aboutsummaryrefslogtreecommitdiffstats
path: root/lib/CodeGen/SelectionDAG/FastISel.cpp
blob: 223a149924b19d87b4edd663d296e375f3f0d3da (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
//===-- FastISel.cpp - Implementation of the FastISel class ---------------===//
//
//                     The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file contains the implementation of the FastISel class.
//
// "Fast" instruction selection is designed to emit very poor code quickly.
// Also, it is not designed to be able to do much lowering, so most illegal
// types (e.g. i64 on 32-bit targets) and operations are not supported.  It is
// also not intended to be able to do much optimization, except in a few cases
// where doing optimizations reduces overall compile time.  For example, folding
// constants into immediate fields is often done, because it's cheap and it
// reduces the number of instructions later phases have to examine.
//
// "Fast" instruction selection is able to fail gracefully and transfer
// control to the SelectionDAG selector for operations that it doesn't
// support.  In many cases, this allows us to avoid duplicating a lot of
// the complicated lowering logic that SelectionDAG currently has.
//
// The intended use for "fast" instruction selection is "-O0" mode
// compilation, where the quality of the generated code is irrelevant when
// weighed against the speed at which the code can be generated.  Also,
// at -O0, the LLVM optimizers are not running, and this makes the
// compile time of codegen a much higher portion of the overall compile
// time.  Despite its limitations, "fast" instruction selection is able to
// handle enough code on its own to provide noticeable overall speedups
// in -O0 compiles.
//
// Basic operations are supported in a target-independent way, by reading
// the same instruction descriptions that the SelectionDAG selector reads,
// and identifying simple arithmetic operations that can be directly selected
// from simple operators.  More complicated operations currently require
// target-specific code.
//
//===----------------------------------------------------------------------===//

#include "llvm/CodeGen/Analysis.h"
#include "llvm/ADT/Optional.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/Analysis/BranchProbabilityInfo.h"
#include "llvm/Analysis/Loads.h"
#include "llvm/Analysis/TargetLibraryInfo.h"
#include "llvm/CodeGen/Analysis.h"
#include "llvm/CodeGen/FastISel.h"
#include "llvm/CodeGen/FunctionLoweringInfo.h"
#include "llvm/CodeGen/MachineFrameInfo.h"
#include "llvm/CodeGen/MachineInstrBuilder.h"
#include "llvm/CodeGen/MachineModuleInfo.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
#include "llvm/CodeGen/StackMaps.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/IR/DebugInfo.h"
#include "llvm/IR/Function.h"
#include "llvm/IR/GlobalVariable.h"
#include "llvm/IR/Instructions.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/IR/Operator.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Target/TargetInstrInfo.h"
#include "llvm/Target/TargetLowering.h"
#include "llvm/Target/TargetMachine.h"
#include "llvm/Target/TargetSubtargetInfo.h"
using namespace llvm;

#define DEBUG_TYPE "isel"

STATISTIC(NumFastIselSuccessIndependent, "Number of insts selected by "
                                         "target-independent selector");
STATISTIC(NumFastIselSuccessTarget, "Number of insts selected by "
                                    "target-specific selector");
STATISTIC(NumFastIselDead, "Number of dead insts removed on failure");

void FastISel::ArgListEntry::setAttributes(ImmutableCallSite *CS,
                                           unsigned AttrIdx) {
  IsSExt = CS->paramHasAttr(AttrIdx, Attribute::SExt);
  IsZExt = CS->paramHasAttr(AttrIdx, Attribute::ZExt);
  IsInReg = CS->paramHasAttr(AttrIdx, Attribute::InReg);
  IsSRet = CS->paramHasAttr(AttrIdx, Attribute::StructRet);
  IsNest = CS->paramHasAttr(AttrIdx, Attribute::Nest);
  IsByVal = CS->paramHasAttr(AttrIdx, Attribute::ByVal);
  IsInAlloca = CS->paramHasAttr(AttrIdx, Attribute::InAlloca);
  IsReturned = CS->paramHasAttr(AttrIdx, Attribute::Returned);
  Alignment = CS->getParamAlignment(AttrIdx);
}

/// Set the current block to which generated machine instructions will be
/// appended, and clear the local CSE map.
void FastISel::startNewBlock() {
  LocalValueMap.clear();

  // Instructions are appended to FuncInfo.MBB. If the basic block already
  // contains labels or copies, use the last instruction as the last local
  // value.
  EmitStartPt = nullptr;
  if (!FuncInfo.MBB->empty())
    EmitStartPt = &FuncInfo.MBB->back();
  LastLocalValue = EmitStartPt;
}

bool FastISel::lowerArguments() {
  if (!FuncInfo.CanLowerReturn)
    // Fallback to SDISel argument lowering code to deal with sret pointer
    // parameter.
    return false;

  if (!fastLowerArguments())
    return false;

  // Enter arguments into ValueMap for uses in non-entry BBs.
  for (Function::const_arg_iterator I = FuncInfo.Fn->arg_begin(),
                                    E = FuncInfo.Fn->arg_end();
       I != E; ++I) {
    DenseMap<const Value *, unsigned>::iterator VI = LocalValueMap.find(I);
    assert(VI != LocalValueMap.end() && "Missed an argument?");
    FuncInfo.ValueMap[I] = VI->second;
  }
  return true;
}

void FastISel::flushLocalValueMap() {
  LocalValueMap.clear();
  LastLocalValue = EmitStartPt;
  recomputeInsertPt();
  SavedInsertPt = FuncInfo.InsertPt;
}

bool FastISel::hasTrivialKill(const Value *V) {
  // Don't consider constants or arguments to have trivial kills.
  const Instruction *I = dyn_cast<Instruction>(V);
  if (!I)
    return false;

  // No-op casts are trivially coalesced by fast-isel.
  if (const auto *Cast = dyn_cast<CastInst>(I))
    if (Cast->isNoopCast(DL.getIntPtrType(Cast->getContext())) &&
        !hasTrivialKill(Cast->getOperand(0)))
      return false;

  // Even the value might have only one use in the LLVM IR, it is possible that
  // FastISel might fold the use into another instruction and now there is more
  // than one use at the Machine Instruction level.
  unsigned Reg = lookUpRegForValue(V);
  if (Reg && !MRI.use_empty(Reg))
    return false;

  // GEPs with all zero indices are trivially coalesced by fast-isel.
  if (const auto *GEP = dyn_cast<GetElementPtrInst>(I))
    if (GEP->hasAllZeroIndices() && !hasTrivialKill(GEP->getOperand(0)))
      return false;

  // Only instructions with a single use in the same basic block are considered
  // to have trivial kills.
  return I->hasOneUse() &&
         !(I->getOpcode() == Instruction::BitCast ||
           I->getOpcode() == Instruction::PtrToInt ||
           I->getOpcode() == Instruction::IntToPtr) &&
         cast<Instruction>(*I->user_begin())->getParent() == I->getParent();
}

unsigned FastISel::getRegForValue(const Value *V) {
  EVT RealVT = TLI.getValueType(V->getType(), /*AllowUnknown=*/true);
  // Don't handle non-simple values in FastISel.
  if (!RealVT.isSimple())
    return 0;

  // Ignore illegal types. We must do this before looking up the value
  // in ValueMap because Arguments are given virtual registers regardless
  // of whether FastISel can handle them.
  MVT VT = RealVT.getSimpleVT();
  if (!TLI.isTypeLegal(VT)) {
    // Handle integer promotions, though, because they're common and easy.
    if (VT == MVT::i1 || VT == MVT::i8 || VT == MVT::i16)
      VT = TLI.getTypeToTransformTo(V->getContext(), VT).getSimpleVT();
    else
      return 0;
  }

  // Look up the value to see if we already have a register for it.
  unsigned Reg = lookUpRegForValue(V);
  if (Reg)
    return Reg;

  // In bottom-up mode, just create the virtual register which will be used
  // to hold the value. It will be materialized later.
  if (isa<Instruction>(V) &&
      (!isa<AllocaInst>(V) ||
       !FuncInfo.StaticAllocaMap.count(cast<AllocaInst>(V))))
    return FuncInfo.InitializeRegForValue(V);

  SavePoint SaveInsertPt = enterLocalValueArea();

  // Materialize the value in a register. Emit any instructions in the
  // local value area.
  Reg = materializeRegForValue(V, VT);

  leaveLocalValueArea(SaveInsertPt);

  return Reg;
}

unsigned FastISel::materializeConstant(const Value *V, MVT VT) {
  unsigned Reg = 0;
  if (const auto *CI = dyn_cast<ConstantInt>(V)) {
    if (CI->getValue().getActiveBits() <= 64)
      Reg = fastEmit_i(VT, VT, ISD::Constant, CI->getZExtValue());
  } else if (isa<AllocaInst>(V))
    Reg = fastMaterializeAlloca(cast<AllocaInst>(V));
  else if (isa<ConstantPointerNull>(V))
    // Translate this as an integer zero so that it can be
    // local-CSE'd with actual integer zeros.
    Reg = getRegForValue(
        Constant::getNullValue(DL.getIntPtrType(V->getContext())));
  else if (const auto *CF = dyn_cast<ConstantFP>(V)) {
    if (CF->isNullValue())
      Reg = fastMaterializeFloatZero(CF);
    else
      // Try to emit the constant directly.
      Reg = fastEmit_f(VT, VT, ISD::ConstantFP, CF);

    if (!Reg) {
      // Try to emit the constant by using an integer constant with a cast.
      const APFloat &Flt = CF->getValueAPF();
      EVT IntVT = TLI.getPointerTy();

      uint64_t x[2];
      uint32_t IntBitWidth = IntVT.getSizeInBits();
      bool isExact;
      (void)Flt.convertToInteger(x, IntBitWidth, /*isSigned=*/true,
                                 APFloat::rmTowardZero, &isExact);
      if (isExact) {
        APInt IntVal(IntBitWidth, x);

        unsigned IntegerReg =
            getRegForValue(ConstantInt::get(V->getContext(), IntVal));
        if (IntegerReg != 0)
          Reg = fastEmit_r(IntVT.getSimpleVT(), VT, ISD::SINT_TO_FP, IntegerReg,
                           /*Kill=*/false);
      }
    }
  } else if (const auto *Op = dyn_cast<Operator>(V)) {
    if (!selectOperator(Op, Op->getOpcode()))
      if (!isa<Instruction>(Op) ||
          !fastSelectInstruction(cast<Instruction>(Op)))
        return 0;
    Reg = lookUpRegForValue(Op);
  } else if (isa<UndefValue>(V)) {
    Reg = createResultReg(TLI.getRegClassFor(VT));
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc,
            TII.get(TargetOpcode::IMPLICIT_DEF), Reg);
  }
  return Reg;
}

/// Helper for getRegForValue. This function is called when the value isn't
/// already available in a register and must be materialized with new
/// instructions.
unsigned FastISel::materializeRegForValue(const Value *V, MVT VT) {
  unsigned Reg = 0;
  // Give the target-specific code a try first.
  if (isa<Constant>(V))
    Reg = fastMaterializeConstant(cast<Constant>(V));

  // If target-specific code couldn't or didn't want to handle the value, then
  // give target-independent code a try.
  if (!Reg)
    Reg = materializeConstant(V, VT);

