClass);
      addRegisterClass(MVT::f64, &SystemZ::FP64BitRegClass);
    }
    if (Subtarget.hasVectorEnhancements1())
      addRegisterClass(MVT::f128, &SystemZ::VR128BitRegClass);
    else
      addRegisterClass(MVT::f128, &SystemZ::FP128BitRegClass);

    if (Subtarget.hasVector()) {
      addRegisterClass(MVT::v16i8, &SystemZ::VR128BitRegClass);
      addRegisterClass(MVT::v8i16, &SystemZ::VR128BitRegClass);
      addRegisterClass(MVT::v4i32, &SystemZ::VR128BitRegClass);
      addRegisterClass(MVT::v2i64, &SystemZ::VR128BitRegClass);
      addRegisterClass(MVT::v4f32, &SystemZ::VR128BitRegClass);
      addRegisterClass(MVT::v2f64, &SystemZ::VR128BitRegClass);
    }
  }

  // Compute derived properties from the register classes
  computeRegisterProperties(Subtarget.getRegisterInfo());

  // Set up special registers.
  setStackPointerRegisterToSaveRestore(SystemZ::R15D);

  // TODO: It may be better to default to latency-oriented scheduling, however
  // LLVM's current latency-oriented scheduler can't handle physreg definitions
  // such as SystemZ has with CC, so set this to the register-pressure
  // scheduler, because it can.
  setSchedulingPreference(Sched::RegPressure);

  setBooleanContents(ZeroOrOneBooleanContent);
  setBooleanVectorContents(ZeroOrNegativeOneBooleanContent);

  // Instructions are strings of 2-byte aligned 2-byte values.
  setMinFunctionAlignment(Align(2));
  // For performance reasons we prefer 16-byte alignment.
  setPrefFunctionAlignment(Align(16));

  // Handle operations that are handled in a similar way for all types.
  for (unsigned I = MVT::FIRST_INTEGER_VALUETYPE;
       I <= MVT::LAST_FP_VALUETYPE;
       ++I) {
    MVT VT = MVT::SimpleValueType(I);
    if (isTypeLegal(VT)) {
      // Lower SET_CC into an IPM-based sequence.
      setOperationAction(ISD::SETCC, VT, Custom);
      setOperationAction(ISD::STRICT_FSETCC, VT, Custom);
      setOperationAction(ISD::STRICT_FSETCCS, VT, Custom);

      // Expand SELECT(C, A, B) into SELECT_CC(X, 0, A, B, NE).
      setOperationAction(ISD::SELECT, VT, Expand);

      // Lower SELECT_CC and BR_CC into separate comparisons and branches.
      setOperationAction(ISD::SELECT_CC, VT, Custom);
      setOperationAction(ISD::BR_CC,     VT, Custom);
    }
  }

  // Expand jump table branches as address arithmetic followed by an
  // indirect jump.
  setOperationAction(ISD::BR_JT, MVT::Other, Expand);

  // Expand BRCOND into a BR_CC (see above).
  setOperationAction(ISD::BRCOND, MVT::Other, Expand);

  // Handle integer types.
  for (unsigned I = MVT::FIRST_INTEGER_VALUETYPE;
       I <= MVT::LAST_INTEGER_VALUETYPE;
       ++I) {
    MVT VT = MVT::SimpleValueType(I);
    if (isTypeLegal(VT)) {
      // Expand individual DIV and REMs into DIVREMs.
      setOperationAction(ISD::SDIV, VT, Expand);
      setOperationAction(ISD::UDIV, VT, Expand);
      setOperationAction(ISD::SREM, VT, Expand);
      setOperationAction(ISD::UREM, VT, Expand);
      setOperationAction(ISD::SDIVREM, VT, Custom);
      setOperationAction(ISD::UDIVREM, VT, Custom);

      // Support addition/subtraction with overflow.
      setOperationAction(ISD::SADDO, VT, Custom);
      setOperationAction(ISD::SSUBO, VT, Custom);

      // Support addition/subtraction with carry.
      setOperationAction(ISD::UADDO, VT, Custom);
      setOperationAction(ISD::USUBO, VT, Custom);

      // Support carry in as value rather than glue.
      setOperationAction(ISD::ADDCARRY, VT, Custom);
      setOperationAction(ISD::SUBCARRY, VT, Custom);

      // Lower ATOMIC_LOAD and ATOMIC_STORE into normal volatile loads and
      // stores, putting a serialization instruction after the stores.
      setOperationAction(ISD::ATOMIC_LOAD,  VT, Custom);
      setOperationAction(ISD::ATOMIC_STORE, VT, Custom);

      // Lower ATOMIC_LOAD_SUB into ATOMIC_LOAD_ADD if LAA and LAAG are
      // available, or if the operand is constant.
      setOperationAction(ISD::ATOMIC_LOAD_SUB, VT, Custom);

      // Use POPCNT on z196 and above.
      if (Subtarget.hasPopulationCount())
        setOperationAction(ISD::CTPOP, VT, Custom);
      else
        setOperationAction(ISD::CTPOP, VT, Expand);

      // No special instructions for these.
      setOperationAction(ISD::CTTZ,            VT, Expand);
      setOperationAction(ISD::ROTR,            VT, Expand);

      // Use *MUL_LOHI where possible instead of MULH*.
      setOperationAction(ISD::MULHS, VT, Expand);
      setOperationAction(ISD::MULHU, VT, Expand);
      setOperationAction(ISD::SMUL_LOHI, VT, Custom);
      setOperationAction(ISD::UMUL_LOHI, VT, Custom);

      // Only z196 and above have native support for conversions to unsigned.
      // On z10, promoting to i64 doesn't generate an inexact condition for
      // values that are outside the i32 range but in the i64 range, so use
      // the default expansion.
      if (!Subtarget.hasFPExtension())
        setOperationAction(ISD::FP_TO_UINT, VT, Expand);

      // Mirror those settings for STRICT_FP_TO_[SU]INT.  Note that these all
      // default to Expand, so need to be modified to Legal where appropriate.
      setOperationAction(ISD::STRICT_FP_TO_SINT, VT, Legal);
      if (Subtarget.hasFPExtension())
        setOperationAction(ISD::STRICT_FP_TO_UINT, VT, Legal);

      // And similarly for STRICT_[SU]INT_TO_FP.
      setOperationAction(ISD::STRICT_SINT_TO_FP, VT, Legal);
      if (Subtarget.hasFPExtension())
        setOperationAction(ISD::STRICT_UINT_TO_FP, VT, Legal);
    }
  }

  // Type legalization will convert 8- and 16-bit atomic operations into
  // forms that operate on i32s (but still keeping the original memory VT).
  // Lower them into full i32 operations.
  setOperationAction(ISD::ATOMIC_SWAP,      MVT::i32, Custom);
  setOperationAction(ISD::ATOMIC_LOAD_ADD,  MVT::i32, Custom);
  setOperationAction(ISD::ATOMIC_LOAD_SUB,  MVT::i32, Custom);
  setOperationAction(ISD::ATOMIC_LOAD_AND,  MVT::i32, Custom);
  setOperationAction(ISD::ATOMIC_LOAD_OR,   MVT::i32, Custom);
  setOperationAction(ISD::ATOMIC_LOAD_XOR,  MVT::i32, Custom);
  setOperationAction(ISD::ATOMIC_LOAD_NAND, MVT::i32, Custom);
  setOperationAction(ISD::ATOMIC_LOAD_MIN,  MVT::i32, Custom);
  setOperationAction(ISD::ATOMIC_LOAD_MAX,  MVT::i32, Custom);
  setOperationAction(ISD::ATOMIC_LOAD_UMIN, MVT::i32, Custom);
  setOperationAction(ISD::ATOMIC_LOAD_UMAX, MVT::i32, Custom);

  // Even though i128 is not a legal type, we still need to custom lower
  // the atomic operations in order to exploit SystemZ instructions.
  setOperationAction(ISD::ATOMIC_LOAD,     MVT::i128, Custom);
  setOperationAction(ISD::ATOMIC_STORE,    MVT::i128, Custom);

  // We can use the CC result of compare-and-swap to implement
  // the "success" result of ATOMIC_CMP_SWAP_WITH_SUCCESS.
  setOperationAction(ISD::ATOMIC_CMP_SWAP_WITH_SUCCESS, MVT::i32, Custom);
  setOperationAction(ISD::ATOMIC_CMP_SWAP_WITH_SUCCESS, MVT::i64, Custom);
  setOperationAction(ISD::ATOMIC_CMP_SWAP_WITH_SUCCESS, MVT::i128, Custom);

  setOperationAction(ISD::ATOMIC_FENCE, MVT::Other, Custom);

  // Traps are legal, as we will convert them to "j .+2".
  setOperationAction(ISD::TRAP, MVT::Other, Legal);

  // z10 has instructions for signed but not unsigned FP conversion.
  // Handle unsigned 32-bit types as signed 64-bit types.
  if (!Subtarget.hasFPExtension()) {
    setOperationAction(ISD::UINT_TO_FP, MVT::i32, Promote);
    setOperationAction(ISD::UINT_TO_FP, MVT::i64, Expand);
    setOperationAction(ISD::STRICT_UINT_TO_FP, MVT::i32, Promote);
    setOperationAction(ISD::STRICT_UINT_TO_FP, MVT::i64, Expand);
  }

  // We have native support for a 64-bit CTLZ, via FLOGR.
  setOperationAction(ISD::CTLZ, MVT::i32, Promote);
  setOperationAction(ISD::CTLZ_ZERO_UNDEF, MVT::i32, Promote);
  setOperationAction(ISD::CTLZ, MVT::i64, Legal);

  // On z15 we have native support for a 64-bit CTPOP.
  if (Subtarget.hasMiscellaneousExtensions3()) {
    setOperationAction(ISD::CTPOP, MVT::i32, Promote);
    setOperationAction(ISD::CTPOP, MVT::i64, Legal);
  }

  // Give LowerOperation the chance to replace 64-bit ORs with subregs.
  setOperationAction(ISD::OR, MVT::i64, Custom);

  // FIXME: Can we support these natively?
  setOperationAction(ISD::SRL_PARTS, MVT::i64, Expand);
  setOperationAction(ISD::SHL_PARTS, MVT::i64, Expand);
  setOperationAction(ISD::SRA_PARTS, MVT::i64, Expand);

  // We have native instructions for i8, i16 and i32 extensions, but not i1.
  setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::i1, Expand);
  for (MVT VT : MVT::integer_valuetypes()) {
    setLoadExtAction(ISD::SEXTLOAD, VT, MVT::i1, Promote);
    setLoadExtAction(ISD::ZEXTLOAD, VT, MVT::i1, Promote);
    setLoadExtAction(ISD::EXTLOAD,  VT, MVT::i1, Promote);
  }

  // Handle the various types of symbolic address.
  setOperationAction(ISD::ConstantPool,     PtrVT, Custom);
  setOperationAction(ISD::GlobalAddress,    PtrVT, Custom);
  setOperationAction(ISD::GlobalTLSAddress, PtrVT, Custom);
  setOperationAction(ISD::BlockAddress,     PtrVT, Custom);
  setOperationAction(ISD::JumpTable,        PtrVT, Custom);

  // We need to handle dynamic allocations specially because of the
  // 160-byte area at the bottom of the stack.
  setOperationAction(ISD::DYNAMIC_STACKALLOC, PtrVT, Custom);
  setOperationAction(ISD::GET_DYNAMIC_AREA_OFFSET, PtrVT, Custom);

  // Use custom expanders so that we can force the function to use
  // a frame pointer.
  setOperationAction(ISD::STACKSAVE,    MVT::Other, Custom);
  setOperationAction(ISD::STACKRESTORE, MVT::Other, Custom);

  // Handle prefetches with PFD or PFDRL.
  setOperationAction(ISD::PREFETCH, MVT::Other, Custom);

  for (MVT VT : MVT::fixedlen_vector_valuetypes()) {
    // Assume by default that all vector operations need to be expanded.
    for (unsigned Opcode = 0; Opcode < ISD::BUILTIN_OP_END; ++Opcode)
      if (getOperationAction(Opcode, VT) == Legal)
        setOperationAction(Opcode, VT, Expand);

    // Likewise all truncating stores and extending loads.
    for (MVT InnerVT : MVT::fixedlen_vector_valuetypes()) {
      setTruncStoreAction(VT, InnerVT, Expand);
      setLoadExtAction(ISD::SEXTLOAD, VT, InnerVT, Expand);
      setLoadExtAction(ISD::ZEXTLOAD, VT, InnerVT, Expand);
      setLoadExtAction(ISD::EXTLOAD, VT, InnerVT, Expand);
    }

    if (isTypeLegal(VT)) {
      // These operations are legal for anything that can be stored in a
      // vector register, even if there is no native support for the format
      // as such.  In particular, we can do these for v4f32 even though there
      // are no specific instructions for that format.
      setOperationAction(ISD::LOAD, VT, Legal);
      setOperationAction(ISD::STORE, VT, Legal);
      setOperationAction(ISD::VSELECT, VT, Legal);
      setOperationAction(ISD::BITCAST, VT, Legal);
      setOperationAction(ISD::UNDEF, VT, Legal);

      // Likewise, except that we need to replace the nodes with something
      // more specific.
      setOperationAction(ISD::BUILD_VECTOR, VT, Custom);
      setOperationAction(ISD::VECTOR_SHUFFLE, VT, Custom);
    }
  }

  // Handle integer vector types.
  for (MVT VT : MVT::integer_fixedlen_vector_valuetypes()) {
    if (isTypeLegal(VT)) {
      // These operations have direct equivalents.
      setOperationAction(ISD::EXTRACT_VECTOR_ELT, VT, Legal);
      setOperationAction(ISD::INSERT_VECTOR_ELT, VT, Legal);
      setOperationAction(ISD::ADD, VT, Legal);
      setOperationAction(ISD::SUB, VT, Legal);
      if (VT != MVT::v2i64)
        setOperationAction(ISD::MUL, VT, Legal);
      setOperationAction(ISD::AND, VT, Legal);
      setOperationAction(ISD::OR, VT, Legal);
      setOperationAction(ISD::XOR, VT, Legal);
      if (Subtarget.hasVectorEnhancements1())
        setOperationAction(ISD::CTPOP, VT, Legal);
      else
        setOperationAction(ISD::CTPOP, VT, Custom);
      setOperationAction(ISD::CTTZ, VT, Legal);
      setOperationAction(ISD::CTLZ, VT, Legal);

      // Convert a GPR scalar to a vector by inserting it into element 0.
      setOperationAction(ISD::SCALAR_TO_VECTOR, VT, Custom);

      // Use a series of unpacks for extensions.
      setOperationAction(ISD::SIGN_EXTEND_VECTOR_INREG, VT, Custom);
      setOperationAction(ISD::ZERO_EXTEND_VECTOR_INREG, VT, Custom);

      // Detect shifts by a scalar amount and convert them into
      // V*_BY_SCALAR.
      setOperationAction(ISD::SHL, VT, Custom);
      setOperationAction(ISD::SRA, VT, Custom);
      setOperationAction(ISD::SRL, VT, Custom);

      // At present ROTL isn't matched by DAGCombiner.  ROTR should be
      // converted into ROTL.
      setOperationAction(ISD::ROTL, VT, Expand);
      setOperationAction(ISD::ROTR, VT, Expand);

      // Map SETCCs onto one of VCE, VCH or VCHL, swapping the operands
      // and inverting the result as necessary.
      setOperationAction(ISD::SETCC, VT, Custom);
      setOperationAction(ISD::STRICT_FSETCC, VT, Custom);
      if (Subtarget.hasVectorEnhancements1())
        setOperationAction(ISD::STRICT_FSETCCS, VT, Custom);
    }
  }

  if (Subtarget.hasVector()) {
    // There should be no need to check for float types other than v2f64
    // since <2 x f32> isn't a legal type.
    setOperationAction(ISD::FP_TO_SINT, MVT::v2i64, Legal);
    setOperationAction(ISD::FP_TO_SINT, MVT::v2f64, Legal);
    setOperationAction(ISD::FP_TO_UINT, MVT::v2i64, Legal);
    setOperationAction(ISD::FP_TO_UINT, MVT::v2f64, Legal);
    setOperationAction(ISD::SINT_TO_FP, MVT::v2i64, Legal);
    setOperationAction(ISD::SINT_TO_FP, MVT::v2f64, Legal);
    setOperationAction(ISD::UINT_TO_FP, MVT::v2i64, Legal);
    setOperationAction(ISD::UINT_TO_FP, MVT::v2f64, Legal);

    setOperationAction(ISD::STRICT_FP_TO_SINT, MVT::v2i64, Legal);
    setOperationAction(ISD::STRICT_FP_TO_SINT, MVT::v2f64, Legal);
    setOperationAction(ISD::STRICT_FP_TO_UINT, MVT::v2i64, Legal);
    setOperationAction(ISD::STRICT_FP_TO_UINT, MVT::v2f64, Legal);
    setOperationAction(ISD::STRICT_SINT_TO_FP, MVT::v2i64, Legal);
    setOperationAction(ISD::STRICT_SINT_TO_FP, MVT::v2f64, Legal);
    setOperationAction(ISD::STRICT_UINT_TO_FP, MVT::v2i64, Legal);
    setOperationAction(ISD::STRICT_UINT_TO_FP, MVT::v2f64, Legal);
  }

  if (Subtarget.hasVectorEnhancements2()) {
    setOperationAction(ISD::FP_TO_SINT, MVT::v4i32, Legal);
    setOperationAction(ISD::FP_TO_SINT, MVT::v4f32, Legal);
    setOperationAction(ISD::FP_TO_UINT, MVT::v4i32, Legal);
    setOperationAction(ISD::FP_TO_UINT, MVT::v4f32, Legal);
    setOperationAction(ISD::SINT_TO_FP, MVT::v4i32, Legal);
    setOperationAction(ISD::SINT_TO_FP, MVT::v4f32, Legal);
    setOperationAction(ISD::UINT_TO_FP, MVT::v4i32, Legal);
    setOperationAction(ISD::UINT_TO_FP, MVT::v4f32, Legal);

    setOperationAction(ISD::STRICT_FP_TO_SINT, MVT::v4i32, Legal);
    setOperationAction(ISD::STRICT_FP_TO_SINT, MVT::v4f32, Legal);
    setOperationAction(ISD::STRICT_FP_TO_UINT, MVT::v4i32, Legal);
    setOperationAction(ISD::STRICT_FP_TO_UINT, MVT::v4f32, Legal);
    setOperationAction(ISD::STRICT_SINT_TO_FP, MVT::v4i32, Legal);
    setOperationAction(ISD::STRICT_SINT_TO_FP, MVT::v4f32, Legal);
    setOperationAction(ISD::STRICT_UINT_TO_FP, MVT::v4i32, Legal);
    setOperationAction(ISD::STRICT_UINT_TO_FP, MVT::v4f32, Legal);
  }

  // Handle floating-point types.
  for (unsigned I = MVT::FIRST_FP_VALUETYPE;
       I <= MVT::LAST_FP_VALUETYPE;
       ++I) {
    MVT VT = MVT::SimpleValueType(I);
    if (isTypeLegal(VT)) {
      // We can use FI for FRINT.
      setOperationAction(ISD::FRINT, VT, Legal);

      // We can use the extended form of FI for other rounding operations.
      if (Subtarget.hasFPExtension()) {
        setOperationAction(ISD::FNEARBYINT, VT, Legal);
        setOperationAction(ISD::FFLOOR, VT, Legal);
        setOperationAction(ISD::FCEIL, VT, Legal);
        setOperationAction(ISD::FTRUNC, VT, Legal);
        setOperationAction(ISD::FROUND, VT, Legal);
      }

      // No special instructions for these.
      setOperationAction(ISD::FSIN, VT, Expand);
      setOperationAction(ISD::FCOS, VT, Expand);
      setOperationAction(ISD::FSINCOS, VT, Expand);
      setOperationAction(ISD::FREM, VT, Expand);
      setOperationAction(ISD::FPOW, VT, Expand);

      // Handle constrained floating-point operations.
      setOperationAction(ISD::STRICT_FADD, VT, Legal);
      setOperationAction(ISD::STRICT_FSUB, VT, Legal);
      setOperationAction(ISD::STRICT_FMUL, VT, Legal);
      setOperationAction(ISD::STRICT_FDIV, VT, Legal);
      setOperationAction(ISD::STRICT_FMA, VT, Legal);
      setOperationAction(ISD::STRICT_FSQRT, VT, Legal);
      setOperationAction(ISD::STRICT_FRINT, VT, Legal);
      setOperationAction(ISD::STRICT_FP_ROUND, VT, Legal);
      setOperationAction(ISD::STRICT_FP_EXTEND, VT, Legal);
      if (Subtarget.hasFPExtension()) {
        setOperationAction(ISD::STRICT_FNEARBYINT, VT, Legal);
        setOperationAction(ISD::STRICT_FFLOOR, VT, Legal);
        setOperationAction(ISD::STRICT_FCEIL, VT, Legal);
        setOperationAction(ISD::STRICT_FROUND, VT, Legal);
        setOperationAction(ISD::STRICT_FTRUNC, VT, Legal);
      }
    }
  }

  // Handle floating-point vector types.
  if (Subtarget.hasVector()) {
    // Scalar-to-vector conversion is just a subreg.
    setOperationAction(ISD::SCALAR_TO_VECTOR, MVT::v4f32, Legal);
    setOperationAction(ISD::SCALAR_TO_VECTOR, MVT::v2f64, Legal);

    // Some insertions and extractions can be done directly but others
    // need to go via integers.
    setOperationAction(ISD::INSERT_VECTOR_ELT, MVT::v4f32, Custom);
    setOperationAction(ISD::INSERT_VECTOR_ELT, MVT::v2f64, Custom);
    setOperationAction(ISD::EXTRACT_VECTOR_ELT, MVT::v4f32, Custom);
    setOperationAction(ISD::EXTRACT_VECTOR_ELT, MVT::v2f64, Custom);

    // These operations have direct equivalents.
    setOperationAction(ISD::FADD, MVT::v2f64, Legal);
    setOperationAction(ISD::FNEG, MVT::v2f64, Legal);
    setOperationAction(ISD::FSUB, MVT::v2f64, Legal);
    setOperationAction(ISD::FMUL, MVT::v2f64, Legal);
    setOperationAction(ISD::FMA, MVT::v2f64, Legal);
    setOperationAction(ISD::FDIV, MVT::v2f64, Legal);
    setOperationAction(ISD::FABS, MVT::v2f64, Legal);
    setOperationAction(ISD::FSQRT, MVT::v2f64, Legal);
    setOperationAction(ISD::FRINT, MVT::v2f64, Legal);
    setOperationAction(ISD::FNEARBYINT, MVT::v2f64, Legal);
    setOperationAction(ISD::FFLOOR, MVT::v2f64, Legal);
    setOperationAction(ISD::FCEIL, MVT::v2f64, Legal);
    setOperationAction(ISD::FTRUNC, MVT::v2f64, Legal);
    setOperationAction(ISD::FROUND, MVT::v2f64, Legal);

    // Handle constrained floating-point operations.
    setOperationAction(ISD::STRICT_FADD, MVT::v2f64, Legal);
    setOperationAction(ISD::STRICT_FSUB, MVT::v2f64, Legal);
    setOperationAction(ISD::STRICT_FMUL, MVT::v2f64, Legal);
    setOperationAction(ISD::STRICT_FMA, MVT::v2f64, Legal);
    setOperationAction(ISD::STRICT_FDIV, MVT::v2f64, Legal);
    setOperationAction(ISD::STRICT_FSQRT, MVT::v2f64, Legal);
    setOperationAction(ISD::STRICT_FRINT, MVT::v2f64, Legal);
    setOperationAction(ISD::STRICT_FNEARBYINT, MVT::v2f64, Legal);
    setOperationAction(ISD::STRICT_FFLOOR, MVT::v2f64, Legal);
    setOperationAction(ISD::STRICT_FCEIL, MVT::v2f64, Legal);
    setOperationAction(ISD::STRICT_FTRUNC, MVT::v2f64, Legal);
    setOperationAction(ISD::STRICT_FROUND, MVT::v2f64, Legal);
  }

  // The vector enhancements facility 1 has instructions for these.
  if (Subtarget.hasVectorEnhancements1()) {
    setOperationAction(ISD::FADD, MVT::v4f32, Legal);
    setOperationAction(ISD::FNEG, MVT::v4f32, Legal);
    setOperationAction(ISD::FSUB, MVT::v4f32, Legal);
    setOperationAction(ISD::FMUL, MVT::v4f32, Legal);
    setOperationAction(ISD::FMA, MVT::v4f32, Legal);
    setOperationAction(ISD::FDIV, MVT::v4f32, Legal);
    setOperationAction(ISD::FABS, MVT::v4f32, Legal);
    setOperationAction(ISD::FSQRT, MVT::v4f32, Legal);
    setOperationAction(ISD::FRINT, MVT::v4f32, Legal);
    setOperationAction(ISD::FNEARBYINT, MVT::v4f32, Legal);
    setOperationAction(ISD::FFLOOR, MVT::v4f32, Legal);
    setOperationAction(ISD::FCEIL, MVT::v4f32, Legal);
    setOperationAction(ISD::FTRUNC, MVT::v4f32, Legal);
    setOperationAction(ISD::FROUND, MVT::v4f32, Legal);

    setOperationAction(ISD::FMAXNUM, MVT::f64, Legal);
    setOperationAction(ISD::FMAXIMUM, MVT::f64, Legal);
    setOperationAction(ISD::FMINNUM, MVT::f64, Legal);
    setOperationAction(ISD::FMINIMUM, MVT::f64, Legal);

    setOperationAction(ISD::FMAXNUM, MVT::v2f64, Legal);
    setOperationAction(ISD::FMAXIMUM, MVT::v2f64, Legal);
    setOperationAction(ISD::FMINNUM, MVT::v2f64, Legal);
    setOperationAction(ISD::FMINIMUM, MVT::v2f64, Legal);

    setOperationAction(ISD::FMAXNUM, MVT::f32, Legal);
    setOperationAction(ISD::FMAXIMUM, MVT::f32, Legal);
    setOperationAction(ISD::FMINNUM, MVT::f32, Legal);
    setOperationAction(ISD::FMINIMUM, MVT::f32, Legal);

    setOperationAction(ISD::FMAXNUM, MVT::v4f32, Legal);
    setOperationAction(ISD::FMAXIMUM, MVT::v4f32, Legal);
    setOperationAction(ISD::FMINNUM, MVT::v4f32, Legal);
    setOperationAction(ISD::FMINIMUM, MVT::v4f32, Legal);

    setOperationAction(ISD::FMAXNUM, MVT::f128, Legal);
    setOperationAction(ISD::FMAXIMUM, MVT::f128, Legal);
    setOperationAction(ISD::FMINNUM, MVT::f128, Legal);
    setOperationAction(ISD::FMINIMUM, MVT::f128, Legal);

    // Handle constrained floating-point operations.
    setOperationAction(ISD::STRICT_FADD, MVT::v4f32, Legal);
    setOperationAction(ISD::STRICT_FSUB, MVT::v4f32, Legal);
    setOperationAction(ISD::STRICT_FMUL, MVT::v4f32, Legal);
    setOperationAction(ISD::STRICT_FMA, MVT::v4f32, Legal);
    setOperationAction(ISD::STRICT_FDIV, MVT::v4f32, Legal);
    setOperationAction(ISD::STRICT_FSQRT, MVT::v4f32, Legal);
    setOperationAction(ISD::STRICT_FRINT, MVT::v4f32, Legal);
    setOperationAction(ISD::STRICT_FNEARBYINT, MVT::v4f32, Legal);
    setOperationAction(ISD::STRICT_FFLOOR, MVT::v4f32, Legal);
    setOperationAction(ISD::STRICT_FCEIL, MVT::v4f32, Legal);
    setOperationAction(ISD::STRICT_FROUND, MVT::v4f32, Legal);
    setOperationAction(ISD::STRICT_FTRUNC, MVT::v4f32, Legal);
    for (auto VT : { MVT::f32, MVT::f64, MVT::f128,
                     MVT::v4f32, MVT::v2f64 }) {
      setOperationAction(ISD::STRICT_FMAXNUM, VT, Legal);
      setOperationAction(ISD::STRICT_FMINNUM, VT, Legal);
      setOperationAction(ISD::STRICT_FMAXIMUM, VT, Legal);
      setOperationAction(ISD::STRICT_FMINIMUM, VT, Legal);
    }
  }

  // We only have fused f128 multiply-addition on vector registers.
  if (!Subtarget.hasVectorEnhancements1()) {
    setOperationAction(ISD::FMA, MVT::f128, Expand);
    setOperationAction(ISD::STRICT_FMA, MVT::f128, Expand);
  }

  // We don't have a copysign instruction on vector registers.
  if (Subtarget.hasVectorEnhancements1())
    setOperationAction(ISD::FCOPYSIGN, MVT::f128, Expand);

  // Needed so that we don't try to implement f128 constant loads using
  // a load-and-extend of a f80 constant (in cases where the constant
  // would fit in an f80).
  for (MVT VT : MVT::fp_valuetypes())
    setLoadExtAction(ISD::EXTLOAD, VT, MVT::f80, Expand);

  // We don't have extending load instruction on vector registers.
  if (Subtarget.hasVectorEnhancements1()) {
    setLoadExtAction(ISD::EXTLOAD, MVT::f128, MVT::f32, Expand);
    setLoadExtAction(ISD::EXTLOAD, MVT::f128, MVT::f64, Expand);
  }

  // Floating-point truncation and stores need to be done separately.
  setTruncStoreAction(MVT::f64,  MVT::f32, Expand);
  setTruncStoreAction(MVT::f128, MVT::f32, Expand);
  setTruncStoreAction(MVT::f128, MVT::f64, Expand);

  // We have 64-bit FPR<->GPR moves, but need special handling for
  // 32-bit forms.
  if (!Subtarget.hasVector()) {
    setOperationAction(ISD::BITCAST, MVT::i32, Custom);
    setOperationAction(ISD::BITCAST, MVT::f32, Custom);
  }

  // VASTART and VACOPY need to deal with the SystemZ-specific varargs
  // structure, but VAEND is a no-op.
  setOperationAction(ISD::VASTART, MVT::Other, Custom);
  setOperationAction(ISD::VACOPY,  MVT::Other, Custom);
  setOperationAction(ISD::VAEND,   MVT::Other, Expand);

  // Codes for which we want to perform some z-specific combinations.
  setTargetDAGCombine(ISD::ZERO_EXTEND);
  setTargetDAGCombine(ISD::SIGN_EXTEND);
  setTargetDAGCombine(ISD::SIGN_EXTEND_INREG);
  setTargetDAGCombine(ISD::LOAD);
  setTargetDAGCombine(ISD::STORE);
  setTargetDAGCombine(ISD::VECTOR_SHUFFLE);
  setTargetDAGCombine(ISD::EXTRACT_VECTOR_ELT);
  setTargetDAGCombine(ISD::FP_ROUND);
  setTargetDAGCombine(ISD::STRICT_FP_ROUND);
  setTargetDAGCombine(ISD::FP_EXTEND);
  setTargetDAGCombine(ISD::SINT_TO_FP);
  setTargetDAGCombine(ISD::UINT_TO_FP);
  setTargetDAGCombine(ISD::STRICT_FP_EXTEND);
  setTargetDAGCombine(ISD::BSWAP);
  setTargetDAGCombine(ISD::SDIV);
  setTargetDAGCombine(ISD::UDIV);
  setTargetDAGCombine(ISD::SREM);
  setTargetDAGCombine(ISD::UREM);
  setTargetDAGCombine(ISD::INTRINSIC_VOID);
  setTargetDAGCombine(ISD::INTRINSIC_W_CHAIN);

  // Handle intrinsics.
  setOperationAction(ISD::INTRINSIC_W_CHAIN, MVT::Other, Custom);
  setOperationAction(ISD::INTRINSIC_WO_CHAIN, MVT::Other, Custom);

  // We want to use MVC in preference to even a single load/store pair.
  MaxStoresPerMemcpy = 0;
  MaxStoresPerMemcpyOptSize = 0;

  // The main memset sequence is a byte store followed by an MVC.
  // Two STC or MV..I stores win over that, but the kind of fused stores
  // generated by target-independent code don't when the byte value is
  // variable.  E.g.  "STC <reg>;MHI <reg>,257;STH <reg>" is not better
  // than "STC;MVC".  Handle the choice in target-specific code instead.
  MaxStoresPerMemset = 0;
  MaxStoresPerMemsetOptSize = 0;

  // Default to having -disable-strictnode-mutation on
  IsStrictFPEnabled = true;
}

bool SystemZTargetLowering::useSoftFloat() const {
  return Subtarget.hasSoftFloat();
}

EVT SystemZTargetLowering::getSetCCResultType(const DataLayout &DL,
                                              LLVMContext &, EVT VT) const {
  if (!VT.isVector())
    return MVT::i32;
  return VT.changeVectorElementTypeToInteger();
}

bool SystemZTargetLowering::isFMAFasterThanFMulAndFAdd(
    const MachineFunction &MF, EVT VT) const {
  VT = VT.getScalarType();

  if (!VT.isSimple())
    return false;

  switch (VT.getSimpleVT().SimpleTy) {
  case MVT::f32:
  case MVT::f64:
    return true;
  case MVT::f128:
    return Subtarget.hasVectorEnhancements1();
  default:
    break;
  }

  return false;
}

// Return true if the constant can be generated with a vector instruction,
// such as VGM, VGMB or VREPI.
bool SystemZVectorConstantInfo::isVectorConstantLegal(
    const SystemZSubtarget &Subtarget) {
  const SystemZInstrInfo *TII =
      static_cast<const SystemZInstrInfo *>(Subtarget.getInstrInfo());
  if (!Subtarget.hasVector() ||
      (isFP128 && !Subtarget.hasVectorEnhancements1()))
    return false;

  // Try using VECTOR GENERATE BYTE MASK.  This is the architecturally-
  // preferred way of creating all-zero and all-one vectors so give it
  // priority over other methods below.
  unsigned Mask = 0;
  unsigned I = 0;
  for (; I < SystemZ::VectorBytes; ++I) {
    uint64_t Byte = IntBits.lshr(I * 8).trunc(8).getZExtValue();
    if (Byte == 0xff)
      Mask |= 1ULL << I;
    else if (Byte != 0)
      break;
  }
  if (I == SystemZ::VectorBytes) {
    Opcode = SystemZISD::BYTE_MASK;
    OpVals.push_back(Mask);
    VecVT = MVT::getVectorVT(MVT::getIntegerVT(8), 16);
    return true;
  }

  if (SplatBitSize > 64)
    return false;

  auto tryValue = [&](uint64_t Value) -> bool {
    // Try VECTOR REPLICATE IMMEDIATE
    int64_t SignedValue = SignExtend64(Value, SplatBitSize);
    if (isInt<16>(SignedValue)) {
      OpVals.push_back(((unsigned) SignedValue));
      Opcode = SystemZISD::REPLICATE;
      VecVT = MVT::getVectorVT(MVT::getIntegerVT(SplatBitSize),
                               SystemZ::VectorBits / SplatBitSize);
      return true;
    }
    // Try VECTOR GENERATE MASK
    unsigned Start, End;
    if (TII->isRxSBGMask(Value, SplatBitSize, Start, End)) {
      // isRxSBGMask returns the bit numbers for a full 64-bit value, with 0
      // denoting 1 << 63 and 63 denoting 1.  Convert them to bit numbers for
      // an SplatBitSize value, so that 0 denotes 1 << (SplatBitSize-1).
      OpVals.push_back(Start - (64 - SplatBitSize));
      OpVals.push_back(End - (64 - SplatBitSize));
      Opcode = SystemZISD::ROTATE_MASK;
      VecVT = MVT::getVectorVT(MVT::getIntegerVT(SplatBitSize),
                               SystemZ::VectorBits / SplatBitSize);
      return true;
    }
    return false;
  };

  // First try assuming that any undefined bits above the highest set bit
  // and below the lowest set bit are 1s.  This increases the likelihood of
  // being able to use a sign-extended element value in VECTOR REPLICATE
  // IMMEDIATE or a wraparound mask in VECTOR GENERATE MASK.
  uint64_t SplatBitsZ = SplatBits.getZExtValue();
  uint64_t SplatUndefZ = SplatUndef.getZExtValue();
  uint64_t Lower =
      (SplatUndefZ & ((uint64_t(1) << findFirstSet(SplatBitsZ)) - 1));
  uint64_t Upper =
      (SplatUndefZ & ~((uint64_t(1) << findLastSet(SplatBitsZ)) - 1));
  if (tryValue(SplatBitsZ | Upper | Lower))
    return true;

  // Now try assuming that any undefined bits between the first and
  // last defined set bits are set.  This increases the chances of
  // using a non-wraparound mask.
  uint64_t Middle = SplatUndefZ & ~Upper & ~Lower;
  return tryValue(SplatBitsZ | Middle);
}

SystemZVectorConstantInfo::SystemZVectorConstantInfo(APFloat FPImm) {
  IntBits = FPImm.bitcastToAPInt().zextOrSelf(128);
  isFP128 = (&FPImm.getSemantics() == &APFloat::IEEEquad());

  // Find the smallest splat.
  SplatBits = FPImm.bitcastToAPInt();
  unsigned Width = SplatBits.getBitWidth();
  while (Width > 8) {
    unsigned HalfSize = Width / 2;
    APInt HighValue = SplatBits.lshr(HalfSize).trunc(HalfSize);
    APInt LowValue = SplatBits.trunc(HalfSize);

    // If the two halves do not match, stop here.
    if (HighValue != LowValue || 8 > HalfSize)
      break;

    SplatBits = HighValue;
    Width = HalfSize;
  }
  SplatUndef = 0;
  SplatBitSize = Width;
}

SystemZVectorConstantInfo::SystemZVectorConstantInfo(BuildVectorSDNode *BVN) {
  assert(BVN->isConstant() && "Expected a constant BUILD_VECTOR");
  bool HasAnyUndefs;

  // Get IntBits by finding the 128 bit splat.
  BVN->isConstantSplat(IntBits, SplatUndef, SplatBitSize, HasAnyUndefs, 128,
                       true);

  // Get SplatBits by finding the 8 bit or greater splat.
  BVN->isConstantSplat(SplatBits, SplatUndef, SplatBitSize, HasAnyUndefs, 8,
                       true);
}

bool SystemZTargetLowering::isFPImmLegal(const APFloat &Imm, EVT VT,
                                         bool ForCodeSize) const {
  // We can load zero using LZ?R and negative zero using LZ?R;LC?BR.
  if (Imm.isZero() || Imm.isNegZero())
    return true;

  return SystemZVectorConstantInfo(Imm).isVectorConstantLegal(Subtarget);
}

/// Returns true if stack probing through inline assembly is requested.
bool SystemZTargetLowering::hasInlineStackProbe(MachineFunction &MF) const {
  // If the function specifically requests inline stack probes, emit them.
  if (MF.getFunction().hasFnAttribute("probe-stack"))
    return MF.getFunction().getFnAttribute("probe-stack").getValueAsString() ==
           "inline-asm";
  return false;
}

bool SystemZTargetLowering::isLegalICmpImmediate(int64_t Imm) const {
  // We can use CGFI or CLGFI.
  return isInt<32>(Imm) || isUInt<32>(Imm);
}

bool SystemZTargetLowering::isLegalAddImmediate(int64_t Imm) const {
  // We can use ALGFI or SLGFI.
  return isUInt<32>(Imm) || isUInt<32>(-Imm);
}

bool SystemZTargetLowering::allowsMisalignedMemoryAccesses(
    EVT VT, unsigned, unsigned, MachineMemOperand::Flags, bool *Fast) const {
  // Unaligned accesses should never be slower than the expanded version.
  // We check specifically for aligned accesses in the few cases where
  // they are required.
  if (Fast)
    *Fast = true;
  return true;
}

// Information about the addressing mode for a memory access.
struct AddressingMode {
  // True if a long displacement is supported.
  bool LongDisplacement;

  // True if use of index register is supported.
  bool IndexReg;

  AddressingMode(bool LongDispl, bool IdxReg) :
    LongDisplacement(LongDispl), IndexReg(IdxReg) {}
};

// Return the desired addressing mode for a Load which has only one use (in
// the same block) which is a Store.
static AddressingMode getLoadStoreAddrMode(bool HasVector,
                                          Type *Ty) {
  // With vector support a Load->Store combination may be combined to either
  // an MVC or vector operations and it seems to work best to allow the
  // vector addressing mode.
  if (HasVector)
    return AddressingMode(false/*LongDispl*/, true/*IdxReg*/);

  // Otherwise only the MVC case is special.
  bool MVC = Ty->isIntegerTy(8);
  return AddressingMode(!MVC/*LongDispl*/, !MVC/*IdxReg*/);
}

// Return the addressing mode which seems most desirable given an LLVM
// Instruction pointer.
static AddressingMode
supportedAddressingMode(Instruction *I, bool HasVector) {
  if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(I)) {
    switch (II->getIntrinsicID()) {
    default: break;
    case Intrinsic::memset:
    case Intrinsic::memmove:
    case Intrinsic::memcpy:
      return AddressingMode(false/*LongDispl*/, false/*IdxReg*/);
    }
  }

  if (isa<LoadInst>(I) && I->hasOneUse()) {
    auto *SingleUser = cast<Instruction>(*I->user_begin());
    if (SingleUser->getParent() == I->getParent()) {
      if (isa<ICmpInst>(SingleUser)) {
        if (auto *C = dyn_cast<ConstantInt>(SingleUser->getOperand(1)))
          if (C->getBitWidth() <= 64 &&
              (isInt<16>(C->getSExtValue()) || isUInt<16>(C->getZExtValue())))
            // Comparison of memory with 16 bit signed / unsigned immediate
            return AddressingMode(false/*LongDispl*/, false/*IdxReg*/);
      } else if (isa<StoreInst>(SingleUser))
        // Load->Store
        return getLoadStoreAddrMode(HasVector, I->getType());
    }
  } else if (auto *StoreI = dyn_cast<StoreInst>(I)) {
    if (auto *LoadI = dyn_cast<LoadInst>(StoreI->getValueOperand()))
      if (LoadI->hasOneUse() && LoadI->getParent() == I->getParent())
        // Load->Store
        return getLoadStoreAddrMode(HasVector, LoadI->getType());
  }

  if (HasVector && (isa<LoadInst>(I) || isa<StoreInst>(I))) {

    // * Use LDE instead of LE/LEY for z13 to avoid partial register
    //   dependencies (LDE only supports small offsets).
    // * Utilize the vector registers to hold floating point
    //   values (vector load / store instructions only support small
    //   offsets).

