1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
|
//===- InstCombineCompares.cpp --------------------------------------------===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file implements the visitICmp and visitFCmp functions.
//
//===----------------------------------------------------------------------===//
#include "InstCombine.h"
#include "llvm/IntrinsicInst.h"
#include "llvm/Analysis/InstructionSimplify.h"
#include "llvm/Analysis/MemoryBuiltins.h"
#include "llvm/Target/TargetData.h"
#include "llvm/Support/ConstantRange.h"
#include "llvm/Support/GetElementPtrTypeIterator.h"
#include "llvm/Support/PatternMatch.h"
using namespace llvm;
using namespace PatternMatch;
/// AddOne - Add one to a ConstantInt
static Constant *AddOne(Constant *C) {
return ConstantExpr::getAdd(C, ConstantInt::get(C->getType(), 1));
}
/// SubOne - Subtract one from a ConstantInt
static Constant *SubOne(ConstantInt *C) {
return ConstantExpr::getSub(C, ConstantInt::get(C->getType(), 1));
}
static ConstantInt *ExtractElement(Constant *V, Constant *Idx) {
return cast<ConstantInt>(ConstantExpr::getExtractElement(V, Idx));
}
static bool HasAddOverflow(ConstantInt *Result,
ConstantInt *In1, ConstantInt *In2,
bool IsSigned) {
if (IsSigned)
if (In2->getValue().isNegative())
return Result->getValue().sgt(In1->getValue());
else
return Result->getValue().slt(In1->getValue());
else
return Result->getValue().ult(In1->getValue());
}
/// AddWithOverflow - Compute Result = In1+In2, returning true if the result
/// overflowed for this type.
static bool AddWithOverflow(Constant *&Result, Constant *In1,
Constant *In2, bool IsSigned = false) {
Result = ConstantExpr::getAdd(In1, In2);
if (const VectorType *VTy = dyn_cast<VectorType>(In1->getType())) {
for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i) {
Constant *Idx = ConstantInt::get(Type::getInt32Ty(In1->getContext()), i);
if (HasAddOverflow(ExtractElement(Result, Idx),
ExtractElement(In1, Idx),
ExtractElement(In2, Idx),
IsSigned))
return true;
}
return false;
}
return HasAddOverflow(cast<ConstantInt>(Result),
cast<ConstantInt>(In1), cast<ConstantInt>(In2),
IsSigned);
}
static bool HasSubOverflow(ConstantInt *Result,
ConstantInt *In1, ConstantInt *In2,
bool IsSigned) {
if (IsSigned)
if (In2->getValue().isNegative())
return Result->getValue().slt(In1->getValue());
else
return Result->getValue().sgt(In1->getValue());
else
return Result->getValue().ugt(In1->getValue());
}
/// SubWithOverflow - Compute Result = In1-In2, returning true if the result
/// overflowed for this type.
static bool SubWithOverflow(Constant *&Result, Constant *In1,
Constant *In2, bool IsSigned = false) {
Result = ConstantExpr::getSub(In1, In2);
if (const VectorType *VTy = dyn_cast<VectorType>(In1->getType())) {
for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i) {
Constant *Idx = ConstantInt::get(Type::getInt32Ty(In1->getContext()), i);
if (HasSubOverflow(ExtractElement(Result, Idx),
ExtractElement(In1, Idx),
ExtractElement(In2, Idx),
IsSigned))
return true;
}
return false;
}
return HasSubOverflow(cast<ConstantInt>(Result),
cast<ConstantInt>(In1), cast<ConstantInt>(In2),
IsSigned);
}
/// isSignBitCheck - Given an exploded icmp instruction, return true if the
/// comparison only checks the sign bit. If it only checks the sign bit, set
/// TrueIfSigned if the result of the comparison is true when the input value is
/// signed.
static bool isSignBitCheck(ICmpInst::Predicate pred, ConstantInt *RHS,
bool &TrueIfSigned) {
switch (pred) {
case ICmpInst::ICMP_SLT: // True if LHS s< 0
TrueIfSigned = true;
return RHS->isZero();
case ICmpInst::ICMP_SLE: // True if LHS s<= RHS and RHS == -1
TrueIfSigned = true;
return RHS->isAllOnesValue();
case ICmpInst::ICMP_SGT: // True if LHS s> -1
TrueIfSigned = false;
return RHS->isAllOnesValue();
case ICmpInst::ICMP_UGT:
// True if LHS u> RHS and RHS == high-bit-mask - 1
TrueIfSigned = true;
return RHS->getValue() ==
APInt::getSignedMaxValue(RHS->getType()->getPrimitiveSizeInBits());
case ICmpInst::ICMP_UGE:
// True if LHS u>= RHS and RHS == high-bit-mask (2^7, 2^15, 2^31, etc)
TrueIfSigned = true;
return RHS->getValue().isSignBit();
default:
return false;
}
}
// isHighOnes - Return true if the constant is of the form 1+0+.
// This is the same as lowones(~X).
static bool isHighOnes(const ConstantInt *CI) {
return (~CI->getValue() + 1).isPowerOf2();
}
/// ComputeSignedMinMaxValuesFromKnownBits - Given a signed integer type and a
/// set of known zero and one bits, compute the maximum and minimum values that
/// could have the specified known zero and known one bits, returning them in
/// min/max.
static void ComputeSignedMinMaxValuesFromKnownBits(const APInt& KnownZero,
const APInt& KnownOne,
APInt& Min, APInt& Max) {
assert(KnownZero.getBitWidth() == KnownOne.getBitWidth() &&
KnownZero.getBitWidth() == Min.getBitWidth() &&
KnownZero.getBitWidth() == Max.getBitWidth() &&
"KnownZero, KnownOne and Min, Max must have equal bitwidth.");
APInt UnknownBits = ~(KnownZero|KnownOne);
// The minimum value is when all unknown bits are zeros, EXCEPT for the sign
// bit if it is unknown.
Min = KnownOne;
Max = KnownOne|UnknownBits;
if (UnknownBits.isNegative()) { // Sign bit is unknown
Min.set(Min.getBitWidth()-1);
Max.clear(Max.getBitWidth()-1);
}
}
// ComputeUnsignedMinMaxValuesFromKnownBits - Given an unsigned integer type and
// a set of known zero and one bits, compute the maximum and minimum values that
// could have the specified known zero and known one bits, returning them in
// min/max.
static void ComputeUnsignedMinMaxValuesFromKnownBits(const APInt &KnownZero,
const APInt &KnownOne,
APInt &Min, APInt &Max) {
assert(KnownZero.getBitWidth() == KnownOne.getBitWidth() &&
KnownZero.getBitWidth() == Min.getBitWidth() &&
KnownZero.getBitWidth() == Max.getBitWidth() &&
"Ty, KnownZero, KnownOne and Min, Max must have equal bitwidth.");
APInt UnknownBits = ~(KnownZero|KnownOne);
// The minimum value is when the unknown bits are all zeros.
Min = KnownOne;
// The maximum value is when the unknown bits are all ones.
Max = KnownOne|UnknownBits;
}
/// FoldCmpLoadFromIndexedGlobal - Called we see this pattern:
/// cmp pred (load (gep GV, ...)), cmpcst
/// where GV is a global variable with a constant initializer. Try to simplify
/// this into some simple computation that does not need the load. For example
/// we can optimize "icmp eq (load (gep "foo", 0, i)), 0" into "icmp eq i, 3".
///
/// If AndCst is non-null, then the loaded value is masked with that constant
/// before doing the comparison. This handles cases like "A[i]&4 == 0".
Instruction *InstCombiner::
FoldCmpLoadFromIndexedGlobal(GetElementPtrInst *GEP, GlobalVariable *GV,
CmpInst &ICI, ConstantInt *AndCst) {
// We need TD information to know the pointer size unless this is inbounds.
if (!GEP->isInBounds() && TD == 0) return 0;
ConstantArray *Init = dyn_cast<ConstantArray>(GV->getInitializer());
if (Init == 0 || Init->getNumOperands() > 1024) return 0;
// There are many forms of this optimization we can handle, for now, just do
// the simple index into a single-dimensional array.
//
// Require: GEP GV, 0, i {{, constant indices}}
if (GEP->getNumOperands() < 3 ||
!isa<ConstantInt>(GEP->getOperand(1)) ||
!cast<ConstantInt>(GEP->getOperand(1))->isZero() ||
isa<Constant>(GEP->getOperand(2)))
return 0;
// Check that indices after the variable are constants and in-range for the
// type they index. Collect the indices. This is typically for arrays of
// structs.
SmallVector<unsigned, 4> LaterIndices;
const Type *EltTy = cast<ArrayType>(Init->getType())->getElementType();
for (unsigned i = 3, e = GEP->getNumOperands(); i != e; ++i) {
ConstantInt *Idx = dyn_cast<ConstantInt>(GEP->getOperand(i));
if (Idx == 0) return 0; // Variable index.
uint64_t IdxVal = Idx->getZExtValue();
if ((unsigned)IdxVal != IdxVal) return 0; // Too large array index.
if (const StructType *STy = dyn_cast<StructType>(EltTy))
EltTy = STy->getElementType(IdxVal);
else if (const ArrayType *ATy = dyn_cast<ArrayType>(EltTy)) {
if (IdxVal >= ATy->getNumElements()) return 0;
EltTy = ATy->getElementType();
} else {
return 0; // Unknown type.
}
LaterIndices.push_back(IdxVal);
}
enum { Overdefined = -3, Undefined = -2 };
// Variables for our state machines.
// FirstTrueElement/SecondTrueElement - Used to emit a comparison of the form
// "i == 47 | i == 87", where 47 is the first index the condition is true for,
// and 87 is the second (and last) index. FirstTrueElement is -2 when
// undefined, otherwise set to the first true element. SecondTrueElement is
// -2 when undefined, -3 when overdefined and >= 0 when that index is true.
int FirstTrueElement = Undefined, SecondTrueElement = Undefined;
// FirstFalseElement/SecondFalseElement - Used to emit a comparison of the
// form "i != 47 & i != 87". Same state transitions as for true elements.
int FirstFalseElement = Undefined, SecondFalseElement = Undefined;
/// TrueRangeEnd/FalseRangeEnd - In conjunction with First*Element, these
/// define a state machine that triggers for ranges of values that the index
/// is true or false for. This triggers on things like "abbbbc"[i] == 'b'.
/// This is -2 when undefined, -3 when overdefined, and otherwise the last
/// index in the range (inclusive). We use -2 for undefined here because we
/// use relative comparisons and don't want 0-1 to match -1.
int TrueRangeEnd = Undefined, FalseRangeEnd = Undefined;
// MagicBitvector - This is a magic bitvector where we set a bit if the
// comparison is true for element 'i'. If there are 64 elements or less in
// the array, this will fully represent all the comparison results.
uint64_t MagicBitvector = 0;
// Scan the array and see if one of our patterns matches.
Constant *CompareRHS = cast<Constant>(ICI.getOperand(1));
for (unsigned i = 0, e = Init->getNumOperands(); i != e; ++i) {
Constant *Elt = Init->getOperand(i);
// If this is indexing an array of structures, get the structure element.
if (!LaterIndices.empty())
Elt = ConstantExpr::getExtractValue(Elt, LaterIndices.data(),
LaterIndices.size());
// If the element is masked, handle it.
if (AndCst) Elt = ConstantExpr::getAnd(Elt, AndCst);
// Find out if the comparison would be true or false for the i'th element.
Constant *C = ConstantFoldCompareInstOperands(ICI.getPredicate(), Elt,
CompareRHS, TD);
// If the result is undef for this element, ignore it.
if (isa<UndefValue>(C)) {
// Extend range state machines to cover this element in case there is an
// undef in the middle of the range.
if (TrueRangeEnd == (int)i-1)
TrueRangeEnd = i;
if (FalseRangeEnd == (int)i-1)
FalseRangeEnd = i;
continue;
}
// If we can't compute the result for any of the elements, we have to give
// up evaluating the entire conditional.
if (!isa<ConstantInt>(C)) return 0;
// Otherwise, we know if the comparison is true or false for this element,
// update our state machines.
bool IsTrueForElt = !cast<ConstantInt>(C)->isZero();
// State machine for single/double/range index comparison.
if (IsTrueForElt) {
// Update the TrueElement state machine.
if (FirstTrueElement == Undefined)
FirstTrueElement = TrueRangeEnd = i; // First true element.
else {
// Update double-compare state machine.
if (SecondTrueElement == Undefined)
SecondTrueElement = i;
else
SecondTrueElement = Overdefined;
// Update range state machine.
if (TrueRangeEnd == (int)i-1)
TrueRangeEnd = i;
else
TrueRangeEnd = Overdefined;
}
} else {
// Update the FalseElement state machine.
if (FirstFalseElement == Undefined)
FirstFalseElement = FalseRangeEnd = i; // First false element.
else {
// Update double-compare state machine.
if (SecondFalseElement == Undefined)
SecondFalseElement = i;
else
SecondFalseElement = Overdefined;
// Update range state machine.
if (FalseRangeEnd == (int)i-1)
FalseRangeEnd = i;
else
FalseRangeEnd = Overdefined;
}
}
// If this element is in range, update our magic bitvector.
if (i < 64 && IsTrueForElt)
MagicBitvector |= 1ULL << i;
// If all of our states become overdefined, bail out early. Since the
// predicate is expensive, only check it every 8 elements. This is only
// really useful for really huge arrays.
if ((i & 8) == 0 && i >= 64 && SecondTrueElement == Overdefined &&
SecondFalseElement == Overdefined && TrueRangeEnd == Overdefined &&
FalseRangeEnd == Overdefined)
return 0;
}
// Now that we've scanned the entire array, emit our new comparison(s). We
// order the state machines in complexity of the generated code.
