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
|
//===- ValueTracking.cpp - Walk computations to compute properties --------===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file contains routines that help analyze properties that chains of
// computations have.
//
//===----------------------------------------------------------------------===//
#include "llvm/Analysis/ValueTracking.h"
#include "llvm/Constants.h"
#include "llvm/Instructions.h"
#include "llvm/GlobalVariable.h"
#include "llvm/GlobalAlias.h"
#include "llvm/IntrinsicInst.h"
#include "llvm/LLVMContext.h"
#include "llvm/Operator.h"
#include "llvm/Target/TargetData.h"
#include "llvm/Support/GetElementPtrTypeIterator.h"
#include "llvm/Support/MathExtras.h"
#include <cstring>
using namespace llvm;
/// ComputeMaskedBits - Determine which of the bits specified in Mask are
/// known to be either zero or one and return them in the KnownZero/KnownOne
/// bit sets. This code only analyzes bits in Mask, in order to short-circuit
/// processing.
/// NOTE: we cannot consider 'undef' to be "IsZero" here. The problem is that
/// we cannot optimize based on the assumption that it is zero without changing
/// it to be an explicit zero. If we don't change it to zero, other code could
/// optimized based on the contradictory assumption that it is non-zero.
/// Because instcombine aggressively folds operations with undef args anyway,
/// this won't lose us code quality.
///
/// This function is defined on values with integer type, values with pointer
/// type (but only if TD is non-null), and vectors of integers. In the case
/// where V is a vector, the mask, known zero, and known one values are the
/// same width as the vector element, and the bit is set only if it is true
/// for all of the elements in the vector.
void llvm::ComputeMaskedBits(Value *V, const APInt &Mask,
APInt &KnownZero, APInt &KnownOne,
const TargetData *TD, unsigned Depth) {
const unsigned MaxDepth = 6;
assert(V && "No Value?");
assert(Depth <= MaxDepth && "Limit Search Depth");
unsigned BitWidth = Mask.getBitWidth();
assert((V->getType()->isIntOrIntVector() || isa<PointerType>(V->getType())) &&
"Not integer or pointer type!");
assert((!TD ||
TD->getTypeSizeInBits(V->getType()->getScalarType()) == BitWidth) &&
(!V->getType()->isIntOrIntVector() ||
V->getType()->getScalarSizeInBits() == BitWidth) &&
KnownZero.getBitWidth() == BitWidth &&
KnownOne.getBitWidth() == BitWidth &&
"V, Mask, KnownOne and KnownZero should have same BitWidth");
if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) {
// We know all of the bits for a constant!
KnownOne = CI->getValue() & Mask;
KnownZero = ~KnownOne & Mask;
return;
}
// Null and aggregate-zero are all-zeros.
if (isa<ConstantPointerNull>(V) ||
isa<ConstantAggregateZero>(V)) {
KnownOne.clear();
KnownZero = Mask;
return;
}
// Handle a constant vector by taking the intersection of the known bits of
// each element.
if (ConstantVector *CV = dyn_cast<ConstantVector>(V)) {
KnownZero.set(); KnownOne.set();
for (unsigned i = 0, e = CV->getNumOperands(); i != e; ++i) {
APInt KnownZero2(BitWidth, 0), KnownOne2(BitWidth, 0);
ComputeMaskedBits(CV->getOperand(i), Mask, KnownZero2, KnownOne2,
TD, Depth);
KnownZero &= KnownZero2;
KnownOne &= KnownOne2;
}
return;
}
// The address of an aligned GlobalValue has trailing zeros.
if (GlobalValue *GV = dyn_cast<GlobalValue>(V)) {
unsigned Align = GV->getAlignment();
if (Align == 0 && TD && GV->getType()->getElementType()->isSized()) {
const Type *ObjectType = GV->getType()->getElementType();
// If the object is defined in the current Module, we'll be giving
// it the preferred alignment. Otherwise, we have to assume that it
// may only have the minimum ABI alignment.
if (!GV->isDeclaration() && !GV->mayBeOverridden())
Align = TD->getPrefTypeAlignment(ObjectType);
else
Align = TD->getABITypeAlignment(ObjectType);
}
if (Align > 0)
KnownZero = Mask & APInt::getLowBitsSet(BitWidth,
CountTrailingZeros_32(Align));
else
KnownZero.clear();
KnownOne.clear();
return;
}
// A weak GlobalAlias is totally unknown. A non-weak GlobalAlias has
// the bits of its aliasee.
if (GlobalAlias *GA = dyn_cast<GlobalAlias>(V)) {
if (GA->mayBeOverridden()) {
KnownZero.clear(); KnownOne.clear();
} else {
ComputeMaskedBits(GA->getAliasee(), Mask, KnownZero, KnownOne,
TD, Depth+1);
}
return;
}
KnownZero.clear(); KnownOne.clear(); // Start out not knowing anything.
if (Depth == MaxDepth || Mask == 0)
return; // Limit search depth.