  // Don't cache constant materializations in the general ValueMap.
  // To do so would require tracking what uses they dominate.
  if (Reg) {
    LocalValueMap[V] = Reg;
    LastLocalValue = MRI.getVRegDef(Reg);
  }
  return Reg;
}

unsigned FastISel::lookUpRegForValue(const Value *V) {
  // Look up the value to see if we already have a register for it. We
  // cache values defined by Instructions across blocks, and other values
  // only locally. This is because Instructions already have the SSA
  // def-dominates-use requirement enforced.
  DenseMap<const Value *, unsigned>::iterator I = FuncInfo.ValueMap.find(V);
  if (I != FuncInfo.ValueMap.end())
    return I->second;
  return LocalValueMap[V];
}

void FastISel::updateValueMap(const Value *I, unsigned Reg, unsigned NumRegs) {
  if (!isa<Instruction>(I)) {
    LocalValueMap[I] = Reg;
    return;
  }

  unsigned &AssignedReg = FuncInfo.ValueMap[I];
  if (AssignedReg == 0)
    // Use the new register.
    AssignedReg = Reg;
  else if (Reg != AssignedReg) {
    // Arrange for uses of AssignedReg to be replaced by uses of Reg.
    for (unsigned i = 0; i < NumRegs; i++)
      FuncInfo.RegFixups[AssignedReg + i] = Reg + i;

    AssignedReg = Reg;
  }
}

std::pair<unsigned, bool> FastISel::getRegForGEPIndex(const Value *Idx) {
  unsigned IdxN = getRegForValue(Idx);
  if (IdxN == 0)
    // Unhandled operand. Halt "fast" selection and bail.
    return std::pair<unsigned, bool>(0, false);

  bool IdxNIsKill = hasTrivialKill(Idx);

  // If the index is smaller or larger than intptr_t, truncate or extend it.
  MVT PtrVT = TLI.getPointerTy();
  EVT IdxVT = EVT::getEVT(Idx->getType(), /*HandleUnknown=*/false);
  if (IdxVT.bitsLT(PtrVT)) {
    IdxN = fastEmit_r(IdxVT.getSimpleVT(), PtrVT, ISD::SIGN_EXTEND, IdxN,
                      IdxNIsKill);
    IdxNIsKill = true;
  } else if (IdxVT.bitsGT(PtrVT)) {
    IdxN =
        fastEmit_r(IdxVT.getSimpleVT(), PtrVT, ISD::TRUNCATE, IdxN, IdxNIsKill);
    IdxNIsKill = true;
  }
  return std::pair<unsigned, bool>(IdxN, IdxNIsKill);
}

void FastISel::recomputeInsertPt() {
  if (getLastLocalValue()) {
    FuncInfo.InsertPt = getLastLocalValue();
    FuncInfo.MBB = FuncInfo.InsertPt->getParent();
    ++FuncInfo.InsertPt;
  } else
    FuncInfo.InsertPt = FuncInfo.MBB->getFirstNonPHI();

  // Now skip past any EH_LABELs, which must remain at the beginning.
  while (FuncInfo.InsertPt != FuncInfo.MBB->end() &&
         FuncInfo.InsertPt->getOpcode() == TargetOpcode::EH_LABEL)
    ++FuncInfo.InsertPt;
}

void FastISel::removeDeadCode(MachineBasicBlock::iterator I,
                              MachineBasicBlock::iterator E) {
  assert(I && E && std::distance(I, E) > 0 && "Invalid iterator!");
  while (I != E) {
    MachineInstr *Dead = &*I;
    ++I;
    Dead->eraseFromParent();
    ++NumFastIselDead;
  }
  recomputeInsertPt();
}

FastISel::SavePoint FastISel::enterLocalValueArea() {
  MachineBasicBlock::iterator OldInsertPt = FuncInfo.InsertPt;
  DebugLoc OldDL = DbgLoc;
  recomputeInsertPt();
  DbgLoc = DebugLoc();
  SavePoint SP = {OldInsertPt, OldDL};
  return SP;
}

void FastISel::leaveLocalValueArea(SavePoint OldInsertPt) {
  if (FuncInfo.InsertPt != FuncInfo.MBB->begin())
    LastLocalValue = std::prev(FuncInfo.InsertPt);

  // Restore the previous insert position.
  FuncInfo.InsertPt = OldInsertPt.InsertPt;
  DbgLoc = OldInsertPt.DL;
}

bool FastISel::selectBinaryOp(const User *I, unsigned ISDOpcode) {
  EVT VT = EVT::getEVT(I->getType(), /*HandleUnknown=*/true);
  if (VT == MVT::Other || !VT.isSimple())
    // Unhandled type. Halt "fast" selection and bail.
    return false;

  // We only handle legal types. For example, on x86-32 the instruction
  // selector contains all of the 64-bit instructions from x86-64,
  // under the assumption that i64 won't be used if the target doesn't
  // support it.
  if (!TLI.isTypeLegal(VT)) {
    // MVT::i1 is special. Allow AND, OR, or XOR because they
    // don't require additional zeroing, which makes them easy.
    if (VT == MVT::i1 && (ISDOpcode == ISD::AND || ISDOpcode == ISD::OR ||
                          ISDOpcode == ISD::XOR))
      VT = TLI.getTypeToTransformTo(I->getContext(), VT);
    else
      return false;
  }

  // Check if the first operand is a constant, and handle it as "ri".  At -O0,
  // we don't have anything that canonicalizes operand order.
  if (const auto *CI = dyn_cast<ConstantInt>(I->getOperand(0)))
    if (isa<Instruction>(I) && cast<Instruction>(I)->isCommutative()) {
      unsigned Op1 = getRegForValue(I->getOperand(1));
      if (!Op1)
        return false;
      bool Op1IsKill = hasTrivialKill(I->getOperand(1));

      unsigned ResultReg =
          fastEmit_ri_(VT.getSimpleVT(), ISDOpcode, Op1, Op1IsKill,
                       CI->getZExtValue(), VT.getSimpleVT());
      if (!ResultReg)
        return false;

      // We successfully emitted code for the given LLVM Instruction.
      updateValueMap(I, ResultReg);
      return true;
    }

  unsigned Op0 = getRegForValue(I->getOperand(0));
  if (!Op0) // Unhandled operand. Halt "fast" selection and bail.
    return false;
  bool Op0IsKill = hasTrivialKill(I->getOperand(0));

  // Check if the second operand is a constant and handle it appropriately.
  if (const auto *CI = dyn_cast<ConstantInt>(I->getOperand(1))) {
    uint64_t Imm = CI->getZExtValue();

    // Transform "sdiv exact X, 8" -> "sra X, 3".
    if (ISDOpcode == ISD::SDIV && isa<BinaryOperator>(I) &&
        cast<BinaryOperator>(I)->isExact() && isPowerOf2_64(Imm)) {
      Imm = Log2_64(Imm);
      ISDOpcode = ISD::SRA;
    }

    // Transform "urem x, pow2" -> "and x, pow2-1".
    if (ISDOpcode == ISD::UREM && isa<BinaryOperator>(I) &&
        isPowerOf2_64(Imm)) {
      --Imm;
      ISDOpcode = ISD::AND;
    }

    unsigned ResultReg = fastEmit_ri_(VT.getSimpleVT(), ISDOpcode, Op0,
                                      Op0IsKill, Imm, VT.getSimpleVT());
    if (!ResultReg)
      return false;

    // We successfully emitted code for the given LLVM Instruction.
    updateValueMap(I, ResultReg);
    return true;
  }

  // Check if the second operand is a constant float.
  if (const auto *CF = dyn_cast<ConstantFP>(I->getOperand(1))) {
    unsigned ResultReg = fastEmit_rf(VT.getSimpleVT(), VT.getSimpleVT(),
                                     ISDOpcode, Op0, Op0IsKill, CF);
    if (ResultReg) {
      // We successfully emitted code for the given LLVM Instruction.
      updateValueMap(I, ResultReg);
      return true;
    }
  }

  unsigned Op1 = getRegForValue(I->getOperand(1));
  if (!Op1) // Unhandled operand. Halt "fast" selection and bail.
    return false;
  bool Op1IsKill = hasTrivialKill(I->getOperand(1));

  // Now we have both operands in registers. Emit the instruction.
  unsigned ResultReg = fastEmit_rr(VT.getSimpleVT(), VT.getSimpleVT(),
                                   ISDOpcode, Op0, Op0IsKill, Op1, Op1IsKill);
  if (!ResultReg)
    // Target-specific code wasn't able to find a machine opcode for
    // the given ISD opcode and type. Halt "fast" selection and bail.
    return false;

  // We successfully emitted code for the given LLVM Instruction.
  updateValueMap(I, ResultReg);
  return true;
}

bool FastISel::selectGetElementPtr(const User *I) {
  unsigned N = getRegForValue(I->getOperand(0));
  if (!N) // Unhandled operand. Halt "fast" selection and bail.
    return false;
  bool NIsKill = hasTrivialKill(I->getOperand(0));

  // Keep a running tab of the total offset to coalesce multiple N = N + Offset
  // into a single N = N + TotalOffset.
  uint64_t TotalOffs = 0;
  // FIXME: What's a good SWAG number for MaxOffs?
  uint64_t MaxOffs = 2048;
  Type *Ty = I->getOperand(0)->getType();
  MVT VT = TLI.getPointerTy();
  for (GetElementPtrInst::const_op_iterator OI = I->op_begin() + 1,
                                            E = I->op_end();
       OI != E; ++OI) {
    const Value *Idx = *OI;
    if (auto *StTy = dyn_cast<StructType>(Ty)) {
      uint64_t Field = cast<ConstantInt>(Idx)->getZExtValue();
      if (Field) {
        // N = N + Offset
        TotalOffs += DL.getStructLayout(StTy)->getElementOffset(Field);
        if (TotalOffs >= MaxOffs) {
          N = fastEmit_ri_(VT, ISD::ADD, N, NIsKill, TotalOffs, VT);
          if (!N) // Unhandled operand. Halt "fast" selection and bail.
            return false;
          NIsKill = true;
          TotalOffs = 0;
        }
      }
      Ty = StTy->getElementType(Field);
    } else {
      Ty = cast<SequentialType>(Ty)->getElementType();

      // If this is a constant subscript, handle it quickly.
      if (const auto *CI = dyn_cast<ConstantInt>(Idx)) {
        if (CI->isZero())
          continue;
        // N = N + Offset
        uint64_t IdxN = CI->getValue().sextOrTrunc(64).getSExtValue();
        TotalOffs += DL.getTypeAllocSize(Ty) * IdxN;
        if (TotalOffs >= MaxOffs) {
          N = fastEmit_ri_(VT, ISD::ADD, N, NIsKill, TotalOffs, VT);
          if (!N) // Unhandled operand. Halt "fast" selection and bail.
            return false;
          NIsKill = true;
          TotalOffs = 0;
        }
        continue;
      }
      if (TotalOffs) {
        N = fastEmit_ri_(VT, ISD::ADD, N, NIsKill, TotalOffs, VT);
        if (!N) // Unhandled operand. Halt "fast" selection and bail.
          return false;
        NIsKill = true;
        TotalOffs = 0;
      }