Value *Idx = GEP->getOperand(2);
// If the index is larger than the pointer size of the target, truncate the
// index down like the GEP would do implicitly. We don't have to do this for
// an inbounds GEP because the index can't be out of range.
if (!GEP->isInBounds() &&
Idx->getType()->getPrimitiveSizeInBits() > TD->getPointerSizeInBits())
Idx = Builder->CreateTrunc(Idx, TD->getIntPtrType(Idx->getContext()));
// If the comparison is only true for one or two elements, emit direct
// comparisons.
if (SecondTrueElement != Overdefined) {
// None true -> false.
if (FirstTrueElement == Undefined)
return ReplaceInstUsesWith(ICI, ConstantInt::getFalse(GEP->getContext()));
Value *FirstTrueIdx = ConstantInt::get(Idx->getType(), FirstTrueElement);
// True for one element -> 'i == 47'.
if (SecondTrueElement == Undefined)
return new ICmpInst(ICmpInst::ICMP_EQ, Idx, FirstTrueIdx);
// True for two elements -> 'i == 47 | i == 72'.
Value *C1 = Builder->CreateICmpEQ(Idx, FirstTrueIdx);
Value *SecondTrueIdx = ConstantInt::get(Idx->getType(), SecondTrueElement);
Value *C2 = Builder->CreateICmpEQ(Idx, SecondTrueIdx);
return BinaryOperator::CreateOr(C1, C2);
}
// If the comparison is only false for one or two elements, emit direct
// comparisons.
if (SecondFalseElement != Overdefined) {
// None false -> true.
if (FirstFalseElement == Undefined)
return ReplaceInstUsesWith(ICI, ConstantInt::getTrue(GEP->getContext()));
Value *FirstFalseIdx = ConstantInt::get(Idx->getType(), FirstFalseElement);
// False for one element -> 'i != 47'.
if (SecondFalseElement == Undefined)
return new ICmpInst(ICmpInst::ICMP_NE, Idx, FirstFalseIdx);
// False for two elements -> 'i != 47 & i != 72'.
Value *C1 = Builder->CreateICmpNE(Idx, FirstFalseIdx);
Value *SecondFalseIdx = ConstantInt::get(Idx->getType(),SecondFalseElement);
Value *C2 = Builder->CreateICmpNE(Idx, SecondFalseIdx);
return BinaryOperator::CreateAnd(C1, C2);
}
// If the comparison can be replaced with a range comparison for the elements
// where it is true, emit the range check.
if (TrueRangeEnd != Overdefined) {
assert(TrueRangeEnd != FirstTrueElement && "Should emit single compare");
// Generate (i-FirstTrue) <u (TrueRangeEnd-FirstTrue+1).
if (FirstTrueElement) {
Value *Offs = ConstantInt::get(Idx->getType(), -FirstTrueElement);
Idx = Builder->CreateAdd(Idx, Offs);
}
Value *End = ConstantInt::get(Idx->getType(),
TrueRangeEnd-FirstTrueElement+1);
return new ICmpInst(ICmpInst::ICMP_ULT, Idx, End);
}
// False range check.
if (FalseRangeEnd != Overdefined) {
assert(FalseRangeEnd != FirstFalseElement && "Should emit single compare");
// Generate (i-FirstFalse) >u (FalseRangeEnd-FirstFalse).
if (FirstFalseElement) {
Value *Offs = ConstantInt::get(Idx->getType(), -FirstFalseElement);
Idx = Builder->CreateAdd(Idx, Offs);
}
Value *End = ConstantInt::get(Idx->getType(),
FalseRangeEnd-FirstFalseElement);
return new ICmpInst(ICmpInst::ICMP_UGT, Idx, End);
}
// If a 32-bit or 64-bit magic bitvector captures the entire comparison state
// of this load, replace it with computation that does:
// ((magic_cst >> i) & 1) != 0
if (Init->getNumOperands() <= 32 ||
(TD && Init->getNumOperands() <= 64 && TD->isLegalInteger(64))) {
const Type *Ty;
if (Init->getNumOperands() <= 32)
Ty = Type::getInt32Ty(Init->getContext());
else
Ty = Type::getInt64Ty(Init->getContext());
Value *V = Builder->CreateIntCast(Idx, Ty, false);
V = Builder->CreateLShr(ConstantInt::get(Ty, MagicBitvector), V);
V = Builder->CreateAnd(ConstantInt::get(Ty, 1), V);
return new ICmpInst(ICmpInst::ICMP_NE, V, ConstantInt::get(Ty, 0));
}
return 0;
}
/// EvaluateGEPOffsetExpression - Return a value that can be used to compare
/// the *offset* implied by a GEP to zero. For example, if we have &A[i], we
/// want to return 'i' for "icmp ne i, 0". Note that, in general, indices can
/// be complex, and scales are involved. The above expression would also be
/// legal to codegen as "icmp ne (i*4), 0" (assuming A is a pointer to i32).
/// This later form is less amenable to optimization though, and we are allowed
/// to generate the first by knowing that pointer arithmetic doesn't overflow.
///
/// If we can't emit an optimized form for this expression, this returns null.
///
static Value *EvaluateGEPOffsetExpression(User *GEP, Instruction &I,
InstCombiner &IC) {
TargetData &TD = *IC.getTargetData();
gep_type_iterator GTI = gep_type_begin(GEP);
// Check to see if this gep only has a single variable index. If so, and if
// any constant indices are a multiple of its scale, then we can compute this
// in terms of the scale of the variable index. For example, if the GEP
// implies an offset of "12 + i*4", then we can codegen this as "3 + i",
// because the expression will cross zero at the same point.
unsigned i, e = GEP->getNumOperands();
int64_t Offset = 0;
for (i = 1; i != e; ++i, ++GTI) {
if (ConstantInt *CI = dyn_cast<ConstantInt>(GEP->getOperand(i))) {
// Compute the aggregate offset of constant indices.
if (CI->isZero()) continue;
// Handle a struct index, which adds its field offset to the pointer.
if (const StructType *STy = dyn_cast<StructType>(*GTI)) {
Offset += TD.getStructLayout(STy)->getElementOffset(CI->getZExtValue());
} else {
uint64_t Size = TD.getTypeAllocSize(GTI.getIndexedType());
Offset += Size*CI->getSExtValue();
}
} else {
// Found our variable index.
break;
}
}
// If there are no variable indices, we must have a constant offset, just
// evaluate it the general way.
if (i == e) return 0;
Value *VariableIdx = GEP->getOperand(i);
// Determine the scale factor of the variable element. For example, this is
// 4 if the variable index is into an array of i32.
uint64_t VariableScale = TD.getTypeAllocSize(GTI.getIndexedType());
// Verify that there are no other variable indices. If so, emit the hard way.
for (++i, ++GTI; i != e; ++i, ++GTI) {
ConstantInt *CI = dyn_cast<ConstantInt>(GEP->getOperand(i));
if (!CI) return 0;
// Compute the aggregate offset of constant indices.
if (CI->isZero()) continue;
// Handle a struct index, which adds its field offset to the pointer.
if (const StructType *STy = dyn_cast<StructType>(*GTI)) {
Offset += TD.getStructLayout(STy)->getElementOffset(CI->getZExtValue());
} else {
uint64_t Size = TD.getTypeAllocSize(GTI.getIndexedType());
Offset += Size*CI->getSExtValue();
}
}
// Okay, we know we have a single variable index, which must be a
// pointer/array/vector index. If there is no offset, life is simple, return
// the index.
unsigned IntPtrWidth = TD.getPointerSizeInBits();
if (Offset == 0) {
// Cast to intptrty in case a truncation occurs. If an extension is needed,
// we don't need to bother extending: the extension won't affect where the
// computation crosses zero.
if (VariableIdx->getType()->getPrimitiveSizeInBits() > IntPtrWidth)
VariableIdx = new TruncInst(VariableIdx,
TD.getIntPtrType(VariableIdx->getContext()),
VariableIdx->getName(), &I);
return VariableIdx;
}
// Otherwise, there is an index. The computation we will do will be modulo
// the pointer size, so get it.
uint64_t PtrSizeMask = ~0ULL >> (64-IntPtrWidth);
Offset &= PtrSizeMask;
VariableScale &= PtrSizeMask;
// To do this transformation, any constant index must be a multiple of the
// variable scale factor. For example, we can evaluate "12 + 4*i" as "3 + i",
// but we can't evaluate "10 + 3*i" in terms of i. Check that the offset is a
// multiple of the variable scale.
int64_t NewOffs = Offset / (int64_t)VariableScale;
if (Offset != NewOffs*(int64_t)VariableScale)
return 0;
// Okay, we can do this evaluation. Start by converting the index to intptr.
const Type *IntPtrTy = TD.getIntPtrType(VariableIdx->getContext());
if (VariableIdx->getType() != IntPtrTy)
VariableIdx = CastInst::CreateIntegerCast(VariableIdx, IntPtrTy,
true /*SExt*/,
VariableIdx->getName(), &I);
Constant *OffsetVal = ConstantInt::get(IntPtrTy, NewOffs);
return BinaryOperator::CreateAdd(VariableIdx, OffsetVal, "offset", &I);
}
/// FoldGEPICmp - Fold comparisons between a GEP instruction and something
/// else. At this point we know that the GEP is on the LHS of the comparison.
Instruction *InstCombiner::FoldGEPICmp(GEPOperator *GEPLHS, Value *RHS,
ICmpInst::Predicate Cond,
Instruction &I) {
// Look through bitcasts.
if (BitCastInst *BCI = dyn_cast<BitCastInst>(RHS))
RHS = BCI->getOperand(0);
Value *PtrBase = GEPLHS->getOperand(0);
if (TD && PtrBase == RHS && GEPLHS->isInBounds()) {
// ((gep Ptr, OFFSET) cmp Ptr) ---> (OFFSET cmp 0).
// This transformation (ignoring the base and scales) is valid because we
// know pointers can't overflow since the gep is inbounds. See if we can
// output an optimized form.
Value *Offset = EvaluateGEPOffsetExpression(GEPLHS, I, *this);
// If not, synthesize the offset the hard way.
if (Offset == 0)
Offset = EmitGEPOffset(GEPLHS);
return new ICmpInst(ICmpInst::getSignedPredicate(Cond), Offset,
Constant::getNullValue(Offset->getType()));
} else if (GEPOperator *GEPRHS = dyn_cast<GEPOperator>(RHS)) {
// If the base pointers are different, but the indices are the same, just
// compare the base pointer.
if (PtrBase != GEPRHS->getOperand(0)) {
bool IndicesTheSame = GEPLHS->getNumOperands()==GEPRHS->getNumOperands();
IndicesTheSame &= GEPLHS->getOperand(0)->getType() ==
GEPRHS->getOperand(0)->getType();
if (IndicesTheSame)
for (unsigned i = 1, e = GEPLHS->getNumOperands(); i != e; ++i)
if (GEPLHS->getOperand(i) != GEPRHS->getOperand(i)) {
IndicesTheSame = false;
break;
}
// If all indices are the same, just compare the base pointers.
if (IndicesTheSame)
return new ICmpInst(ICmpInst::getSignedPredicate(Cond),
GEPLHS->getOperand(0), GEPRHS->getOperand(0));
// Otherwise, the base pointers are different and the indices are
// different, bail out.
return 0;
}
// If one of the GEPs has all zero indices, recurse.
bool AllZeros = true;
for (unsigned i = 1, e = GEPLHS->getNumOperands(); i != e; ++i)
if (!isa<Constant>(GEPLHS->getOperand(i)) ||
!cast<Constant>(GEPLHS->getOperand(i))->isNullValue()) {
AllZeros = false;
break;
}
if (AllZeros)
return FoldGEPICmp(GEPRHS, GEPLHS->getOperand(0),
ICmpInst::getSwappedPredicate(Cond), I);
// If the other GEP has all zero indices, recurse.
AllZeros = true;
for (unsigned i = 1, e = GEPRHS->getNumOperands(); i != e; ++i)
if (!isa<Constant>(GEPRHS->getOperand(i)) ||
!cast<Constant>(GEPRHS->getOperand(i))->isNullValue()) {
AllZeros = false;
break;
}
if (AllZeros)
return FoldGEPICmp(GEPLHS, GEPRHS->getOperand(0), Cond, I);
if (GEPLHS->getNumOperands() == GEPRHS->getNumOperands()) {
// If the GEPs only differ by one index, compare it.
unsigned NumDifferences = 0; // Keep track of # differences.
unsigned DiffOperand = 0; // The operand that differs.
for (unsigned i = 1, e = GEPRHS->getNumOperands(); i != e; ++i)
if (GEPLHS->getOperand(i) != GEPRHS->getOperand(i)) {
if (GEPLHS->getOperand(i)->getType()->getPrimitiveSizeInBits() !=
GEPRHS->getOperand(i)->getType()->getPrimitiveSizeInBits()) {
// Irreconcilable differences.
NumDifferences = 2;
break;
} else {
if (NumDifferences++) break;
DiffOperand = i;
}
}
if (NumDifferences == 0) // SAME GEP?
return ReplaceInstUsesWith(I, // No comparison is needed here.
ConstantInt::get(Type::getInt1Ty(I.getContext()),
ICmpInst::isTrueWhenEqual(Cond)));
else if (NumDifferences == 1) {
Value *LHSV = GEPLHS->getOperand(DiffOperand);
Value *RHSV = GEPRHS->getOperand(DiffOperand);
// Make sure we do a signed comparison here.
return new ICmpInst(ICmpInst::getSignedPredicate(Cond), LHSV, RHSV);
}
}
// Only lower this if the icmp is the only user of the GEP or if we expect
// the result to fold to a constant!
if (TD &&
(isa<ConstantExpr>(GEPLHS) || GEPLHS->hasOneUse()) &&
(isa<ConstantExpr>(GEPRHS) || GEPRHS->hasOneUse())) {
// ((gep Ptr, OFFSET1) cmp (gep Ptr, OFFSET2) ---> (OFFSET1 cmp OFFSET2)
Value *L = EmitGEPOffset(GEPLHS);
Value *R = EmitGEPOffset(GEPRHS);
return new ICmpInst(ICmpInst::getSignedPredicate(Cond), L, R);
}
}
return 0;
}
/// FoldICmpAddOpCst - Fold "icmp pred (X+CI), X".