    Type *MemAccessTy = (isa<LoadInst>(I) ? I->getType() :
                         I->getOperand(0)->getType());
    bool IsFPAccess = MemAccessTy->isFloatingPointTy();
    bool IsVectorAccess = MemAccessTy->isVectorTy();

    // A store of an extracted vector element will be combined into a VSTE type
    // instruction.
    if (!IsVectorAccess && isa<StoreInst>(I)) {
      Value *DataOp = I->getOperand(0);
      if (isa<ExtractElementInst>(DataOp))
        IsVectorAccess = true;
    }

    // A load which gets inserted into a vector element will be combined into a
    // VLE type instruction.
    if (!IsVectorAccess && isa<LoadInst>(I) && I->hasOneUse()) {
      User *LoadUser = *I->user_begin();
      if (isa<InsertElementInst>(LoadUser))
        IsVectorAccess = true;
    }

    if (IsFPAccess || IsVectorAccess)
      return AddressingMode(false/*LongDispl*/, true/*IdxReg*/);
  }

  return AddressingMode(true/*LongDispl*/, true/*IdxReg*/);
}

bool SystemZTargetLowering::isLegalAddressingMode(const DataLayout &DL,
       const AddrMode &AM, Type *Ty, unsigned AS, Instruction *I) const {
  // Punt on globals for now, although they can be used in limited
  // RELATIVE LONG cases.
  if (AM.BaseGV)
    return false;

  // Require a 20-bit signed offset.
  if (!isInt<20>(AM.BaseOffs))
    return false;

  AddressingMode SupportedAM(true, true);
  if (I != nullptr)
    SupportedAM = supportedAddressingMode(I, Subtarget.hasVector());

  if (!SupportedAM.LongDisplacement && !isUInt<12>(AM.BaseOffs))
    return false;

  if (!SupportedAM.IndexReg)
    // No indexing allowed.
    return AM.Scale == 0;
  else
    // Indexing is OK but no scale factor can be applied.
    return AM.Scale == 0 || AM.Scale == 1;
}

bool SystemZTargetLowering::isTruncateFree(Type *FromType, Type *ToType) const {
  if (!FromType->isIntegerTy() || !ToType->isIntegerTy())
    return false;
  unsigned FromBits = FromType->getPrimitiveSizeInBits();
  unsigned ToBits = ToType->getPrimitiveSizeInBits();
  return FromBits > ToBits;
}

bool SystemZTargetLowering::isTruncateFree(EVT FromVT, EVT ToVT) const {
  if (!FromVT.isInteger() || !ToVT.isInteger())
    return false;
  unsigned FromBits = FromVT.getSizeInBits();
  unsigned ToBits = ToVT.getSizeInBits();
  return FromBits > ToBits;
}

//===----------------------------------------------------------------------===//
// Inline asm support
//===----------------------------------------------------------------------===//

TargetLowering::ConstraintType
SystemZTargetLowering::getConstraintType(StringRef Constraint) const {
  if (Constraint.size() == 1) {
    switch (Constraint[0]) {
    case 'a': // Address register
    case 'd': // Data register (equivalent to 'r')
    case 'f': // Floating-point register
    case 'h': // High-part register
    case 'r': // General-purpose register
    case 'v': // Vector register
      return C_RegisterClass;

    case 'Q': // Memory with base and unsigned 12-bit displacement
    case 'R': // Likewise, plus an index
    case 'S': // Memory with base and signed 20-bit displacement
    case 'T': // Likewise, plus an index
    case 'm': // Equivalent to 'T'.
      return C_Memory;

    case 'I': // Unsigned 8-bit constant
    case 'J': // Unsigned 12-bit constant
    case 'K': // Signed 16-bit constant
    case 'L': // Signed 20-bit displacement (on all targets we support)
    case 'M': // 0x7fffffff
      return C_Immediate;

    default:
      break;
    }
  }
  return TargetLowering::getConstraintType(Constraint);
}

TargetLowering::ConstraintWeight SystemZTargetLowering::
getSingleConstraintMatchWeight(AsmOperandInfo &info,
                               const char *constraint) const {
  ConstraintWeight weight = CW_Invalid;
  Value *CallOperandVal = info.CallOperandVal;
  // If we don't have a value, we can't do a match,
  // but allow it at the lowest weight.
  if (!CallOperandVal)
    return CW_Default;
  Type *type = CallOperandVal->getType();
  // Look at the constraint type.
  switch (*constraint) {
  default:
    weight = TargetLowering::getSingleConstraintMatchWeight(info, constraint);
    break;

  case 'a': // Address register
  case 'd': // Data register (equivalent to 'r')
  case 'h': // High-part register
  case 'r': // General-purpose register
    if (CallOperandVal->getType()->isIntegerTy())
      weight = CW_Register;
    break;

  case 'f': // Floating-point register
    if (type->isFloatingPointTy())
      weight = CW_Register;
    break;

  case 'v': // Vector register
    if ((type->isVectorTy() || type->isFloatingPointTy()) &&
        Subtarget.hasVector())
      weight = CW_Register;
    break;

  case 'I': // Unsigned 8-bit constant
    if (auto *C = dyn_cast<ConstantInt>(CallOperandVal))
      if (isUInt<8>(C->getZExtValue()))
        weight = CW_Constant;
    break;

  case 'J': // Unsigned 12-bit constant
    if (auto *C = dyn_cast<ConstantInt>(CallOperandVal))
      if (isUInt<12>(C->getZExtValue()))
        weight = CW_Constant;
    break;

  case 'K': // Signed 16-bit constant
    if (auto *C = dyn_cast<ConstantInt>(CallOperandVal))
      if (isInt<16>(C->getSExtValue()))
        weight = CW_Constant;
    break;

  case 'L': // Signed 20-bit displacement (on all targets we support)
    if (auto *C = dyn_cast<ConstantInt>(CallOperandVal))
      if (isInt<20>(C->getSExtValue()))
        weight = CW_Constant;
    break;

  case 'M': // 0x7fffffff
    if (auto *C = dyn_cast<ConstantInt>(CallOperandVal))
      if (C->getZExtValue() == 0x7fffffff)
        weight = CW_Constant;
    break;
  }
  return weight;
}

// Parse a "{tNNN}" register constraint for which the register type "t"
// has already been verified.  MC is the class associated with "t" and
// Map maps 0-based register numbers to LLVM register numbers.
static std::pair<unsigned, const TargetRegisterClass *>
parseRegisterNumber(StringRef Constraint, const TargetRegisterClass *RC,
                    const unsigned *Map, unsigned Size) {
  assert(*(Constraint.end()-1) == '}' && "Missing '}'");
  if (isdigit(Constraint[2])) {
    unsigned Index;
    bool Failed =
        Constraint.slice(2, Constraint.size() - 1).getAsInteger(10, Index);
    if (!Failed && Index < Size && Map[Index])
      return std::make_pair(Map[Index], RC);
  }
  return std::make_pair(0U, nullptr);
}

std::pair<unsigned, const TargetRegisterClass *>
SystemZTargetLowering::getRegForInlineAsmConstraint(
    const TargetRegisterInfo *TRI, StringRef Constraint, MVT VT) const {
  if (Constraint.size() == 1) {
    // GCC Constraint Letters
    switch (Constraint[0]) {
    default: break;
    case 'd': // Data register (equivalent to 'r')
    case 'r': // General-purpose register
      if (VT == MVT::i64)
        return std::make_pair(0U, &SystemZ::GR64BitRegClass);
      else if (VT == MVT::i128)
        return std::make_pair(0U, &SystemZ::GR128BitRegClass);
      return std::make_pair(0U, &SystemZ::GR32BitRegClass);

    case 'a': // Address register
      if (VT == MVT::i64)
        return std::make_pair(0U, &SystemZ::ADDR64BitRegClass);
      else if (VT == MVT::i128)
        return std::make_pair(0U, &SystemZ::ADDR128BitRegClass);
      return std::make_pair(0U, &SystemZ::ADDR32BitRegClass);

    case 'h': // High-part register (an LLVM extension)
      return std::make_pair(0U, &SystemZ::GRH32BitRegClass);

    case 'f': // Floating-point register
      if (!useSoftFloat()) {
        if (VT == MVT::f64)
          return std::make_pair(0U, &SystemZ::FP64BitRegClass);
        else if (VT == MVT::f128)
          return std::make_pair(0U, &SystemZ::FP128BitRegClass);
        return std::make_pair(0U, &SystemZ::FP32BitRegClass);
      }
      break;
    case 'v': // Vector register
      if (Subtarget.hasVector()) {
        if (VT == MVT::f32)
          return std::make_pair(0U, &SystemZ::VR32BitRegClass);
        if (VT == MVT::f64)
          return std::make_pair(0U, &SystemZ::VR64BitRegClass);
        return std::make_pair(0U, &SystemZ::VR128BitRegClass);
      }
      break;
    }
  }
  if (Constraint.size() > 0 && Constraint[0] == '{') {
    // We need to override the default register parsing for GPRs and FPRs
    // because the interpretation depends on VT.  The internal names of
    // the registers are also different from the external names
    // (F0D and F0S instead of F0, etc.).
    if (Constraint[1] == 'r') {
      if (VT == MVT::i32)
        return parseRegisterNumber(Constraint, &SystemZ::GR32BitRegClass,
                                   SystemZMC::GR32Regs, 16);
      if (VT == MVT::i128)
        return parseRegisterNumber(Constraint, &SystemZ::GR128BitRegClass,
                                   SystemZMC::GR128Regs, 16);
      return parseRegisterNumber(Constraint, &SystemZ::GR64BitRegClass,
                                 SystemZMC::GR64Regs, 16);
    }
    if (Constraint[1] == 'f') {
      if (useSoftFloat())
        return std::make_pair(
            0u, static_cast<const TargetRegisterClass *>(nullptr));
      if (VT == MVT::f32)
        return parseRegisterNumber(Constraint, &SystemZ::FP32BitRegClass,
                                   SystemZMC::FP32Regs, 16);
      if (VT == MVT::f128)
        return parseRegisterNumber(Constraint, &SystemZ::FP128BitRegClass,
                                   SystemZMC::FP128Regs, 16);
      return parseRegisterNumber(Constraint, &SystemZ::FP64BitRegClass,
                                 SystemZMC::FP64Regs, 16);
    }
    if (Constraint[1] == 'v') {
      if (!Subtarget.hasVector())
        return std::make_pair(
            0u, static_cast<const TargetRegisterClass *>(nullptr));
      if (VT == MVT::f32)
        return parseRegisterNumber(Constraint, &SystemZ::VR32BitRegClass,
                                   SystemZMC::VR32Regs, 32);
      if (VT == MVT::f64)
        return parseRegisterNumber(Constraint, &SystemZ::VR64BitRegClass,
                                   SystemZMC::VR64Regs, 32);
      return parseRegisterNumber(Constraint, &SystemZ::VR128BitRegClass,
                                 SystemZMC::VR128Regs, 32);
    }
  }
  return TargetLowering::getRegForInlineAsmConstraint(TRI, Constraint, VT);
}

// FIXME? Maybe this could be a TableGen attribute on some registers and
// this table could be generated automatically from RegInfo.
Register SystemZTargetLowering::getRegisterByName(const char *RegName, LLT VT,
                                                  const MachineFunction &MF) const {

  Register Reg = StringSwitch<Register>(RegName)
                   .Case("r15", SystemZ::R15D)
                   .Default(0);
  if (Reg)
    return Reg;
  report_fatal_error("Invalid register name global variable");
}

void SystemZTargetLowering::
LowerAsmOperandForConstraint(SDValue Op, std::string &Constraint,
                             std::vector<SDValue> &Ops,
                             SelectionDAG &DAG) const {
  // Only support length 1 constraints for now.
  if (Constraint.length() == 1) {
    switch (Constraint[0]) {
    case 'I': // Unsigned 8-bit constant
      if (auto *C = dyn_cast<ConstantSDNode>(Op))
        if (isUInt<8>(C->getZExtValue()))
          Ops.push_back(DAG.getTargetConstant(C->getZExtValue(), SDLoc(Op),
                                              Op.getValueType()));
      return;

    case 'J': // Unsigned 12-bit constant
      if (auto *C = dyn_cast<ConstantSDNode>(Op))
        if (isUInt<12>(C->getZExtValue()))
          Ops.push_back(DAG.getTargetConstant(C->getZExtValue(), SDLoc(Op),
                                              Op.getValueType()));
      return;

    case 'K': // Signed 16-bit constant
      if (auto *C = dyn_cast<ConstantSDNode>(Op))
        if (isInt<16>(C->getSExtValue()))
          Ops.push_back(DAG.getTargetConstant(C->getSExtValue(), SDLoc(Op),
                                              Op.getValueType()));
      return;

    case 'L': // Signed 20-bit displacement (on all targets we support)
      if (auto *C = dyn_cast<ConstantSDNode>(Op))
        if (isInt<20>(C->getSExtValue()))
          Ops.push_back(DAG.getTargetConstant(C->getSExtValue(), SDLoc(Op),
                                              Op.getValueType()));
      return;

    case 'M': // 0x7fffffff
      if (auto *C = dyn_cast<ConstantSDNode>(Op))
        if (C->getZExtValue() == 0x7fffffff)
          Ops.push_back(DAG.getTargetConstant(C->getZExtValue(), SDLoc(Op),
                                              Op.getValueType()));
      return;
    }
  }
  TargetLowering::LowerAsmOperandForConstraint(Op, Constraint, Ops, DAG);
}

//===----------------------------------------------------------------------===//
// Calling conventions
//===----------------------------------------------------------------------===//

#include "SystemZGenCallingConv.inc"

const MCPhysReg *SystemZTargetLowering::getScratchRegisters(
  CallingConv::ID) const {
  static const MCPhysReg ScratchRegs[] = { SystemZ::R0D, SystemZ::R1D,
                                           SystemZ::R14D, 0 };
  return ScratchRegs;
}

bool SystemZTargetLowering::allowTruncateForTailCall(Type *FromType,
                                                     Type *ToType) const {
  return isTruncateFree(FromType, ToType);
}

bool SystemZTargetLowering::mayBeEmittedAsTailCall(const CallInst *CI) const {
  return CI->isTailCall();
}

// We do not yet support 128-bit single-element vector types.  If the user
// attempts to use such types as function argument or return type, prefer
// to error out instead of emitting code violating the ABI.
static void VerifyVectorType(MVT VT, EVT ArgVT) {
  if (ArgVT.isVector() && !VT.isVector())
    report_fatal_error("Unsupported vector argument or return type");
}

static void VerifyVectorTypes(const SmallVectorImpl<ISD::InputArg> &Ins) {
  for (unsigned i = 0; i < Ins.size(); ++i)
    VerifyVectorType(Ins[i].VT, Ins[i].ArgVT);
}

static void VerifyVectorTypes(const SmallVectorImpl<ISD::OutputArg> &Outs) {
  for (unsigned i = 0; i < Outs.size(); ++i)
    VerifyVectorType(Outs[i].VT, Outs[i].ArgVT);
}

// Value is a value that has been passed to us in the location described by VA
// (and so has type VA.getLocVT()).  Convert Value to VA.getValVT(), chaining
// any loads onto Chain.
static SDValue convertLocVTToValVT(SelectionDAG &DAG, const SDLoc &DL,
                                   CCValAssign &VA, SDValue Chain,
                                   SDValue Value) {
  // If the argument has been promoted from a smaller type, insert an
  // assertion to capture this.
  if (VA.getLocInfo() == CCValAssign::SExt)
    Value = DAG.getNode(ISD::AssertSext, DL, VA.getLocVT(), Value,
                        DAG.getValueType(VA.getValVT()));
  else if (VA.getLocInfo() == CCValAssign::ZExt)
    Value = DAG.getNode(ISD::AssertZext, DL, VA.getLocVT(), Value,
                        DAG.getValueType(VA.getValVT()));

  if (VA.isExtInLoc())
    Value = DAG.getNode(ISD::TRUNCATE, DL, VA.getValVT(), Value);
  else if (VA.getLocInfo() == CCValAssign::BCvt) {
    // If this is a short vector argument loaded from the stack,
    // extend from i64 to full vector size and then bitcast.
    assert(VA.getLocVT() == MVT::i64);
    assert(VA.getValVT().isVector());
    Value = DAG.getBuildVector(MVT::v2i64, DL, {Value, DAG.getUNDEF(MVT::i64)});
    Value = DAG.getNode(ISD::BITCAST, DL, VA.getValVT(), Value);
  } else
    assert(VA.getLocInfo() == CCValAssign::Full && "Unsupported getLocInfo");
  return Value;
}

// Value is a value of type VA.getValVT() that we need to copy into
// the location described by VA.  Return a copy of Value converted to
// VA.getValVT().  The caller is responsible for handling indirect values.
static SDValue convertValVTToLocVT(SelectionDAG &DAG, const SDLoc &DL,
                                   CCValAssign &VA, SDValue Value) {
  switch (VA.getLocInfo()) {
  case CCValAssign::SExt:
    return DAG.getNode(ISD::SIGN_EXTEND, DL, VA.getLocVT(), Value);
  case CCValAssign::ZExt:
    return DAG.getNode(ISD::ZERO_EXTEND, DL, VA.getLocVT(), Value);
  case CCValAssign::AExt:
    return DAG.getNode(ISD::ANY_EXTEND, DL, VA.getLocVT(), Value);
  case CCValAssign::BCvt:
    // If this is a short vector argument to be stored to the stack,
    // bitcast to v2i64 and then extract first element.
    assert(VA.getLocVT() == MVT::i64);
    assert(VA.getValVT().isVector());
    Value = DAG.getNode(ISD::BITCAST, DL, MVT::v2i64, Value);
    return DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, VA.getLocVT(), Value,
                       DAG.getConstant(0, DL, MVT::i32));
  case CCValAssign::Full:
    return Value;
  default:
    llvm_unreachable("Unhandled getLocInfo()");
  }
}

SDValue SystemZTargetLowering::LowerFormalArguments(
    SDValue Chain, CallingConv::ID CallConv, bool IsVarArg,
    const SmallVectorImpl<ISD::InputArg> &Ins, const SDLoc &DL,
    SelectionDAG &DAG, SmallVectorImpl<SDValue> &InVals) const {
  MachineFunction &MF = DAG.getMachineFunction();
  MachineFrameInfo &MFI = MF.getFrameInfo();
  MachineRegisterInfo &MRI = MF.getRegInfo();
  SystemZMachineFunctionInfo *FuncInfo =
      MF.getInfo<SystemZMachineFunctionInfo>();
  auto *TFL =
      static_cast<const SystemZFrameLowering *>(Subtarget.getFrameLowering());
  EVT PtrVT = getPointerTy(DAG.getDataLayout());

  // Detect unsupported vector argument types.
  if (Subtarget.hasVector())
    VerifyVectorTypes(Ins);

  // Assign locations to all of the incoming arguments.
  SmallVector<CCValAssign, 16> ArgLocs;
  SystemZCCState CCInfo(CallConv, IsVarArg, MF, ArgLocs, *DAG.getContext());
  CCInfo.AnalyzeFormalArguments(Ins, CC_SystemZ);

  unsigned NumFixedGPRs = 0;
  unsigned NumFixedFPRs = 0;
  for (unsigned I = 0, E = ArgLocs.size(); I != E; ++I) {
    SDValue ArgValue;
    CCValAssign &VA = ArgLocs[I];
    EVT LocVT = VA.getLocVT();
    if (VA.isRegLoc()) {
      // Arguments passed in registers
      const TargetRegisterClass *RC;
      switch (LocVT.getSimpleVT().SimpleTy) {
      default:
        // Integers smaller than i64 should be promoted to i64.
        llvm_unreachable("Unexpected argument type");
      case MVT::i32:
        NumFixedGPRs += 1;
        RC = &SystemZ::GR32BitRegClass;
        break;
      case MVT::i64:
        NumFixedGPRs += 1;
        RC = &SystemZ::GR64BitRegClass;
        break;
      case MVT::f32:
        NumFixedFPRs += 1;
        RC = &SystemZ::FP32BitRegClass;
        break;
      case MVT::f64:
        NumFixedFPRs += 1;
        RC = &SystemZ::FP64BitRegClass;
        break;
      case MVT::v16i8:
      case MVT::v8i16:
      case MVT::v4i32:
      case MVT::v2i64:
      case MVT::v4f32:
      case MVT::v2f64:
        RC = &SystemZ::VR128BitRegClass;
        break;
      }

      Register VReg = MRI.createVirtualRegister(RC);
      MRI.addLiveIn(VA.getLocReg(), VReg);
      ArgValue = DAG.getCopyFromReg(Chain, DL, VReg, LocVT);
    } else {
      assert(VA.isMemLoc() && "Argument not register or memory");

      // Create the frame index object for this incoming parameter.
      int FI = MFI.CreateFixedObject(LocVT.getSizeInBits() / 8,
                                     VA.getLocMemOffset(), true);

      // Create the SelectionDAG nodes corresponding to a load
      // from this parameter.  Unpromoted ints and floats are
      // passed as right-justified 8-byte values.
      SDValue FIN = DAG.getFrameIndex(FI, PtrVT);
      if (VA.getLocVT() == MVT::i32 || VA.getLocVT() == MVT::f32)
        FIN = DAG.getNode(ISD::ADD, DL, PtrVT, FIN,
                          DAG.getIntPtrConstant(4, DL));
      ArgValue = DAG.getLoad(LocVT, DL, Chain, FIN,
                             MachinePointerInfo::getFixedStack(MF, FI));
    }

    // Convert the value of the argument register into the value that's
    // being passed.
    if (VA.getLocInfo() == CCValAssign::Indirect) {
      InVals.push_back(DAG.getLoad(VA.getValVT(), DL, Chain, ArgValue,
                                   MachinePointerInfo()));
      // If the original argument was split (e.g. i128), we need
      // to load all parts of it here (using the same address).
      unsigned ArgIndex = Ins[I].OrigArgIndex;
      assert (Ins[I].PartOffset == 0);
      while (I + 1 != E && Ins[I + 1].OrigArgIndex == ArgIndex) {
        CCValAssign &PartVA = ArgLocs[I + 1];
        unsigned PartOffset = Ins[I + 1].PartOffset;
        SDValue Address = DAG.getNode(ISD::ADD, DL, PtrVT, ArgValue,
                                      DAG.getIntPtrConstant(PartOffset, DL));
        InVals.push_back(DAG.getLoad(PartVA.getValVT(), DL, Chain, Address,
                                     MachinePointerInfo()));
        ++I;
      }
    } else
      InVals.push_back(convertLocVTToValVT(DAG, DL, VA, Chain, ArgValue));
  }

  if (IsVarArg) {
    // Save the number of non-varargs registers for later use by va_start, etc.
    FuncInfo->setVarArgsFirstGPR(NumFixedGPRs);
    FuncInfo->setVarArgsFirstFPR(NumFixedFPRs);

    // Likewise the address (in the form of a frame index) of where the
    // first stack vararg would be.  The 1-byte size here is arbitrary.
    int64_t StackSize = CCInfo.getNextStackOffset();
    FuncInfo->setVarArgsFrameIndex(MFI.CreateFixedObject(1, StackSize, true));

    // ...and a similar frame index for the caller-allocated save area
    // that will be used to store the incoming registers.
    int64_t RegSaveOffset =
      -SystemZMC::CallFrameSize + TFL->getRegSpillOffset(MF, SystemZ::R2D) - 16;
    unsigned RegSaveIndex = MFI.CreateFixedObject(1, RegSaveOffset, true);
    FuncInfo->setRegSaveFrameIndex(RegSaveIndex);

    // Store the FPR varargs in the reserved frame slots.  (We store the
    // GPRs as part of the prologue.)
    if (NumFixedFPRs < SystemZ::NumArgFPRs && !useSoftFloat()) {
      SDValue MemOps[SystemZ::NumArgFPRs];
      for (unsigned I = NumFixedFPRs; I < SystemZ::NumArgFPRs; ++I) {
        unsigned Offset = TFL->getRegSpillOffset(MF, SystemZ::ArgFPRs[I]);
        int FI =
          MFI.CreateFixedObject(8, -SystemZMC::CallFrameSize + Offset, true);
        SDValue FIN = DAG.getFrameIndex(FI, getPointerTy(DAG.getDataLayout()));
        unsigned VReg = MF.addLiveIn(SystemZ::ArgFPRs[I],
                                     &SystemZ::FP64BitRegClass);
        SDValue ArgValue = DAG.getCopyFromReg(Chain, DL, VReg, MVT::f64);
        MemOps[I] = DAG.getStore(ArgValue.getValue(1), DL, ArgValue, FIN,
                                 MachinePointerInfo::getFixedStack(MF, FI));
      }
      // Join the stores, which are independent of one another.
      Chain = DAG.getNode(ISD::TokenFactor, DL, MVT::Other,
                          makeArrayRef(&MemOps[NumFixedFPRs],
                                       SystemZ::NumArgFPRs-NumFixedFPRs));
    }
  }

  return Chain;
}

static bool canUseSiblingCall(const CCState &ArgCCInfo,
                              SmallVectorImpl<CCValAssign> &ArgLocs,
                              SmallVectorImpl<ISD::OutputArg> &Outs) {
  // Punt if there are any indirect or stack arguments, or if the call
  // needs the callee-saved argument register R6, or if the call uses
  // the callee-saved register arguments SwiftSelf and SwiftError.
  for (unsigned I = 0, E = ArgLocs.size(); I != E; ++I) {
    CCValAssign &VA = ArgLocs[I];
    if (VA.getLocInfo() == CCValAssign::Indirect)
      return false;
    if (!VA.isRegLoc())
      return false;
    Register Reg = VA.getLocReg();
    if (Reg == SystemZ::R6H || Reg == SystemZ::R6L || Reg == SystemZ::R6D)
      return false;
    if (Outs[I].Flags.isSwiftSelf() || Outs[I].Flags.isSwiftError())
      return false;
  }
  return true;
}

SDValue
SystemZTargetLowering::LowerCall(CallLoweringInfo &CLI,
                                 SmallVectorImpl<SDValue> &InVals) const {
  SelectionDAG &DAG = CLI.DAG;
  SDLoc &DL = CLI.DL;
  SmallVectorImpl<ISD::OutputArg> &Outs = CLI.Outs;
  SmallVectorImpl<SDValue> &OutVals = CLI.OutVals;
  SmallVectorImpl<ISD::InputArg> &Ins = CLI.Ins;
  SDValue Chain = CLI.Chain;
  SDValue Callee = CLI.Callee;
  bool &IsTailCall = CLI.IsTailCall;
  CallingConv::ID CallConv = CLI.CallConv;
  bool IsVarArg = CLI.IsVarArg;
  MachineFunction &MF = DAG.getMachineFunction();
  EVT PtrVT = getPointerTy(MF.getDataLayout());

  // Detect unsupported vector argument and return types.
  if (Subtarget.hasVector()) {
    VerifyVectorTypes(Outs);
    VerifyVectorTypes(Ins);
  }

  // Analyze the operands of the call, assigning locations to each operand.
  SmallVector<CCValAssign, 16> ArgLocs;
  SystemZCCState ArgCCInfo(CallConv, IsVarArg, MF, ArgLocs, *DAG.getContext());
  ArgCCInfo.AnalyzeCallOperands(Outs, CC_SystemZ);

  // We don't support GuaranteedTailCallOpt, only automatically-detected
  // sibling calls.
  if (IsTailCall && !canUseSiblingCall(ArgCCInfo, ArgLocs, Outs))
    IsTailCall = false;

  // Get a count of how many bytes are to be pushed on the stack.
  unsigned NumBytes = ArgCCInfo.getNextStackOffset();

  // Mark the start of the call.
  if (!IsTailCall)
    Chain = DAG.getCALLSEQ_START(Chain, NumBytes, 0, DL);

  // Copy argument values to their designated locations.
  SmallVector<std::pair<unsigned, SDValue>, 9> RegsToPass;
  SmallVector<SDValue, 8> MemOpChains;
  SDValue StackPtr;
  for (unsigned I = 0, E = ArgLocs.size(); I != E; ++I) {
    CCValAssign &VA = ArgLocs[I];
    SDValue ArgValue = OutVals[I];

    if (VA.getLocInfo() == CCValAssign::Indirect) {
      // Store the argument in a stack slot and pass its address.
      SDValue SpillSlot = DAG.CreateStackTemporary(Outs[I].ArgVT);
      int FI = cast<FrameIndexSDNode>(SpillSlot)->getIndex();
      MemOpChains.push_back(
          DAG.getStore(Chain, DL, ArgValue, SpillSlot,
                       MachinePointerInfo::getFixedStack(MF, FI)));
      // If the original argument was split (e.g. i128), we need
      // to store all parts of it here (and pass just one address).
      unsigned ArgIndex = Outs[I].OrigArgIndex;
      assert (Outs[I].PartOffset == 0);
      while (I + 1 != E && Outs[I + 1].OrigArgIndex == ArgIndex) {
        SDValue PartValue = OutVals[I + 1];
        unsigned PartOffset = Outs[I + 1].PartOffset;
        SDValue Address = DAG.getNode(ISD::ADD, DL, PtrVT, SpillSlot,
                                      DAG.getIntPtrConstant(PartOffset, DL));
        MemOpChains.push_back(
            DAG.getStore(Chain, DL, PartValue, Address,
                         MachinePointerInfo::getFixedStack(MF, FI)));
        ++I;
      }
      ArgValue = SpillSlot;
    } else
      ArgValue = convertValVTToLocVT(DAG, DL, VA, ArgValue);

    if (VA.isRegLoc())
      // Queue up the argument copies and emit them at the end.
      RegsToPass.push_back(std::make_pair(VA.getLocReg(), ArgValue));
    else {
      assert(VA.isMemLoc() && "Argument not register or memory");

      // Work out the address of the stack slot.  Unpromoted ints and
      // floats are passed as right-justified 8-byte values.
      if (!StackPtr.getNode())
        StackPtr = DAG.getCopyFromReg(Chain, DL, SystemZ::R15D, PtrVT);
      unsigned Offset = SystemZMC::CallFrameSize + VA.getLocMemOffset();
      if (VA.getLocVT() == MVT::i32 || VA.getLocVT() == MVT::f32)
        Offset += 4;
      SDValue Address = DAG.getNode(ISD::ADD, DL, PtrVT, StackPtr,
                                    DAG.getIntPtrConstant(Offset, DL));

      // Emit the store.
      MemOpChains.push_back(
          DAG.getStore(Chain, DL, ArgValue, Address, MachinePointerInfo()));
    }
  }

  // Join the stores, which are independent of one another.
  if (!MemOpChains.empty())
    Chain = DAG.getNode(ISD::TokenFactor, DL, MVT::Other, MemOpChains);

  // Accept direct calls by converting symbolic call addresses to the
  // associated Target* opcodes.  Force %r1 to be used for indirect
  // tail calls.
  SDValue Glue;
  if (auto *G = dyn_cast<GlobalAddressSDNode>(Callee)) {
    Callee = DAG.getTargetGlobalAddress(G->getGlobal(), DL, PtrVT);
    Callee = DAG.getNode(SystemZISD::PCREL_WRAPPER, DL, PtrVT, Callee);
  } else if (auto *E = dyn_cast<ExternalSymbolSDNode>(Callee)) {
    Callee = DAG.getTargetExternalSymbol(E->getSymbol(), PtrVT);
    Callee = DAG.getNode(SystemZISD::PCREL_WRAPPER, DL, PtrVT, Callee);
  } else if (IsTailCall) {
    Chain = DAG.getCopyToReg(Chain, DL, SystemZ::R1D, Callee, Glue);
    Glue = Chain.getValue(1);
    Callee = DAG.getRegister(SystemZ::R1D, Callee.getValueType());
  }

  // Build a sequence of copy-to-reg nodes, chained and glued together.
  for (unsigned I = 0, E = RegsToPass.size(); I != E; ++I) {
    Chain = DAG.getCopyToReg(Chain, DL, RegsToPass[I].first,
                             RegsToPass[I].second, Glue);
    Glue = Chain.getValue(1);
  }

  // The first call operand is the chain and the second is the target address.
  SmallVector<SDValue, 8> Ops;
  Ops.push_back(Chain);
  Ops.push_back(Callee);

  // Add argument registers to the end of the list so that they are
  // known live into the call.
  for (unsigned I = 0, E = RegsToPass.size(); I != E; ++I)
    Ops.push_back(DAG.getRegister(RegsToPass[I].first,
                                  RegsToPass[I].second.getValueType()));

  // Add a register mask operand representing the call-preserved registers.
  const TargetRegisterInfo *TRI = Subtarget.getRegisterInfo();
  const uint32_t *Mask = TRI->getCallPreservedMask(MF, CallConv);
  assert(Mask && "Missing call preserved mask for calling convention");
  Ops.push_back(DAG.getRegisterMask(Mask));

  // Glue the call to the argument copies, if any.
  if (Glue.getNode())
    Ops.push_back(Glue);

  // Emit the call.
  SDVTList NodeTys = DAG.getVTList(MVT::Other, MVT::Glue);
  if (IsTailCall)
    return DAG.getNode(SystemZISD::SIBCALL, DL, NodeTys, Ops);
  Chain = DAG.getNode(SystemZISD::CALL, DL, NodeTys, Ops);
  DAG.addNoMergeSiteInfo(Chain.getNode(), CLI.NoMerge);
  Glue = Chain.getValue(1);

  // Mark the end of the call, which is glued to the call itself.
  Chain = DAG.getCALLSEQ_END(Chain,
                             DAG.getConstant(NumBytes, DL, PtrVT, true),
                             DAG.getConstant(0, DL, PtrVT, true),
                             Glue, DL);
  Glue = Chain.getValue(1);

  // Assign locations to each value returned by this call.
  SmallVector<CCValAssign, 16> RetLocs;
  CCState RetCCInfo(CallConv, IsVarArg, MF, RetLocs, *DAG.getContext());
  RetCCInfo.AnalyzeCallResult(Ins, RetCC_SystemZ);

  // Copy all of the result registers out of their specified physreg.
  for (unsigned I = 0, E = RetLocs.size(); I != E; ++I) {
    CCValAssign &VA = RetLocs[I];

    // Copy the value out, gluing the copy to the end of the call sequence.
    SDValue RetValue = DAG.getCopyFromReg(Chain, DL, VA.getLocReg(),
                                          VA.getLocVT(), Glue);
    Chain = RetValue.getValue(1);
    Glue = RetValue.getValue(2);

    // Convert the value of the return register into the value that's
    // being returned.
    InVals.push_back(convertLocVTToValVT(DAG, DL, VA, Chain, RetValue));
  }

  return Chain;
}

bool SystemZTargetLowering::
CanLowerReturn(CallingConv::ID CallConv,
               MachineFunction &MF, bool isVarArg,
               const SmallVectorImpl<ISD::OutputArg> &Outs,
               LLVMContext &Context) const {
  // Detect unsupported vector return types.
  if (Subtarget.hasVector())
    VerifyVectorTypes(Outs);

  // Special case that we cannot easily detect in RetCC_SystemZ since
  // i128 is not a legal type.
  for (auto &Out : Outs)
    if (Out.ArgVT == MVT::i128)
      return false;

  SmallVector<CCValAssign, 16> RetLocs;
  CCState RetCCInfo(CallConv, isVarArg, MF, RetLocs, Context);
  return RetCCInfo.CheckReturn(Outs, RetCC_SystemZ);
}

SDValue
SystemZTargetLowering::LowerReturn(SDValue Chain, CallingConv::ID CallConv,
                                   bool IsVarArg,
                                   const SmallVectorImpl<ISD::OutputArg> &Outs,
                                   const SmallVectorImpl<SDValue> &OutVals,
                                   const SDLoc &DL, SelectionDAG &DAG) const {
  MachineFunction &MF = DAG.getMachineFunction();

  // Detect unsupported vector return types.
  if (Subtarget.hasVector())
    VerifyVectorTypes(Outs);

  // Assign locations to each returned value.
  SmallVector<CCValAssign, 16> RetLocs;
  CCState RetCCInfo(CallConv, IsVarArg, MF, RetLocs, *DAG.getContext());
  RetCCInfo.AnalyzeReturn(Outs, RetCC_SystemZ);

  // Quick exit for void returns
  if (RetLocs.empty())
    return DAG.getNode(SystemZISD::RET_FLAG, DL, MVT::Other, Chain);

  if (CallConv == CallingConv::GHC)
    report_fatal_error("GHC functions return void only");

  // Copy the result values into the output registers.
  SDValue Glue;
  SmallVector<SDValue, 4> RetOps;
  RetOps.push_back(Chain);
  for (unsigned I = 0, E = RetLocs.size(); I != E; ++I) {
    CCValAssign &VA = RetLocs[I];
    SDValue RetValue = OutVals[I];

    // Make the return register live on exit.
    assert(VA.isRegLoc() && "Can only return in registers!");

    // Promote the value as required.
    RetValue = convertValVTToLocVT(DAG, DL, VA, RetValue);

    // Chain and glue the copies together.
    Register Reg = VA.getLocReg();
    Chain = DAG.getCopyToReg(Chain, DL, Reg, RetValue, Glue);
    Glue = Chain.getValue(1);
    RetOps.push_back(DAG.getRegister(Reg, VA.getLocVT()));
  }

  // Update chain and glue.
  RetOps[0] = Chain;
  if (Glue.getNode())
    RetOps.push_back(Glue);

  return DAG.getNode(SystemZISD::RET_FLAG, DL, MVT::Other, RetOps);
}

// Return true if Op is an intrinsic node with chain that returns the CC value
// as its only (other) argument.  Provide the associated SystemZISD opcode and
// the mask of valid CC values if so.
static bool isIntrinsicWithCCAndChain(SDValue Op, unsigned &Opcode,
                                      unsigned &CCValid) {
  unsigned Id = cast<ConstantSDNode>(Op.getOperand(1))->getZExtValue();
  switch (Id) {
  case Intrinsic::s390_tbegin:
    Opcode = SystemZISD::TBEGIN;
    CCValid = SystemZ::CCMASK_TBEGIN;
    return true;

  case Intrinsic::s390_tbegin_nofloat:
    Opcode = SystemZISD::TBEGIN_NOFLOAT;
    CCValid = SystemZ::CCMASK_TBEGIN;
    return true;

  case Intrinsic::s390_tend:
    Opcode = SystemZISD::TEND;
    CCValid = SystemZ::CCMASK_TEND;
    return true;

  default:
    return false;
  }
}

// Return true if Op is an intrinsic node without chain that returns the
// CC value as its final argument.  Provide the associated SystemZISD
// opcode and the mask of valid CC values if so.
static bool isIntrinsicWithCC(SDValue Op, unsigned &Opcode, unsigned &CCValid) {
  unsigned Id = cast<ConstantSDNode>(Op.getOperand(0))->getZExtValue();
  switch (Id) {
  case Intrinsic::s390_vpkshs:
  case Intrinsic::s390_vpksfs:
  case Intrinsic::s390_vpksgs:
    Opcode = SystemZISD::PACKS_CC;
    CCValid = SystemZ::CCMASK_VCMP;
    return true;

  case Intrinsic::s390_vpklshs:
  case Intrinsic::s390_vpklsfs:
  case Intrinsic::s390_vpklsgs:
    Opcode = SystemZISD::PACKLS_CC;
    CCValid = SystemZ::CCMASK_VCMP;
    return true;

  case Intrinsic::s390_vceqbs:
  case Intrinsic::s390_vceqhs:
  case Intrinsic::s390_vceqfs:
  case Intrinsic::s390_vceqgs:
    Opcode = SystemZISD::VICMPES;
    CCValid = SystemZ::CCMASK_VCMP;
    return true;

  case Intrinsic::s390_vchbs:
  case Intrinsic::s390_vchhs:
  case Intrinsic::s390_vchfs:
  case Intrinsic::s390_vchgs:
    Opcode = SystemZISD::VICMPHS;
    CCValid = SystemZ::CCMASK_VCMP;
    return true;

  case Intrinsic::s390_vchlbs:
  case Intrinsic::s390_vchlhs:
  case Intrinsic::s390_vchlfs:
  case Intrinsic::s390_vchlgs:
    Opcode = SystemZISD::VICMPHLS;
    CCValid = SystemZ::CCMASK_VCMP;
    return true;

  case Intrinsic::s390_vtm:
    Opcode = SystemZISD::VTM;
    CCValid = SystemZ::CCMASK_VCMP;
    return true;

  case Intrinsic::s390_vfaebs:
  case Intrinsic::s390_vfaehs:
  case Intrinsic::s390_vfaefs:
    Opcode = SystemZISD::VFAE_CC;
    CCValid = SystemZ::CCMASK_ANY;
    return true;

  case Intrinsic::s390_vfaezbs:
  case Intrinsic::s390_vfaezhs:
  case Intrinsic::s390_vfaezfs:
    Opcode = SystemZISD::VFAEZ_CC;
    CCValid = SystemZ::CCMASK_ANY;
    return true;

  case Intrinsic::s390_vfeebs:
  case Intrinsic::s390_vfeehs:
  case Intrinsic::s390_vfeefs:
    Opcode = SystemZISD::VFEE_CC;
    CCValid = SystemZ::CCMASK_ANY;
    return true;

  case Intrinsic::s390_vfeezbs:
  case Intrinsic::s390_vfeezhs:
  case Intrinsic::s390_vfeezfs:
    Opcode = SystemZISD::VFEEZ_CC;
    CCValid = SystemZ::CCMASK_ANY;
    return true;

  case Intrinsic::s390_vfenebs:
  case Intrinsic::s390_vfenehs:
  case Intrinsic::s390_vfenefs:
    Opcode = SystemZISD::VFENE_CC;
    CCValid = SystemZ::CCMASK_ANY;
    return true;

  case Intrinsic::s390_vfenezbs:
  case Intrinsic::s390_vfenezhs:
  case Intrinsic::s390_vfenezfs:
    Opcode = SystemZISD::VFENEZ_CC;
    CCValid = SystemZ::CCMASK_ANY;
    return true;

  case Intrinsic::s390_vistrbs:
  case Intrinsic::s390_vistrhs:
  case Intrinsic::s390_vistrfs:
    Opcode = SystemZISD::VISTR_CC;
    CCValid = SystemZ::CCMASK_0 | SystemZ::CCMASK_3;
    return true;

  case Intrinsic::s390_vstrcbs:
  case Intrinsic::s390_vstrchs:
  case Intrinsic::s390_vstrcfs:
    Opcode = SystemZISD::VSTRC_CC;
    CCValid = SystemZ::CCMASK_ANY;
    return true;

  case Intrinsic::s390_vstrczbs:
  case Intrinsic::s390_vstrczhs:
  case Intrinsic::s390_vstrczfs:
    Opcode = SystemZISD::VSTRCZ_CC;
    CCValid = SystemZ::CCMASK_ANY;
    return true;

  case Intrinsic::s390_vstrsb:
  case Intrinsic::s390_vstrsh:
  case Intrinsic::s390_vstrsf:
    Opcode = SystemZISD::VSTRS_CC;
    CCValid = SystemZ::CCMASK_ANY;
    return true;

  case Intrinsic::s390_vstrszb:
  case Intrinsic::s390_vstrszh:
  case Intrinsic::s390_vstrszf:
    Opcode = SystemZISD::VSTRSZ_CC;
    CCValid = SystemZ::CCMASK_ANY;
    return true;

  case Intrinsic::s390_vfcedbs:
  case Intrinsic::s390_vfcesbs:
    Opcode = SystemZISD::VFCMPES;
    CCValid = SystemZ::CCMASK_VCMP;
    return true;

  case Intrinsic::s390_vfchdbs:
  case Intrinsic::s390_vfchsbs:
    Opcode = SystemZISD::VFCMPHS;
    CCValid = SystemZ::CCMASK_VCMP;
    return true;

  case Intrinsic::s390_vfchedbs:
  case Intrinsic::s390_vfchesbs:
    Opcode = SystemZISD::VFCMPHES;
    CCValid = SystemZ::CCMASK_VCMP;
    return true;

  case Intrinsic::s390_vftcidb:
  case Intrinsic::s390_vftcisb:
    Opcode = SystemZISD::VFTCI;
    CCValid = SystemZ::CCMASK_VCMP;
    return true;

  case Intrinsic::s390_tdc:
    Opcode = SystemZISD::TDC;
    CCValid = SystemZ::CCMASK_TDC;
    return true;

  default:
    return false;
  }
}

// Emit an intrinsic with chain and an explicit CC register result.
static SDNode *emitIntrinsicWithCCAndChain(SelectionDAG &DAG, SDValue Op,
                                           unsigned Opcode) {
  // Copy all operands except the intrinsic ID.
  unsigned NumOps = Op.getNumOperands();
  SmallVector<SDValue, 6> Ops;
  Ops.reserve(NumOps - 1);
  Ops.push_back(Op.getOperand(0));
  for (unsigned I = 2; I < NumOps; ++I)
    Ops.push_back(Op.getOperand(I));

  assert(Op->getNumValues() == 2 && "Expected only CC result and chain");
  SDVTList RawVTs = DAG.getVTList(MVT::i32, MVT::Other);
  SDValue Intr = DAG.getNode(Opcode, SDLoc(Op), RawVTs, Ops);
  SDValue OldChain = SDValue(Op.getNode(), 1);
  SDValue NewChain = SDValue(Intr.getNode(), 1);
  DAG.ReplaceAllUsesOfValueWith(OldChain, NewChain);
  return Intr.getNode();
}

// Emit an intrinsic with an explicit CC register result.
static SDNode *emitIntrinsicWithCC(SelectionDAG &DAG, SDValue Op,
                                   unsigned Opcode) {
  // Copy all operands except the intrinsic ID.
  unsigned NumOps = Op.getNumOperands();
  SmallVector<SDValue, 6> Ops;
  Ops.reserve(NumOps - 1);
  for (unsigned I = 1; I < NumOps; ++I)
    Ops.push_back(Op.getOperand(I));

  SDValue Intr = DAG.getNode(Opcode, SDLoc(Op), Op->getVTList(), Ops);
  return Intr.getNode();
}

// CC is a comparison that will be implemented using an integer or
// floating-point comparison.  Return the condition code mask for
// a branch on true.  In the integer case, CCMASK_CMP_UO is set for
// unsigned comparisons and clear for signed ones.  In the floating-point
// case, CCMASK_CMP_UO has its normal mask meaning (unordered).
static unsigned CCMaskForCondCode(ISD::CondCode CC) {
#define CONV(X) \
  case ISD::SET##X: return SystemZ::CCMASK_CMP_##X; \
  case ISD::SETO##X: return SystemZ::CCMASK_CMP_##X; \
  case ISD::SETU##X: return SystemZ::CCMASK_CMP_UO | SystemZ::CCMASK_CMP_##X

  switch (CC) {
  default:
    llvm_unreachable("Invalid integer condition!");

  CONV(EQ);
  CONV(NE);
  CONV(GT);
  CONV(GE);
  CONV(LT);
  CONV(LE);

  case ISD::SETO:  return SystemZ::CCMASK_CMP_O;
  case ISD::SETUO: return SystemZ::CCMASK_CMP_UO;
  }
#undef CONV
}

// If C can be converted to a comparison against zero, adjust the operands
// as necessary.
static void adjustZeroCmp(SelectionDAG &DAG, const SDLoc &DL, Comparison &C) {
  if (C.ICmpType == SystemZICMP::UnsignedOnly)
    return;

  auto *ConstOp1 = dyn_cast<ConstantSDNode>(C.Op1.getNode());
  if (!ConstOp1)
    return;

  int64_t Value = ConstOp1->getSExtValue();
  if ((Value == -1 && C.CCMask == SystemZ::CCMASK_CMP_GT) ||
      (Value == -1 && C.CCMask == SystemZ::CCMASK_CMP_LE) ||
      (Value == 1 && C.CCMask == SystemZ::CCMASK_CMP_LT) ||
      (Value == 1 && C.CCMask == SystemZ::CCMASK_CMP_GE)) {
    C.CCMask ^= SystemZ::CCMASK_CMP_EQ;
    C.Op1 = DAG.getConstant(0, DL, C.Op1.getValueType());
  }
}

// If a comparison described by C is suitable for CLI(Y), CHHSI or CLHHSI,
// adjust the operands as necessary.
static void adjustSubwordCmp(SelectionDAG &DAG, const SDLoc &DL,
                             Comparison &C) {
  // For us to make any changes, it must a comparison between a single-use
  // load and a constant.
  if (!C.Op0.hasOneUse() ||
      C.Op0.getOpcode() != ISD::LOAD ||
      C.Op1.getOpcode() != ISD::Constant)
    return;

  // We must have an 8- or 16-bit load.
  auto *Load = cast<LoadSDNode>(C.Op0);
  unsigned NumBits = Load->getMemoryVT().getSizeInBits();
  if ((NumBits != 8 && NumBits != 16) ||
      NumBits != Load->getMemoryVT().getStoreSizeInBits())
    return;

  // The load must be an extending one and the constant must be within the
  // range of the unextended value.
  auto *ConstOp1 = cast<ConstantSDNode>(C.Op1);
  uint64_t Value = ConstOp1->getZExtValue();
  uint64_t Mask = (1 << NumBits) - 1;
  if (Load->getExtensionType() == ISD::SEXTLOAD) {
    // Make sure that ConstOp1 is in range of C.Op0.
    int64_t SignedValue = ConstOp1->getSExtValue();
    if (uint64_t(SignedValue) + (uint64_t(1) << (NumBits - 1)) > Mask)
      return;
    if (C.ICmpType != SystemZICMP::SignedOnly) {
      // Unsigned comparison between two sign-extended values is equivalent
      // to unsigned comparison between two zero-extended values.
      Value &= Mask;
    } else if (NumBits == 8) {
      // Try to treat the comparison as unsigned, so that we can use CLI.
      // Adjust CCMask and Value as necessary.
      if (Value == 0 && C.CCMask == SystemZ::CCMASK_CMP_LT)
        // Test whether the high bit of the byte is set.
        Value = 127, C.CCMask = SystemZ::CCMASK_CMP_GT;
      else if (Value == 0 && C.CCMask == SystemZ::CCMASK_CMP_GE)
        // Test whether the high bit of the byte is clear.
        Value = 128, C.CCMask = SystemZ::CCMASK_CMP_LT;
      else
        // No instruction exists for this combination.
        return;
      C.ICmpType = SystemZICMP::UnsignedOnly;
    }
  } else if (Load->getExtensionType() == ISD::ZEXTLOAD) {
    if (Value > Mask)
      return;
    // If the constant is in range, we can use any comparison.
    C.ICmpType = SystemZICMP::Any;
  } else
    return;

  // Make sure that the first operand is an i32 of the right extension type.
  ISD::LoadExtType ExtType = (C.ICmpType == SystemZICMP::SignedOnly ?
                              ISD::SEXTLOAD :
                              ISD::ZEXTLOAD);
  if (C.Op0.getValueType() != MVT::i32 ||
      Load->getExtensionType() != ExtType) {
    C.Op0 = DAG.getExtLoad(ExtType, SDLoc(Load), MVT::i32, Load->getChain(),
                           Load->getBasePtr(), Load->getPointerInfo(),
                           Load->getMemoryVT(), Load->getAlignment(),
                           Load->getMemOperand()->getFlags());
    // Update the chain uses.
    DAG.ReplaceAllUsesOfValueWith(SDValue(Load, 1), C.Op0.getValue(1));
  }

  // Make sure that the second operand is an i32 with the right value.
  if (C.Op1.getValueType() != MVT::i32 ||
      Value != ConstOp1->getZExtValue())
    C.Op1 = DAG.getConstant(Value, DL, MVT::i32);
}

// Return true if Op is either an unextended load, or a load suitable
// for integer register-memory comparisons of type ICmpType.
static bool isNaturalMemoryOperand(SDValue Op, unsigned ICmpType) {
  auto *Load = dyn_cast<LoadSDNode>(Op.getNode());
  if (Load) {
    // There are no instructions to compare a register with a memory byte.
    if (Load->getMemoryVT() == MVT::i8)
      return false;
    // Otherwise decide on extension type.
    switch (Load->getExtensionType()) {
    case ISD::NON_EXTLOAD:
      return true;
    case ISD::SEXTLOAD:
      return ICmpType != SystemZICMP::UnsignedOnly;
    case ISD::ZEXTLOAD:
      return ICmpType != SystemZICMP::SignedOnly;
    default:
      break;
    }
  }
  return false;
}

// Return true if it is better to swap the operands of C.
static bool shouldSwapCmpOperands(const Comparison &C) {
  // Leave f128 comparisons alone, since they have no memory forms.
  if (C.Op0.getValueType() == MVT::f128)
    return false;

  // Always keep a floating-point constant second, since comparisons with
  // zero can use LOAD TEST and comparisons with other constants make a
  // natural memory operand.
  if (isa<ConstantFPSDNode>(C.Op1))
    return false;

  // Never swap comparisons with zero since there are many ways to optimize
  // those later.
  auto *ConstOp1 = dyn_cast<ConstantSDNode>(C.Op1);
  if (ConstOp1 && ConstOp1->getZExtValue() == 0)
    return false;

  // Also keep natural memory operands second if the loaded value is
  // only used here.  Several comparisons have memory forms.
  if (isNaturalMemoryOperand(C.Op1, C.ICmpType) && C.Op1.hasOneUse())
    return false;

  // Look for cases where Cmp0 is a single-use load and Cmp1 isn't.
  // In that case we generally prefer the memory to be second.
  if (isNaturalMemoryOperand(C.Op0, C.ICmpType) && C.Op0.hasOneUse()) {
    // The only exceptions are when the second operand is a constant and
    // we can use things like CHHSI.
    if (!ConstOp1)
      return true;
    // The unsigned memory-immediate instructions can handle 16-bit
    // unsigned integers.
    if (C.ICmpType != SystemZICMP::SignedOnly &&
        isUInt<16>(ConstOp1->getZExtValue()))
      return false;
    // The signed memory-immediate instructions can handle 16-bit
    // signed integers.
    if (C.ICmpType != SystemZICMP::UnsignedOnly &&
        isInt<16>(ConstOp1->getSExtValue()))
      return false;
    return true;
  }

  // Try to promote the use of CGFR and CLGFR.
  unsigned Opcode0 = C.Op0.getOpcode();
  if (C.ICmpType != SystemZICMP::UnsignedOnly && Opcode0 == ISD::SIGN_EXTEND)
    return true;
  if (C.ICmpType != SystemZICMP::SignedOnly && Opcode0 == ISD::ZERO_EXTEND)
    return true;
  if (C.ICmpType != SystemZICMP::SignedOnly &&
      Opcode0 == ISD::AND &&
      C.Op0.getOperand(1).getOpcode() == ISD::Constant &&
      cast<ConstantSDNode>(C.Op0.getOperand(1))->getZExtValue() == 0xffffffff)
    return true;

  return false;
}

// Check whether C tests for equality between X and Y and whether X - Y
// or Y - X is also computed.  In that case it's better to compare the
// result of the subtraction against zero.
static void adjustForSubtraction(SelectionDAG &DAG, const SDLoc &DL,
                                 Comparison &C) {
  if (C.CCMask == SystemZ::CCMASK_CMP_EQ ||
      C.CCMask == SystemZ::CCMASK_CMP_NE) {
    for (auto I = C.Op0->use_begin(), E = C.Op0->use_end(); I != E; ++I) {
      SDNode *N = *I;
      if (N->getOpcode() == ISD::SUB &&
          ((N->getOperand(0) == C.Op0 && N->getOperand(1) == C.Op1) ||
           (N->getOperand(0) == C.Op1 && N->getOperand(1) == C.Op0))) {
        C.Op0 = SDValue(N, 0);
        C.Op1 = DAG.getConstant(0, DL, N->getValueType(0));
        return;
      }
    }
  }
}

// Check whether C compares a floating-point value with zero and if that
// floating-point value is also negated.  In this case we can use the
// negation to set CC, so avoiding separate LOAD AND TEST and
// LOAD (NEGATIVE/COMPLEMENT) instructions.
static void adjustForFNeg(Comparison &C) {
  // This optimization is invalid for strict comparisons, since FNEG
  // does not raise any exceptions.
  if (C.Chain)
    return;
  auto *C1 = dyn_cast<ConstantFPSDNode>(C.Op1);
  if (C1 && C1->isZero()) {
    for (auto I = C.Op0->use_begin(), E = C.Op0->use_end(); I != E; ++I) {
      SDNode *N = *I;
      if (N->getOpcode() == ISD::FNEG) {
        C.Op0 = SDValue(N, 0);
        C.CCMask = SystemZ::reverseCCMask(C.CCMask);
        return;
      }
    }
  }
}

