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
|
//===- ConstantFold.cpp - LLVM constant folder ----------------------------===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file implements folding of constants for LLVM. This implements the
// (internal) ConstantFold.h interface, which is used by the
// ConstantExpr::get* methods to automatically fold constants when possible.
//
// The current constant folding implementation is implemented in two pieces: the
// pieces that don't need DataLayout, and the pieces that do. This is to avoid
// a dependence in IR on Target.
//
//===----------------------------------------------------------------------===//
#include "ConstantFold.h"
#include "llvm/ADT/SmallVector.h"
#include "llvm/IR/Constants.h"
#include "llvm/IR/DerivedTypes.h"
#include "llvm/IR/Function.h"
#include "llvm/IR/GetElementPtrTypeIterator.h"
#include "llvm/IR/GlobalAlias.h"
#include "llvm/IR/GlobalVariable.h"
#include "llvm/IR/Instructions.h"
#include "llvm/IR/Operator.h"
#include "llvm/IR/PatternMatch.h"
#include "llvm/Support/Compiler.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/ManagedStatic.h"
#include "llvm/Support/MathExtras.h"
#include <limits>
using namespace llvm;
using namespace llvm::PatternMatch;
//===----------------------------------------------------------------------===//
// ConstantFold*Instruction Implementations
//===----------------------------------------------------------------------===//
/// BitCastConstantVector - Convert the specified vector Constant node to the
/// specified vector type. At this point, we know that the elements of the
/// input vector constant are all simple integer or FP values.
static Constant *BitCastConstantVector(Constant *CV, VectorType *DstTy) {
if (CV->isAllOnesValue()) return Constant::getAllOnesValue(DstTy);
if (CV->isNullValue()) return Constant::getNullValue(DstTy);
// If this cast changes element count then we can't handle it here:
// doing so requires endianness information. This should be handled by
// Analysis/ConstantFolding.cpp
unsigned NumElts = DstTy->getNumElements();
if (NumElts != CV->getType()->getVectorNumElements())
return nullptr;
Type *DstEltTy = DstTy->getElementType();
SmallVector<Constant*, 16> Result;
Type *Ty = IntegerType::get(CV->getContext(), 32);
for (unsigned i = 0; i != NumElts; ++i) {
Constant *C =
ConstantExpr::getExtractElement(CV, ConstantInt::get(Ty, i));
C = ConstantExpr::getBitCast(C, DstEltTy);
Result.push_back(C);
}
return ConstantVector::get(Result);
}
/// This function determines which opcode to use to fold two constant cast
/// expressions together. It uses CastInst::isEliminableCastPair to determine
/// the opcode. Consequently its just a wrapper around that function.
/// @brief Determine if it is valid to fold a cast of a cast
static unsigned
foldConstantCastPair(
unsigned opc, ///< opcode of the second cast constant expression
ConstantExpr *Op, ///< the first cast constant expression
Type *DstTy ///< destination type of the first cast
) {
assert(Op && Op->isCast() && "Can't fold cast of cast without a cast!");
assert(DstTy && DstTy->isFirstClassType() && "Invalid cast destination type");
assert(CastInst::isCast(opc) && "Invalid cast opcode");
// The the types and opcodes for the two Cast constant expressions
Type *SrcTy = Op->getOperand(0)->getType();
Type *MidTy = Op->getType();
Instruction::CastOps firstOp = Instruction::CastOps(Op->getOpcode());
Instruction::CastOps secondOp = Instruction::CastOps(opc);
// Assume that pointers are never more than 64 bits wide, and only use this
// for the middle type. Otherwise we could end up folding away illegal
// bitcasts between address spaces with different sizes.
IntegerType *FakeIntPtrTy = Type::getInt64Ty(DstTy->getContext());
// Let CastInst::isEliminableCastPair do the heavy lifting.
return CastInst::isEliminableCastPair(firstOp, secondOp, SrcTy, MidTy, DstTy,
nullptr, FakeIntPtrTy, nullptr);
}
static Constant *FoldBitCast(Constant *V, Type *DestTy) {
Type *SrcTy = V->getType();
if (SrcTy == DestTy)
return V; // no-op cast
// Check to see if we are casting a pointer to an aggregate to a pointer to
// the first element. If so, return the appropriate GEP instruction.
if (PointerType *PTy = dyn_cast<PointerType>(V->getType()))
if (PointerType *DPTy = dyn_cast<PointerType>(DestTy))
if (PTy->getAddressSpace() == DPTy->getAddressSpace()
&& DPTy->getElementType()->isSized()) {
SmallVector<Value*, 8> IdxList;
Value *Zero =
Constant::getNullValue(Type::getInt32Ty(DPTy->getContext()));
IdxList.push_back(Zero);
Type *ElTy = PTy->getElementType();
while (ElTy != DPTy->getElementType()) {
if (StructType *STy = dyn_cast<StructType>(ElTy)) {
if (STy->getNumElements() == 0) break;
ElTy = STy->getElementType(0);
IdxList.push_back(Zero);
} else if (SequentialType *STy =
dyn_cast<SequentialType>(ElTy)) {
if (ElTy->isPointerTy()) break; // Can't index into pointers!
ElTy = STy->getElementType();
IdxList.push_back(Zero);
} else {
break;
}
}
if (ElTy == DPTy->getElementType())
// This GEP is inbounds because all indices are zero.
return ConstantExpr::getInBoundsGetElementPtr(V, IdxList);
}
// Handle casts from one vector constant to another. We know that the src
// and dest type have the same size (otherwise its an illegal cast).
if (VectorType *DestPTy = dyn_cast<VectorType>(DestTy)) {
if (VectorType *SrcTy = dyn_cast<VectorType>(V->getType())) {
assert(DestPTy->getBitWidth() == SrcTy->getBitWidth() &&
"Not cast between same sized vectors!");
SrcTy = nullptr;
// First, check for null. Undef is already handled.
if (isa<ConstantAggregateZero>(V))
return Constant::getNullValue(DestTy);
// Handle ConstantVector and ConstantAggregateVector.
return BitCastConstantVector(V, DestPTy);
}
// Canonicalize scalar-to-vector bitcasts into vector-to-vector bitcasts
// This allows for other simplifications (although some of them
// can only be handled by Analysis/ConstantFolding.cpp).
if (isa<ConstantInt>(V) || isa<ConstantFP>(V))
return ConstantExpr::getBitCast(ConstantVector::get(V), DestPTy);
}
// Finally, implement bitcast folding now. The code below doesn't handle
// bitcast right.
if (isa<ConstantPointerNull>(V)) // ptr->ptr cast.
return ConstantPointerNull::get(cast<PointerType>(DestTy));
// Handle integral constant input.
if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) {
if (DestTy->isIntegerTy())
// Integral -> Integral. This is a no-op because the bit widths must
// be the same. Consequently, we just fold to V.
return V;
if (DestTy->isFloatingPointTy())
return ConstantFP::get(DestTy->getContext(),
APFloat(DestTy->getFltSemantics(),
CI->getValue()));
// Otherwise, can't fold this (vector?)
return nullptr;
}
// Handle ConstantFP input: FP -> Integral.
if (ConstantFP *FP = dyn_cast<ConstantFP>(V))
return ConstantInt::get(FP->getContext(),
FP->getValueAPF().bitcastToAPInt());
return nullptr;
}
/// ExtractConstantBytes - V is an integer constant which only has a subset of
/// its bytes used. The bytes used are indicated by ByteStart (which is the
/// first byte used, counting from the least significant byte) and ByteSize,
/// which is the number of bytes used.
///
/// This function analyzes the specified constant to see if the specified byte
/// range can be returned as a simplified constant. If so, the constant is
/// returned, otherwise null is returned.
///
static Constant *ExtractConstantBytes(Constant *C, unsigned ByteStart,
unsigned ByteSize) {
assert(C->getType()->isIntegerTy() &&
(cast<IntegerType>(C->getType())->getBitWidth() & 7) == 0 &&
"Non-byte sized integer input");
unsigned CSize = cast<IntegerType>(C->getType())->getBitWidth()/8;
assert(ByteSize && "Must be accessing some piece");
assert(ByteStart+ByteSize <= CSize && "Extracting invalid piece from input");
assert(ByteSize != CSize && "Should not extract everything");
// Constant Integers are simple.
if (ConstantInt *CI = dyn_cast<ConstantInt>(C)) {
APInt V = CI->getValue();
if (ByteStart)
V = V.lshr(ByteStart*8);
V = V.trunc(ByteSize*8);
return ConstantInt::get(CI->getContext(), V);
}
// In the input is a constant expr, we might be able to recursively simplify.
// If not, we definitely can't do anything.
ConstantExpr *CE = dyn_cast<ConstantExpr>(C);
if (!CE) return nullptr;
switch (CE->getOpcode()) {
default: return nullptr;
case Instruction::Or: {
Constant *RHS = ExtractConstantBytes(CE->getOperand(1), ByteStart,ByteSize);
if (!RHS)
return nullptr;
// X | -1 -> -1.
if (ConstantInt *RHSC = dyn_cast<ConstantInt>(RHS))
if (RHSC->isAllOnesValue())
return RHSC;
Constant *LHS = ExtractConstantBytes(CE->getOperand(0), ByteStart,ByteSize);
if (!LHS)
return nullptr;
return ConstantExpr::getOr(LHS, RHS);
}
case Instruction::And: {
Constant *RHS = ExtractConstantBytes(CE->getOperand(1), ByteStart,ByteSize);
if (!RHS)
return nullptr;
// X & 0 -> 0.
if (RHS->isNullValue())
return RHS;
Constant *LHS = ExtractConstantBytes(CE->getOperand(0), ByteStart,ByteSize);
if (!LHS)
return nullptr;
return ConstantExpr::getAnd(LHS, RHS);
}
case Instruction::LShr: {
ConstantInt *Amt = dyn_cast<ConstantInt>(CE->getOperand(1));
if (!Amt)
return nullptr;
unsigned ShAmt = Amt->getZExtValue();
// Cannot analyze non-byte shifts.
if ((ShAmt & 7) != 0)
return nullptr;
ShAmt >>= 3;
// If the extract is known to be all zeros, return zero.
if (ByteStart >= CSize-ShAmt)
return Constant::getNullValue(IntegerType::get(CE->getContext(),
ByteSize*8));
// If the extract is known to be fully in the input, extract it.
if (ByteStart+ByteSize+ShAmt <= CSize)
return ExtractConstantBytes(CE->getOperand(0), ByteStart+ShAmt, ByteSize);
// TODO: Handle the 'partially zero' case.
return nullptr;
}
case Instruction::Shl: {
ConstantInt *Amt = dyn_cast<ConstantInt>(CE->getOperand(1));
if (!Amt)
return nullptr;
unsigned ShAmt = Amt->getZExtValue();
// Cannot analyze non-byte shifts.
if ((ShAmt & 7) != 0)
return nullptr;
ShAmt >>= 3;
// If the extract is known to be all zeros, return zero.
if (ByteStart+ByteSize <= ShAmt)
return Constant::getNullValue(IntegerType::get(CE->getContext(),
ByteSize*8));
// If the extract is known to be fully in the input, extract it.
if (ByteStart >= ShAmt)
return ExtractConstantBytes(CE->getOperand(0), ByteStart-ShAmt, ByteSize);
// TODO: Handle the 'partially zero' case.
return nullptr;
}
case Instruction::ZExt: {
unsigned SrcBitSize =
cast<IntegerType>(CE->getOperand(0)->getType())->getBitWidth();
// If extracting something that is completely zero, return 0.
if (ByteStart*8 >= SrcBitSize)
return Constant::getNullValue(IntegerType::get(CE->getContext(),
ByteSize*8));
// If exactly extracting the input, return it.
if (ByteStart == 0 && ByteSize*8 == SrcBitSize)
return CE->getOperand(0);
// If extracting something completely in the input, if if the input is a
// multiple of 8 bits, recurse.
if ((SrcBitSize&7) == 0 && (ByteStart+ByteSize)*8 <= SrcBitSize)
return ExtractConstantBytes(CE->getOperand(0), ByteStart, ByteSize);
// Otherwise, if extracting a subset of the input, which is not multiple of
// 8 bits, do a shift and trunc to get the bits.
if ((ByteStart+ByteSize)*8 < SrcBitSize) {
assert((SrcBitSize&7) && "Shouldn't get byte sized case here");
Constant *Res = CE->getOperand(0);
if (ByteStart)
Res = ConstantExpr::getLShr(Res,
ConstantInt::get(Res->getType(), ByteStart*8));
return ConstantExpr::getTrunc(Res, IntegerType::get(C->getContext(),
ByteSize*8));
}
// TODO: Handle the 'partially zero' case.
return nullptr;
}
}
}
/// getFoldedSizeOf - Return a ConstantExpr with type DestTy for sizeof
/// on Ty, with any known factors factored out. If Folded is false,
/// return null if no factoring was possible, to avoid endlessly
/// bouncing an unfoldable expression back into the top-level folder.