      // N = N + Idx * ElementSize;
      uint64_t ElementSize = DL.getTypeAllocSize(Ty);
      std::pair<unsigned, bool> Pair = getRegForGEPIndex(Idx);
      unsigned IdxN = Pair.first;
      bool IdxNIsKill = Pair.second;
      if (!IdxN) // Unhandled operand. Halt "fast" selection and bail.
        return false;

      if (ElementSize != 1) {
        IdxN = fastEmit_ri_(VT, ISD::MUL, IdxN, IdxNIsKill, ElementSize, VT);
        if (!IdxN) // Unhandled operand. Halt "fast" selection and bail.
          return false;
        IdxNIsKill = true;
      }
      N = fastEmit_rr(VT, VT, ISD::ADD, N, NIsKill, IdxN, IdxNIsKill);
      if (!N) // Unhandled operand. Halt "fast" selection and bail.
        return false;
    }
  }
  if (TotalOffs) {
    N = fastEmit_ri_(VT, ISD::ADD, N, NIsKill, TotalOffs, VT);
    if (!N) // Unhandled operand. Halt "fast" selection and bail.
      return false;
  }

  // We successfully emitted code for the given LLVM Instruction.
  updateValueMap(I, N);
  return true;
}

bool FastISel::addStackMapLiveVars(SmallVectorImpl<MachineOperand> &Ops,
                                   const CallInst *CI, unsigned StartIdx) {
  for (unsigned i = StartIdx, e = CI->getNumArgOperands(); i != e; ++i) {
    Value *Val = CI->getArgOperand(i);
    // Check for constants and encode them with a StackMaps::ConstantOp prefix.
    if (const auto *C = dyn_cast<ConstantInt>(Val)) {
      Ops.push_back(MachineOperand::CreateImm(StackMaps::ConstantOp));
      Ops.push_back(MachineOperand::CreateImm(C->getSExtValue()));
    } else if (isa<ConstantPointerNull>(Val)) {
      Ops.push_back(MachineOperand::CreateImm(StackMaps::ConstantOp));
      Ops.push_back(MachineOperand::CreateImm(0));
    } else if (auto *AI = dyn_cast<AllocaInst>(Val)) {
      // Values coming from a stack location also require a sepcial encoding,
      // but that is added later on by the target specific frame index
      // elimination implementation.
      auto SI = FuncInfo.StaticAllocaMap.find(AI);
      if (SI != FuncInfo.StaticAllocaMap.end())
        Ops.push_back(MachineOperand::CreateFI(SI->second));
      else
        return false;
    } else {
      unsigned Reg = getRegForValue(Val);
      if (!Reg)
        return false;
      Ops.push_back(MachineOperand::CreateReg(Reg, /*IsDef=*/false));
    }
  }
  return true;
}

bool FastISel::selectStackmap(const CallInst *I) {
  // void @llvm.experimental.stackmap(i64 <id>, i32 <numShadowBytes>,
  //                                  [live variables...])
  assert(I->getCalledFunction()->getReturnType()->isVoidTy() &&
         "Stackmap cannot return a value.");

  // The stackmap intrinsic only records the live variables (the arguments
  // passed to it) and emits NOPS (if requested). Unlike the patchpoint
  // intrinsic, this won't be lowered to a function call. This means we don't
  // have to worry about calling conventions and target-specific lowering code.
  // Instead we perform the call lowering right here.
  //
  // CALLSEQ_START(0)
  // STACKMAP(id, nbytes, ...)
  // CALLSEQ_END(0, 0)
  //
  SmallVector<MachineOperand, 32> Ops;

  // Add the <id> and <numBytes> constants.
  assert(isa<ConstantInt>(I->getOperand(PatchPointOpers::IDPos)) &&
         "Expected a constant integer.");
  const auto *ID = cast<ConstantInt>(I->getOperand(PatchPointOpers::IDPos));
  Ops.push_back(MachineOperand::CreateImm(ID->getZExtValue()));

  assert(isa<ConstantInt>(I->getOperand(PatchPointOpers::NBytesPos)) &&
         "Expected a constant integer.");
  const auto *NumBytes =
      cast<ConstantInt>(I->getOperand(PatchPointOpers::NBytesPos));
  Ops.push_back(MachineOperand::CreateImm(NumBytes->getZExtValue()));

  // Push live variables for the stack map (skipping the first two arguments
  // <id> and <numBytes>).
  if (!addStackMapLiveVars(Ops, I, 2))
    return false;

  // We are not adding any register mask info here, because the stackmap doesn't
  // clobber anything.

  // Add scratch registers as implicit def and early clobber.
  CallingConv::ID CC = I->getCallingConv();
  const MCPhysReg *ScratchRegs = TLI.getScratchRegisters(CC);
  for (unsigned i = 0; ScratchRegs[i]; ++i)
    Ops.push_back(MachineOperand::CreateReg(
        ScratchRegs[i], /*IsDef=*/true, /*IsImp=*/true, /*IsKill=*/false,
        /*IsDead=*/false, /*IsUndef=*/false, /*IsEarlyClobber=*/true));

  // Issue CALLSEQ_START
  unsigned AdjStackDown = TII.getCallFrameSetupOpcode();
  BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(AdjStackDown))
      .addImm(0);

  // Issue STACKMAP.
  MachineInstrBuilder MIB = BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc,
                                    TII.get(TargetOpcode::STACKMAP));
  for (auto const &MO : Ops)
    MIB.addOperand(MO);

  // Issue CALLSEQ_END
  unsigned AdjStackUp = TII.getCallFrameDestroyOpcode();
  BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(AdjStackUp))
      .addImm(0)
      .addImm(0);

  // Inform the Frame Information that we have a stackmap in this function.
  FuncInfo.MF->getFrameInfo()->setHasStackMap();

  return true;
}

/// \brief Lower an argument list according to the target calling convention.
///
/// This is a helper for lowering intrinsics that follow a target calling
/// convention or require stack pointer adjustment. Only a subset of the
/// intrinsic's operands need to participate in the calling convention.
bool FastISel::lowerCallOperands(const CallInst *CI, unsigned ArgIdx,
                                 unsigned NumArgs, const Value *Callee,
                                 bool ForceRetVoidTy, CallLoweringInfo &CLI) {
  ArgListTy Args;
  Args.reserve(NumArgs);

  // Populate the argument list.
  // Attributes for args start at offset 1, after the return attribute.
  ImmutableCallSite CS(CI);
  for (unsigned ArgI = ArgIdx, ArgE = ArgIdx + NumArgs, AttrI = ArgIdx + 1;
       ArgI != ArgE; ++ArgI) {
    Value *V = CI->getOperand(ArgI);

    assert(!V->getType()->isEmptyTy() && "Empty type passed to intrinsic.");

    ArgListEntry Entry;
    Entry.Val = V;
    Entry.Ty = V->getType();
    Entry.setAttributes(&CS, AttrI);
    Args.push_back(Entry);
  }

  Type *RetTy = ForceRetVoidTy ? Type::getVoidTy(CI->getType()->getContext())
                               : CI->getType();
  CLI.setCallee(CI->getCallingConv(), RetTy, Callee, std::move(Args), NumArgs);

  return lowerCallTo(CLI);
}

bool FastISel::selectPatchpoint(const CallInst *I) {
  // void|i64 @llvm.experimental.patchpoint.void|i64(i64 <id>,
  //                                                 i32 <numBytes>,
  //                                                 i8* <target>,
  //                                                 i32 <numArgs>,
  //                                                 [Args...],
  //                                                 [live variables...])
  CallingConv::ID CC = I->getCallingConv();
  bool IsAnyRegCC = CC == CallingConv::AnyReg;
  bool HasDef = !I->getType()->isVoidTy();
  Value *Callee = I->getOperand(PatchPointOpers::TargetPos);

  // Get the real number of arguments participating in the call <numArgs>
  assert(isa<ConstantInt>(I->getOperand(PatchPointOpers::NArgPos)) &&
         "Expected a constant integer.");
  const auto *NumArgsVal =
      cast<ConstantInt>(I->getOperand(PatchPointOpers::NArgPos));
  unsigned NumArgs = NumArgsVal->getZExtValue();

  // Skip the four meta args: <id>, <numNopBytes>, <target>, <numArgs>
  // This includes all meta-operands up to but not including CC.
  unsigned NumMetaOpers = PatchPointOpers::CCPos;
  assert(I->getNumArgOperands() >= NumMetaOpers + NumArgs &&
         "Not enough arguments provided to the patchpoint intrinsic");

  // For AnyRegCC the arguments are lowered later on manually.
  unsigned NumCallArgs = IsAnyRegCC ? 0 : NumArgs;
  CallLoweringInfo CLI;
  CLI.setIsPatchPoint();
  if (!lowerCallOperands(I, NumMetaOpers, NumCallArgs, Callee, IsAnyRegCC, CLI))
    return false;

  assert(CLI.Call && "No call instruction specified.");

  SmallVector<MachineOperand, 32> Ops;

  // Add an explicit result reg if we use the anyreg calling convention.
  if (IsAnyRegCC && HasDef) {
    assert(CLI.NumResultRegs == 0 && "Unexpected result register.");
    CLI.ResultReg = createResultReg(TLI.getRegClassFor(MVT::i64));
    CLI.NumResultRegs = 1;
    Ops.push_back(MachineOperand::CreateReg(CLI.ResultReg, /*IsDef=*/true));
  }

  // Add the <id> and <numBytes> constants.
  assert(isa<ConstantInt>(I->getOperand(PatchPointOpers::IDPos)) &&
         "Expected a constant integer.");
  const auto *ID = cast<ConstantInt>(I->getOperand(PatchPointOpers::IDPos));
  Ops.push_back(MachineOperand::CreateImm(ID->getZExtValue()));

  assert(isa<ConstantInt>(I->getOperand(PatchPointOpers::NBytesPos)) &&
         "Expected a constant integer.");
  const auto *NumBytes =
      cast<ConstantInt>(I->getOperand(PatchPointOpers::NBytesPos));
  Ops.push_back(MachineOperand::CreateImm(NumBytes->getZExtValue()));

  // Assume that the callee is a constant address or null pointer.
  // FIXME: handle function symbols in the future.
  uint64_t CalleeAddr;
  if (const auto *C = dyn_cast<IntToPtrInst>(Callee))
    CalleeAddr = cast<ConstantInt>(C->getOperand(0))->getZExtValue();
  else if (const auto *C = dyn_cast<ConstantExpr>(Callee)) {
    if (C->getOpcode() == Instruction::IntToPtr)
      CalleeAddr = cast<ConstantInt>(C->getOperand(0))->getZExtValue();
    else
      llvm_unreachable("Unsupported ConstantExpr.");
  } else if (isa<ConstantPointerNull>(Callee))
    CalleeAddr = 0;
  else
    llvm_unreachable("Unsupported callee address.");

  Ops.push_back(MachineOperand::CreateImm(CalleeAddr));

  // Adjust <numArgs> to account for any arguments that have been passed on
  // the stack instead.
  unsigned NumCallRegArgs = IsAnyRegCC ? NumArgs : CLI.OutRegs.size();
  Ops.push_back(MachineOperand::CreateImm(NumCallRegArgs));

  // Add the calling convention
  Ops.push_back(MachineOperand::CreateImm((unsigned)CC));

  // Add the arguments we omitted previously. The register allocator should
  // place these in any free register.
  if (IsAnyRegCC) {
    for (unsigned i = NumMetaOpers, e = NumMetaOpers + NumArgs; i != e; ++i) {
      unsigned Reg = getRegForValue(I->getArgOperand(i));
      if (!Reg)
        return false;
      Ops.push_back(MachineOperand::CreateReg(Reg, /*IsDef=*/false));
    }
  }

  // Push the arguments from the call instruction.
  for (auto Reg : CLI.OutRegs)
    Ops.push_back(MachineOperand::CreateReg(Reg, /*IsDef=*/false));

  // Push live variables for the stack map.
  if (!addStackMapLiveVars(Ops, I, NumMetaOpers + NumArgs))
    return false;