Instruction *InstCombiner::FoldICmpAddOpCst(ICmpInst &ICI,
Value *X, ConstantInt *CI,
ICmpInst::Predicate Pred,
Value *TheAdd) {
// If we have X+0, exit early (simplifying logic below) and let it get folded
// elsewhere. icmp X+0, X -> icmp X, X
if (CI->isZero()) {
bool isTrue = ICmpInst::isTrueWhenEqual(Pred);
return ReplaceInstUsesWith(ICI, ConstantInt::get(ICI.getType(), isTrue));
}
// (X+4) == X -> false.
if (Pred == ICmpInst::ICMP_EQ)
return ReplaceInstUsesWith(ICI, ConstantInt::getFalse(X->getContext()));
// (X+4) != X -> true.
if (Pred == ICmpInst::ICMP_NE)
return ReplaceInstUsesWith(ICI, ConstantInt::getTrue(X->getContext()));
// If this is an instruction (as opposed to constantexpr) get NUW/NSW info.
bool isNUW = false, isNSW = false;
if (BinaryOperator *Add = dyn_cast<BinaryOperator>(TheAdd)) {
isNUW = Add->hasNoUnsignedWrap();
isNSW = Add->hasNoSignedWrap();
}
// From this point on, we know that (X+C <= X) --> (X+C < X) because C != 0,
// so the values can never be equal. Similiarly for all other "or equals"
// operators.
// (X+1) <u X --> X >u (MAXUINT-1) --> X == 255
// (X+2) <u X --> X >u (MAXUINT-2) --> X > 253
// (X+MAXUINT) <u X --> X >u (MAXUINT-MAXUINT) --> X != 0
if (Pred == ICmpInst::ICMP_ULT || Pred == ICmpInst::ICMP_ULE) {
// If this is an NUW add, then this is always false.
if (isNUW)
return ReplaceInstUsesWith(ICI, ConstantInt::getFalse(X->getContext()));
Value *R =
ConstantExpr::getSub(ConstantInt::getAllOnesValue(CI->getType()), CI);
return new ICmpInst(ICmpInst::ICMP_UGT, X, R);
}
// (X+1) >u X --> X <u (0-1) --> X != 255
// (X+2) >u X --> X <u (0-2) --> X <u 254
// (X+MAXUINT) >u X --> X <u (0-MAXUINT) --> X <u 1 --> X == 0
if (Pred == ICmpInst::ICMP_UGT || Pred == ICmpInst::ICMP_UGE) {
// If this is an NUW add, then this is always true.
if (isNUW)
return ReplaceInstUsesWith(ICI, ConstantInt::getTrue(X->getContext()));
return new ICmpInst(ICmpInst::ICMP_ULT, X, ConstantExpr::getNeg(CI));
}
unsigned BitWidth = CI->getType()->getPrimitiveSizeInBits();
ConstantInt *SMax = ConstantInt::get(X->getContext(),
APInt::getSignedMaxValue(BitWidth));
// (X+ 1) <s X --> X >s (MAXSINT-1) --> X == 127
// (X+ 2) <s X --> X >s (MAXSINT-2) --> X >s 125
// (X+MAXSINT) <s X --> X >s (MAXSINT-MAXSINT) --> X >s 0
// (X+MINSINT) <s X --> X >s (MAXSINT-MINSINT) --> X >s -1
// (X+ -2) <s X --> X >s (MAXSINT- -2) --> X >s 126
// (X+ -1) <s X --> X >s (MAXSINT- -1) --> X != 127
if (Pred == ICmpInst::ICMP_SLT || Pred == ICmpInst::ICMP_SLE) {
// If this is an NSW add, then we have two cases: if the constant is
// positive, then this is always false, if negative, this is always true.
if (isNSW) {
bool isTrue = CI->getValue().isNegative();
return ReplaceInstUsesWith(ICI, ConstantInt::get(ICI.getType(), isTrue));
}
return new ICmpInst(ICmpInst::ICMP_SGT, X, ConstantExpr::getSub(SMax, CI));
}
// (X+ 1) >s X --> X <s (MAXSINT-(1-1)) --> X != 127
// (X+ 2) >s X --> X <s (MAXSINT-(2-1)) --> X <s 126
// (X+MAXSINT) >s X --> X <s (MAXSINT-(MAXSINT-1)) --> X <s 1
// (X+MINSINT) >s X --> X <s (MAXSINT-(MINSINT-1)) --> X <s -2
// (X+ -2) >s X --> X <s (MAXSINT-(-2-1)) --> X <s -126
// (X+ -1) >s X --> X <s (MAXSINT-(-1-1)) --> X == -128
// If this is an NSW add, then we have two cases: if the constant is
// positive, then this is always true, if negative, this is always false.
if (isNSW) {
bool isTrue = !CI->getValue().isNegative();
return ReplaceInstUsesWith(ICI, ConstantInt::get(ICI.getType(), isTrue));
}
assert(Pred == ICmpInst::ICMP_SGT || Pred == ICmpInst::ICMP_SGE);
Constant *C = ConstantInt::get(X->getContext(), CI->getValue()-1);
return new ICmpInst(ICmpInst::ICMP_SLT, X, ConstantExpr::getSub(SMax, C));
}
/// FoldICmpDivCst - Fold "icmp pred, ([su]div X, DivRHS), CmpRHS" where DivRHS
/// and CmpRHS are both known to be integer constants.
Instruction *InstCombiner::FoldICmpDivCst(ICmpInst &ICI, BinaryOperator *DivI,
ConstantInt *DivRHS) {
ConstantInt *CmpRHS = cast<ConstantInt>(ICI.getOperand(1));
const APInt &CmpRHSV = CmpRHS->getValue();
// FIXME: If the operand types don't match the type of the divide
// then don't attempt this transform. The code below doesn't have the
// logic to deal with a signed divide and an unsigned compare (and
// vice versa). This is because (x /s C1) <s C2 produces different
// results than (x /s C1) <u C2 or (x /u C1) <s C2 or even
// (x /u C1) <u C2. Simply casting the operands and result won't
// work. :( The if statement below tests that condition and bails
// if it finds it.
bool DivIsSigned = DivI->getOpcode() == Instruction::SDiv;
if (!ICI.isEquality() && DivIsSigned != ICI.isSigned())
return 0;
if (DivRHS->isZero())
return 0; // The ProdOV computation fails on divide by zero.
if (DivIsSigned && DivRHS->isAllOnesValue())
return 0; // The overflow computation also screws up here
if (DivRHS->isOne())
return 0; // Not worth bothering, and eliminates some funny cases
// with INT_MIN.
// Compute Prod = CI * DivRHS. We are essentially solving an equation
// of form X/C1=C2. We solve for X by multiplying C1 (DivRHS) and
// C2 (CI). By solving for X we can turn this into a range check
// instead of computing a divide.
Constant *Prod = ConstantExpr::getMul(CmpRHS, DivRHS);
// Determine if the product overflows by seeing if the product is
// not equal to the divide. Make sure we do the same kind of divide
// as in the LHS instruction that we're folding.
bool ProdOV = (DivIsSigned ? ConstantExpr::getSDiv(Prod, DivRHS) :
ConstantExpr::getUDiv(Prod, DivRHS)) != CmpRHS;
// Get the ICmp opcode
ICmpInst::Predicate Pred = ICI.getPredicate();
// Figure out the interval that is being checked. For example, a comparison
// like "X /u 5 == 0" is really checking that X is in the interval [0, 5).
// Compute this interval based on the constants involved and the signedness of
// the compare/divide. This computes a half-open interval, keeping track of
// whether either value in the interval overflows. After analysis each
// overflow variable is set to 0 if it's corresponding bound variable is valid
// -1 if overflowed off the bottom end, or +1 if overflowed off the top end.
int LoOverflow = 0, HiOverflow = 0;
Constant *LoBound = 0, *HiBound = 0;
if (!DivIsSigned) { // udiv
// e.g. X/5 op 3 --> [15, 20)
LoBound = Prod;
HiOverflow = LoOverflow = ProdOV;
if (!HiOverflow)
HiOverflow = AddWithOverflow(HiBound, LoBound, DivRHS, false);
} else if (DivRHS->getValue().isStrictlyPositive()) { // Divisor is > 0.
if (CmpRHSV == 0) { // (X / pos) op 0
// Can't overflow. e.g. X/2 op 0 --> [-1, 2)
LoBound = cast<ConstantInt>(ConstantExpr::getNeg(SubOne(DivRHS)));
HiBound = DivRHS;
} else if (CmpRHSV.isStrictlyPositive()) { // (X / pos) op pos
LoBound = Prod; // e.g. X/5 op 3 --> [15, 20)
HiOverflow = LoOverflow = ProdOV;
if (!HiOverflow)
HiOverflow = AddWithOverflow(HiBound, Prod, DivRHS, true);
} else { // (X / pos) op neg
// e.g. X/5 op -3 --> [-15-4, -15+1) --> [-19, -14)
HiBound = AddOne(Prod);
LoOverflow = HiOverflow = ProdOV ? -1 : 0;
if (!LoOverflow) {
ConstantInt* DivNeg =
cast<ConstantInt>(ConstantExpr::getNeg(DivRHS));
LoOverflow = AddWithOverflow(LoBound, HiBound, DivNeg, true) ? -1 : 0;
}
}
} else if (DivRHS->getValue().isNegative()) { // Divisor is < 0.
if (CmpRHSV == 0) { // (X / neg) op 0
// e.g. X/-5 op 0 --> [-4, 5)
LoBound = AddOne(DivRHS);
HiBound = cast<ConstantInt>(ConstantExpr::getNeg(DivRHS));
if (HiBound == DivRHS) { // -INTMIN = INTMIN
HiOverflow = 1; // [INTMIN+1, overflow)
HiBound = 0; // e.g. X/INTMIN = 0 --> X > INTMIN
}
} else if (CmpRHSV.isStrictlyPositive()) { // (X / neg) op pos
// e.g. X/-5 op 3 --> [-19, -14)
HiBound = AddOne(Prod);
HiOverflow = LoOverflow = ProdOV ? -1 : 0;
if (!LoOverflow)
LoOverflow = AddWithOverflow(LoBound, HiBound, DivRHS, true) ? -1 : 0;
} else { // (X / neg) op neg
LoBound = Prod; // e.g. X/-5 op -3 --> [15, 20)
LoOverflow = HiOverflow = ProdOV;
if (!HiOverflow)
HiOverflow = SubWithOverflow(HiBound, Prod, DivRHS, true);
}
// Dividing by a negative swaps the condition. LT <-> GT
Pred = ICmpInst::getSwappedPredicate(Pred);
}
Value *X = DivI->getOperand(0);
switch (Pred) {
default: llvm_unreachable("Unhandled icmp opcode!");
case ICmpInst::ICMP_EQ:
if (LoOverflow && HiOverflow)
return ReplaceInstUsesWith(ICI, ConstantInt::getFalse(ICI.getContext()));
if (HiOverflow)
return new ICmpInst(DivIsSigned ? ICmpInst::ICMP_SGE :
ICmpInst::ICMP_UGE, X, LoBound);
if (LoOverflow)
return new ICmpInst(DivIsSigned ? ICmpInst::ICMP_SLT :
ICmpInst::ICMP_ULT, X, HiBound);
return ReplaceInstUsesWith(ICI,
InsertRangeTest(X, LoBound, HiBound, DivIsSigned,
true));
case ICmpInst::ICMP_NE:
if (LoOverflow && HiOverflow)
return ReplaceInstUsesWith(ICI, ConstantInt::getTrue(ICI.getContext()));
if (HiOverflow)
return new ICmpInst(DivIsSigned ? ICmpInst::ICMP_SLT :
ICmpInst::ICMP_ULT, X, LoBound);
if (LoOverflow)
return new ICmpInst(DivIsSigned ? ICmpInst::ICMP_SGE :
ICmpInst::ICMP_UGE, X, HiBound);
return ReplaceInstUsesWith(ICI, InsertRangeTest(X, LoBound, HiBound,
DivIsSigned, false));
case ICmpInst::ICMP_ULT:
case ICmpInst::ICMP_SLT:
if (LoOverflow == +1) // Low bound is greater than input range.
return ReplaceInstUsesWith(ICI, ConstantInt::getTrue(ICI.getContext()));
if (LoOverflow == -1) // Low bound is less than input range.
return ReplaceInstUsesWith(ICI, ConstantInt::getFalse(ICI.getContext()));
return new ICmpInst(Pred, X, LoBound);
case ICmpInst::ICMP_UGT:
case ICmpInst::ICMP_SGT:
if (HiOverflow == +1) // High bound greater than input range.
return ReplaceInstUsesWith(ICI, ConstantInt::getFalse(ICI.getContext()));
else if (HiOverflow == -1) // High bound less than input range.
return ReplaceInstUsesWith(ICI, ConstantInt::getTrue(ICI.getContext()));
if (Pred == ICmpInst::ICMP_UGT)
return new ICmpInst(ICmpInst::ICMP_UGE, X, HiBound);
else
return new ICmpInst(ICmpInst::ICMP_SGE, X, HiBound);
}
}
/// visitICmpInstWithInstAndIntCst - Handle "icmp (instr, intcst)".