// Check whether C compares (shl X, 32) with 0 and whether X is
// also sign-extended.  In that case it is better to test the result
// of the sign extension using LTGFR.
//
// This case is important because InstCombine transforms a comparison
// with (sext (trunc X)) into a comparison with (shl X, 32).
static void adjustForLTGFR(Comparison &C) {
  // Check for a comparison between (shl X, 32) and 0.
  if (C.Op0.getOpcode() == ISD::SHL &&
      C.Op0.getValueType() == MVT::i64 &&
      C.Op1.getOpcode() == ISD::Constant &&
      cast<ConstantSDNode>(C.Op1)->getZExtValue() == 0) {
    auto *C1 = dyn_cast<ConstantSDNode>(C.Op0.getOperand(1));
    if (C1 && C1->getZExtValue() == 32) {
      SDValue ShlOp0 = C.Op0.getOperand(0);
      // See whether X has any SIGN_EXTEND_INREG uses.
      for (auto I = ShlOp0->use_begin(), E = ShlOp0->use_end(); I != E; ++I) {
        SDNode *N = *I;
        if (N->getOpcode() == ISD::SIGN_EXTEND_INREG &&
            cast<VTSDNode>(N->getOperand(1))->getVT() == MVT::i32) {
          C.Op0 = SDValue(N, 0);
          return;
        }
      }
    }
  }
}

// If C compares the truncation of an extending load, try to compare
// the untruncated value instead.  This exposes more opportunities to
// reuse CC.
static void adjustICmpTruncate(SelectionDAG &DAG, const SDLoc &DL,
                               Comparison &C) {
  if (C.Op0.getOpcode() == ISD::TRUNCATE &&
      C.Op0.getOperand(0).getOpcode() == ISD::LOAD &&
      C.Op1.getOpcode() == ISD::Constant &&
      cast<ConstantSDNode>(C.Op1)->getZExtValue() == 0) {
    auto *L = cast<LoadSDNode>(C.Op0.getOperand(0));
    if (L->getMemoryVT().getStoreSizeInBits() <= C.Op0.getValueSizeInBits()) {
      unsigned Type = L->getExtensionType();
      if ((Type == ISD::ZEXTLOAD && C.ICmpType != SystemZICMP::SignedOnly) ||
          (Type == ISD::SEXTLOAD && C.ICmpType != SystemZICMP::UnsignedOnly)) {
        C.Op0 = C.Op0.getOperand(0);
        C.Op1 = DAG.getConstant(0, DL, C.Op0.getValueType());
      }
    }
  }
}

// Return true if shift operation N has an in-range constant shift value.
// Store it in ShiftVal if so.
static bool isSimpleShift(SDValue N, unsigned &ShiftVal) {
  auto *Shift = dyn_cast<ConstantSDNode>(N.getOperand(1));
  if (!Shift)
    return false;

  uint64_t Amount = Shift->getZExtValue();
  if (Amount >= N.getValueSizeInBits())
    return false;

  ShiftVal = Amount;
  return true;
}

// Check whether an AND with Mask is suitable for a TEST UNDER MASK
// instruction and whether the CC value is descriptive enough to handle
// a comparison of type Opcode between the AND result and CmpVal.
// CCMask says which comparison result is being tested and BitSize is
// the number of bits in the operands.  If TEST UNDER MASK can be used,
// return the corresponding CC mask, otherwise return 0.
static unsigned getTestUnderMaskCond(unsigned BitSize, unsigned CCMask,
                                     uint64_t Mask, uint64_t CmpVal,
                                     unsigned ICmpType) {
  assert(Mask != 0 && "ANDs with zero should have been removed by now");

  // Check whether the mask is suitable for TMHH, TMHL, TMLH or TMLL.
  if (!SystemZ::isImmLL(Mask) && !SystemZ::isImmLH(Mask) &&
      !SystemZ::isImmHL(Mask) && !SystemZ::isImmHH(Mask))
    return 0;

  // Work out the masks for the lowest and highest bits.
  unsigned HighShift = 63 - countLeadingZeros(Mask);
  uint64_t High = uint64_t(1) << HighShift;
  uint64_t Low = uint64_t(1) << countTrailingZeros(Mask);

  // Signed ordered comparisons are effectively unsigned if the sign
  // bit is dropped.
  bool EffectivelyUnsigned = (ICmpType != SystemZICMP::SignedOnly);

  // Check for equality comparisons with 0, or the equivalent.
  if (CmpVal == 0) {
    if (CCMask == SystemZ::CCMASK_CMP_EQ)
      return SystemZ::CCMASK_TM_ALL_0;
    if (CCMask == SystemZ::CCMASK_CMP_NE)
      return SystemZ::CCMASK_TM_SOME_1;
  }
  if (EffectivelyUnsigned && CmpVal > 0 && CmpVal <= Low) {
    if (CCMask == SystemZ::CCMASK_CMP_LT)
      return SystemZ::CCMASK_TM_ALL_0;
    if (CCMask == SystemZ::CCMASK_CMP_GE)
      return SystemZ::CCMASK_TM_SOME_1;
  }
  if (EffectivelyUnsigned && CmpVal < Low) {
    if (CCMask == SystemZ::CCMASK_CMP_LE)
      return SystemZ::CCMASK_TM_ALL_0;
    if (CCMask == SystemZ::CCMASK_CMP_GT)
      return SystemZ::CCMASK_TM_SOME_1;
  }

  // Check for equality comparisons with the mask, or the equivalent.
  if (CmpVal == Mask) {
    if (CCMask == SystemZ::CCMASK_CMP_EQ)
      return SystemZ::CCMASK_TM_ALL_1;
    if (CCMask == SystemZ::CCMASK_CMP_NE)
      return SystemZ::CCMASK_TM_SOME_0;
  }
  if (EffectivelyUnsigned && CmpVal >= Mask - Low && CmpVal < Mask) {
    if (CCMask == SystemZ::CCMASK_CMP_GT)
      return SystemZ::CCMASK_TM_ALL_1;
    if (CCMask == SystemZ::CCMASK_CMP_LE)
      return SystemZ::CCMASK_TM_SOME_0;
  }
  if (EffectivelyUnsigned && CmpVal > Mask - Low && CmpVal <= Mask) {
    if (CCMask == SystemZ::CCMASK_CMP_GE)
      return SystemZ::CCMASK_TM_ALL_1;
    if (CCMask == SystemZ::CCMASK_CMP_LT)
      return SystemZ::CCMASK_TM_SOME_0;
  }

  // Check for ordered comparisons with the top bit.
  if (EffectivelyUnsigned && CmpVal >= Mask - High && CmpVal < High) {
    if (CCMask == SystemZ::CCMASK_CMP_LE)
      return SystemZ::CCMASK_TM_MSB_0;
    if (CCMask == SystemZ::CCMASK_CMP_GT)
      return SystemZ::CCMASK_TM_MSB_1;
  }
  if (EffectivelyUnsigned && CmpVal > Mask - High && CmpVal <= High) {
    if (CCMask == SystemZ::CCMASK_CMP_LT)
      return SystemZ::CCMASK_TM_MSB_0;
    if (CCMask == SystemZ::CCMASK_CMP_GE)
      return SystemZ::CCMASK_TM_MSB_1;
  }

  // If there are just two bits, we can do equality checks for Low and High
  // as well.
  if (Mask == Low + High) {
    if (CCMask == SystemZ::CCMASK_CMP_EQ && CmpVal == Low)
      return SystemZ::CCMASK_TM_MIXED_MSB_0;
    if (CCMask == SystemZ::CCMASK_CMP_NE && CmpVal == Low)
      return SystemZ::CCMASK_TM_MIXED_MSB_0 ^ SystemZ::CCMASK_ANY;
    if (CCMask == SystemZ::CCMASK_CMP_EQ && CmpVal == High)
      return SystemZ::CCMASK_TM_MIXED_MSB_1;
    if (CCMask == SystemZ::CCMASK_CMP_NE && CmpVal == High)
      return SystemZ::CCMASK_TM_MIXED_MSB_1 ^ SystemZ::CCMASK_ANY;
  }

  // Looks like we've exhausted our options.
  return 0;
}

// See whether C can be implemented as a TEST UNDER MASK instruction.
// Update the arguments with the TM version if so.
static void adjustForTestUnderMask(SelectionDAG &DAG, const SDLoc &DL,
                                   Comparison &C) {
  // Check that we have a comparison with a constant.
  auto *ConstOp1 = dyn_cast<ConstantSDNode>(C.Op1);
  if (!ConstOp1)
    return;
  uint64_t CmpVal = ConstOp1->getZExtValue();

  // Check whether the nonconstant input is an AND with a constant mask.
  Comparison NewC(C);
  uint64_t MaskVal;
  ConstantSDNode *Mask = nullptr;
  if (C.Op0.getOpcode() == ISD::AND) {
    NewC.Op0 = C.Op0.getOperand(0);
    NewC.Op1 = C.Op0.getOperand(1);
    Mask = dyn_cast<ConstantSDNode>(NewC.Op1);
    if (!Mask)
      return;
    MaskVal = Mask->getZExtValue();
  } else {
    // There is no instruction to compare with a 64-bit immediate
    // so use TMHH instead if possible.  We need an unsigned ordered
    // comparison with an i64 immediate.
    if (NewC.Op0.getValueType() != MVT::i64 ||
        NewC.CCMask == SystemZ::CCMASK_CMP_EQ ||
        NewC.CCMask == SystemZ::CCMASK_CMP_NE ||
        NewC.ICmpType == SystemZICMP::SignedOnly)
      return;
    // Convert LE and GT comparisons into LT and GE.
    if (NewC.CCMask == SystemZ::CCMASK_CMP_LE ||
        NewC.CCMask == SystemZ::CCMASK_CMP_GT) {
      if (CmpVal == uint64_t(-1))
        return;
      CmpVal += 1;
      NewC.CCMask ^= SystemZ::CCMASK_CMP_EQ;
    }
    // If the low N bits of Op1 are zero than the low N bits of Op0 can
    // be masked off without changing the result.
    MaskVal = -(CmpVal & -CmpVal);
    NewC.ICmpType = SystemZICMP::UnsignedOnly;
  }
  if (!MaskVal)
    return;

  // Check whether the combination of mask, comparison value and comparison
  // type are suitable.
  unsigned BitSize = NewC.Op0.getValueSizeInBits();
  unsigned NewCCMask, ShiftVal;
  if (NewC.ICmpType != SystemZICMP::SignedOnly &&
      NewC.Op0.getOpcode() == ISD::SHL &&
      isSimpleShift(NewC.Op0, ShiftVal) &&
      (MaskVal >> ShiftVal != 0) &&
      ((CmpVal >> ShiftVal) << ShiftVal) == CmpVal &&
      (NewCCMask = getTestUnderMaskCond(BitSize, NewC.CCMask,
                                        MaskVal >> ShiftVal,
                                        CmpVal >> ShiftVal,
                                        SystemZICMP::Any))) {
    NewC.Op0 = NewC.Op0.getOperand(0);
    MaskVal >>= ShiftVal;
  } else if (NewC.ICmpType != SystemZICMP::SignedOnly &&
             NewC.Op0.getOpcode() == ISD::SRL &&
             isSimpleShift(NewC.Op0, ShiftVal) &&
             (MaskVal << ShiftVal != 0) &&
             ((CmpVal << ShiftVal) >> ShiftVal) == CmpVal &&
             (NewCCMask = getTestUnderMaskCond(BitSize, NewC.CCMask,
                                               MaskVal << ShiftVal,
                                               CmpVal << ShiftVal,
                                               SystemZICMP::UnsignedOnly))) {
    NewC.Op0 = NewC.Op0.getOperand(0);
    MaskVal <<= ShiftVal;
  } else {
    NewCCMask = getTestUnderMaskCond(BitSize, NewC.CCMask, MaskVal, CmpVal,
                                     NewC.ICmpType);
    if (!NewCCMask)
      return;
  }

  // Go ahead and make the change.
  C.Opcode = SystemZISD::TM;
  C.Op0 = NewC.Op0;
  if (Mask && Mask->getZExtValue() == MaskVal)
    C.Op1 = SDValue(Mask, 0);
  else
    C.Op1 = DAG.getConstant(MaskVal, DL, C.Op0.getValueType());
  C.CCValid = SystemZ::CCMASK_TM;
  C.CCMask = NewCCMask;
}

// See whether the comparison argument contains a redundant AND
// and remove it if so.  This sometimes happens due to the generic
// BRCOND expansion.
static void adjustForRedundantAnd(SelectionDAG &DAG, const SDLoc &DL,
                                  Comparison &C) {
  if (C.Op0.getOpcode() != ISD::AND)
    return;
  auto *Mask = dyn_cast<ConstantSDNode>(C.Op0.getOperand(1));
  if (!Mask)
    return;
  KnownBits Known = DAG.computeKnownBits(C.Op0.getOperand(0));
  if ((~Known.Zero).getZExtValue() & ~Mask->getZExtValue())
    return;

  C.Op0 = C.Op0.getOperand(0);
}

// Return a Comparison that tests the condition-code result of intrinsic
// node Call against constant integer CC using comparison code Cond.
// Opcode is the opcode of the SystemZISD operation for the intrinsic
// and CCValid is the set of possible condition-code results.
static Comparison getIntrinsicCmp(SelectionDAG &DAG, unsigned Opcode,
                                  SDValue Call, unsigned CCValid, uint64_t CC,
                                  ISD::CondCode Cond) {
  Comparison C(Call, SDValue(), SDValue());
  C.Opcode = Opcode;
  C.CCValid = CCValid;
  if (Cond == ISD::SETEQ)
    // bit 3 for CC==0, bit 0 for CC==3, always false for CC>3.
    C.CCMask = CC < 4 ? 1 << (3 - CC) : 0;
  else if (Cond == ISD::SETNE)
    // ...and the inverse of that.
    C.CCMask = CC < 4 ? ~(1 << (3 - CC)) : -1;
  else if (Cond == ISD::SETLT || Cond == ISD::SETULT)
    // bits above bit 3 for CC==0 (always false), bits above bit 0 for CC==3,
    // always true for CC>3.
    C.CCMask = CC < 4 ? ~0U << (4 - CC) : -1;
  else if (Cond == ISD::SETGE || Cond == ISD::SETUGE)
    // ...and the inverse of that.
    C.CCMask = CC < 4 ? ~(~0U << (4 - CC)) : 0;
  else if (Cond == ISD::SETLE || Cond == ISD::SETULE)
    // bit 3 and above for CC==0, bit 0 and above for CC==3 (always true),
    // always true for CC>3.
    C.CCMask = CC < 4 ? ~0U << (3 - CC) : -1;
  else if (Cond == ISD::SETGT || Cond == ISD::SETUGT)
    // ...and the inverse of that.
    C.CCMask = CC < 4 ? ~(~0U << (3 - CC)) : 0;
  else
    llvm_unreachable("Unexpected integer comparison type");
  C.CCMask &= CCValid;
  return C;
}

// Decide how to implement a comparison of type Cond between CmpOp0 with CmpOp1.
static Comparison getCmp(SelectionDAG &DAG, SDValue CmpOp0, SDValue CmpOp1,
                         ISD::CondCode Cond, const SDLoc &DL,
                         SDValue Chain = SDValue(),
                         bool IsSignaling = false) {
  if (CmpOp1.getOpcode() == ISD::Constant) {
    assert(!Chain);
    uint64_t Constant = cast<ConstantSDNode>(CmpOp1)->getZExtValue();
    unsigned Opcode, CCValid;
    if (CmpOp0.getOpcode() == ISD::INTRINSIC_W_CHAIN &&
        CmpOp0.getResNo() == 0 && CmpOp0->hasNUsesOfValue(1, 0) &&
        isIntrinsicWithCCAndChain(CmpOp0, Opcode, CCValid))
      return getIntrinsicCmp(DAG, Opcode, CmpOp0, CCValid, Constant, Cond);
    if (CmpOp0.getOpcode() == ISD::INTRINSIC_WO_CHAIN &&
        CmpOp0.getResNo() == CmpOp0->getNumValues() - 1 &&
        isIntrinsicWithCC(CmpOp0, Opcode, CCValid))
      return getIntrinsicCmp(DAG, Opcode, CmpOp0, CCValid, Constant, Cond);
  }
  Comparison C(CmpOp0, CmpOp1, Chain);
  C.CCMask = CCMaskForCondCode(Cond);
  if (C.Op0.getValueType().isFloatingPoint()) {
    C.CCValid = SystemZ::CCMASK_FCMP;
    if (!C.Chain)
      C.Opcode = SystemZISD::FCMP;
    else if (!IsSignaling)
      C.Opcode = SystemZISD::STRICT_FCMP;
    else
      C.Opcode = SystemZISD::STRICT_FCMPS;
    adjustForFNeg(C);
  } else {
    assert(!C.Chain);
    C.CCValid = SystemZ::CCMASK_ICMP;
    C.Opcode = SystemZISD::ICMP;
    // Choose the type of comparison.  Equality and inequality tests can
    // use either signed or unsigned comparisons.  The choice also doesn't
    // matter if both sign bits are known to be clear.  In those cases we
    // want to give the main isel code the freedom to choose whichever
    // form fits best.
    if (C.CCMask == SystemZ::CCMASK_CMP_EQ ||
        C.CCMask == SystemZ::CCMASK_CMP_NE ||
        (DAG.SignBitIsZero(C.Op0) && DAG.SignBitIsZero(C.Op1)))
      C.ICmpType = SystemZICMP::Any;
    else if (C.CCMask & SystemZ::CCMASK_CMP_UO)
      C.ICmpType = SystemZICMP::UnsignedOnly;
    else
      C.ICmpType = SystemZICMP::SignedOnly;
    C.CCMask &= ~SystemZ::CCMASK_CMP_UO;
    adjustForRedundantAnd(DAG, DL, C);
    adjustZeroCmp(DAG, DL, C);
    adjustSubwordCmp(DAG, DL, C);
    adjustForSubtraction(DAG, DL, C);
    adjustForLTGFR(C);
    adjustICmpTruncate(DAG, DL, C);
  }

  if (shouldSwapCmpOperands(C)) {
    std::swap(C.Op0, C.Op1);
    C.CCMask = SystemZ::reverseCCMask(C.CCMask);
  }

  adjustForTestUnderMask(DAG, DL, C);
  return C;
}

// Emit the comparison instruction described by C.
static SDValue emitCmp(SelectionDAG &DAG, const SDLoc &DL, Comparison &C) {
  if (!C.Op1.getNode()) {
    SDNode *Node;
    switch (C.Op0.getOpcode()) {
    case ISD::INTRINSIC_W_CHAIN:
      Node = emitIntrinsicWithCCAndChain(DAG, C.Op0, C.Opcode);
      return SDValue(Node, 0);
    case ISD::INTRINSIC_WO_CHAIN:
      Node = emitIntrinsicWithCC(DAG, C.Op0, C.Opcode);
      return SDValue(Node, Node->getNumValues() - 1);
    default:
      llvm_unreachable("Invalid comparison operands");
    }
  }
  if (C.Opcode == SystemZISD::ICMP)
    return DAG.getNode(SystemZISD::ICMP, DL, MVT::i32, C.Op0, C.Op1,
                       DAG.getTargetConstant(C.ICmpType, DL, MVT::i32));
  if (C.Opcode == SystemZISD::TM) {
    bool RegisterOnly = (bool(C.CCMask & SystemZ::CCMASK_TM_MIXED_MSB_0) !=
                         bool(C.CCMask & SystemZ::CCMASK_TM_MIXED_MSB_1));
    return DAG.getNode(SystemZISD::TM, DL, MVT::i32, C.Op0, C.Op1,
                       DAG.getTargetConstant(RegisterOnly, DL, MVT::i32));
  }
  if (C.Chain) {
    SDVTList VTs = DAG.getVTList(MVT::i32, MVT::Other);
    return DAG.getNode(C.Opcode, DL, VTs, C.Chain, C.Op0, C.Op1);
  }
  return DAG.getNode(C.Opcode, DL, MVT::i32, C.Op0, C.Op1);
}

// Implement a 32-bit *MUL_LOHI operation by extending both operands to
// 64 bits.  Extend is the extension type to use.  Store the high part
// in Hi and the low part in Lo.
static void lowerMUL_LOHI32(SelectionDAG &DAG, const SDLoc &DL, unsigned Extend,
                            SDValue Op0, SDValue Op1, SDValue &Hi,
                            SDValue &Lo) {
  Op0 = DAG.getNode(Extend, DL, MVT::i64, Op0);
  Op1 = DAG.getNode(Extend, DL, MVT::i64, Op1);
  SDValue Mul = DAG.getNode(ISD::MUL, DL, MVT::i64, Op0, Op1);
  Hi = DAG.getNode(ISD::SRL, DL, MVT::i64, Mul,
                   DAG.getConstant(32, DL, MVT::i64));
  Hi = DAG.getNode(ISD::TRUNCATE, DL, MVT::i32, Hi);
  Lo = DAG.getNode(ISD::TRUNCATE, DL, MVT::i32, Mul);
}

// Lower a binary operation that produces two VT results, one in each
// half of a GR128 pair.  Op0 and Op1 are the VT operands to the operation,
// and Opcode performs the GR128 operation.  Store the even register result
// in Even and the odd register result in Odd.
static void lowerGR128Binary(SelectionDAG &DAG, const SDLoc &DL, EVT VT,
                             unsigned Opcode, SDValue Op0, SDValue Op1,
                             SDValue &Even, SDValue &Odd) {
  SDValue Result = DAG.getNode(Opcode, DL, MVT::Untyped, Op0, Op1);
  bool Is32Bit = is32Bit(VT);
  Even = DAG.getTargetExtractSubreg(SystemZ::even128(Is32Bit), DL, VT, Result);
  Odd = DAG.getTargetExtractSubreg(SystemZ::odd128(Is32Bit), DL, VT, Result);
}

// Return an i32 value that is 1 if the CC value produced by CCReg is
// in the mask CCMask and 0 otherwise.  CC is known to have a value
// in CCValid, so other values can be ignored.
static SDValue emitSETCC(SelectionDAG &DAG, const SDLoc &DL, SDValue CCReg,
                         unsigned CCValid, unsigned CCMask) {
  SDValue Ops[] = {DAG.getConstant(1, DL, MVT::i32),
                   DAG.getConstant(0, DL, MVT::i32),
                   DAG.getTargetConstant(CCValid, DL, MVT::i32),
                   DAG.getTargetConstant(CCMask, DL, MVT::i32), CCReg};
  return DAG.getNode(SystemZISD::SELECT_CCMASK, DL, MVT::i32, Ops);
}

// Return the SystemISD vector comparison operation for CC, or 0 if it cannot
// be done directly.  Mode is CmpMode::Int for integer comparisons, CmpMode::FP
// for regular floating-point comparisons, CmpMode::StrictFP for strict (quiet)
// floating-point comparisons, and CmpMode::SignalingFP for strict signaling
// floating-point comparisons.
enum class CmpMode { Int, FP, StrictFP, SignalingFP };
static unsigned getVectorComparison(ISD::CondCode CC, CmpMode Mode) {
  switch (CC) {
  case ISD::SETOEQ:
  case ISD::SETEQ:
    switch (Mode) {
    case CmpMode::Int:         return SystemZISD::VICMPE;
    case CmpMode::FP:          return SystemZISD::VFCMPE;
    case CmpMode::StrictFP:    return SystemZISD::STRICT_VFCMPE;
    case CmpMode::SignalingFP: return SystemZISD::STRICT_VFCMPES;
    }
    llvm_unreachable("Bad mode");

  case ISD::SETOGE:
  case ISD::SETGE:
    switch (Mode) {
    case CmpMode::Int:         return 0;
    case CmpMode::FP:          return SystemZISD::VFCMPHE;
    case CmpMode::StrictFP:    return SystemZISD::STRICT_VFCMPHE;
    case CmpMode::SignalingFP: return SystemZISD::STRICT_VFCMPHES;
    }
    llvm_unreachable("Bad mode");

  case ISD::SETOGT:
  case ISD::SETGT:
    switch (Mode) {
    case CmpMode::Int:         return SystemZISD::VICMPH;
    case CmpMode::FP:          return SystemZISD::VFCMPH;
    case CmpMode::StrictFP:    return SystemZISD::STRICT_VFCMPH;
    case CmpMode::SignalingFP: return SystemZISD::STRICT_VFCMPHS;
    }
    llvm_unreachable("Bad mode");

  case ISD::SETUGT:
    switch (Mode) {
    case CmpMode::Int:         return SystemZISD::VICMPHL;
    case CmpMode::FP:          return 0;
    case CmpMode::StrictFP:    return 0;
    case CmpMode::SignalingFP: return 0;
    }
    llvm_unreachable("Bad mode");

  default:
    return 0;
  }
}

// Return the SystemZISD vector comparison operation for CC or its inverse,
// or 0 if neither can be done directly.  Indicate in Invert whether the
// result is for the inverse of CC.  Mode is as above.
static unsigned getVectorComparisonOrInvert(ISD::CondCode CC, CmpMode Mode,
                                            bool &Invert) {
  if (unsigned Opcode = getVectorComparison(CC, Mode)) {
    Invert = false;
    return Opcode;
  }

  CC = ISD::getSetCCInverse(CC, Mode == CmpMode::Int ? MVT::i32 : MVT::f32);
  if (unsigned Opcode = getVectorComparison(CC, Mode)) {
    Invert = true;
    return Opcode;
  }

  return 0;
}

// Return a v2f64 that contains the extended form of elements Start and Start+1
// of v4f32 value Op.  If Chain is nonnull, return the strict form.
static SDValue expandV4F32ToV2F64(SelectionDAG &DAG, int Start, const SDLoc &DL,
                                  SDValue Op, SDValue Chain) {
  int Mask[] = { Start, -1, Start + 1, -1 };
  Op = DAG.getVectorShuffle(MVT::v4f32, DL, Op, DAG.getUNDEF(MVT::v4f32), Mask);
  if (Chain) {
    SDVTList VTs = DAG.getVTList(MVT::v2f64, MVT::Other);
    return DAG.getNode(SystemZISD::STRICT_VEXTEND, DL, VTs, Chain, Op);
  }
  return DAG.getNode(SystemZISD::VEXTEND, DL, MVT::v2f64, Op);
}

// Build a comparison of vectors CmpOp0 and CmpOp1 using opcode Opcode,
// producing a result of type VT.  If Chain is nonnull, return the strict form.
SDValue SystemZTargetLowering::getVectorCmp(SelectionDAG &DAG, unsigned Opcode,
                                            const SDLoc &DL, EVT VT,
                                            SDValue CmpOp0,
                                            SDValue CmpOp1,
                                            SDValue Chain) const {
  // There is no hardware support for v4f32 (unless we have the vector
  // enhancements facility 1), so extend the vector into two v2f64s
  // and compare those.
  if (CmpOp0.getValueType() == MVT::v4f32 &&
      !Subtarget.hasVectorEnhancements1()) {
    SDValue H0 = expandV4F32ToV2F64(DAG, 0, DL, CmpOp0, Chain);
    SDValue L0 = expandV4F32ToV2F64(DAG, 2, DL, CmpOp0, Chain);
    SDValue H1 = expandV4F32ToV2F64(DAG, 0, DL, CmpOp1, Chain);
    SDValue L1 = expandV4F32ToV2F64(DAG, 2, DL, CmpOp1, Chain);
    if (Chain) {
      SDVTList VTs = DAG.getVTList(MVT::v2i64, MVT::Other);
      SDValue HRes = DAG.getNode(Opcode, DL, VTs, Chain, H0, H1);
      SDValue LRes = DAG.getNode(Opcode, DL, VTs, Chain, L0, L1);
      SDValue Res = DAG.getNode(SystemZISD::PACK, DL, VT, HRes, LRes);
      SDValue Chains[6] = { H0.getValue(1), L0.getValue(1),
                            H1.getValue(1), L1.getValue(1),
                            HRes.getValue(1), LRes.getValue(1) };
      SDValue NewChain = DAG.getNode(ISD::TokenFactor, DL, MVT::Other, Chains);
      SDValue Ops[2] = { Res, NewChain };
      return DAG.getMergeValues(Ops, DL);
    }
    SDValue HRes = DAG.getNode(Opcode, DL, MVT::v2i64, H0, H1);
    SDValue LRes = DAG.getNode(Opcode, DL, MVT::v2i64, L0, L1);
    return DAG.getNode(SystemZISD::PACK, DL, VT, HRes, LRes);
  }
  if (Chain) {
    SDVTList VTs = DAG.getVTList(VT, MVT::Other);
    return DAG.getNode(Opcode, DL, VTs, Chain, CmpOp0, CmpOp1);
  }
  return DAG.getNode(Opcode, DL, VT, CmpOp0, CmpOp1);
}

// Lower a vector comparison of type CC between CmpOp0 and CmpOp1, producing
// an integer mask of type VT.  If Chain is nonnull, we have a strict
// floating-point comparison.  If in addition IsSignaling is true, we have
// a strict signaling floating-point comparison.
SDValue SystemZTargetLowering::lowerVectorSETCC(SelectionDAG &DAG,
                                                const SDLoc &DL, EVT VT,
                                                ISD::CondCode CC,
                                                SDValue CmpOp0,
                                                SDValue CmpOp1,
                                                SDValue Chain,
                                                bool IsSignaling) const {
  bool IsFP = CmpOp0.getValueType().isFloatingPoint();
  assert (!Chain || IsFP);
  assert (!IsSignaling || Chain);
  CmpMode Mode = IsSignaling ? CmpMode::SignalingFP :
                 Chain ? CmpMode::StrictFP : IsFP ? CmpMode::FP : CmpMode::Int;
  bool Invert = false;
  SDValue Cmp;
  switch (CC) {
    // Handle tests for order using (or (ogt y x) (oge x y)).
  case ISD::SETUO:
    Invert = true;
    LLVM_FALLTHROUGH;
  case ISD::SETO: {
    assert(IsFP && "Unexpected integer comparison");
    SDValue LT = getVectorCmp(DAG, getVectorComparison(ISD::SETOGT, Mode),
                              DL, VT, CmpOp1, CmpOp0, Chain);
    SDValue GE = getVectorCmp(DAG, getVectorComparison(ISD::SETOGE, Mode),
                              DL, VT, CmpOp0, CmpOp1, Chain);
    Cmp = DAG.getNode(ISD::OR, DL, VT, LT, GE);
    if (Chain)
      Chain = DAG.getNode(ISD::TokenFactor, DL, MVT::Other,
                          LT.getValue(1), GE.getValue(1));
    break;
  }

    // Handle <> tests using (or (ogt y x) (ogt x y)).
  case ISD::SETUEQ:
    Invert = true;
    LLVM_FALLTHROUGH;
  case ISD::SETONE: {
    assert(IsFP && "Unexpected integer comparison");
    SDValue LT = getVectorCmp(DAG, getVectorComparison(ISD::SETOGT, Mode),
                              DL, VT, CmpOp1, CmpOp0, Chain);
    SDValue GT = getVectorCmp(DAG, getVectorComparison(ISD::SETOGT, Mode),
                              DL, VT, CmpOp0, CmpOp1, Chain);
    Cmp = DAG.getNode(ISD::OR, DL, VT, LT, GT);
    if (Chain)
      Chain = DAG.getNode(ISD::TokenFactor, DL, MVT::Other,
                          LT.getValue(1), GT.getValue(1));
    break;
  }

    // Otherwise a single comparison is enough.  It doesn't really
    // matter whether we try the inversion or the swap first, since
    // there are no cases where both work.
  default:
    if (unsigned Opcode = getVectorComparisonOrInvert(CC, Mode, Invert))
      Cmp = getVectorCmp(DAG, Opcode, DL, VT, CmpOp0, CmpOp1, Chain);
    else {
      CC = ISD::getSetCCSwappedOperands(CC);
      if (unsigned Opcode = getVectorComparisonOrInvert(CC, Mode, Invert))
        Cmp = getVectorCmp(DAG, Opcode, DL, VT, CmpOp1, CmpOp0, Chain);
      else
        llvm_unreachable("Unhandled comparison");
    }
    if (Chain)
      Chain = Cmp.getValue(1);
    break;
  }
  if (Invert) {
    SDValue Mask =
      DAG.getSplatBuildVector(VT, DL, DAG.getConstant(-1, DL, MVT::i64));
    Cmp = DAG.getNode(ISD::XOR, DL, VT, Cmp, Mask);
  }
  if (Chain && Chain.getNode() != Cmp.getNode()) {
    SDValue Ops[2] = { Cmp, Chain };
    Cmp = DAG.getMergeValues(Ops, DL);
  }
  return Cmp;
}

SDValue SystemZTargetLowering::lowerSETCC(SDValue Op,
                                          SelectionDAG &DAG) const {
  SDValue CmpOp0   = Op.getOperand(0);
  SDValue CmpOp1   = Op.getOperand(1);
  ISD::CondCode CC = cast<CondCodeSDNode>(Op.getOperand(2))->get();
  SDLoc DL(Op);
  EVT VT = Op.getValueType();
  if (VT.isVector())
    return lowerVectorSETCC(DAG, DL, VT, CC, CmpOp0, CmpOp1);

  Comparison C(getCmp(DAG, CmpOp0, CmpOp1, CC, DL));
  SDValue CCReg = emitCmp(DAG, DL, C);
  return emitSETCC(DAG, DL, CCReg, C.CCValid, C.CCMask);
}

SDValue SystemZTargetLowering::lowerSTRICT_FSETCC(SDValue Op,
                                                  SelectionDAG &DAG,
                                                  bool IsSignaling) const {
  SDValue Chain    = Op.getOperand(0);
  SDValue CmpOp0   = Op.getOperand(1);
  SDValue CmpOp1   = Op.getOperand(2);
  ISD::CondCode CC = cast<CondCodeSDNode>(Op.getOperand(3))->get();
  SDLoc DL(Op);
  EVT VT = Op.getNode()->getValueType(0);
  if (VT.isVector()) {
    SDValue Res = lowerVectorSETCC(DAG, DL, VT, CC, CmpOp0, CmpOp1,
                                   Chain, IsSignaling);
    return Res.getValue(Op.getResNo());
  }

  Comparison C(getCmp(DAG, CmpOp0, CmpOp1, CC, DL, Chain, IsSignaling));
  SDValue CCReg = emitCmp(DAG, DL, C);
  CCReg->setFlags(Op->getFlags());
  SDValue Result = emitSETCC(DAG, DL, CCReg, C.CCValid, C.CCMask);
  SDValue Ops[2] = { Result, CCReg.getValue(1) };
  return DAG.getMergeValues(Ops, DL);
}

SDValue SystemZTargetLowering::lowerBR_CC(SDValue Op, SelectionDAG &DAG) const {
  ISD::CondCode CC = cast<CondCodeSDNode>(Op.getOperand(1))->get();
  SDValue CmpOp0   = Op.getOperand(2);
  SDValue CmpOp1   = Op.getOperand(3);
  SDValue Dest     = Op.getOperand(4);
  SDLoc DL(Op);

  Comparison C(getCmp(DAG, CmpOp0, CmpOp1, CC, DL));
  SDValue CCReg = emitCmp(DAG, DL, C);
  return DAG.getNode(
      SystemZISD::BR_CCMASK, DL, Op.getValueType(), Op.getOperand(0),
      DAG.getTargetConstant(C.CCValid, DL, MVT::i32),
      DAG.getTargetConstant(C.CCMask, DL, MVT::i32), Dest, CCReg);
}

// Return true if Pos is CmpOp and Neg is the negative of CmpOp,
// allowing Pos and Neg to be wider than CmpOp.
static bool isAbsolute(SDValue CmpOp, SDValue Pos, SDValue Neg) {
  return (Neg.getOpcode() == ISD::SUB &&
          Neg.getOperand(0).getOpcode() == ISD::Constant &&
          cast<ConstantSDNode>(Neg.getOperand(0))->getZExtValue() == 0 &&
          Neg.getOperand(1) == Pos &&
          (Pos == CmpOp ||
           (Pos.getOpcode() == ISD::SIGN_EXTEND &&
            Pos.getOperand(0) == CmpOp)));
}

// Return the absolute or negative absolute of Op; IsNegative decides which.
static SDValue getAbsolute(SelectionDAG &DAG, const SDLoc &DL, SDValue Op,
                           bool IsNegative) {
  Op = DAG.getNode(SystemZISD::IABS, DL, Op.getValueType(), Op);
  if (IsNegative)
    Op = DAG.getNode(ISD::SUB, DL, Op.getValueType(),
                     DAG.getConstant(0, DL, Op.getValueType()), Op);
  return Op;
}

SDValue SystemZTargetLowering::lowerSELECT_CC(SDValue Op,
                                              SelectionDAG &DAG) const {
  SDValue CmpOp0   = Op.getOperand(0);
  SDValue CmpOp1   = Op.getOperand(1);
  SDValue TrueOp   = Op.getOperand(2);
  SDValue FalseOp  = Op.getOperand(3);
  ISD::CondCode CC = cast<CondCodeSDNode>(Op.getOperand(4))->get();
  SDLoc DL(Op);

  Comparison C(getCmp(DAG, CmpOp0, CmpOp1, CC, DL));

  // Check for absolute and negative-absolute selections, including those
  // where the comparison value is sign-extended (for LPGFR and LNGFR).
  // This check supplements the one in DAGCombiner.
  if (C.Opcode == SystemZISD::ICMP &&
      C.CCMask != SystemZ::CCMASK_CMP_EQ &&
      C.CCMask != SystemZ::CCMASK_CMP_NE &&
      C.Op1.getOpcode() == ISD::Constant &&
      cast<ConstantSDNode>(C.Op1)->getZExtValue() == 0) {
    if (isAbsolute(C.Op0, TrueOp, FalseOp))
      return getAbsolute(DAG, DL, TrueOp, C.CCMask & SystemZ::CCMASK_CMP_LT);
    if (isAbsolute(C.Op0, FalseOp, TrueOp))
      return getAbsolute(DAG, DL, FalseOp, C.CCMask & SystemZ::CCMASK_CMP_GT);
  }

  SDValue CCReg = emitCmp(DAG, DL, C);
  SDValue Ops[] = {TrueOp, FalseOp,
                   DAG.getTargetConstant(C.CCValid, DL, MVT::i32),
                   DAG.getTargetConstant(C.CCMask, DL, MVT::i32), CCReg};

  return DAG.getNode(SystemZISD::SELECT_CCMASK, DL, Op.getValueType(), Ops);
}

SDValue SystemZTargetLowering::lowerGlobalAddress(GlobalAddressSDNode *Node,
                                                  SelectionDAG &DAG) const {
  SDLoc DL(Node);
  const GlobalValue *GV = Node->getGlobal();
  int64_t Offset = Node->getOffset();
  EVT PtrVT = getPointerTy(DAG.getDataLayout());
  CodeModel::Model CM = DAG.getTarget().getCodeModel();

  SDValue Result;
  if (Subtarget.isPC32DBLSymbol(GV, CM)) {
    if (isInt<32>(Offset)) {
      // Assign anchors at 1<<12 byte boundaries.
      uint64_t Anchor = Offset & ~uint64_t(0xfff);
      Result = DAG.getTargetGlobalAddress(GV, DL, PtrVT, Anchor);
      Result = DAG.getNode(SystemZISD::PCREL_WRAPPER, DL, PtrVT, Result);

      // The offset can be folded into the address if it is aligned to a
      // halfword.
      Offset -= Anchor;
      if (Offset != 0 && (Offset & 1) == 0) {
        SDValue Full =
          DAG.getTargetGlobalAddress(GV, DL, PtrVT, Anchor + Offset);
        Result = DAG.getNode(SystemZISD::PCREL_OFFSET, DL, PtrVT, Full, Result);
        Offset = 0;
      }
    } else {
      // Conservatively load a constant offset greater than 32 bits into a
      // register below.
      Result = DAG.getTargetGlobalAddress(GV, DL, PtrVT);
      Result = DAG.getNode(SystemZISD::PCREL_WRAPPER, DL, PtrVT, Result);
    }
  } else {
    Result = DAG.getTargetGlobalAddress(GV, DL, PtrVT, 0, SystemZII::MO_GOT);
    Result = DAG.getNode(SystemZISD::PCREL_WRAPPER, DL, PtrVT, Result);
    Result = DAG.getLoad(PtrVT, DL, DAG.getEntryNode(), Result,
                         MachinePointerInfo::getGOT(DAG.getMachineFunction()));
  }

  // If there was a non-zero offset that we didn't fold, create an explicit
  // addition for it.
  if (Offset != 0)
    Result = DAG.getNode(ISD::ADD, DL, PtrVT, Result,
                         DAG.getConstant(Offset, DL, PtrVT));

  return Result;
}

SDValue SystemZTargetLowering::lowerTLSGetOffset(GlobalAddressSDNode *Node,
                                                 SelectionDAG &DAG,
                                                 unsigned Opcode,
                                                 SDValue GOTOffset) const {
  SDLoc DL(Node);
  EVT PtrVT = getPointerTy(DAG.getDataLayout());
  SDValue Chain = DAG.getEntryNode();
  SDValue Glue;

  if (DAG.getMachineFunction().getFunction().getCallingConv() ==
      CallingConv::GHC)
    report_fatal_error("In GHC calling convention TLS is not supported");

  // __tls_get_offset takes the GOT offset in %r2 and the GOT in %r12.
  SDValue GOT = DAG.getGLOBAL_OFFSET_TABLE(PtrVT);
  Chain = DAG.getCopyToReg(Chain, DL, SystemZ::R12D, GOT, Glue);
  Glue = Chain.getValue(1);
  Chain = DAG.getCopyToReg(Chain, DL, SystemZ::R2D, GOTOffset, Glue);
  Glue = Chain.getValue(1);

  // The first call operand is the chain and the second is the TLS symbol.
  SmallVector<SDValue, 8> Ops;
  Ops.push_back(Chain);
  Ops.push_back(DAG.getTargetGlobalAddress(Node->getGlobal(), DL,
                                           Node->getValueType(0),
                                           0, 0));

  // Add argument registers to the end of the list so that they are
  // known live into the call.
  Ops.push_back(DAG.getRegister(SystemZ::R2D, PtrVT));
  Ops.push_back(DAG.getRegister(SystemZ::R12D, PtrVT));

  // Add a register mask operand representing the call-preserved registers.
  const TargetRegisterInfo *TRI = Subtarget.getRegisterInfo();
  const uint32_t *Mask =
      TRI->getCallPreservedMask(DAG.getMachineFunction(), CallingConv::C);
  assert(Mask && "Missing call preserved mask for calling convention");
  Ops.push_back(DAG.getRegisterMask(Mask));

  // Glue the call to the argument copies.
  Ops.push_back(Glue);

  // Emit the call.
  SDVTList NodeTys = DAG.getVTList(MVT::Other, MVT::Glue);
  Chain = DAG.getNode(Opcode, DL, NodeTys, Ops);
  Glue = Chain.getValue(1);

  // Copy the return value from %r2.
  return DAG.getCopyFromReg(Chain, DL, SystemZ::R2D, PtrVT, Glue);
}

SDValue SystemZTargetLowering::lowerThreadPointer(const SDLoc &DL,
                                                  SelectionDAG &DAG) const {
  SDValue Chain = DAG.getEntryNode();
  EVT PtrVT = getPointerTy(DAG.getDataLayout());

  // The high part of the thread pointer is in access register 0.
  SDValue TPHi = DAG.getCopyFromReg(Chain, DL, SystemZ::A0, MVT::i32);
  TPHi = DAG.getNode(ISD::ANY_EXTEND, DL, PtrVT, TPHi);

  // The low part of the thread pointer is in access register 1.
  SDValue TPLo = DAG.getCopyFromReg(Chain, DL, SystemZ::A1, MVT::i32);
  TPLo = DAG.getNode(ISD::ZERO_EXTEND, DL, PtrVT, TPLo);

  // Merge them into a single 64-bit address.
  SDValue TPHiShifted = DAG.getNode(ISD::SHL, DL, PtrVT, TPHi,
                                    DAG.getConstant(32, DL, PtrVT));
  return DAG.getNode(ISD::OR, DL, PtrVT, TPHiShifted, TPLo);
}

SDValue SystemZTargetLowering::lowerGlobalTLSAddress(GlobalAddressSDNode *Node,
                                                     SelectionDAG &DAG) const {
  if (DAG.getTarget().useEmulatedTLS())
    return LowerToTLSEmulatedModel(Node, DAG);
  SDLoc DL(Node);
  const GlobalValue *GV = Node->getGlobal();
  EVT PtrVT = getPointerTy(DAG.getDataLayout());
  TLSModel::Model model = DAG.getTarget().getTLSModel(GV);

  if (DAG.getMachineFunction().getFunction().getCallingConv() ==
      CallingConv::GHC)
    report_fatal_error("In GHC calling convention TLS is not supported");

  SDValue TP = lowerThreadPointer(DL, DAG);

  // Get the offset of GA from the thread pointer, based on the TLS model.
  SDValue Offset;
  switch (model) {
    case TLSModel::GeneralDynamic: {
      // Load the GOT offset of the tls_index (module ID / per-symbol offset).
      SystemZConstantPoolValue *CPV =
        SystemZConstantPoolValue::Create(GV, SystemZCP::TLSGD);

      Offset = DAG.getConstantPool(CPV, PtrVT, Align(8));
      Offset = DAG.getLoad(
          PtrVT, DL, DAG.getEntryNode(), Offset,
          MachinePointerInfo::getConstantPool(DAG.getMachineFunction()));

      // Call __tls_get_offset to retrieve the offset.
      Offset = lowerTLSGetOffset(Node, DAG, SystemZISD::TLS_GDCALL, Offset);
      break;
    }

    case TLSModel::LocalDynamic: {
      // Load the GOT offset of the module ID.
      SystemZConstantPoolValue *CPV =
        SystemZConstantPoolValue::Create(GV, SystemZCP::TLSLDM);

      Offset = DAG.getConstantPool(CPV, PtrVT, Align(8));
      Offset = DAG.getLoad(
          PtrVT, DL, DAG.getEntryNode(), Offset,
          MachinePointerInfo::getConstantPool(DAG.getMachineFunction()));

      // Call __tls_get_offset to retrieve the module base offset.
      Offset = lowerTLSGetOffset(Node, DAG, SystemZISD::TLS_LDCALL, Offset);

      // Note: The SystemZLDCleanupPass will remove redundant computations
      // of the module base offset.  Count total number of local-dynamic
      // accesses to trigger execution of that pass.
      SystemZMachineFunctionInfo* MFI =
        DAG.getMachineFunction().getInfo<SystemZMachineFunctionInfo>();
      MFI->incNumLocalDynamicTLSAccesses();

      // Add the per-symbol offset.
      CPV = SystemZConstantPoolValue::Create(GV, SystemZCP::DTPOFF);

      SDValue DTPOffset = DAG.getConstantPool(CPV, PtrVT, Align(8));
      DTPOffset = DAG.getLoad(
          PtrVT, DL, DAG.getEntryNode(), DTPOffset,
          MachinePointerInfo::getConstantPool(DAG.getMachineFunction()));

      Offset = DAG.getNode(ISD::ADD, DL, PtrVT, Offset, DTPOffset);
      break;
    }

    case TLSModel::InitialExec: {
      // Load the offset from the GOT.
      Offset = DAG.getTargetGlobalAddress(GV, DL, PtrVT, 0,
                                          SystemZII::MO_INDNTPOFF);
      Offset = DAG.getNode(SystemZISD::PCREL_WRAPPER, DL, PtrVT, Offset);
      Offset =
          DAG.getLoad(PtrVT, DL, DAG.getEntryNode(), Offset,
                      MachinePointerInfo::getGOT(DAG.getMachineFunction()));
      break;
    }

    case TLSModel::LocalExec: {
      // Force the offset into the constant pool and load it from there.
      SystemZConstantPoolValue *CPV =
        SystemZConstantPoolValue::Create(GV, SystemZCP::NTPOFF);

      Offset = DAG.getConstantPool(CPV, PtrVT, Align(8));
      Offset = DAG.getLoad(
          PtrVT, DL, DAG.getEntryNode(), Offset,
          MachinePointerInfo::getConstantPool(DAG.getMachineFunction()));
      break;
    }
  }

  // Add the base and offset together.
  return DAG.getNode(ISD::ADD, DL, PtrVT, TP, Offset);
}

SDValue SystemZTargetLowering::lowerBlockAddress(BlockAddressSDNode *Node,
                                                 SelectionDAG &DAG) const {
  SDLoc DL(Node);
  const BlockAddress *BA = Node->getBlockAddress();
  int64_t Offset = Node->getOffset();
  EVT PtrVT = getPointerTy(DAG.getDataLayout());

  SDValue Result = DAG.getTargetBlockAddress(BA, PtrVT, Offset);
  Result = DAG.getNode(SystemZISD::PCREL_WRAPPER, DL, PtrVT, Result);
  return Result;
}

SDValue SystemZTargetLowering::lowerJumpTable(JumpTableSDNode *JT,
                                              SelectionDAG &DAG) const {
  SDLoc DL(JT);
  EVT PtrVT = getPointerTy(DAG.getDataLayout());
  SDValue Result = DAG.getTargetJumpTable(JT->getIndex(), PtrVT);

  // Use LARL to load the address of the table.
  return DAG.getNode(SystemZISD::PCREL_WRAPPER, DL, PtrVT, Result);
}

SDValue SystemZTargetLowering::lowerConstantPool(ConstantPoolSDNode *CP,
                                                 SelectionDAG &DAG) const {
  SDLoc DL(CP);
  EVT PtrVT = getPointerTy(DAG.getDataLayout());

  SDValue Result;
  if (CP->isMachineConstantPoolEntry())
    Result =
        DAG.getTargetConstantPool(CP->getMachineCPVal(), PtrVT, CP->getAlign());
  else
    Result = DAG.getTargetConstantPool(CP->getConstVal(), PtrVT, CP->getAlign(),
                                       CP->getOffset());