  // Push the register mask info.
  Ops.push_back(MachineOperand::CreateRegMask(
      TRI.getCallPreservedMask(*FuncInfo.MF, CC)));

  // Add scratch registers as implicit def and early clobber.
  const MCPhysReg *ScratchRegs = TLI.getScratchRegisters(CC);
  for (unsigned i = 0; ScratchRegs[i]; ++i)
    Ops.push_back(MachineOperand::CreateReg(
        ScratchRegs[i], /*IsDef=*/true, /*IsImp=*/true, /*IsKill=*/false,
        /*IsDead=*/false, /*IsUndef=*/false, /*IsEarlyClobber=*/true));

  // Add implicit defs (return values).
  for (auto Reg : CLI.InRegs)
    Ops.push_back(MachineOperand::CreateReg(Reg, /*IsDef=*/true,
                                            /*IsImpl=*/true));

  // Insert the patchpoint instruction before the call generated by the target.
  MachineInstrBuilder MIB = BuildMI(*FuncInfo.MBB, CLI.Call, DbgLoc,
                                    TII.get(TargetOpcode::PATCHPOINT));

  for (auto &MO : Ops)
    MIB.addOperand(MO);

  MIB->setPhysRegsDeadExcept(CLI.InRegs, TRI);

  // Delete the original call instruction.
  CLI.Call->eraseFromParent();

  // Inform the Frame Information that we have a patchpoint in this function.
  FuncInfo.MF->getFrameInfo()->setHasPatchPoint();

  if (CLI.NumResultRegs)
    updateValueMap(I, CLI.ResultReg, CLI.NumResultRegs);
  return true;
}

/// Returns an AttributeSet representing the attributes applied to the return
/// value of the given call.
static AttributeSet getReturnAttrs(FastISel::CallLoweringInfo &CLI) {
  SmallVector<Attribute::AttrKind, 2> Attrs;
  if (CLI.RetSExt)
    Attrs.push_back(Attribute::SExt);
  if (CLI.RetZExt)
    Attrs.push_back(Attribute::ZExt);
  if (CLI.IsInReg)
    Attrs.push_back(Attribute::InReg);

  return AttributeSet::get(CLI.RetTy->getContext(), AttributeSet::ReturnIndex,
                           Attrs);
}

bool FastISel::lowerCallTo(const CallInst *CI, const char *SymName,
                           unsigned NumArgs) {
  ImmutableCallSite CS(CI);

  PointerType *PT = cast<PointerType>(CS.getCalledValue()->getType());
  FunctionType *FTy = cast<FunctionType>(PT->getElementType());
  Type *RetTy = FTy->getReturnType();

  ArgListTy Args;
  Args.reserve(NumArgs);

  // Populate the argument list.
  // Attributes for args start at offset 1, after the return attribute.
  for (unsigned ArgI = 0; ArgI != NumArgs; ++ArgI) {
    Value *V = CI->getOperand(ArgI);

    assert(!V->getType()->isEmptyTy() && "Empty type passed to intrinsic.");

    ArgListEntry Entry;
    Entry.Val = V;
    Entry.Ty = V->getType();
    Entry.setAttributes(&CS, ArgI + 1);
    Args.push_back(Entry);
  }

  CallLoweringInfo CLI;
  CLI.setCallee(RetTy, FTy, SymName, std::move(Args), CS, NumArgs);

  return lowerCallTo(CLI);
}

bool FastISel::lowerCallTo(CallLoweringInfo &CLI) {
  // Handle the incoming return values from the call.
  CLI.clearIns();
  SmallVector<EVT, 4> RetTys;
  ComputeValueVTs(TLI, CLI.RetTy, RetTys);

  SmallVector<ISD::OutputArg, 4> Outs;
  GetReturnInfo(CLI.RetTy, getReturnAttrs(CLI), Outs, TLI);

  bool CanLowerReturn = TLI.CanLowerReturn(
      CLI.CallConv, *FuncInfo.MF, CLI.IsVarArg, Outs, CLI.RetTy->getContext());

  // FIXME: sret demotion isn't supported yet - bail out.
  if (!CanLowerReturn)
    return false;

  for (unsigned I = 0, E = RetTys.size(); I != E; ++I) {
    EVT VT = RetTys[I];
    MVT RegisterVT = TLI.getRegisterType(CLI.RetTy->getContext(), VT);
    unsigned NumRegs = TLI.getNumRegisters(CLI.RetTy->getContext(), VT);
    for (unsigned i = 0; i != NumRegs; ++i) {
      ISD::InputArg MyFlags;
      MyFlags.VT = RegisterVT;
      MyFlags.ArgVT = VT;
      MyFlags.Used = CLI.IsReturnValueUsed;
      if (CLI.RetSExt)
        MyFlags.Flags.setSExt();
      if (CLI.RetZExt)
        MyFlags.Flags.setZExt();
      if (CLI.IsInReg)
        MyFlags.Flags.setInReg();
      CLI.Ins.push_back(MyFlags);
    }
  }

  // Handle all of the outgoing arguments.
  CLI.clearOuts();
  for (auto &Arg : CLI.getArgs()) {
    Type *FinalType = Arg.Ty;
    if (Arg.IsByVal)
      FinalType = cast<PointerType>(Arg.Ty)->getElementType();
    bool NeedsRegBlock = TLI.functionArgumentNeedsConsecutiveRegisters(
        FinalType, CLI.CallConv, CLI.IsVarArg);

    ISD::ArgFlagsTy Flags;
    if (Arg.IsZExt)
      Flags.setZExt();
    if (Arg.IsSExt)
      Flags.setSExt();
    if (Arg.IsInReg)
      Flags.setInReg();
    if (Arg.IsSRet)
      Flags.setSRet();
    if (Arg.IsByVal)
      Flags.setByVal();
    if (Arg.IsInAlloca) {
      Flags.setInAlloca();
      // Set the byval flag for CCAssignFn callbacks that don't know about
      // inalloca. This way we can know how many bytes we should've allocated
      // and how many bytes a callee cleanup function will pop.  If we port
      // inalloca to more targets, we'll have to add custom inalloca handling in
      // the various CC lowering callbacks.
      Flags.setByVal();
    }
    if (Arg.IsByVal || Arg.IsInAlloca) {
      PointerType *Ty = cast<PointerType>(Arg.Ty);
      Type *ElementTy = Ty->getElementType();
      unsigned FrameSize = DL.getTypeAllocSize(ElementTy);
      // For ByVal, alignment should come from FE. BE will guess if this info is
      // not there, but there are cases it cannot get right.
      unsigned FrameAlign = Arg.Alignment;
      if (!FrameAlign)
        FrameAlign = TLI.getByValTypeAlignment(ElementTy);
      Flags.setByValSize(FrameSize);
      Flags.setByValAlign(FrameAlign);
    }
    if (Arg.IsNest)
      Flags.setNest();
    if (NeedsRegBlock)
      Flags.setInConsecutiveRegs();
    unsigned OriginalAlignment = DL.getABITypeAlignment(Arg.Ty);
    Flags.setOrigAlign(OriginalAlignment);

    CLI.OutVals.push_back(Arg.Val);
    CLI.OutFlags.push_back(Flags);
  }

  if (!fastLowerCall(CLI))
    return false;

  // Set all unused physreg defs as dead.
  assert(CLI.Call && "No call instruction specified.");
  CLI.Call->setPhysRegsDeadExcept(CLI.InRegs, TRI);

  if (CLI.NumResultRegs && CLI.CS)
    updateValueMap(CLI.CS->getInstruction(), CLI.ResultReg, CLI.NumResultRegs);

  return true;
}

bool FastISel::lowerCall(const CallInst *CI) {
  ImmutableCallSite CS(CI);

  PointerType *PT = cast<PointerType>(CS.getCalledValue()->getType());
  FunctionType *FuncTy = cast<FunctionType>(PT->getElementType());
  Type *RetTy = FuncTy->getReturnType();

  ArgListTy Args;
  ArgListEntry Entry;
  Args.reserve(CS.arg_size());

  for (ImmutableCallSite::arg_iterator i = CS.arg_begin(), e = CS.arg_end();
       i != e; ++i) {
    Value *V = *i;

    // Skip empty types
    if (V->getType()->isEmptyTy())
      continue;

    Entry.Val = V;
    Entry.Ty = V->getType();

    // Skip the first return-type Attribute to get to params.
    Entry.setAttributes(&CS, i - CS.arg_begin() + 1);
    Args.push_back(Entry);
  }

  // Check if target-independent constraints permit a tail call here.
  // Target-dependent constraints are checked within fastLowerCall.
  bool IsTailCall = CI->isTailCall();
  if (IsTailCall && !isInTailCallPosition(CS, TM))
    IsTailCall = false;

  CallLoweringInfo CLI;
  CLI.setCallee(RetTy, FuncTy, CI->getCalledValue(), std::move(Args), CS)
      .setTailCall(IsTailCall);

  return lowerCallTo(CLI);
}

bool FastISel::selectCall(const User *I) {
  const CallInst *Call = cast<CallInst>(I);

  // Handle simple inline asms.
  if (const InlineAsm *IA = dyn_cast<InlineAsm>(Call->getCalledValue())) {
    // If the inline asm has side effects, then make sure that no local value
    // lives across by flushing the local value map.
    if (IA->hasSideEffects())
      flushLocalValueMap();

    // Don't attempt to handle constraints.
    if (!IA->getConstraintString().empty())
      return false;

    unsigned ExtraInfo = 0;
    if (IA->hasSideEffects())
      ExtraInfo |= InlineAsm::Extra_HasSideEffects;
    if (IA->isAlignStack())
      ExtraInfo |= InlineAsm::Extra_IsAlignStack;

    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc,
            TII.get(TargetOpcode::INLINEASM))
        .addExternalSymbol(IA->getAsmString().c_str())
        .addImm(ExtraInfo);
    return true;
  }

  MachineModuleInfo &MMI = FuncInfo.MF->getMMI();
  ComputeUsesVAFloatArgument(*Call, &MMI);

  // Handle intrinsic function calls.
  if (const auto *II = dyn_cast<IntrinsicInst>(Call))
    return selectIntrinsicCall(II);

  // Usually, it does not make sense to initialize a value,
  // make an unrelated function call and use the value, because
  // it tends to be spilled on the stack. So, we move the pointer
  // to the last local value to the beginning of the block, so that
  // all the values which have already been materialized,
  // appear after the call. It also makes sense to skip intrinsics
  // since they tend to be inlined.
  flushLocalValueMap();

  return lowerCall(Call);
}

bool FastISel::selectIntrinsicCall(const IntrinsicInst *II) {
  switch (II->getIntrinsicID()) {
  default:
    break;
  // At -O0 we don't care about the lifetime intrinsics.
  case Intrinsic::lifetime_start:
  case Intrinsic::lifetime_end:
  // The donothing intrinsic does, well, nothing.
  case Intrinsic::donothing:
    return true;
  case Intrinsic::dbg_declare: {
    const DbgDeclareInst *DI = cast<DbgDeclareInst>(II);
    DIVariable DIVar(DI->getVariable());
    assert((!DIVar || DIVar.isVariable()) &&
           "Variable in DbgDeclareInst should be either null or a DIVariable.");
    if (!DIVar || !FuncInfo.MF->getMMI().hasDebugInfo()) {
      DEBUG(dbgs() << "Dropping debug info for " << *DI << "\n");
      return true;
    }

    const Value *Address = DI->getAddress();
    if (!Address || isa<UndefValue>(Address)) {
      DEBUG(dbgs() << "Dropping debug info for " << *DI << "\n");
      return true;
    }

    unsigned Offset = 0;
    Optional<MachineOperand> Op;
    if (const auto *Arg = dyn_cast<Argument>(Address))
      // Some arguments' frame index is recorded during argument lowering.
      Offset = FuncInfo.getArgumentFrameIndex(Arg);
    if (Offset)
      Op = MachineOperand::CreateFI(Offset);
    if (!Op)
      if (unsigned Reg = lookUpRegForValue(Address))
        Op = MachineOperand::CreateReg(Reg, false);