///
static Constant *getFoldedSizeOf(Type *Ty, Type *DestTy,
bool Folded) {
if (ArrayType *ATy = dyn_cast<ArrayType>(Ty)) {
Constant *N = ConstantInt::get(DestTy, ATy->getNumElements());
Constant *E = getFoldedSizeOf(ATy->getElementType(), DestTy, true);
return ConstantExpr::getNUWMul(E, N);
}
if (StructType *STy = dyn_cast<StructType>(Ty))
if (!STy->isPacked()) {
unsigned NumElems = STy->getNumElements();
// An empty struct has size zero.
if (NumElems == 0)
return ConstantExpr::getNullValue(DestTy);
// Check for a struct with all members having the same size.
Constant *MemberSize =
getFoldedSizeOf(STy->getElementType(0), DestTy, true);
bool AllSame = true;
for (unsigned i = 1; i != NumElems; ++i)
if (MemberSize !=
getFoldedSizeOf(STy->getElementType(i), DestTy, true)) {
AllSame = false;
break;
}
if (AllSame) {
Constant *N = ConstantInt::get(DestTy, NumElems);
return ConstantExpr::getNUWMul(MemberSize, N);
}
}
// Pointer size doesn't depend on the pointee type, so canonicalize them
// to an arbitrary pointee.
if (PointerType *PTy = dyn_cast<PointerType>(Ty))
if (!PTy->getElementType()->isIntegerTy(1))
return
getFoldedSizeOf(PointerType::get(IntegerType::get(PTy->getContext(), 1),
PTy->getAddressSpace()),
DestTy, true);
// If there's no interesting folding happening, bail so that we don't create
// a constant that looks like it needs folding but really doesn't.
if (!Folded)
return nullptr;
// Base case: Get a regular sizeof expression.
Constant *C = ConstantExpr::getSizeOf(Ty);
C = ConstantExpr::getCast(CastInst::getCastOpcode(C, false,
DestTy, false),
C, DestTy);
return C;
}
/// getFoldedAlignOf - Return a ConstantExpr with type DestTy for alignof
/// on Ty, with any known factors factored out. If Folded is false,
/// return null if no factoring was possible, to avoid endlessly
/// bouncing an unfoldable expression back into the top-level folder.
///
static Constant *getFoldedAlignOf(Type *Ty, Type *DestTy,
bool Folded) {
// The alignment of an array is equal to the alignment of the
// array element. Note that this is not always true for vectors.
if (ArrayType *ATy = dyn_cast<ArrayType>(Ty)) {
Constant *C = ConstantExpr::getAlignOf(ATy->getElementType());
C = ConstantExpr::getCast(CastInst::getCastOpcode(C, false,
DestTy,
false),
C, DestTy);
return C;
}
if (StructType *STy = dyn_cast<StructType>(Ty)) {
// Packed structs always have an alignment of 1.
if (STy->isPacked())
return ConstantInt::get(DestTy, 1);
// Otherwise, struct alignment is the maximum alignment of any member.
// Without target data, we can't compare much, but we can check to see
// if all the members have the same alignment.
unsigned NumElems = STy->getNumElements();
// An empty struct has minimal alignment.
if (NumElems == 0)
return ConstantInt::get(DestTy, 1);
// Check for a struct with all members having the same alignment.
Constant *MemberAlign =
getFoldedAlignOf(STy->getElementType(0), DestTy, true);
bool AllSame = true;
for (unsigned i = 1; i != NumElems; ++i)
if (MemberAlign != getFoldedAlignOf(STy->getElementType(i), DestTy, true)) {
AllSame = false;
break;
}
if (AllSame)
return MemberAlign;
}
// Pointer alignment doesn't depend on the pointee type, so canonicalize them
// to an arbitrary pointee.
if (PointerType *PTy = dyn_cast<PointerType>(Ty))
if (!PTy->getElementType()->isIntegerTy(1))
return
getFoldedAlignOf(PointerType::get(IntegerType::get(PTy->getContext(),
1),
PTy->getAddressSpace()),
DestTy, true);
// If there's no interesting folding happening, bail so that we don't create
// a constant that looks like it needs folding but really doesn't.
if (!Folded)
return nullptr;
// Base case: Get a regular alignof expression.
Constant *C = ConstantExpr::getAlignOf(Ty);
C = ConstantExpr::getCast(CastInst::getCastOpcode(C, false,
DestTy, false),
C, DestTy);
return C;
}
/// getFoldedOffsetOf - Return a ConstantExpr with type DestTy for offsetof
/// on Ty and FieldNo, with any known factors factored out. If Folded is false,
/// return null if no factoring was possible, to avoid endlessly
/// bouncing an unfoldable expression back into the top-level folder.
///
static Constant *getFoldedOffsetOf(Type *Ty, Constant *FieldNo,
Type *DestTy,
bool Folded) {
if (ArrayType *ATy = dyn_cast<ArrayType>(Ty)) {
Constant *N = ConstantExpr::getCast(CastInst::getCastOpcode(FieldNo, false,
DestTy, false),
FieldNo, DestTy);
Constant *E = getFoldedSizeOf(ATy->getElementType(), DestTy, true);
return ConstantExpr::getNUWMul(E, N);
}
if (StructType *STy = dyn_cast<StructType>(Ty))
if (!STy->isPacked()) {
unsigned NumElems = STy->getNumElements();
// An empty struct has no members.
if (NumElems == 0)
return nullptr;
// Check for a struct with all members having the same size.
Constant *MemberSize =
getFoldedSizeOf(STy->getElementType(0), DestTy, true);
bool AllSame = true;
for (unsigned i = 1; i != NumElems; ++i)
if (MemberSize !=
getFoldedSizeOf(STy->getElementType(i), DestTy, true)) {
AllSame = false;
break;
}
if (AllSame) {
Constant *N = ConstantExpr::getCast(CastInst::getCastOpcode(FieldNo,
false,
DestTy,
false),
FieldNo, DestTy);
return ConstantExpr::getNUWMul(MemberSize, N);
}
}
// If there's no interesting folding happening, bail so that we don't create
// a constant that looks like it needs folding but really doesn't.
if (!Folded)
return nullptr;
// Base case: Get a regular offsetof expression.
Constant *C = ConstantExpr::getOffsetOf(Ty, FieldNo);
C = ConstantExpr::getCast(CastInst::getCastOpcode(C, false,
DestTy, false),
C, DestTy);
return C;
}
Constant *llvm::ConstantFoldCastInstruction(unsigned opc, Constant *V,
Type *DestTy) {
if (isa<UndefValue>(V)) {
// zext(undef) = 0, because the top bits will be zero.
// sext(undef) = 0, because the top bits will all be the same.
// [us]itofp(undef) = 0, because the result value is bounded.
if (opc == Instruction::ZExt || opc == Instruction::SExt ||
opc == Instruction::UIToFP || opc == Instruction::SIToFP)
return Constant::getNullValue(DestTy);
return UndefValue::get(DestTy);
}
if (V->isNullValue() && !DestTy->isX86_MMXTy())
return Constant::getNullValue(DestTy);
// If the cast operand is a constant expression, there's a few things we can
// do to try to simplify it.
if (ConstantExpr *CE = dyn_cast<ConstantExpr>(V)) {
if (CE->isCast()) {
// Try hard to fold cast of cast because they are often eliminable.
if (unsigned newOpc = foldConstantCastPair(opc, CE, DestTy))
return ConstantExpr::getCast(newOpc, CE->getOperand(0), DestTy);
} else if (CE->getOpcode() == Instruction::GetElementPtr &&
// Do not fold addrspacecast (gep 0, .., 0). It might make the
// addrspacecast uncanonicalized.
opc != Instruction::AddrSpaceCast) {
// If all of the indexes in the GEP are null values, there is no pointer
// adjustment going on. We might as well cast the source pointer.
bool isAllNull = true;
for (unsigned i = 1, e = CE->getNumOperands(); i != e; ++i)
if (!CE->getOperand(i)->isNullValue()) {
isAllNull = false;
break;
}
if (isAllNull)
// This is casting one pointer type to another, always BitCast
return ConstantExpr::getPointerCast(CE->getOperand(0), DestTy);
}
}
// If the cast operand is a constant vector, perform the cast by
// operating on each element. In the cast of bitcasts, the element
// count may be mismatched; don't attempt to handle that here.
if ((isa<ConstantVector>(V) || isa<ConstantDataVector>(V)) &&
DestTy->isVectorTy() &&
DestTy->getVectorNumElements() == V->getType()->getVectorNumElements()) {
SmallVector<Constant*, 16> res;
VectorType *DestVecTy = cast<VectorType>(DestTy);
Type *DstEltTy = DestVecTy->getElementType();
Type *Ty = IntegerType::get(V->getContext(), 32);
for (unsigned i = 0, e = V->getType()->getVectorNumElements(); i != e; ++i) {
Constant *C =
ConstantExpr::getExtractElement(V, ConstantInt::get(Ty, i));
res.push_back(ConstantExpr::getCast(opc, C, DstEltTy));
}
return ConstantVector::get(res);
}
// We actually have to do a cast now. Perform the cast according to the
// opcode specified.
switch (opc) {
default:
llvm_unreachable("Failed to cast constant expression");
case Instruction::FPTrunc:
case Instruction::FPExt:
if (ConstantFP *FPC = dyn_cast<ConstantFP>(V)) {
bool ignored;
APFloat Val = FPC->getValueAPF();
Val.convert(DestTy->isHalfTy() ? APFloat::IEEEhalf :
DestTy->isFloatTy() ? APFloat::IEEEsingle :
DestTy->isDoubleTy() ? APFloat::IEEEdouble :
DestTy->isX86_FP80Ty() ? APFloat::x87DoubleExtended :
DestTy->isFP128Ty() ? APFloat::IEEEquad :
DestTy->isPPC_FP128Ty() ? APFloat::PPCDoubleDouble :
APFloat::Bogus,
APFloat::rmNearestTiesToEven, &ignored);
return ConstantFP::get(V->getContext(), Val);
}
return nullptr; // Can't fold.
case Instruction::FPToUI:
case Instruction::FPToSI:
if (ConstantFP *FPC = dyn_cast<ConstantFP>(V)) {
const APFloat &V = FPC->getValueAPF();
bool ignored;
uint64_t x[2];
uint32_t DestBitWidth = cast<IntegerType>(DestTy)->getBitWidth();
if (APFloat::opInvalidOp ==
V.convertToInteger(x, DestBitWidth, opc==Instruction::FPToSI,
APFloat::rmTowardZero, &ignored)) {
// Undefined behavior invoked - the destination type can't represent
// the input constant.
return UndefValue::get(DestTy);
}
APInt Val(DestBitWidth, x);
return ConstantInt::get(FPC->getContext(), Val);
}
return nullptr; // Can't fold.
case Instruction::IntToPtr: //always treated as unsigned
if (V->isNullValue()) // Is it an integral null value?
return ConstantPointerNull::get(cast<PointerType>(DestTy));
return nullptr; // Other pointer types cannot be casted
case Instruction::PtrToInt: // always treated as unsigned
// Is it a null pointer value?
if (V->isNullValue())
return ConstantInt::get(DestTy, 0);
// If this is a sizeof-like expression, pull out multiplications by
// known factors to expose them to subsequent folding. If it's an
// alignof-like expression, factor out known factors.
if (ConstantExpr *CE = dyn_cast<ConstantExpr>(V))
if (CE->getOpcode() == Instruction::GetElementPtr &&
CE->getOperand(0)->isNullValue()) {
Type *Ty =
cast<PointerType>(CE->getOperand(0)->getType())->getElementType();
if (CE->getNumOperands() == 2) {
// Handle a sizeof-like expression.