///
Instruction *InstCombiner::visitICmpInstWithInstAndIntCst(ICmpInst &ICI,
Instruction *LHSI,
ConstantInt *RHS) {
const APInt &RHSV = RHS->getValue();
switch (LHSI->getOpcode()) {
case Instruction::Trunc:
if (ICI.isEquality() && LHSI->hasOneUse()) {
// Simplify icmp eq (trunc x to i8), 42 -> icmp eq x, 42|highbits if all
// of the high bits truncated out of x are known.
unsigned DstBits = LHSI->getType()->getPrimitiveSizeInBits(),
SrcBits = LHSI->getOperand(0)->getType()->getPrimitiveSizeInBits();
APInt Mask(APInt::getHighBitsSet(SrcBits, SrcBits-DstBits));
APInt KnownZero(SrcBits, 0), KnownOne(SrcBits, 0);
ComputeMaskedBits(LHSI->getOperand(0), Mask, KnownZero, KnownOne);
// If all the high bits are known, we can do this xform.
if ((KnownZero|KnownOne).countLeadingOnes() >= SrcBits-DstBits) {
// Pull in the high bits from known-ones set.
APInt NewRHS(RHS->getValue());
NewRHS.zext(SrcBits);
NewRHS |= KnownOne;
return new ICmpInst(ICI.getPredicate(), LHSI->getOperand(0),
ConstantInt::get(ICI.getContext(), NewRHS));
}
}
break;
case Instruction::Xor: // (icmp pred (xor X, XorCST), CI)
if (ConstantInt *XorCST = dyn_cast<ConstantInt>(LHSI->getOperand(1))) {
// If this is a comparison that tests the signbit (X < 0) or (x > -1),
// fold the xor.
if ((ICI.getPredicate() == ICmpInst::ICMP_SLT && RHSV == 0) ||
(ICI.getPredicate() == ICmpInst::ICMP_SGT && RHSV.isAllOnesValue())) {
Value *CompareVal = LHSI->getOperand(0);
// If the sign bit of the XorCST is not set, there is no change to
// the operation, just stop using the Xor.
if (!XorCST->getValue().isNegative()) {
ICI.setOperand(0, CompareVal);
Worklist.Add(LHSI);
return &ICI;
}
// Was the old condition true if the operand is positive?
bool isTrueIfPositive = ICI.getPredicate() == ICmpInst::ICMP_SGT;
// If so, the new one isn't.
isTrueIfPositive ^= true;
if (isTrueIfPositive)
return new ICmpInst(ICmpInst::ICMP_SGT, CompareVal,
SubOne(RHS));
else
return new ICmpInst(ICmpInst::ICMP_SLT, CompareVal,
AddOne(RHS));
}
if (LHSI->hasOneUse()) {
// (icmp u/s (xor A SignBit), C) -> (icmp s/u A, (xor C SignBit))
if (!ICI.isEquality() && XorCST->getValue().isSignBit()) {
const APInt &SignBit = XorCST->getValue();
ICmpInst::Predicate Pred = ICI.isSigned()
? ICI.getUnsignedPredicate()
: ICI.getSignedPredicate();
return new ICmpInst(Pred, LHSI->getOperand(0),
ConstantInt::get(ICI.getContext(),
RHSV ^ SignBit));
}
// (icmp u/s (xor A ~SignBit), C) -> (icmp s/u (xor C ~SignBit), A)
if (!ICI.isEquality() && XorCST->getValue().isMaxSignedValue()) {
const APInt &NotSignBit = XorCST->getValue();
ICmpInst::Predicate Pred = ICI.isSigned()
? ICI.getUnsignedPredicate()
: ICI.getSignedPredicate();
Pred = ICI.getSwappedPredicate(Pred);
return new ICmpInst(Pred, LHSI->getOperand(0),
ConstantInt::get(ICI.getContext(),
RHSV ^ NotSignBit));
}
}
}
break;
case Instruction::And: // (icmp pred (and X, AndCST), RHS)
if (LHSI->hasOneUse() && isa<ConstantInt>(LHSI->getOperand(1)) &&
LHSI->getOperand(0)->hasOneUse()) {
ConstantInt *AndCST = cast<ConstantInt>(LHSI->getOperand(1));
// If the LHS is an AND of a truncating cast, we can widen the
// and/compare to be the input width without changing the value
// produced, eliminating a cast.
if (TruncInst *Cast = dyn_cast<TruncInst>(LHSI->getOperand(0))) {
// We can do this transformation if either the AND constant does not
// have its sign bit set or if it is an equality comparison.
// Extending a relational comparison when we're checking the sign
// bit would not work.
if (Cast->hasOneUse() &&
(ICI.isEquality() ||
(AndCST->getValue().isNonNegative() && RHSV.isNonNegative()))) {
uint32_t BitWidth =
cast<IntegerType>(Cast->getOperand(0)->getType())->getBitWidth();
APInt NewCST = AndCST->getValue();
NewCST.zext(BitWidth);
APInt NewCI = RHSV;
NewCI.zext(BitWidth);
Value *NewAnd =
Builder->CreateAnd(Cast->getOperand(0),
ConstantInt::get(ICI.getContext(), NewCST),
LHSI->getName());
return new ICmpInst(ICI.getPredicate(), NewAnd,
ConstantInt::get(ICI.getContext(), NewCI));
}
}
// If this is: (X >> C1) & C2 != C3 (where any shift and any compare
// could exist), turn it into (X & (C2 << C1)) != (C3 << C1). This
// happens a LOT in code produced by the C front-end, for bitfield
// access.
BinaryOperator *Shift = dyn_cast<BinaryOperator>(LHSI->getOperand(0));
if (Shift && !Shift->isShift())
Shift = 0;
ConstantInt *ShAmt;
ShAmt = Shift ? dyn_cast<ConstantInt>(Shift->getOperand(1)) : 0;
const Type *Ty = Shift ? Shift->getType() : 0; // Type of the shift.
const Type *AndTy = AndCST->getType(); // Type of the and.
// We can fold this as long as we can't shift unknown bits
// into the mask. This can only happen with signed shift
// rights, as they sign-extend.
if (ShAmt) {
bool CanFold = Shift->isLogicalShift();
if (!CanFold) {
// To test for the bad case of the signed shr, see if any
// of the bits shifted in could be tested after the mask.
uint32_t TyBits = Ty->getPrimitiveSizeInBits();
int ShAmtVal = TyBits - ShAmt->getLimitedValue(TyBits);
uint32_t BitWidth = AndTy->getPrimitiveSizeInBits();
if ((APInt::getHighBitsSet(BitWidth, BitWidth-ShAmtVal) &
AndCST->getValue()) == 0)
CanFold = true;
}
if (CanFold) {
Constant *NewCst;
if (Shift->getOpcode() == Instruction::Shl)
NewCst = ConstantExpr::getLShr(RHS, ShAmt);
else
NewCst = ConstantExpr::getShl(RHS, ShAmt);
// Check to see if we are shifting out any of the bits being
// compared.
if (ConstantExpr::get(Shift->getOpcode(),
NewCst, ShAmt) != RHS) {
// If we shifted bits out, the fold is not going to work out.
// As a special case, check to see if this means that the
// result is always true or false now.
if (ICI.getPredicate() == ICmpInst::ICMP_EQ)
return ReplaceInstUsesWith(ICI,
ConstantInt::getFalse(ICI.getContext()));
if (ICI.getPredicate() == ICmpInst::ICMP_NE)
return ReplaceInstUsesWith(ICI,
ConstantInt::getTrue(ICI.getContext()));
} else {
ICI.setOperand(1, NewCst);
Constant *NewAndCST;
if (Shift->getOpcode() == Instruction::Shl)
NewAndCST = ConstantExpr::getLShr(AndCST, ShAmt);
else
NewAndCST = ConstantExpr::getShl(AndCST, ShAmt);
LHSI->setOperand(1, NewAndCST);
LHSI->setOperand(0, Shift->getOperand(0));
Worklist.Add(Shift); // Shift is dead.
return &ICI;
}
}
}
// Turn ((X >> Y) & C) == 0 into (X & (C << Y)) == 0. The later is
// preferable because it allows the C<<Y expression to be hoisted out
// of a loop if Y is invariant and X is not.
if (Shift && Shift->hasOneUse() && RHSV == 0 &&
ICI.isEquality() && !Shift->isArithmeticShift() &&
!isa<Constant>(Shift->getOperand(0))) {
// Compute C << Y.
Value *NS;
if (Shift->getOpcode() == Instruction::LShr) {
NS = Builder->CreateShl(AndCST, Shift->getOperand(1), "tmp");
} else {
// Insert a logical shift.
NS = Builder->CreateLShr(AndCST, Shift->getOperand(1), "tmp");
}
// Compute X & (C << Y).
Value *NewAnd =
Builder->CreateAnd(Shift->getOperand(0), NS, LHSI->getName());
ICI.setOperand(0, NewAnd);
return &ICI;
}
}
// Try to optimize things like "A[i]&42 == 0" to index computations.
if (LoadInst *LI = dyn_cast<LoadInst>(LHSI->getOperand(0))) {
if (GetElementPtrInst *GEP =
dyn_cast<GetElementPtrInst>(LI->getOperand(0)))
if (GlobalVariable *GV = dyn_cast<GlobalVariable>(GEP->getOperand(0)))
if (GV->isConstant() && GV->hasDefinitiveInitializer() &&
!LI->isVolatile() && isa<ConstantInt>(LHSI->getOperand(1))) {
ConstantInt *C = cast<ConstantInt>(LHSI->getOperand(1));
if (Instruction *Res = FoldCmpLoadFromIndexedGlobal(GEP, GV,ICI, C))
return Res;
}
}
break;
case Instruction::Or: {
if (!ICI.isEquality() || !RHS->isNullValue() || !LHSI->hasOneUse())
break;
Value *P, *Q;
if (match(LHSI, m_Or(m_PtrToInt(m_Value(P)), m_PtrToInt(m_Value(Q))))) {
// Simplify icmp eq (or (ptrtoint P), (ptrtoint Q)), 0
// -> and (icmp eq P, null), (icmp eq Q, null).
Value *ICIP = Builder->CreateICmp(ICI.getPredicate(), P,
Constant::getNullValue(P->getType()));
Value *ICIQ = Builder->CreateICmp(ICI.getPredicate(), Q,
Constant::getNullValue(Q->getType()));
Instruction *Op;
if (ICI.getPredicate() == ICmpInst::ICMP_EQ)
Op = BinaryOperator::CreateAnd(ICIP, ICIQ);
else
Op = BinaryOperator::CreateOr(ICIP, ICIQ);
return Op;
}
break;
}
case Instruction::Shl: { // (icmp pred (shl X, ShAmt), CI)
ConstantInt *ShAmt = dyn_cast<ConstantInt>(LHSI->getOperand(1));
if (!ShAmt) break;
uint32_t TypeBits = RHSV.getBitWidth();
// Check that the shift amount is in range. If not, don't perform
// undefined shifts. When the shift is visited it will be
// simplified.
if (ShAmt->uge(TypeBits))
break;
if (ICI.isEquality()) {
// If we are comparing against bits always shifted out, the
// comparison cannot succeed.
Constant *Comp =
ConstantExpr::getShl(ConstantExpr::getLShr(RHS, ShAmt),
ShAmt);
if (Comp != RHS) {// Comparing against a bit that we know is zero.
bool IsICMP_NE = ICI.getPredicate() == ICmpInst::ICMP_NE;
Constant *Cst =
ConstantInt::get(Type::getInt1Ty(ICI.getContext()), IsICMP_NE);
return ReplaceInstUsesWith(ICI, Cst);
}
if (LHSI->hasOneUse()) {
// Otherwise strength reduce the shift into an and.
uint32_t ShAmtVal = (uint32_t)ShAmt->getLimitedValue(TypeBits);
Constant *Mask =
ConstantInt::get(ICI.getContext(), APInt::getLowBitsSet(TypeBits,
TypeBits-ShAmtVal));
Value *And =
Builder->CreateAnd(LHSI->getOperand(0),Mask, LHSI->getName()+".mask");
return new ICmpInst(ICI.getPredicate(), And,
ConstantInt::get(ICI.getContext(),
RHSV.lshr(ShAmtVal)));
}
}
// Otherwise, if this is a comparison of the sign bit, simplify to and/test.
bool TrueIfSigned = false;
if (LHSI->hasOneUse() &&
isSignBitCheck(ICI.getPredicate(), RHS, TrueIfSigned)) {
// (X << 31) <s 0 --> (X&1) != 0
Constant *Mask = ConstantInt::get(ICI.getContext(), APInt(TypeBits, 1) <<
(TypeBits-ShAmt->getZExtValue()-1));
Value *And =
Builder->CreateAnd(LHSI->getOperand(0), Mask, LHSI->getName()+".mask");
return new ICmpInst(TrueIfSigned ? ICmpInst::ICMP_NE : ICmpInst::ICMP_EQ,
And, Constant::getNullValue(And->getType()));
}
break;
}
case Instruction::LShr: // (icmp pred (shr X, ShAmt), CI)
case Instruction::AShr: {
// Only handle equality comparisons of shift-by-constant.
ConstantInt *ShAmt = dyn_cast<ConstantInt>(LHSI->getOperand(1));
if (!ShAmt || !ICI.isEquality()) break;
// Check that the shift amount is in range. If not, don't perform
// undefined shifts. When the shift is visited it will be
// simplified.
uint32_t TypeBits = RHSV.getBitWidth();
if (ShAmt->uge(TypeBits))
break;
uint32_t ShAmtVal = (uint32_t)ShAmt->getLimitedValue(TypeBits);
// If we are comparing against bits always shifted out, the
// comparison cannot succeed.
APInt Comp = RHSV << ShAmtVal;
if (LHSI->getOpcode() == Instruction::LShr)
Comp = Comp.lshr(ShAmtVal);
else
Comp = Comp.ashr(ShAmtVal);
if (Comp != RHSV) { // Comparing against a bit that we know is zero.
bool IsICMP_NE = ICI.getPredicate() == ICmpInst::ICMP_NE;
Constant *Cst = ConstantInt::get(Type::getInt1Ty(ICI.getContext()),
IsICMP_NE);
return ReplaceInstUsesWith(ICI, Cst);
}
// Otherwise, check to see if the bits shifted out are known to be zero.