    // If we have a VLA that has a "use" in a metadata node that's then used
    // here but it has no other uses, then we have a problem. E.g.,
    //
    //   int foo (const int *x) {
    //     char a[*x];
    //     return 0;
    //   }
    //
    // If we assign 'a' a vreg and fast isel later on has to use the selection
    // DAG isel, it will want to copy the value to the vreg. However, there are
    // no uses, which goes counter to what selection DAG isel expects.
    if (!Op && !Address->use_empty() && isa<Instruction>(Address) &&
        (!isa<AllocaInst>(Address) ||
         !FuncInfo.StaticAllocaMap.count(cast<AllocaInst>(Address))))
      Op = MachineOperand::CreateReg(FuncInfo.InitializeRegForValue(Address),
                                     false);

    if (Op) {
      if (Op->isReg()) {
        Op->setIsDebug(true);
        BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc,
                TII.get(TargetOpcode::DBG_VALUE), false, Op->getReg(), 0,
                DI->getVariable(), DI->getExpression());
      } else
        BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc,
                TII.get(TargetOpcode::DBG_VALUE))
            .addOperand(*Op)
            .addImm(0)
            .addMetadata(DI->getVariable())
            .addMetadata(DI->getExpression());
    } else {
      // We can't yet handle anything else here because it would require
      // generating code, thus altering codegen because of debug info.
      DEBUG(dbgs() << "Dropping debug info for " << *DI << "\n");
    }
    return true;
  }
  case Intrinsic::dbg_value: {
    // This form of DBG_VALUE is target-independent.
    const DbgValueInst *DI = cast<DbgValueInst>(II);
    const MCInstrDesc &II = TII.get(TargetOpcode::DBG_VALUE);
    const Value *V = DI->getValue();
    if (!V) {
      // Currently the optimizer can produce this; insert an undef to
      // help debugging.  Probably the optimizer should not do this.
      BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II)
          .addReg(0U)
          .addImm(DI->getOffset())
          .addMetadata(DI->getVariable())
          .addMetadata(DI->getExpression());
    } else if (const auto *CI = dyn_cast<ConstantInt>(V)) {
      if (CI->getBitWidth() > 64)
        BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II)
            .addCImm(CI)
            .addImm(DI->getOffset())
            .addMetadata(DI->getVariable())
            .addMetadata(DI->getExpression());
      else
        BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II)
            .addImm(CI->getZExtValue())
            .addImm(DI->getOffset())
            .addMetadata(DI->getVariable())
            .addMetadata(DI->getExpression());
    } else if (const auto *CF = dyn_cast<ConstantFP>(V)) {
      BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II)
          .addFPImm(CF)
          .addImm(DI->getOffset())
          .addMetadata(DI->getVariable())
          .addMetadata(DI->getExpression());
    } else if (unsigned Reg = lookUpRegForValue(V)) {
      // FIXME: This does not handle register-indirect values at offset 0.
      bool IsIndirect = DI->getOffset() != 0;
      BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II, IsIndirect, Reg,
              DI->getOffset(), DI->getVariable(), DI->getExpression());
    } else {
      // We can't yet handle anything else here because it would require
      // generating code, thus altering codegen because of debug info.
      DEBUG(dbgs() << "Dropping debug info for " << *DI << "\n");
    }
    return true;
  }
  case Intrinsic::objectsize: {
    ConstantInt *CI = cast<ConstantInt>(II->getArgOperand(1));
    unsigned long long Res = CI->isZero() ? -1ULL : 0;
    Constant *ResCI = ConstantInt::get(II->getType(), Res);
    unsigned ResultReg = getRegForValue(ResCI);
    if (!ResultReg)
      return false;
    updateValueMap(II, ResultReg);
    return true;
  }
  case Intrinsic::expect: {
    unsigned ResultReg = getRegForValue(II->getArgOperand(0));
    if (!ResultReg)
      return false;
    updateValueMap(II, ResultReg);
    return true;
  }
  case Intrinsic::experimental_stackmap:
    return selectStackmap(II);
  case Intrinsic::experimental_patchpoint_void:
  case Intrinsic::experimental_patchpoint_i64:
    return selectPatchpoint(II);
  }

  return fastLowerIntrinsicCall(II);
}

bool FastISel::selectCast(const User *I, unsigned Opcode) {
  EVT SrcVT = TLI.getValueType(I->getOperand(0)->getType());
  EVT DstVT = TLI.getValueType(I->getType());

  if (SrcVT == MVT::Other || !SrcVT.isSimple() || DstVT == MVT::Other ||
      !DstVT.isSimple())
    // Unhandled type. Halt "fast" selection and bail.
    return false;

  // Check if the destination type is legal.
  if (!TLI.isTypeLegal(DstVT))
    return false;

  // Check if the source operand is legal.
  if (!TLI.isTypeLegal(SrcVT))
    return false;

  unsigned InputReg = getRegForValue(I->getOperand(0));
  if (!InputReg)
    // Unhandled operand.  Halt "fast" selection and bail.
    return false;

  bool InputRegIsKill = hasTrivialKill(I->getOperand(0));

  unsigned ResultReg = fastEmit_r(SrcVT.getSimpleVT(), DstVT.getSimpleVT(),
                                  Opcode, InputReg, InputRegIsKill);
  if (!ResultReg)
    return false;

  updateValueMap(I, ResultReg);
  return true;
}

bool FastISel::selectBitCast(const User *I) {
  // If the bitcast doesn't change the type, just use the operand value.
  if (I->getType() == I->getOperand(0)->getType()) {
    unsigned Reg = getRegForValue(I->getOperand(0));
    if (!Reg)
      return false;
    updateValueMap(I, Reg);
    return true;
  }

  // Bitcasts of other values become reg-reg copies or BITCAST operators.
  EVT SrcEVT = TLI.getValueType(I->getOperand(0)->getType());
  EVT DstEVT = TLI.getValueType(I->getType());
  if (SrcEVT == MVT::Other || DstEVT == MVT::Other ||
      !TLI.isTypeLegal(SrcEVT) || !TLI.isTypeLegal(DstEVT))
    // Unhandled type. Halt "fast" selection and bail.
    return false;

  MVT SrcVT = SrcEVT.getSimpleVT();
  MVT DstVT = DstEVT.getSimpleVT();
  unsigned Op0 = getRegForValue(I->getOperand(0));
  if (!Op0) // Unhandled operand. Halt "fast" selection and bail.
    return false;
  bool Op0IsKill = hasTrivialKill(I->getOperand(0));

  // First, try to perform the bitcast by inserting a reg-reg copy.
  unsigned ResultReg = 0;
  if (SrcVT == DstVT) {
    const TargetRegisterClass *SrcClass = TLI.getRegClassFor(SrcVT);
    const TargetRegisterClass *DstClass = TLI.getRegClassFor(DstVT);
    // Don't attempt a cross-class copy. It will likely fail.
    if (SrcClass == DstClass) {
      ResultReg = createResultReg(DstClass);
      BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc,
              TII.get(TargetOpcode::COPY), ResultReg).addReg(Op0);
    }
  }

  // If the reg-reg copy failed, select a BITCAST opcode.
  if (!ResultReg)
    ResultReg = fastEmit_r(SrcVT, DstVT, ISD::BITCAST, Op0, Op0IsKill);

  if (!ResultReg)
    return false;

  updateValueMap(I, ResultReg);
  return true;
}

bool FastISel::selectInstruction(const Instruction *I) {
  // Just before the terminator instruction, insert instructions to
  // feed PHI nodes in successor blocks.
  if (isa<TerminatorInst>(I))
    if (!handlePHINodesInSuccessorBlocks(I->getParent()))
      return false;

  DbgLoc = I->getDebugLoc();

  SavedInsertPt = FuncInfo.InsertPt;

  if (const auto *Call = dyn_cast<CallInst>(I)) {
    const Function *F = Call->getCalledFunction();
    LibFunc::Func Func;

    // As a special case, don't handle calls to builtin library functions that
    // may be translated directly to target instructions.
    if (F && !F->hasLocalLinkage() && F->hasName() &&
        LibInfo->getLibFunc(F->getName(), Func) &&
        LibInfo->hasOptimizedCodeGen(Func))
      return false;

    // Don't handle Intrinsic::trap if a trap funciton is specified.
    if (F && F->getIntrinsicID() == Intrinsic::trap &&
        !TM.Options.getTrapFunctionName().empty())
      return false;
  }

  // First, try doing target-independent selection.
  if (!SkipTargetIndependentISel) {
    if (selectOperator(I, I->getOpcode())) {
      ++NumFastIselSuccessIndependent;
      DbgLoc = DebugLoc();
      return true;
    }
    // Remove dead code.
    recomputeInsertPt();
    if (SavedInsertPt != FuncInfo.InsertPt)
      removeDeadCode(FuncInfo.InsertPt, SavedInsertPt);
    SavedInsertPt = FuncInfo.InsertPt;
  }
  // Next, try calling the target to attempt to handle the instruction.
  if (fastSelectInstruction(I)) {
    ++NumFastIselSuccessTarget;
    DbgLoc = DebugLoc();
    return true;
  }
  // Remove dead code.
  recomputeInsertPt();
  if (SavedInsertPt != FuncInfo.InsertPt)
    removeDeadCode(FuncInfo.InsertPt, SavedInsertPt);

  DbgLoc = DebugLoc();
  // Undo phi node updates, because they will be added again by SelectionDAG.
  if (isa<TerminatorInst>(I))
    FuncInfo.PHINodesToUpdate.resize(FuncInfo.OrigNumPHINodesToUpdate);
  return false;
}

/// Emit an unconditional branch to the given block, unless it is the immediate
/// (fall-through) successor, and update the CFG.
void FastISel::fastEmitBranch(MachineBasicBlock *MSucc, DebugLoc DbgLoc) {
  if (FuncInfo.MBB->getBasicBlock()->size() > 1 &&
      FuncInfo.MBB->isLayoutSuccessor(MSucc)) {
    // For more accurate line information if this is the only instruction
    // in the block then emit it, otherwise we have the unconditional
    // fall-through case, which needs no instructions.
  } else {
    // The unconditional branch case.
    TII.InsertBranch(*FuncInfo.MBB, MSucc, nullptr,
                     SmallVector<MachineOperand, 0>(), DbgLoc);
  }
  uint32_t BranchWeight = 0;
  if (FuncInfo.BPI)
    BranchWeight = FuncInfo.BPI->getEdgeWeight(FuncInfo.MBB->getBasicBlock(),
                                               MSucc->getBasicBlock());
  FuncInfo.MBB->addSuccessor(MSucc, BranchWeight);
}

/// Emit an FNeg operation.
bool FastISel::selectFNeg(const User *I) {
  unsigned OpReg = getRegForValue(BinaryOperator::getFNegArgument(I));
  if (!OpReg)
    return false;
  bool OpRegIsKill = hasTrivialKill(I);