Constant *Idx = CE->getOperand(1);
bool isOne = isa<ConstantInt>(Idx) && cast<ConstantInt>(Idx)->isOne();
if (Constant *C = getFoldedSizeOf(Ty, DestTy, !isOne)) {
Idx = ConstantExpr::getCast(CastInst::getCastOpcode(Idx, true,
DestTy, false),
Idx, DestTy);
return ConstantExpr::getMul(C, Idx);
}
} else if (CE->getNumOperands() == 3 &&
CE->getOperand(1)->isNullValue()) {
// Handle an alignof-like expression.
if (StructType *STy = dyn_cast<StructType>(Ty))
if (!STy->isPacked()) {
ConstantInt *CI = cast<ConstantInt>(CE->getOperand(2));
if (CI->isOne() &&
STy->getNumElements() == 2 &&
STy->getElementType(0)->isIntegerTy(1)) {
return getFoldedAlignOf(STy->getElementType(1), DestTy, false);
}
}
// Handle an offsetof-like expression.
if (Ty->isStructTy() || Ty->isArrayTy()) {
if (Constant *C = getFoldedOffsetOf(Ty, CE->getOperand(2),
DestTy, false))
return C;
}
}
}
// Other pointer types cannot be casted
return nullptr;
case Instruction::UIToFP:
case Instruction::SIToFP:
if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) {
APInt api = CI->getValue();
APFloat apf(DestTy->getFltSemantics(),
APInt::getNullValue(DestTy->getPrimitiveSizeInBits()));
if (APFloat::opOverflow &
apf.convertFromAPInt(api, opc==Instruction::SIToFP,
APFloat::rmNearestTiesToEven)) {
// Undefined behavior invoked - the destination type can't represent
// the input constant.
return UndefValue::get(DestTy);
}
return ConstantFP::get(V->getContext(), apf);
}
return nullptr;
case Instruction::ZExt:
if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) {
uint32_t BitWidth = cast<IntegerType>(DestTy)->getBitWidth();
return ConstantInt::get(V->getContext(),
CI->getValue().zext(BitWidth));
}
return nullptr;
case Instruction::SExt:
if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) {
uint32_t BitWidth = cast<IntegerType>(DestTy)->getBitWidth();
return ConstantInt::get(V->getContext(),
CI->getValue().sext(BitWidth));
}
return nullptr;
case Instruction::Trunc: {
if (V->getType()->isVectorTy())
return nullptr;
uint32_t DestBitWidth = cast<IntegerType>(DestTy)->getBitWidth();
if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) {
return ConstantInt::get(V->getContext(),
CI->getValue().trunc(DestBitWidth));
}
// The input must be a constantexpr. See if we can simplify this based on
// the bytes we are demanding. Only do this if the source and dest are an
// even multiple of a byte.
if ((DestBitWidth & 7) == 0 &&
(cast<IntegerType>(V->getType())->getBitWidth() & 7) == 0)
if (Constant *Res = ExtractConstantBytes(V, 0, DestBitWidth / 8))
return Res;
return nullptr;
}
case Instruction::BitCast:
return FoldBitCast(V, DestTy);
case Instruction::AddrSpaceCast:
return nullptr;
}
}
Constant *llvm::ConstantFoldSelectInstruction(Constant *Cond,
Constant *V1, Constant *V2) {
// Check for i1 and vector true/false conditions.
if (Cond->isNullValue()) return V2;
if (Cond->isAllOnesValue()) return V1;
// If the condition is a vector constant, fold the result elementwise.
if (ConstantVector *CondV = dyn_cast<ConstantVector>(Cond)) {
SmallVector<Constant*, 16> Result;
Type *Ty = IntegerType::get(CondV->getContext(), 32);
for (unsigned i = 0, e = V1->getType()->getVectorNumElements(); i != e;++i){
Constant *V;
Constant *V1Element = ConstantExpr::getExtractElement(V1,
ConstantInt::get(Ty, i));
Constant *V2Element = ConstantExpr::getExtractElement(V2,
ConstantInt::get(Ty, i));
Constant *Cond = dyn_cast<Constant>(CondV->getOperand(i));
if (V1Element == V2Element) {
V = V1Element;
} else if (isa<UndefValue>(Cond)) {
V = isa<UndefValue>(V1Element) ? V1Element : V2Element;
} else {
if (!isa<ConstantInt>(Cond)) break;
V = Cond->isNullValue() ? V2Element : V1Element;
}
Result.push_back(V);
}
// If we were able to build the vector, return it.
if (Result.size() == V1->getType()->getVectorNumElements())
return ConstantVector::get(Result);
}
if (isa<UndefValue>(Cond)) {
if (isa<UndefValue>(V1)) return V1;
return V2;
}
if (isa<UndefValue>(V1)) return V2;
if (isa<UndefValue>(V2)) return V1;
if (V1 == V2) return V1;
if (ConstantExpr *TrueVal = dyn_cast<ConstantExpr>(V1)) {
if (TrueVal->getOpcode() == Instruction::Select)
if (TrueVal->getOperand(0) == Cond)
return ConstantExpr::getSelect(Cond, TrueVal->getOperand(1), V2);
}
if (ConstantExpr *FalseVal = dyn_cast<ConstantExpr>(V2)) {
if (FalseVal->getOpcode() == Instruction::Select)
if (FalseVal->getOperand(0) == Cond)
return ConstantExpr::getSelect(Cond, V1, FalseVal->getOperand(2));
}
return nullptr;
}
Constant *llvm::ConstantFoldExtractElementInstruction(Constant *Val,
Constant *Idx) {
if (isa<UndefValue>(Val)) // ee(undef, x) -> undef
return UndefValue::get(Val->getType()->getVectorElementType());
if (Val->isNullValue()) // ee(zero, x) -> zero
return Constant::getNullValue(Val->getType()->getVectorElementType());
// ee({w,x,y,z}, undef) -> undef
if (isa<UndefValue>(Idx))
return UndefValue::get(Val->getType()->getVectorElementType());
if (ConstantInt *CIdx = dyn_cast<ConstantInt>(Idx)) {
uint64_t Index = CIdx->getZExtValue();
// ee({w,x,y,z}, wrong_value) -> undef
if (Index >= Val->getType()->getVectorNumElements())
return UndefValue::get(Val->getType()->getVectorElementType());
return Val->getAggregateElement(Index);
}
return nullptr;
}
Constant *llvm::ConstantFoldInsertElementInstruction(Constant *Val,
Constant *Elt,
Constant *Idx) {
ConstantInt *CIdx = dyn_cast<ConstantInt>(Idx);
if (!CIdx) return nullptr;
const APInt &IdxVal = CIdx->getValue();
SmallVector<Constant*, 16> Result;
Type *Ty = IntegerType::get(Val->getContext(), 32);
for (unsigned i = 0, e = Val->getType()->getVectorNumElements(); i != e; ++i){
if (i == IdxVal) {
Result.push_back(Elt);
continue;
}
Constant *C =
ConstantExpr::getExtractElement(Val, ConstantInt::get(Ty, i));
Result.push_back(C);
}
return ConstantVector::get(Result);
}
Constant *llvm::ConstantFoldShuffleVectorInstruction(Constant *V1,
Constant *V2,
Constant *Mask) {
unsigned MaskNumElts = Mask->getType()->getVectorNumElements();
Type *EltTy = V1->getType()->getVectorElementType();
// Undefined shuffle mask -> undefined value.
if (isa<UndefValue>(Mask))
return UndefValue::get(VectorType::get(EltTy, MaskNumElts));
// Don't break the bitcode reader hack.
if (isa<ConstantExpr>(Mask)) return nullptr;
unsigned SrcNumElts = V1->getType()->getVectorNumElements();
// Loop over the shuffle mask, evaluating each element.
SmallVector<Constant*, 32> Result;
for (unsigned i = 0; i != MaskNumElts; ++i) {
int Elt = ShuffleVectorInst::getMaskValue(Mask, i);
if (Elt == -1) {
Result.push_back(UndefValue::get(EltTy));
continue;
}
Constant *InElt;
if (unsigned(Elt) >= SrcNumElts*2)
InElt = UndefValue::get(EltTy);
else if (unsigned(Elt) >= SrcNumElts) {
Type *Ty = IntegerType::get(V2->getContext(), 32);
InElt =
ConstantExpr::getExtractElement(V2,
ConstantInt::get(Ty, Elt - SrcNumElts));
} else {
Type *Ty = IntegerType::get(V1->getContext(), 32);
InElt = ConstantExpr::getExtractElement(V1, ConstantInt::get(Ty, Elt));
}
Result.push_back(InElt);
}
return ConstantVector::get(Result);
}
Constant *llvm::ConstantFoldExtractValueInstruction(Constant *Agg,
ArrayRef<unsigned> Idxs) {
// Base case: no indices, so return the entire value.
if (Idxs.empty())
return Agg;
if (Constant *C = Agg->getAggregateElement(Idxs[0]))
return ConstantFoldExtractValueInstruction(C, Idxs.slice(1));
return nullptr;
}
Constant *llvm::ConstantFoldInsertValueInstruction(Constant *Agg,
Constant *Val,
ArrayRef<unsigned> Idxs) {
// Base case: no indices, so replace the entire value.
if (Idxs.empty())
return Val;
unsigned NumElts;
if (StructType *ST = dyn_cast<StructType>(Agg->getType()))
NumElts = ST->getNumElements();
else if (ArrayType *AT = dyn_cast<ArrayType>(Agg->getType()))
NumElts = AT->getNumElements();
else
NumElts = Agg->getType()->getVectorNumElements();
SmallVector<Constant*, 32> Result;
for (unsigned i = 0; i != NumElts; ++i) {
Constant *C = Agg->getAggregateElement(i);
if (!C) return nullptr;
if (Idxs[0] == i)
C = ConstantFoldInsertValueInstruction(C, Val, Idxs.slice(1));
Result.push_back(C);
}
if (StructType *ST = dyn_cast<StructType>(Agg->getType()))
return ConstantStruct::get(ST, Result);
if (ArrayType *AT = dyn_cast<ArrayType>(Agg->getType()))
return ConstantArray::get(AT, Result);
return ConstantVector::get(Result);
}
Constant *llvm::ConstantFoldBinaryInstruction(unsigned Opcode,
Constant *C1, Constant *C2) {
// Handle UndefValue up front.