// If so, we can compare against the unshifted value:
// (X & 4) >> 1 == 2 --> (X & 4) == 4.
if (LHSI->hasOneUse() &&
MaskedValueIsZero(LHSI->getOperand(0),
APInt::getLowBitsSet(Comp.getBitWidth(), ShAmtVal))) {
return new ICmpInst(ICI.getPredicate(), LHSI->getOperand(0),
ConstantExpr::getShl(RHS, ShAmt));
}
if (LHSI->hasOneUse()) {
// Otherwise strength reduce the shift into an and.
APInt Val(APInt::getHighBitsSet(TypeBits, TypeBits - ShAmtVal));
Constant *Mask = ConstantInt::get(ICI.getContext(), Val);
Value *And = Builder->CreateAnd(LHSI->getOperand(0),
Mask, LHSI->getName()+".mask");
return new ICmpInst(ICI.getPredicate(), And,
ConstantExpr::getShl(RHS, ShAmt));
}
break;
}
case Instruction::SDiv:
case Instruction::UDiv:
// Fold: icmp pred ([us]div X, C1), C2 -> range test
// Fold this div into the comparison, producing a range check.
// Determine, based on the divide type, what the range is being
// checked. If there is an overflow on the low or high side, remember
// it, otherwise compute the range [low, hi) bounding the new value.
// See: InsertRangeTest above for the kinds of replacements possible.
if (ConstantInt *DivRHS = dyn_cast<ConstantInt>(LHSI->getOperand(1)))
if (Instruction *R = FoldICmpDivCst(ICI, cast<BinaryOperator>(LHSI),
DivRHS))
return R;
break;
case Instruction::Add:
// Fold: icmp pred (add X, C1), C2
if (!ICI.isEquality()) {
ConstantInt *LHSC = dyn_cast<ConstantInt>(LHSI->getOperand(1));
if (!LHSC) break;
const APInt &LHSV = LHSC->getValue();
ConstantRange CR = ICI.makeConstantRange(ICI.getPredicate(), RHSV)
.subtract(LHSV);
if (ICI.isSigned()) {
if (CR.getLower().isSignBit()) {
return new ICmpInst(ICmpInst::ICMP_SLT, LHSI->getOperand(0),
ConstantInt::get(ICI.getContext(),CR.getUpper()));
} else if (CR.getUpper().isSignBit()) {
return new ICmpInst(ICmpInst::ICMP_SGE, LHSI->getOperand(0),
ConstantInt::get(ICI.getContext(),CR.getLower()));
}
} else {
if (CR.getLower().isMinValue()) {
return new ICmpInst(ICmpInst::ICMP_ULT, LHSI->getOperand(0),
ConstantInt::get(ICI.getContext(),CR.getUpper()));
} else if (CR.getUpper().isMinValue()) {
return new ICmpInst(ICmpInst::ICMP_UGE, LHSI->getOperand(0),
ConstantInt::get(ICI.getContext(),CR.getLower()));
}
}
}
break;
}
// Simplify icmp_eq and icmp_ne instructions with integer constant RHS.
if (ICI.isEquality()) {
bool isICMP_NE = ICI.getPredicate() == ICmpInst::ICMP_NE;
// If the first operand is (add|sub|and|or|xor|rem) with a constant, and
// the second operand is a constant, simplify a bit.
if (BinaryOperator *BO = dyn_cast<BinaryOperator>(LHSI)) {
switch (BO->getOpcode()) {
case Instruction::SRem:
// If we have a signed (X % (2^c)) == 0, turn it into an unsigned one.
if (RHSV == 0 && isa<ConstantInt>(BO->getOperand(1)) &&BO->hasOneUse()){
const APInt &V = cast<ConstantInt>(BO->getOperand(1))->getValue();
if (V.sgt(1) && V.isPowerOf2()) {
Value *NewRem =
Builder->CreateURem(BO->getOperand(0), BO->getOperand(1),
BO->getName());
return new ICmpInst(ICI.getPredicate(), NewRem,
Constant::getNullValue(BO->getType()));
}
}
break;
case Instruction::Add:
// Replace ((add A, B) != C) with (A != C-B) if B & C are constants.
if (ConstantInt *BOp1C = dyn_cast<ConstantInt>(BO->getOperand(1))) {
if (BO->hasOneUse())
return new ICmpInst(ICI.getPredicate(), BO->getOperand(0),
ConstantExpr::getSub(RHS, BOp1C));
} else if (RHSV == 0) {
// Replace ((add A, B) != 0) with (A != -B) if A or B is
// efficiently invertible, or if the add has just this one use.
Value *BOp0 = BO->getOperand(0), *BOp1 = BO->getOperand(1);
if (Value *NegVal = dyn_castNegVal(BOp1))
return new ICmpInst(ICI.getPredicate(), BOp0, NegVal);
else if (Value *NegVal = dyn_castNegVal(BOp0))
return new ICmpInst(ICI.getPredicate(), NegVal, BOp1);
else if (BO->hasOneUse()) {
Value *Neg = Builder->CreateNeg(BOp1);
Neg->takeName(BO);
return new ICmpInst(ICI.getPredicate(), BOp0, Neg);
}
}
break;
case Instruction::Xor:
// For the xor case, we can xor two constants together, eliminating
// the explicit xor.
if (Constant *BOC = dyn_cast<Constant>(BO->getOperand(1)))
return new ICmpInst(ICI.getPredicate(), BO->getOperand(0),
ConstantExpr::getXor(RHS, BOC));
// FALLTHROUGH
case Instruction::Sub:
// Replace (([sub|xor] A, B) != 0) with (A != B)
if (RHSV == 0)
return new ICmpInst(ICI.getPredicate(), BO->getOperand(0),
BO->getOperand(1));
break;
case Instruction::Or:
// If bits are being or'd in that are not present in the constant we
// are comparing against, then the comparison could never succeed!
if (Constant *BOC = dyn_cast<Constant>(BO->getOperand(1))) {
Constant *NotCI = ConstantExpr::getNot(RHS);
if (!ConstantExpr::getAnd(BOC, NotCI)->isNullValue())
return ReplaceInstUsesWith(ICI,
ConstantInt::get(Type::getInt1Ty(ICI.getContext()),
isICMP_NE));
}
break;
case Instruction::And:
if (ConstantInt *BOC = dyn_cast<ConstantInt>(BO->getOperand(1))) {
// If bits are being compared against that are and'd out, then the
// comparison can never succeed!
if ((RHSV & ~BOC->getValue()) != 0)
return ReplaceInstUsesWith(ICI,
ConstantInt::get(Type::getInt1Ty(ICI.getContext()),
isICMP_NE));
// If we have ((X & C) == C), turn it into ((X & C) != 0).
if (RHS == BOC && RHSV.isPowerOf2())
return new ICmpInst(isICMP_NE ? ICmpInst::ICMP_EQ :
ICmpInst::ICMP_NE, LHSI,
Constant::getNullValue(RHS->getType()));
// Replace (and X, (1 << size(X)-1) != 0) with x s< 0
if (BOC->getValue().isSignBit()) {
Value *X = BO->getOperand(0);
Constant *Zero = Constant::getNullValue(X->getType());
ICmpInst::Predicate pred = isICMP_NE ?
ICmpInst::ICMP_SLT : ICmpInst::ICMP_SGE;
return new ICmpInst(pred, X, Zero);
}
// ((X & ~7) == 0) --> X < 8
if (RHSV == 0 && isHighOnes(BOC)) {
Value *X = BO->getOperand(0);
Constant *NegX = ConstantExpr::getNeg(BOC);
ICmpInst::Predicate pred = isICMP_NE ?
ICmpInst::ICMP_UGE : ICmpInst::ICMP_ULT;
return new ICmpInst(pred, X, NegX);
}
}
default: break;
}
} else if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(LHSI)) {
// Handle icmp {eq|ne} <intrinsic>, intcst.
switch (II->getIntrinsicID()) {
case Intrinsic::bswap:
Worklist.Add(II);
ICI.setOperand(0, II->getOperand(1));
ICI.setOperand(1, ConstantInt::get(II->getContext(), RHSV.byteSwap()));
return &ICI;
case Intrinsic::ctlz:
case Intrinsic::cttz:
// ctz(A) == bitwidth(a) -> A == 0 and likewise for !=
if (RHSV == RHS->getType()->getBitWidth()) {
Worklist.Add(II);
ICI.setOperand(0, II->getOperand(1));
ICI.setOperand(1, ConstantInt::get(RHS->getType(), 0));
return &ICI;
}
break;
case Intrinsic::ctpop:
// popcount(A) == 0 -> A == 0 and likewise for !=
if (RHS->isZero()) {
Worklist.Add(II);
ICI.setOperand(0, II->getOperand(1));
ICI.setOperand(1, RHS);
return &ICI;
}
break;
default:
break;
}
}
}
return 0;
}
/// visitICmpInstWithCastAndCast - Handle icmp (cast x to y), (cast/cst).
/// We only handle extending casts so far.
///
Instruction *InstCombiner::visitICmpInstWithCastAndCast(ICmpInst &ICI) {
const CastInst *LHSCI = cast<CastInst>(ICI.getOperand(0));
Value *LHSCIOp = LHSCI->getOperand(0);
const Type *SrcTy = LHSCIOp->getType();
const Type *DestTy = LHSCI->getType();
Value *RHSCIOp;
// Turn icmp (ptrtoint x), (ptrtoint/c) into a compare of the input if the
// integer type is the same size as the pointer type.
if (TD && LHSCI->getOpcode() == Instruction::PtrToInt &&
TD->getPointerSizeInBits() ==
cast<IntegerType>(DestTy)->getBitWidth()) {
Value *RHSOp = 0;
if (Constant *RHSC = dyn_cast<Constant>(ICI.getOperand(1))) {
RHSOp = ConstantExpr::getIntToPtr(RHSC, SrcTy);
} else if (PtrToIntInst *RHSC = dyn_cast<PtrToIntInst>(ICI.getOperand(1))) {
RHSOp = RHSC->getOperand(0);
// If the pointer types don't match, insert a bitcast.
if (LHSCIOp->getType() != RHSOp->getType())
RHSOp = Builder->CreateBitCast(RHSOp, LHSCIOp->getType());
}
if (RHSOp)
return new ICmpInst(ICI.getPredicate(), LHSCIOp, RHSOp);
}
// The code below only handles extension cast instructions, so far.
// Enforce this.
if (LHSCI->getOpcode() != Instruction::ZExt &&
LHSCI->getOpcode() != Instruction::SExt)
return 0;
bool isSignedExt = LHSCI->getOpcode() == Instruction::SExt;
bool isSignedCmp = ICI.isSigned();
if (CastInst *CI = dyn_cast<CastInst>(ICI.getOperand(1))) {
// Not an extension from the same type?
RHSCIOp = CI->getOperand(0);
if (RHSCIOp->getType() != LHSCIOp->getType())
return 0;
// If the signedness of the two casts doesn't agree (i.e. one is a sext
// and the other is a zext), then we can't handle this.
if (CI->getOpcode() != LHSCI->getOpcode())
return 0;
// Deal with equality cases early.
if (ICI.isEquality())
return new ICmpInst(ICI.getPredicate(), LHSCIOp, RHSCIOp);
// A signed comparison of sign extended values simplifies into a
// signed comparison.
if (isSignedCmp && isSignedExt)
return new ICmpInst(ICI.getPredicate(), LHSCIOp, RHSCIOp);
// The other three cases all fold into an unsigned comparison.
return new ICmpInst(ICI.getUnsignedPredicate(), LHSCIOp, RHSCIOp);
}
// If we aren't dealing with a constant on the RHS, exit early
ConstantInt *CI = dyn_cast<ConstantInt>(ICI.getOperand(1));
if (!CI)
return 0;
// Compute the constant that would happen if we truncated to SrcTy then
// reextended to DestTy.
Constant *Res1 = ConstantExpr::getTrunc(CI, SrcTy);
Constant *Res2 = ConstantExpr::getCast(LHSCI->getOpcode(),
Res1, DestTy);
// If the re-extended constant didn't change...
if (Res2 == CI) {
// Deal with equality cases early.
if (ICI.isEquality())
return new ICmpInst(ICI.getPredicate(), LHSCIOp, Res1);
// A signed comparison of sign extended values simplifies into a
// signed comparison.
if (isSignedExt && isSignedCmp)
return new ICmpInst(ICI.getPredicate(), LHSCIOp, Res1);
// The other three cases all fold into an unsigned comparison.
return new ICmpInst(ICI.getUnsignedPredicate(), LHSCIOp, Res1);
}
// The re-extended constant changed so the constant cannot be represented
// in the shorter type. Consequently, we cannot emit a simple comparison.
// First, handle some easy cases. We know the result cannot be equal at this
// point so handle the ICI.isEquality() cases
if (ICI.getPredicate() == ICmpInst::ICMP_EQ)
return ReplaceInstUsesWith(ICI, ConstantInt::getFalse(ICI.getContext()));
if (ICI.getPredicate() == ICmpInst::ICMP_NE)
return ReplaceInstUsesWith(ICI, ConstantInt::getTrue(ICI.getContext()));
// Evaluate the comparison for LT (we invert for GT below). LE and GE cases
// should have been folded away previously and not enter in here.
Value *Result;
if (isSignedCmp) {
// We're performing a signed comparison.
if (cast<ConstantInt>(CI)->getValue().isNegative())
Result = ConstantInt::getFalse(ICI.getContext()); // X < (small) --> false
else
Result = ConstantInt::getTrue(ICI.getContext()); // X < (large) --> true
} else {
// We're performing an unsigned comparison.
if (isSignedExt) {
// We're performing an unsigned comp with a sign extended value.