  // Use LARL to load the address of the constant pool entry.
  return DAG.getNode(SystemZISD::PCREL_WRAPPER, DL, PtrVT, Result);
}

SDValue SystemZTargetLowering::lowerFRAMEADDR(SDValue Op,
                                              SelectionDAG &DAG) const {
  auto *TFL =
      static_cast<const SystemZFrameLowering *>(Subtarget.getFrameLowering());
  MachineFunction &MF = DAG.getMachineFunction();
  MachineFrameInfo &MFI = MF.getFrameInfo();
  MFI.setFrameAddressIsTaken(true);

  SDLoc DL(Op);
  unsigned Depth = cast<ConstantSDNode>(Op.getOperand(0))->getZExtValue();
  EVT PtrVT = getPointerTy(DAG.getDataLayout());

  // Return null if the back chain is not present.
  bool HasBackChain = MF.getFunction().hasFnAttribute("backchain");
  if (TFL->usePackedStack(MF) && !HasBackChain)
    return DAG.getConstant(0, DL, PtrVT);

  // By definition, the frame address is the address of the back chain.
  int BackChainIdx = TFL->getOrCreateFramePointerSaveIndex(MF);
  SDValue BackChain = DAG.getFrameIndex(BackChainIdx, PtrVT);

  // FIXME The frontend should detect this case.
  if (Depth > 0) {
    report_fatal_error("Unsupported stack frame traversal count");
  }

  return BackChain;
}

SDValue SystemZTargetLowering::lowerRETURNADDR(SDValue Op,
                                               SelectionDAG &DAG) const {
  MachineFunction &MF = DAG.getMachineFunction();
  MachineFrameInfo &MFI = MF.getFrameInfo();
  MFI.setReturnAddressIsTaken(true);

  if (verifyReturnAddressArgumentIsConstant(Op, DAG))
    return SDValue();

  SDLoc DL(Op);
  unsigned Depth = cast<ConstantSDNode>(Op.getOperand(0))->getZExtValue();
  EVT PtrVT = getPointerTy(DAG.getDataLayout());

  // FIXME The frontend should detect this case.
  if (Depth > 0) {
    report_fatal_error("Unsupported stack frame traversal count");
  }

  // Return R14D, which has the return address. Mark it an implicit live-in.
  unsigned LinkReg = MF.addLiveIn(SystemZ::R14D, &SystemZ::GR64BitRegClass);
  return DAG.getCopyFromReg(DAG.getEntryNode(), DL, LinkReg, PtrVT);
}

SDValue SystemZTargetLowering::lowerBITCAST(SDValue Op,
                                            SelectionDAG &DAG) const {
  SDLoc DL(Op);
  SDValue In = Op.getOperand(0);
  EVT InVT = In.getValueType();
  EVT ResVT = Op.getValueType();

  // Convert loads directly.  This is normally done by DAGCombiner,
  // but we need this case for bitcasts that are created during lowering
  // and which are then lowered themselves.
  if (auto *LoadN = dyn_cast<LoadSDNode>(In))
    if (ISD::isNormalLoad(LoadN)) {
      SDValue NewLoad = DAG.getLoad(ResVT, DL, LoadN->getChain(),
                                    LoadN->getBasePtr(), LoadN->getMemOperand());
      // Update the chain uses.
      DAG.ReplaceAllUsesOfValueWith(SDValue(LoadN, 1), NewLoad.getValue(1));
      return NewLoad;
    }

  if (InVT == MVT::i32 && ResVT == MVT::f32) {
    SDValue In64;
    if (Subtarget.hasHighWord()) {
      SDNode *U64 = DAG.getMachineNode(TargetOpcode::IMPLICIT_DEF, DL,
                                       MVT::i64);
      In64 = DAG.getTargetInsertSubreg(SystemZ::subreg_h32, DL,
                                       MVT::i64, SDValue(U64, 0), In);
    } else {
      In64 = DAG.getNode(ISD::ANY_EXTEND, DL, MVT::i64, In);
      In64 = DAG.getNode(ISD::SHL, DL, MVT::i64, In64,
                         DAG.getConstant(32, DL, MVT::i64));
    }
    SDValue Out64 = DAG.getNode(ISD::BITCAST, DL, MVT::f64, In64);
    return DAG.getTargetExtractSubreg(SystemZ::subreg_h32,
                                      DL, MVT::f32, Out64);
  }
  if (InVT == MVT::f32 && ResVT == MVT::i32) {
    SDNode *U64 = DAG.getMachineNode(TargetOpcode::IMPLICIT_DEF, DL, MVT::f64);
    SDValue In64 = DAG.getTargetInsertSubreg(SystemZ::subreg_h32, DL,
                                             MVT::f64, SDValue(U64, 0), In);
    SDValue Out64 = DAG.getNode(ISD::BITCAST, DL, MVT::i64, In64);
    if (Subtarget.hasHighWord())
      return DAG.getTargetExtractSubreg(SystemZ::subreg_h32, DL,
                                        MVT::i32, Out64);
    SDValue Shift = DAG.getNode(ISD::SRL, DL, MVT::i64, Out64,
                                DAG.getConstant(32, DL, MVT::i64));
    return DAG.getNode(ISD::TRUNCATE, DL, MVT::i32, Shift);
  }
  llvm_unreachable("Unexpected bitcast combination");
}

SDValue SystemZTargetLowering::lowerVASTART(SDValue Op,
                                            SelectionDAG &DAG) const {
  MachineFunction &MF = DAG.getMachineFunction();
  SystemZMachineFunctionInfo *FuncInfo =
    MF.getInfo<SystemZMachineFunctionInfo>();
  EVT PtrVT = getPointerTy(DAG.getDataLayout());

  SDValue Chain   = Op.getOperand(0);
  SDValue Addr    = Op.getOperand(1);
  const Value *SV = cast<SrcValueSDNode>(Op.getOperand(2))->getValue();
  SDLoc DL(Op);

  // The initial values of each field.
  const unsigned NumFields = 4;
  SDValue Fields[NumFields] = {
    DAG.getConstant(FuncInfo->getVarArgsFirstGPR(), DL, PtrVT),
    DAG.getConstant(FuncInfo->getVarArgsFirstFPR(), DL, PtrVT),
    DAG.getFrameIndex(FuncInfo->getVarArgsFrameIndex(), PtrVT),
    DAG.getFrameIndex(FuncInfo->getRegSaveFrameIndex(), PtrVT)
  };

  // Store each field into its respective slot.
  SDValue MemOps[NumFields];
  unsigned Offset = 0;
  for (unsigned I = 0; I < NumFields; ++I) {
    SDValue FieldAddr = Addr;
    if (Offset != 0)
      FieldAddr = DAG.getNode(ISD::ADD, DL, PtrVT, FieldAddr,
                              DAG.getIntPtrConstant(Offset, DL));
    MemOps[I] = DAG.getStore(Chain, DL, Fields[I], FieldAddr,
                             MachinePointerInfo(SV, Offset));
    Offset += 8;
  }
  return DAG.getNode(ISD::TokenFactor, DL, MVT::Other, MemOps);
}

SDValue SystemZTargetLowering::lowerVACOPY(SDValue Op,
                                           SelectionDAG &DAG) const {
  SDValue Chain      = Op.getOperand(0);
  SDValue DstPtr     = Op.getOperand(1);
  SDValue SrcPtr     = Op.getOperand(2);
  const Value *DstSV = cast<SrcValueSDNode>(Op.getOperand(3))->getValue();
  const Value *SrcSV = cast<SrcValueSDNode>(Op.getOperand(4))->getValue();
  SDLoc DL(Op);

  return DAG.getMemcpy(Chain, DL, DstPtr, SrcPtr, DAG.getIntPtrConstant(32, DL),
                       Align(8), /*isVolatile*/ false, /*AlwaysInline*/ false,
                       /*isTailCall*/ false, MachinePointerInfo(DstSV),
                       MachinePointerInfo(SrcSV));
}

SDValue SystemZTargetLowering::
lowerDYNAMIC_STACKALLOC(SDValue Op, SelectionDAG &DAG) const {
  const TargetFrameLowering *TFI = Subtarget.getFrameLowering();
  MachineFunction &MF = DAG.getMachineFunction();
  bool RealignOpt = !MF.getFunction().hasFnAttribute("no-realign-stack");
  bool StoreBackchain = MF.getFunction().hasFnAttribute("backchain");

  SDValue Chain = Op.getOperand(0);
  SDValue Size  = Op.getOperand(1);
  SDValue Align = Op.getOperand(2);
  SDLoc DL(Op);

  // If user has set the no alignment function attribute, ignore
  // alloca alignments.
  uint64_t AlignVal = (RealignOpt ?
                       dyn_cast<ConstantSDNode>(Align)->getZExtValue() : 0);

  uint64_t StackAlign = TFI->getStackAlignment();
  uint64_t RequiredAlign = std::max(AlignVal, StackAlign);
  uint64_t ExtraAlignSpace = RequiredAlign - StackAlign;

  unsigned SPReg = getStackPointerRegisterToSaveRestore();
  SDValue NeededSpace = Size;

  // Get a reference to the stack pointer.
  SDValue OldSP = DAG.getCopyFromReg(Chain, DL, SPReg, MVT::i64);

  // If we need a backchain, save it now.
  SDValue Backchain;
  if (StoreBackchain)
    Backchain = DAG.getLoad(MVT::i64, DL, Chain, OldSP, MachinePointerInfo());

  // Add extra space for alignment if needed.
  if (ExtraAlignSpace)
    NeededSpace = DAG.getNode(ISD::ADD, DL, MVT::i64, NeededSpace,
                              DAG.getConstant(ExtraAlignSpace, DL, MVT::i64));

  // Get the new stack pointer value.
  SDValue NewSP;
  if (hasInlineStackProbe(MF)) {
    NewSP = DAG.getNode(SystemZISD::PROBED_ALLOCA, DL,
                DAG.getVTList(MVT::i64, MVT::Other), Chain, OldSP, NeededSpace);
    Chain = NewSP.getValue(1);
  }
  else {
    NewSP = DAG.getNode(ISD::SUB, DL, MVT::i64, OldSP, NeededSpace);
    // Copy the new stack pointer back.
    Chain = DAG.getCopyToReg(Chain, DL, SPReg, NewSP);
  }

  // The allocated data lives above the 160 bytes allocated for the standard
  // frame, plus any outgoing stack arguments.  We don't know how much that
  // amounts to yet, so emit a special ADJDYNALLOC placeholder.
  SDValue ArgAdjust = DAG.getNode(SystemZISD::ADJDYNALLOC, DL, MVT::i64);
  SDValue Result = DAG.getNode(ISD::ADD, DL, MVT::i64, NewSP, ArgAdjust);

  // Dynamically realign if needed.
  if (RequiredAlign > StackAlign) {
    Result =
      DAG.getNode(ISD::ADD, DL, MVT::i64, Result,
                  DAG.getConstant(ExtraAlignSpace, DL, MVT::i64));
    Result =
      DAG.getNode(ISD::AND, DL, MVT::i64, Result,
                  DAG.getConstant(~(RequiredAlign - 1), DL, MVT::i64));
  }

  if (StoreBackchain)
    Chain = DAG.getStore(Chain, DL, Backchain, NewSP, MachinePointerInfo());

  SDValue Ops[2] = { Result, Chain };
  return DAG.getMergeValues(Ops, DL);
}

SDValue SystemZTargetLowering::lowerGET_DYNAMIC_AREA_OFFSET(
    SDValue Op, SelectionDAG &DAG) const {
  SDLoc DL(Op);

  return DAG.getNode(SystemZISD::ADJDYNALLOC, DL, MVT::i64);
}

SDValue SystemZTargetLowering::lowerSMUL_LOHI(SDValue Op,
                                              SelectionDAG &DAG) const {
  EVT VT = Op.getValueType();
  SDLoc DL(Op);
  SDValue Ops[2];
  if (is32Bit(VT))
    // Just do a normal 64-bit multiplication and extract the results.
    // We define this so that it can be used for constant division.
    lowerMUL_LOHI32(DAG, DL, ISD::SIGN_EXTEND, Op.getOperand(0),
                    Op.getOperand(1), Ops[1], Ops[0]);
  else if (Subtarget.hasMiscellaneousExtensions2())
    // SystemZISD::SMUL_LOHI returns the low result in the odd register and
    // the high result in the even register.  ISD::SMUL_LOHI is defined to
    // return the low half first, so the results are in reverse order.
    lowerGR128Binary(DAG, DL, VT, SystemZISD::SMUL_LOHI,
                     Op.getOperand(0), Op.getOperand(1), Ops[1], Ops[0]);
  else {
    // Do a full 128-bit multiplication based on SystemZISD::UMUL_LOHI:
    //
    //   (ll * rl) + ((lh * rl) << 64) + ((ll * rh) << 64)
    //
    // but using the fact that the upper halves are either all zeros
    // or all ones:
    //
    //   (ll * rl) - ((lh & rl) << 64) - ((ll & rh) << 64)
    //
    // and grouping the right terms together since they are quicker than the
    // multiplication:
    //
    //   (ll * rl) - (((lh & rl) + (ll & rh)) << 64)
    SDValue C63 = DAG.getConstant(63, DL, MVT::i64);
    SDValue LL = Op.getOperand(0);
    SDValue RL = Op.getOperand(1);
    SDValue LH = DAG.getNode(ISD::SRA, DL, VT, LL, C63);
    SDValue RH = DAG.getNode(ISD::SRA, DL, VT, RL, C63);
    // SystemZISD::UMUL_LOHI returns the low result in the odd register and
    // the high result in the even register.  ISD::SMUL_LOHI is defined to
    // return the low half first, so the results are in reverse order.
    lowerGR128Binary(DAG, DL, VT, SystemZISD::UMUL_LOHI,
                     LL, RL, Ops[1], Ops[0]);
    SDValue NegLLTimesRH = DAG.getNode(ISD::AND, DL, VT, LL, RH);
    SDValue NegLHTimesRL = DAG.getNode(ISD::AND, DL, VT, LH, RL);
    SDValue NegSum = DAG.getNode(ISD::ADD, DL, VT, NegLLTimesRH, NegLHTimesRL);
    Ops[1] = DAG.getNode(ISD::SUB, DL, VT, Ops[1], NegSum);
  }
  return DAG.getMergeValues(Ops, DL);
}

SDValue SystemZTargetLowering::lowerUMUL_LOHI(SDValue Op,
                                              SelectionDAG &DAG) const {
  EVT VT = Op.getValueType();
  SDLoc DL(Op);
  SDValue Ops[2];
  if (is32Bit(VT))
    // Just do a normal 64-bit multiplication and extract the results.
    // We define this so that it can be used for constant division.
    lowerMUL_LOHI32(DAG, DL, ISD::ZERO_EXTEND, Op.getOperand(0),
                    Op.getOperand(1), Ops[1], Ops[0]);
  else
    // SystemZISD::UMUL_LOHI returns the low result in the odd register and
    // the high result in the even register.  ISD::UMUL_LOHI is defined to
    // return the low half first, so the results are in reverse order.
    lowerGR128Binary(DAG, DL, VT, SystemZISD::UMUL_LOHI,
                     Op.getOperand(0), Op.getOperand(1), Ops[1], Ops[0]);
  return DAG.getMergeValues(Ops, DL);
}

SDValue SystemZTargetLowering::lowerSDIVREM(SDValue Op,
                                            SelectionDAG &DAG) const {
  SDValue Op0 = Op.getOperand(0);
  SDValue Op1 = Op.getOperand(1);
  EVT VT = Op.getValueType();
  SDLoc DL(Op);

  // We use DSGF for 32-bit division.  This means the first operand must
  // always be 64-bit, and the second operand should be 32-bit whenever
  // that is possible, to improve performance.
  if (is32Bit(VT))
    Op0 = DAG.getNode(ISD::SIGN_EXTEND, DL, MVT::i64, Op0);
  else if (DAG.ComputeNumSignBits(Op1) > 32)
    Op1 = DAG.getNode(ISD::TRUNCATE, DL, MVT::i32, Op1);

  // DSG(F) returns the remainder in the even register and the
  // quotient in the odd register.
  SDValue Ops[2];
  lowerGR128Binary(DAG, DL, VT, SystemZISD::SDIVREM, Op0, Op1, Ops[1], Ops[0]);
  return DAG.getMergeValues(Ops, DL);
}

SDValue SystemZTargetLowering::lowerUDIVREM(SDValue Op,
                                            SelectionDAG &DAG) const {
  EVT VT = Op.getValueType();
  SDLoc DL(Op);

  // DL(G) returns the remainder in the even register and the
  // quotient in the odd register.
  SDValue Ops[2];
  lowerGR128Binary(DAG, DL, VT, SystemZISD::UDIVREM,
                   Op.getOperand(0), Op.getOperand(1), Ops[1], Ops[0]);
  return DAG.getMergeValues(Ops, DL);
}

SDValue SystemZTargetLowering::lowerOR(SDValue Op, SelectionDAG &DAG) const {
  assert(Op.getValueType() == MVT::i64 && "Should be 64-bit operation");

  // Get the known-zero masks for each operand.
  SDValue Ops[] = {Op.getOperand(0), Op.getOperand(1)};
  KnownBits Known[2] = {DAG.computeKnownBits(Ops[0]),
                        DAG.computeKnownBits(Ops[1])};

  // See if the upper 32 bits of one operand and the lower 32 bits of the
  // other are known zero.  They are the low and high operands respectively.
  uint64_t Masks[] = { Known[0].Zero.getZExtValue(),
                       Known[1].Zero.getZExtValue() };
  unsigned High, Low;
  if ((Masks[0] >> 32) == 0xffffffff && uint32_t(Masks[1]) == 0xffffffff)
    High = 1, Low = 0;
  else if ((Masks[1] >> 32) == 0xffffffff && uint32_t(Masks[0]) == 0xffffffff)
    High = 0, Low = 1;
  else
    return Op;

  SDValue LowOp = Ops[Low];
  SDValue HighOp = Ops[High];

  // If the high part is a constant, we're better off using IILH.
  if (HighOp.getOpcode() == ISD::Constant)
    return Op;

  // If the low part is a constant that is outside the range of LHI,
  // then we're better off using IILF.
  if (LowOp.getOpcode() == ISD::Constant) {
    int64_t Value = int32_t(cast<ConstantSDNode>(LowOp)->getZExtValue());
    if (!isInt<16>(Value))
      return Op;
  }

  // Check whether the high part is an AND that doesn't change the
  // high 32 bits and just masks out low bits.  We can skip it if so.
  if (HighOp.getOpcode() == ISD::AND &&
      HighOp.getOperand(1).getOpcode() == ISD::Constant) {
    SDValue HighOp0 = HighOp.getOperand(0);
    uint64_t Mask = cast<ConstantSDNode>(HighOp.getOperand(1))->getZExtValue();
    if (DAG.MaskedValueIsZero(HighOp0, APInt(64, ~(Mask | 0xffffffff))))
      HighOp = HighOp0;
  }

  // Take advantage of the fact that all GR32 operations only change the
  // low 32 bits by truncating Low to an i32 and inserting it directly
  // using a subreg.  The interesting cases are those where the truncation
  // can be folded.
  SDLoc DL(Op);
  SDValue Low32 = DAG.getNode(ISD::TRUNCATE, DL, MVT::i32, LowOp);
  return DAG.getTargetInsertSubreg(SystemZ::subreg_l32, DL,
                                   MVT::i64, HighOp, Low32);
}

// Lower SADDO/SSUBO/UADDO/USUBO nodes.
SDValue SystemZTargetLowering::lowerXALUO(SDValue Op,
                                          SelectionDAG &DAG) const {
  SDNode *N = Op.getNode();
  SDValue LHS = N->getOperand(0);
  SDValue RHS = N->getOperand(1);
  SDLoc DL(N);
  unsigned BaseOp = 0;
  unsigned CCValid = 0;
  unsigned CCMask = 0;

  switch (Op.getOpcode()) {
  default: llvm_unreachable("Unknown instruction!");
  case ISD::SADDO:
    BaseOp = SystemZISD::SADDO;
    CCValid = SystemZ::CCMASK_ARITH;
    CCMask = SystemZ::CCMASK_ARITH_OVERFLOW;
    break;
  case ISD::SSUBO:
    BaseOp = SystemZISD::SSUBO;
    CCValid = SystemZ::CCMASK_ARITH;
    CCMask = SystemZ::CCMASK_ARITH_OVERFLOW;
    break;
  case ISD::UADDO:
    BaseOp = SystemZISD::UADDO;
    CCValid = SystemZ::CCMASK_LOGICAL;
    CCMask = SystemZ::CCMASK_LOGICAL_CARRY;
    break;
  case ISD::USUBO:
    BaseOp = SystemZISD::USUBO;
    CCValid = SystemZ::CCMASK_LOGICAL;
    CCMask = SystemZ::CCMASK_LOGICAL_BORROW;
    break;
  }

  SDVTList VTs = DAG.getVTList(N->getValueType(0), MVT::i32);
  SDValue Result = DAG.getNode(BaseOp, DL, VTs, LHS, RHS);

  SDValue SetCC = emitSETCC(DAG, DL, Result.getValue(1), CCValid, CCMask);
  if (N->getValueType(1) == MVT::i1)
    SetCC = DAG.getNode(ISD::TRUNCATE, DL, MVT::i1, SetCC);

  return DAG.getNode(ISD::MERGE_VALUES, DL, N->getVTList(), Result, SetCC);
}

static bool isAddCarryChain(SDValue Carry) {
  while (Carry.getOpcode() == ISD::ADDCARRY)
    Carry = Carry.getOperand(2);
  return Carry.getOpcode() == ISD::UADDO;
}

static bool isSubBorrowChain(SDValue Carry) {
  while (Carry.getOpcode() == ISD::SUBCARRY)
    Carry = Carry.getOperand(2);
  return Carry.getOpcode() == ISD::USUBO;
}

// Lower ADDCARRY/SUBCARRY nodes.
SDValue SystemZTargetLowering::lowerADDSUBCARRY(SDValue Op,
                                                SelectionDAG &DAG) const {

  SDNode *N = Op.getNode();
  MVT VT = N->getSimpleValueType(0);

  // Let legalize expand this if it isn't a legal type yet.
  if (!DAG.getTargetLoweringInfo().isTypeLegal(VT))
    return SDValue();

  SDValue LHS = N->getOperand(0);
  SDValue RHS = N->getOperand(1);
  SDValue Carry = Op.getOperand(2);
  SDLoc DL(N);
  unsigned BaseOp = 0;
  unsigned CCValid = 0;
  unsigned CCMask = 0;

  switch (Op.getOpcode()) {
  default: llvm_unreachable("Unknown instruction!");
  case ISD::ADDCARRY:
    if (!isAddCarryChain(Carry))
      return SDValue();

    BaseOp = SystemZISD::ADDCARRY;
    CCValid = SystemZ::CCMASK_LOGICAL;
    CCMask = SystemZ::CCMASK_LOGICAL_CARRY;
    break;
  case ISD::SUBCARRY:
    if (!isSubBorrowChain(Carry))
      return SDValue();

    BaseOp = SystemZISD::SUBCARRY;
    CCValid = SystemZ::CCMASK_LOGICAL;
    CCMask = SystemZ::CCMASK_LOGICAL_BORROW;
    break;
  }

  // Set the condition code from the carry flag.
  Carry = DAG.getNode(SystemZISD::GET_CCMASK, DL, MVT::i32, Carry,
                      DAG.getConstant(CCValid, DL, MVT::i32),
                      DAG.getConstant(CCMask, DL, MVT::i32));

  SDVTList VTs = DAG.getVTList(VT, MVT::i32);
  SDValue Result = DAG.getNode(BaseOp, DL, VTs, LHS, RHS, Carry);

  SDValue SetCC = emitSETCC(DAG, DL, Result.getValue(1), CCValid, CCMask);
  if (N->getValueType(1) == MVT::i1)
    SetCC = DAG.getNode(ISD::TRUNCATE, DL, MVT::i1, SetCC);

  return DAG.getNode(ISD::MERGE_VALUES, DL, N->getVTList(), Result, SetCC);
}

SDValue SystemZTargetLowering::lowerCTPOP(SDValue Op,
                                          SelectionDAG &DAG) const {
  EVT VT = Op.getValueType();
  SDLoc DL(Op);
  Op = Op.getOperand(0);

  // Handle vector types via VPOPCT.
  if (VT.isVector()) {
    Op = DAG.getNode(ISD::BITCAST, DL, MVT::v16i8, Op);
    Op = DAG.getNode(SystemZISD::POPCNT, DL, MVT::v16i8, Op);
    switch (VT.getScalarSizeInBits()) {
    case 8:
      break;
    case 16: {
      Op = DAG.getNode(ISD::BITCAST, DL, VT, Op);
      SDValue Shift = DAG.getConstant(8, DL, MVT::i32);
      SDValue Tmp = DAG.getNode(SystemZISD::VSHL_BY_SCALAR, DL, VT, Op, Shift);
      Op = DAG.getNode(ISD::ADD, DL, VT, Op, Tmp);
      Op = DAG.getNode(SystemZISD::VSRL_BY_SCALAR, DL, VT, Op, Shift);
      break;
    }
    case 32: {
      SDValue Tmp = DAG.getSplatBuildVector(MVT::v16i8, DL,
                                            DAG.getConstant(0, DL, MVT::i32));
      Op = DAG.getNode(SystemZISD::VSUM, DL, VT, Op, Tmp);
      break;
    }
    case 64: {
      SDValue Tmp = DAG.getSplatBuildVector(MVT::v16i8, DL,
                                            DAG.getConstant(0, DL, MVT::i32));
      Op = DAG.getNode(SystemZISD::VSUM, DL, MVT::v4i32, Op, Tmp);
      Op = DAG.getNode(SystemZISD::VSUM, DL, VT, Op, Tmp);
      break;
    }
    default:
      llvm_unreachable("Unexpected type");
    }
    return Op;
  }

  // Get the known-zero mask for the operand.
  KnownBits Known = DAG.computeKnownBits(Op);
  unsigned NumSignificantBits = Known.getMaxValue().getActiveBits();
  if (NumSignificantBits == 0)
    return DAG.getConstant(0, DL, VT);

  // Skip known-zero high parts of the operand.
  int64_t OrigBitSize = VT.getSizeInBits();
  int64_t BitSize = (int64_t)1 << Log2_32_Ceil(NumSignificantBits);
  BitSize = std::min(BitSize, OrigBitSize);

  // The POPCNT instruction counts the number of bits in each byte.
  Op = DAG.getNode(ISD::ANY_EXTEND, DL, MVT::i64, Op);
  Op = DAG.getNode(SystemZISD::POPCNT, DL, MVT::i64, Op);
  Op = DAG.getNode(ISD::TRUNCATE, DL, VT, Op);

  // Add up per-byte counts in a binary tree.  All bits of Op at
  // position larger than BitSize remain zero throughout.
  for (int64_t I = BitSize / 2; I >= 8; I = I / 2) {
    SDValue Tmp = DAG.getNode(ISD::SHL, DL, VT, Op, DAG.getConstant(I, DL, VT));
    if (BitSize != OrigBitSize)
      Tmp = DAG.getNode(ISD::AND, DL, VT, Tmp,
                        DAG.getConstant(((uint64_t)1 << BitSize) - 1, DL, VT));
    Op = DAG.getNode(ISD::ADD, DL, VT, Op, Tmp);
  }

  // Extract overall result from high byte.
  if (BitSize > 8)
    Op = DAG.getNode(ISD::SRL, DL, VT, Op,
                     DAG.getConstant(BitSize - 8, DL, VT));

  return Op;
}

SDValue SystemZTargetLowering::lowerATOMIC_FENCE(SDValue Op,
                                                 SelectionDAG &DAG) const {
  SDLoc DL(Op);
  AtomicOrdering FenceOrdering = static_cast<AtomicOrdering>(
    cast<ConstantSDNode>(Op.getOperand(1))->getZExtValue());
  SyncScope::ID FenceSSID = static_cast<SyncScope::ID>(
    cast<ConstantSDNode>(Op.getOperand(2))->getZExtValue());

  // The only fence that needs an instruction is a sequentially-consistent
  // cross-thread fence.
  if (FenceOrdering == AtomicOrdering::SequentiallyConsistent &&
      FenceSSID == SyncScope::System) {
    return SDValue(DAG.getMachineNode(SystemZ::Serialize, DL, MVT::Other,
                                      Op.getOperand(0)),
                   0);
  }

  // MEMBARRIER is a compiler barrier; it codegens to a no-op.
  return DAG.getNode(SystemZISD::MEMBARRIER, DL, MVT::Other, Op.getOperand(0));
}

// Op is an atomic load.  Lower it into a normal volatile load.
SDValue SystemZTargetLowering::lowerATOMIC_LOAD(SDValue Op,
                                                SelectionDAG &DAG) const {
  auto *Node = cast<AtomicSDNode>(Op.getNode());
  return DAG.getExtLoad(ISD::EXTLOAD, SDLoc(Op), Op.getValueType(),
                        Node->getChain(), Node->getBasePtr(),
                        Node->getMemoryVT(), Node->getMemOperand());
}

// Op is an atomic store.  Lower it into a normal volatile store.
SDValue SystemZTargetLowering::lowerATOMIC_STORE(SDValue Op,
                                                 SelectionDAG &DAG) const {
  auto *Node = cast<AtomicSDNode>(Op.getNode());
  SDValue Chain = DAG.getTruncStore(Node->getChain(), SDLoc(Op), Node->getVal(),
                                    Node->getBasePtr(), Node->getMemoryVT(),
                                    Node->getMemOperand());
  // We have to enforce sequential consistency by performing a
  // serialization operation after the store.
  if (Node->getOrdering() == AtomicOrdering::SequentiallyConsistent)
    Chain = SDValue(DAG.getMachineNode(SystemZ::Serialize, SDLoc(Op),
                                       MVT::Other, Chain), 0);
  return Chain;
}

// Op is an 8-, 16-bit or 32-bit ATOMIC_LOAD_* operation.  Lower the first
// two into the fullword ATOMIC_LOADW_* operation given by Opcode.
SDValue SystemZTargetLowering::lowerATOMIC_LOAD_OP(SDValue Op,
                                                   SelectionDAG &DAG,
                                                   unsigned Opcode) const {
  auto *Node = cast<AtomicSDNode>(Op.getNode());

  // 32-bit operations need no code outside the main loop.
  EVT NarrowVT = Node->getMemoryVT();
  EVT WideVT = MVT::i32;
  if (NarrowVT == WideVT)
    return Op;

  int64_t BitSize = NarrowVT.getSizeInBits();
  SDValue ChainIn = Node->getChain();
  SDValue Addr = Node->getBasePtr();
  SDValue Src2 = Node->getVal();
  MachineMemOperand *MMO = Node->getMemOperand();
  SDLoc DL(Node);
  EVT PtrVT = Addr.getValueType();

  // Convert atomic subtracts of constants into additions.
  if (Opcode == SystemZISD::ATOMIC_LOADW_SUB)
    if (auto *Const = dyn_cast<ConstantSDNode>(Src2)) {
      Opcode = SystemZISD::ATOMIC_LOADW_ADD;
      Src2 = DAG.getConstant(-Const->getSExtValue(), DL, Src2.getValueType());
    }

  // Get the address of the containing word.
  SDValue AlignedAddr = DAG.getNode(ISD::AND, DL, PtrVT, Addr,
                                    DAG.getConstant(-4, DL, PtrVT));

  // Get the number of bits that the word must be rotated left in order
  // to bring the field to the top bits of a GR32.
  SDValue BitShift = DAG.getNode(ISD::SHL, DL, PtrVT, Addr,
                                 DAG.getConstant(3, DL, PtrVT));
  BitShift = DAG.getNode(ISD::TRUNCATE, DL, WideVT, BitShift);

  // Get the complementing shift amount, for rotating a field in the top
  // bits back to its proper position.
  SDValue NegBitShift = DAG.getNode(ISD::SUB, DL, WideVT,
                                    DAG.getConstant(0, DL, WideVT), BitShift);

  // Extend the source operand to 32 bits and prepare it for the inner loop.
  // ATOMIC_SWAPW uses RISBG to rotate the field left, but all other
  // operations require the source to be shifted in advance.  (This shift
  // can be folded if the source is constant.)  For AND and NAND, the lower
  // bits must be set, while for other opcodes they should be left clear.
  if (Opcode != SystemZISD::ATOMIC_SWAPW)
    Src2 = DAG.getNode(ISD::SHL, DL, WideVT, Src2,
                       DAG.getConstant(32 - BitSize, DL, WideVT));
  if (Opcode == SystemZISD::ATOMIC_LOADW_AND ||
      Opcode == SystemZISD::ATOMIC_LOADW_NAND)
    Src2 = DAG.getNode(ISD::OR, DL, WideVT, Src2,
                       DAG.getConstant(uint32_t(-1) >> BitSize, DL, WideVT));

  // Construct the ATOMIC_LOADW_* node.
  SDVTList VTList = DAG.getVTList(WideVT, MVT::Other);
  SDValue Ops[] = { ChainIn, AlignedAddr, Src2, BitShift, NegBitShift,
                    DAG.getConstant(BitSize, DL, WideVT) };
  SDValue AtomicOp = DAG.getMemIntrinsicNode(Opcode, DL, VTList, Ops,
                                             NarrowVT, MMO);

  // Rotate the result of the final CS so that the field is in the lower
  // bits of a GR32, then truncate it.
  SDValue ResultShift = DAG.getNode(ISD::ADD, DL, WideVT, BitShift,
                                    DAG.getConstant(BitSize, DL, WideVT));
  SDValue Result = DAG.getNode(ISD::ROTL, DL, WideVT, AtomicOp, ResultShift);

  SDValue RetOps[2] = { Result, AtomicOp.getValue(1) };
  return DAG.getMergeValues(RetOps, DL);
}

// Op is an ATOMIC_LOAD_SUB operation.  Lower 8- and 16-bit operations
// into ATOMIC_LOADW_SUBs and decide whether to convert 32- and 64-bit
// operations into additions.
SDValue SystemZTargetLowering::lowerATOMIC_LOAD_SUB(SDValue Op,
                                                    SelectionDAG &DAG) const {
  auto *Node = cast<AtomicSDNode>(Op.getNode());
  EVT MemVT = Node->getMemoryVT();
  if (MemVT == MVT::i32 || MemVT == MVT::i64) {
    // A full-width operation.
    assert(Op.getValueType() == MemVT && "Mismatched VTs");
    SDValue Src2 = Node->getVal();
    SDValue NegSrc2;
    SDLoc DL(Src2);

    if (auto *Op2 = dyn_cast<ConstantSDNode>(Src2)) {
      // Use an addition if the operand is constant and either LAA(G) is
      // available or the negative value is in the range of A(G)FHI.
      int64_t Value = (-Op2->getAPIntValue()).getSExtValue();
      if (isInt<32>(Value) || Subtarget.hasInterlockedAccess1())
        NegSrc2 = DAG.getConstant(Value, DL, MemVT);
    } else if (Subtarget.hasInterlockedAccess1())
      // Use LAA(G) if available.
      NegSrc2 = DAG.getNode(ISD::SUB, DL, MemVT, DAG.getConstant(0, DL, MemVT),
                            Src2);

    if (NegSrc2.getNode())
      return DAG.getAtomic(ISD::ATOMIC_LOAD_ADD, DL, MemVT,
                           Node->getChain(), Node->getBasePtr(), NegSrc2,
                           Node->getMemOperand());

    // Use the node as-is.
    return Op;
  }

  return lowerATOMIC_LOAD_OP(Op, DAG, SystemZISD::ATOMIC_LOADW_SUB);
}

// Lower 8/16/32/64-bit ATOMIC_CMP_SWAP_WITH_SUCCESS node.
SDValue SystemZTargetLowering::lowerATOMIC_CMP_SWAP(SDValue Op,
                                                    SelectionDAG &DAG) const {
  auto *Node = cast<AtomicSDNode>(Op.getNode());
  SDValue ChainIn = Node->getOperand(0);
  SDValue Addr = Node->getOperand(1);
  SDValue CmpVal = Node->getOperand(2);
  SDValue SwapVal = Node->getOperand(3);
  MachineMemOperand *MMO = Node->getMemOperand();
  SDLoc DL(Node);

  // We have native support for 32-bit and 64-bit compare and swap, but we
  // still need to expand extracting the "success" result from the CC.
  EVT NarrowVT = Node->getMemoryVT();
  EVT WideVT = NarrowVT == MVT::i64 ? MVT::i64 : MVT::i32;
  if (NarrowVT == WideVT) {
    SDVTList Tys = DAG.getVTList(WideVT, MVT::i32, MVT::Other);
    SDValue Ops[] = { ChainIn, Addr, CmpVal, SwapVal };
    SDValue AtomicOp = DAG.getMemIntrinsicNode(SystemZISD::ATOMIC_CMP_SWAP,
                                               DL, Tys, Ops, NarrowVT, MMO);
    SDValue Success = emitSETCC(DAG, DL, AtomicOp.getValue(1),
                                SystemZ::CCMASK_CS, SystemZ::CCMASK_CS_EQ);

    DAG.ReplaceAllUsesOfValueWith(Op.getValue(0), AtomicOp.getValue(0));
    DAG.ReplaceAllUsesOfValueWith(Op.getValue(1), Success);
    DAG.ReplaceAllUsesOfValueWith(Op.getValue(2), AtomicOp.getValue(2));
    return SDValue();
  }

  // Convert 8-bit and 16-bit compare and swap to a loop, implemented
  // via a fullword ATOMIC_CMP_SWAPW operation.
  int64_t BitSize = NarrowVT.getSizeInBits();
  EVT PtrVT = Addr.getValueType();

  // Get the address of the containing word.
  SDValue AlignedAddr = DAG.getNode(ISD::AND, DL, PtrVT, Addr,
                                    DAG.getConstant(-4, DL, PtrVT));

  // Get the number of bits that the word must be rotated left in order
  // to bring the field to the top bits of a GR32.
  SDValue BitShift = DAG.getNode(ISD::SHL, DL, PtrVT, Addr,
                                 DAG.getConstant(3, DL, PtrVT));
  BitShift = DAG.getNode(ISD::TRUNCATE, DL, WideVT, BitShift);

  // Get the complementing shift amount, for rotating a field in the top
  // bits back to its proper position.
  SDValue NegBitShift = DAG.getNode(ISD::SUB, DL, WideVT,
                                    DAG.getConstant(0, DL, WideVT), BitShift);

  // Construct the ATOMIC_CMP_SWAPW node.
  SDVTList VTList = DAG.getVTList(WideVT, MVT::i32, MVT::Other);
  SDValue Ops[] = { ChainIn, AlignedAddr, CmpVal, SwapVal, BitShift,
                    NegBitShift, DAG.getConstant(BitSize, DL, WideVT) };
  SDValue AtomicOp = DAG.getMemIntrinsicNode(SystemZISD::ATOMIC_CMP_SWAPW, DL,
                                             VTList, Ops, NarrowVT, MMO);
  SDValue Success = emitSETCC(DAG, DL, AtomicOp.getValue(1),
                              SystemZ::CCMASK_ICMP, SystemZ::CCMASK_CMP_EQ);

  DAG.ReplaceAllUsesOfValueWith(Op.getValue(0), AtomicOp.getValue(0));
  DAG.ReplaceAllUsesOfValueWith(Op.getValue(1), Success);
  DAG.ReplaceAllUsesOfValueWith(Op.getValue(2), AtomicOp.getValue(2));
  return SDValue();
}

MachineMemOperand::Flags
SystemZTargetLowering::getTargetMMOFlags(const Instruction &I) const {
  // Because of how we convert atomic_load and atomic_store to normal loads and
  // stores in the DAG, we need to ensure that the MMOs are marked volatile
  // since DAGCombine hasn't been updated to account for atomic, but non
  // volatile loads.  (See D57601)
  if (auto *SI = dyn_cast<StoreInst>(&I))
    if (SI->isAtomic())
      return MachineMemOperand::MOVolatile;
  if (auto *LI = dyn_cast<LoadInst>(&I))
    if (LI->isAtomic())
      return MachineMemOperand::MOVolatile;
  if (auto *AI = dyn_cast<AtomicRMWInst>(&I))
    if (AI->isAtomic())
      return MachineMemOperand::MOVolatile;
  if (auto *AI = dyn_cast<AtomicCmpXchgInst>(&I))
    if (AI->isAtomic())
      return MachineMemOperand::MOVolatile;
  return MachineMemOperand::MONone;
}

SDValue SystemZTargetLowering::lowerSTACKSAVE(SDValue Op,
                                              SelectionDAG &DAG) const {
  MachineFunction &MF = DAG.getMachineFunction();
  MF.getInfo<SystemZMachineFunctionInfo>()->setManipulatesSP(true);
  if (MF.getFunction().getCallingConv() == CallingConv::GHC)
    report_fatal_error("Variable-sized stack allocations are not supported "
                       "in GHC calling convention");
  return DAG.getCopyFromReg(Op.getOperand(0), SDLoc(Op),
                            SystemZ::R15D, Op.getValueType());
}

SDValue SystemZTargetLowering::lowerSTACKRESTORE(SDValue Op,
                                                 SelectionDAG &DAG) const {
  MachineFunction &MF = DAG.getMachineFunction();
  MF.getInfo<SystemZMachineFunctionInfo>()->setManipulatesSP(true);
  bool StoreBackchain = MF.getFunction().hasFnAttribute("backchain");

  if (MF.getFunction().getCallingConv() == CallingConv::GHC)
    report_fatal_error("Variable-sized stack allocations are not supported "
                       "in GHC calling convention");

  SDValue Chain = Op.getOperand(0);
  SDValue NewSP = Op.getOperand(1);
  SDValue Backchain;
  SDLoc DL(Op);

  if (StoreBackchain) {
    SDValue OldSP = DAG.getCopyFromReg(Chain, DL, SystemZ::R15D, MVT::i64);
    Backchain = DAG.getLoad(MVT::i64, DL, Chain, OldSP, MachinePointerInfo());
  }

  Chain = DAG.getCopyToReg(Chain, DL, SystemZ::R15D, NewSP);

  if (StoreBackchain)
    Chain = DAG.getStore(Chain, DL, Backchain, NewSP, MachinePointerInfo());

  return Chain;
}

SDValue SystemZTargetLowering::lowerPREFETCH(SDValue Op,
                                             SelectionDAG &DAG) const {
  bool IsData = cast<ConstantSDNode>(Op.getOperand(4))->getZExtValue();
  if (!IsData)
    // Just preserve the chain.
    return Op.getOperand(0);

  SDLoc DL(Op);
  bool IsWrite = cast<ConstantSDNode>(Op.getOperand(2))->getZExtValue();
  unsigned Code = IsWrite ? SystemZ::PFD_WRITE : SystemZ::PFD_READ;
  auto *Node = cast<MemIntrinsicSDNode>(Op.getNode());
  SDValue Ops[] = {Op.getOperand(0), DAG.getTargetConstant(Code, DL, MVT::i32),
                   Op.getOperand(1)};
  return DAG.getMemIntrinsicNode(SystemZISD::PREFETCH, DL,
                                 Node->getVTList(), Ops,
                                 Node->getMemoryVT(), Node->getMemOperand());
}

// Convert condition code in CCReg to an i32 value.
static SDValue getCCResult(SelectionDAG &DAG, SDValue CCReg) {
  SDLoc DL(CCReg);
  SDValue IPM = DAG.getNode(SystemZISD::IPM, DL, MVT::i32, CCReg);
  return DAG.getNode(ISD::SRL, DL, MVT::i32, IPM,
                     DAG.getConstant(SystemZ::IPM_CC, DL, MVT::i32));
}

SDValue
SystemZTargetLowering::lowerINTRINSIC_W_CHAIN(SDValue Op,
                                              SelectionDAG &DAG) const {
  unsigned Opcode, CCValid;
  if (isIntrinsicWithCCAndChain(Op, Opcode, CCValid)) {
    assert(Op->getNumValues() == 2 && "Expected only CC result and chain");
    SDNode *Node = emitIntrinsicWithCCAndChain(DAG, Op, Opcode);
    SDValue CC = getCCResult(DAG, SDValue(Node, 0));
    DAG.ReplaceAllUsesOfValueWith(SDValue(Op.getNode(), 0), CC);
    return SDValue();
  }

  return SDValue();
}

SDValue
SystemZTargetLowering::lowerINTRINSIC_WO_CHAIN(SDValue Op,
                                               SelectionDAG &DAG) const {
  unsigned Opcode, CCValid;
  if (isIntrinsicWithCC(Op, Opcode, CCValid)) {
    SDNode *Node = emitIntrinsicWithCC(DAG, Op, Opcode);
    if (Op->getNumValues() == 1)
      return getCCResult(DAG, SDValue(Node, 0));
    assert(Op->getNumValues() == 2 && "Expected a CC and non-CC result");
    return DAG.getNode(ISD::MERGE_VALUES, SDLoc(Op), Op->getVTList(),
                       SDValue(Node, 0), getCCResult(DAG, SDValue(Node, 1)));
  }

  unsigned Id = cast<ConstantSDNode>(Op.getOperand(0))->getZExtValue();
  switch (Id) {
  case Intrinsic::thread_pointer:
    return lowerThreadPointer(SDLoc(Op), DAG);

  case Intrinsic::s390_vpdi:
    return DAG.getNode(SystemZISD::PERMUTE_DWORDS, SDLoc(Op), Op.getValueType(),
                       Op.getOperand(1), Op.getOperand(2), Op.getOperand(3));

  case Intrinsic::s390_vperm:
    return DAG.getNode(SystemZISD::PERMUTE, SDLoc(Op), Op.getValueType(),
                       Op.getOperand(1), Op.getOperand(2), Op.getOperand(3));

  case Intrinsic::s390_vuphb:
  case Intrinsic::s390_vuphh:
  case Intrinsic::s390_vuphf:
    return DAG.getNode(SystemZISD::UNPACK_HIGH, SDLoc(Op), Op.getValueType(),
                       Op.getOperand(1));

  case Intrinsic::s390_vuplhb:
  case Intrinsic::s390_vuplhh:
  case Intrinsic::s390_vuplhf:
    return DAG.getNode(SystemZISD::UNPACKL_HIGH, SDLoc(Op), Op.getValueType(),
                       Op.getOperand(1));

  case Intrinsic::s390_vuplb:
  case Intrinsic::s390_vuplhw:
  case Intrinsic::s390_vuplf:
    return DAG.getNode(SystemZISD::UNPACK_LOW, SDLoc(Op), Op.getValueType(),
                       Op.getOperand(1));

  case Intrinsic::s390_vupllb:
  case Intrinsic::s390_vupllh:
  case Intrinsic::s390_vupllf:
    return DAG.getNode(SystemZISD::UNPACKL_LOW, SDLoc(Op), Op.getValueType(),
                       Op.getOperand(1));

  case Intrinsic::s390_vsumb:
  case Intrinsic::s390_vsumh:
  case Intrinsic::s390_vsumgh:
  case Intrinsic::s390_vsumgf:
  case Intrinsic::s390_vsumqf:
  case Intrinsic::s390_vsumqg:
    return DAG.getNode(SystemZISD::VSUM, SDLoc(Op), Op.getValueType(),
                       Op.getOperand(1), Op.getOperand(2));
  }

  return SDValue();
}

namespace {
// Says that SystemZISD operation Opcode can be used to perform the equivalent
// of a VPERM with permute vector Bytes.  If Opcode takes three operands,
// Operand is the constant third operand, otherwise it is the number of
// bytes in each element of the result.
struct Permute {
  unsigned Opcode;
  unsigned Operand;
  unsigned char Bytes[SystemZ::VectorBytes];
};
}

static const Permute PermuteForms[] = {
  // VMRHG
  { SystemZISD::MERGE_HIGH, 8,
    { 0, 1, 2, 3, 4, 5, 6, 7, 16, 17, 18, 19, 20, 21, 22, 23 } },
  // VMRHF
  { SystemZISD::MERGE_HIGH, 4,
    { 0, 1, 2, 3, 16, 17, 18, 19, 4, 5, 6, 7, 20, 21, 22, 23 } },
  // VMRHH
  { SystemZISD::MERGE_HIGH, 2,
    { 0, 1, 16, 17, 2, 3, 18, 19, 4, 5, 20, 21, 6, 7, 22, 23 } },
  // VMRHB
  { SystemZISD::MERGE_HIGH, 1,
    { 0, 16, 1, 17, 2, 18, 3, 19, 4, 20, 5, 21, 6, 22, 7, 23 } },
  // VMRLG
  { SystemZISD::MERGE_LOW, 8,
    { 8, 9, 10, 11, 12, 13, 14, 15, 24, 25, 26, 27, 28, 29, 30, 31 } },
  // VMRLF
  { SystemZISD::MERGE_LOW, 4,
    { 8, 9, 10, 11, 24, 25, 26, 27, 12, 13, 14, 15, 28, 29, 30, 31 } },
  // VMRLH
  { SystemZISD::MERGE_LOW, 2,
    { 8, 9, 24, 25, 10, 11, 26, 27, 12, 13, 28, 29, 14, 15, 30, 31 } },
  // VMRLB
  { SystemZISD::MERGE_LOW, 1,
    { 8, 24, 9, 25, 10, 26, 11, 27, 12, 28, 13, 29, 14, 30, 15, 31 } },
  // VPKG
  { SystemZISD::PACK, 4,
    { 4, 5, 6, 7, 12, 13, 14, 15, 20, 21, 22, 23, 28, 29, 30, 31 } },
  // VPKF
  { SystemZISD::PACK, 2,
    { 2, 3, 6, 7, 10, 11, 14, 15, 18, 19, 22, 23, 26, 27, 30, 31 } },
  // VPKH
  { SystemZISD::PACK, 1,
    { 1, 3, 5, 7, 9, 11, 13, 15, 17, 19, 21, 23, 25, 27, 29, 31 } },
  // VPDI V1, V2, 4  (low half of V1, high half of V2)
  { SystemZISD::PERMUTE_DWORDS, 4,
    { 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23 } },
  // VPDI V1, V2, 1  (high half of V1, low half of V2)
  { SystemZISD::PERMUTE_DWORDS, 1,
    { 0, 1, 2, 3, 4, 5, 6, 7, 24, 25, 26, 27, 28, 29, 30, 31 } }
};