if (isa<UndefValue>(C1) || isa<UndefValue>(C2)) {
switch (Opcode) {
case Instruction::Xor:
if (isa<UndefValue>(C1) && isa<UndefValue>(C2))
// Handle undef ^ undef -> 0 special case. This is a common
// idiom (misuse).
return Constant::getNullValue(C1->getType());
// Fallthrough
case Instruction::Add:
case Instruction::Sub:
return UndefValue::get(C1->getType());
case Instruction::And:
if (isa<UndefValue>(C1) && isa<UndefValue>(C2)) // undef & undef -> undef
return C1;
return Constant::getNullValue(C1->getType()); // undef & X -> 0
case Instruction::Mul: {
// undef * undef -> undef
if (isa<UndefValue>(C1) && isa<UndefValue>(C2))
return C1;
const APInt *CV;
// X * undef -> undef if X is odd
if (match(C1, m_APInt(CV)) || match(C2, m_APInt(CV)))
if ((*CV)[0])
return UndefValue::get(C1->getType());
// X * undef -> 0 otherwise
return Constant::getNullValue(C1->getType());
}
case Instruction::SDiv:
case Instruction::UDiv:
// X / undef -> undef
if (match(C1, m_Zero()))
return C2;
// undef / 0 -> undef
// undef / 1 -> undef
if (match(C2, m_Zero()) || match(C2, m_One()))
return C1;
// undef / X -> 0 otherwise
return Constant::getNullValue(C1->getType());
case Instruction::URem:
case Instruction::SRem:
// X % undef -> undef
if (match(C2, m_Undef()))
return C2;
// undef % 0 -> undef
if (match(C2, m_Zero()))
return C1;
// undef % X -> 0 otherwise
return Constant::getNullValue(C1->getType());
case Instruction::Or: // X | undef -> -1
if (isa<UndefValue>(C1) && isa<UndefValue>(C2)) // undef | undef -> undef
return C1;
return Constant::getAllOnesValue(C1->getType()); // undef | X -> ~0
case Instruction::LShr:
// X >>l undef -> undef
if (isa<UndefValue>(C2))
return C2;
// undef >>l 0 -> undef
if (match(C2, m_Zero()))
return C1;
// undef >>l X -> 0
return Constant::getNullValue(C1->getType());
case Instruction::AShr:
// X >>a undef -> undef
if (isa<UndefValue>(C2))
return C2;
// undef >>a 0 -> undef
if (match(C2, m_Zero()))
return C1;
// TODO: undef >>a X -> undef if the shift is exact
// undef >>a X -> 0
return Constant::getNullValue(C1->getType());
case Instruction::Shl:
// X << undef -> undef
if (isa<UndefValue>(C2))
return C2;
// undef << 0 -> undef
if (match(C2, m_Zero()))
return C1;
// undef << X -> 0
return Constant::getNullValue(C1->getType());
}
}
// Handle simplifications when the RHS is a constant int.
if (ConstantInt *CI2 = dyn_cast<ConstantInt>(C2)) {
switch (Opcode) {
case Instruction::Add:
if (CI2->equalsInt(0)) return C1; // X + 0 == X
break;
case Instruction::Sub:
if (CI2->equalsInt(0)) return C1; // X - 0 == X
break;
case Instruction::Mul:
if (CI2->equalsInt(0)) return C2; // X * 0 == 0
if (CI2->equalsInt(1))
return C1; // X * 1 == X
break;
case Instruction::UDiv:
case Instruction::SDiv:
if (CI2->equalsInt(1))
return C1; // X / 1 == X
if (CI2->equalsInt(0))
return UndefValue::get(CI2->getType()); // X / 0 == undef
break;
case Instruction::URem:
case Instruction::SRem:
if (CI2->equalsInt(1))
return Constant::getNullValue(CI2->getType()); // X % 1 == 0
if (CI2->equalsInt(0))
return UndefValue::get(CI2->getType()); // X % 0 == undef
break;
case Instruction::And:
if (CI2->isZero()) return C2; // X & 0 == 0
if (CI2->isAllOnesValue())
return C1; // X & -1 == X
if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(C1)) {
// (zext i32 to i64) & 4294967295 -> (zext i32 to i64)
if (CE1->getOpcode() == Instruction::ZExt) {
unsigned DstWidth = CI2->getType()->getBitWidth();
unsigned SrcWidth =
CE1->getOperand(0)->getType()->getPrimitiveSizeInBits();
APInt PossiblySetBits(APInt::getLowBitsSet(DstWidth, SrcWidth));
if ((PossiblySetBits & CI2->getValue()) == PossiblySetBits)
return C1;
}
// If and'ing the address of a global with a constant, fold it.
if (CE1->getOpcode() == Instruction::PtrToInt &&
isa<GlobalValue>(CE1->getOperand(0))) {
GlobalValue *GV = cast<GlobalValue>(CE1->getOperand(0));
// Functions are at least 4-byte aligned.
unsigned GVAlign = GV->getAlignment();
if (isa<Function>(GV))
GVAlign = std::max(GVAlign, 4U);
if (GVAlign > 1) {
unsigned DstWidth = CI2->getType()->getBitWidth();
unsigned SrcWidth = std::min(DstWidth, Log2_32(GVAlign));
APInt BitsNotSet(APInt::getLowBitsSet(DstWidth, SrcWidth));
// If checking bits we know are clear, return zero.
if ((CI2->getValue() & BitsNotSet) == CI2->getValue())
return Constant::getNullValue(CI2->getType());
}
}
}
break;
case Instruction::Or:
if (CI2->equalsInt(0)) return C1; // X | 0 == X
if (CI2->isAllOnesValue())
return C2; // X | -1 == -1
break;
case Instruction::Xor:
if (CI2->equalsInt(0)) return C1; // X ^ 0 == X
if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(C1)) {
switch (CE1->getOpcode()) {
default: break;
case Instruction::ICmp:
case Instruction::FCmp:
// cmp pred ^ true -> cmp !pred
assert(CI2->equalsInt(1));
CmpInst::Predicate pred = (CmpInst::Predicate)CE1->getPredicate();
pred = CmpInst::getInversePredicate(pred);
return ConstantExpr::getCompare(pred, CE1->getOperand(0),
CE1->getOperand(1));
}
}
break;
case Instruction::AShr:
// ashr (zext C to Ty), C2 -> lshr (zext C, CSA), C2
if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(C1))
if (CE1->getOpcode() == Instruction::ZExt) // Top bits known zero.
return ConstantExpr::getLShr(C1, C2);
break;
}
} else if (isa<ConstantInt>(C1)) {
// If C1 is a ConstantInt and C2 is not, swap the operands.
if (Instruction::isCommutative(Opcode))
return ConstantExpr::get(Opcode, C2, C1);
}
// At this point we know neither constant is an UndefValue.
if (ConstantInt *CI1 = dyn_cast<ConstantInt>(C1)) {
if (ConstantInt *CI2 = dyn_cast<ConstantInt>(C2)) {
const APInt &C1V = CI1->getValue();
const APInt &C2V = CI2->getValue();
switch (Opcode) {
default:
break;
case Instruction::Add:
return ConstantInt::get(CI1->getContext(), C1V + C2V);
case Instruction::Sub:
return ConstantInt::get(CI1->getContext(), C1V - C2V);
case Instruction::Mul:
return ConstantInt::get(CI1->getContext(), C1V * C2V);
case Instruction::UDiv:
assert(!CI2->isNullValue() && "Div by zero handled above");
return ConstantInt::get(CI1->getContext(), C1V.udiv(C2V));
case Instruction::SDiv:
assert(!CI2->isNullValue() && "Div by zero handled above");
if (C2V.isAllOnesValue() && C1V.isMinSignedValue())
return UndefValue::get(CI1->getType()); // MIN_INT / -1 -> undef
return ConstantInt::get(CI1->getContext(), C1V.sdiv(C2V));
case Instruction::URem:
assert(!CI2->isNullValue() && "Div by zero handled above");
return ConstantInt::get(CI1->getContext(), C1V.urem(C2V));
case Instruction::SRem:
assert(!CI2->isNullValue() && "Div by zero handled above");
if (C2V.isAllOnesValue() && C1V.isMinSignedValue())
return UndefValue::get(CI1->getType()); // MIN_INT % -1 -> undef
return ConstantInt::get(CI1->getContext(), C1V.srem(C2V));
case Instruction::And:
return ConstantInt::get(CI1->getContext(), C1V & C2V);
case Instruction::Or:
return ConstantInt::get(CI1->getContext(), C1V | C2V);
case Instruction::Xor:
return ConstantInt::get(CI1->getContext(), C1V ^ C2V);
case Instruction::Shl: {
uint32_t shiftAmt = C2V.getZExtValue();
if (shiftAmt < C1V.getBitWidth())
return ConstantInt::get(CI1->getContext(), C1V.shl(shiftAmt));
else
return UndefValue::get(C1->getType()); // too big shift is undef
}
case Instruction::LShr: {
uint32_t shiftAmt = C2V.getZExtValue();
if (shiftAmt < C1V.getBitWidth())
return ConstantInt::get(CI1->getContext(), C1V.lshr(shiftAmt));
else
return UndefValue::get(C1->getType()); // too big shift is undef
}
case Instruction::AShr: {
uint32_t shiftAmt = C2V.getZExtValue();
if (shiftAmt < C1V.getBitWidth())
return ConstantInt::get(CI1->getContext(), C1V.ashr(shiftAmt));
else
return UndefValue::get(C1->getType()); // too big shift is undef
}
}
}
switch (Opcode) {
case Instruction::SDiv:
case Instruction::UDiv:
case Instruction::URem:
case Instruction::SRem:
case Instruction::LShr:
case Instruction::AShr:
case Instruction::Shl:
if (CI1->equalsInt(0)) return C1;
break;
default:
break;
}
} else if (ConstantFP *CFP1 = dyn_cast<ConstantFP>(C1)) {
if (ConstantFP *CFP2 = dyn_cast<ConstantFP>(C2)) {
APFloat C1V = CFP1->getValueAPF();
APFloat C2V = CFP2->getValueAPF();
APFloat C3V = C1V; // copy for modification
switch (Opcode) {
default:
break;
case Instruction::FAdd:
(void)C3V.add(C2V, APFloat::rmNearestTiesToEven);
return ConstantFP::get(C1->getContext(), C3V);
case Instruction::FSub:
(void)C3V.subtract(C2V, APFloat::rmNearestTiesToEven);
return ConstantFP::get(C1->getContext(), C3V);
case Instruction::FMul:
(void)C3V.multiply(C2V, APFloat::rmNearestTiesToEven);
return ConstantFP::get(C1->getContext(), C3V);
case Instruction::FDiv:
(void)C3V.divide(C2V, APFloat::rmNearestTiesToEven);
return ConstantFP::get(C1->getContext(), C3V);
case Instruction::FRem:
(void)C3V.mod(C2V, APFloat::rmNearestTiesToEven);
return ConstantFP::get(C1->getContext(), C3V);
}
}
} else if (VectorType *VTy = dyn_cast<VectorType>(C1->getType())) {
// Perform elementwise folding.
SmallVector<Constant*, 16> Result;
Type *Ty = IntegerType::get(VTy->getContext(), 32);
for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i) {
Constant *LHS =
ConstantExpr::getExtractElement(C1, ConstantInt::get(Ty, i));
Constant *RHS =
ConstantExpr::getExtractElement(C2, ConstantInt::get(Ty, i));
Result.push_back(ConstantExpr::get(Opcode, LHS, RHS));
}
return ConstantVector::get(Result);
}
if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(C1)) {
// There are many possible foldings we could do here. We should probably
// at least fold add of a pointer with an integer into the appropriate
// getelementptr. This will improve alias analysis a bit.