// This is true if the input is >= 0. [aka >s -1]
Constant *NegOne = Constant::getAllOnesValue(SrcTy);
Result = Builder->CreateICmpSGT(LHSCIOp, NegOne, ICI.getName());
} else {
// Unsigned extend & unsigned compare -> always true.
Result = ConstantInt::getTrue(ICI.getContext());
}
}
// Finally, return the value computed.
if (ICI.getPredicate() == ICmpInst::ICMP_ULT ||
ICI.getPredicate() == ICmpInst::ICMP_SLT)
return ReplaceInstUsesWith(ICI, Result);
assert((ICI.getPredicate()==ICmpInst::ICMP_UGT ||
ICI.getPredicate()==ICmpInst::ICMP_SGT) &&
"ICmp should be folded!");
if (Constant *CI = dyn_cast<Constant>(Result))
return ReplaceInstUsesWith(ICI, ConstantExpr::getNot(CI));
return BinaryOperator::CreateNot(Result);
}
Instruction *InstCombiner::visitICmpInst(ICmpInst &I) {
bool Changed = false;
Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
/// Orders the operands of the compare so that they are listed from most
/// complex to least complex. This puts constants before unary operators,
/// before binary operators.
if (getComplexity(Op0) < getComplexity(Op1)) {
I.swapOperands();
std::swap(Op0, Op1);
Changed = true;
}
if (Value *V = SimplifyICmpInst(I.getPredicate(), Op0, Op1, TD))
return ReplaceInstUsesWith(I, V);
const Type *Ty = Op0->getType();
// icmp's with boolean values can always be turned into bitwise operations
if (Ty->isIntegerTy(1)) {
switch (I.getPredicate()) {
default: llvm_unreachable("Invalid icmp instruction!");
case ICmpInst::ICMP_EQ: { // icmp eq i1 A, B -> ~(A^B)
Value *Xor = Builder->CreateXor(Op0, Op1, I.getName()+"tmp");
return BinaryOperator::CreateNot(Xor);
}
case ICmpInst::ICMP_NE: // icmp eq i1 A, B -> A^B
return BinaryOperator::CreateXor(Op0, Op1);
case ICmpInst::ICMP_UGT:
std::swap(Op0, Op1); // Change icmp ugt -> icmp ult
// FALL THROUGH
case ICmpInst::ICMP_ULT:{ // icmp ult i1 A, B -> ~A & B
Value *Not = Builder->CreateNot(Op0, I.getName()+"tmp");
return BinaryOperator::CreateAnd(Not, Op1);
}
case ICmpInst::ICMP_SGT:
std::swap(Op0, Op1); // Change icmp sgt -> icmp slt
// FALL THROUGH
case ICmpInst::ICMP_SLT: { // icmp slt i1 A, B -> A & ~B
Value *Not = Builder->CreateNot(Op1, I.getName()+"tmp");
return BinaryOperator::CreateAnd(Not, Op0);
}
case ICmpInst::ICMP_UGE:
std::swap(Op0, Op1); // Change icmp uge -> icmp ule
// FALL THROUGH
case ICmpInst::ICMP_ULE: { // icmp ule i1 A, B -> ~A | B
Value *Not = Builder->CreateNot(Op0, I.getName()+"tmp");
return BinaryOperator::CreateOr(Not, Op1);
}
case ICmpInst::ICMP_SGE:
std::swap(Op0, Op1); // Change icmp sge -> icmp sle
// FALL THROUGH
case ICmpInst::ICMP_SLE: { // icmp sle i1 A, B -> A | ~B
Value *Not = Builder->CreateNot(Op1, I.getName()+"tmp");
return BinaryOperator::CreateOr(Not, Op0);
}
}
}
unsigned BitWidth = 0;
if (TD)
BitWidth = TD->getTypeSizeInBits(Ty->getScalarType());
else if (Ty->isIntOrIntVectorTy())
BitWidth = Ty->getScalarSizeInBits();
bool isSignBit = false;
// See if we are doing a comparison with a constant.
if (ConstantInt *CI = dyn_cast<ConstantInt>(Op1)) {
Value *A = 0, *B = 0;
// (icmp ne/eq (sub A B) 0) -> (icmp ne/eq A, B)
if (I.isEquality() && CI->isZero() &&
match(Op0, m_Sub(m_Value(A), m_Value(B)))) {
// (icmp cond A B) if cond is equality
return new ICmpInst(I.getPredicate(), A, B);
}
// If we have an icmp le or icmp ge instruction, turn it into the
// appropriate icmp lt or icmp gt instruction. This allows us to rely on
// them being folded in the code below. The SimplifyICmpInst code has
// already handled the edge cases for us, so we just assert on them.
switch (I.getPredicate()) {
default: break;
case ICmpInst::ICMP_ULE:
assert(!CI->isMaxValue(false)); // A <=u MAX -> TRUE
return new ICmpInst(ICmpInst::ICMP_ULT, Op0,
ConstantInt::get(CI->getContext(), CI->getValue()+1));
case ICmpInst::ICMP_SLE:
assert(!CI->isMaxValue(true)); // A <=s MAX -> TRUE
return new ICmpInst(ICmpInst::ICMP_SLT, Op0,
ConstantInt::get(CI->getContext(), CI->getValue()+1));
case ICmpInst::ICMP_UGE:
assert(!CI->isMinValue(false)); // A >=u MIN -> TRUE
return new ICmpInst(ICmpInst::ICMP_UGT, Op0,
ConstantInt::get(CI->getContext(), CI->getValue()-1));
case ICmpInst::ICMP_SGE:
assert(!CI->isMinValue(true)); // A >=s MIN -> TRUE
return new ICmpInst(ICmpInst::ICMP_SGT, Op0,
ConstantInt::get(CI->getContext(), CI->getValue()-1));
}
// If this comparison is a normal comparison, it demands all
// bits, if it is a sign bit comparison, it only demands the sign bit.
bool UnusedBit;
isSignBit = isSignBitCheck(I.getPredicate(), CI, UnusedBit);
}
// See if we can fold the comparison based on range information we can get
// by checking whether bits are known to be zero or one in the input.
if (BitWidth != 0) {
APInt Op0KnownZero(BitWidth, 0), Op0KnownOne(BitWidth, 0);
APInt Op1KnownZero(BitWidth, 0), Op1KnownOne(BitWidth, 0);
if (SimplifyDemandedBits(I.getOperandUse(0),
isSignBit ? APInt::getSignBit(BitWidth)
: APInt::getAllOnesValue(BitWidth),
Op0KnownZero, Op0KnownOne, 0))
return &I;
if (SimplifyDemandedBits(I.getOperandUse(1),
APInt::getAllOnesValue(BitWidth),
Op1KnownZero, Op1KnownOne, 0))
return &I;
// Given the known and unknown bits, compute a range that the LHS could be
// in. Compute the Min, Max and RHS values based on the known bits. For the
// EQ and NE we use unsigned values.
APInt Op0Min(BitWidth, 0), Op0Max(BitWidth, 0);
APInt Op1Min(BitWidth, 0), Op1Max(BitWidth, 0);
if (I.isSigned()) {
ComputeSignedMinMaxValuesFromKnownBits(Op0KnownZero, Op0KnownOne,
Op0Min, Op0Max);
ComputeSignedMinMaxValuesFromKnownBits(Op1KnownZero, Op1KnownOne,
Op1Min, Op1Max);
} else {
ComputeUnsignedMinMaxValuesFromKnownBits(Op0KnownZero, Op0KnownOne,
Op0Min, Op0Max);
ComputeUnsignedMinMaxValuesFromKnownBits(Op1KnownZero, Op1KnownOne,
Op1Min, Op1Max);
}
// If Min and Max are known to be the same, then SimplifyDemandedBits
// figured out that the LHS is a constant. Just constant fold this now so
// that code below can assume that Min != Max.
if (!isa<Constant>(Op0) && Op0Min == Op0Max)
return new ICmpInst(I.getPredicate(),
ConstantInt::get(I.getContext(), Op0Min), Op1);
if (!isa<Constant>(Op1) && Op1Min == Op1Max)
return new ICmpInst(I.getPredicate(), Op0,
ConstantInt::get(I.getContext(), Op1Min));
// Based on the range information we know about the LHS, see if we can
// simplify this comparison. For example, (x&4) < 8 is always true.
switch (I.getPredicate()) {
default: llvm_unreachable("Unknown icmp opcode!");
case ICmpInst::ICMP_EQ:
if (Op0Max.ult(Op1Min) || Op0Min.ugt(Op1Max))
return ReplaceInstUsesWith(I, ConstantInt::getFalse(I.getContext()));
break;
case ICmpInst::ICMP_NE:
if (Op0Max.ult(Op1Min) || Op0Min.ugt(Op1Max))
return ReplaceInstUsesWith(I, ConstantInt::getTrue(I.getContext()));
break;
case ICmpInst::ICMP_ULT:
if (Op0Max.ult(Op1Min)) // A <u B -> true if max(A) < min(B)
return ReplaceInstUsesWith(I, ConstantInt::getTrue(I.getContext()));
if (Op0Min.uge(Op1Max)) // A <u B -> false if min(A) >= max(B)
return ReplaceInstUsesWith(I, ConstantInt::getFalse(I.getContext()));
if (Op1Min == Op0Max) // A <u B -> A != B if max(A) == min(B)
return new ICmpInst(ICmpInst::ICMP_NE, Op0, Op1);
if (ConstantInt *CI = dyn_cast<ConstantInt>(Op1)) {
if (Op1Max == Op0Min+1) // A <u C -> A == C-1 if min(A)+1 == C
return new ICmpInst(ICmpInst::ICMP_EQ, Op0,
ConstantInt::get(CI->getContext(), CI->getValue()-1));
// (x <u 2147483648) -> (x >s -1) -> true if sign bit clear
if (CI->isMinValue(true))
return new ICmpInst(ICmpInst::ICMP_SGT, Op0,
Constant::getAllOnesValue(Op0->getType()));
}
break;
case ICmpInst::ICMP_UGT:
if (Op0Min.ugt(Op1Max)) // A >u B -> true if min(A) > max(B)
return ReplaceInstUsesWith(I, ConstantInt::getTrue(I.getContext()));
if (Op0Max.ule(Op1Min)) // A >u B -> false if max(A) <= max(B)
return ReplaceInstUsesWith(I, ConstantInt::getFalse(I.getContext()));
if (Op1Max == Op0Min) // A >u B -> A != B if min(A) == max(B)
return new ICmpInst(ICmpInst::ICMP_NE, Op0, Op1);
if (ConstantInt *CI = dyn_cast<ConstantInt>(Op1)) {
if (Op1Min == Op0Max-1) // A >u C -> A == C+1 if max(a)-1 == C
return new ICmpInst(ICmpInst::ICMP_EQ, Op0,
ConstantInt::get(CI->getContext(), CI->getValue()+1));
// (x >u 2147483647) -> (x <s 0) -> true if sign bit set
if (CI->isMaxValue(true))
return new ICmpInst(ICmpInst::ICMP_SLT, Op0,
Constant::getNullValue(Op0->getType()));
}
break;
case ICmpInst::ICMP_SLT:
if (Op0Max.slt(Op1Min)) // A <s B -> true if max(A) < min(C)
return ReplaceInstUsesWith(I, ConstantInt::getTrue(I.getContext()));
if (Op0Min.sge(Op1Max)) // A <s B -> false if min(A) >= max(C)
return ReplaceInstUsesWith(I, ConstantInt::getFalse(I.getContext()));
if (Op1Min == Op0Max) // A <s B -> A != B if max(A) == min(B)
return new ICmpInst(ICmpInst::ICMP_NE, Op0, Op1);
if (ConstantInt *CI = dyn_cast<ConstantInt>(Op1)) {
if (Op1Max == Op0Min+1) // A <s C -> A == C-1 if min(A)+1 == C
return new ICmpInst(ICmpInst::ICMP_EQ, Op0,
ConstantInt::get(CI->getContext(), CI->getValue()-1));
}
break;
case ICmpInst::ICMP_SGT:
if (Op0Min.sgt(Op1Max)) // A >s B -> true if min(A) > max(B)
return ReplaceInstUsesWith(I, ConstantInt::getTrue(I.getContext()));
if (Op0Max.sle(Op1Min)) // A >s B -> false if max(A) <= min(B)
return ReplaceInstUsesWith(I, ConstantInt::getFalse(I.getContext()));
if (Op1Max == Op0Min) // A >s B -> A != B if min(A) == max(B)
return new ICmpInst(ICmpInst::ICMP_NE, Op0, Op1);
if (ConstantInt *CI = dyn_cast<ConstantInt>(Op1)) {
if (Op1Min == Op0Max-1) // A >s C -> A == C+1 if max(A)-1 == C
return new ICmpInst(ICmpInst::ICMP_EQ, Op0,
ConstantInt::get(CI->getContext(), CI->getValue()+1));
}
break;
case ICmpInst::ICMP_SGE:
assert(!isa<ConstantInt>(Op1) && "ICMP_SGE with ConstantInt not folded!");
if (Op0Min.sge(Op1Max)) // A >=s B -> true if min(A) >= max(B)
return ReplaceInstUsesWith(I, ConstantInt::getTrue(I.getContext()));
if (Op0Max.slt(Op1Min)) // A >=s B -> false if max(A) < min(B)
return ReplaceInstUsesWith(I, ConstantInt::getFalse(I.getContext()));
break;
case ICmpInst::ICMP_SLE:
assert(!isa<ConstantInt>(Op1) && "ICMP_SLE with ConstantInt not folded!");
if (Op0Max.sle(Op1Min)) // A <=s B -> true if max(A) <= min(B)
return ReplaceInstUsesWith(I, ConstantInt::getTrue(I.getContext()));
if (Op0Min.sgt(Op1Max)) // A <=s B -> false if min(A) > max(B)
return ReplaceInstUsesWith(I, ConstantInt::getFalse(I.getContext()));
break;
case ICmpInst::ICMP_UGE:
assert(!isa<ConstantInt>(Op1) && "ICMP_UGE with ConstantInt not folded!");
if (Op0Min.uge(Op1Max)) // A >=u B -> true if min(A) >= max(B)
return ReplaceInstUsesWith(I, ConstantInt::getTrue(I.getContext()));
if (Op0Max.ult(Op1Min)) // A >=u B -> false if max(A) < min(B)
return ReplaceInstUsesWith(I, ConstantInt::getFalse(I.getContext()));
break;
case ICmpInst::ICMP_ULE:
assert(!isa<ConstantInt>(Op1) && "ICMP_ULE with ConstantInt not folded!");
if (Op0Max.ule(Op1Min)) // A <=u B -> true if max(A) <= min(B)
return ReplaceInstUsesWith(I, ConstantInt::getTrue(I.getContext()));
if (Op0Min.ugt(Op1Max)) // A <=u B -> false if min(A) > max(B)
return ReplaceInstUsesWith(I, ConstantInt::getFalse(I.getContext()));
break;
}
// Turn a signed comparison into an unsigned one if both operands
// are known to have the same sign.