// Called after matching a vector shuffle against a particular pattern.
// Both the original shuffle and the pattern have two vector operands.
// OpNos[0] is the operand of the original shuffle that should be used for
// operand 0 of the pattern, or -1 if operand 0 of the pattern can be anything.
// OpNos[1] is the same for operand 1 of the pattern.  Resolve these -1s and
// set OpNo0 and OpNo1 to the shuffle operands that should actually be used
// for operands 0 and 1 of the pattern.
static bool chooseShuffleOpNos(int *OpNos, unsigned &OpNo0, unsigned &OpNo1) {
  if (OpNos[0] < 0) {
    if (OpNos[1] < 0)
      return false;
    OpNo0 = OpNo1 = OpNos[1];
  } else if (OpNos[1] < 0) {
    OpNo0 = OpNo1 = OpNos[0];
  } else {
    OpNo0 = OpNos[0];
    OpNo1 = OpNos[1];
  }
  return true;
}

// Bytes is a VPERM-like permute vector, except that -1 is used for
// undefined bytes.  Return true if the VPERM can be implemented using P.
// When returning true set OpNo0 to the VPERM operand that should be
// used for operand 0 of P and likewise OpNo1 for operand 1 of P.
//
// For example, if swapping the VPERM operands allows P to match, OpNo0
// will be 1 and OpNo1 will be 0.  If instead Bytes only refers to one
// operand, but rewriting it to use two duplicated operands allows it to
// match P, then OpNo0 and OpNo1 will be the same.
static bool matchPermute(const SmallVectorImpl<int> &Bytes, const Permute &P,
                         unsigned &OpNo0, unsigned &OpNo1) {
  int OpNos[] = { -1, -1 };
  for (unsigned I = 0; I < SystemZ::VectorBytes; ++I) {
    int Elt = Bytes[I];
    if (Elt >= 0) {
      // Make sure that the two permute vectors use the same suboperand
      // byte number.  Only the operand numbers (the high bits) are
      // allowed to differ.
      if ((Elt ^ P.Bytes[I]) & (SystemZ::VectorBytes - 1))
        return false;
      int ModelOpNo = P.Bytes[I] / SystemZ::VectorBytes;
      int RealOpNo = unsigned(Elt) / SystemZ::VectorBytes;
      // Make sure that the operand mappings are consistent with previous
      // elements.
      if (OpNos[ModelOpNo] == 1 - RealOpNo)
        return false;
      OpNos[ModelOpNo] = RealOpNo;
    }
  }
  return chooseShuffleOpNos(OpNos, OpNo0, OpNo1);
}

// As above, but search for a matching permute.
static const Permute *matchPermute(const SmallVectorImpl<int> &Bytes,
                                   unsigned &OpNo0, unsigned &OpNo1) {
  for (auto &P : PermuteForms)
    if (matchPermute(Bytes, P, OpNo0, OpNo1))
      return &P;
  return nullptr;
}

// Bytes is a VPERM-like permute vector, except that -1 is used for
// undefined bytes.  This permute is an operand of an outer permute.
// See whether redistributing the -1 bytes gives a shuffle that can be
// implemented using P.  If so, set Transform to a VPERM-like permute vector
// that, when applied to the result of P, gives the original permute in Bytes.
static bool matchDoublePermute(const SmallVectorImpl<int> &Bytes,
                               const Permute &P,
                               SmallVectorImpl<int> &Transform) {
  unsigned To = 0;
  for (unsigned From = 0; From < SystemZ::VectorBytes; ++From) {
    int Elt = Bytes[From];
    if (Elt < 0)
      // Byte number From of the result is undefined.
      Transform[From] = -1;
    else {
      while (P.Bytes[To] != Elt) {
        To += 1;
        if (To == SystemZ::VectorBytes)
          return false;
      }
      Transform[From] = To;
    }
  }
  return true;
}

// As above, but search for a matching permute.
static const Permute *matchDoublePermute(const SmallVectorImpl<int> &Bytes,
                                         SmallVectorImpl<int> &Transform) {
  for (auto &P : PermuteForms)
    if (matchDoublePermute(Bytes, P, Transform))
      return &P;
  return nullptr;
}

// Convert the mask of the given shuffle op into a byte-level mask,
// as if it had type vNi8.
static bool getVPermMask(SDValue ShuffleOp,
                         SmallVectorImpl<int> &Bytes) {
  EVT VT = ShuffleOp.getValueType();
  unsigned NumElements = VT.getVectorNumElements();
  unsigned BytesPerElement = VT.getVectorElementType().getStoreSize();

  if (auto *VSN = dyn_cast<ShuffleVectorSDNode>(ShuffleOp)) {
    Bytes.resize(NumElements * BytesPerElement, -1);
    for (unsigned I = 0; I < NumElements; ++I) {
      int Index = VSN->getMaskElt(I);
      if (Index >= 0)
        for (unsigned J = 0; J < BytesPerElement; ++J)
          Bytes[I * BytesPerElement + J] = Index * BytesPerElement + J;
    }
    return true;
  }
  if (SystemZISD::SPLAT == ShuffleOp.getOpcode() &&
      isa<ConstantSDNode>(ShuffleOp.getOperand(1))) {
    unsigned Index = ShuffleOp.getConstantOperandVal(1);
    Bytes.resize(NumElements * BytesPerElement, -1);
    for (unsigned I = 0; I < NumElements; ++I)
      for (unsigned J = 0; J < BytesPerElement; ++J)
        Bytes[I * BytesPerElement + J] = Index * BytesPerElement + J;
    return true;
  }
  return false;
}

// Bytes is a VPERM-like permute vector, except that -1 is used for
// undefined bytes.  See whether bytes [Start, Start + BytesPerElement) of
// the result come from a contiguous sequence of bytes from one input.
// Set Base to the selector for the first byte if so.
static bool getShuffleInput(const SmallVectorImpl<int> &Bytes, unsigned Start,
                            unsigned BytesPerElement, int &Base) {
  Base = -1;
  for (unsigned I = 0; I < BytesPerElement; ++I) {
    if (Bytes[Start + I] >= 0) {
      unsigned Elem = Bytes[Start + I];
      if (Base < 0) {
        Base = Elem - I;
        // Make sure the bytes would come from one input operand.
        if (unsigned(Base) % Bytes.size() + BytesPerElement > Bytes.size())
          return false;
      } else if (unsigned(Base) != Elem - I)
        return false;
    }
  }
  return true;
}

// Bytes is a VPERM-like permute vector, except that -1 is used for
// undefined bytes.  Return true if it can be performed using VSLDB.
// When returning true, set StartIndex to the shift amount and OpNo0
// and OpNo1 to the VPERM operands that should be used as the first
// and second shift operand respectively.
static bool isShlDoublePermute(const SmallVectorImpl<int> &Bytes,
                               unsigned &StartIndex, unsigned &OpNo0,
                               unsigned &OpNo1) {
  int OpNos[] = { -1, -1 };
  int Shift = -1;
  for (unsigned I = 0; I < 16; ++I) {
    int Index = Bytes[I];
    if (Index >= 0) {
      int ExpectedShift = (Index - I) % SystemZ::VectorBytes;
      int ModelOpNo = unsigned(ExpectedShift + I) / SystemZ::VectorBytes;
      int RealOpNo = unsigned(Index) / SystemZ::VectorBytes;
      if (Shift < 0)
        Shift = ExpectedShift;
      else if (Shift != ExpectedShift)
        return false;
      // Make sure that the operand mappings are consistent with previous
      // elements.
      if (OpNos[ModelOpNo] == 1 - RealOpNo)
        return false;
      OpNos[ModelOpNo] = RealOpNo;
    }
  }
  StartIndex = Shift;
  return chooseShuffleOpNos(OpNos, OpNo0, OpNo1);
}

// Create a node that performs P on operands Op0 and Op1, casting the
// operands to the appropriate type.  The type of the result is determined by P.
static SDValue getPermuteNode(SelectionDAG &DAG, const SDLoc &DL,
                              const Permute &P, SDValue Op0, SDValue Op1) {
  // VPDI (PERMUTE_DWORDS) always operates on v2i64s.  The input
  // elements of a PACK are twice as wide as the outputs.
  unsigned InBytes = (P.Opcode == SystemZISD::PERMUTE_DWORDS ? 8 :
                      P.Opcode == SystemZISD::PACK ? P.Operand * 2 :
                      P.Operand);
  // Cast both operands to the appropriate type.
  MVT InVT = MVT::getVectorVT(MVT::getIntegerVT(InBytes * 8),
                              SystemZ::VectorBytes / InBytes);
  Op0 = DAG.getNode(ISD::BITCAST, DL, InVT, Op0);
  Op1 = DAG.getNode(ISD::BITCAST, DL, InVT, Op1);
  SDValue Op;
  if (P.Opcode == SystemZISD::PERMUTE_DWORDS) {
    SDValue Op2 = DAG.getTargetConstant(P.Operand, DL, MVT::i32);
    Op = DAG.getNode(SystemZISD::PERMUTE_DWORDS, DL, InVT, Op0, Op1, Op2);
  } else if (P.Opcode == SystemZISD::PACK) {
    MVT OutVT = MVT::getVectorVT(MVT::getIntegerVT(P.Operand * 8),
                                 SystemZ::VectorBytes / P.Operand);
    Op = DAG.getNode(SystemZISD::PACK, DL, OutVT, Op0, Op1);
  } else {
    Op = DAG.getNode(P.Opcode, DL, InVT, Op0, Op1);
  }
  return Op;
}

static bool isZeroVector(SDValue N) {
  if (N->getOpcode() == ISD::BITCAST)
    N = N->getOperand(0);
  if (N->getOpcode() == ISD::SPLAT_VECTOR)
    if (auto *Op = dyn_cast<ConstantSDNode>(N->getOperand(0)))
      return Op->getZExtValue() == 0;
  return ISD::isBuildVectorAllZeros(N.getNode());
}

// Return the index of the zero/undef vector, or UINT32_MAX if not found.
static uint32_t findZeroVectorIdx(SDValue *Ops, unsigned Num) {
  for (unsigned I = 0; I < Num ; I++)
    if (isZeroVector(Ops[I]))
      return I;
  return UINT32_MAX;
}

// Bytes is a VPERM-like permute vector, except that -1 is used for
// undefined bytes.  Implement it on operands Ops[0] and Ops[1] using
// VSLDB or VPERM.
static SDValue getGeneralPermuteNode(SelectionDAG &DAG, const SDLoc &DL,
                                     SDValue *Ops,
                                     const SmallVectorImpl<int> &Bytes) {
  for (unsigned I = 0; I < 2; ++I)
    Ops[I] = DAG.getNode(ISD::BITCAST, DL, MVT::v16i8, Ops[I]);

  // First see whether VSLDB can be used.
  unsigned StartIndex, OpNo0, OpNo1;
  if (isShlDoublePermute(Bytes, StartIndex, OpNo0, OpNo1))
    return DAG.getNode(SystemZISD::SHL_DOUBLE, DL, MVT::v16i8, Ops[OpNo0],
                       Ops[OpNo1],
                       DAG.getTargetConstant(StartIndex, DL, MVT::i32));

  // Fall back on VPERM.  Construct an SDNode for the permute vector.  Try to
  // eliminate a zero vector by reusing any zero index in the permute vector.
  unsigned ZeroVecIdx = findZeroVectorIdx(&Ops[0], 2);
  if (ZeroVecIdx != UINT32_MAX) {
    bool MaskFirst = true;
    int ZeroIdx = -1;
    for (unsigned I = 0; I < SystemZ::VectorBytes; ++I) {
      unsigned OpNo = unsigned(Bytes[I]) / SystemZ::VectorBytes;
      unsigned Byte = unsigned(Bytes[I]) % SystemZ::VectorBytes;
      if (OpNo == ZeroVecIdx && I == 0) {
        // If the first byte is zero, use mask as first operand.
        ZeroIdx = 0;
        break;
      }
      if (OpNo != ZeroVecIdx && Byte == 0) {
        // If mask contains a zero, use it by placing that vector first.
        ZeroIdx = I + SystemZ::VectorBytes;
        MaskFirst = false;
        break;
      }
    }
    if (ZeroIdx != -1) {
      SDValue IndexNodes[SystemZ::VectorBytes];
      for (unsigned I = 0; I < SystemZ::VectorBytes; ++I) {
        if (Bytes[I] >= 0) {
          unsigned OpNo = unsigned(Bytes[I]) / SystemZ::VectorBytes;
          unsigned Byte = unsigned(Bytes[I]) % SystemZ::VectorBytes;
          if (OpNo == ZeroVecIdx)
            IndexNodes[I] = DAG.getConstant(ZeroIdx, DL, MVT::i32);
          else {
            unsigned BIdx = MaskFirst ? Byte + SystemZ::VectorBytes : Byte;
            IndexNodes[I] = DAG.getConstant(BIdx, DL, MVT::i32);
          }
        } else
          IndexNodes[I] = DAG.getUNDEF(MVT::i32);
      }
      SDValue Mask = DAG.getBuildVector(MVT::v16i8, DL, IndexNodes);
      SDValue Src = ZeroVecIdx == 0 ? Ops[1] : Ops[0];
      if (MaskFirst)
        return DAG.getNode(SystemZISD::PERMUTE, DL, MVT::v16i8, Mask, Src,
                           Mask);
      else
        return DAG.getNode(SystemZISD::PERMUTE, DL, MVT::v16i8, Src, Mask,
                           Mask);
    }
  }

  SDValue IndexNodes[SystemZ::VectorBytes];
  for (unsigned I = 0; I < SystemZ::VectorBytes; ++I)
    if (Bytes[I] >= 0)
      IndexNodes[I] = DAG.getConstant(Bytes[I], DL, MVT::i32);
    else
      IndexNodes[I] = DAG.getUNDEF(MVT::i32);
  SDValue Op2 = DAG.getBuildVector(MVT::v16i8, DL, IndexNodes);
  return DAG.getNode(SystemZISD::PERMUTE, DL, MVT::v16i8, Ops[0],
                     (!Ops[1].isUndef() ? Ops[1] : Ops[0]), Op2);
}

namespace {
// Describes a general N-operand vector shuffle.
struct GeneralShuffle {
  GeneralShuffle(EVT vt) : VT(vt), UnpackFromEltSize(UINT_MAX) {}
  void addUndef();
  bool add(SDValue, unsigned);
  SDValue getNode(SelectionDAG &, const SDLoc &);
  void tryPrepareForUnpack();
  bool unpackWasPrepared() { return UnpackFromEltSize <= 4; }
  SDValue insertUnpackIfPrepared(SelectionDAG &DAG, const SDLoc &DL, SDValue Op);

  // The operands of the shuffle.
  SmallVector<SDValue, SystemZ::VectorBytes> Ops;

  // Index I is -1 if byte I of the result is undefined.  Otherwise the
  // result comes from byte Bytes[I] % SystemZ::VectorBytes of operand
  // Bytes[I] / SystemZ::VectorBytes.
  SmallVector<int, SystemZ::VectorBytes> Bytes;

  // The type of the shuffle result.
  EVT VT;

  // Holds a value of 1, 2 or 4 if a final unpack has been prepared for.
  unsigned UnpackFromEltSize;
};
}

// Add an extra undefined element to the shuffle.
void GeneralShuffle::addUndef() {
  unsigned BytesPerElement = VT.getVectorElementType().getStoreSize();
  for (unsigned I = 0; I < BytesPerElement; ++I)
    Bytes.push_back(-1);
}

// Add an extra element to the shuffle, taking it from element Elem of Op.
// A null Op indicates a vector input whose value will be calculated later;
// there is at most one such input per shuffle and it always has the same
// type as the result. Aborts and returns false if the source vector elements
// of an EXTRACT_VECTOR_ELT are smaller than the destination elements. Per
// LLVM they become implicitly extended, but this is rare and not optimized.
bool GeneralShuffle::add(SDValue Op, unsigned Elem) {
  unsigned BytesPerElement = VT.getVectorElementType().getStoreSize();

  // The source vector can have wider elements than the result,
  // either through an explicit TRUNCATE or because of type legalization.
  // We want the least significant part.
  EVT FromVT = Op.getNode() ? Op.getValueType() : VT;
  unsigned FromBytesPerElement = FromVT.getVectorElementType().getStoreSize();

  // Return false if the source elements are smaller than their destination
  // elements.
  if (FromBytesPerElement < BytesPerElement)
    return false;

  unsigned Byte = ((Elem * FromBytesPerElement) % SystemZ::VectorBytes +
                   (FromBytesPerElement - BytesPerElement));

  // Look through things like shuffles and bitcasts.
  while (Op.getNode()) {
    if (Op.getOpcode() == ISD::BITCAST)
      Op = Op.getOperand(0);
    else if (Op.getOpcode() == ISD::VECTOR_SHUFFLE && Op.hasOneUse()) {
      // See whether the bytes we need come from a contiguous part of one
      // operand.
      SmallVector<int, SystemZ::VectorBytes> OpBytes;
      if (!getVPermMask(Op, OpBytes))
        break;
      int NewByte;
      if (!getShuffleInput(OpBytes, Byte, BytesPerElement, NewByte))
        break;
      if (NewByte < 0) {
        addUndef();
        return true;
      }
      Op = Op.getOperand(unsigned(NewByte) / SystemZ::VectorBytes);
      Byte = unsigned(NewByte) % SystemZ::VectorBytes;
    } else if (Op.isUndef()) {
      addUndef();
      return true;
    } else
      break;
  }

  // Make sure that the source of the extraction is in Ops.
  unsigned OpNo = 0;
  for (; OpNo < Ops.size(); ++OpNo)
    if (Ops[OpNo] == Op)
      break;
  if (OpNo == Ops.size())
    Ops.push_back(Op);

  // Add the element to Bytes.
  unsigned Base = OpNo * SystemZ::VectorBytes + Byte;
  for (unsigned I = 0; I < BytesPerElement; ++I)
    Bytes.push_back(Base + I);

  return true;
}

// Return SDNodes for the completed shuffle.
SDValue GeneralShuffle::getNode(SelectionDAG &DAG, const SDLoc &DL) {
  assert(Bytes.size() == SystemZ::VectorBytes && "Incomplete vector");

  if (Ops.size() == 0)
    return DAG.getUNDEF(VT);

  // Use a single unpack if possible as the last operation.
  tryPrepareForUnpack();

  // Make sure that there are at least two shuffle operands.
  if (Ops.size() == 1)
    Ops.push_back(DAG.getUNDEF(MVT::v16i8));

  // Create a tree of shuffles, deferring root node until after the loop.
  // Try to redistribute the undefined elements of non-root nodes so that
  // the non-root shuffles match something like a pack or merge, then adjust
  // the parent node's permute vector to compensate for the new order.
  // Among other things, this copes with vectors like <2 x i16> that were
  // padded with undefined elements during type legalization.
  //
  // In the best case this redistribution will lead to the whole tree
  // using packs and merges.  It should rarely be a loss in other cases.
  unsigned Stride = 1;
  for (; Stride * 2 < Ops.size(); Stride *= 2) {
    for (unsigned I = 0; I < Ops.size() - Stride; I += Stride * 2) {
      SDValue SubOps[] = { Ops[I], Ops[I + Stride] };

      // Create a mask for just these two operands.
      SmallVector<int, SystemZ::VectorBytes> NewBytes(SystemZ::VectorBytes);
      for (unsigned J = 0; J < SystemZ::VectorBytes; ++J) {
        unsigned OpNo = unsigned(Bytes[J]) / SystemZ::VectorBytes;
        unsigned Byte = unsigned(Bytes[J]) % SystemZ::VectorBytes;
        if (OpNo == I)
          NewBytes[J] = Byte;
        else if (OpNo == I + Stride)
          NewBytes[J] = SystemZ::VectorBytes + Byte;
        else
          NewBytes[J] = -1;
      }
      // See if it would be better to reorganize NewMask to avoid using VPERM.
      SmallVector<int, SystemZ::VectorBytes> NewBytesMap(SystemZ::VectorBytes);
      if (const Permute *P = matchDoublePermute(NewBytes, NewBytesMap)) {
        Ops[I] = getPermuteNode(DAG, DL, *P, SubOps[0], SubOps[1]);
        // Applying NewBytesMap to Ops[I] gets back to NewBytes.
        for (unsigned J = 0; J < SystemZ::VectorBytes; ++J) {
          if (NewBytes[J] >= 0) {
            assert(unsigned(NewBytesMap[J]) < SystemZ::VectorBytes &&
                   "Invalid double permute");
            Bytes[J] = I * SystemZ::VectorBytes + NewBytesMap[J];
          } else
            assert(NewBytesMap[J] < 0 && "Invalid double permute");
        }
      } else {
        // Just use NewBytes on the operands.
        Ops[I] = getGeneralPermuteNode(DAG, DL, SubOps, NewBytes);
        for (unsigned J = 0; J < SystemZ::VectorBytes; ++J)
          if (NewBytes[J] >= 0)
            Bytes[J] = I * SystemZ::VectorBytes + J;
      }
    }
  }

  // Now we just have 2 inputs.  Put the second operand in Ops[1].
  if (Stride > 1) {
    Ops[1] = Ops[Stride];
    for (unsigned I = 0; I < SystemZ::VectorBytes; ++I)
      if (Bytes[I] >= int(SystemZ::VectorBytes))
        Bytes[I] -= (Stride - 1) * SystemZ::VectorBytes;
  }

  // Look for an instruction that can do the permute without resorting
  // to VPERM.
  unsigned OpNo0, OpNo1;
  SDValue Op;
  if (unpackWasPrepared() && Ops[1].isUndef())
    Op = Ops[0];
  else if (const Permute *P = matchPermute(Bytes, OpNo0, OpNo1))
    Op = getPermuteNode(DAG, DL, *P, Ops[OpNo0], Ops[OpNo1]);
  else
    Op = getGeneralPermuteNode(DAG, DL, &Ops[0], Bytes);

  Op = insertUnpackIfPrepared(DAG, DL, Op);

  return DAG.getNode(ISD::BITCAST, DL, VT, Op);
}

#ifndef NDEBUG
static void dumpBytes(const SmallVectorImpl<int> &Bytes, std::string Msg) {
  dbgs() << Msg.c_str() << " { ";
  for (unsigned i = 0; i < Bytes.size(); i++)
    dbgs() << Bytes[i] << " ";
  dbgs() << "}\n";
}
#endif

// If the Bytes vector matches an unpack operation, prepare to do the unpack
// after all else by removing the zero vector and the effect of the unpack on
// Bytes.
void GeneralShuffle::tryPrepareForUnpack() {
  uint32_t ZeroVecOpNo = findZeroVectorIdx(&Ops[0], Ops.size());
  if (ZeroVecOpNo == UINT32_MAX || Ops.size() == 1)
    return;

  // Only do this if removing the zero vector reduces the depth, otherwise
  // the critical path will increase with the final unpack.
  if (Ops.size() > 2 &&
      Log2_32_Ceil(Ops.size()) == Log2_32_Ceil(Ops.size() - 1))
    return;

  // Find an unpack that would allow removing the zero vector from Ops.
  UnpackFromEltSize = 1;
  for (; UnpackFromEltSize <= 4; UnpackFromEltSize *= 2) {
    bool MatchUnpack = true;
    SmallVector<int, SystemZ::VectorBytes> SrcBytes;
    for (unsigned Elt = 0; Elt < SystemZ::VectorBytes; Elt++) {
      unsigned ToEltSize = UnpackFromEltSize * 2;
      bool IsZextByte = (Elt % ToEltSize) < UnpackFromEltSize;
      if (!IsZextByte)
        SrcBytes.push_back(Bytes[Elt]);
      if (Bytes[Elt] != -1) {
        unsigned OpNo = unsigned(Bytes[Elt]) / SystemZ::VectorBytes;
        if (IsZextByte != (OpNo == ZeroVecOpNo)) {
          MatchUnpack = false;
          break;
        }
      }
    }
    if (MatchUnpack) {
      if (Ops.size() == 2) {
        // Don't use unpack if a single source operand needs rearrangement.
        for (unsigned i = 0; i < SystemZ::VectorBytes / 2; i++)
          if (SrcBytes[i] != -1 && SrcBytes[i] % 16 != int(i)) {
            UnpackFromEltSize = UINT_MAX;
            return;
          }
      }
      break;
    }
  }
  if (UnpackFromEltSize > 4)
    return;

  LLVM_DEBUG(dbgs() << "Preparing for final unpack of element size "
             << UnpackFromEltSize << ". Zero vector is Op#" << ZeroVecOpNo
             << ".\n";
             dumpBytes(Bytes, "Original Bytes vector:"););

  // Apply the unpack in reverse to the Bytes array.
  unsigned B = 0;
  for (unsigned Elt = 0; Elt < SystemZ::VectorBytes;) {
    Elt += UnpackFromEltSize;
    for (unsigned i = 0; i < UnpackFromEltSize; i++, Elt++, B++)
      Bytes[B] = Bytes[Elt];
  }
  while (B < SystemZ::VectorBytes)
    Bytes[B++] = -1;

  // Remove the zero vector from Ops
  Ops.erase(&Ops[ZeroVecOpNo]);
  for (unsigned I = 0; I < SystemZ::VectorBytes; ++I)
    if (Bytes[I] >= 0) {
      unsigned OpNo = unsigned(Bytes[I]) / SystemZ::VectorBytes;
      if (OpNo > ZeroVecOpNo)
        Bytes[I] -= SystemZ::VectorBytes;
    }

  LLVM_DEBUG(dumpBytes(Bytes, "Resulting Bytes vector, zero vector removed:");
             dbgs() << "\n";);
}

SDValue GeneralShuffle::insertUnpackIfPrepared(SelectionDAG &DAG,
                                               const SDLoc &DL,
                                               SDValue Op) {
  if (!unpackWasPrepared())
    return Op;
  unsigned InBits = UnpackFromEltSize * 8;
  EVT InVT = MVT::getVectorVT(MVT::getIntegerVT(InBits),
                                SystemZ::VectorBits / InBits);
  SDValue PackedOp = DAG.getNode(ISD::BITCAST, DL, InVT, Op);
  unsigned OutBits = InBits * 2;
  EVT OutVT = MVT::getVectorVT(MVT::getIntegerVT(OutBits),
                               SystemZ::VectorBits / OutBits);
  return DAG.getNode(SystemZISD::UNPACKL_HIGH, DL, OutVT, PackedOp);
}

// Return true if the given BUILD_VECTOR is a scalar-to-vector conversion.
static bool isScalarToVector(SDValue Op) {
  for (unsigned I = 1, E = Op.getNumOperands(); I != E; ++I)
    if (!Op.getOperand(I).isUndef())
      return false;
  return true;
}

// Return a vector of type VT that contains Value in the first element.
// The other elements don't matter.
static SDValue buildScalarToVector(SelectionDAG &DAG, const SDLoc &DL, EVT VT,
                                   SDValue Value) {
  // If we have a constant, replicate it to all elements and let the
  // BUILD_VECTOR lowering take care of it.
  if (Value.getOpcode() == ISD::Constant ||
      Value.getOpcode() == ISD::ConstantFP) {
    SmallVector<SDValue, 16> Ops(VT.getVectorNumElements(), Value);
    return DAG.getBuildVector(VT, DL, Ops);
  }
  if (Value.isUndef())
    return DAG.getUNDEF(VT);
  return DAG.getNode(ISD::SCALAR_TO_VECTOR, DL, VT, Value);
}

// Return a vector of type VT in which Op0 is in element 0 and Op1 is in
// element 1.  Used for cases in which replication is cheap.
static SDValue buildMergeScalars(SelectionDAG &DAG, const SDLoc &DL, EVT VT,
                                 SDValue Op0, SDValue Op1) {
  if (Op0.isUndef()) {
    if (Op1.isUndef())
      return DAG.getUNDEF(VT);
    return DAG.getNode(SystemZISD::REPLICATE, DL, VT, Op1);
  }
  if (Op1.isUndef())
    return DAG.getNode(SystemZISD::REPLICATE, DL, VT, Op0);
  return DAG.getNode(SystemZISD::MERGE_HIGH, DL, VT,
                     buildScalarToVector(DAG, DL, VT, Op0),
                     buildScalarToVector(DAG, DL, VT, Op1));
}

// Extend GPR scalars Op0 and Op1 to doublewords and return a v2i64
// vector for them.
static SDValue joinDwords(SelectionDAG &DAG, const SDLoc &DL, SDValue Op0,
                          SDValue Op1) {
  if (Op0.isUndef() && Op1.isUndef())
    return DAG.getUNDEF(MVT::v2i64);
  // If one of the two inputs is undefined then replicate the other one,
  // in order to avoid using another register unnecessarily.
  if (Op0.isUndef())
    Op0 = Op1 = DAG.getNode(ISD::ANY_EXTEND, DL, MVT::i64, Op1);
  else if (Op1.isUndef())
    Op0 = Op1 = DAG.getNode(ISD::ANY_EXTEND, DL, MVT::i64, Op0);
  else {
    Op0 = DAG.getNode(ISD::ANY_EXTEND, DL, MVT::i64, Op0);
    Op1 = DAG.getNode(ISD::ANY_EXTEND, DL, MVT::i64, Op1);
  }
  return DAG.getNode(SystemZISD::JOIN_DWORDS, DL, MVT::v2i64, Op0, Op1);
}

// If a BUILD_VECTOR contains some EXTRACT_VECTOR_ELTs, it's usually
// better to use VECTOR_SHUFFLEs on them, only using BUILD_VECTOR for
// the non-EXTRACT_VECTOR_ELT elements.  See if the given BUILD_VECTOR
// would benefit from this representation and return it if so.
static SDValue tryBuildVectorShuffle(SelectionDAG &DAG,
                                     BuildVectorSDNode *BVN) {
  EVT VT = BVN->getValueType(0);
  unsigned NumElements = VT.getVectorNumElements();

  // Represent the BUILD_VECTOR as an N-operand VECTOR_SHUFFLE-like operation
  // on byte vectors.  If there are non-EXTRACT_VECTOR_ELT elements that still
  // need a BUILD_VECTOR, add an additional placeholder operand for that
  // BUILD_VECTOR and store its operands in ResidueOps.
  GeneralShuffle GS(VT);
  SmallVector<SDValue, SystemZ::VectorBytes> ResidueOps;
  bool FoundOne = false;
  for (unsigned I = 0; I < NumElements; ++I) {
    SDValue Op = BVN->getOperand(I);
    if (Op.getOpcode() == ISD::TRUNCATE)
      Op = Op.getOperand(0);
    if (Op.getOpcode() == ISD::EXTRACT_VECTOR_ELT &&
        Op.getOperand(1).getOpcode() == ISD::Constant) {
      unsigned Elem = cast<ConstantSDNode>(Op.getOperand(1))->getZExtValue();
      if (!GS.add(Op.getOperand(0), Elem))
        return SDValue();
      FoundOne = true;
    } else if (Op.isUndef()) {
      GS.addUndef();
    } else {
      if (!GS.add(SDValue(), ResidueOps.size()))
        return SDValue();
      ResidueOps.push_back(BVN->getOperand(I));
    }
  }

  // Nothing to do if there are no EXTRACT_VECTOR_ELTs.
  if (!FoundOne)
    return SDValue();

  // Create the BUILD_VECTOR for the remaining elements, if any.
  if (!ResidueOps.empty()) {
    while (ResidueOps.size() < NumElements)
      ResidueOps.push_back(DAG.getUNDEF(ResidueOps[0].getValueType()));
    for (auto &Op : GS.Ops) {
      if (!Op.getNode()) {
        Op = DAG.getBuildVector(VT, SDLoc(BVN), ResidueOps);
        break;
      }
    }
  }
  return GS.getNode(DAG, SDLoc(BVN));
}

bool SystemZTargetLowering::isVectorElementLoad(SDValue Op) const {
  if (Op.getOpcode() == ISD::LOAD && cast<LoadSDNode>(Op)->isUnindexed())
    return true;
  if (Subtarget.hasVectorEnhancements2() && Op.getOpcode() == SystemZISD::LRV)
    return true;
  return false;
}

// Combine GPR scalar values Elems into a vector of type VT.
SDValue
SystemZTargetLowering::buildVector(SelectionDAG &DAG, const SDLoc &DL, EVT VT,
                                   SmallVectorImpl<SDValue> &Elems) const {
  // See whether there is a single replicated value.
  SDValue Single;
  unsigned int NumElements = Elems.size();
  unsigned int Count = 0;
  for (auto Elem : Elems) {
    if (!Elem.isUndef()) {
      if (!Single.getNode())
        Single = Elem;
      else if (Elem != Single) {
        Single = SDValue();
        break;
      }
      Count += 1;
    }
  }
  // There are three cases here:
  //
  // - if the only defined element is a loaded one, the best sequence
  //   is a replicating load.
  //
  // - otherwise, if the only defined element is an i64 value, we will
  //   end up with the same VLVGP sequence regardless of whether we short-cut
  //   for replication or fall through to the later code.
  //
  // - otherwise, if the only defined element is an i32 or smaller value,
  //   we would need 2 instructions to replicate it: VLVGP followed by VREPx.
  //   This is only a win if the single defined element is used more than once.
  //   In other cases we're better off using a single VLVGx.
  if (Single.getNode() && (Count > 1 || isVectorElementLoad(Single)))
    return DAG.getNode(SystemZISD::REPLICATE, DL, VT, Single);

  // If all elements are loads, use VLREP/VLEs (below).
  bool AllLoads = true;
  for (auto Elem : Elems)
    if (!isVectorElementLoad(Elem)) {
      AllLoads = false;
      break;
    }

  // The best way of building a v2i64 from two i64s is to use VLVGP.
  if (VT == MVT::v2i64 && !AllLoads)
    return joinDwords(DAG, DL, Elems[0], Elems[1]);

  // Use a 64-bit merge high to combine two doubles.
  if (VT == MVT::v2f64 && !AllLoads)
    return buildMergeScalars(DAG, DL, VT, Elems[0], Elems[1]);

  // Build v4f32 values directly from the FPRs:
  //
  //   <Axxx> <Bxxx> <Cxxxx> <Dxxx>
  //         V              V         VMRHF
  //      <ABxx>         <CDxx>
  //                V                 VMRHG
  //              <ABCD>
  if (VT == MVT::v4f32 && !AllLoads) {
    SDValue Op01 = buildMergeScalars(DAG, DL, VT, Elems[0], Elems[1]);
    SDValue Op23 = buildMergeScalars(DAG, DL, VT, Elems[2], Elems[3]);
    // Avoid unnecessary undefs by reusing the other operand.
    if (Op01.isUndef())
      Op01 = Op23;
    else if (Op23.isUndef())
      Op23 = Op01;
    // Merging identical replications is a no-op.
    if (Op01.getOpcode() == SystemZISD::REPLICATE && Op01 == Op23)
      return Op01;
    Op01 = DAG.getNode(ISD::BITCAST, DL, MVT::v2i64, Op01);
    Op23 = DAG.getNode(ISD::BITCAST, DL, MVT::v2i64, Op23);
    SDValue Op = DAG.getNode(SystemZISD::MERGE_HIGH,
                             DL, MVT::v2i64, Op01, Op23);
    return DAG.getNode(ISD::BITCAST, DL, VT, Op);
  }

  // Collect the constant terms.
  SmallVector<SDValue, SystemZ::VectorBytes> Constants(NumElements, SDValue());
  SmallVector<bool, SystemZ::VectorBytes> Done(NumElements, false);

  unsigned NumConstants = 0;
  for (unsigned I = 0; I < NumElements; ++I) {
    SDValue Elem = Elems[I];
    if (Elem.getOpcode() == ISD::Constant ||
        Elem.getOpcode() == ISD::ConstantFP) {
      NumConstants += 1;
      Constants[I] = Elem;
      Done[I] = true;
    }
  }
  // If there was at least one constant, fill in the other elements of
  // Constants with undefs to get a full vector constant and use that
  // as the starting point.
  SDValue Result;
  SDValue ReplicatedVal;
  if (NumConstants > 0) {
    for (unsigned I = 0; I < NumElements; ++I)
      if (!Constants[I].getNode())
        Constants[I] = DAG.getUNDEF(Elems[I].getValueType());
    Result = DAG.getBuildVector(VT, DL, Constants);
  } else {
    // Otherwise try to use VLREP or VLVGP to start the sequence in order to
    // avoid a false dependency on any previous contents of the vector
    // register.

    // Use a VLREP if at least one element is a load. Make sure to replicate
    // the load with the most elements having its value.
    std::map<const SDNode*, unsigned> UseCounts;
    SDNode *LoadMaxUses = nullptr;
    for (unsigned I = 0; I < NumElements; ++I)
      if (isVectorElementLoad(Elems[I])) {
        SDNode *Ld = Elems[I].getNode();
        UseCounts[Ld]++;
        if (LoadMaxUses == nullptr || UseCounts[LoadMaxUses] < UseCounts[Ld])
          LoadMaxUses = Ld;
      }
    if (LoadMaxUses != nullptr) {
      ReplicatedVal = SDValue(LoadMaxUses, 0);
      Result = DAG.getNode(SystemZISD::REPLICATE, DL, VT, ReplicatedVal);
    } else {
      // Try to use VLVGP.
      unsigned I1 = NumElements / 2 - 1;
      unsigned I2 = NumElements - 1;
      bool Def1 = !Elems[I1].isUndef();
      bool Def2 = !Elems[I2].isUndef();
      if (Def1 || Def2) {
        SDValue Elem1 = Elems[Def1 ? I1 : I2];
        SDValue Elem2 = Elems[Def2 ? I2 : I1];
        Result = DAG.getNode(ISD::BITCAST, DL, VT,
                             joinDwords(DAG, DL, Elem1, Elem2));
        Done[I1] = true;
        Done[I2] = true;
      } else
        Result = DAG.getUNDEF(VT);
    }
  }

  // Use VLVGx to insert the other elements.
  for (unsigned I = 0; I < NumElements; ++I)
    if (!Done[I] && !Elems[I].isUndef() && Elems[I] != ReplicatedVal)
      Result = DAG.getNode(ISD::INSERT_VECTOR_ELT, DL, VT, Result, Elems[I],
                           DAG.getConstant(I, DL, MVT::i32));
  return Result;
}

SDValue SystemZTargetLowering::lowerBUILD_VECTOR(SDValue Op,
                                                 SelectionDAG &DAG) const {
  auto *BVN = cast<BuildVectorSDNode>(Op.getNode());
  SDLoc DL(Op);
  EVT VT = Op.getValueType();

  if (BVN->isConstant()) {
    if (SystemZVectorConstantInfo(BVN).isVectorConstantLegal(Subtarget))
      return Op;

    // Fall back to loading it from memory.
    return SDValue();
  }

  // See if we should use shuffles to construct the vector from other vectors.
  if (SDValue Res = tryBuildVectorShuffle(DAG, BVN))
    return Res;

  // Detect SCALAR_TO_VECTOR conversions.
  if (isOperationLegal(ISD::SCALAR_TO_VECTOR, VT) && isScalarToVector(Op))
    return buildScalarToVector(DAG, DL, VT, Op.getOperand(0));

  // Otherwise use buildVector to build the vector up from GPRs.
  unsigned NumElements = Op.getNumOperands();
  SmallVector<SDValue, SystemZ::VectorBytes> Ops(NumElements);
  for (unsigned I = 0; I < NumElements; ++I)
    Ops[I] = Op.getOperand(I);
  return buildVector(DAG, DL, VT, Ops);
}

SDValue SystemZTargetLowering::lowerVECTOR_SHUFFLE(SDValue Op,
                                                   SelectionDAG &DAG) const {
  auto *VSN = cast<ShuffleVectorSDNode>(Op.getNode());
  SDLoc DL(Op);
  EVT VT = Op.getValueType();
  unsigned NumElements = VT.getVectorNumElements();

  if (VSN->isSplat()) {
    SDValue Op0 = Op.getOperand(0);
    unsigned Index = VSN->getSplatIndex();
    assert(Index < VT.getVectorNumElements() &&
           "Splat index should be defined and in first operand");
    // See whether the value we're splatting is directly available as a scalar.
    if ((Index == 0 && Op0.getOpcode() == ISD::SCALAR_TO_VECTOR) ||
        Op0.getOpcode() == ISD::BUILD_VECTOR)
      return DAG.getNode(SystemZISD::REPLICATE, DL, VT, Op0.getOperand(Index));
    // Otherwise keep it as a vector-to-vector operation.
    return DAG.getNode(SystemZISD::SPLAT, DL, VT, Op.getOperand(0),
                       DAG.getTargetConstant(Index, DL, MVT::i32));
  }

  GeneralShuffle GS(VT);
  for (unsigned I = 0; I < NumElements; ++I) {
    int Elt = VSN->getMaskElt(I);
    if (Elt < 0)
      GS.addUndef();
    else if (!GS.add(Op.getOperand(unsigned(Elt) / NumElements),
                     unsigned(Elt) % NumElements))
      return SDValue();
  }
  return GS.getNode(DAG, SDLoc(VSN));
}

SDValue SystemZTargetLowering::lowerSCALAR_TO_VECTOR(SDValue Op,
                                                     SelectionDAG &DAG) const {
  SDLoc DL(Op);
  // Just insert the scalar into element 0 of an undefined vector.
  return DAG.getNode(ISD::INSERT_VECTOR_ELT, DL,
                     Op.getValueType(), DAG.getUNDEF(Op.getValueType()),
                     Op.getOperand(0), DAG.getConstant(0, DL, MVT::i32));
}

SDValue SystemZTargetLowering::lowerINSERT_VECTOR_ELT(SDValue Op,
                                                      SelectionDAG &DAG) const {
  // Handle insertions of floating-point values.
  SDLoc DL(Op);
  SDValue Op0 = Op.getOperand(0);
  SDValue Op1 = Op.getOperand(1);
  SDValue Op2 = Op.getOperand(2);
  EVT VT = Op.getValueType();

  // Insertions into constant indices of a v2f64 can be done using VPDI.
  // However, if the inserted value is a bitcast or a constant then it's
  // better to use GPRs, as below.
  if (VT == MVT::v2f64 &&
      Op1.getOpcode() != ISD::BITCAST &&
      Op1.getOpcode() != ISD::ConstantFP &&
      Op2.getOpcode() == ISD::Constant) {
    uint64_t Index = cast<ConstantSDNode>(Op2)->getZExtValue();
    unsigned Mask = VT.getVectorNumElements() - 1;
    if (Index <= Mask)
      return Op;
  }

  // Otherwise bitcast to the equivalent integer form and insert via a GPR.
  MVT IntVT = MVT::getIntegerVT(VT.getScalarSizeInBits());
  MVT IntVecVT = MVT::getVectorVT(IntVT, VT.getVectorNumElements());
  SDValue Res = DAG.getNode(ISD::INSERT_VECTOR_ELT, DL, IntVecVT,
                            DAG.getNode(ISD::BITCAST, DL, IntVecVT, Op0),
                            DAG.getNode(ISD::BITCAST, DL, IntVT, Op1), Op2);
  return DAG.getNode(ISD::BITCAST, DL, VT, Res);
}

SDValue
SystemZTargetLowering::lowerEXTRACT_VECTOR_ELT(SDValue Op,
                                               SelectionDAG &DAG) const {
  // Handle extractions of floating-point values.
  SDLoc DL(Op);
  SDValue Op0 = Op.getOperand(0);
  SDValue Op1 = Op.getOperand(1);
  EVT VT = Op.getValueType();
  EVT VecVT = Op0.getValueType();

  // Extractions of constant indices can be done directly.
  if (auto *CIndexN = dyn_cast<ConstantSDNode>(Op1)) {
    uint64_t Index = CIndexN->getZExtValue();
    unsigned Mask = VecVT.getVectorNumElements() - 1;
    if (Index <= Mask)
      return Op;
  }

  // Otherwise bitcast to the equivalent integer form and extract via a GPR.
  MVT IntVT = MVT::getIntegerVT(VT.getSizeInBits());
  MVT IntVecVT = MVT::getVectorVT(IntVT, VecVT.getVectorNumElements());
  SDValue Res = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, IntVT,
                            DAG.getNode(ISD::BITCAST, DL, IntVecVT, Op0), Op1);
  return DAG.getNode(ISD::BITCAST, DL, VT, Res);
}

SDValue SystemZTargetLowering::
lowerSIGN_EXTEND_VECTOR_INREG(SDValue Op, SelectionDAG &DAG) const {
  SDValue PackedOp = Op.getOperand(0);
  EVT OutVT = Op.getValueType();
  EVT InVT = PackedOp.getValueType();
  unsigned ToBits = OutVT.getScalarSizeInBits();
  unsigned FromBits = InVT.getScalarSizeInBits();
  do {
    FromBits *= 2;
    EVT OutVT = MVT::getVectorVT(MVT::getIntegerVT(FromBits),
                                 SystemZ::VectorBits / FromBits);
    PackedOp =
      DAG.getNode(SystemZISD::UNPACK_HIGH, SDLoc(PackedOp), OutVT, PackedOp);
  } while (FromBits != ToBits);
  return PackedOp;
}

// Lower a ZERO_EXTEND_VECTOR_INREG to a vector shuffle with a zero vector.
SDValue SystemZTargetLowering::
lowerZERO_EXTEND_VECTOR_INREG(SDValue Op, SelectionDAG &DAG) const {
  SDValue PackedOp = Op.getOperand(0);
  SDLoc DL(Op);
  EVT OutVT = Op.getValueType();
  EVT InVT = PackedOp.getValueType();
  unsigned InNumElts = InVT.getVectorNumElements();
  unsigned OutNumElts = OutVT.getVectorNumElements();
  unsigned NumInPerOut = InNumElts / OutNumElts;

  SDValue ZeroVec =
    DAG.getSplatVector(InVT, DL, DAG.getConstant(0, DL, InVT.getScalarType()));

  SmallVector<int, 16> Mask(InNumElts);
  unsigned ZeroVecElt = InNumElts;
  for (unsigned PackedElt = 0; PackedElt < OutNumElts; PackedElt++) {
    unsigned MaskElt = PackedElt * NumInPerOut;
    unsigned End = MaskElt + NumInPerOut - 1;
    for (; MaskElt < End; MaskElt++)
      Mask[MaskElt] = ZeroVecElt++;
    Mask[MaskElt] = PackedElt;
  }
  SDValue Shuf = DAG.getVectorShuffle(InVT, DL, PackedOp, ZeroVec, Mask);
  return DAG.getNode(ISD::BITCAST, DL, OutVT, Shuf);
}

SDValue SystemZTargetLowering::lowerShift(SDValue Op, SelectionDAG &DAG,
                                          unsigned ByScalar) const {
  // Look for cases where a vector shift can use the *_BY_SCALAR form.
  SDValue Op0 = Op.getOperand(0);
  SDValue Op1 = Op.getOperand(1);
  SDLoc DL(Op);
  EVT VT = Op.getValueType();
  unsigned ElemBitSize = VT.getScalarSizeInBits();

  // See whether the shift vector is a splat represented as BUILD_VECTOR.
  if (auto *BVN = dyn_cast<BuildVectorSDNode>(Op1)) {
    APInt SplatBits, SplatUndef;
    unsigned SplatBitSize;
    bool HasAnyUndefs;
    // Check for constant splats.  Use ElemBitSize as the minimum element
    // width and reject splats that need wider elements.
    if (BVN->isConstantSplat(SplatBits, SplatUndef, SplatBitSize, HasAnyUndefs,
                             ElemBitSize, true) &&
        SplatBitSize == ElemBitSize) {
      SDValue Shift = DAG.getConstant(SplatBits.getZExtValue() & 0xfff,
                                      DL, MVT::i32);
      return DAG.getNode(ByScalar, DL, VT, Op0, Shift);
    }
    // Check for variable splats.
    BitVector UndefElements;
    SDValue Splat = BVN->getSplatValue(&UndefElements);
    if (Splat) {
      // Since i32 is the smallest legal type, we either need a no-op
      // or a truncation.
      SDValue Shift = DAG.getNode(ISD::TRUNCATE, DL, MVT::i32, Splat);
      return DAG.getNode(ByScalar, DL, VT, Op0, Shift);
    }
  }

  // See whether the shift vector is a splat represented as SHUFFLE_VECTOR,
  // and the shift amount is directly available in a GPR.
  if (auto *VSN = dyn_cast<ShuffleVectorSDNode>(Op1)) {
    if (VSN->isSplat()) {
      SDValue VSNOp0 = VSN->getOperand(0);
      unsigned Index = VSN->getSplatIndex();
      assert(Index < VT.getVectorNumElements() &&
             "Splat index should be defined and in first operand");
      if ((Index == 0 && VSNOp0.getOpcode() == ISD::SCALAR_TO_VECTOR) ||
          VSNOp0.getOpcode() == ISD::BUILD_VECTOR) {
        // Since i32 is the smallest legal type, we either need a no-op
        // or a truncation.
        SDValue Shift = DAG.getNode(ISD::TRUNCATE, DL, MVT::i32,
                                    VSNOp0.getOperand(Index));
        return DAG.getNode(ByScalar, DL, VT, Op0, Shift);
      }
    }
  }