  // If the target has ISD::FNEG, use it.
  EVT VT = TLI.getValueType(I->getType());
  unsigned ResultReg = fastEmit_r(VT.getSimpleVT(), VT.getSimpleVT(), ISD::FNEG,
                                  OpReg, OpRegIsKill);
  if (ResultReg) {
    updateValueMap(I, ResultReg);
    return true;
  }

  // Bitcast the value to integer, twiddle the sign bit with xor,
  // and then bitcast it back to floating-point.
  if (VT.getSizeInBits() > 64)
    return false;
  EVT IntVT = EVT::getIntegerVT(I->getContext(), VT.getSizeInBits());
  if (!TLI.isTypeLegal(IntVT))
    return false;

  unsigned IntReg = fastEmit_r(VT.getSimpleVT(), IntVT.getSimpleVT(),
                               ISD::BITCAST, OpReg, OpRegIsKill);
  if (!IntReg)
    return false;

  unsigned IntResultReg = fastEmit_ri_(
      IntVT.getSimpleVT(), ISD::XOR, IntReg, /*IsKill=*/true,
      UINT64_C(1) << (VT.getSizeInBits() - 1), IntVT.getSimpleVT());
  if (!IntResultReg)
    return false;

  ResultReg = fastEmit_r(IntVT.getSimpleVT(), VT.getSimpleVT(), ISD::BITCAST,
                         IntResultReg, /*IsKill=*/true);
  if (!ResultReg)
    return false;

  updateValueMap(I, ResultReg);
  return true;
}

bool FastISel::selectExtractValue(const User *U) {
  const ExtractValueInst *EVI = dyn_cast<ExtractValueInst>(U);
  if (!EVI)
    return false;

  // Make sure we only try to handle extracts with a legal result.  But also
  // allow i1 because it's easy.
  EVT RealVT = TLI.getValueType(EVI->getType(), /*AllowUnknown=*/true);
  if (!RealVT.isSimple())
    return false;
  MVT VT = RealVT.getSimpleVT();
  if (!TLI.isTypeLegal(VT) && VT != MVT::i1)
    return false;

  const Value *Op0 = EVI->getOperand(0);
  Type *AggTy = Op0->getType();

  // Get the base result register.
  unsigned ResultReg;
  DenseMap<const Value *, unsigned>::iterator I = FuncInfo.ValueMap.find(Op0);
  if (I != FuncInfo.ValueMap.end())
    ResultReg = I->second;
  else if (isa<Instruction>(Op0))
    ResultReg = FuncInfo.InitializeRegForValue(Op0);
  else
    return false; // fast-isel can't handle aggregate constants at the moment

  // Get the actual result register, which is an offset from the base register.
  unsigned VTIndex = ComputeLinearIndex(AggTy, EVI->getIndices());

  SmallVector<EVT, 4> AggValueVTs;
  ComputeValueVTs(TLI, AggTy, AggValueVTs);

  for (unsigned i = 0; i < VTIndex; i++)
    ResultReg += TLI.getNumRegisters(FuncInfo.Fn->getContext(), AggValueVTs[i]);

  updateValueMap(EVI, ResultReg);
  return true;
}

bool FastISel::selectOperator(const User *I, unsigned Opcode) {
  switch (Opcode) {
  case Instruction::Add:
    return selectBinaryOp(I, ISD::ADD);
  case Instruction::FAdd:
    return selectBinaryOp(I, ISD::FADD);
  case Instruction::Sub:
    return selectBinaryOp(I, ISD::SUB);
  case Instruction::FSub:
    // FNeg is currently represented in LLVM IR as a special case of FSub.
    if (BinaryOperator::isFNeg(I))
      return selectFNeg(I);
    return selectBinaryOp(I, ISD::FSUB);
  case Instruction::Mul:
    return selectBinaryOp(I, ISD::MUL);
  case Instruction::FMul:
    return selectBinaryOp(I, ISD::FMUL);
  case Instruction::SDiv:
    return selectBinaryOp(I, ISD::SDIV);
  case Instruction::UDiv:
    return selectBinaryOp(I, ISD::UDIV);
  case Instruction::FDiv:
    return selectBinaryOp(I, ISD::FDIV);
  case Instruction::SRem:
    return selectBinaryOp(I, ISD::SREM);
  case Instruction::URem:
    return selectBinaryOp(I, ISD::UREM);
  case Instruction::FRem:
    return selectBinaryOp(I, ISD::FREM);
  case Instruction::Shl:
    return selectBinaryOp(I, ISD::SHL);
  case Instruction::LShr:
    return selectBinaryOp(I, ISD::SRL);
  case Instruction::AShr:
    return selectBinaryOp(I, ISD::SRA);
  case Instruction::And:
    return selectBinaryOp(I, ISD::AND);
  case Instruction::Or:
    return selectBinaryOp(I, ISD::OR);
  case Instruction::Xor:
    return selectBinaryOp(I, ISD::XOR);

  case Instruction::GetElementPtr:
    return selectGetElementPtr(I);

  case Instruction::Br: {
    const BranchInst *BI = cast<BranchInst>(I);

    if (BI->isUnconditional()) {
      const BasicBlock *LLVMSucc = BI->getSuccessor(0);
      MachineBasicBlock *MSucc = FuncInfo.MBBMap[LLVMSucc];
      fastEmitBranch(MSucc, BI->getDebugLoc());
      return true;
    }

    // Conditional branches are not handed yet.
    // Halt "fast" selection and bail.
    return false;
  }

  case Instruction::Unreachable:
    if (TM.Options.TrapUnreachable)
      return fastEmit_(MVT::Other, MVT::Other, ISD::TRAP) != 0;
    else
      return true;

  case Instruction::Alloca:
    // FunctionLowering has the static-sized case covered.
    if (FuncInfo.StaticAllocaMap.count(cast<AllocaInst>(I)))
      return true;

    // Dynamic-sized alloca is not handled yet.
    return false;

  case Instruction::Call:
    return selectCall(I);

  case Instruction::BitCast:
    return selectBitCast(I);

  case Instruction::FPToSI:
    return selectCast(I, ISD::FP_TO_SINT);
  case Instruction::ZExt:
    return selectCast(I, ISD::ZERO_EXTEND);
  case Instruction::SExt:
    return selectCast(I, ISD::SIGN_EXTEND);
  case Instruction::Trunc:
    return selectCast(I, ISD::TRUNCATE);
  case Instruction::SIToFP:
    return selectCast(I, ISD::SINT_TO_FP);

  case Instruction::IntToPtr: // Deliberate fall-through.
  case Instruction::PtrToInt: {
    EVT SrcVT = TLI.getValueType(I->getOperand(0)->getType());
    EVT DstVT = TLI.getValueType(I->getType());
    if (DstVT.bitsGT(SrcVT))
      return selectCast(I, ISD::ZERO_EXTEND);
    if (DstVT.bitsLT(SrcVT))
      return selectCast(I, ISD::TRUNCATE);
    unsigned Reg = getRegForValue(I->getOperand(0));
    if (!Reg)
      return false;
    updateValueMap(I, Reg);
    return true;
  }

  case Instruction::ExtractValue:
    return selectExtractValue(I);

  case Instruction::PHI:
    llvm_unreachable("FastISel shouldn't visit PHI nodes!");

  default:
    // Unhandled instruction. Halt "fast" selection and bail.
    return false;
  }
}

FastISel::FastISel(FunctionLoweringInfo &FuncInfo,
                   const TargetLibraryInfo *LibInfo,
                   bool SkipTargetIndependentISel)
    : FuncInfo(FuncInfo), MF(FuncInfo.MF), MRI(FuncInfo.MF->getRegInfo()),
      MFI(*FuncInfo.MF->getFrameInfo()), MCP(*FuncInfo.MF->getConstantPool()),
      TM(FuncInfo.MF->getTarget()), DL(*TM.getDataLayout()),
      TII(*MF->getSubtarget().getInstrInfo()),
      TLI(*MF->getSubtarget().getTargetLowering()),
      TRI(*MF->getSubtarget().getRegisterInfo()), LibInfo(LibInfo),
      SkipTargetIndependentISel(SkipTargetIndependentISel) {}

FastISel::~FastISel() {}

bool FastISel::fastLowerArguments() { return false; }

bool FastISel::fastLowerCall(CallLoweringInfo & /*CLI*/) { return false; }

bool FastISel::fastLowerIntrinsicCall(const IntrinsicInst * /*II*/) {
  return false;
}

unsigned FastISel::fastEmit_(MVT, MVT, unsigned) { return 0; }

unsigned FastISel::fastEmit_r(MVT, MVT, unsigned, unsigned /*Op0*/,
                              bool /*Op0IsKill*/) {
  return 0;
}

unsigned FastISel::fastEmit_rr(MVT, MVT, unsigned, unsigned /*Op0*/,
                               bool /*Op0IsKill*/, unsigned /*Op1*/,
                               bool /*Op1IsKill*/) {
  return 0;
}

unsigned FastISel::fastEmit_i(MVT, MVT, unsigned, uint64_t /*Imm*/) {
  return 0;
}

unsigned FastISel::fastEmit_f(MVT, MVT, unsigned,
                              const ConstantFP * /*FPImm*/) {
  return 0;
}

unsigned FastISel::fastEmit_ri(MVT, MVT, unsigned, unsigned /*Op0*/,
                               bool /*Op0IsKill*/, uint64_t /*Imm*/) {
  return 0;
}

unsigned FastISel::fastEmit_rf(MVT, MVT, unsigned, unsigned /*Op0*/,
                               bool /*Op0IsKill*/,
                               const ConstantFP * /*FPImm*/) {
  return 0;
}

unsigned FastISel::fastEmit_rri(MVT, MVT, unsigned, unsigned /*Op0*/,
                                bool /*Op0IsKill*/, unsigned /*Op1*/,
                                bool /*Op1IsKill*/, uint64_t /*Imm*/) {
  return 0;
}

/// This method is a wrapper of fastEmit_ri. It first tries to emit an
/// instruction with an immediate operand using fastEmit_ri.
/// If that fails, it materializes the immediate into a register and try
/// fastEmit_rr instead.
unsigned FastISel::fastEmit_ri_(MVT VT, unsigned Opcode, unsigned Op0,
                                bool Op0IsKill, uint64_t Imm, MVT ImmType) {
  // If this is a multiply by a power of two, emit this as a shift left.
  if (Opcode == ISD::MUL && isPowerOf2_64(Imm)) {
    Opcode = ISD::SHL;
    Imm = Log2_64(Imm);
  } else if (Opcode == ISD::UDIV && isPowerOf2_64(Imm)) {
    // div x, 8 -> srl x, 3
    Opcode = ISD::SRL;
    Imm = Log2_64(Imm);
  }

  // Horrible hack (to be removed), check to make sure shift amounts are
  // in-range.
  if ((Opcode == ISD::SHL || Opcode == ISD::SRA || Opcode == ISD::SRL) &&
      Imm >= VT.getSizeInBits())
    return 0;