// Given ((a + b) + c), if (b + c) folds to something interesting, return
// (a + (b + c)).
if (Instruction::isAssociative(Opcode) && CE1->getOpcode() == Opcode) {
Constant *T = ConstantExpr::get(Opcode, CE1->getOperand(1), C2);
if (!isa<ConstantExpr>(T) || cast<ConstantExpr>(T)->getOpcode() != Opcode)
return ConstantExpr::get(Opcode, CE1->getOperand(0), T);
}
} else if (isa<ConstantExpr>(C2)) {
// If C2 is a constant expr and C1 isn't, flop them around and fold the
// other way if possible.
if (Instruction::isCommutative(Opcode))
return ConstantFoldBinaryInstruction(Opcode, C2, C1);
}
// i1 can be simplified in many cases.
if (C1->getType()->isIntegerTy(1)) {
switch (Opcode) {
case Instruction::Add:
case Instruction::Sub:
return ConstantExpr::getXor(C1, C2);
case Instruction::Mul:
return ConstantExpr::getAnd(C1, C2);
case Instruction::Shl:
case Instruction::LShr:
case Instruction::AShr:
// We can assume that C2 == 0. If it were one the result would be
// undefined because the shift value is as large as the bitwidth.
return C1;
case Instruction::SDiv:
case Instruction::UDiv:
// We can assume that C2 == 1. If it were zero the result would be
// undefined through division by zero.
return C1;
case Instruction::URem:
case Instruction::SRem:
// We can assume that C2 == 1. If it were zero the result would be
// undefined through division by zero.
return ConstantInt::getFalse(C1->getContext());
default:
break;
}
}
// We don't know how to fold this.
return nullptr;
}
/// isZeroSizedType - This type is zero sized if its an array or structure of
/// zero sized types. The only leaf zero sized type is an empty structure.
static bool isMaybeZeroSizedType(Type *Ty) {
if (StructType *STy = dyn_cast<StructType>(Ty)) {
if (STy->isOpaque()) return true; // Can't say.
// If all of elements have zero size, this does too.
for (unsigned i = 0, e = STy->getNumElements(); i != e; ++i)
if (!isMaybeZeroSizedType(STy->getElementType(i))) return false;
return true;
} else if (ArrayType *ATy = dyn_cast<ArrayType>(Ty)) {
return isMaybeZeroSizedType(ATy->getElementType());
}
return false;
}
/// IdxCompare - Compare the two constants as though they were getelementptr
/// indices. This allows coersion of the types to be the same thing.
///
/// If the two constants are the "same" (after coersion), return 0. If the
/// first is less than the second, return -1, if the second is less than the
/// first, return 1. If the constants are not integral, return -2.
///
static int IdxCompare(Constant *C1, Constant *C2, Type *ElTy) {
if (C1 == C2) return 0;
// Ok, we found a different index. If they are not ConstantInt, we can't do
// anything with them.
if (!isa<ConstantInt>(C1) || !isa<ConstantInt>(C2))
return -2; // don't know!
// We cannot compare the indices if they don't fit in an int64_t.
if (cast<ConstantInt>(C1)->getValue().getActiveBits() > 64 ||
cast<ConstantInt>(C2)->getValue().getActiveBits() > 64)
return -2; // don't know!
// Ok, we have two differing integer indices. Sign extend them to be the same
// type.
int64_t C1Val = cast<ConstantInt>(C1)->getSExtValue();
int64_t C2Val = cast<ConstantInt>(C2)->getSExtValue();
if (C1Val == C2Val) return 0; // They are equal
// If the type being indexed over is really just a zero sized type, there is
// no pointer difference being made here.
if (isMaybeZeroSizedType(ElTy))
return -2; // dunno.
// If they are really different, now that they are the same type, then we
// found a difference!
if (C1Val < C2Val)
return -1;
else
return 1;
}
/// evaluateFCmpRelation - This function determines if there is anything we can
/// decide about the two constants provided. This doesn't need to handle simple
/// things like ConstantFP comparisons, but should instead handle ConstantExprs.
/// If we can determine that the two constants have a particular relation to
/// each other, we should return the corresponding FCmpInst predicate,
/// otherwise return FCmpInst::BAD_FCMP_PREDICATE. This is used below in
/// ConstantFoldCompareInstruction.
///
/// To simplify this code we canonicalize the relation so that the first
/// operand is always the most "complex" of the two. We consider ConstantFP
/// to be the simplest, and ConstantExprs to be the most complex.
static FCmpInst::Predicate evaluateFCmpRelation(Constant *V1, Constant *V2) {
assert(V1->getType() == V2->getType() &&
"Cannot compare values of different types!");
// Handle degenerate case quickly
if (V1 == V2) return FCmpInst::FCMP_OEQ;
if (!isa<ConstantExpr>(V1)) {
if (!isa<ConstantExpr>(V2)) {
// We distilled thisUse the standard constant folder for a few cases
ConstantInt *R = nullptr;
R = dyn_cast<ConstantInt>(
ConstantExpr::getFCmp(FCmpInst::FCMP_OEQ, V1, V2));
if (R && !R->isZero())
return FCmpInst::FCMP_OEQ;
R = dyn_cast<ConstantInt>(
ConstantExpr::getFCmp(FCmpInst::FCMP_OLT, V1, V2));
if (R && !R->isZero())
return FCmpInst::FCMP_OLT;
R = dyn_cast<ConstantInt>(
ConstantExpr::getFCmp(FCmpInst::FCMP_OGT, V1, V2));
if (R && !R->isZero())
return FCmpInst::FCMP_OGT;
// Nothing more we can do
return FCmpInst::BAD_FCMP_PREDICATE;
}
// If the first operand is simple and second is ConstantExpr, swap operands.
FCmpInst::Predicate SwappedRelation = evaluateFCmpRelation(V2, V1);
if (SwappedRelation != FCmpInst::BAD_FCMP_PREDICATE)
return FCmpInst::getSwappedPredicate(SwappedRelation);
} else {
// Ok, the LHS is known to be a constantexpr. The RHS can be any of a
// constantexpr or a simple constant.
ConstantExpr *CE1 = cast<ConstantExpr>(V1);
switch (CE1->getOpcode()) {
case Instruction::FPTrunc:
case Instruction::FPExt:
case Instruction::UIToFP:
case Instruction::SIToFP:
// We might be able to do something with these but we don't right now.
break;
default:
break;
}
}
// There are MANY other foldings that we could perform here. They will
// probably be added on demand, as they seem needed.
return FCmpInst::BAD_FCMP_PREDICATE;
}
static ICmpInst::Predicate areGlobalsPotentiallyEqual(const GlobalValue *GV1,
const GlobalValue *GV2) {
auto isGlobalUnsafeForEquality = [](const GlobalValue *GV) {
if (GV->hasExternalWeakLinkage() || GV->hasWeakAnyLinkage())
return true;
if (const auto *GVar = dyn_cast<GlobalVariable>(GV)) {
Type *Ty = GVar->getType()->getPointerElementType();
// A global with opaque type might end up being zero sized.
if (!Ty->isSized())
return true;
// A global with an empty type might lie at the address of any other
// global.
if (Ty->isEmptyTy())
return true;
}
return false;
};
// Don't try to decide equality of aliases.
if (!isa<GlobalAlias>(GV1) && !isa<GlobalAlias>(GV2))
if (!isGlobalUnsafeForEquality(GV1) && !isGlobalUnsafeForEquality(GV2))
return ICmpInst::ICMP_NE;
return ICmpInst::BAD_ICMP_PREDICATE;
}
/// evaluateICmpRelation - This function determines if there is anything we can
/// decide about the two constants provided. This doesn't need to handle simple
/// things like integer comparisons, but should instead handle ConstantExprs
/// and GlobalValues. If we can determine that the two constants have a
/// particular relation to each other, we should return the corresponding ICmp
/// predicate, otherwise return ICmpInst::BAD_ICMP_PREDICATE.
///
/// To simplify this code we canonicalize the relation so that the first
/// operand is always the most "complex" of the two. We consider simple
/// constants (like ConstantInt) to be the simplest, followed by
/// GlobalValues, followed by ConstantExpr's (the most complex).
///
static ICmpInst::Predicate evaluateICmpRelation(Constant *V1, Constant *V2,
bool isSigned) {
assert(V1->getType() == V2->getType() &&
"Cannot compare different types of values!");
if (V1 == V2) return ICmpInst::ICMP_EQ;
if (!isa<ConstantExpr>(V1) && !isa<GlobalValue>(V1) &&
!isa<BlockAddress>(V1)) {
if (!isa<GlobalValue>(V2) && !isa<ConstantExpr>(V2) &&
!isa<BlockAddress>(V2)) {
// We distilled this down to a simple case, use the standard constant
// folder.
ConstantInt *R = nullptr;
ICmpInst::Predicate pred = ICmpInst::ICMP_EQ;
R = dyn_cast<ConstantInt>(ConstantExpr::getICmp(pred, V1, V2));
if (R && !R->isZero())
return pred;
pred = isSigned ? ICmpInst::ICMP_SLT : ICmpInst::ICMP_ULT;
R = dyn_cast<ConstantInt>(ConstantExpr::getICmp(pred, V1, V2));
if (R && !R->isZero())
return pred;
pred = isSigned ? ICmpInst::ICMP_SGT : ICmpInst::ICMP_UGT;
R = dyn_cast<ConstantInt>(ConstantExpr::getICmp(pred, V1, V2));
if (R && !R->isZero())
return pred;
// If we couldn't figure it out, bail.
return ICmpInst::BAD_ICMP_PREDICATE;
}
// If the first operand is simple, swap operands.
ICmpInst::Predicate SwappedRelation =
evaluateICmpRelation(V2, V1, isSigned);
if (SwappedRelation != ICmpInst::BAD_ICMP_PREDICATE)
return ICmpInst::getSwappedPredicate(SwappedRelation);
} else if (const GlobalValue *GV = dyn_cast<GlobalValue>(V1)) {
if (isa<ConstantExpr>(V2)) { // Swap as necessary.
ICmpInst::Predicate SwappedRelation =
evaluateICmpRelation(V2, V1, isSigned);
if (SwappedRelation != ICmpInst::BAD_ICMP_PREDICATE)
return ICmpInst::getSwappedPredicate(SwappedRelation);
return ICmpInst::BAD_ICMP_PREDICATE;
}
// Now we know that the RHS is a GlobalValue, BlockAddress or simple
// constant (which, since the types must match, means that it's a
// ConstantPointerNull).
if (const GlobalValue *GV2 = dyn_cast<GlobalValue>(V2)) {
return areGlobalsPotentiallyEqual(GV, GV2);
} else if (isa<BlockAddress>(V2)) {
return ICmpInst::ICMP_NE; // Globals never equal labels.
} else {
assert(isa<ConstantPointerNull>(V2) && "Canonicalization guarantee!");
// GlobalVals can never be null unless they have external weak linkage.
// We don't try to evaluate aliases here.
if (!GV->hasExternalWeakLinkage() && !isa<GlobalAlias>(GV))
return ICmpInst::ICMP_NE;
}
} else if (const BlockAddress *BA = dyn_cast<BlockAddress>(V1)) {
if (isa<ConstantExpr>(V2)) { // Swap as necessary.