  // Otherwise just treat the current form as legal.
  return Op;
}

SDValue SystemZTargetLowering::LowerOperation(SDValue Op,
                                              SelectionDAG &DAG) const {
  switch (Op.getOpcode()) {
  case ISD::FRAMEADDR:
    return lowerFRAMEADDR(Op, DAG);
  case ISD::RETURNADDR:
    return lowerRETURNADDR(Op, DAG);
  case ISD::BR_CC:
    return lowerBR_CC(Op, DAG);
  case ISD::SELECT_CC:
    return lowerSELECT_CC(Op, DAG);
  case ISD::SETCC:
    return lowerSETCC(Op, DAG);
  case ISD::STRICT_FSETCC:
    return lowerSTRICT_FSETCC(Op, DAG, false);
  case ISD::STRICT_FSETCCS:
    return lowerSTRICT_FSETCC(Op, DAG, true);
  case ISD::GlobalAddress:
    return lowerGlobalAddress(cast<GlobalAddressSDNode>(Op), DAG);
  case ISD::GlobalTLSAddress:
    return lowerGlobalTLSAddress(cast<GlobalAddressSDNode>(Op), DAG);
  case ISD::BlockAddress:
    return lowerBlockAddress(cast<BlockAddressSDNode>(Op), DAG);
  case ISD::JumpTable:
    return lowerJumpTable(cast<JumpTableSDNode>(Op), DAG);
  case ISD::ConstantPool:
    return lowerConstantPool(cast<ConstantPoolSDNode>(Op), DAG);
  case ISD::BITCAST:
    return lowerBITCAST(Op, DAG);
  case ISD::VASTART:
    return lowerVASTART(Op, DAG);
  case ISD::VACOPY:
    return lowerVACOPY(Op, DAG);
  case ISD::DYNAMIC_STACKALLOC:
    return lowerDYNAMIC_STACKALLOC(Op, DAG);
  case ISD::GET_DYNAMIC_AREA_OFFSET:
    return lowerGET_DYNAMIC_AREA_OFFSET(Op, DAG);
  case ISD::SMUL_LOHI:
    return lowerSMUL_LOHI(Op, DAG);
  case ISD::UMUL_LOHI:
    return lowerUMUL_LOHI(Op, DAG);
  case ISD::SDIVREM:
    return lowerSDIVREM(Op, DAG);
  case ISD::UDIVREM:
    return lowerUDIVREM(Op, DAG);
  case ISD::SADDO:
  case ISD::SSUBO:
  case ISD::UADDO:
  case ISD::USUBO:
    return lowerXALUO(Op, DAG);
  case ISD::ADDCARRY:
  case ISD::SUBCARRY:
    return lowerADDSUBCARRY(Op, DAG);
  case ISD::OR:
    return lowerOR(Op, DAG);
  case ISD::CTPOP:
    return lowerCTPOP(Op, DAG);
  case ISD::ATOMIC_FENCE:
    return lowerATOMIC_FENCE(Op, DAG);
  case ISD::ATOMIC_SWAP:
    return lowerATOMIC_LOAD_OP(Op, DAG, SystemZISD::ATOMIC_SWAPW);
  case ISD::ATOMIC_STORE:
    return lowerATOMIC_STORE(Op, DAG);
  case ISD::ATOMIC_LOAD:
    return lowerATOMIC_LOAD(Op, DAG);
  case ISD::ATOMIC_LOAD_ADD:
    return lowerATOMIC_LOAD_OP(Op, DAG, SystemZISD::ATOMIC_LOADW_ADD);
  case ISD::ATOMIC_LOAD_SUB:
    return lowerATOMIC_LOAD_SUB(Op, DAG);
  case ISD::ATOMIC_LOAD_AND:
    return lowerATOMIC_LOAD_OP(Op, DAG, SystemZISD::ATOMIC_LOADW_AND);
  case ISD::ATOMIC_LOAD_OR:
    return lowerATOMIC_LOAD_OP(Op, DAG, SystemZISD::ATOMIC_LOADW_OR);
  case ISD::ATOMIC_LOAD_XOR:
    return lowerATOMIC_LOAD_OP(Op, DAG, SystemZISD::ATOMIC_LOADW_XOR);
  case ISD::ATOMIC_LOAD_NAND:
    return lowerATOMIC_LOAD_OP(Op, DAG, SystemZISD::ATOMIC_LOADW_NAND);
  case ISD::ATOMIC_LOAD_MIN:
    return lowerATOMIC_LOAD_OP(Op, DAG, SystemZISD::ATOMIC_LOADW_MIN);
  case ISD::ATOMIC_LOAD_MAX:
    return lowerATOMIC_LOAD_OP(Op, DAG, SystemZISD::ATOMIC_LOADW_MAX);
  case ISD::ATOMIC_LOAD_UMIN:
    return lowerATOMIC_LOAD_OP(Op, DAG, SystemZISD::ATOMIC_LOADW_UMIN);
  case ISD::ATOMIC_LOAD_UMAX:
    return lowerATOMIC_LOAD_OP(Op, DAG, SystemZISD::ATOMIC_LOADW_UMAX);
  case ISD::ATOMIC_CMP_SWAP_WITH_SUCCESS:
    return lowerATOMIC_CMP_SWAP(Op, DAG);
  case ISD::STACKSAVE:
    return lowerSTACKSAVE(Op, DAG);
  case ISD::STACKRESTORE:
    return lowerSTACKRESTORE(Op, DAG);
  case ISD::PREFETCH:
    return lowerPREFETCH(Op, DAG);
  case ISD::INTRINSIC_W_CHAIN:
    return lowerINTRINSIC_W_CHAIN(Op, DAG);
  case ISD::INTRINSIC_WO_CHAIN:
    return lowerINTRINSIC_WO_CHAIN(Op, DAG);
  case ISD::BUILD_VECTOR:
    return lowerBUILD_VECTOR(Op, DAG);
  case ISD::VECTOR_SHUFFLE:
    return lowerVECTOR_SHUFFLE(Op, DAG);
  case ISD::SCALAR_TO_VECTOR:
    return lowerSCALAR_TO_VECTOR(Op, DAG);
  case ISD::INSERT_VECTOR_ELT:
    return lowerINSERT_VECTOR_ELT(Op, DAG);
  case ISD::EXTRACT_VECTOR_ELT:
    return lowerEXTRACT_VECTOR_ELT(Op, DAG);
  case ISD::SIGN_EXTEND_VECTOR_INREG:
    return lowerSIGN_EXTEND_VECTOR_INREG(Op, DAG);
  case ISD::ZERO_EXTEND_VECTOR_INREG:
    return lowerZERO_EXTEND_VECTOR_INREG(Op, DAG);
  case ISD::SHL:
    return lowerShift(Op, DAG, SystemZISD::VSHL_BY_SCALAR);
  case ISD::SRL:
    return lowerShift(Op, DAG, SystemZISD::VSRL_BY_SCALAR);
  case ISD::SRA:
    return lowerShift(Op, DAG, SystemZISD::VSRA_BY_SCALAR);
  default:
    llvm_unreachable("Unexpected node to lower");
  }
}

// Lower operations with invalid operand or result types (currently used
// only for 128-bit integer types).

static SDValue lowerI128ToGR128(SelectionDAG &DAG, SDValue In) {
  SDLoc DL(In);
  SDValue Lo = DAG.getNode(ISD::EXTRACT_ELEMENT, DL, MVT::i64, In,
                           DAG.getIntPtrConstant(0, DL));
  SDValue Hi = DAG.getNode(ISD::EXTRACT_ELEMENT, DL, MVT::i64, In,
                           DAG.getIntPtrConstant(1, DL));
  SDNode *Pair = DAG.getMachineNode(SystemZ::PAIR128, DL,
                                    MVT::Untyped, Hi, Lo);
  return SDValue(Pair, 0);
}

static SDValue lowerGR128ToI128(SelectionDAG &DAG, SDValue In) {
  SDLoc DL(In);
  SDValue Hi = DAG.getTargetExtractSubreg(SystemZ::subreg_h64,
                                          DL, MVT::i64, In);
  SDValue Lo = DAG.getTargetExtractSubreg(SystemZ::subreg_l64,
                                          DL, MVT::i64, In);
  return DAG.getNode(ISD::BUILD_PAIR, DL, MVT::i128, Lo, Hi);
}

void
SystemZTargetLowering::LowerOperationWrapper(SDNode *N,
                                             SmallVectorImpl<SDValue> &Results,
                                             SelectionDAG &DAG) const {
  switch (N->getOpcode()) {
  case ISD::ATOMIC_LOAD: {
    SDLoc DL(N);
    SDVTList Tys = DAG.getVTList(MVT::Untyped, MVT::Other);
    SDValue Ops[] = { N->getOperand(0), N->getOperand(1) };
    MachineMemOperand *MMO = cast<AtomicSDNode>(N)->getMemOperand();
    SDValue Res = DAG.getMemIntrinsicNode(SystemZISD::ATOMIC_LOAD_128,
                                          DL, Tys, Ops, MVT::i128, MMO);
    Results.push_back(lowerGR128ToI128(DAG, Res));
    Results.push_back(Res.getValue(1));
    break;
  }
  case ISD::ATOMIC_STORE: {
    SDLoc DL(N);
    SDVTList Tys = DAG.getVTList(MVT::Other);
    SDValue Ops[] = { N->getOperand(0),
                      lowerI128ToGR128(DAG, N->getOperand(2)),
                      N->getOperand(1) };
    MachineMemOperand *MMO = cast<AtomicSDNode>(N)->getMemOperand();
    SDValue Res = DAG.getMemIntrinsicNode(SystemZISD::ATOMIC_STORE_128,
                                          DL, Tys, Ops, MVT::i128, MMO);
    // We have to enforce sequential consistency by performing a
    // serialization operation after the store.
    if (cast<AtomicSDNode>(N)->getOrdering() ==
        AtomicOrdering::SequentiallyConsistent)
      Res = SDValue(DAG.getMachineNode(SystemZ::Serialize, DL,
                                       MVT::Other, Res), 0);
    Results.push_back(Res);
    break;
  }
  case ISD::ATOMIC_CMP_SWAP_WITH_SUCCESS: {
    SDLoc DL(N);
    SDVTList Tys = DAG.getVTList(MVT::Untyped, MVT::i32, MVT::Other);
    SDValue Ops[] = { N->getOperand(0), N->getOperand(1),
                      lowerI128ToGR128(DAG, N->getOperand(2)),
                      lowerI128ToGR128(DAG, N->getOperand(3)) };
    MachineMemOperand *MMO = cast<AtomicSDNode>(N)->getMemOperand();
    SDValue Res = DAG.getMemIntrinsicNode(SystemZISD::ATOMIC_CMP_SWAP_128,
                                          DL, Tys, Ops, MVT::i128, MMO);
    SDValue Success = emitSETCC(DAG, DL, Res.getValue(1),
                                SystemZ::CCMASK_CS, SystemZ::CCMASK_CS_EQ);
    Success = DAG.getZExtOrTrunc(Success, DL, N->getValueType(1));
    Results.push_back(lowerGR128ToI128(DAG, Res));
    Results.push_back(Success);
    Results.push_back(Res.getValue(2));
    break;
  }
  default:
    llvm_unreachable("Unexpected node to lower");
  }
}

void
SystemZTargetLowering::ReplaceNodeResults(SDNode *N,
                                          SmallVectorImpl<SDValue> &Results,
                                          SelectionDAG &DAG) const {
  return LowerOperationWrapper(N, Results, DAG);
}

const char *SystemZTargetLowering::getTargetNodeName(unsigned Opcode) const {
#define OPCODE(NAME) case SystemZISD::NAME: return "SystemZISD::" #NAME
  switch ((SystemZISD::NodeType)Opcode) {
    case SystemZISD::FIRST_NUMBER: break;
    OPCODE(RET_FLAG);
    OPCODE(CALL);
    OPCODE(SIBCALL);
    OPCODE(TLS_GDCALL);
    OPCODE(TLS_LDCALL);
    OPCODE(PCREL_WRAPPER);
    OPCODE(PCREL_OFFSET);
    OPCODE(IABS);
    OPCODE(ICMP);
    OPCODE(FCMP);
    OPCODE(STRICT_FCMP);
    OPCODE(STRICT_FCMPS);
    OPCODE(TM);
    OPCODE(BR_CCMASK);
    OPCODE(SELECT_CCMASK);
    OPCODE(ADJDYNALLOC);
    OPCODE(PROBED_ALLOCA);
    OPCODE(POPCNT);
    OPCODE(SMUL_LOHI);
    OPCODE(UMUL_LOHI);
    OPCODE(SDIVREM);
    OPCODE(UDIVREM);
    OPCODE(SADDO);
    OPCODE(SSUBO);
    OPCODE(UADDO);
    OPCODE(USUBO);
    OPCODE(ADDCARRY);
    OPCODE(SUBCARRY);
    OPCODE(GET_CCMASK);
    OPCODE(MVC);
    OPCODE(MVC_LOOP);
    OPCODE(NC);
    OPCODE(NC_LOOP);
    OPCODE(OC);
    OPCODE(OC_LOOP);
    OPCODE(XC);
    OPCODE(XC_LOOP);
    OPCODE(CLC);
    OPCODE(CLC_LOOP);
    OPCODE(STPCPY);
    OPCODE(STRCMP);
    OPCODE(SEARCH_STRING);
    OPCODE(IPM);
    OPCODE(MEMBARRIER);
    OPCODE(TBEGIN);
    OPCODE(TBEGIN_NOFLOAT);
    OPCODE(TEND);
    OPCODE(BYTE_MASK);
    OPCODE(ROTATE_MASK);
    OPCODE(REPLICATE);
    OPCODE(JOIN_DWORDS);
    OPCODE(SPLAT);
    OPCODE(MERGE_HIGH);
    OPCODE(MERGE_LOW);
    OPCODE(SHL_DOUBLE);
    OPCODE(PERMUTE_DWORDS);
    OPCODE(PERMUTE);
    OPCODE(PACK);
    OPCODE(PACKS_CC);
    OPCODE(PACKLS_CC);
    OPCODE(UNPACK_HIGH);
    OPCODE(UNPACKL_HIGH);
    OPCODE(UNPACK_LOW);
    OPCODE(UNPACKL_LOW);
    OPCODE(VSHL_BY_SCALAR);
    OPCODE(VSRL_BY_SCALAR);
    OPCODE(VSRA_BY_SCALAR);
    OPCODE(VSUM);
    OPCODE(VICMPE);
    OPCODE(VICMPH);
    OPCODE(VICMPHL);
    OPCODE(VICMPES);
    OPCODE(VICMPHS);
    OPCODE(VICMPHLS);
    OPCODE(VFCMPE);
    OPCODE(STRICT_VFCMPE);
    OPCODE(STRICT_VFCMPES);
    OPCODE(VFCMPH);
    OPCODE(STRICT_VFCMPH);
    OPCODE(STRICT_VFCMPHS);
    OPCODE(VFCMPHE);
    OPCODE(STRICT_VFCMPHE);
    OPCODE(STRICT_VFCMPHES);
    OPCODE(VFCMPES);
    OPCODE(VFCMPHS);
    OPCODE(VFCMPHES);
    OPCODE(VFTCI);
    OPCODE(VEXTEND);
    OPCODE(STRICT_VEXTEND);
    OPCODE(VROUND);
    OPCODE(STRICT_VROUND);
    OPCODE(VTM);
    OPCODE(VFAE_CC);
    OPCODE(VFAEZ_CC);
    OPCODE(VFEE_CC);
    OPCODE(VFEEZ_CC);
    OPCODE(VFENE_CC);
    OPCODE(VFENEZ_CC);
    OPCODE(VISTR_CC);
    OPCODE(VSTRC_CC);
    OPCODE(VSTRCZ_CC);
    OPCODE(VSTRS_CC);
    OPCODE(VSTRSZ_CC);
    OPCODE(TDC);
    OPCODE(ATOMIC_SWAPW);
    OPCODE(ATOMIC_LOADW_ADD);
    OPCODE(ATOMIC_LOADW_SUB);
    OPCODE(ATOMIC_LOADW_AND);
    OPCODE(ATOMIC_LOADW_OR);
    OPCODE(ATOMIC_LOADW_XOR);
    OPCODE(ATOMIC_LOADW_NAND);
    OPCODE(ATOMIC_LOADW_MIN);
    OPCODE(ATOMIC_LOADW_MAX);
    OPCODE(ATOMIC_LOADW_UMIN);
    OPCODE(ATOMIC_LOADW_UMAX);
    OPCODE(ATOMIC_CMP_SWAPW);
    OPCODE(ATOMIC_CMP_SWAP);
    OPCODE(ATOMIC_LOAD_128);
    OPCODE(ATOMIC_STORE_128);
    OPCODE(ATOMIC_CMP_SWAP_128);
    OPCODE(LRV);
    OPCODE(STRV);
    OPCODE(VLER);
    OPCODE(VSTER);
    OPCODE(PREFETCH);
  }
  return nullptr;
#undef OPCODE
}

// Return true if VT is a vector whose elements are a whole number of bytes
// in width. Also check for presence of vector support.
bool SystemZTargetLowering::canTreatAsByteVector(EVT VT) const {
  if (!Subtarget.hasVector())
    return false;

  return VT.isVector() && VT.getScalarSizeInBits() % 8 == 0 && VT.isSimple();
}

// Try to simplify an EXTRACT_VECTOR_ELT from a vector of type VecVT
// producing a result of type ResVT.  Op is a possibly bitcast version
// of the input vector and Index is the index (based on type VecVT) that
// should be extracted.  Return the new extraction if a simplification
// was possible or if Force is true.
SDValue SystemZTargetLowering::combineExtract(const SDLoc &DL, EVT ResVT,
                                              EVT VecVT, SDValue Op,
                                              unsigned Index,
                                              DAGCombinerInfo &DCI,
                                              bool Force) const {
  SelectionDAG &DAG = DCI.DAG;

  // The number of bytes being extracted.
  unsigned BytesPerElement = VecVT.getVectorElementType().getStoreSize();

  for (;;) {
    unsigned Opcode = Op.getOpcode();
    if (Opcode == ISD::BITCAST)
      // Look through bitcasts.
      Op = Op.getOperand(0);
    else if ((Opcode == ISD::VECTOR_SHUFFLE || Opcode == SystemZISD::SPLAT) &&
             canTreatAsByteVector(Op.getValueType())) {
      // Get a VPERM-like permute mask and see whether the bytes covered
      // by the extracted element are a contiguous sequence from one
      // source operand.
      SmallVector<int, SystemZ::VectorBytes> Bytes;
      if (!getVPermMask(Op, Bytes))
        break;
      int First;
      if (!getShuffleInput(Bytes, Index * BytesPerElement,
                           BytesPerElement, First))
        break;
      if (First < 0)
        return DAG.getUNDEF(ResVT);
      // Make sure the contiguous sequence starts at a multiple of the
      // original element size.
      unsigned Byte = unsigned(First) % Bytes.size();
      if (Byte % BytesPerElement != 0)
        break;
      // We can get the extracted value directly from an input.
      Index = Byte / BytesPerElement;
      Op = Op.getOperand(unsigned(First) / Bytes.size());
      Force = true;
    } else if (Opcode == ISD::BUILD_VECTOR &&
               canTreatAsByteVector(Op.getValueType())) {
      // We can only optimize this case if the BUILD_VECTOR elements are
      // at least as wide as the extracted value.
      EVT OpVT = Op.getValueType();
      unsigned OpBytesPerElement = OpVT.getVectorElementType().getStoreSize();
      if (OpBytesPerElement < BytesPerElement)
        break;
      // Make sure that the least-significant bit of the extracted value
      // is the least significant bit of an input.
      unsigned End = (Index + 1) * BytesPerElement;
      if (End % OpBytesPerElement != 0)
        break;
      // We're extracting the low part of one operand of the BUILD_VECTOR.
      Op = Op.getOperand(End / OpBytesPerElement - 1);
      if (!Op.getValueType().isInteger()) {
        EVT VT = MVT::getIntegerVT(Op.getValueSizeInBits());
        Op = DAG.getNode(ISD::BITCAST, DL, VT, Op);
        DCI.AddToWorklist(Op.getNode());
      }
      EVT VT = MVT::getIntegerVT(ResVT.getSizeInBits());
      Op = DAG.getNode(ISD::TRUNCATE, DL, VT, Op);
      if (VT != ResVT) {
        DCI.AddToWorklist(Op.getNode());
        Op = DAG.getNode(ISD::BITCAST, DL, ResVT, Op);
      }
      return Op;
    } else if ((Opcode == ISD::SIGN_EXTEND_VECTOR_INREG ||
                Opcode == ISD::ZERO_EXTEND_VECTOR_INREG ||
                Opcode == ISD::ANY_EXTEND_VECTOR_INREG) &&
               canTreatAsByteVector(Op.getValueType()) &&
               canTreatAsByteVector(Op.getOperand(0).getValueType())) {
      // Make sure that only the unextended bits are significant.
      EVT ExtVT = Op.getValueType();
      EVT OpVT = Op.getOperand(0).getValueType();
      unsigned ExtBytesPerElement = ExtVT.getVectorElementType().getStoreSize();
      unsigned OpBytesPerElement = OpVT.getVectorElementType().getStoreSize();
      unsigned Byte = Index * BytesPerElement;
      unsigned SubByte = Byte % ExtBytesPerElement;
      unsigned MinSubByte = ExtBytesPerElement - OpBytesPerElement;
      if (SubByte < MinSubByte ||
          SubByte + BytesPerElement > ExtBytesPerElement)
        break;
      // Get the byte offset of the unextended element
      Byte = Byte / ExtBytesPerElement * OpBytesPerElement;
      // ...then add the byte offset relative to that element.
      Byte += SubByte - MinSubByte;
      if (Byte % BytesPerElement != 0)
        break;
      Op = Op.getOperand(0);
      Index = Byte / BytesPerElement;
      Force = true;
    } else
      break;
  }
  if (Force) {
    if (Op.getValueType() != VecVT) {
      Op = DAG.getNode(ISD::BITCAST, DL, VecVT, Op);
      DCI.AddToWorklist(Op.getNode());
    }
    return DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, ResVT, Op,
                       DAG.getConstant(Index, DL, MVT::i32));
  }
  return SDValue();
}

// Optimize vector operations in scalar value Op on the basis that Op
// is truncated to TruncVT.
SDValue SystemZTargetLowering::combineTruncateExtract(
    const SDLoc &DL, EVT TruncVT, SDValue Op, DAGCombinerInfo &DCI) const {
  // If we have (trunc (extract_vector_elt X, Y)), try to turn it into
  // (extract_vector_elt (bitcast X), Y'), where (bitcast X) has elements
  // of type TruncVT.
  if (Op.getOpcode() == ISD::EXTRACT_VECTOR_ELT &&
      TruncVT.getSizeInBits() % 8 == 0) {
    SDValue Vec = Op.getOperand(0);
    EVT VecVT = Vec.getValueType();
    if (canTreatAsByteVector(VecVT)) {
      if (auto *IndexN = dyn_cast<ConstantSDNode>(Op.getOperand(1))) {
        unsigned BytesPerElement = VecVT.getVectorElementType().getStoreSize();
        unsigned TruncBytes = TruncVT.getStoreSize();
        if (BytesPerElement % TruncBytes == 0) {
          // Calculate the value of Y' in the above description.  We are
          // splitting the original elements into Scale equal-sized pieces
          // and for truncation purposes want the last (least-significant)
          // of these pieces for IndexN.  This is easiest to do by calculating
          // the start index of the following element and then subtracting 1.
          unsigned Scale = BytesPerElement / TruncBytes;
          unsigned NewIndex = (IndexN->getZExtValue() + 1) * Scale - 1;

          // Defer the creation of the bitcast from X to combineExtract,
          // which might be able to optimize the extraction.
          VecVT = MVT::getVectorVT(MVT::getIntegerVT(TruncBytes * 8),
                                   VecVT.getStoreSize() / TruncBytes);
          EVT ResVT = (TruncBytes < 4 ? MVT::i32 : TruncVT);
          return combineExtract(DL, ResVT, VecVT, Vec, NewIndex, DCI, true);
        }
      }
    }
  }
  return SDValue();
}

SDValue SystemZTargetLowering::combineZERO_EXTEND(
    SDNode *N, DAGCombinerInfo &DCI) const {
  // Convert (zext (select_ccmask C1, C2)) into (select_ccmask C1', C2')
  SelectionDAG &DAG = DCI.DAG;
  SDValue N0 = N->getOperand(0);
  EVT VT = N->getValueType(0);
  if (N0.getOpcode() == SystemZISD::SELECT_CCMASK) {
    auto *TrueOp = dyn_cast<ConstantSDNode>(N0.getOperand(0));
    auto *FalseOp = dyn_cast<ConstantSDNode>(N0.getOperand(1));
    if (TrueOp && FalseOp) {
      SDLoc DL(N0);
      SDValue Ops[] = { DAG.getConstant(TrueOp->getZExtValue(), DL, VT),
                        DAG.getConstant(FalseOp->getZExtValue(), DL, VT),
                        N0.getOperand(2), N0.getOperand(3), N0.getOperand(4) };
      SDValue NewSelect = DAG.getNode(SystemZISD::SELECT_CCMASK, DL, VT, Ops);
      // If N0 has multiple uses, change other uses as well.
      if (!N0.hasOneUse()) {
        SDValue TruncSelect =
          DAG.getNode(ISD::TRUNCATE, DL, N0.getValueType(), NewSelect);
        DCI.CombineTo(N0.getNode(), TruncSelect);
      }
      return NewSelect;
    }
  }
  return SDValue();
}

SDValue SystemZTargetLowering::combineSIGN_EXTEND_INREG(
    SDNode *N, DAGCombinerInfo &DCI) const {
  // Convert (sext_in_reg (setcc LHS, RHS, COND), i1)
  // and (sext_in_reg (any_extend (setcc LHS, RHS, COND)), i1)
  // into (select_cc LHS, RHS, -1, 0, COND)
  SelectionDAG &DAG = DCI.DAG;
  SDValue N0 = N->getOperand(0);
  EVT VT = N->getValueType(0);
  EVT EVT = cast<VTSDNode>(N->getOperand(1))->getVT();
  if (N0.hasOneUse() && N0.getOpcode() == ISD::ANY_EXTEND)
    N0 = N0.getOperand(0);
  if (EVT == MVT::i1 && N0.hasOneUse() && N0.getOpcode() == ISD::SETCC) {
    SDLoc DL(N0);
    SDValue Ops[] = { N0.getOperand(0), N0.getOperand(1),
                      DAG.getConstant(-1, DL, VT), DAG.getConstant(0, DL, VT),
                      N0.getOperand(2) };
    return DAG.getNode(ISD::SELECT_CC, DL, VT, Ops);
  }
  return SDValue();
}

SDValue SystemZTargetLowering::combineSIGN_EXTEND(
    SDNode *N, DAGCombinerInfo &DCI) const {
  // Convert (sext (ashr (shl X, C1), C2)) to
  // (ashr (shl (anyext X), C1'), C2')), since wider shifts are as
  // cheap as narrower ones.
  SelectionDAG &DAG = DCI.DAG;
  SDValue N0 = N->getOperand(0);
  EVT VT = N->getValueType(0);
  if (N0.hasOneUse() && N0.getOpcode() == ISD::SRA) {
    auto *SraAmt = dyn_cast<ConstantSDNode>(N0.getOperand(1));
    SDValue Inner = N0.getOperand(0);
    if (SraAmt && Inner.hasOneUse() && Inner.getOpcode() == ISD::SHL) {
      if (auto *ShlAmt = dyn_cast<ConstantSDNode>(Inner.getOperand(1))) {
        unsigned Extra = (VT.getSizeInBits() - N0.getValueSizeInBits());
        unsigned NewShlAmt = ShlAmt->getZExtValue() + Extra;
        unsigned NewSraAmt = SraAmt->getZExtValue() + Extra;
        EVT ShiftVT = N0.getOperand(1).getValueType();
        SDValue Ext = DAG.getNode(ISD::ANY_EXTEND, SDLoc(Inner), VT,
                                  Inner.getOperand(0));
        SDValue Shl = DAG.getNode(ISD::SHL, SDLoc(Inner), VT, Ext,
                                  DAG.getConstant(NewShlAmt, SDLoc(Inner),
                                                  ShiftVT));
        return DAG.getNode(ISD::SRA, SDLoc(N0), VT, Shl,
                           DAG.getConstant(NewSraAmt, SDLoc(N0), ShiftVT));
      }
    }
  }
  return SDValue();
}

SDValue SystemZTargetLowering::combineMERGE(
    SDNode *N, DAGCombinerInfo &DCI) const {
  SelectionDAG &DAG = DCI.DAG;
  unsigned Opcode = N->getOpcode();
  SDValue Op0 = N->getOperand(0);
  SDValue Op1 = N->getOperand(1);
  if (Op0.getOpcode() == ISD::BITCAST)
    Op0 = Op0.getOperand(0);
  if (ISD::isBuildVectorAllZeros(Op0.getNode())) {
    // (z_merge_* 0, 0) -> 0.  This is mostly useful for using VLLEZF
    // for v4f32.
    if (Op1 == N->getOperand(0))
      return Op1;
    // (z_merge_? 0, X) -> (z_unpackl_? 0, X).
    EVT VT = Op1.getValueType();
    unsigned ElemBytes = VT.getVectorElementType().getStoreSize();
    if (ElemBytes <= 4) {
      Opcode = (Opcode == SystemZISD::MERGE_HIGH ?
                SystemZISD::UNPACKL_HIGH : SystemZISD::UNPACKL_LOW);
      EVT InVT = VT.changeVectorElementTypeToInteger();
      EVT OutVT = MVT::getVectorVT(MVT::getIntegerVT(ElemBytes * 16),
                                   SystemZ::VectorBytes / ElemBytes / 2);
      if (VT != InVT) {
        Op1 = DAG.getNode(ISD::BITCAST, SDLoc(N), InVT, Op1);
        DCI.AddToWorklist(Op1.getNode());
      }
      SDValue Op = DAG.getNode(Opcode, SDLoc(N), OutVT, Op1);
      DCI.AddToWorklist(Op.getNode());
      return DAG.getNode(ISD::BITCAST, SDLoc(N), VT, Op);
    }
  }
  return SDValue();
}

SDValue SystemZTargetLowering::combineLOAD(
    SDNode *N, DAGCombinerInfo &DCI) const {
  SelectionDAG &DAG = DCI.DAG;
  EVT LdVT = N->getValueType(0);
  if (LdVT.isVector() || LdVT.isInteger())
    return SDValue();
  // Transform a scalar load that is REPLICATEd as well as having other
  // use(s) to the form where the other use(s) use the first element of the
  // REPLICATE instead of the load. Otherwise instruction selection will not
  // produce a VLREP. Avoid extracting to a GPR, so only do this for floating
  // point loads.

  SDValue Replicate;
  SmallVector<SDNode*, 8> OtherUses;
  for (SDNode::use_iterator UI = N->use_begin(), UE = N->use_end();
       UI != UE; ++UI) {
    if (UI->getOpcode() == SystemZISD::REPLICATE) {
      if (Replicate)
        return SDValue(); // Should never happen
      Replicate = SDValue(*UI, 0);
    }
    else if (UI.getUse().getResNo() == 0)
      OtherUses.push_back(*UI);
  }
  if (!Replicate || OtherUses.empty())
    return SDValue();

  SDLoc DL(N);
  SDValue Extract0 = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, LdVT,
                              Replicate, DAG.getConstant(0, DL, MVT::i32));
  // Update uses of the loaded Value while preserving old chains.
  for (SDNode *U : OtherUses) {
    SmallVector<SDValue, 8> Ops;
    for (SDValue Op : U->ops())
      Ops.push_back((Op.getNode() == N && Op.getResNo() == 0) ? Extract0 : Op);
    DAG.UpdateNodeOperands(U, Ops);
  }
  return SDValue(N, 0);
}

bool SystemZTargetLowering::canLoadStoreByteSwapped(EVT VT) const {
  if (VT == MVT::i16 || VT == MVT::i32 || VT == MVT::i64)
    return true;
  if (Subtarget.hasVectorEnhancements2())
    if (VT == MVT::v8i16 || VT == MVT::v4i32 || VT == MVT::v2i64)
      return true;
  return false;
}

static bool isVectorElementSwap(ArrayRef<int> M, EVT VT) {
  if (!VT.isVector() || !VT.isSimple() ||
      VT.getSizeInBits() != 128 ||
      VT.getScalarSizeInBits() % 8 != 0)
    return false;

  unsigned NumElts = VT.getVectorNumElements();
  for (unsigned i = 0; i < NumElts; ++i) {
    if (M[i] < 0) continue; // ignore UNDEF indices
    if ((unsigned) M[i] != NumElts - 1 - i)
      return false;
  }

  return true;
}

SDValue SystemZTargetLowering::combineSTORE(
    SDNode *N, DAGCombinerInfo &DCI) const {
  SelectionDAG &DAG = DCI.DAG;
  auto *SN = cast<StoreSDNode>(N);
  auto &Op1 = N->getOperand(1);
  EVT MemVT = SN->getMemoryVT();
  // If we have (truncstoreiN (extract_vector_elt X, Y), Z) then it is better
  // for the extraction to be done on a vMiN value, so that we can use VSTE.
  // If X has wider elements then convert it to:
  // (truncstoreiN (extract_vector_elt (bitcast X), Y2), Z).
  if (MemVT.isInteger() && SN->isTruncatingStore()) {
    if (SDValue Value =
            combineTruncateExtract(SDLoc(N), MemVT, SN->getValue(), DCI)) {
      DCI.AddToWorklist(Value.getNode());

      // Rewrite the store with the new form of stored value.
      return DAG.getTruncStore(SN->getChain(), SDLoc(SN), Value,
                               SN->getBasePtr(), SN->getMemoryVT(),
                               SN->getMemOperand());
    }
  }
  // Combine STORE (BSWAP) into STRVH/STRV/STRVG/VSTBR
  if (!SN->isTruncatingStore() &&
      Op1.getOpcode() == ISD::BSWAP &&
      Op1.getNode()->hasOneUse() &&
      canLoadStoreByteSwapped(Op1.getValueType())) {

      SDValue BSwapOp = Op1.getOperand(0);

      if (BSwapOp.getValueType() == MVT::i16)
        BSwapOp = DAG.getNode(ISD::ANY_EXTEND, SDLoc(N), MVT::i32, BSwapOp);

      SDValue Ops[] = {
        N->getOperand(0), BSwapOp, N->getOperand(2)
      };

      return
        DAG.getMemIntrinsicNode(SystemZISD::STRV, SDLoc(N), DAG.getVTList(MVT::Other),
                                Ops, MemVT, SN->getMemOperand());
    }
  // Combine STORE (element-swap) into VSTER
  if (!SN->isTruncatingStore() &&
      Op1.getOpcode() == ISD::VECTOR_SHUFFLE &&
      Op1.getNode()->hasOneUse() &&
      Subtarget.hasVectorEnhancements2()) {
    ShuffleVectorSDNode *SVN = cast<ShuffleVectorSDNode>(Op1.getNode());
    ArrayRef<int> ShuffleMask = SVN->getMask();
    if (isVectorElementSwap(ShuffleMask, Op1.getValueType())) {
      SDValue Ops[] = {
        N->getOperand(0), Op1.getOperand(0), N->getOperand(2)
      };

      return DAG.getMemIntrinsicNode(SystemZISD::VSTER, SDLoc(N),
                                     DAG.getVTList(MVT::Other),
                                     Ops, MemVT, SN->getMemOperand());
    }
  }

  return SDValue();
}

SDValue SystemZTargetLowering::combineVECTOR_SHUFFLE(
    SDNode *N, DAGCombinerInfo &DCI) const {
  SelectionDAG &DAG = DCI.DAG;
  // Combine element-swap (LOAD) into VLER
  if (ISD::isNON_EXTLoad(N->getOperand(0).getNode()) &&
      N->getOperand(0).hasOneUse() &&
      Subtarget.hasVectorEnhancements2()) {
    ShuffleVectorSDNode *SVN = cast<ShuffleVectorSDNode>(N);
    ArrayRef<int> ShuffleMask = SVN->getMask();
    if (isVectorElementSwap(ShuffleMask, N->getValueType(0))) {
      SDValue Load = N->getOperand(0);
      LoadSDNode *LD = cast<LoadSDNode>(Load);

      // Create the element-swapping load.
      SDValue Ops[] = {
        LD->getChain(),    // Chain
        LD->getBasePtr()   // Ptr
      };
      SDValue ESLoad =
        DAG.getMemIntrinsicNode(SystemZISD::VLER, SDLoc(N),
                                DAG.getVTList(LD->getValueType(0), MVT::Other),
                                Ops, LD->getMemoryVT(), LD->getMemOperand());

      // First, combine the VECTOR_SHUFFLE away.  This makes the value produced
      // by the load dead.
      DCI.CombineTo(N, ESLoad);

      // Next, combine the load away, we give it a bogus result value but a real
      // chain result.  The result value is dead because the shuffle is dead.
      DCI.CombineTo(Load.getNode(), ESLoad, ESLoad.getValue(1));

      // Return N so it doesn't get rechecked!
      return SDValue(N, 0);
    }
  }

  return SDValue();
}

SDValue SystemZTargetLowering::combineEXTRACT_VECTOR_ELT(
    SDNode *N, DAGCombinerInfo &DCI) const {
  SelectionDAG &DAG = DCI.DAG;

  if (!Subtarget.hasVector())
    return SDValue();

  // Look through bitcasts that retain the number of vector elements.
  SDValue Op = N->getOperand(0);
  if (Op.getOpcode() == ISD::BITCAST &&
      Op.getValueType().isVector() &&
      Op.getOperand(0).getValueType().isVector() &&
      Op.getValueType().getVectorNumElements() ==
      Op.getOperand(0).getValueType().getVectorNumElements())
    Op = Op.getOperand(0);

  // Pull BSWAP out of a vector extraction.
  if (Op.getOpcode() == ISD::BSWAP && Op.hasOneUse()) {
    EVT VecVT = Op.getValueType();
    EVT EltVT = VecVT.getVectorElementType();
    Op = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, SDLoc(N), EltVT,
                     Op.getOperand(0), N->getOperand(1));
    DCI.AddToWorklist(Op.getNode());
    Op = DAG.getNode(ISD::BSWAP, SDLoc(N), EltVT, Op);
    if (EltVT != N->getValueType(0)) {
      DCI.AddToWorklist(Op.getNode());
      Op = DAG.getNode(ISD::BITCAST, SDLoc(N), N->getValueType(0), Op);
    }
    return Op;
  }

  // Try to simplify a vector extraction.
  if (auto *IndexN = dyn_cast<ConstantSDNode>(N->getOperand(1))) {
    SDValue Op0 = N->getOperand(0);
    EVT VecVT = Op0.getValueType();
    return combineExtract(SDLoc(N), N->getValueType(0), VecVT, Op0,
                          IndexN->getZExtValue(), DCI, false);
  }
  return SDValue();
}

SDValue SystemZTargetLowering::combineJOIN_DWORDS(
    SDNode *N, DAGCombinerInfo &DCI) const {
  SelectionDAG &DAG = DCI.DAG;
  // (join_dwords X, X) == (replicate X)
  if (N->getOperand(0) == N->getOperand(1))
    return DAG.getNode(SystemZISD::REPLICATE, SDLoc(N), N->getValueType(0),
                       N->getOperand(0));
  return SDValue();
}

static SDValue MergeInputChains(SDNode *N1, SDNode *N2) {
  SDValue Chain1 = N1->getOperand(0);
  SDValue Chain2 = N2->getOperand(0);

  // Trivial case: both nodes take the same chain.
  if (Chain1 == Chain2)
    return Chain1;

  // FIXME - we could handle more complex cases via TokenFactor,
  // assuming we can verify that this would not create a cycle.
  return SDValue();
}

SDValue SystemZTargetLowering::combineFP_ROUND(
    SDNode *N, DAGCombinerInfo &DCI) const {

  if (!Subtarget.hasVector())
    return SDValue();

  // (fpround (extract_vector_elt X 0))
  // (fpround (extract_vector_elt X 1)) ->
  // (extract_vector_elt (VROUND X) 0)
  // (extract_vector_elt (VROUND X) 2)
  //
  // This is a special case since the target doesn't really support v2f32s.
  unsigned OpNo = N->isStrictFPOpcode() ? 1 : 0;
  SelectionDAG &DAG = DCI.DAG;
  SDValue Op0 = N->getOperand(OpNo);
  if (N->getValueType(0) == MVT::f32 &&
      Op0.hasOneUse() &&
      Op0.getOpcode() == ISD::EXTRACT_VECTOR_ELT &&
      Op0.getOperand(0).getValueType() == MVT::v2f64 &&
      Op0.getOperand(1).getOpcode() == ISD::Constant &&
      cast<ConstantSDNode>(Op0.getOperand(1))->getZExtValue() == 0) {
    SDValue Vec = Op0.getOperand(0);
    for (auto *U : Vec->uses()) {
      if (U != Op0.getNode() &&
          U->hasOneUse() &&
          U->getOpcode() == ISD::EXTRACT_VECTOR_ELT &&
          U->getOperand(0) == Vec &&
          U->getOperand(1).getOpcode() == ISD::Constant &&
          cast<ConstantSDNode>(U->getOperand(1))->getZExtValue() == 1) {
        SDValue OtherRound = SDValue(*U->use_begin(), 0);
        if (OtherRound.getOpcode() == N->getOpcode() &&
            OtherRound.getOperand(OpNo) == SDValue(U, 0) &&
            OtherRound.getValueType() == MVT::f32) {
          SDValue VRound, Chain;
          if (N->isStrictFPOpcode()) {
            Chain = MergeInputChains(N, OtherRound.getNode());
            if (!Chain)
              continue;
            VRound = DAG.getNode(SystemZISD::STRICT_VROUND, SDLoc(N),
                                 {MVT::v4f32, MVT::Other}, {Chain, Vec});
            Chain = VRound.getValue(1);
          } else
            VRound = DAG.getNode(SystemZISD::VROUND, SDLoc(N),
                                 MVT::v4f32, Vec);
          DCI.AddToWorklist(VRound.getNode());
          SDValue Extract1 =
            DAG.getNode(ISD::EXTRACT_VECTOR_ELT, SDLoc(U), MVT::f32,
                        VRound, DAG.getConstant(2, SDLoc(U), MVT::i32));
          DCI.AddToWorklist(Extract1.getNode());
          DAG.ReplaceAllUsesOfValueWith(OtherRound, Extract1);
          if (Chain)
            DAG.ReplaceAllUsesOfValueWith(OtherRound.getValue(1), Chain);
          SDValue Extract0 =
            DAG.getNode(ISD::EXTRACT_VECTOR_ELT, SDLoc(Op0), MVT::f32,
                        VRound, DAG.getConstant(0, SDLoc(Op0), MVT::i32));
          if (Chain)
            return DAG.getNode(ISD::MERGE_VALUES, SDLoc(Op0),
                               N->getVTList(), Extract0, Chain);
          return Extract0;
        }
      }
    }
  }
  return SDValue();
}

SDValue SystemZTargetLowering::combineFP_EXTEND(
    SDNode *N, DAGCombinerInfo &DCI) const {

  if (!Subtarget.hasVector())
    return SDValue();

  // (fpextend (extract_vector_elt X 0))
  // (fpextend (extract_vector_elt X 2)) ->
  // (extract_vector_elt (VEXTEND X) 0)
  // (extract_vector_elt (VEXTEND X) 1)
  //
  // This is a special case since the target doesn't really support v2f32s.
  unsigned OpNo = N->isStrictFPOpcode() ? 1 : 0;
  SelectionDAG &DAG = DCI.DAG;
  SDValue Op0 = N->getOperand(OpNo);
  if (N->getValueType(0) == MVT::f64 &&
      Op0.hasOneUse() &&
      Op0.getOpcode() == ISD::EXTRACT_VECTOR_ELT &&
      Op0.getOperand(0).getValueType() == MVT::v4f32 &&
      Op0.getOperand(1).getOpcode() == ISD::Constant &&
      cast<ConstantSDNode>(Op0.getOperand(1))->getZExtValue() == 0) {
    SDValue Vec = Op0.getOperand(0);
    for (auto *U : Vec->uses()) {
      if (U != Op0.getNode() &&
          U->hasOneUse() &&
          U->getOpcode() == ISD::EXTRACT_VECTOR_ELT &&
          U->getOperand(0) == Vec &&
          U->getOperand(1).getOpcode() == ISD::Constant &&
          cast<ConstantSDNode>(U->getOperand(1))->getZExtValue() == 2) {
        SDValue OtherExtend = SDValue(*U->use_begin(), 0);
        if (OtherExtend.getOpcode() == N->getOpcode() &&
            OtherExtend.getOperand(OpNo) == SDValue(U, 0) &&
            OtherExtend.getValueType() == MVT::f64) {
          SDValue VExtend, Chain;
          if (N->isStrictFPOpcode()) {
            Chain = MergeInputChains(N, OtherExtend.getNode());
            if (!Chain)
              continue;
            VExtend = DAG.getNode(SystemZISD::STRICT_VEXTEND, SDLoc(N),
                                  {MVT::v2f64, MVT::Other}, {Chain, Vec});
            Chain = VExtend.getValue(1);
          } else
            VExtend = DAG.getNode(SystemZISD::VEXTEND, SDLoc(N),
                                  MVT::v2f64, Vec);
          DCI.AddToWorklist(VExtend.getNode());
          SDValue Extract1 =
            DAG.getNode(ISD::EXTRACT_VECTOR_ELT, SDLoc(U), MVT::f64,
                        VExtend, DAG.getConstant(1, SDLoc(U), MVT::i32));
          DCI.AddToWorklist(Extract1.getNode());
          DAG.ReplaceAllUsesOfValueWith(OtherExtend, Extract1);
          if (Chain)
            DAG.ReplaceAllUsesOfValueWith(OtherExtend.getValue(1), Chain);
          SDValue Extract0 =
            DAG.getNode(ISD::EXTRACT_VECTOR_ELT, SDLoc(Op0), MVT::f64,
                        VExtend, DAG.getConstant(0, SDLoc(Op0), MVT::i32));
          if (Chain)
            return DAG.getNode(ISD::MERGE_VALUES, SDLoc(Op0),
                               N->getVTList(), Extract0, Chain);
          return Extract0;
        }
      }
    }
  }
  return SDValue();
}

SDValue SystemZTargetLowering::combineINT_TO_FP(
    SDNode *N, DAGCombinerInfo &DCI) const {
  if (DCI.Level != BeforeLegalizeTypes)
    return SDValue();
  unsigned Opcode = N->getOpcode();
  EVT OutVT = N->getValueType(0);
  SelectionDAG &DAG = DCI.DAG;
  SDValue Op = N->getOperand(0);
  unsigned OutScalarBits = OutVT.getScalarSizeInBits();
  unsigned InScalarBits = Op->getValueType(0).getScalarSizeInBits();

  // Insert an extension before type-legalization to avoid scalarization, e.g.:
  // v2f64 = uint_to_fp v2i16
  // =>
  // v2f64 = uint_to_fp (v2i64 zero_extend v2i16)
  if (OutVT.isVector() && OutScalarBits > InScalarBits) {
    MVT ExtVT = MVT::getVectorVT(MVT::getIntegerVT(OutVT.getScalarSizeInBits()),
                                 OutVT.getVectorNumElements());
    unsigned ExtOpcode =
      (Opcode == ISD::UINT_TO_FP ? ISD::ZERO_EXTEND : ISD::SIGN_EXTEND);
    SDValue ExtOp = DAG.getNode(ExtOpcode, SDLoc(N), ExtVT, Op);
    return DAG.getNode(Opcode, SDLoc(N), OutVT, ExtOp);
  }
  return SDValue();
}

SDValue SystemZTargetLowering::combineBSWAP(
    SDNode *N, DAGCombinerInfo &DCI) const {
  SelectionDAG &DAG = DCI.DAG;
  // Combine BSWAP (LOAD) into LRVH/LRV/LRVG/VLBR
  if (ISD::isNON_EXTLoad(N->getOperand(0).getNode()) &&
      N->getOperand(0).hasOneUse() &&
      canLoadStoreByteSwapped(N->getValueType(0))) {
      SDValue Load = N->getOperand(0);
      LoadSDNode *LD = cast<LoadSDNode>(Load);

      // Create the byte-swapping load.
      SDValue Ops[] = {
        LD->getChain(),    // Chain
        LD->getBasePtr()   // Ptr
      };
      EVT LoadVT = N->getValueType(0);
      if (LoadVT == MVT::i16)
        LoadVT = MVT::i32;
      SDValue BSLoad =
        DAG.getMemIntrinsicNode(SystemZISD::LRV, SDLoc(N),
                                DAG.getVTList(LoadVT, MVT::Other),
                                Ops, LD->getMemoryVT(), LD->getMemOperand());

      // If this is an i16 load, insert the truncate.
      SDValue ResVal = BSLoad;
      if (N->getValueType(0) == MVT::i16)
        ResVal = DAG.getNode(ISD::TRUNCATE, SDLoc(N), MVT::i16, BSLoad);

      // First, combine the bswap away.  This makes the value produced by the
      // load dead.
      DCI.CombineTo(N, ResVal);

      // Next, combine the load away, we give it a bogus result value but a real
      // chain result.  The result value is dead because the bswap is dead.
      DCI.CombineTo(Load.getNode(), ResVal, BSLoad.getValue(1));

      // Return N so it doesn't get rechecked!
      return SDValue(N, 0);
    }

  // Look through bitcasts that retain the number of vector elements.
  SDValue Op = N->getOperand(0);
  if (Op.getOpcode() == ISD::BITCAST &&
      Op.getValueType().isVector() &&
      Op.getOperand(0).getValueType().isVector() &&
      Op.getValueType().getVectorNumElements() ==
      Op.getOperand(0).getValueType().getVectorNumElements())
    Op = Op.getOperand(0);

  // Push BSWAP into a vector insertion if at least one side then simplifies.
  if (Op.getOpcode() == ISD::INSERT_VECTOR_ELT && Op.hasOneUse()) {
    SDValue Vec = Op.getOperand(0);
    SDValue Elt = Op.getOperand(1);
    SDValue Idx = Op.getOperand(2);

    if (DAG.isConstantIntBuildVectorOrConstantInt(Vec) ||
        Vec.getOpcode() == ISD::BSWAP || Vec.isUndef() ||
        DAG.isConstantIntBuildVectorOrConstantInt(Elt) ||
        Elt.getOpcode() == ISD::BSWAP || Elt.isUndef() ||
        (canLoadStoreByteSwapped(N->getValueType(0)) &&
         ISD::isNON_EXTLoad(Elt.getNode()) && Elt.hasOneUse())) {
      EVT VecVT = N->getValueType(0);
      EVT EltVT = N->getValueType(0).getVectorElementType();
      if (VecVT != Vec.getValueType()) {
        Vec = DAG.getNode(ISD::BITCAST, SDLoc(N), VecVT, Vec);
        DCI.AddToWorklist(Vec.getNode());
      }
      if (EltVT != Elt.getValueType()) {
        Elt = DAG.getNode(ISD::BITCAST, SDLoc(N), EltVT, Elt);
        DCI.AddToWorklist(Elt.getNode());
      }
      Vec = DAG.getNode(ISD::BSWAP, SDLoc(N), VecVT, Vec);
      DCI.AddToWorklist(Vec.getNode());
      Elt = DAG.getNode(ISD::BSWAP, SDLoc(N), EltVT, Elt);
      DCI.AddToWorklist(Elt.getNode());
      return DAG.getNode(ISD::INSERT_VECTOR_ELT, SDLoc(N), VecVT,
                         Vec, Elt, Idx);
    }
  }

  // Push BSWAP into a vector shuffle if at least one side then simplifies.
  ShuffleVectorSDNode *SV = dyn_cast<ShuffleVectorSDNode>(Op);
  if (SV && Op.hasOneUse()) {
    SDValue Op0 = Op.getOperand(0);
    SDValue Op1 = Op.getOperand(1);

    if (DAG.isConstantIntBuildVectorOrConstantInt(Op0) ||
        Op0.getOpcode() == ISD::BSWAP || Op0.isUndef() ||
        DAG.isConstantIntBuildVectorOrConstantInt(Op1) ||
        Op1.getOpcode() == ISD::BSWAP || Op1.isUndef()) {
      EVT VecVT = N->getValueType(0);
      if (VecVT != Op0.getValueType()) {
        Op0 = DAG.getNode(ISD::BITCAST, SDLoc(N), VecVT, Op0);
        DCI.AddToWorklist(Op0.getNode());
      }
      if (VecVT != Op1.getValueType()) {
        Op1 = DAG.getNode(ISD::BITCAST, SDLoc(N), VecVT, Op1);
        DCI.AddToWorklist(Op1.getNode());
      }
      Op0 = DAG.getNode(ISD::BSWAP, SDLoc(N), VecVT, Op0);
      DCI.AddToWorklist(Op0.getNode());
      Op1 = DAG.getNode(ISD::BSWAP, SDLoc(N), VecVT, Op1);
      DCI.AddToWorklist(Op1.getNode());
      return DAG.getVectorShuffle(VecVT, SDLoc(N), Op0, Op1, SV->getMask());
    }
  }

  return SDValue();
}

static bool combineCCMask(SDValue &CCReg, int &CCValid, int &CCMask) {
  // We have a SELECT_CCMASK or BR_CCMASK comparing the condition code
  // set by the CCReg instruction using the CCValid / CCMask masks,
  // If the CCReg instruction is itself a ICMP testing the condition
  // code set by some other instruction, see whether we can directly
  // use that condition code.