if (I.isSigned() &&
((Op0KnownZero.isNegative() && Op1KnownZero.isNegative()) ||
(Op0KnownOne.isNegative() && Op1KnownOne.isNegative())))
return new ICmpInst(I.getUnsignedPredicate(), Op0, Op1);
}
// Test if the ICmpInst instruction is used exclusively by a select as
// part of a minimum or maximum operation. If so, refrain from doing
// any other folding. This helps out other analyses which understand
// non-obfuscated minimum and maximum idioms, such as ScalarEvolution
// and CodeGen. And in this case, at least one of the comparison
// operands has at least one user besides the compare (the select),
// which would often largely negate the benefit of folding anyway.
if (I.hasOneUse())
if (SelectInst *SI = dyn_cast<SelectInst>(*I.use_begin()))
if ((SI->getOperand(1) == Op0 && SI->getOperand(2) == Op1) ||
(SI->getOperand(2) == Op0 && SI->getOperand(1) == Op1))
return 0;
// See if we are doing a comparison between a constant and an instruction that
// can be folded into the comparison.
if (ConstantInt *CI = dyn_cast<ConstantInt>(Op1)) {
// Since the RHS is a ConstantInt (CI), if the left hand side is an
// instruction, see if that instruction also has constants so that the
// instruction can be folded into the icmp
if (Instruction *LHSI = dyn_cast<Instruction>(Op0))
if (Instruction *Res = visitICmpInstWithInstAndIntCst(I, LHSI, CI))
return Res;
}
// Handle icmp with constant (but not simple integer constant) RHS
if (Constant *RHSC = dyn_cast<Constant>(Op1)) {
if (Instruction *LHSI = dyn_cast<Instruction>(Op0))
switch (LHSI->getOpcode()) {
case Instruction::GetElementPtr:
// icmp pred GEP (P, int 0, int 0, int 0), null -> icmp pred P, null
if (RHSC->isNullValue() &&
cast<GetElementPtrInst>(LHSI)->hasAllZeroIndices())
return new ICmpInst(I.getPredicate(), LHSI->getOperand(0),
Constant::getNullValue(LHSI->getOperand(0)->getType()));
break;
case Instruction::PHI:
// Only fold icmp into the PHI if the phi and icmp are in the same
// block. If in the same block, we're encouraging jump threading. If
// not, we are just pessimizing the code by making an i1 phi.
if (LHSI->getParent() == I.getParent())
if (Instruction *NV = FoldOpIntoPhi(I, true))
return NV;
break;
case Instruction::Select: {
// If either operand of the select is a constant, we can fold the
// comparison into the select arms, which will cause one to be
// constant folded and the select turned into a bitwise or.
Value *Op1 = 0, *Op2 = 0;
if (Constant *C = dyn_cast<Constant>(LHSI->getOperand(1)))
Op1 = ConstantExpr::getICmp(I.getPredicate(), C, RHSC);
if (Constant *C = dyn_cast<Constant>(LHSI->getOperand(2)))
Op2 = ConstantExpr::getICmp(I.getPredicate(), C, RHSC);
// We only want to perform this transformation if it will not lead to
// additional code. This is true if either both sides of the select
// fold to a constant (in which case the icmp is replaced with a select
// which will usually simplify) or this is the only user of the
// select (in which case we are trading a select+icmp for a simpler
// select+icmp).
if ((Op1 && Op2) || (LHSI->hasOneUse() && (Op1 || Op2))) {
if (!Op1)
Op1 = Builder->CreateICmp(I.getPredicate(), LHSI->getOperand(1),
RHSC, I.getName());
if (!Op2)
Op2 = Builder->CreateICmp(I.getPredicate(), LHSI->getOperand(2),
RHSC, I.getName());
return SelectInst::Create(LHSI->getOperand(0), Op1, Op2);
}
break;
}
case Instruction::IntToPtr:
// icmp pred inttoptr(X), null -> icmp pred X, 0
if (RHSC->isNullValue() && TD &&
TD->getIntPtrType(RHSC->getContext()) ==
LHSI->getOperand(0)->getType())
return new ICmpInst(I.getPredicate(), LHSI->getOperand(0),
Constant::getNullValue(LHSI->getOperand(0)->getType()));
break;
case Instruction::Load:
// Try to optimize things like "A[i] > 4" to index computations.
if (GetElementPtrInst *GEP =
dyn_cast<GetElementPtrInst>(LHSI->getOperand(0))) {
if (GlobalVariable *GV = dyn_cast<GlobalVariable>(GEP->getOperand(0)))
if (GV->isConstant() && GV->hasDefinitiveInitializer() &&
!cast<LoadInst>(LHSI)->isVolatile())
if (Instruction *Res = FoldCmpLoadFromIndexedGlobal(GEP, GV, I))
return Res;
}
break;
}
}
// If we can optimize a 'icmp GEP, P' or 'icmp P, GEP', do so now.
if (GEPOperator *GEP = dyn_cast<GEPOperator>(Op0))
if (Instruction *NI = FoldGEPICmp(GEP, Op1, I.getPredicate(), I))
return NI;
if (GEPOperator *GEP = dyn_cast<GEPOperator>(Op1))
if (Instruction *NI = FoldGEPICmp(GEP, Op0,
ICmpInst::getSwappedPredicate(I.getPredicate()), I))
return NI;
// Test to see if the operands of the icmp are casted versions of other
// values. If the ptr->ptr cast can be stripped off both arguments, we do so
// now.
if (BitCastInst *CI = dyn_cast<BitCastInst>(Op0)) {
if (Op0->getType()->isPointerTy() &&
(isa<Constant>(Op1) || isa<BitCastInst>(Op1))) {
// We keep moving the cast from the left operand over to the right
// operand, where it can often be eliminated completely.
Op0 = CI->getOperand(0);
// If operand #1 is a bitcast instruction, it must also be a ptr->ptr cast
// so eliminate it as well.
if (BitCastInst *CI2 = dyn_cast<BitCastInst>(Op1))
Op1 = CI2->getOperand(0);
// If Op1 is a constant, we can fold the cast into the constant.
if (Op0->getType() != Op1->getType()) {
if (Constant *Op1C = dyn_cast<Constant>(Op1)) {
Op1 = ConstantExpr::getBitCast(Op1C, Op0->getType());
} else {
// Otherwise, cast the RHS right before the icmp
Op1 = Builder->CreateBitCast(Op1, Op0->getType());
}
}
return new ICmpInst(I.getPredicate(), Op0, Op1);
}
}
if (isa<CastInst>(Op0)) {
// Handle the special case of: icmp (cast bool to X), <cst>
// This comes up when you have code like
// int X = A < B;
// if (X) ...
// For generality, we handle any zero-extension of any operand comparison
// with a constant or another cast from the same type.
if (isa<Constant>(Op1) || isa<CastInst>(Op1))
if (Instruction *R = visitICmpInstWithCastAndCast(I))
return R;
}
// See if it's the same type of instruction on the left and right.
if (BinaryOperator *Op0I = dyn_cast<BinaryOperator>(Op0)) {
if (BinaryOperator *Op1I = dyn_cast<BinaryOperator>(Op1)) {
if (Op0I->getOpcode() == Op1I->getOpcode() && Op0I->hasOneUse() &&
Op1I->hasOneUse() && Op0I->getOperand(1) == Op1I->getOperand(1)) {
switch (Op0I->getOpcode()) {
default: break;
case Instruction::Add:
case Instruction::Sub:
case Instruction::Xor:
if (I.isEquality()) // a+x icmp eq/ne b+x --> a icmp b
return new ICmpInst(I.getPredicate(), Op0I->getOperand(0),
Op1I->getOperand(0));
// icmp u/s (a ^ signbit), (b ^ signbit) --> icmp s/u a, b
if (ConstantInt *CI = dyn_cast<ConstantInt>(Op0I->getOperand(1))) {
if (CI->getValue().isSignBit()) {
ICmpInst::Predicate Pred = I.isSigned()
? I.getUnsignedPredicate()
: I.getSignedPredicate();
return new ICmpInst(Pred, Op0I->getOperand(0),
Op1I->getOperand(0));
}
if (CI->getValue().isMaxSignedValue()) {
ICmpInst::Predicate Pred = I.isSigned()
? I.getUnsignedPredicate()
: I.getSignedPredicate();
Pred = I.getSwappedPredicate(Pred);
return new ICmpInst(Pred, Op0I->getOperand(0),
Op1I->getOperand(0));
}
}
break;
case Instruction::Mul:
if (!I.isEquality())
break;
if (ConstantInt *CI = dyn_cast<ConstantInt>(Op0I->getOperand(1))) {
// a * Cst icmp eq/ne b * Cst --> a & Mask icmp b & Mask
// Mask = -1 >> count-trailing-zeros(Cst).
if (!CI->isZero() && !CI->isOne()) {
const APInt &AP = CI->getValue();
ConstantInt *Mask = ConstantInt::get(I.getContext(),
APInt::getLowBitsSet(AP.getBitWidth(),
AP.getBitWidth() -
AP.countTrailingZeros()));
Value *And1 = Builder->CreateAnd(Op0I->getOperand(0), Mask);
Value *And2 = Builder->CreateAnd(Op1I->getOperand(0), Mask);
return new ICmpInst(I.getPredicate(), And1, And2);
}
}
break;
}
}
}
}
// ~x < ~y --> y < x
{ Value *A, *B;
if (match(Op0, m_Not(m_Value(A))) &&
match(Op1, m_Not(m_Value(B))))
return new ICmpInst(I.getPredicate(), B, A);
}
if (I.isEquality()) {
Value *A, *B, *C, *D;
// -x == -y --> x == y
if (match(Op0, m_Neg(m_Value(A))) &&
match(Op1, m_Neg(m_Value(B))))
return new ICmpInst(I.getPredicate(), A, B);
if (match(Op0, m_Xor(m_Value(A), m_Value(B)))) {
if (A == Op1 || B == Op1) { // (A^B) == A -> B == 0
Value *OtherVal = A == Op1 ? B : A;
return new ICmpInst(I.getPredicate(), OtherVal,
Constant::getNullValue(A->getType()));
}
if (match(Op1, m_Xor(m_Value(C), m_Value(D)))) {
// A^c1 == C^c2 --> A == C^(c1^c2)
ConstantInt *C1, *C2;
if (match(B, m_ConstantInt(C1)) &&
match(D, m_ConstantInt(C2)) && Op1->hasOneUse()) {
Constant *NC = ConstantInt::get(I.getContext(),
C1->getValue() ^ C2->getValue());
Value *Xor = Builder->CreateXor(C, NC, "tmp");
return new ICmpInst(I.getPredicate(), A, Xor);
}
// A^B == A^D -> B == D
if (A == C) return new ICmpInst(I.getPredicate(), B, D);
if (A == D) return new ICmpInst(I.getPredicate(), B, C);
if (B == C) return new ICmpInst(I.getPredicate(), A, D);
if (B == D) return new ICmpInst(I.getPredicate(), A, C);
}
}
if (match(Op1, m_Xor(m_Value(A), m_Value(B))) &&
(A == Op0 || B == Op0)) {
// A == (A^B) -> B == 0
Value *OtherVal = A == Op0 ? B : A;
return new ICmpInst(I.getPredicate(), OtherVal,
Constant::getNullValue(A->getType()));
}
// (A-B) == A -> B == 0
if (match(Op0, m_Sub(m_Specific(Op1), m_Value(B))))
return new ICmpInst(I.getPredicate(), B,
Constant::getNullValue(B->getType()));
// A == (A-B) -> B == 0
if (match(Op1, m_Sub(m_Specific(Op0), m_Value(B))))
return new ICmpInst(I.getPredicate(), B,
Constant::getNullValue(B->getType()));
// (X&Z) == (Y&Z) -> (X^Y) & Z == 0
if (Op0->hasOneUse() && Op1->hasOneUse() &&
match(Op0, m_And(m_Value(A), m_Value(B))) &&
match(Op1, m_And(m_Value(C), m_Value(D)))) {
Value *X = 0, *Y = 0, *Z = 0;
if (A == C) {
X = B; Y = D; Z = A;
} else if (A == D) {
X = B; Y = C; Z = A;
} else if (B == C) {
X = A; Y = D; Z = B;
} else if (B == D) {
X = A; Y = C; Z = B;
}
if (X) { // Build (X^Y) & Z
Op1 = Builder->CreateXor(X, Y, "tmp");
Op1 = Builder->CreateAnd(Op1, Z, "tmp");
I.setOperand(0, Op1);
I.setOperand(1, Constant::getNullValue(Op1->getType()));
return &I;
}
}
}
{
Value *X; ConstantInt *Cst;
// icmp X+Cst, X
if (match(Op0, m_Add(m_Value(X), m_ConstantInt(Cst))) && Op1 == X)
return FoldICmpAddOpCst(I, X, Cst, I.getPredicate(), Op0);
// icmp X, X+Cst
if (match(Op1, m_Add(m_Value(X), m_ConstantInt(Cst))) && Op0 == X)
return FoldICmpAddOpCst(I, X, Cst, I.getSwappedPredicate(), Op1);
}
return Changed ? &I : 0;
}
/// FoldFCmp_IntToFP_Cst - Fold fcmp ([us]itofp x, cst) if possible.