ICmpInst::Predicate SwappedRelation =
evaluateICmpRelation(V2, V1, isSigned);
if (SwappedRelation != ICmpInst::BAD_ICMP_PREDICATE)
return ICmpInst::getSwappedPredicate(SwappedRelation);
return ICmpInst::BAD_ICMP_PREDICATE;
}
// Now we know that the RHS is a GlobalValue, BlockAddress or simple
// constant (which, since the types must match, means that it is a
// ConstantPointerNull).
if (const BlockAddress *BA2 = dyn_cast<BlockAddress>(V2)) {
// Block address in another function can't equal this one, but block
// addresses in the current function might be the same if blocks are
// empty.
if (BA2->getFunction() != BA->getFunction())
return ICmpInst::ICMP_NE;
} else {
// Block addresses aren't null, don't equal the address of globals.
assert((isa<ConstantPointerNull>(V2) || isa<GlobalValue>(V2)) &&
"Canonicalization guarantee!");
return ICmpInst::ICMP_NE;
}
} else {
// Ok, the LHS is known to be a constantexpr. The RHS can be any of a
// constantexpr, a global, block address, or a simple constant.
ConstantExpr *CE1 = cast<ConstantExpr>(V1);
Constant *CE1Op0 = CE1->getOperand(0);
switch (CE1->getOpcode()) {
case Instruction::Trunc:
case Instruction::FPTrunc:
case Instruction::FPExt:
case Instruction::FPToUI:
case Instruction::FPToSI:
break; // We can't evaluate floating point casts or truncations.
case Instruction::UIToFP:
case Instruction::SIToFP:
case Instruction::BitCast:
case Instruction::ZExt:
case Instruction::SExt:
// If the cast is not actually changing bits, and the second operand is a
// null pointer, do the comparison with the pre-casted value.
if (V2->isNullValue() &&
(CE1->getType()->isPointerTy() || CE1->getType()->isIntegerTy())) {
if (CE1->getOpcode() == Instruction::ZExt) isSigned = false;
if (CE1->getOpcode() == Instruction::SExt) isSigned = true;
return evaluateICmpRelation(CE1Op0,
Constant::getNullValue(CE1Op0->getType()),
isSigned);
}
break;
case Instruction::GetElementPtr: {
GEPOperator *CE1GEP = cast<GEPOperator>(CE1);
// Ok, since this is a getelementptr, we know that the constant has a
// pointer type. Check the various cases.
if (isa<ConstantPointerNull>(V2)) {
// If we are comparing a GEP to a null pointer, check to see if the base
// of the GEP equals the null pointer.
if (const GlobalValue *GV = dyn_cast<GlobalValue>(CE1Op0)) {
if (GV->hasExternalWeakLinkage())
// Weak linkage GVals could be zero or not. We're comparing that
// to null pointer so its greater-or-equal
return isSigned ? ICmpInst::ICMP_SGE : ICmpInst::ICMP_UGE;
else
// If its not weak linkage, the GVal must have a non-zero address
// so the result is greater-than
return isSigned ? ICmpInst::ICMP_SGT : ICmpInst::ICMP_UGT;
} else if (isa<ConstantPointerNull>(CE1Op0)) {
// If we are indexing from a null pointer, check to see if we have any
// non-zero indices.
for (unsigned i = 1, e = CE1->getNumOperands(); i != e; ++i)
if (!CE1->getOperand(i)->isNullValue())
// Offsetting from null, must not be equal.
return isSigned ? ICmpInst::ICMP_SGT : ICmpInst::ICMP_UGT;
// Only zero indexes from null, must still be zero.
return ICmpInst::ICMP_EQ;
}
// Otherwise, we can't really say if the first operand is null or not.
} else if (const GlobalValue *GV2 = dyn_cast<GlobalValue>(V2)) {
if (isa<ConstantPointerNull>(CE1Op0)) {
if (GV2->hasExternalWeakLinkage())
// Weak linkage GVals could be zero or not. We're comparing it to
// a null pointer, so its less-or-equal
return isSigned ? ICmpInst::ICMP_SLE : ICmpInst::ICMP_ULE;
else
// If its not weak linkage, the GVal must have a non-zero address
// so the result is less-than
return isSigned ? ICmpInst::ICMP_SLT : ICmpInst::ICMP_ULT;
} else if (const GlobalValue *GV = dyn_cast<GlobalValue>(CE1Op0)) {
if (GV == GV2) {
// If this is a getelementptr of the same global, then it must be
// different. Because the types must match, the getelementptr could
// only have at most one index, and because we fold getelementptr's
// with a single zero index, it must be nonzero.
assert(CE1->getNumOperands() == 2 &&
!CE1->getOperand(1)->isNullValue() &&
"Surprising getelementptr!");
return isSigned ? ICmpInst::ICMP_SGT : ICmpInst::ICMP_UGT;
} else {
if (CE1GEP->hasAllZeroIndices())
return areGlobalsPotentiallyEqual(GV, GV2);
return ICmpInst::BAD_ICMP_PREDICATE;
}
}
} else {
ConstantExpr *CE2 = cast<ConstantExpr>(V2);
Constant *CE2Op0 = CE2->getOperand(0);
// There are MANY other foldings that we could perform here. They will
// probably be added on demand, as they seem needed.
switch (CE2->getOpcode()) {
default: break;
case Instruction::GetElementPtr:
// By far the most common case to handle is when the base pointers are
// obviously to the same global.
if (isa<GlobalValue>(CE1Op0) && isa<GlobalValue>(CE2Op0)) {
// Don't know relative ordering, but check for inequality.
if (CE1Op0 != CE2Op0) {
GEPOperator *CE2GEP = cast<GEPOperator>(CE2);
if (CE1GEP->hasAllZeroIndices() && CE2GEP->hasAllZeroIndices())
return areGlobalsPotentiallyEqual(cast<GlobalValue>(CE1Op0),
cast<GlobalValue>(CE2Op0));
return ICmpInst::BAD_ICMP_PREDICATE;
}
// Ok, we know that both getelementptr instructions are based on the
// same global. From this, we can precisely determine the relative
// ordering of the resultant pointers.
unsigned i = 1;
// The logic below assumes that the result of the comparison
// can be determined by finding the first index that differs.
// This doesn't work if there is over-indexing in any
// subsequent indices, so check for that case first.
if (!CE1->isGEPWithNoNotionalOverIndexing() ||
!CE2->isGEPWithNoNotionalOverIndexing())
return ICmpInst::BAD_ICMP_PREDICATE; // Might be equal.
// Compare all of the operands the GEP's have in common.
gep_type_iterator GTI = gep_type_begin(CE1);
for (;i != CE1->getNumOperands() && i != CE2->getNumOperands();
++i, ++GTI)
switch (IdxCompare(CE1->getOperand(i),
CE2->getOperand(i), GTI.getIndexedType())) {
case -1: return isSigned ? ICmpInst::ICMP_SLT:ICmpInst::ICMP_ULT;
case 1: return isSigned ? ICmpInst::ICMP_SGT:ICmpInst::ICMP_UGT;
case -2: return ICmpInst::BAD_ICMP_PREDICATE;
}
// Ok, we ran out of things they have in common. If any leftovers
// are non-zero then we have a difference, otherwise we are equal.
for (; i < CE1->getNumOperands(); ++i)
if (!CE1->getOperand(i)->isNullValue()) {
if (isa<ConstantInt>(CE1->getOperand(i)))
return isSigned ? ICmpInst::ICMP_SGT : ICmpInst::ICMP_UGT;
else
return ICmpInst::BAD_ICMP_PREDICATE; // Might be equal.
}
for (; i < CE2->getNumOperands(); ++i)
if (!CE2->getOperand(i)->isNullValue()) {
if (isa<ConstantInt>(CE2->getOperand(i)))
return isSigned ? ICmpInst::ICMP_SLT : ICmpInst::ICMP_ULT;
else
return ICmpInst::BAD_ICMP_PREDICATE; // Might be equal.
}
return ICmpInst::ICMP_EQ;
}
}
}
}
default:
break;
}
}
return ICmpInst::BAD_ICMP_PREDICATE;
}
Constant *llvm::ConstantFoldCompareInstruction(unsigned short pred,
Constant *C1, Constant *C2) {
Type *ResultTy;
if (VectorType *VT = dyn_cast<VectorType>(C1->getType()))
ResultTy = VectorType::get(Type::getInt1Ty(C1->getContext()),
VT->getNumElements());
else
ResultTy = Type::getInt1Ty(C1->getContext());
// Fold FCMP_FALSE/FCMP_TRUE unconditionally.
if (pred == FCmpInst::FCMP_FALSE)
return Constant::getNullValue(ResultTy);
if (pred == FCmpInst::FCMP_TRUE)
return Constant::getAllOnesValue(ResultTy);
// Handle some degenerate cases first
if (isa<UndefValue>(C1) || isa<UndefValue>(C2)) {
// For EQ and NE, we can always pick a value for the undef to make the
// predicate pass or fail, so we can return undef.
// Also, if both operands are undef, we can return undef.
if (ICmpInst::isEquality(ICmpInst::Predicate(pred)) ||
(isa<UndefValue>(C1) && isa<UndefValue>(C2)))
return UndefValue::get(ResultTy);
// Otherwise, pick the same value as the non-undef operand, and fold
// it to true or false.
return ConstantInt::get(ResultTy, CmpInst::isTrueWhenEqual(pred));
}
// icmp eq/ne(null,GV) -> false/true
if (C1->isNullValue()) {
if (const GlobalValue *GV = dyn_cast<GlobalValue>(C2))
// Don't try to evaluate aliases. External weak GV can be null.
if (!isa<GlobalAlias>(GV) && !GV->hasExternalWeakLinkage()) {
if (pred == ICmpInst::ICMP_EQ)
return ConstantInt::getFalse(C1->getContext());
else if (pred == ICmpInst::ICMP_NE)
return ConstantInt::getTrue(C1->getContext());
}
// icmp eq/ne(GV,null) -> false/true
} else if (C2->isNullValue()) {
if (const GlobalValue *GV = dyn_cast<GlobalValue>(C1))
// Don't try to evaluate aliases. External weak GV can be null.
if (!isa<GlobalAlias>(GV) && !GV->hasExternalWeakLinkage()) {
if (pred == ICmpInst::ICMP_EQ)
return ConstantInt::getFalse(C1->getContext());
else if (pred == ICmpInst::ICMP_NE)
return ConstantInt::getTrue(C1->getContext());
}
}
// If the comparison is a comparison between two i1's, simplify it.