  // First check if immediate type is legal. If not, we can't use the ri form.
  unsigned ResultReg = fastEmit_ri(VT, VT, Opcode, Op0, Op0IsKill, Imm);
  if (ResultReg)
    return ResultReg;
  unsigned MaterialReg = fastEmit_i(ImmType, ImmType, ISD::Constant, Imm);
  if (!MaterialReg) {
    // This is a bit ugly/slow, but failing here means falling out of
    // fast-isel, which would be very slow.
    IntegerType *ITy =
        IntegerType::get(FuncInfo.Fn->getContext(), VT.getSizeInBits());
    MaterialReg = getRegForValue(ConstantInt::get(ITy, Imm));
    if (!MaterialReg)
      return 0;
  }
  return fastEmit_rr(VT, VT, Opcode, Op0, Op0IsKill, MaterialReg,
                     /*IsKill=*/true);
}

unsigned FastISel::createResultReg(const TargetRegisterClass *RC) {
  return MRI.createVirtualRegister(RC);
}

unsigned FastISel::constrainOperandRegClass(const MCInstrDesc &II, unsigned Op,
                                            unsigned OpNum) {
  if (TargetRegisterInfo::isVirtualRegister(Op)) {
    const TargetRegisterClass *RegClass =
        TII.getRegClass(II, OpNum, &TRI, *FuncInfo.MF);
    if (!MRI.constrainRegClass(Op, RegClass)) {
      // If it's not legal to COPY between the register classes, something
      // has gone very wrong before we got here.
      unsigned NewOp = createResultReg(RegClass);
      BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc,
              TII.get(TargetOpcode::COPY), NewOp).addReg(Op);
      return NewOp;
    }
  }
  return Op;
}

unsigned FastISel::fastEmitInst_(unsigned MachineInstOpcode,
                                 const TargetRegisterClass *RC) {
  unsigned ResultReg = createResultReg(RC);
  const MCInstrDesc &II = TII.get(MachineInstOpcode);

  BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II, ResultReg);
  return ResultReg;
}

unsigned FastISel::fastEmitInst_r(unsigned MachineInstOpcode,
                                  const TargetRegisterClass *RC, unsigned Op0,
                                  bool Op0IsKill) {
  const MCInstrDesc &II = TII.get(MachineInstOpcode);

  unsigned ResultReg = createResultReg(RC);
  Op0 = constrainOperandRegClass(II, Op0, II.getNumDefs());

  if (II.getNumDefs() >= 1)
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II, ResultReg)
        .addReg(Op0, getKillRegState(Op0IsKill));
  else {
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II)
        .addReg(Op0, getKillRegState(Op0IsKill));
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc,
            TII.get(TargetOpcode::COPY), ResultReg).addReg(II.ImplicitDefs[0]);
  }

  return ResultReg;
}

unsigned FastISel::fastEmitInst_rr(unsigned MachineInstOpcode,
                                   const TargetRegisterClass *RC, unsigned Op0,
                                   bool Op0IsKill, unsigned Op1,
                                   bool Op1IsKill) {
  const MCInstrDesc &II = TII.get(MachineInstOpcode);

  unsigned ResultReg = createResultReg(RC);
  Op0 = constrainOperandRegClass(II, Op0, II.getNumDefs());
  Op1 = constrainOperandRegClass(II, Op1, II.getNumDefs() + 1);

  if (II.getNumDefs() >= 1)
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II, ResultReg)
        .addReg(Op0, getKillRegState(Op0IsKill))
        .addReg(Op1, getKillRegState(Op1IsKill));
  else {
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II)
        .addReg(Op0, getKillRegState(Op0IsKill))
        .addReg(Op1, getKillRegState(Op1IsKill));
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc,
            TII.get(TargetOpcode::COPY), ResultReg).addReg(II.ImplicitDefs[0]);
  }
  return ResultReg;
}

unsigned FastISel::fastEmitInst_rrr(unsigned MachineInstOpcode,
                                    const TargetRegisterClass *RC, unsigned Op0,
                                    bool Op0IsKill, unsigned Op1,
                                    bool Op1IsKill, unsigned Op2,
                                    bool Op2IsKill) {
  const MCInstrDesc &II = TII.get(MachineInstOpcode);

  unsigned ResultReg = createResultReg(RC);
  Op0 = constrainOperandRegClass(II, Op0, II.getNumDefs());
  Op1 = constrainOperandRegClass(II, Op1, II.getNumDefs() + 1);
  Op2 = constrainOperandRegClass(II, Op2, II.getNumDefs() + 2);

  if (II.getNumDefs() >= 1)
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II, ResultReg)
        .addReg(Op0, getKillRegState(Op0IsKill))
        .addReg(Op1, getKillRegState(Op1IsKill))
        .addReg(Op2, getKillRegState(Op2IsKill));
  else {
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II)
        .addReg(Op0, getKillRegState(Op0IsKill))
        .addReg(Op1, getKillRegState(Op1IsKill))
        .addReg(Op2, getKillRegState(Op2IsKill));
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc,
            TII.get(TargetOpcode::COPY), ResultReg).addReg(II.ImplicitDefs[0]);
  }
  return ResultReg;
}

unsigned FastISel::fastEmitInst_ri(unsigned MachineInstOpcode,
                                   const TargetRegisterClass *RC, unsigned Op0,
                                   bool Op0IsKill, uint64_t Imm) {
  const MCInstrDesc &II = TII.get(MachineInstOpcode);

  unsigned ResultReg = createResultReg(RC);
  Op0 = constrainOperandRegClass(II, Op0, II.getNumDefs());

  if (II.getNumDefs() >= 1)
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II, ResultReg)
        .addReg(Op0, getKillRegState(Op0IsKill))
        .addImm(Imm);
  else {
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II)
        .addReg(Op0, getKillRegState(Op0IsKill))
        .addImm(Imm);
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc,
            TII.get(TargetOpcode::COPY), ResultReg).addReg(II.ImplicitDefs[0]);
  }
  return ResultReg;
}

unsigned FastISel::fastEmitInst_rii(unsigned MachineInstOpcode,
                                    const TargetRegisterClass *RC, unsigned Op0,
                                    bool Op0IsKill, uint64_t Imm1,
                                    uint64_t Imm2) {
  const MCInstrDesc &II = TII.get(MachineInstOpcode);

  unsigned ResultReg = createResultReg(RC);
  Op0 = constrainOperandRegClass(II, Op0, II.getNumDefs());

  if (II.getNumDefs() >= 1)
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II, ResultReg)
        .addReg(Op0, getKillRegState(Op0IsKill))
        .addImm(Imm1)
        .addImm(Imm2);
  else {
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II)
        .addReg(Op0, getKillRegState(Op0IsKill))
        .addImm(Imm1)
        .addImm(Imm2);
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc,
            TII.get(TargetOpcode::COPY), ResultReg).addReg(II.ImplicitDefs[0]);
  }
  return ResultReg;
}

unsigned FastISel::fastEmitInst_rf(unsigned MachineInstOpcode,
                                   const TargetRegisterClass *RC, unsigned Op0,
                                   bool Op0IsKill, const ConstantFP *FPImm) {
  const MCInstrDesc &II = TII.get(MachineInstOpcode);

  unsigned ResultReg = createResultReg(RC);
  Op0 = constrainOperandRegClass(II, Op0, II.getNumDefs());

  if (II.getNumDefs() >= 1)
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II, ResultReg)
        .addReg(Op0, getKillRegState(Op0IsKill))
        .addFPImm(FPImm);
  else {
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II)
        .addReg(Op0, getKillRegState(Op0IsKill))
        .addFPImm(FPImm);
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc,
            TII.get(TargetOpcode::COPY), ResultReg).addReg(II.ImplicitDefs[0]);
  }
  return ResultReg;
}

unsigned FastISel::fastEmitInst_rri(unsigned MachineInstOpcode,
                                    const TargetRegisterClass *RC, unsigned Op0,
                                    bool Op0IsKill, unsigned Op1,
                                    bool Op1IsKill, uint64_t Imm) {
  const MCInstrDesc &II = TII.get(MachineInstOpcode);

  unsigned ResultReg = createResultReg(RC);
  Op0 = constrainOperandRegClass(II, Op0, II.getNumDefs());
  Op1 = constrainOperandRegClass(II, Op1, II.getNumDefs() + 1);

  if (II.getNumDefs() >= 1)
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II, ResultReg)
        .addReg(Op0, getKillRegState(Op0IsKill))
        .addReg(Op1, getKillRegState(Op1IsKill))
        .addImm(Imm);
  else {
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II)
        .addReg(Op0, getKillRegState(Op0IsKill))
        .addReg(Op1, getKillRegState(Op1IsKill))
        .addImm(Imm);
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc,
            TII.get(TargetOpcode::COPY), ResultReg).addReg(II.ImplicitDefs[0]);
  }
  return ResultReg;
}

unsigned FastISel::fastEmitInst_rrii(unsigned MachineInstOpcode,
                                     const TargetRegisterClass *RC,
                                     unsigned Op0, bool Op0IsKill, unsigned Op1,
                                     bool Op1IsKill, uint64_t Imm1,
                                     uint64_t Imm2) {
  const MCInstrDesc &II = TII.get(MachineInstOpcode);

  unsigned ResultReg = createResultReg(RC);
  Op0 = constrainOperandRegClass(II, Op0, II.getNumDefs());
  Op1 = constrainOperandRegClass(II, Op1, II.getNumDefs() + 1);

  if (II.getNumDefs() >= 1)
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II, ResultReg)
        .addReg(Op0, getKillRegState(Op0IsKill))
        .addReg(Op1, getKillRegState(Op1IsKill))
        .addImm(Imm1)
        .addImm(Imm2);
  else {
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II)
        .addReg(Op0, getKillRegState(Op0IsKill))
        .addReg(Op1, getKillRegState(Op1IsKill))
        .addImm(Imm1)
        .addImm(Imm2);
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc,
            TII.get(TargetOpcode::COPY), ResultReg).addReg(II.ImplicitDefs[0]);
  }
  return ResultReg;
}

unsigned FastISel::fastEmitInst_i(unsigned MachineInstOpcode,
                                  const TargetRegisterClass *RC, uint64_t Imm) {
  unsigned ResultReg = createResultReg(RC);
  const MCInstrDesc &II = TII.get(MachineInstOpcode);

  if (II.getNumDefs() >= 1)
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II, ResultReg)
        .addImm(Imm);
  else {
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II).addImm(Imm);
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc,
            TII.get(TargetOpcode::COPY), ResultReg).addReg(II.ImplicitDefs[0]);
  }
  return ResultReg;
}

unsigned FastISel::fastEmitInst_ii(unsigned MachineInstOpcode,
                                   const TargetRegisterClass *RC, uint64_t Imm1,
                                   uint64_t Imm2) {
  unsigned ResultReg = createResultReg(RC);
  const MCInstrDesc &II = TII.get(MachineInstOpcode);

  if (II.getNumDefs() >= 1)
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II, ResultReg)
        .addImm(Imm1)
        .addImm(Imm2);
  else {
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II).addImm(Imm1)
        .addImm(Imm2);
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc,
            TII.get(TargetOpcode::COPY), ResultReg).addReg(II.ImplicitDefs[0]);
  }
  return ResultReg;
}

unsigned FastISel::fastEmitInst_extractsubreg(MVT RetVT, unsigned Op0,
                                              bool Op0IsKill, uint32_t Idx) {
  unsigned ResultReg = createResultReg(TLI.getRegClassFor(RetVT));
  assert(TargetRegisterInfo::isVirtualRegister(Op0) &&
         "Cannot yet extract from physregs");
  const TargetRegisterClass *RC = MRI.getRegClass(Op0);
  MRI.constrainRegClass(Op0, TRI.getSubClassWithSubReg(RC, Idx));
  BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(TargetOpcode::COPY),
          ResultReg).addReg(Op0, getKillRegState(Op0IsKill), Idx);
  return ResultReg;
}