  // Verify that we have an ICMP against some constant.
  if (CCValid != SystemZ::CCMASK_ICMP)
    return false;
  auto *ICmp = CCReg.getNode();
  if (ICmp->getOpcode() != SystemZISD::ICMP)
    return false;
  auto *CompareLHS = ICmp->getOperand(0).getNode();
  auto *CompareRHS = dyn_cast<ConstantSDNode>(ICmp->getOperand(1));
  if (!CompareRHS)
    return false;

  // Optimize the case where CompareLHS is a SELECT_CCMASK.
  if (CompareLHS->getOpcode() == SystemZISD::SELECT_CCMASK) {
    // Verify that we have an appropriate mask for a EQ or NE comparison.
    bool Invert = false;
    if (CCMask == SystemZ::CCMASK_CMP_NE)
      Invert = !Invert;
    else if (CCMask != SystemZ::CCMASK_CMP_EQ)
      return false;

    // Verify that the ICMP compares against one of select values.
    auto *TrueVal = dyn_cast<ConstantSDNode>(CompareLHS->getOperand(0));
    if (!TrueVal)
      return false;
    auto *FalseVal = dyn_cast<ConstantSDNode>(CompareLHS->getOperand(1));
    if (!FalseVal)
      return false;
    if (CompareRHS->getZExtValue() == FalseVal->getZExtValue())
      Invert = !Invert;
    else if (CompareRHS->getZExtValue() != TrueVal->getZExtValue())
      return false;

    // Compute the effective CC mask for the new branch or select.
    auto *NewCCValid = dyn_cast<ConstantSDNode>(CompareLHS->getOperand(2));
    auto *NewCCMask = dyn_cast<ConstantSDNode>(CompareLHS->getOperand(3));
    if (!NewCCValid || !NewCCMask)
      return false;
    CCValid = NewCCValid->getZExtValue();
    CCMask = NewCCMask->getZExtValue();
    if (Invert)
      CCMask ^= CCValid;

    // Return the updated CCReg link.
    CCReg = CompareLHS->getOperand(4);
    return true;
  }

  // Optimize the case where CompareRHS is (SRA (SHL (IPM))).
  if (CompareLHS->getOpcode() == ISD::SRA) {
    auto *SRACount = dyn_cast<ConstantSDNode>(CompareLHS->getOperand(1));
    if (!SRACount || SRACount->getZExtValue() != 30)
      return false;
    auto *SHL = CompareLHS->getOperand(0).getNode();
    if (SHL->getOpcode() != ISD::SHL)
      return false;
    auto *SHLCount = dyn_cast<ConstantSDNode>(SHL->getOperand(1));
    if (!SHLCount || SHLCount->getZExtValue() != 30 - SystemZ::IPM_CC)
      return false;
    auto *IPM = SHL->getOperand(0).getNode();
    if (IPM->getOpcode() != SystemZISD::IPM)
      return false;

    // Avoid introducing CC spills (because SRA would clobber CC).
    if (!CompareLHS->hasOneUse())
      return false;
    // Verify that the ICMP compares against zero.
    if (CompareRHS->getZExtValue() != 0)
      return false;

    // Compute the effective CC mask for the new branch or select.
    CCMask = SystemZ::reverseCCMask(CCMask);

    // Return the updated CCReg link.
    CCReg = IPM->getOperand(0);
    return true;
  }

  return false;
}

SDValue SystemZTargetLowering::combineBR_CCMASK(
    SDNode *N, DAGCombinerInfo &DCI) const {
  SelectionDAG &DAG = DCI.DAG;

  // Combine BR_CCMASK (ICMP (SELECT_CCMASK)) into a single BR_CCMASK.
  auto *CCValid = dyn_cast<ConstantSDNode>(N->getOperand(1));
  auto *CCMask = dyn_cast<ConstantSDNode>(N->getOperand(2));
  if (!CCValid || !CCMask)
    return SDValue();

  int CCValidVal = CCValid->getZExtValue();
  int CCMaskVal = CCMask->getZExtValue();
  SDValue Chain = N->getOperand(0);
  SDValue CCReg = N->getOperand(4);

  if (combineCCMask(CCReg, CCValidVal, CCMaskVal))
    return DAG.getNode(SystemZISD::BR_CCMASK, SDLoc(N), N->getValueType(0),
                       Chain,
                       DAG.getTargetConstant(CCValidVal, SDLoc(N), MVT::i32),
                       DAG.getTargetConstant(CCMaskVal, SDLoc(N), MVT::i32),
                       N->getOperand(3), CCReg);
  return SDValue();
}

SDValue SystemZTargetLowering::combineSELECT_CCMASK(
    SDNode *N, DAGCombinerInfo &DCI) const {
  SelectionDAG &DAG = DCI.DAG;

  // Combine SELECT_CCMASK (ICMP (SELECT_CCMASK)) into a single SELECT_CCMASK.
  auto *CCValid = dyn_cast<ConstantSDNode>(N->getOperand(2));
  auto *CCMask = dyn_cast<ConstantSDNode>(N->getOperand(3));
  if (!CCValid || !CCMask)
    return SDValue();

  int CCValidVal = CCValid->getZExtValue();
  int CCMaskVal = CCMask->getZExtValue();
  SDValue CCReg = N->getOperand(4);

  if (combineCCMask(CCReg, CCValidVal, CCMaskVal))
    return DAG.getNode(SystemZISD::SELECT_CCMASK, SDLoc(N), N->getValueType(0),
                       N->getOperand(0), N->getOperand(1),
                       DAG.getTargetConstant(CCValidVal, SDLoc(N), MVT::i32),
                       DAG.getTargetConstant(CCMaskVal, SDLoc(N), MVT::i32),
                       CCReg);
  return SDValue();
}


SDValue SystemZTargetLowering::combineGET_CCMASK(
    SDNode *N, DAGCombinerInfo &DCI) const {

  // Optimize away GET_CCMASK (SELECT_CCMASK) if the CC masks are compatible
  auto *CCValid = dyn_cast<ConstantSDNode>(N->getOperand(1));
  auto *CCMask = dyn_cast<ConstantSDNode>(N->getOperand(2));
  if (!CCValid || !CCMask)
    return SDValue();
  int CCValidVal = CCValid->getZExtValue();
  int CCMaskVal = CCMask->getZExtValue();

  SDValue Select = N->getOperand(0);
  if (Select->getOpcode() != SystemZISD::SELECT_CCMASK)
    return SDValue();

  auto *SelectCCValid = dyn_cast<ConstantSDNode>(Select->getOperand(2));
  auto *SelectCCMask = dyn_cast<ConstantSDNode>(Select->getOperand(3));
  if (!SelectCCValid || !SelectCCMask)
    return SDValue();
  int SelectCCValidVal = SelectCCValid->getZExtValue();
  int SelectCCMaskVal = SelectCCMask->getZExtValue();

  auto *TrueVal = dyn_cast<ConstantSDNode>(Select->getOperand(0));
  auto *FalseVal = dyn_cast<ConstantSDNode>(Select->getOperand(1));
  if (!TrueVal || !FalseVal)
    return SDValue();
  if (TrueVal->getZExtValue() != 0 && FalseVal->getZExtValue() == 0)
    ;
  else if (TrueVal->getZExtValue() == 0 && FalseVal->getZExtValue() != 0)
    SelectCCMaskVal ^= SelectCCValidVal;
  else
    return SDValue();

  if (SelectCCValidVal & ~CCValidVal)
    return SDValue();
  if (SelectCCMaskVal != (CCMaskVal & SelectCCValidVal))
    return SDValue();

  return Select->getOperand(4);
}

SDValue SystemZTargetLowering::combineIntDIVREM(
    SDNode *N, DAGCombinerInfo &DCI) const {
  SelectionDAG &DAG = DCI.DAG;
  EVT VT = N->getValueType(0);
  // In the case where the divisor is a vector of constants a cheaper
  // sequence of instructions can replace the divide. BuildSDIV is called to
  // do this during DAG combining, but it only succeeds when it can build a
  // multiplication node. The only option for SystemZ is ISD::SMUL_LOHI, and
  // since it is not Legal but Custom it can only happen before
  // legalization. Therefore we must scalarize this early before Combine
  // 1. For widened vectors, this is already the result of type legalization.
  if (DCI.Level == BeforeLegalizeTypes && VT.isVector() && isTypeLegal(VT) &&
      DAG.isConstantIntBuildVectorOrConstantInt(N->getOperand(1)))
    return DAG.UnrollVectorOp(N);
  return SDValue();
}

SDValue SystemZTargetLowering::combineINTRINSIC(
    SDNode *N, DAGCombinerInfo &DCI) const {
  SelectionDAG &DAG = DCI.DAG;

  unsigned Id = cast<ConstantSDNode>(N->getOperand(1))->getZExtValue();
  switch (Id) {
  // VECTOR LOAD (RIGHTMOST) WITH LENGTH with a length operand of 15
  // or larger is simply a vector load.
  case Intrinsic::s390_vll:
  case Intrinsic::s390_vlrl:
    if (auto *C = dyn_cast<ConstantSDNode>(N->getOperand(2)))
      if (C->getZExtValue() >= 15)
        return DAG.getLoad(N->getValueType(0), SDLoc(N), N->getOperand(0),
                           N->getOperand(3), MachinePointerInfo());
    break;
  // Likewise for VECTOR STORE (RIGHTMOST) WITH LENGTH.
  case Intrinsic::s390_vstl:
  case Intrinsic::s390_vstrl:
    if (auto *C = dyn_cast<ConstantSDNode>(N->getOperand(3)))
      if (C->getZExtValue() >= 15)
        return DAG.getStore(N->getOperand(0), SDLoc(N), N->getOperand(2),
                            N->getOperand(4), MachinePointerInfo());
    break;
  }

  return SDValue();
}

SDValue SystemZTargetLowering::unwrapAddress(SDValue N) const {
  if (N->getOpcode() == SystemZISD::PCREL_WRAPPER)
    return N->getOperand(0);
  return N;
}

SDValue SystemZTargetLowering::PerformDAGCombine(SDNode *N,
                                                 DAGCombinerInfo &DCI) const {
  switch(N->getOpcode()) {
  default: break;
  case ISD::ZERO_EXTEND:        return combineZERO_EXTEND(N, DCI);
  case ISD::SIGN_EXTEND:        return combineSIGN_EXTEND(N, DCI);
  case ISD::SIGN_EXTEND_INREG:  return combineSIGN_EXTEND_INREG(N, DCI);
  case SystemZISD::MERGE_HIGH:
  case SystemZISD::MERGE_LOW:   return combineMERGE(N, DCI);
  case ISD::LOAD:               return combineLOAD(N, DCI);
  case ISD::STORE:              return combineSTORE(N, DCI);
  case ISD::VECTOR_SHUFFLE:     return combineVECTOR_SHUFFLE(N, DCI);
  case ISD::EXTRACT_VECTOR_ELT: return combineEXTRACT_VECTOR_ELT(N, DCI);
  case SystemZISD::JOIN_DWORDS: return combineJOIN_DWORDS(N, DCI);
  case ISD::STRICT_FP_ROUND:
  case ISD::FP_ROUND:           return combineFP_ROUND(N, DCI);
  case ISD::STRICT_FP_EXTEND:
  case ISD::FP_EXTEND:          return combineFP_EXTEND(N, DCI);
  case ISD::SINT_TO_FP:
  case ISD::UINT_TO_FP:         return combineINT_TO_FP(N, DCI);
  case ISD::BSWAP:              return combineBSWAP(N, DCI);
  case SystemZISD::BR_CCMASK:   return combineBR_CCMASK(N, DCI);
  case SystemZISD::SELECT_CCMASK: return combineSELECT_CCMASK(N, DCI);
  case SystemZISD::GET_CCMASK:  return combineGET_CCMASK(N, DCI);
  case ISD::SDIV:
  case ISD::UDIV:
  case ISD::SREM:
  case ISD::UREM:               return combineIntDIVREM(N, DCI);
  case ISD::INTRINSIC_W_CHAIN:
  case ISD::INTRINSIC_VOID:     return combineINTRINSIC(N, DCI);
  }

  return SDValue();
}

// Return the demanded elements for the OpNo source operand of Op. DemandedElts
// are for Op.
static APInt getDemandedSrcElements(SDValue Op, const APInt &DemandedElts,
                                    unsigned OpNo) {
  EVT VT = Op.getValueType();
  unsigned NumElts = (VT.isVector() ? VT.getVectorNumElements() : 1);
  APInt SrcDemE;
  unsigned Opcode = Op.getOpcode();
  if (Opcode == ISD::INTRINSIC_WO_CHAIN) {
    unsigned Id = cast<ConstantSDNode>(Op.getOperand(0))->getZExtValue();
    switch (Id) {
    case Intrinsic::s390_vpksh:   // PACKS
    case Intrinsic::s390_vpksf:
    case Intrinsic::s390_vpksg:
    case Intrinsic::s390_vpkshs:  // PACKS_CC
    case Intrinsic::s390_vpksfs:
    case Intrinsic::s390_vpksgs:
    case Intrinsic::s390_vpklsh:  // PACKLS
    case Intrinsic::s390_vpklsf:
    case Intrinsic::s390_vpklsg:
    case Intrinsic::s390_vpklshs: // PACKLS_CC
    case Intrinsic::s390_vpklsfs:
    case Intrinsic::s390_vpklsgs:
      // VECTOR PACK truncates the elements of two source vectors into one.
      SrcDemE = DemandedElts;
      if (OpNo == 2)
        SrcDemE.lshrInPlace(NumElts / 2);
      SrcDemE = SrcDemE.trunc(NumElts / 2);
      break;
      // VECTOR UNPACK extends half the elements of the source vector.
    case Intrinsic::s390_vuphb:  // VECTOR UNPACK HIGH
    case Intrinsic::s390_vuphh:
    case Intrinsic::s390_vuphf:
    case Intrinsic::s390_vuplhb: // VECTOR UNPACK LOGICAL HIGH
    case Intrinsic::s390_vuplhh:
    case Intrinsic::s390_vuplhf:
      SrcDemE = APInt(NumElts * 2, 0);
      SrcDemE.insertBits(DemandedElts, 0);
      break;
    case Intrinsic::s390_vuplb:  // VECTOR UNPACK LOW
    case Intrinsic::s390_vuplhw:
    case Intrinsic::s390_vuplf:
    case Intrinsic::s390_vupllb: // VECTOR UNPACK LOGICAL LOW
    case Intrinsic::s390_vupllh:
    case Intrinsic::s390_vupllf:
      SrcDemE = APInt(NumElts * 2, 0);
      SrcDemE.insertBits(DemandedElts, NumElts);
      break;
    case Intrinsic::s390_vpdi: {
      // VECTOR PERMUTE DWORD IMMEDIATE selects one element from each source.
      SrcDemE = APInt(NumElts, 0);
      if (!DemandedElts[OpNo - 1])
        break;
      unsigned Mask = cast<ConstantSDNode>(Op.getOperand(3))->getZExtValue();
      unsigned MaskBit = ((OpNo - 1) ? 1 : 4);
      // Demand input element 0 or 1, given by the mask bit value.
      SrcDemE.setBit((Mask & MaskBit)? 1 : 0);
      break;
    }
    case Intrinsic::s390_vsldb: {
      // VECTOR SHIFT LEFT DOUBLE BY BYTE
      assert(VT == MVT::v16i8 && "Unexpected type.");
      unsigned FirstIdx = cast<ConstantSDNode>(Op.getOperand(3))->getZExtValue();
      assert (FirstIdx > 0 && FirstIdx < 16 && "Unused operand.");
      unsigned NumSrc0Els = 16 - FirstIdx;
      SrcDemE = APInt(NumElts, 0);
      if (OpNo == 1) {
        APInt DemEls = DemandedElts.trunc(NumSrc0Els);
        SrcDemE.insertBits(DemEls, FirstIdx);
      } else {
        APInt DemEls = DemandedElts.lshr(NumSrc0Els);
        SrcDemE.insertBits(DemEls, 0);
      }
      break;
    }
    case Intrinsic::s390_vperm:
      SrcDemE = APInt(NumElts, 1);
      break;
    default:
      llvm_unreachable("Unhandled intrinsic.");
      break;
    }
  } else {
    switch (Opcode) {
    case SystemZISD::JOIN_DWORDS:
      // Scalar operand.
      SrcDemE = APInt(1, 1);
      break;
    case SystemZISD::SELECT_CCMASK:
      SrcDemE = DemandedElts;
      break;
    default:
      llvm_unreachable("Unhandled opcode.");
      break;
    }
  }
  return SrcDemE;
}

static void computeKnownBitsBinOp(const SDValue Op, KnownBits &Known,
                                  const APInt &DemandedElts,
                                  const SelectionDAG &DAG, unsigned Depth,
                                  unsigned OpNo) {
  APInt Src0DemE = getDemandedSrcElements(Op, DemandedElts, OpNo);
  APInt Src1DemE = getDemandedSrcElements(Op, DemandedElts, OpNo + 1);
  KnownBits LHSKnown =
      DAG.computeKnownBits(Op.getOperand(OpNo), Src0DemE, Depth + 1);
  KnownBits RHSKnown =
      DAG.computeKnownBits(Op.getOperand(OpNo + 1), Src1DemE, Depth + 1);
  Known.Zero = LHSKnown.Zero & RHSKnown.Zero;
  Known.One = LHSKnown.One & RHSKnown.One;
}

void
SystemZTargetLowering::computeKnownBitsForTargetNode(const SDValue Op,
                                                     KnownBits &Known,
                                                     const APInt &DemandedElts,
                                                     const SelectionDAG &DAG,
                                                     unsigned Depth) const {
  Known.resetAll();

  // Intrinsic CC result is returned in the two low bits.
  unsigned tmp0, tmp1; // not used
  if (Op.getResNo() == 1 && isIntrinsicWithCC(Op, tmp0, tmp1)) {
    Known.Zero.setBitsFrom(2);
    return;
  }
  EVT VT = Op.getValueType();
  if (Op.getResNo() != 0 || VT == MVT::Untyped)
    return;
  assert (Known.getBitWidth() == VT.getScalarSizeInBits() &&
          "KnownBits does not match VT in bitwidth");
  assert ((!VT.isVector() ||
           (DemandedElts.getBitWidth() == VT.getVectorNumElements())) &&
          "DemandedElts does not match VT number of elements");
  unsigned BitWidth = Known.getBitWidth();
  unsigned Opcode = Op.getOpcode();
  if (Opcode == ISD::INTRINSIC_WO_CHAIN) {
    bool IsLogical = false;
    unsigned Id = cast<ConstantSDNode>(Op.getOperand(0))->getZExtValue();
    switch (Id) {
    case Intrinsic::s390_vpksh:   // PACKS
    case Intrinsic::s390_vpksf:
    case Intrinsic::s390_vpksg:
    case Intrinsic::s390_vpkshs:  // PACKS_CC
    case Intrinsic::s390_vpksfs:
    case Intrinsic::s390_vpksgs:
    case Intrinsic::s390_vpklsh:  // PACKLS
    case Intrinsic::s390_vpklsf:
    case Intrinsic::s390_vpklsg:
    case Intrinsic::s390_vpklshs: // PACKLS_CC
    case Intrinsic::s390_vpklsfs:
    case Intrinsic::s390_vpklsgs:
    case Intrinsic::s390_vpdi:
    case Intrinsic::s390_vsldb:
    case Intrinsic::s390_vperm:
      computeKnownBitsBinOp(Op, Known, DemandedElts, DAG, Depth, 1);
      break;
    case Intrinsic::s390_vuplhb: // VECTOR UNPACK LOGICAL HIGH
    case Intrinsic::s390_vuplhh:
    case Intrinsic::s390_vuplhf:
    case Intrinsic::s390_vupllb: // VECTOR UNPACK LOGICAL LOW
    case Intrinsic::s390_vupllh:
    case Intrinsic::s390_vupllf:
      IsLogical = true;
      LLVM_FALLTHROUGH;
    case Intrinsic::s390_vuphb:  // VECTOR UNPACK HIGH
    case Intrinsic::s390_vuphh:
    case Intrinsic::s390_vuphf:
    case Intrinsic::s390_vuplb:  // VECTOR UNPACK LOW
    case Intrinsic::s390_vuplhw:
    case Intrinsic::s390_vuplf: {
      SDValue SrcOp = Op.getOperand(1);
      APInt SrcDemE = getDemandedSrcElements(Op, DemandedElts, 0);
      Known = DAG.computeKnownBits(SrcOp, SrcDemE, Depth + 1);
      if (IsLogical) {
        Known = Known.zext(BitWidth);
      } else
        Known = Known.sext(BitWidth);
      break;
    }
    default:
      break;
    }
  } else {
    switch (Opcode) {
    case SystemZISD::JOIN_DWORDS:
    case SystemZISD::SELECT_CCMASK:
      computeKnownBitsBinOp(Op, Known, DemandedElts, DAG, Depth, 0);
      break;
    case SystemZISD::REPLICATE: {
      SDValue SrcOp = Op.getOperand(0);
      Known = DAG.computeKnownBits(SrcOp, Depth + 1);
      if (Known.getBitWidth() < BitWidth && isa<ConstantSDNode>(SrcOp))
        Known = Known.sext(BitWidth); // VREPI sign extends the immedate.
      break;
    }
    default:
      break;
    }
  }

  // Known has the width of the source operand(s). Adjust if needed to match
  // the passed bitwidth.
  if (Known.getBitWidth() != BitWidth)
    Known = Known.anyextOrTrunc(BitWidth);
}

static unsigned computeNumSignBitsBinOp(SDValue Op, const APInt &DemandedElts,
                                        const SelectionDAG &DAG, unsigned Depth,
                                        unsigned OpNo) {
  APInt Src0DemE = getDemandedSrcElements(Op, DemandedElts, OpNo);
  unsigned LHS = DAG.ComputeNumSignBits(Op.getOperand(OpNo), Src0DemE, Depth + 1);
  if (LHS == 1) return 1; // Early out.
  APInt Src1DemE = getDemandedSrcElements(Op, DemandedElts, OpNo + 1);
  unsigned RHS = DAG.ComputeNumSignBits(Op.getOperand(OpNo + 1), Src1DemE, Depth + 1);
  if (RHS == 1) return 1; // Early out.
  unsigned Common = std::min(LHS, RHS);
  unsigned SrcBitWidth = Op.getOperand(OpNo).getScalarValueSizeInBits();
  EVT VT = Op.getValueType();
  unsigned VTBits = VT.getScalarSizeInBits();
  if (SrcBitWidth > VTBits) { // PACK
    unsigned SrcExtraBits = SrcBitWidth - VTBits;
    if (Common > SrcExtraBits)
      return (Common - SrcExtraBits);
    return 1;
  }
  assert (SrcBitWidth == VTBits && "Expected operands of same bitwidth.");
  return Common;
}

unsigned
SystemZTargetLowering::ComputeNumSignBitsForTargetNode(
    SDValue Op, const APInt &DemandedElts, const SelectionDAG &DAG,
    unsigned Depth) const {
  if (Op.getResNo() != 0)
    return 1;
  unsigned Opcode = Op.getOpcode();
  if (Opcode == ISD::INTRINSIC_WO_CHAIN) {
    unsigned Id = cast<ConstantSDNode>(Op.getOperand(0))->getZExtValue();
    switch (Id) {
    case Intrinsic::s390_vpksh:   // PACKS
    case Intrinsic::s390_vpksf:
    case Intrinsic::s390_vpksg:
    case Intrinsic::s390_vpkshs:  // PACKS_CC
    case Intrinsic::s390_vpksfs:
    case Intrinsic::s390_vpksgs:
    case Intrinsic::s390_vpklsh:  // PACKLS
    case Intrinsic::s390_vpklsf:
    case Intrinsic::s390_vpklsg:
    case Intrinsic::s390_vpklshs: // PACKLS_CC
    case Intrinsic::s390_vpklsfs:
    case Intrinsic::s390_vpklsgs:
    case Intrinsic::s390_vpdi:
    case Intrinsic::s390_vsldb:
    case Intrinsic::s390_vperm:
      return computeNumSignBitsBinOp(Op, DemandedElts, DAG, Depth, 1);
    case Intrinsic::s390_vuphb:  // VECTOR UNPACK HIGH
    case Intrinsic::s390_vuphh:
    case Intrinsic::s390_vuphf:
    case Intrinsic::s390_vuplb:  // VECTOR UNPACK LOW
    case Intrinsic::s390_vuplhw:
    case Intrinsic::s390_vuplf: {
      SDValue PackedOp = Op.getOperand(1);
      APInt SrcDemE = getDemandedSrcElements(Op, DemandedElts, 1);
      unsigned Tmp = DAG.ComputeNumSignBits(PackedOp, SrcDemE, Depth + 1);
      EVT VT = Op.getValueType();
      unsigned VTBits = VT.getScalarSizeInBits();
      Tmp += VTBits - PackedOp.getScalarValueSizeInBits();
      return Tmp;
    }
    default:
      break;
    }
  } else {
    switch (Opcode) {
    case SystemZISD::SELECT_CCMASK:
      return computeNumSignBitsBinOp(Op, DemandedElts, DAG, Depth, 0);
    default:
      break;
    }
  }

  return 1;
}

unsigned
SystemZTargetLowering::getStackProbeSize(MachineFunction &MF) const {
  const TargetFrameLowering *TFI = Subtarget.getFrameLowering();
  unsigned StackAlign = TFI->getStackAlignment();
  assert(StackAlign >=1 && isPowerOf2_32(StackAlign) &&
         "Unexpected stack alignment");
  // The default stack probe size is 4096 if the function has no
  // stack-probe-size attribute.
  unsigned StackProbeSize = 4096;
  const Function &Fn = MF.getFunction();
  if (Fn.hasFnAttribute("stack-probe-size"))
    Fn.getFnAttribute("stack-probe-size")
        .getValueAsString()
        .getAsInteger(0, StackProbeSize);
  // Round down to the stack alignment.
  StackProbeSize &= ~(StackAlign - 1);
  return StackProbeSize ? StackProbeSize : StackAlign;
}

//===----------------------------------------------------------------------===//
// Custom insertion
//===----------------------------------------------------------------------===//

// Force base value Base into a register before MI.  Return the register.
static Register forceReg(MachineInstr &MI, MachineOperand &Base,
                         const SystemZInstrInfo *TII) {
  if (Base.isReg())
    return Base.getReg();

  MachineBasicBlock *MBB = MI.getParent();
  MachineFunction &MF = *MBB->getParent();
  MachineRegisterInfo &MRI = MF.getRegInfo();

  Register Reg = MRI.createVirtualRegister(&SystemZ::ADDR64BitRegClass);
  BuildMI(*MBB, MI, MI.getDebugLoc(), TII->get(SystemZ::LA), Reg)
      .add(Base)
      .addImm(0)
      .addReg(0);
  return Reg;
}

// The CC operand of MI might be missing a kill marker because there
// were multiple uses of CC, and ISel didn't know which to mark.
// Figure out whether MI should have had a kill marker.
static bool checkCCKill(MachineInstr &MI, MachineBasicBlock *MBB) {
  // Scan forward through BB for a use/def of CC.
  MachineBasicBlock::iterator miI(std::next(MachineBasicBlock::iterator(MI)));
  for (MachineBasicBlock::iterator miE = MBB->end(); miI != miE; ++miI) {
    const MachineInstr& mi = *miI;
    if (mi.readsRegister(SystemZ::CC))
      return false;
    if (mi.definesRegister(SystemZ::CC))
      break; // Should have kill-flag - update below.
  }

  // If we hit the end of the block, check whether CC is live into a
  // successor.
  if (miI == MBB->end()) {
    for (auto SI = MBB->succ_begin(), SE = MBB->succ_end(); SI != SE; ++SI)
      if ((*SI)->isLiveIn(SystemZ::CC))
        return false;
  }

  return true;
}

// Return true if it is OK for this Select pseudo-opcode to be cascaded
// together with other Select pseudo-opcodes into a single basic-block with
// a conditional jump around it.
static bool isSelectPseudo(MachineInstr &MI) {
  switch (MI.getOpcode()) {
  case SystemZ::Select32:
  case SystemZ::Select64:
  case SystemZ::SelectF32:
  case SystemZ::SelectF64:
  case SystemZ::SelectF128:
  case SystemZ::SelectVR32:
  case SystemZ::SelectVR64:
  case SystemZ::SelectVR128:
    return true;

  default:
    return false;
  }
}

// Helper function, which inserts PHI functions into SinkMBB:
//   %Result(i) = phi [ %FalseValue(i), FalseMBB ], [ %TrueValue(i), TrueMBB ],
// where %FalseValue(i) and %TrueValue(i) are taken from Selects.
static void createPHIsForSelects(SmallVector<MachineInstr*, 8> &Selects,
                                 MachineBasicBlock *TrueMBB,
                                 MachineBasicBlock *FalseMBB,
                                 MachineBasicBlock *SinkMBB) {
  MachineFunction *MF = TrueMBB->getParent();
  const TargetInstrInfo *TII = MF->getSubtarget().getInstrInfo();

  MachineInstr *FirstMI = Selects.front();
  unsigned CCValid = FirstMI->getOperand(3).getImm();
  unsigned CCMask = FirstMI->getOperand(4).getImm();

  MachineBasicBlock::iterator SinkInsertionPoint = SinkMBB->begin();

  // As we are creating the PHIs, we have to be careful if there is more than
  // one.  Later Selects may reference the results of earlier Selects, but later
  // PHIs have to reference the individual true/false inputs from earlier PHIs.
  // That also means that PHI construction must work forward from earlier to
  // later, and that the code must maintain a mapping from earlier PHI's
  // destination registers, and the registers that went into the PHI.
  DenseMap<unsigned, std::pair<unsigned, unsigned>> RegRewriteTable;

  for (auto MI : Selects) {
    Register DestReg = MI->getOperand(0).getReg();
    Register TrueReg = MI->getOperand(1).getReg();
    Register FalseReg = MI->getOperand(2).getReg();

    // If this Select we are generating is the opposite condition from
    // the jump we generated, then we have to swap the operands for the
    // PHI that is going to be generated.
    if (MI->getOperand(4).getImm() == (CCValid ^ CCMask))
      std::swap(TrueReg, FalseReg);

    if (RegRewriteTable.find(TrueReg) != RegRewriteTable.end())
      TrueReg = RegRewriteTable[TrueReg].first;

    if (RegRewriteTable.find(FalseReg) != RegRewriteTable.end())
      FalseReg = RegRewriteTable[FalseReg].second;

    DebugLoc DL = MI->getDebugLoc();
    BuildMI(*SinkMBB, SinkInsertionPoint, DL, TII->get(SystemZ::PHI), DestReg)
      .addReg(TrueReg).addMBB(TrueMBB)
      .addReg(FalseReg).addMBB(FalseMBB);

    // Add this PHI to the rewrite table.
    RegRewriteTable[DestReg] = std::make_pair(TrueReg, FalseReg);
  }

  MF->getProperties().reset(MachineFunctionProperties::Property::NoPHIs);
}

// Implement EmitInstrWithCustomInserter for pseudo Select* instruction MI.
MachineBasicBlock *
SystemZTargetLowering::emitSelect(MachineInstr &MI,
                                  MachineBasicBlock *MBB) const {
  assert(isSelectPseudo(MI) && "Bad call to emitSelect()");
  const SystemZInstrInfo *TII =
      static_cast<const SystemZInstrInfo *>(Subtarget.getInstrInfo());

  unsigned CCValid = MI.getOperand(3).getImm();
  unsigned CCMask = MI.getOperand(4).getImm();

  // If we have a sequence of Select* pseudo instructions using the
  // same condition code value, we want to expand all of them into
  // a single pair of basic blocks using the same condition.
  SmallVector<MachineInstr*, 8> Selects;
  SmallVector<MachineInstr*, 8> DbgValues;
  Selects.push_back(&MI);
  unsigned Count = 0;
  for (MachineBasicBlock::iterator NextMIIt =
         std::next(MachineBasicBlock::iterator(MI));
       NextMIIt != MBB->end(); ++NextMIIt) {
    if (isSelectPseudo(*NextMIIt)) {
      assert(NextMIIt->getOperand(3).getImm() == CCValid &&
             "Bad CCValid operands since CC was not redefined.");
      if (NextMIIt->getOperand(4).getImm() == CCMask ||
          NextMIIt->getOperand(4).getImm() == (CCValid ^ CCMask)) {
        Selects.push_back(&*NextMIIt);
        continue;
      }
      break;
    }
    if (NextMIIt->definesRegister(SystemZ::CC) ||
        NextMIIt->usesCustomInsertionHook())
      break;
    bool User = false;
    for (auto SelMI : Selects)
      if (NextMIIt->readsVirtualRegister(SelMI->getOperand(0).getReg())) {
        User = true;
        break;
      }
    if (NextMIIt->isDebugInstr()) {
      if (User) {
        assert(NextMIIt->isDebugValue() && "Unhandled debug opcode.");
        DbgValues.push_back(&*NextMIIt);
      }
    }
    else if (User || ++Count > 20)
      break;
  }

  MachineInstr *LastMI = Selects.back();
  bool CCKilled =
      (LastMI->killsRegister(SystemZ::CC) || checkCCKill(*LastMI, MBB));
  MachineBasicBlock *StartMBB = MBB;
  MachineBasicBlock *JoinMBB  = SystemZ::splitBlockAfter(LastMI, MBB);
  MachineBasicBlock *FalseMBB = SystemZ::emitBlockAfter(StartMBB);

  // Unless CC was killed in the last Select instruction, mark it as
  // live-in to both FalseMBB and JoinMBB.
  if (!CCKilled) {
    FalseMBB->addLiveIn(SystemZ::CC);
    JoinMBB->addLiveIn(SystemZ::CC);
  }

  //  StartMBB:
  //   BRC CCMask, JoinMBB
  //   # fallthrough to FalseMBB
  MBB = StartMBB;
  BuildMI(MBB, MI.getDebugLoc(), TII->get(SystemZ::BRC))
    .addImm(CCValid).addImm(CCMask).addMBB(JoinMBB);
  MBB->addSuccessor(JoinMBB);
  MBB->addSuccessor(FalseMBB);

  //  FalseMBB:
  //   # fallthrough to JoinMBB
  MBB = FalseMBB;
  MBB->addSuccessor(JoinMBB);

  //  JoinMBB:
  //   %Result = phi [ %FalseReg, FalseMBB ], [ %TrueReg, StartMBB ]
  //  ...
  MBB = JoinMBB;
  createPHIsForSelects(Selects, StartMBB, FalseMBB, MBB);
  for (auto SelMI : Selects)
    SelMI->eraseFromParent();

  MachineBasicBlock::iterator InsertPos = MBB->getFirstNonPHI();
  for (auto DbgMI : DbgValues)
    MBB->splice(InsertPos, StartMBB, DbgMI);

  return JoinMBB;
}

// Implement EmitInstrWithCustomInserter for pseudo CondStore* instruction MI.
// StoreOpcode is the store to use and Invert says whether the store should
// happen when the condition is false rather than true.  If a STORE ON
// CONDITION is available, STOCOpcode is its opcode, otherwise it is 0.
MachineBasicBlock *SystemZTargetLowering::emitCondStore(MachineInstr &MI,
                                                        MachineBasicBlock *MBB,
                                                        unsigned StoreOpcode,
                                                        unsigned STOCOpcode,
                                                        bool Invert) const {
  const SystemZInstrInfo *TII =
      static_cast<const SystemZInstrInfo *>(Subtarget.getInstrInfo());

  Register SrcReg = MI.getOperand(0).getReg();
  MachineOperand Base = MI.getOperand(1);
  int64_t Disp = MI.getOperand(2).getImm();
  Register IndexReg = MI.getOperand(3).getReg();
  unsigned CCValid = MI.getOperand(4).getImm();
  unsigned CCMask = MI.getOperand(5).getImm();
  DebugLoc DL = MI.getDebugLoc();

  StoreOpcode = TII->getOpcodeForOffset(StoreOpcode, Disp);

  // Use STOCOpcode if possible.  We could use different store patterns in
  // order to avoid matching the index register, but the performance trade-offs
  // might be more complicated in that case.
  if (STOCOpcode && !IndexReg && Subtarget.hasLoadStoreOnCond()) {
    if (Invert)
      CCMask ^= CCValid;

    // ISel pattern matching also adds a load memory operand of the same
    // address, so take special care to find the storing memory operand.
    MachineMemOperand *MMO = nullptr;
    for (auto *I : MI.memoperands())
      if (I->isStore()) {
          MMO = I;
          break;
        }

    BuildMI(*MBB, MI, DL, TII->get(STOCOpcode))
      .addReg(SrcReg)
      .add(Base)
      .addImm(Disp)
      .addImm(CCValid)
      .addImm(CCMask)
      .addMemOperand(MMO);

    MI.eraseFromParent();
    return MBB;
  }

  // Get the condition needed to branch around the store.
  if (!Invert)
    CCMask ^= CCValid;

  MachineBasicBlock *StartMBB = MBB;
  MachineBasicBlock *JoinMBB  = SystemZ::splitBlockBefore(MI, MBB);
  MachineBasicBlock *FalseMBB = SystemZ::emitBlockAfter(StartMBB);

  // Unless CC was killed in the CondStore instruction, mark it as
  // live-in to both FalseMBB and JoinMBB.
  if (!MI.killsRegister(SystemZ::CC) && !checkCCKill(MI, JoinMBB)) {
    FalseMBB->addLiveIn(SystemZ::CC);
    JoinMBB->addLiveIn(SystemZ::CC);
  }

  //  StartMBB:
  //   BRC CCMask, JoinMBB
  //   # fallthrough to FalseMBB
  MBB = StartMBB;
  BuildMI(MBB, DL, TII->get(SystemZ::BRC))
    .addImm(CCValid).addImm(CCMask).addMBB(JoinMBB);
  MBB->addSuccessor(JoinMBB);
  MBB->addSuccessor(FalseMBB);

  //  FalseMBB:
  //   store %SrcReg, %Disp(%Index,%Base)
  //   # fallthrough to JoinMBB
  MBB = FalseMBB;
  BuildMI(MBB, DL, TII->get(StoreOpcode))
      .addReg(SrcReg)
      .add(Base)
      .addImm(Disp)
      .addReg(IndexReg);
  MBB->addSuccessor(JoinMBB);

  MI.eraseFromParent();
  return JoinMBB;
}

// Implement EmitInstrWithCustomInserter for pseudo ATOMIC_LOAD{,W}_*
// or ATOMIC_SWAP{,W} instruction MI.  BinOpcode is the instruction that
// performs the binary operation elided by "*", or 0 for ATOMIC_SWAP{,W}.
// BitSize is the width of the field in bits, or 0 if this is a partword
// ATOMIC_LOADW_* or ATOMIC_SWAPW instruction, in which case the bitsize
// is one of the operands.  Invert says whether the field should be
// inverted after performing BinOpcode (e.g. for NAND).
MachineBasicBlock *SystemZTargetLowering::emitAtomicLoadBinary(
    MachineInstr &MI, MachineBasicBlock *MBB, unsigned BinOpcode,
    unsigned BitSize, bool Invert) const {
  MachineFunction &MF = *MBB->getParent();
  const SystemZInstrInfo *TII =
      static_cast<const SystemZInstrInfo *>(Subtarget.getInstrInfo());
  MachineRegisterInfo &MRI = MF.getRegInfo();
  bool IsSubWord = (BitSize < 32);

  // Extract the operands.  Base can be a register or a frame index.
  // Src2 can be a register or immediate.
  Register Dest = MI.getOperand(0).getReg();
  MachineOperand Base = earlyUseOperand(MI.getOperand(1));
  int64_t Disp = MI.getOperand(2).getImm();
  MachineOperand Src2 = earlyUseOperand(MI.getOperand(3));
  Register BitShift = IsSubWord ? MI.getOperand(4).getReg() : Register();
  Register NegBitShift = IsSubWord ? MI.getOperand(5).getReg() : Register();
  DebugLoc DL = MI.getDebugLoc();
  if (IsSubWord)
    BitSize = MI.getOperand(6).getImm();

  // Subword operations use 32-bit registers.
  const TargetRegisterClass *RC = (BitSize <= 32 ?
                                   &SystemZ::GR32BitRegClass :
                                   &SystemZ::GR64BitRegClass);
  unsigned LOpcode  = BitSize <= 32 ? SystemZ::L  : SystemZ::LG;
  unsigned CSOpcode = BitSize <= 32 ? SystemZ::CS : SystemZ::CSG;

  // Get the right opcodes for the displacement.
  LOpcode  = TII->getOpcodeForOffset(LOpcode,  Disp);
  CSOpcode = TII->getOpcodeForOffset(CSOpcode, Disp);
  assert(LOpcode && CSOpcode && "Displacement out of range");

  // Create virtual registers for temporary results.
  Register OrigVal       = MRI.createVirtualRegister(RC);
  Register OldVal        = MRI.createVirtualRegister(RC);
  Register NewVal        = (BinOpcode || IsSubWord ?
                            MRI.createVirtualRegister(RC) : Src2.getReg());
  Register RotatedOldVal = (IsSubWord ? MRI.createVirtualRegister(RC) : OldVal);
  Register RotatedNewVal = (IsSubWord ? MRI.createVirtualRegister(RC) : NewVal);

  // Insert a basic block for the main loop.
  MachineBasicBlock *StartMBB = MBB;
  MachineBasicBlock *DoneMBB  = SystemZ::splitBlockBefore(MI, MBB);
  MachineBasicBlock *LoopMBB  = SystemZ::emitBlockAfter(StartMBB);

  //  StartMBB:
  //   ...
  //   %OrigVal = L Disp(%Base)
  //   # fall through to LoopMMB
  MBB = StartMBB;
  BuildMI(MBB, DL, TII->get(LOpcode), OrigVal).add(Base).addImm(Disp).addReg(0);
  MBB->addSuccessor(LoopMBB);

  //  LoopMBB:
  //   %OldVal        = phi [ %OrigVal, StartMBB ], [ %Dest, LoopMBB ]
  //   %RotatedOldVal = RLL %OldVal, 0(%BitShift)
  //   %RotatedNewVal = OP %RotatedOldVal, %Src2
  //   %NewVal        = RLL %RotatedNewVal, 0(%NegBitShift)
  //   %Dest          = CS %OldVal, %NewVal, Disp(%Base)
  //   JNE LoopMBB
  //   # fall through to DoneMMB
  MBB = LoopMBB;
  BuildMI(MBB, DL, TII->get(SystemZ::PHI), OldVal)
    .addReg(OrigVal).addMBB(StartMBB)
    .addReg(Dest).addMBB(LoopMBB);
  if (IsSubWord)
    BuildMI(MBB, DL, TII->get(SystemZ::RLL), RotatedOldVal)
      .addReg(OldVal).addReg(BitShift).addImm(0);
  if (Invert) {
    // Perform the operation normally and then invert every bit of the field.
    Register Tmp = MRI.createVirtualRegister(RC);
    BuildMI(MBB, DL, TII->get(BinOpcode), Tmp).addReg(RotatedOldVal).add(Src2);
    if (BitSize <= 32)
      // XILF with the upper BitSize bits set.
      BuildMI(MBB, DL, TII->get(SystemZ::XILF), RotatedNewVal)
        .addReg(Tmp).addImm(-1U << (32 - BitSize));
    else {
      // Use LCGR and add -1 to the result, which is more compact than
      // an XILF, XILH pair.
      Register Tmp2 = MRI.createVirtualRegister(RC);
      BuildMI(MBB, DL, TII->get(SystemZ::LCGR), Tmp2).addReg(Tmp);
      BuildMI(MBB, DL, TII->get(SystemZ::AGHI), RotatedNewVal)
        .addReg(Tmp2).addImm(-1);
    }
  } else if (BinOpcode)
    // A simply binary operation.
    BuildMI(MBB, DL, TII->get(BinOpcode), RotatedNewVal)
        .addReg(RotatedOldVal)
        .add(Src2);
  else if (IsSubWord)
    // Use RISBG to rotate Src2 into position and use it to replace the
    // field in RotatedOldVal.
    BuildMI(MBB, DL, TII->get(SystemZ::RISBG32), RotatedNewVal)
      .addReg(RotatedOldVal).addReg(Src2.getReg())
      .addImm(32).addImm(31 + BitSize).addImm(32 - BitSize);
  if (IsSubWord)
    BuildMI(MBB, DL, TII->get(SystemZ::RLL), NewVal)
      .addReg(RotatedNewVal).addReg(NegBitShift).addImm(0);
  BuildMI(MBB, DL, TII->get(CSOpcode), Dest)
      .addReg(OldVal)
      .addReg(NewVal)
      .add(Base)
      .addImm(Disp);
  BuildMI(MBB, DL, TII->get(SystemZ::BRC))
    .addImm(SystemZ::CCMASK_CS).addImm(SystemZ::CCMASK_CS_NE).addMBB(LoopMBB);
  MBB->addSuccessor(LoopMBB);
  MBB->addSuccessor(DoneMBB);

  MI.eraseFromParent();
  return DoneMBB;
}

// Implement EmitInstrWithCustomInserter for pseudo
// ATOMIC_LOAD{,W}_{,U}{MIN,MAX} instruction MI.  CompareOpcode is the
// instruction that should be used to compare the current field with the
// minimum or maximum value.  KeepOldMask is the BRC condition-code mask
// for when the current field should be kept.  BitSize is the width of
// the field in bits, or 0 if this is a partword ATOMIC_LOADW_* instruction.
MachineBasicBlock *SystemZTargetLowering::emitAtomicLoadMinMax(
    MachineInstr &MI, MachineBasicBlock *MBB, unsigned CompareOpcode,
    unsigned KeepOldMask, unsigned BitSize) const {
  MachineFunction &MF = *MBB->getParent();
  const SystemZInstrInfo *TII =
      static_cast<const SystemZInstrInfo *>(Subtarget.getInstrInfo());
  MachineRegisterInfo &MRI = MF.getRegInfo();
  bool IsSubWord = (BitSize < 32);