if (C1->getType()->isIntegerTy(1)) {
switch(pred) {
case ICmpInst::ICMP_EQ:
if (isa<ConstantInt>(C2))
return ConstantExpr::getXor(C1, ConstantExpr::getNot(C2));
return ConstantExpr::getXor(ConstantExpr::getNot(C1), C2);
case ICmpInst::ICMP_NE:
return ConstantExpr::getXor(C1, C2);
default:
break;
}
}
if (isa<ConstantInt>(C1) && isa<ConstantInt>(C2)) {
APInt V1 = cast<ConstantInt>(C1)->getValue();
APInt V2 = cast<ConstantInt>(C2)->getValue();
switch (pred) {
default: llvm_unreachable("Invalid ICmp Predicate");
case ICmpInst::ICMP_EQ: return ConstantInt::get(ResultTy, V1 == V2);
case ICmpInst::ICMP_NE: return ConstantInt::get(ResultTy, V1 != V2);
case ICmpInst::ICMP_SLT: return ConstantInt::get(ResultTy, V1.slt(V2));
case ICmpInst::ICMP_SGT: return ConstantInt::get(ResultTy, V1.sgt(V2));
case ICmpInst::ICMP_SLE: return ConstantInt::get(ResultTy, V1.sle(V2));
case ICmpInst::ICMP_SGE: return ConstantInt::get(ResultTy, V1.sge(V2));
case ICmpInst::ICMP_ULT: return ConstantInt::get(ResultTy, V1.ult(V2));
case ICmpInst::ICMP_UGT: return ConstantInt::get(ResultTy, V1.ugt(V2));
case ICmpInst::ICMP_ULE: return ConstantInt::get(ResultTy, V1.ule(V2));
case ICmpInst::ICMP_UGE: return ConstantInt::get(ResultTy, V1.uge(V2));
}
} else if (isa<ConstantFP>(C1) && isa<ConstantFP>(C2)) {
APFloat C1V = cast<ConstantFP>(C1)->getValueAPF();
APFloat C2V = cast<ConstantFP>(C2)->getValueAPF();
APFloat::cmpResult R = C1V.compare(C2V);
switch (pred) {
default: llvm_unreachable("Invalid FCmp Predicate");
case FCmpInst::FCMP_FALSE: return Constant::getNullValue(ResultTy);
case FCmpInst::FCMP_TRUE: return Constant::getAllOnesValue(ResultTy);
case FCmpInst::FCMP_UNO:
return ConstantInt::get(ResultTy, R==APFloat::cmpUnordered);
case FCmpInst::FCMP_ORD:
return ConstantInt::get(ResultTy, R!=APFloat::cmpUnordered);
case FCmpInst::FCMP_UEQ:
return ConstantInt::get(ResultTy, R==APFloat::cmpUnordered ||
R==APFloat::cmpEqual);
case FCmpInst::FCMP_OEQ:
return ConstantInt::get(ResultTy, R==APFloat::cmpEqual);
case FCmpInst::FCMP_UNE:
return ConstantInt::get(ResultTy, R!=APFloat::cmpEqual);
case FCmpInst::FCMP_ONE:
return ConstantInt::get(ResultTy, R==APFloat::cmpLessThan ||
R==APFloat::cmpGreaterThan);
case FCmpInst::FCMP_ULT:
return ConstantInt::get(ResultTy, R==APFloat::cmpUnordered ||
R==APFloat::cmpLessThan);
case FCmpInst::FCMP_OLT:
return ConstantInt::get(ResultTy, R==APFloat::cmpLessThan);
case FCmpInst::FCMP_UGT:
return ConstantInt::get(ResultTy, R==APFloat::cmpUnordered ||
R==APFloat::cmpGreaterThan);
case FCmpInst::FCMP_OGT:
return ConstantInt::get(ResultTy, R==APFloat::cmpGreaterThan);
case FCmpInst::FCMP_ULE:
return ConstantInt::get(ResultTy, R!=APFloat::cmpGreaterThan);
case FCmpInst::FCMP_OLE:
return ConstantInt::get(ResultTy, R==APFloat::cmpLessThan ||
R==APFloat::cmpEqual);
case FCmpInst::FCMP_UGE:
return ConstantInt::get(ResultTy, R!=APFloat::cmpLessThan);
case FCmpInst::FCMP_OGE:
return ConstantInt::get(ResultTy, R==APFloat::cmpGreaterThan ||
R==APFloat::cmpEqual);
}
} else if (C1->getType()->isVectorTy()) {
// If we can constant fold the comparison of each element, constant fold
// the whole vector comparison.
SmallVector<Constant*, 4> ResElts;
Type *Ty = IntegerType::get(C1->getContext(), 32);
// Compare the elements, producing an i1 result or constant expr.
for (unsigned i = 0, e = C1->getType()->getVectorNumElements(); i != e;++i){
Constant *C1E =
ConstantExpr::getExtractElement(C1, ConstantInt::get(Ty, i));
Constant *C2E =
ConstantExpr::getExtractElement(C2, ConstantInt::get(Ty, i));
ResElts.push_back(ConstantExpr::getCompare(pred, C1E, C2E));
}
return ConstantVector::get(ResElts);
}
if (C1->getType()->isFloatingPointTy()) {
int Result = -1; // -1 = unknown, 0 = known false, 1 = known true.
switch (evaluateFCmpRelation(C1, C2)) {
default: llvm_unreachable("Unknown relation!");
case FCmpInst::FCMP_UNO:
case FCmpInst::FCMP_ORD:
case FCmpInst::FCMP_UEQ:
case FCmpInst::FCMP_UNE:
case FCmpInst::FCMP_ULT:
case FCmpInst::FCMP_UGT:
case FCmpInst::FCMP_ULE:
case FCmpInst::FCMP_UGE:
case FCmpInst::FCMP_TRUE:
case FCmpInst::FCMP_FALSE:
case FCmpInst::BAD_FCMP_PREDICATE:
break; // Couldn't determine anything about these constants.
case FCmpInst::FCMP_OEQ: // We know that C1 == C2
Result = (pred == FCmpInst::FCMP_UEQ || pred == FCmpInst::FCMP_OEQ ||
pred == FCmpInst::FCMP_ULE || pred == FCmpInst::FCMP_OLE ||
pred == FCmpInst::FCMP_UGE || pred == FCmpInst::FCMP_OGE);
break;
case FCmpInst::FCMP_OLT: // We know that C1 < C2
Result = (pred == FCmpInst::FCMP_UNE || pred == FCmpInst::FCMP_ONE ||
pred == FCmpInst::FCMP_ULT || pred == FCmpInst::FCMP_OLT ||
pred == FCmpInst::FCMP_ULE || pred == FCmpInst::FCMP_OLE);
break;
case FCmpInst::FCMP_OGT: // We know that C1 > C2
Result = (pred == FCmpInst::FCMP_UNE || pred == FCmpInst::FCMP_ONE ||
pred == FCmpInst::FCMP_UGT || pred == FCmpInst::FCMP_OGT ||
pred == FCmpInst::FCMP_UGE || pred == FCmpInst::FCMP_OGE);
break;
case FCmpInst::FCMP_OLE: // We know that C1 <= C2
// We can only partially decide this relation.
if (pred == FCmpInst::FCMP_UGT || pred == FCmpInst::FCMP_OGT)
Result = 0;
else if (pred == FCmpInst::FCMP_ULT || pred == FCmpInst::FCMP_OLT)
Result = 1;
break;
case FCmpInst::FCMP_OGE: // We known that C1 >= C2
// We can only partially decide this relation.
if (pred == FCmpInst::FCMP_ULT || pred == FCmpInst::FCMP_OLT)
Result = 0;
else if (pred == FCmpInst::FCMP_UGT || pred == FCmpInst::FCMP_OGT)
Result = 1;
break;
case FCmpInst::FCMP_ONE: // We know that C1 != C2
// We can only partially decide this relation.
if (pred == FCmpInst::FCMP_OEQ || pred == FCmpInst::FCMP_UEQ)
Result = 0;
else if (pred == FCmpInst::FCMP_ONE || pred == FCmpInst::FCMP_UNE)
Result = 1;
break;
}
// If we evaluated the result, return it now.
if (Result != -1)
return ConstantInt::get(ResultTy, Result);
} else {
// Evaluate the relation between the two constants, per the predicate.
int Result = -1; // -1 = unknown, 0 = known false, 1 = known true.
switch (evaluateICmpRelation(C1, C2, CmpInst::isSigned(pred))) {
default: llvm_unreachable("Unknown relational!");
case ICmpInst::BAD_ICMP_PREDICATE:
break; // Couldn't determine anything about these constants.
case ICmpInst::ICMP_EQ: // We know the constants are equal!
// If we know the constants are equal, we can decide the result of this
// computation precisely.
Result = ICmpInst::isTrueWhenEqual((ICmpInst::Predicate)pred);
break;
case ICmpInst::ICMP_ULT:
switch (pred) {
case ICmpInst::ICMP_ULT: case ICmpInst::ICMP_NE: case ICmpInst::ICMP_ULE:
Result = 1; break;
case ICmpInst::ICMP_UGT: case ICmpInst::ICMP_EQ: case ICmpInst::ICMP_UGE:
Result = 0; break;
}
break;
case ICmpInst::ICMP_SLT:
switch (pred) {
case ICmpInst::ICMP_SLT: case ICmpInst::ICMP_NE: case ICmpInst::ICMP_SLE:
Result = 1; break;
case ICmpInst::ICMP_SGT: case ICmpInst::ICMP_EQ: case ICmpInst::ICMP_SGE:
Result = 0; break;
}
break;
case ICmpInst::ICMP_UGT:
switch (pred) {
case ICmpInst::ICMP_UGT: case ICmpInst::ICMP_NE: case ICmpInst::ICMP_UGE:
Result = 1; break;
case ICmpInst::ICMP_ULT: case ICmpInst::ICMP_EQ: case ICmpInst::ICMP_ULE:
Result = 0; break;
}
break;
case ICmpInst::ICMP_SGT:
switch (pred) {
case ICmpInst::ICMP_SGT: case ICmpInst::ICMP_NE: case ICmpInst::ICMP_SGE:
Result = 1; break;
case ICmpInst::ICMP_SLT: case ICmpInst::ICMP_EQ: case ICmpInst::ICMP_SLE:
Result = 0; break;
}
break;
case ICmpInst::ICMP_ULE:
if (pred == ICmpInst::ICMP_UGT) Result = 0;
if (pred == ICmpInst::ICMP_ULT || pred == ICmpInst::ICMP_ULE) Result = 1;
break;
case ICmpInst::ICMP_SLE:
if (pred == ICmpInst::ICMP_SGT) Result = 0;
if (pred == ICmpInst::ICMP_SLT || pred == ICmpInst::ICMP_SLE) Result = 1;
break;
case ICmpInst::ICMP_UGE:
if (pred == ICmpInst::ICMP_ULT) Result = 0;
if (pred == ICmpInst::ICMP_UGT || pred == ICmpInst::ICMP_UGE) Result = 1;
break;
case ICmpInst::ICMP_SGE:
if (pred == ICmpInst::ICMP_SLT) Result = 0;
if (pred == ICmpInst::ICMP_SGT || pred == ICmpInst::ICMP_SGE) Result = 1;
break;
case ICmpInst::ICMP_NE:
if (pred == ICmpInst::ICMP_EQ) Result = 0;
if (pred == ICmpInst::ICMP_NE) Result = 1;
break;
}
// If we evaluated the result, return it now.
if (Result != -1)
return ConstantInt::get(ResultTy, Result);
// If the right hand side is a bitcast, try using its inverse to simplify
// it by moving it to the left hand side. We can't do this if it would turn
// a vector compare into a scalar compare or visa versa.
if (ConstantExpr *CE2 = dyn_cast<ConstantExpr>(C2)) {
Constant *CE2Op0 = CE2->getOperand(0);
if (CE2->getOpcode() == Instruction::BitCast &&
CE2->getType()->isVectorTy() == CE2Op0->getType()->isVectorTy()) {
Constant *Inverse = ConstantExpr::getBitCast(C1, CE2Op0->getType());
return ConstantExpr::getICmp(pred, Inverse, CE2Op0);
}
}
// If the left hand side is an extension, try eliminating it.
if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(C1)) {
if ((CE1->getOpcode() == Instruction::SExt && ICmpInst::isSigned(pred)) ||
(CE1->getOpcode() == Instruction::ZExt && !ICmpInst::isSigned(pred))){
Constant *CE1Op0 = CE1->getOperand(0);
Constant *CE1Inverse = ConstantExpr::getTrunc(CE1, CE1Op0->getType());
if (CE1Inverse == CE1Op0) {
// Check whether we can safely truncate the right hand side.
Constant *C2Inverse = ConstantExpr::getTrunc(C2, CE1Op0->getType());
if (ConstantExpr::getCast(CE1->getOpcode(), C2Inverse,
C2->getType()) == C2)
return ConstantExpr::getICmp(pred, CE1Inverse, C2Inverse);
}
}
}
if ((!isa<ConstantExpr>(C1) && isa<ConstantExpr>(C2)) ||
(C1->isNullValue() && !C2->isNullValue())) {
// If C2 is a constant expr and C1 isn't, flip them around and fold the
// other way if possible.