///
Instruction *InstCombiner::FoldFCmp_IntToFP_Cst(FCmpInst &I,
Instruction *LHSI,
Constant *RHSC) {
if (!isa<ConstantFP>(RHSC)) return 0;
const APFloat &RHS = cast<ConstantFP>(RHSC)->getValueAPF();
// Get the width of the mantissa. We don't want to hack on conversions that
// might lose information from the integer, e.g. "i64 -> float"
int MantissaWidth = LHSI->getType()->getFPMantissaWidth();
if (MantissaWidth == -1) return 0; // Unknown.
// Check to see that the input is converted from an integer type that is small
// enough that preserves all bits. TODO: check here for "known" sign bits.
// This would allow us to handle (fptosi (x >>s 62) to float) if x is i64 f.e.
unsigned InputSize = LHSI->getOperand(0)->getType()->getScalarSizeInBits();
// If this is a uitofp instruction, we need an extra bit to hold the sign.
bool LHSUnsigned = isa<UIToFPInst>(LHSI);
if (LHSUnsigned)
++InputSize;
// If the conversion would lose info, don't hack on this.
if ((int)InputSize > MantissaWidth)
return 0;
// Otherwise, we can potentially simplify the comparison. We know that it
// will always come through as an integer value and we know the constant is
// not a NAN (it would have been previously simplified).
assert(!RHS.isNaN() && "NaN comparison not already folded!");
ICmpInst::Predicate Pred;
switch (I.getPredicate()) {
default: llvm_unreachable("Unexpected predicate!");
case FCmpInst::FCMP_UEQ:
case FCmpInst::FCMP_OEQ:
Pred = ICmpInst::ICMP_EQ;
break;
case FCmpInst::FCMP_UGT:
case FCmpInst::FCMP_OGT:
Pred = LHSUnsigned ? ICmpInst::ICMP_UGT : ICmpInst::ICMP_SGT;
break;
case FCmpInst::FCMP_UGE:
case FCmpInst::FCMP_OGE:
Pred = LHSUnsigned ? ICmpInst::ICMP_UGE : ICmpInst::ICMP_SGE;
break;
case FCmpInst::FCMP_ULT:
case FCmpInst::FCMP_OLT:
Pred = LHSUnsigned ? ICmpInst::ICMP_ULT : ICmpInst::ICMP_SLT;
break;
case FCmpInst::FCMP_ULE:
case FCmpInst::FCMP_OLE:
Pred = LHSUnsigned ? ICmpInst::ICMP_ULE : ICmpInst::ICMP_SLE;
break;
case FCmpInst::FCMP_UNE:
case FCmpInst::FCMP_ONE:
Pred = ICmpInst::ICMP_NE;
break;
case FCmpInst::FCMP_ORD:
return ReplaceInstUsesWith(I, ConstantInt::getTrue(I.getContext()));
case FCmpInst::FCMP_UNO:
return ReplaceInstUsesWith(I, ConstantInt::getFalse(I.getContext()));
}
const IntegerType *IntTy = cast<IntegerType>(LHSI->getOperand(0)->getType());
// Now we know that the APFloat is a normal number, zero or inf.
// See if the FP constant is too large for the integer. For example,
// comparing an i8 to 300.0.
unsigned IntWidth = IntTy->getScalarSizeInBits();
if (!LHSUnsigned) {
// If the RHS value is > SignedMax, fold the comparison. This handles +INF
// and large values.
APFloat SMax(RHS.getSemantics(), APFloat::fcZero, false);
SMax.convertFromAPInt(APInt::getSignedMaxValue(IntWidth), true,
APFloat::rmNearestTiesToEven);
if (SMax.compare(RHS) == APFloat::cmpLessThan) { // smax < 13123.0
if (Pred == ICmpInst::ICMP_NE || Pred == ICmpInst::ICMP_SLT ||
Pred == ICmpInst::ICMP_SLE)
return ReplaceInstUsesWith(I, ConstantInt::getTrue(I.getContext()));
return ReplaceInstUsesWith(I, ConstantInt::getFalse(I.getContext()));
}
} else {
// If the RHS value is > UnsignedMax, fold the comparison. This handles
// +INF and large values.
APFloat UMax(RHS.getSemantics(), APFloat::fcZero, false);
UMax.convertFromAPInt(APInt::getMaxValue(IntWidth), false,
APFloat::rmNearestTiesToEven);
if (UMax.compare(RHS) == APFloat::cmpLessThan) { // umax < 13123.0
if (Pred == ICmpInst::ICMP_NE || Pred == ICmpInst::ICMP_ULT ||
Pred == ICmpInst::ICMP_ULE)
return ReplaceInstUsesWith(I, ConstantInt::getTrue(I.getContext()));
return ReplaceInstUsesWith(I, ConstantInt::getFalse(I.getContext()));
}
}
if (!LHSUnsigned) {
// See if the RHS value is < SignedMin.
APFloat SMin(RHS.getSemantics(), APFloat::fcZero, false);
SMin.convertFromAPInt(APInt::getSignedMinValue(IntWidth), true,
APFloat::rmNearestTiesToEven);
if (SMin.compare(RHS) == APFloat::cmpGreaterThan) { // smin > 12312.0
if (Pred == ICmpInst::ICMP_NE || Pred == ICmpInst::ICMP_SGT ||
Pred == ICmpInst::ICMP_SGE)
return ReplaceInstUsesWith(I, ConstantInt::getTrue(I.getContext()));
return ReplaceInstUsesWith(I, ConstantInt::getFalse(I.getContext()));
}
}
// Okay, now we know that the FP constant fits in the range [SMIN, SMAX] or
// [0, UMAX], but it may still be fractional. See if it is fractional by
// casting the FP value to the integer value and back, checking for equality.
// Don't do this for zero, because -0.0 is not fractional.
Constant *RHSInt = LHSUnsigned
? ConstantExpr::getFPToUI(RHSC, IntTy)
: ConstantExpr::getFPToSI(RHSC, IntTy);
if (!RHS.isZero()) {
bool Equal = LHSUnsigned
? ConstantExpr::getUIToFP(RHSInt, RHSC->getType()) == RHSC
: ConstantExpr::getSIToFP(RHSInt, RHSC->getType()) == RHSC;
if (!Equal) {
// If we had a comparison against a fractional value, we have to adjust
// the compare predicate and sometimes the value. RHSC is rounded towards
// zero at this point.
switch (Pred) {
default: llvm_unreachable("Unexpected integer comparison!");
case ICmpInst::ICMP_NE: // (float)int != 4.4 --> true
return ReplaceInstUsesWith(I, ConstantInt::getTrue(I.getContext()));
case ICmpInst::ICMP_EQ: // (float)int == 4.4 --> false
return ReplaceInstUsesWith(I, ConstantInt::getFalse(I.getContext()));
case ICmpInst::ICMP_ULE:
// (float)int <= 4.4 --> int <= 4
// (float)int <= -4.4 --> false
if (RHS.isNegative())
return ReplaceInstUsesWith(I, ConstantInt::getFalse(I.getContext()));
break;
case ICmpInst::ICMP_SLE:
// (float)int <= 4.4 --> int <= 4
// (float)int <= -4.4 --> int < -4
if (RHS.isNegative())
Pred = ICmpInst::ICMP_SLT;
break;
case ICmpInst::ICMP_ULT:
// (float)int < -4.4 --> false
// (float)int < 4.4 --> int <= 4
if (RHS.isNegative())
return ReplaceInstUsesWith(I, ConstantInt::getFalse(I.getContext()));
Pred = ICmpInst::ICMP_ULE;
break;
case ICmpInst::ICMP_SLT:
// (float)int < -4.4 --> int < -4
// (float)int < 4.4 --> int <= 4
if (!RHS.isNegative())
Pred = ICmpInst::ICMP_SLE;
break;
case ICmpInst::ICMP_UGT:
// (float)int > 4.4 --> int > 4
// (float)int > -4.4 --> true
if (RHS.isNegative())
return ReplaceInstUsesWith(I, ConstantInt::getTrue(I.getContext()));
break;
case ICmpInst::ICMP_SGT:
// (float)int > 4.4 --> int > 4
// (float)int > -4.4 --> int >= -4
if (RHS.isNegative())
Pred = ICmpInst::ICMP_SGE;
break;
case ICmpInst::ICMP_UGE:
// (float)int >= -4.4 --> true
// (float)int >= 4.4 --> int > 4
if (!RHS.isNegative())
return ReplaceInstUsesWith(I, ConstantInt::getTrue(I.getContext()));
Pred = ICmpInst::ICMP_UGT;
break;
case ICmpInst::ICMP_SGE:
// (float)int >= -4.4 --> int >= -4
// (float)int >= 4.4 --> int > 4
if (!RHS.isNegative())
Pred = ICmpInst::ICMP_SGT;
break;
}
}
}
// Lower this FP comparison into an appropriate integer version of the
// comparison.
return new ICmpInst(Pred, LHSI->getOperand(0), RHSInt);
}
Instruction *InstCombiner::visitFCmpInst(FCmpInst &I) {
bool Changed = false;
/// Orders the operands of the compare so that they are listed from most
/// complex to least complex. This puts constants before unary operators,
/// before binary operators.
if (getComplexity(I.getOperand(0)) < getComplexity(I.getOperand(1))) {
I.swapOperands();
Changed = true;
}
Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
if (Value *V = SimplifyFCmpInst(I.getPredicate(), Op0, Op1, TD))
return ReplaceInstUsesWith(I, V);
// Simplify 'fcmp pred X, X'
if (Op0 == Op1) {
switch (I.getPredicate()) {
default: llvm_unreachable("Unknown predicate!");
case FCmpInst::FCMP_UNO: // True if unordered: isnan(X) | isnan(Y)
case FCmpInst::FCMP_ULT: // True if unordered or less than
case FCmpInst::FCMP_UGT: // True if unordered or greater than
case FCmpInst::FCMP_UNE: // True if unordered or not equal
// Canonicalize these to be 'fcmp uno %X, 0.0'.
I.setPredicate(FCmpInst::FCMP_UNO);
I.setOperand(1, Constant::getNullValue(Op0->getType()));
return &I;
case FCmpInst::FCMP_ORD: // True if ordered (no nans)
case FCmpInst::FCMP_OEQ: // True if ordered and equal
case FCmpInst::FCMP_OGE: // True if ordered and greater than or equal
case FCmpInst::FCMP_OLE: // True if ordered and less than or equal
// Canonicalize these to be 'fcmp ord %X, 0.0'.
I.setPredicate(FCmpInst::FCMP_ORD);
I.setOperand(1, Constant::getNullValue(Op0->getType()));
return &I;
}
}
// Handle fcmp with constant RHS
if (Constant *RHSC = dyn_cast<Constant>(Op1)) {
if (Instruction *LHSI = dyn_cast<Instruction>(Op0))
switch (LHSI->getOpcode()) {
case Instruction::PHI:
// Only fold fcmp into the PHI if the phi and fcmp are in the same
// block. If in the same block, we're encouraging jump threading. If
// not, we are just pessimizing the code by making an i1 phi.
if (LHSI->getParent() == I.getParent())
if (Instruction *NV = FoldOpIntoPhi(I, true))
return NV;
break;
case Instruction::SIToFP:
case Instruction::UIToFP:
if (Instruction *NV = FoldFCmp_IntToFP_Cst(I, LHSI, RHSC))
return NV;
break;
case Instruction::Select: {
// If either operand of the select is a constant, we can fold the
// comparison into the select arms, which will cause one to be
// constant folded and the select turned into a bitwise or.
Value *Op1 = 0, *Op2 = 0;
if (LHSI->hasOneUse()) {
if (Constant *C = dyn_cast<Constant>(LHSI->getOperand(1))) {
// Fold the known value into the constant operand.
Op1 = ConstantExpr::getCompare(I.getPredicate(), C, RHSC);
// Insert a new FCmp of the other select operand.
Op2 = Builder->CreateFCmp(I.getPredicate(),
LHSI->getOperand(2), RHSC, I.getName());
} else if (Constant *C = dyn_cast<Constant>(LHSI->getOperand(2))) {
// Fold the known value into the constant operand.
Op2 = ConstantExpr::getCompare(I.getPredicate(), C, RHSC);
// Insert a new FCmp of the other select operand.
Op1 = Builder->CreateFCmp(I.getPredicate(), LHSI->getOperand(1),
RHSC, I.getName());
}
}
if (Op1)
return SelectInst::Create(LHSI->getOperand(0), Op1, Op2);
break;
}
case Instruction::Load:
if (GetElementPtrInst *GEP =
dyn_cast<GetElementPtrInst>(LHSI->getOperand(0))) {
if (GlobalVariable *GV = dyn_cast<GlobalVariable>(GEP->getOperand(0)))
if (GV->isConstant() && GV->hasDefinitiveInitializer() &&
!cast<LoadInst>(LHSI)->isVolatile())
if (Instruction *Res = FoldCmpLoadFromIndexedGlobal(GEP, GV, I))
return Res;
}
break;
}
}
return Changed ? &I : 0;
}
|