  // Extract the operands.  Base can be a register or a frame index.
  Register Dest = MI.getOperand(0).getReg();
  MachineOperand Base = earlyUseOperand(MI.getOperand(1));
  int64_t Disp = MI.getOperand(2).getImm();
  Register Src2 = MI.getOperand(3).getReg();
  Register BitShift = (IsSubWord ? MI.getOperand(4).getReg() : Register());
  Register NegBitShift = (IsSubWord ? MI.getOperand(5).getReg() : Register());
  DebugLoc DL = MI.getDebugLoc();
  if (IsSubWord)
    BitSize = MI.getOperand(6).getImm();

  // Subword operations use 32-bit registers.
  const TargetRegisterClass *RC = (BitSize <= 32 ?
                                   &SystemZ::GR32BitRegClass :
                                   &SystemZ::GR64BitRegClass);
  unsigned LOpcode  = BitSize <= 32 ? SystemZ::L  : SystemZ::LG;
  unsigned CSOpcode = BitSize <= 32 ? SystemZ::CS : SystemZ::CSG;

  // Get the right opcodes for the displacement.
  LOpcode  = TII->getOpcodeForOffset(LOpcode,  Disp);
  CSOpcode = TII->getOpcodeForOffset(CSOpcode, Disp);
  assert(LOpcode && CSOpcode && "Displacement out of range");

  // Create virtual registers for temporary results.
  Register OrigVal       = MRI.createVirtualRegister(RC);
  Register OldVal        = MRI.createVirtualRegister(RC);
  Register NewVal        = MRI.createVirtualRegister(RC);
  Register RotatedOldVal = (IsSubWord ? MRI.createVirtualRegister(RC) : OldVal);
  Register RotatedAltVal = (IsSubWord ? MRI.createVirtualRegister(RC) : Src2);
  Register RotatedNewVal = (IsSubWord ? MRI.createVirtualRegister(RC) : NewVal);

  // Insert 3 basic blocks for the loop.
  MachineBasicBlock *StartMBB  = MBB;
  MachineBasicBlock *DoneMBB   = SystemZ::splitBlockBefore(MI, MBB);
  MachineBasicBlock *LoopMBB   = SystemZ::emitBlockAfter(StartMBB);
  MachineBasicBlock *UseAltMBB = SystemZ::emitBlockAfter(LoopMBB);
  MachineBasicBlock *UpdateMBB = SystemZ::emitBlockAfter(UseAltMBB);

  //  StartMBB:
  //   ...
  //   %OrigVal     = L Disp(%Base)
  //   # fall through to LoopMMB
  MBB = StartMBB;
  BuildMI(MBB, DL, TII->get(LOpcode), OrigVal).add(Base).addImm(Disp).addReg(0);
  MBB->addSuccessor(LoopMBB);

  //  LoopMBB:
  //   %OldVal        = phi [ %OrigVal, StartMBB ], [ %Dest, UpdateMBB ]
  //   %RotatedOldVal = RLL %OldVal, 0(%BitShift)
  //   CompareOpcode %RotatedOldVal, %Src2
  //   BRC KeepOldMask, UpdateMBB
  MBB = LoopMBB;
  BuildMI(MBB, DL, TII->get(SystemZ::PHI), OldVal)
    .addReg(OrigVal).addMBB(StartMBB)
    .addReg(Dest).addMBB(UpdateMBB);
  if (IsSubWord)
    BuildMI(MBB, DL, TII->get(SystemZ::RLL), RotatedOldVal)
      .addReg(OldVal).addReg(BitShift).addImm(0);
  BuildMI(MBB, DL, TII->get(CompareOpcode))
    .addReg(RotatedOldVal).addReg(Src2);
  BuildMI(MBB, DL, TII->get(SystemZ::BRC))
    .addImm(SystemZ::CCMASK_ICMP).addImm(KeepOldMask).addMBB(UpdateMBB);
  MBB->addSuccessor(UpdateMBB);
  MBB->addSuccessor(UseAltMBB);

  //  UseAltMBB:
  //   %RotatedAltVal = RISBG %RotatedOldVal, %Src2, 32, 31 + BitSize, 0
  //   # fall through to UpdateMMB
  MBB = UseAltMBB;
  if (IsSubWord)
    BuildMI(MBB, DL, TII->get(SystemZ::RISBG32), RotatedAltVal)
      .addReg(RotatedOldVal).addReg(Src2)
      .addImm(32).addImm(31 + BitSize).addImm(0);
  MBB->addSuccessor(UpdateMBB);

  //  UpdateMBB:
  //   %RotatedNewVal = PHI [ %RotatedOldVal, LoopMBB ],
  //                        [ %RotatedAltVal, UseAltMBB ]
  //   %NewVal        = RLL %RotatedNewVal, 0(%NegBitShift)
  //   %Dest          = CS %OldVal, %NewVal, Disp(%Base)
  //   JNE LoopMBB
  //   # fall through to DoneMMB
  MBB = UpdateMBB;
  BuildMI(MBB, DL, TII->get(SystemZ::PHI), RotatedNewVal)
    .addReg(RotatedOldVal).addMBB(LoopMBB)
    .addReg(RotatedAltVal).addMBB(UseAltMBB);
  if (IsSubWord)
    BuildMI(MBB, DL, TII->get(SystemZ::RLL), NewVal)
      .addReg(RotatedNewVal).addReg(NegBitShift).addImm(0);
  BuildMI(MBB, DL, TII->get(CSOpcode), Dest)
      .addReg(OldVal)
      .addReg(NewVal)
      .add(Base)
      .addImm(Disp);
  BuildMI(MBB, DL, TII->get(SystemZ::BRC))
    .addImm(SystemZ::CCMASK_CS).addImm(SystemZ::CCMASK_CS_NE).addMBB(LoopMBB);
  MBB->addSuccessor(LoopMBB);
  MBB->addSuccessor(DoneMBB);

  MI.eraseFromParent();
  return DoneMBB;
}

// Implement EmitInstrWithCustomInserter for pseudo ATOMIC_CMP_SWAPW
// instruction MI.
MachineBasicBlock *
SystemZTargetLowering::emitAtomicCmpSwapW(MachineInstr &MI,
                                          MachineBasicBlock *MBB) const {

  MachineFunction &MF = *MBB->getParent();
  const SystemZInstrInfo *TII =
      static_cast<const SystemZInstrInfo *>(Subtarget.getInstrInfo());
  MachineRegisterInfo &MRI = MF.getRegInfo();

  // Extract the operands.  Base can be a register or a frame index.
  Register Dest = MI.getOperand(0).getReg();
  MachineOperand Base = earlyUseOperand(MI.getOperand(1));
  int64_t Disp = MI.getOperand(2).getImm();
  Register OrigCmpVal = MI.getOperand(3).getReg();
  Register OrigSwapVal = MI.getOperand(4).getReg();
  Register BitShift = MI.getOperand(5).getReg();
  Register NegBitShift = MI.getOperand(6).getReg();
  int64_t BitSize = MI.getOperand(7).getImm();
  DebugLoc DL = MI.getDebugLoc();

  const TargetRegisterClass *RC = &SystemZ::GR32BitRegClass;

  // Get the right opcodes for the displacement.
  unsigned LOpcode  = TII->getOpcodeForOffset(SystemZ::L,  Disp);
  unsigned CSOpcode = TII->getOpcodeForOffset(SystemZ::CS, Disp);
  assert(LOpcode && CSOpcode && "Displacement out of range");

  // Create virtual registers for temporary results.
  Register OrigOldVal = MRI.createVirtualRegister(RC);
  Register OldVal = MRI.createVirtualRegister(RC);
  Register CmpVal = MRI.createVirtualRegister(RC);
  Register SwapVal = MRI.createVirtualRegister(RC);
  Register StoreVal = MRI.createVirtualRegister(RC);
  Register RetryOldVal = MRI.createVirtualRegister(RC);
  Register RetryCmpVal = MRI.createVirtualRegister(RC);
  Register RetrySwapVal = MRI.createVirtualRegister(RC);

  // Insert 2 basic blocks for the loop.
  MachineBasicBlock *StartMBB = MBB;
  MachineBasicBlock *DoneMBB  = SystemZ::splitBlockBefore(MI, MBB);
  MachineBasicBlock *LoopMBB  = SystemZ::emitBlockAfter(StartMBB);
  MachineBasicBlock *SetMBB   = SystemZ::emitBlockAfter(LoopMBB);

  //  StartMBB:
  //   ...
  //   %OrigOldVal     = L Disp(%Base)
  //   # fall through to LoopMMB
  MBB = StartMBB;
  BuildMI(MBB, DL, TII->get(LOpcode), OrigOldVal)
      .add(Base)
      .addImm(Disp)
      .addReg(0);
  MBB->addSuccessor(LoopMBB);

  //  LoopMBB:
  //   %OldVal        = phi [ %OrigOldVal, EntryBB ], [ %RetryOldVal, SetMBB ]
  //   %CmpVal        = phi [ %OrigCmpVal, EntryBB ], [ %RetryCmpVal, SetMBB ]
  //   %SwapVal       = phi [ %OrigSwapVal, EntryBB ], [ %RetrySwapVal, SetMBB ]
  //   %Dest          = RLL %OldVal, BitSize(%BitShift)
  //                      ^^ The low BitSize bits contain the field
  //                         of interest.
  //   %RetryCmpVal   = RISBG32 %CmpVal, %Dest, 32, 63-BitSize, 0
  //                      ^^ Replace the upper 32-BitSize bits of the
  //                         comparison value with those that we loaded,
  //                         so that we can use a full word comparison.
  //   CR %Dest, %RetryCmpVal
  //   JNE DoneMBB
  //   # Fall through to SetMBB
  MBB = LoopMBB;
  BuildMI(MBB, DL, TII->get(SystemZ::PHI), OldVal)
    .addReg(OrigOldVal).addMBB(StartMBB)
    .addReg(RetryOldVal).addMBB(SetMBB);
  BuildMI(MBB, DL, TII->get(SystemZ::PHI), CmpVal)
    .addReg(OrigCmpVal).addMBB(StartMBB)
    .addReg(RetryCmpVal).addMBB(SetMBB);
  BuildMI(MBB, DL, TII->get(SystemZ::PHI), SwapVal)
    .addReg(OrigSwapVal).addMBB(StartMBB)
    .addReg(RetrySwapVal).addMBB(SetMBB);
  BuildMI(MBB, DL, TII->get(SystemZ::RLL), Dest)
    .addReg(OldVal).addReg(BitShift).addImm(BitSize);
  BuildMI(MBB, DL, TII->get(SystemZ::RISBG32), RetryCmpVal)
    .addReg(CmpVal).addReg(Dest).addImm(32).addImm(63 - BitSize).addImm(0);
  BuildMI(MBB, DL, TII->get(SystemZ::CR))
    .addReg(Dest).addReg(RetryCmpVal);
  BuildMI(MBB, DL, TII->get(SystemZ::BRC))
    .addImm(SystemZ::CCMASK_ICMP)
    .addImm(SystemZ::CCMASK_CMP_NE).addMBB(DoneMBB);
  MBB->addSuccessor(DoneMBB);
  MBB->addSuccessor(SetMBB);

  //  SetMBB:
  //   %RetrySwapVal = RISBG32 %SwapVal, %Dest, 32, 63-BitSize, 0
  //                      ^^ Replace the upper 32-BitSize bits of the new
  //                         value with those that we loaded.
  //   %StoreVal    = RLL %RetrySwapVal, -BitSize(%NegBitShift)
  //                      ^^ Rotate the new field to its proper position.
  //   %RetryOldVal = CS %Dest, %StoreVal, Disp(%Base)
  //   JNE LoopMBB
  //   # fall through to ExitMMB
  MBB = SetMBB;
  BuildMI(MBB, DL, TII->get(SystemZ::RISBG32), RetrySwapVal)
    .addReg(SwapVal).addReg(Dest).addImm(32).addImm(63 - BitSize).addImm(0);
  BuildMI(MBB, DL, TII->get(SystemZ::RLL), StoreVal)
    .addReg(RetrySwapVal).addReg(NegBitShift).addImm(-BitSize);
  BuildMI(MBB, DL, TII->get(CSOpcode), RetryOldVal)
      .addReg(OldVal)
      .addReg(StoreVal)
      .add(Base)
      .addImm(Disp);
  BuildMI(MBB, DL, TII->get(SystemZ::BRC))
    .addImm(SystemZ::CCMASK_CS).addImm(SystemZ::CCMASK_CS_NE).addMBB(LoopMBB);
  MBB->addSuccessor(LoopMBB);
  MBB->addSuccessor(DoneMBB);

  // If the CC def wasn't dead in the ATOMIC_CMP_SWAPW, mark CC as live-in
  // to the block after the loop.  At this point, CC may have been defined
  // either by the CR in LoopMBB or by the CS in SetMBB.
  if (!MI.registerDefIsDead(SystemZ::CC))
    DoneMBB->addLiveIn(SystemZ::CC);

  MI.eraseFromParent();
  return DoneMBB;
}

// Emit a move from two GR64s to a GR128.
MachineBasicBlock *
SystemZTargetLowering::emitPair128(MachineInstr &MI,
                                   MachineBasicBlock *MBB) const {
  MachineFunction &MF = *MBB->getParent();
  const SystemZInstrInfo *TII =
      static_cast<const SystemZInstrInfo *>(Subtarget.getInstrInfo());
  MachineRegisterInfo &MRI = MF.getRegInfo();
  DebugLoc DL = MI.getDebugLoc();

  Register Dest = MI.getOperand(0).getReg();
  Register Hi = MI.getOperand(1).getReg();
  Register Lo = MI.getOperand(2).getReg();
  Register Tmp1 = MRI.createVirtualRegister(&SystemZ::GR128BitRegClass);
  Register Tmp2 = MRI.createVirtualRegister(&SystemZ::GR128BitRegClass);

  BuildMI(*MBB, MI, DL, TII->get(TargetOpcode::IMPLICIT_DEF), Tmp1);
  BuildMI(*MBB, MI, DL, TII->get(TargetOpcode::INSERT_SUBREG), Tmp2)
    .addReg(Tmp1).addReg(Hi).addImm(SystemZ::subreg_h64);
  BuildMI(*MBB, MI, DL, TII->get(TargetOpcode::INSERT_SUBREG), Dest)
    .addReg(Tmp2).addReg(Lo).addImm(SystemZ::subreg_l64);

  MI.eraseFromParent();
  return MBB;
}

// Emit an extension from a GR64 to a GR128.  ClearEven is true
// if the high register of the GR128 value must be cleared or false if
// it's "don't care".
MachineBasicBlock *SystemZTargetLowering::emitExt128(MachineInstr &MI,
                                                     MachineBasicBlock *MBB,
                                                     bool ClearEven) const {
  MachineFunction &MF = *MBB->getParent();
  const SystemZInstrInfo *TII =
      static_cast<const SystemZInstrInfo *>(Subtarget.getInstrInfo());
  MachineRegisterInfo &MRI = MF.getRegInfo();
  DebugLoc DL = MI.getDebugLoc();

  Register Dest = MI.getOperand(0).getReg();
  Register Src = MI.getOperand(1).getReg();
  Register In128 = MRI.createVirtualRegister(&SystemZ::GR128BitRegClass);

  BuildMI(*MBB, MI, DL, TII->get(TargetOpcode::IMPLICIT_DEF), In128);
  if (ClearEven) {
    Register NewIn128 = MRI.createVirtualRegister(&SystemZ::GR128BitRegClass);
    Register Zero64 = MRI.createVirtualRegister(&SystemZ::GR64BitRegClass);

    BuildMI(*MBB, MI, DL, TII->get(SystemZ::LLILL), Zero64)
      .addImm(0);
    BuildMI(*MBB, MI, DL, TII->get(TargetOpcode::INSERT_SUBREG), NewIn128)
      .addReg(In128).addReg(Zero64).addImm(SystemZ::subreg_h64);
    In128 = NewIn128;
  }
  BuildMI(*MBB, MI, DL, TII->get(TargetOpcode::INSERT_SUBREG), Dest)
    .addReg(In128).addReg(Src).addImm(SystemZ::subreg_l64);

  MI.eraseFromParent();
  return MBB;
}

MachineBasicBlock *SystemZTargetLowering::emitMemMemWrapper(
    MachineInstr &MI, MachineBasicBlock *MBB, unsigned Opcode) const {
  MachineFunction &MF = *MBB->getParent();
  const SystemZInstrInfo *TII =
      static_cast<const SystemZInstrInfo *>(Subtarget.getInstrInfo());
  MachineRegisterInfo &MRI = MF.getRegInfo();
  DebugLoc DL = MI.getDebugLoc();

  MachineOperand DestBase = earlyUseOperand(MI.getOperand(0));
  uint64_t DestDisp = MI.getOperand(1).getImm();
  MachineOperand SrcBase = earlyUseOperand(MI.getOperand(2));
  uint64_t SrcDisp = MI.getOperand(3).getImm();
  uint64_t Length = MI.getOperand(4).getImm();

  // When generating more than one CLC, all but the last will need to
  // branch to the end when a difference is found.
  MachineBasicBlock *EndMBB = (Length > 256 && Opcode == SystemZ::CLC ?
                               SystemZ::splitBlockAfter(MI, MBB) : nullptr);

  // Check for the loop form, in which operand 5 is the trip count.
  if (MI.getNumExplicitOperands() > 5) {
    bool HaveSingleBase = DestBase.isIdenticalTo(SrcBase);

    Register StartCountReg = MI.getOperand(5).getReg();
    Register StartSrcReg   = forceReg(MI, SrcBase, TII);
    Register StartDestReg  = (HaveSingleBase ? StartSrcReg :
                              forceReg(MI, DestBase, TII));

    const TargetRegisterClass *RC = &SystemZ::ADDR64BitRegClass;
    Register ThisSrcReg  = MRI.createVirtualRegister(RC);
    Register ThisDestReg = (HaveSingleBase ? ThisSrcReg :
                            MRI.createVirtualRegister(RC));
    Register NextSrcReg  = MRI.createVirtualRegister(RC);
    Register NextDestReg = (HaveSingleBase ? NextSrcReg :
                            MRI.createVirtualRegister(RC));

    RC = &SystemZ::GR64BitRegClass;
    Register ThisCountReg = MRI.createVirtualRegister(RC);
    Register NextCountReg = MRI.createVirtualRegister(RC);

    MachineBasicBlock *StartMBB = MBB;
    MachineBasicBlock *DoneMBB = SystemZ::splitBlockBefore(MI, MBB);
    MachineBasicBlock *LoopMBB = SystemZ::emitBlockAfter(StartMBB);
    MachineBasicBlock *NextMBB =
        (EndMBB ? SystemZ::emitBlockAfter(LoopMBB) : LoopMBB);

    //  StartMBB:
    //   # fall through to LoopMMB
    MBB->addSuccessor(LoopMBB);

    //  LoopMBB:
    //   %ThisDestReg = phi [ %StartDestReg, StartMBB ],
    //                      [ %NextDestReg, NextMBB ]
    //   %ThisSrcReg = phi [ %StartSrcReg, StartMBB ],
    //                     [ %NextSrcReg, NextMBB ]
    //   %ThisCountReg = phi [ %StartCountReg, StartMBB ],
    //                       [ %NextCountReg, NextMBB ]
    //   ( PFD 2, 768+DestDisp(%ThisDestReg) )
    //   Opcode DestDisp(256,%ThisDestReg), SrcDisp(%ThisSrcReg)
    //   ( JLH EndMBB )
    //
    // The prefetch is used only for MVC.  The JLH is used only for CLC.
    MBB = LoopMBB;

    BuildMI(MBB, DL, TII->get(SystemZ::PHI), ThisDestReg)
      .addReg(StartDestReg).addMBB(StartMBB)
      .addReg(NextDestReg).addMBB(NextMBB);
    if (!HaveSingleBase)
      BuildMI(MBB, DL, TII->get(SystemZ::PHI), ThisSrcReg)
        .addReg(StartSrcReg).addMBB(StartMBB)
        .addReg(NextSrcReg).addMBB(NextMBB);
    BuildMI(MBB, DL, TII->get(SystemZ::PHI), ThisCountReg)
      .addReg(StartCountReg).addMBB(StartMBB)
      .addReg(NextCountReg).addMBB(NextMBB);
    if (Opcode == SystemZ::MVC)
      BuildMI(MBB, DL, TII->get(SystemZ::PFD))
        .addImm(SystemZ::PFD_WRITE)
        .addReg(ThisDestReg).addImm(DestDisp + 768).addReg(0);
    BuildMI(MBB, DL, TII->get(Opcode))
      .addReg(ThisDestReg).addImm(DestDisp).addImm(256)
      .addReg(ThisSrcReg).addImm(SrcDisp);
    if (EndMBB) {
      BuildMI(MBB, DL, TII->get(SystemZ::BRC))
        .addImm(SystemZ::CCMASK_ICMP).addImm(SystemZ::CCMASK_CMP_NE)
        .addMBB(EndMBB);
      MBB->addSuccessor(EndMBB);
      MBB->addSuccessor(NextMBB);
    }

    // NextMBB:
    //   %NextDestReg = LA 256(%ThisDestReg)
    //   %NextSrcReg = LA 256(%ThisSrcReg)
    //   %NextCountReg = AGHI %ThisCountReg, -1
    //   CGHI %NextCountReg, 0
    //   JLH LoopMBB
    //   # fall through to DoneMMB
    //
    // The AGHI, CGHI and JLH should be converted to BRCTG by later passes.
    MBB = NextMBB;

    BuildMI(MBB, DL, TII->get(SystemZ::LA), NextDestReg)
      .addReg(ThisDestReg).addImm(256).addReg(0);
    if (!HaveSingleBase)
      BuildMI(MBB, DL, TII->get(SystemZ::LA), NextSrcReg)
        .addReg(ThisSrcReg).addImm(256).addReg(0);
    BuildMI(MBB, DL, TII->get(SystemZ::AGHI), NextCountReg)
      .addReg(ThisCountReg).addImm(-1);
    BuildMI(MBB, DL, TII->get(SystemZ::CGHI))
      .addReg(NextCountReg).addImm(0);
    BuildMI(MBB, DL, TII->get(SystemZ::BRC))
      .addImm(SystemZ::CCMASK_ICMP).addImm(SystemZ::CCMASK_CMP_NE)
      .addMBB(LoopMBB);
    MBB->addSuccessor(LoopMBB);
    MBB->addSuccessor(DoneMBB);

    DestBase = MachineOperand::CreateReg(NextDestReg, false);
    SrcBase = MachineOperand::CreateReg(NextSrcReg, false);
    Length &= 255;
    if (EndMBB && !Length)
      // If the loop handled the whole CLC range, DoneMBB will be empty with
      // CC live-through into EndMBB, so add it as live-in.
      DoneMBB->addLiveIn(SystemZ::CC);
    MBB = DoneMBB;
  }
  // Handle any remaining bytes with straight-line code.
  while (Length > 0) {
    uint64_t ThisLength = std::min(Length, uint64_t(256));
    // The previous iteration might have created out-of-range displacements.
    // Apply them using LAY if so.
    if (!isUInt<12>(DestDisp)) {
      Register Reg = MRI.createVirtualRegister(&SystemZ::ADDR64BitRegClass);
      BuildMI(*MBB, MI, MI.getDebugLoc(), TII->get(SystemZ::LAY), Reg)
          .add(DestBase)
          .addImm(DestDisp)
          .addReg(0);
      DestBase = MachineOperand::CreateReg(Reg, false);
      DestDisp = 0;
    }
    if (!isUInt<12>(SrcDisp)) {
      Register Reg = MRI.createVirtualRegister(&SystemZ::ADDR64BitRegClass);
      BuildMI(*MBB, MI, MI.getDebugLoc(), TII->get(SystemZ::LAY), Reg)
          .add(SrcBase)
          .addImm(SrcDisp)
          .addReg(0);
      SrcBase = MachineOperand::CreateReg(Reg, false);
      SrcDisp = 0;
    }
    BuildMI(*MBB, MI, DL, TII->get(Opcode))
        .add(DestBase)
        .addImm(DestDisp)
        .addImm(ThisLength)
        .add(SrcBase)
        .addImm(SrcDisp)
        .setMemRefs(MI.memoperands());
    DestDisp += ThisLength;
    SrcDisp += ThisLength;
    Length -= ThisLength;
    // If there's another CLC to go, branch to the end if a difference
    // was found.
    if (EndMBB && Length > 0) {
      MachineBasicBlock *NextMBB = SystemZ::splitBlockBefore(MI, MBB);
      BuildMI(MBB, DL, TII->get(SystemZ::BRC))
        .addImm(SystemZ::CCMASK_ICMP).addImm(SystemZ::CCMASK_CMP_NE)
        .addMBB(EndMBB);
      MBB->addSuccessor(EndMBB);
      MBB->addSuccessor(NextMBB);
      MBB = NextMBB;
    }
  }
  if (EndMBB) {
    MBB->addSuccessor(EndMBB);
    MBB = EndMBB;
    MBB->addLiveIn(SystemZ::CC);
  }

  MI.eraseFromParent();
  return MBB;
}

// Decompose string pseudo-instruction MI into a loop that continually performs
// Opcode until CC != 3.
MachineBasicBlock *SystemZTargetLowering::emitStringWrapper(
    MachineInstr &MI, MachineBasicBlock *MBB, unsigned Opcode) const {
  MachineFunction &MF = *MBB->getParent();
  const SystemZInstrInfo *TII =
      static_cast<const SystemZInstrInfo *>(Subtarget.getInstrInfo());
  MachineRegisterInfo &MRI = MF.getRegInfo();
  DebugLoc DL = MI.getDebugLoc();

  uint64_t End1Reg = MI.getOperand(0).getReg();
  uint64_t Start1Reg = MI.getOperand(1).getReg();
  uint64_t Start2Reg = MI.getOperand(2).getReg();
  uint64_t CharReg = MI.getOperand(3).getReg();

  const TargetRegisterClass *RC = &SystemZ::GR64BitRegClass;
  uint64_t This1Reg = MRI.createVirtualRegister(RC);
  uint64_t This2Reg = MRI.createVirtualRegister(RC);
  uint64_t End2Reg  = MRI.createVirtualRegister(RC);

  MachineBasicBlock *StartMBB = MBB;
  MachineBasicBlock *DoneMBB = SystemZ::splitBlockBefore(MI, MBB);
  MachineBasicBlock *LoopMBB = SystemZ::emitBlockAfter(StartMBB);

  //  StartMBB:
  //   # fall through to LoopMMB
  MBB->addSuccessor(LoopMBB);

  //  LoopMBB:
  //   %This1Reg = phi [ %Start1Reg, StartMBB ], [ %End1Reg, LoopMBB ]
  //   %This2Reg = phi [ %Start2Reg, StartMBB ], [ %End2Reg, LoopMBB ]
  //   R0L = %CharReg
  //   %End1Reg, %End2Reg = CLST %This1Reg, %This2Reg -- uses R0L
  //   JO LoopMBB
  //   # fall through to DoneMMB
  //
  // The load of R0L can be hoisted by post-RA LICM.
  MBB = LoopMBB;

  BuildMI(MBB, DL, TII->get(SystemZ::PHI), This1Reg)
    .addReg(Start1Reg).addMBB(StartMBB)
    .addReg(End1Reg).addMBB(LoopMBB);
  BuildMI(MBB, DL, TII->get(SystemZ::PHI), This2Reg)
    .addReg(Start2Reg).addMBB(StartMBB)
    .addReg(End2Reg).addMBB(LoopMBB);
  BuildMI(MBB, DL, TII->get(TargetOpcode::COPY), SystemZ::R0L).addReg(CharReg);
  BuildMI(MBB, DL, TII->get(Opcode))
    .addReg(End1Reg, RegState::Define).addReg(End2Reg, RegState::Define)
    .addReg(This1Reg).addReg(This2Reg);
  BuildMI(MBB, DL, TII->get(SystemZ::BRC))
    .addImm(SystemZ::CCMASK_ANY).addImm(SystemZ::CCMASK_3).addMBB(LoopMBB);
  MBB->addSuccessor(LoopMBB);
  MBB->addSuccessor(DoneMBB);

  DoneMBB->addLiveIn(SystemZ::CC);

  MI.eraseFromParent();
  return DoneMBB;
}

// Update TBEGIN instruction with final opcode and register clobbers.
MachineBasicBlock *SystemZTargetLowering::emitTransactionBegin(
    MachineInstr &MI, MachineBasicBlock *MBB, unsigned Opcode,
    bool NoFloat) const {
  MachineFunction &MF = *MBB->getParent();
  const TargetFrameLowering *TFI = Subtarget.getFrameLowering();
  const SystemZInstrInfo *TII = Subtarget.getInstrInfo();

  // Update opcode.
  MI.setDesc(TII->get(Opcode));

  // We cannot handle a TBEGIN that clobbers the stack or frame pointer.
  // Make sure to add the corresponding GRSM bits if they are missing.
  uint64_t Control = MI.getOperand(2).getImm();
  static const unsigned GPRControlBit[16] = {
    0x8000, 0x8000, 0x4000, 0x4000, 0x2000, 0x2000, 0x1000, 0x1000,
    0x0800, 0x0800, 0x0400, 0x0400, 0x0200, 0x0200, 0x0100, 0x0100
  };
  Control |= GPRControlBit[15];
  if (TFI->hasFP(MF))
    Control |= GPRControlBit[11];
  MI.getOperand(2).setImm(Control);

  // Add GPR clobbers.
  for (int I = 0; I < 16; I++) {
    if ((Control & GPRControlBit[I]) == 0) {
      unsigned Reg = SystemZMC::GR64Regs[I];
      MI.addOperand(MachineOperand::CreateReg(Reg, true, true));
    }
  }

  // Add FPR/VR clobbers.
  if (!NoFloat && (Control & 4) != 0) {
    if (Subtarget.hasVector()) {
      for (int I = 0; I < 32; I++) {
        unsigned Reg = SystemZMC::VR128Regs[I];
        MI.addOperand(MachineOperand::CreateReg(Reg, true, true));
      }
    } else {
      for (int I = 0; I < 16; I++) {
        unsigned Reg = SystemZMC::FP64Regs[I];
        MI.addOperand(MachineOperand::CreateReg(Reg, true, true));
      }
    }
  }

  return MBB;
}

MachineBasicBlock *SystemZTargetLowering::emitLoadAndTestCmp0(
    MachineInstr &MI, MachineBasicBlock *MBB, unsigned Opcode) const {
  MachineFunction &MF = *MBB->getParent();
  MachineRegisterInfo *MRI = &MF.getRegInfo();
  const SystemZInstrInfo *TII =
      static_cast<const SystemZInstrInfo *>(Subtarget.getInstrInfo());
  DebugLoc DL = MI.getDebugLoc();

  Register SrcReg = MI.getOperand(0).getReg();

  // Create new virtual register of the same class as source.
  const TargetRegisterClass *RC = MRI->getRegClass(SrcReg);
  Register DstReg = MRI->createVirtualRegister(RC);

  // Replace pseudo with a normal load-and-test that models the def as
  // well.
  BuildMI(*MBB, MI, DL, TII->get(Opcode), DstReg)
    .addReg(SrcReg)
    .setMIFlags(MI.getFlags());
  MI.eraseFromParent();

  return MBB;
}

MachineBasicBlock *SystemZTargetLowering::emitProbedAlloca(
    MachineInstr &MI, MachineBasicBlock *MBB) const {
  MachineFunction &MF = *MBB->getParent();
  MachineRegisterInfo *MRI = &MF.getRegInfo();
  const SystemZInstrInfo *TII =
      static_cast<const SystemZInstrInfo *>(Subtarget.getInstrInfo());
  DebugLoc DL = MI.getDebugLoc();
  const unsigned ProbeSize = getStackProbeSize(MF);
  Register DstReg = MI.getOperand(0).getReg();
  Register SizeReg = MI.getOperand(2).getReg();

  MachineBasicBlock *StartMBB = MBB;
  MachineBasicBlock *DoneMBB  = SystemZ::splitBlockAfter(MI, MBB);
  MachineBasicBlock *LoopTestMBB  = SystemZ::emitBlockAfter(StartMBB);
  MachineBasicBlock *LoopBodyMBB = SystemZ::emitBlockAfter(LoopTestMBB);
  MachineBasicBlock *TailTestMBB = SystemZ::emitBlockAfter(LoopBodyMBB);
  MachineBasicBlock *TailMBB = SystemZ::emitBlockAfter(TailTestMBB);

  MachineMemOperand *VolLdMMO = MF.getMachineMemOperand(MachinePointerInfo(),
    MachineMemOperand::MOVolatile | MachineMemOperand::MOLoad, 8, Align(1));

  Register PHIReg = MRI->createVirtualRegister(&SystemZ::ADDR64BitRegClass);
  Register IncReg = MRI->createVirtualRegister(&SystemZ::ADDR64BitRegClass);

  //  LoopTestMBB
  //  BRC TailTestMBB
  //  # fallthrough to LoopBodyMBB
  StartMBB->addSuccessor(LoopTestMBB);
  MBB = LoopTestMBB;
  BuildMI(MBB, DL, TII->get(SystemZ::PHI), PHIReg)
    .addReg(SizeReg)
    .addMBB(StartMBB)
    .addReg(IncReg)
    .addMBB(LoopBodyMBB);
  BuildMI(MBB, DL, TII->get(SystemZ::CLGFI))
    .addReg(PHIReg)
    .addImm(ProbeSize);
  BuildMI(MBB, DL, TII->get(SystemZ::BRC))
    .addImm(SystemZ::CCMASK_ICMP).addImm(SystemZ::CCMASK_CMP_LT)
    .addMBB(TailTestMBB);
  MBB->addSuccessor(LoopBodyMBB);
  MBB->addSuccessor(TailTestMBB);

  //  LoopBodyMBB: Allocate and probe by means of a volatile compare.
  //  J LoopTestMBB
  MBB = LoopBodyMBB;
  BuildMI(MBB, DL, TII->get(SystemZ::SLGFI), IncReg)
    .addReg(PHIReg)
    .addImm(ProbeSize);
  BuildMI(MBB, DL, TII->get(SystemZ::SLGFI), SystemZ::R15D)
    .addReg(SystemZ::R15D)
    .addImm(ProbeSize);
  BuildMI(MBB, DL, TII->get(SystemZ::CG)).addReg(SystemZ::R15D)
    .addReg(SystemZ::R15D).addImm(ProbeSize - 8).addReg(0)
    .setMemRefs(VolLdMMO);
  BuildMI(MBB, DL, TII->get(SystemZ::J)).addMBB(LoopTestMBB);
  MBB->addSuccessor(LoopTestMBB);

  //  TailTestMBB
  //  BRC DoneMBB
  //  # fallthrough to TailMBB
  MBB = TailTestMBB;
  BuildMI(MBB, DL, TII->get(SystemZ::CGHI))
    .addReg(PHIReg)
    .addImm(0);
  BuildMI(MBB, DL, TII->get(SystemZ::BRC))
    .addImm(SystemZ::CCMASK_ICMP).addImm(SystemZ::CCMASK_CMP_EQ)
    .addMBB(DoneMBB);
  MBB->addSuccessor(TailMBB);
  MBB->addSuccessor(DoneMBB);

  //  TailMBB
  //  # fallthrough to DoneMBB
  MBB = TailMBB;
  BuildMI(MBB, DL, TII->get(SystemZ::SLGR), SystemZ::R15D)
    .addReg(SystemZ::R15D)
    .addReg(PHIReg);
  BuildMI(MBB, DL, TII->get(SystemZ::CG)).addReg(SystemZ::R15D)
    .addReg(SystemZ::R15D).addImm(-8).addReg(PHIReg)
    .setMemRefs(VolLdMMO);
  MBB->addSuccessor(DoneMBB);

  //  DoneMBB
  MBB = DoneMBB;
  BuildMI(*MBB, MBB->begin(), DL, TII->get(TargetOpcode::COPY), DstReg)
    .addReg(SystemZ::R15D);

  MI.eraseFromParent();
  return DoneMBB;
}

MachineBasicBlock *SystemZTargetLowering::EmitInstrWithCustomInserter(
    MachineInstr &MI, MachineBasicBlock *MBB) const {
  switch (MI.getOpcode()) {
  case SystemZ::Select32:
  case SystemZ::Select64:
  case SystemZ::SelectF32:
  case SystemZ::SelectF64:
  case SystemZ::SelectF128:
  case SystemZ::SelectVR32:
  case SystemZ::SelectVR64:
  case SystemZ::SelectVR128:
    return emitSelect(MI, MBB);

  case SystemZ::CondStore8Mux:
    return emitCondStore(MI, MBB, SystemZ::STCMux, 0, false);
  case SystemZ::CondStore8MuxInv:
    return emitCondStore(MI, MBB, SystemZ::STCMux, 0, true);
  case SystemZ::CondStore16Mux:
    return emitCondStore(MI, MBB, SystemZ::STHMux, 0, false);
  case SystemZ::CondStore16MuxInv:
    return emitCondStore(MI, MBB, SystemZ::STHMux, 0, true);
  case SystemZ::CondStore32Mux:
    return emitCondStore(MI, MBB, SystemZ::STMux, SystemZ::STOCMux, false);
  case SystemZ::CondStore32MuxInv:
    return emitCondStore(MI, MBB, SystemZ::STMux, SystemZ::STOCMux, true);
  case SystemZ::CondStore8:
    return emitCondStore(MI, MBB, SystemZ::STC, 0, false);
  case SystemZ::CondStore8Inv:
    return emitCondStore(MI, MBB, SystemZ::STC, 0, true);
  case SystemZ::CondStore16:
    return emitCondStore(MI, MBB, SystemZ::STH, 0, false);
  case SystemZ::CondStore16Inv:
    return emitCondStore(MI, MBB, SystemZ::STH, 0, true);
  case SystemZ::CondStore32:
    return emitCondStore(MI, MBB, SystemZ::ST, SystemZ::STOC, false);
  case SystemZ::CondStore32Inv:
    return emitCondStore(MI, MBB, SystemZ::ST, SystemZ::STOC, true);
  case SystemZ::CondStore64:
    return emitCondStore(MI, MBB, SystemZ::STG, SystemZ::STOCG, false);
  case SystemZ::CondStore64Inv:
    return emitCondStore(MI, MBB, SystemZ::STG, SystemZ::STOCG, true);
  case SystemZ::CondStoreF32:
    return emitCondStore(MI, MBB, SystemZ::STE, 0, false);
  case SystemZ::CondStoreF32Inv:
    return emitCondStore(MI, MBB, SystemZ::STE, 0, true);
  case SystemZ::CondStoreF64:
    return emitCondStore(MI, MBB, SystemZ::STD, 0, false);
  case SystemZ::CondStoreF64Inv:
    return emitCondStore(MI, MBB, SystemZ::STD, 0, true);

  case SystemZ::PAIR128:
    return emitPair128(MI, MBB);
  case SystemZ::AEXT128:
    return emitExt128(MI, MBB, false);
  case SystemZ::ZEXT128:
    return emitExt128(MI, MBB, true);

  case SystemZ::ATOMIC_SWAPW:
    return emitAtomicLoadBinary(MI, MBB, 0, 0);
  case SystemZ::ATOMIC_SWAP_32:
    return emitAtomicLoadBinary(MI, MBB, 0, 32);
  case SystemZ::ATOMIC_SWAP_64:
    return emitAtomicLoadBinary(MI, MBB, 0, 64);

  case SystemZ::ATOMIC_LOADW_AR:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::AR, 0);
  case SystemZ::ATOMIC_LOADW_AFI:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::AFI, 0);
  case SystemZ::ATOMIC_LOAD_AR:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::AR, 32);
  case SystemZ::ATOMIC_LOAD_AHI:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::AHI, 32);
  case SystemZ::ATOMIC_LOAD_AFI:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::AFI, 32);
  case SystemZ::ATOMIC_LOAD_AGR:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::AGR, 64);
  case SystemZ::ATOMIC_LOAD_AGHI:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::AGHI, 64);
  case SystemZ::ATOMIC_LOAD_AGFI:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::AGFI, 64);

  case SystemZ::ATOMIC_LOADW_SR:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::SR, 0);
  case SystemZ::ATOMIC_LOAD_SR:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::SR, 32);
  case SystemZ::ATOMIC_LOAD_SGR:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::SGR, 64);

  case SystemZ::ATOMIC_LOADW_NR:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::NR, 0);
  case SystemZ::ATOMIC_LOADW_NILH:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::NILH, 0);
  case SystemZ::ATOMIC_LOAD_NR:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::NR, 32);
  case SystemZ::ATOMIC_LOAD_NILL:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::NILL, 32);
  case SystemZ::ATOMIC_LOAD_NILH:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::NILH, 32);
  case SystemZ::ATOMIC_LOAD_NILF:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::NILF, 32);
  case SystemZ::ATOMIC_LOAD_NGR:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::NGR, 64);
  case SystemZ::ATOMIC_LOAD_NILL64:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::NILL64, 64);
  case SystemZ::ATOMIC_LOAD_NILH64:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::NILH64, 64);
  case SystemZ::ATOMIC_LOAD_NIHL64:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::NIHL64, 64);
  case SystemZ::ATOMIC_LOAD_NIHH64:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::NIHH64, 64);
  case SystemZ::ATOMIC_LOAD_NILF64:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::NILF64, 64);
  case SystemZ::ATOMIC_LOAD_NIHF64:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::NIHF64, 64);

  case SystemZ::ATOMIC_LOADW_OR:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::OR, 0);
  case SystemZ::ATOMIC_LOADW_OILH:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::OILH, 0);
  case SystemZ::ATOMIC_LOAD_OR:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::OR, 32);
  case SystemZ::ATOMIC_LOAD_OILL:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::OILL, 32);
  case SystemZ::ATOMIC_LOAD_OILH:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::OILH, 32);
  case SystemZ::ATOMIC_LOAD_OILF:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::OILF, 32);
  case SystemZ::ATOMIC_LOAD_OGR:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::OGR, 64);
  case SystemZ::ATOMIC_LOAD_OILL64:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::OILL64, 64);
  case SystemZ::ATOMIC_LOAD_OILH64:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::OILH64, 64);
  case SystemZ::ATOMIC_LOAD_OIHL64:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::OIHL64, 64);
  case SystemZ::ATOMIC_LOAD_OIHH64:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::OIHH64, 64);
  case SystemZ::ATOMIC_LOAD_OILF64:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::OILF64, 64);
  case SystemZ::ATOMIC_LOAD_OIHF64:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::OIHF64, 64);

  case SystemZ::ATOMIC_LOADW_XR:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::XR, 0);
  case SystemZ::ATOMIC_LOADW_XILF:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::XILF, 0);
  case SystemZ::ATOMIC_LOAD_XR:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::XR, 32);
  case SystemZ::ATOMIC_LOAD_XILF:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::XILF, 32);
  case SystemZ::ATOMIC_LOAD_XGR:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::XGR, 64);
  case SystemZ::ATOMIC_LOAD_XILF64:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::XILF64, 64);
  case SystemZ::ATOMIC_LOAD_XIHF64:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::XIHF64, 64);

  case SystemZ::ATOMIC_LOADW_NRi:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::NR, 0, true);
  case SystemZ::ATOMIC_LOADW_NILHi:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::NILH, 0, true);
  case SystemZ::ATOMIC_LOAD_NRi:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::NR, 32, true);
  case SystemZ::ATOMIC_LOAD_NILLi:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::NILL, 32, true);
  case SystemZ::ATOMIC_LOAD_NILHi:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::NILH, 32, true);
  case SystemZ::ATOMIC_LOAD_NILFi:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::NILF, 32, true);
  case SystemZ::ATOMIC_LOAD_NGRi:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::NGR, 64, true);
  case SystemZ::ATOMIC_LOAD_NILL64i:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::NILL64, 64, true);
  case SystemZ::ATOMIC_LOAD_NILH64i:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::NILH64, 64, true);
  case SystemZ::ATOMIC_LOAD_NIHL64i:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::NIHL64, 64, true);
  case SystemZ::ATOMIC_LOAD_NIHH64i:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::NIHH64, 64, true);
  case SystemZ::ATOMIC_LOAD_NILF64i:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::NILF64, 64, true);
  case SystemZ::ATOMIC_LOAD_NIHF64i:
    return emitAtomicLoadBinary(MI, MBB, SystemZ::NIHF64, 64, true);

  case SystemZ::ATOMIC_LOADW_MIN:
    return emitAtomicLoadMinMax(MI, MBB, SystemZ::CR,
                                SystemZ::CCMASK_CMP_LE, 0);
  case SystemZ::ATOMIC_LOAD_MIN_32:
    return emitAtomicLoadMinMax(MI, MBB, SystemZ::CR,
                                SystemZ::CCMASK_CMP_LE, 32);
  case SystemZ::ATOMIC_LOAD_MIN_64:
    return emitAtomicLoadMinMax(MI, MBB, SystemZ::CGR,
                                SystemZ::CCMASK_CMP_LE, 64);

  case SystemZ::ATOMIC_LOADW_MAX:
    return emitAtomicLoadMinMax(MI, MBB, SystemZ::CR,
                                SystemZ::CCMASK_CMP_GE, 0);
  case SystemZ::ATOMIC_LOAD_MAX_32:
    return emitAtomicLoadMinMax(MI, MBB, SystemZ::CR,
                                SystemZ::CCMASK_CMP_GE, 32);
  case SystemZ::ATOMIC_LOAD_MAX_64:
    return emitAtomicLoadMinMax(MI, MBB, SystemZ::CGR,
                                SystemZ::CCMASK_CMP_GE, 64);

  case SystemZ::ATOMIC_LOADW_UMIN:
    return emitAtomicLoadMinMax(MI, MBB, SystemZ::CLR,
                                SystemZ::CCMASK_CMP_LE, 0);
  case SystemZ::ATOMIC_LOAD_UMIN_32:
    return emitAtomicLoadMinMax(MI, MBB, SystemZ::CLR,
                                SystemZ::CCMASK_CMP_LE, 32);
  case SystemZ::ATOMIC_LOAD_UMIN_64:
    return emitAtomicLoadMinMax(MI, MBB, SystemZ::CLGR,
                                SystemZ::CCMASK_CMP_LE, 64);

  case SystemZ::ATOMIC_LOADW_UMAX:
    return emitAtomicLoadMinMax(MI, MBB, SystemZ::CLR,
                                SystemZ::CCMASK_CMP_GE, 0);
  case SystemZ::ATOMIC_LOAD_UMAX_32:
    return emitAtomicLoadMinMax(MI, MBB, SystemZ::CLR,
                                SystemZ::CCMASK_CMP_GE, 32);
  case SystemZ::ATOMIC_LOAD_UMAX_64:
    return emitAtomicLoadMinMax(MI, MBB, SystemZ::CLGR,
                                SystemZ::CCMASK_CMP_GE, 64);

  case SystemZ::ATOMIC_CMP_SWAPW:
    return emitAtomicCmpSwapW(MI, MBB);
  case SystemZ::MVCSequence:
  case SystemZ::MVCLoop:
    return emitMemMemWrapper(MI, MBB, SystemZ::MVC);
  case SystemZ::NCSequence:
  case SystemZ::NCLoop:
    return emitMemMemWrapper(MI, MBB, SystemZ::NC);
  case SystemZ::OCSequence:
  case SystemZ::OCLoop:
    return emitMemMemWrapper(MI, MBB, SystemZ::OC);
  case SystemZ::XCSequence:
  case SystemZ::XCLoop:
    return emitMemMemWrapper(MI, MBB, SystemZ::XC);
  case SystemZ::CLCSequence:
  case SystemZ::CLCLoop:
    return emitMemMemWrapper(MI, MBB, SystemZ::CLC);
  case SystemZ::CLSTLoop:
    return emitStringWrapper(MI, MBB, SystemZ::CLST);
  case SystemZ::MVSTLoop:
    return emitStringWrapper(MI, MBB, SystemZ::MVST);
  case SystemZ::SRSTLoop:
    return emitStringWrapper(MI, MBB, SystemZ::SRST);
  case SystemZ::TBEGIN:
    return emitTransactionBegin(MI, MBB, SystemZ::TBEGIN, false);
  case SystemZ::TBEGIN_nofloat:
    return emitTransactionBegin(MI, MBB, SystemZ::TBEGIN, true);
  case SystemZ::TBEGINC:
    return emitTransactionBegin(MI, MBB, SystemZ::TBEGINC, true);
  case SystemZ::LTEBRCompare_VecPseudo:
    return emitLoadAndTestCmp0(MI, MBB, SystemZ::LTEBR);
  case SystemZ::LTDBRCompare_VecPseudo:
    return emitLoadAndTestCmp0(MI, MBB, SystemZ::LTDBR);
  case SystemZ::LTXBRCompare_VecPseudo:
    return emitLoadAndTestCmp0(MI, MBB, SystemZ::LTXBR);

  case SystemZ::PROBED_ALLOCA:
    return emitProbedAlloca(MI, MBB);

  case TargetOpcode::STACKMAP:
  case TargetOpcode::PATCHPOINT:
    return emitPatchPoint(MI, MBB);

  default:
    llvm_unreachable("Unexpected instr type to insert");
  }
}

// This is only used by the isel schedulers, and is needed only to prevent
// compiler from crashing when list-ilp is used.
const TargetRegisterClass *
SystemZTargetLowering::getRepRegClassFor(MVT VT) const {
  if (VT == MVT::Untyped)
    return &SystemZ::ADDR128BitRegClass;
  return TargetLowering::getRepRegClassFor(VT);
}