// Also, if C1 is null and C2 isn't, flip them around.
pred = ICmpInst::getSwappedPredicate((ICmpInst::Predicate)pred);
return ConstantExpr::getICmp(pred, C2, C1);
}
}
return nullptr;
}
/// isInBoundsIndices - Test whether the given sequence of *normalized* indices
/// is "inbounds".
template<typename IndexTy>
static bool isInBoundsIndices(ArrayRef<IndexTy> Idxs) {
// No indices means nothing that could be out of bounds.
if (Idxs.empty()) return true;
// If the first index is zero, it's in bounds.
if (cast<Constant>(Idxs[0])->isNullValue()) return true;
// If the first index is one and all the rest are zero, it's in bounds,
// by the one-past-the-end rule.
if (!cast<ConstantInt>(Idxs[0])->isOne())
return false;
for (unsigned i = 1, e = Idxs.size(); i != e; ++i)
if (!cast<Constant>(Idxs[i])->isNullValue())
return false;
return true;
}
/// \brief Test whether a given ConstantInt is in-range for a SequentialType.
static bool isIndexInRangeOfSequentialType(const SequentialType *STy,
const ConstantInt *CI) {
if (const PointerType *PTy = dyn_cast<PointerType>(STy))
// Only handle pointers to sized types, not pointers to functions.
return PTy->getElementType()->isSized();
uint64_t NumElements = 0;
// Determine the number of elements in our sequential type.
if (const ArrayType *ATy = dyn_cast<ArrayType>(STy))
NumElements = ATy->getNumElements();
else if (const VectorType *VTy = dyn_cast<VectorType>(STy))
NumElements = VTy->getNumElements();
assert((isa<ArrayType>(STy) || NumElements > 0) &&
"didn't expect non-array type to have zero elements!");
// We cannot bounds check the index if it doesn't fit in an int64_t.
if (CI->getValue().getActiveBits() > 64)
return false;
// A negative index or an index past the end of our sequential type is
// considered out-of-range.
int64_t IndexVal = CI->getSExtValue();
if (IndexVal < 0 || (NumElements > 0 && (uint64_t)IndexVal >= NumElements))
return false;
// Otherwise, it is in-range.
return true;
}
template<typename IndexTy>
static Constant *ConstantFoldGetElementPtrImpl(Constant *C,
bool inBounds,
ArrayRef<IndexTy> Idxs) {
if (Idxs.empty()) return C;
Constant *Idx0 = cast<Constant>(Idxs[0]);
if ((Idxs.size() == 1 && Idx0->isNullValue()))
return C;
if (isa<UndefValue>(C)) {
PointerType *Ptr = cast<PointerType>(C->getType());
Type *Ty = GetElementPtrInst::getIndexedType(Ptr, Idxs);
assert(Ty && "Invalid indices for GEP!");
return UndefValue::get(PointerType::get(Ty, Ptr->getAddressSpace()));
}
if (C->isNullValue()) {
bool isNull = true;
for (unsigned i = 0, e = Idxs.size(); i != e; ++i)
if (!cast<Constant>(Idxs[i])->isNullValue()) {
isNull = false;
break;
}
if (isNull) {
PointerType *Ptr = cast<PointerType>(C->getType());
Type *Ty = GetElementPtrInst::getIndexedType(Ptr, Idxs);
assert(Ty && "Invalid indices for GEP!");
return ConstantPointerNull::get(PointerType::get(Ty,
Ptr->getAddressSpace()));
}
}
if (ConstantExpr *CE = dyn_cast<ConstantExpr>(C)) {
// Combine Indices - If the source pointer to this getelementptr instruction
// is a getelementptr instruction, combine the indices of the two
// getelementptr instructions into a single instruction.
//
if (CE->getOpcode() == Instruction::GetElementPtr) {
Type *LastTy = nullptr;
for (gep_type_iterator I = gep_type_begin(CE), E = gep_type_end(CE);
I != E; ++I)
LastTy = *I;
// We cannot combine indices if doing so would take us outside of an
// array or vector. Doing otherwise could trick us if we evaluated such a
// GEP as part of a load.
//
// e.g. Consider if the original GEP was:
// i8* getelementptr ({ [2 x i8], i32, i8, [3 x i8] }* @main.c,
// i32 0, i32 0, i64 0)
//
// If we then tried to offset it by '8' to get to the third element,
// an i8, we should *not* get:
// i8* getelementptr ({ [2 x i8], i32, i8, [3 x i8] }* @main.c,
// i32 0, i32 0, i64 8)
//
// This GEP tries to index array element '8 which runs out-of-bounds.
// Subsequent evaluation would get confused and produce erroneous results.
//
// The following prohibits such a GEP from being formed by checking to see
// if the index is in-range with respect to an array or vector.
bool PerformFold = false;
if (Idx0->isNullValue())
PerformFold = true;
else if (SequentialType *STy = dyn_cast_or_null<SequentialType>(LastTy))
if (ConstantInt *CI = dyn_cast<ConstantInt>(Idx0))
PerformFold = isIndexInRangeOfSequentialType(STy, CI);
if (PerformFold) {
SmallVector<Value*, 16> NewIndices;
NewIndices.reserve(Idxs.size() + CE->getNumOperands());
NewIndices.append(CE->op_begin() + 1, CE->op_end() - 1);
// Add the last index of the source with the first index of the new GEP.
// Make sure to handle the case when they are actually different types.
Constant *Combined = CE->getOperand(CE->getNumOperands()-1);
// Otherwise it must be an array.
if (!Idx0->isNullValue()) {
Type *IdxTy = Combined->getType();
if (IdxTy != Idx0->getType()) {
unsigned CommonExtendedWidth =
std::max(IdxTy->getIntegerBitWidth(),
Idx0->getType()->getIntegerBitWidth());
CommonExtendedWidth = std::max(CommonExtendedWidth, 64U);
Type *CommonTy =
Type::getIntNTy(IdxTy->getContext(), CommonExtendedWidth);
Constant *C1 = ConstantExpr::getSExtOrBitCast(Idx0, CommonTy);
Constant *C2 = ConstantExpr::getSExtOrBitCast(Combined, CommonTy);
Combined = ConstantExpr::get(Instruction::Add, C1, C2);
} else {
Combined =
ConstantExpr::get(Instruction::Add, Idx0, Combined);
}
}
NewIndices.push_back(Combined);
NewIndices.append(Idxs.begin() + 1, Idxs.end());
return
ConstantExpr::getGetElementPtr(CE->getOperand(0), NewIndices,
inBounds &&
cast<GEPOperator>(CE)->isInBounds());
}
}
// Attempt to fold casts to the same type away. For example, folding:
//
// i32* getelementptr ([2 x i32]* bitcast ([3 x i32]* %X to [2 x i32]*),
// i64 0, i64 0)
// into:
//
// i32* getelementptr ([3 x i32]* %X, i64 0, i64 0)
//
// Don't fold if the cast is changing address spaces.
if (CE->isCast() && Idxs.size() > 1 && Idx0->isNullValue()) {
PointerType *SrcPtrTy =
dyn_cast<PointerType>(CE->getOperand(0)->getType());
PointerType *DstPtrTy = dyn_cast<PointerType>(CE->getType());
if (SrcPtrTy && DstPtrTy) {
ArrayType *SrcArrayTy =
dyn_cast<ArrayType>(SrcPtrTy->getElementType());
ArrayType *DstArrayTy =
dyn_cast<ArrayType>(DstPtrTy->getElementType());
if (SrcArrayTy && DstArrayTy
&& SrcArrayTy->getElementType() == DstArrayTy->getElementType()
&& SrcPtrTy->getAddressSpace() == DstPtrTy->getAddressSpace())
return ConstantExpr::getGetElementPtr((Constant*)CE->getOperand(0),
Idxs, inBounds);
}
}
}
// Check to see if any array indices are not within the corresponding
// notional array or vector bounds. If so, try to determine if they can be
// factored out into preceding dimensions.
bool Unknown = false;
SmallVector<Constant *, 8> NewIdxs;
Type *Ty = C->getType();
Type *Prev = nullptr;
for (unsigned i = 0, e = Idxs.size(); i != e;
Prev = Ty, Ty = cast<CompositeType>(Ty)->getTypeAtIndex(Idxs[i]), ++i) {
if (ConstantInt *CI = dyn_cast<ConstantInt>(Idxs[i])) {
if (isa<ArrayType>(Ty) || isa<VectorType>(Ty))
if (CI->getSExtValue() > 0 &&
!isIndexInRangeOfSequentialType(cast<SequentialType>(Ty), CI)) {
if (isa<SequentialType>(Prev)) {
// It's out of range, but we can factor it into the prior
// dimension.
NewIdxs.resize(Idxs.size());
uint64_t NumElements = 0;
if (const ArrayType *ATy = dyn_cast<ArrayType>(Ty))
NumElements = ATy->getNumElements();
else
NumElements = cast<VectorType>(Ty)->getNumElements();
ConstantInt *Factor = ConstantInt::get(CI->getType(), NumElements);
NewIdxs[i] = ConstantExpr::getSRem(CI, Factor);
Constant *PrevIdx = cast<Constant>(Idxs[i-1]);
Constant *Div = ConstantExpr::getSDiv(CI, Factor);
unsigned CommonExtendedWidth =
std::max(PrevIdx->getType()->getIntegerBitWidth(),
Div->getType()->getIntegerBitWidth());
CommonExtendedWidth = std::max(CommonExtendedWidth, 64U);
// Before adding, extend both operands to i64 to avoid
// overflow trouble.
if (!PrevIdx->getType()->isIntegerTy(CommonExtendedWidth))
PrevIdx = ConstantExpr::getSExt(
PrevIdx,
Type::getIntNTy(Div->getContext(), CommonExtendedWidth));
if (!Div->getType()->isIntegerTy(CommonExtendedWidth))
Div = ConstantExpr::getSExt(
Div, Type::getIntNTy(Div->getContext(), CommonExtendedWidth));
NewIdxs[i-1] = ConstantExpr::getAdd(PrevIdx, Div);
} else {
// It's out of range, but the prior dimension is a struct
// so we can't do anything about it.
Unknown = true;
}
}
} else {
// We don't know if it's in range or not.
Unknown = true;
}
}
// If we did any factoring, start over with the adjusted indices.
if (!NewIdxs.empty()) {
for (unsigned i = 0, e = Idxs.size(); i != e; ++i)
if (!NewIdxs[i]) NewIdxs[i] = cast<Constant>(Idxs[i]);
return ConstantExpr::getGetElementPtr(C, NewIdxs, inBounds);
}
// If all indices are known integers and normalized, we can do a simple
// check for the "inbounds" property.
if (!Unknown && !inBounds)
if (auto *GV = dyn_cast<GlobalVariable>(C))
if (!GV->hasExternalWeakLinkage() && isInBoundsIndices(Idxs))
return ConstantExpr::getInBoundsGetElementPtr(C, Idxs);
return nullptr;
}
Constant *llvm::ConstantFoldGetElementPtr(Constant *C,
bool inBounds,
ArrayRef<Constant *> Idxs) {
return ConstantFoldGetElementPtrImpl(C, inBounds, Idxs);
}
Constant *llvm::ConstantFoldGetElementPtr(Constant *C,
bool inBounds,
ArrayRef<Value *> Idxs) {
return ConstantFoldGetElementPtrImpl(C, inBounds, Idxs);
}
|