/// Emit MachineInstrs to compute the value of Op with all but the least
/// significant bit set to zero.
unsigned FastISel::fastEmitZExtFromI1(MVT VT, unsigned Op0, bool Op0IsKill) {
  return fastEmit_ri(VT, VT, ISD::AND, Op0, Op0IsKill, 1);
}

/// HandlePHINodesInSuccessorBlocks - Handle PHI nodes in successor blocks.
/// Emit code to ensure constants are copied into registers when needed.
/// Remember the virtual registers that need to be added to the Machine PHI
/// nodes as input.  We cannot just directly add them, because expansion
/// might result in multiple MBB's for one BB.  As such, the start of the
/// BB might correspond to a different MBB than the end.
bool FastISel::handlePHINodesInSuccessorBlocks(const BasicBlock *LLVMBB) {
  const TerminatorInst *TI = LLVMBB->getTerminator();

  SmallPtrSet<MachineBasicBlock *, 4> SuccsHandled;
  FuncInfo.OrigNumPHINodesToUpdate = FuncInfo.PHINodesToUpdate.size();

  // Check successor nodes' PHI nodes that expect a constant to be available
  // from this block.
  for (unsigned succ = 0, e = TI->getNumSuccessors(); succ != e; ++succ) {
    const BasicBlock *SuccBB = TI->getSuccessor(succ);
    if (!isa<PHINode>(SuccBB->begin()))
      continue;
    MachineBasicBlock *SuccMBB = FuncInfo.MBBMap[SuccBB];

    // If this terminator has multiple identical successors (common for
    // switches), only handle each succ once.
    if (!SuccsHandled.insert(SuccMBB).second)
      continue;

    MachineBasicBlock::iterator MBBI = SuccMBB->begin();

    // At this point we know that there is a 1-1 correspondence between LLVM PHI
    // nodes and Machine PHI nodes, but the incoming operands have not been
    // emitted yet.
    for (BasicBlock::const_iterator I = SuccBB->begin();
         const auto *PN = dyn_cast<PHINode>(I); ++I) {

      // Ignore dead phi's.
      if (PN->use_empty())
        continue;

      // Only handle legal types. Two interesting things to note here. First,
      // by bailing out early, we may leave behind some dead instructions,
      // since SelectionDAG's HandlePHINodesInSuccessorBlocks will insert its
      // own moves. Second, this check is necessary because FastISel doesn't
      // use CreateRegs to create registers, so it always creates
      // exactly one register for each non-void instruction.
      EVT VT = TLI.getValueType(PN->getType(), /*AllowUnknown=*/true);
      if (VT == MVT::Other || !TLI.isTypeLegal(VT)) {
        // Handle integer promotions, though, because they're common and easy.
        if (!(VT == MVT::i1 || VT == MVT::i8 || VT == MVT::i16)) {
          FuncInfo.PHINodesToUpdate.resize(FuncInfo.OrigNumPHINodesToUpdate);
          return false;
        }
      }

      const Value *PHIOp = PN->getIncomingValueForBlock(LLVMBB);

      // Set the DebugLoc for the copy. Prefer the location of the operand
      // if there is one; use the location of the PHI otherwise.
      DbgLoc = PN->getDebugLoc();
      if (const auto *Inst = dyn_cast<Instruction>(PHIOp))
        DbgLoc = Inst->getDebugLoc();

      unsigned Reg = getRegForValue(PHIOp);
      if (!Reg) {
        FuncInfo.PHINodesToUpdate.resize(FuncInfo.OrigNumPHINodesToUpdate);
        return false;
      }
      FuncInfo.PHINodesToUpdate.push_back(std::make_pair(MBBI++, Reg));
      DbgLoc = DebugLoc();
    }
  }

  return true;
}

bool FastISel::tryToFoldLoad(const LoadInst *LI, const Instruction *FoldInst) {
  assert(LI->hasOneUse() &&
         "tryToFoldLoad expected a LoadInst with a single use");
  // We know that the load has a single use, but don't know what it is.  If it
  // isn't one of the folded instructions, then we can't succeed here.  Handle
  // this by scanning the single-use users of the load until we get to FoldInst.
  unsigned MaxUsers = 6; // Don't scan down huge single-use chains of instrs.

  const Instruction *TheUser = LI->user_back();
  while (TheUser != FoldInst && // Scan up until we find FoldInst.
         // Stay in the right block.
         TheUser->getParent() == FoldInst->getParent() &&
         --MaxUsers) { // Don't scan too far.
    // If there are multiple or no uses of this instruction, then bail out.
    if (!TheUser->hasOneUse())
      return false;

    TheUser = TheUser->user_back();
  }

  // If we didn't find the fold instruction, then we failed to collapse the
  // sequence.
  if (TheUser != FoldInst)
    return false;

  // Don't try to fold volatile loads.  Target has to deal with alignment
  // constraints.
  if (LI->isVolatile())
    return false;

  // Figure out which vreg this is going into.  If there is no assigned vreg yet
  // then there actually was no reference to it.  Perhaps the load is referenced
  // by a dead instruction.
  unsigned LoadReg = getRegForValue(LI);
  if (!LoadReg)
    return false;

  // We can't fold if this vreg has no uses or more than one use.  Multiple uses
  // may mean that the instruction got lowered to multiple MIs, or the use of
  // the loaded value ended up being multiple operands of the result.
  if (!MRI.hasOneUse(LoadReg))
    return false;

  MachineRegisterInfo::reg_iterator RI = MRI.reg_begin(LoadReg);
  MachineInstr *User = RI->getParent();

  // Set the insertion point properly.  Folding the load can cause generation of
  // other random instructions (like sign extends) for addressing modes; make
  // sure they get inserted in a logical place before the new instruction.
  FuncInfo.InsertPt = User;
  FuncInfo.MBB = User->getParent();

  // Ask the target to try folding the load.
  return tryToFoldLoadIntoMI(User, RI.getOperandNo(), LI);
}

bool FastISel::canFoldAddIntoGEP(const User *GEP, const Value *Add) {
  // Must be an add.
  if (!isa<AddOperator>(Add))
    return false;
  // Type size needs to match.
  if (DL.getTypeSizeInBits(GEP->getType()) !=
      DL.getTypeSizeInBits(Add->getType()))
    return false;
  // Must be in the same basic block.
  if (isa<Instruction>(Add) &&
      FuncInfo.MBBMap[cast<Instruction>(Add)->getParent()] != FuncInfo.MBB)
    return false;
  // Must have a constant operand.
  return isa<ConstantInt>(cast<AddOperator>(Add)->getOperand(1));
}

MachineMemOperand *
FastISel::createMachineMemOperandFor(const Instruction *I) const {
  const Value *Ptr;
  Type *ValTy;
  unsigned Alignment;
  unsigned Flags;
  bool IsVolatile;

  if (const auto *LI = dyn_cast<LoadInst>(I)) {
    Alignment = LI->getAlignment();
    IsVolatile = LI->isVolatile();
    Flags = MachineMemOperand::MOLoad;
    Ptr = LI->getPointerOperand();
    ValTy = LI->getType();
  } else if (const auto *SI = dyn_cast<StoreInst>(I)) {
    Alignment = SI->getAlignment();
    IsVolatile = SI->isVolatile();
    Flags = MachineMemOperand::MOStore;
    Ptr = SI->getPointerOperand();
    ValTy = SI->getValueOperand()->getType();
  } else
    return nullptr;

  bool IsNonTemporal = I->getMetadata(LLVMContext::MD_nontemporal) != nullptr;
  bool IsInvariant = I->getMetadata(LLVMContext::MD_invariant_load) != nullptr;
  const MDNode *Ranges = I->getMetadata(LLVMContext::MD_range);

  AAMDNodes AAInfo;
  I->getAAMetadata(AAInfo);

  if (Alignment == 0) // Ensure that codegen never sees alignment 0.
    Alignment = DL.getABITypeAlignment(ValTy);

  unsigned Size = DL.getTypeStoreSize(ValTy);

  if (IsVolatile)
    Flags |= MachineMemOperand::MOVolatile;
  if (IsNonTemporal)
    Flags |= MachineMemOperand::MONonTemporal;
  if (IsInvariant)
    Flags |= MachineMemOperand::MOInvariant;

  return FuncInfo.MF->getMachineMemOperand(MachinePointerInfo(Ptr), Flags, Size,
                                           Alignment, AAInfo, Ranges);
}

CmpInst::Predicate FastISel::optimizeCmpPredicate(const CmpInst *CI) const {
  // If both operands are the same, then try to optimize or fold the cmp.
  CmpInst::Predicate Predicate = CI->getPredicate();
  if (CI->getOperand(0) != CI->getOperand(1))
    return Predicate;

  switch (Predicate) {
  default: llvm_unreachable("Invalid predicate!");
  case CmpInst::FCMP_FALSE: Predicate = CmpInst::FCMP_FALSE; break;
  case CmpInst::FCMP_OEQ:   Predicate = CmpInst::FCMP_ORD;   break;
  case CmpInst::FCMP_OGT:   Predicate = CmpInst::FCMP_FALSE; break;
  case CmpInst::FCMP_OGE:   Predicate = CmpInst::FCMP_ORD;   break;
  case CmpInst::FCMP_OLT:   Predicate = CmpInst::FCMP_FALSE; break;
  case CmpInst::FCMP_OLE:   Predicate = CmpInst::FCMP_ORD;   break;
  case CmpInst::FCMP_ONE:   Predicate = CmpInst::FCMP_FALSE; break;
  case CmpInst::FCMP_ORD:   Predicate = CmpInst::FCMP_ORD;   break;
  case CmpInst::FCMP_UNO:   Predicate = CmpInst::FCMP_UNO;   break;
  case CmpInst::FCMP_UEQ:   Predicate = CmpInst::FCMP_TRUE;  break;
  case CmpInst::FCMP_UGT:   Predicate = CmpInst::FCMP_UNO;   break;
  case CmpInst::FCMP_UGE:   Predicate = CmpInst::FCMP_TRUE;  break;
  case CmpInst::FCMP_ULT:   Predicate = CmpInst::FCMP_UNO;   break;
  case CmpInst::FCMP_ULE:   Predicate = CmpInst::FCMP_TRUE;  break;
  case CmpInst::FCMP_UNE:   Predicate = CmpInst::FCMP_UNO;   break;
  case CmpInst::FCMP_TRUE:  Predicate = CmpInst::FCMP_TRUE;  break;

  case CmpInst::ICMP_EQ:    Predicate = CmpInst::FCMP_TRUE;  break;
  case CmpInst::ICMP_NE:    Predicate = CmpInst::FCMP_FALSE; break;
  case CmpInst::ICMP_UGT:   Predicate = CmpInst::FCMP_FALSE; break;
  case CmpInst::ICMP_UGE:   Predicate = CmpInst::FCMP_TRUE;  break;
  case CmpInst::ICMP_ULT:   Predicate = CmpInst::FCMP_FALSE; break;
  case CmpInst::ICMP_ULE:   Predicate = CmpInst::FCMP_TRUE;  break;
  case CmpInst::ICMP_SGT:   Predicate = CmpInst::FCMP_FALSE; break;
  case CmpInst::ICMP_SGE:   Predicate = CmpInst::FCMP_TRUE;  break;
  case CmpInst::ICMP_SLT:   Predicate = CmpInst::FCMP_FALSE; break;
  case CmpInst::ICMP_SLE:   Predicate = CmpInst::FCMP_TRUE;  break;
  }

  return Predicate;
}