Operator *I = dyn_cast<Operator>(V);
if (!I) return;
APInt KnownZero2(KnownZero), KnownOne2(KnownOne);
switch (I->getOpcode()) {
default: break;
case Instruction::And: {
// If either the LHS or the RHS are Zero, the result is zero.
ComputeMaskedBits(I->getOperand(1), Mask, KnownZero, KnownOne, TD, Depth+1);
APInt Mask2(Mask & ~KnownZero);
ComputeMaskedBits(I->getOperand(0), Mask2, KnownZero2, KnownOne2, TD,
Depth+1);
assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?");
assert((KnownZero2 & KnownOne2) == 0 && "Bits known to be one AND zero?");
// Output known-1 bits are only known if set in both the LHS & RHS.
KnownOne &= KnownOne2;
// Output known-0 are known to be clear if zero in either the LHS | RHS.
KnownZero |= KnownZero2;
return;
}
case Instruction::Or: {
ComputeMaskedBits(I->getOperand(1), Mask, KnownZero, KnownOne, TD, Depth+1);
APInt Mask2(Mask & ~KnownOne);
ComputeMaskedBits(I->getOperand(0), Mask2, KnownZero2, KnownOne2, TD,
Depth+1);
assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?");
assert((KnownZero2 & KnownOne2) == 0 && "Bits known to be one AND zero?");
// Output known-0 bits are only known if clear in both the LHS & RHS.
KnownZero &= KnownZero2;
// Output known-1 are known to be set if set in either the LHS | RHS.
KnownOne |= KnownOne2;
return;
}
case Instruction::Xor: {
ComputeMaskedBits(I->getOperand(1), Mask, KnownZero, KnownOne, TD, Depth+1);
ComputeMaskedBits(I->getOperand(0), Mask, KnownZero2, KnownOne2, TD,
Depth+1);
assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?");
assert((KnownZero2 & KnownOne2) == 0 && "Bits known to be one AND zero?");
// Output known-0 bits are known if clear or set in both the LHS & RHS.
APInt KnownZeroOut = (KnownZero & KnownZero2) | (KnownOne & KnownOne2);
// Output known-1 are known to be set if set in only one of the LHS, RHS.
KnownOne = (KnownZero & KnownOne2) | (KnownOne & KnownZero2);
KnownZero = KnownZeroOut;
return;
}
case Instruction::Mul: {
APInt Mask2 = APInt::getAllOnesValue(BitWidth);
ComputeMaskedBits(I->getOperand(1), Mask2, KnownZero, KnownOne, TD,Depth+1);
ComputeMaskedBits(I->getOperand(0), Mask2, KnownZero2, KnownOne2, TD,
Depth+1);
assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?");
assert((KnownZero2 & KnownOne2) == 0 && "Bits known to be one AND zero?");
// If low bits are zero in either operand, output low known-0 bits.
// Also compute a conserative estimate for high known-0 bits.
// More trickiness is possible, but this is sufficient for the
// interesting case of alignment computation.
KnownOne.clear();
unsigned TrailZ = KnownZero.countTrailingOnes() +
KnownZero2.countTrailingOnes();
unsigned LeadZ = std::max(KnownZero.countLeadingOnes() +
KnownZero2.countLeadingOnes(),
BitWidth) - BitWidth;
TrailZ = std::min(TrailZ, BitWidth);
LeadZ = std::min(LeadZ, BitWidth);
KnownZero = APInt::getLowBitsSet(BitWidth, TrailZ) |
APInt::getHighBitsSet(BitWidth, LeadZ);
KnownZero &= Mask;
return;
}
case Instruction::UDiv: {
// For the purposes of computing leading zeros we can conservatively
// treat a udiv as a logical right shift by the power of 2 known to
// be less than the denominator.
APInt AllOnes = APInt::getAllOnesValue(BitWidth);
ComputeMaskedBits(I->getOperand(0),
AllOnes, KnownZero2, KnownOne2, TD, Depth+1);
unsigned LeadZ = KnownZero2.countLeadingOnes();
KnownOne2.clear();
KnownZero2.clear();
ComputeMaskedBits(I->getOperand(1),
AllOnes, KnownZero2, KnownOne2, TD, Depth+1);
unsigned RHSUnknownLeadingOnes = KnownOne2.countLeadingZeros();
if (RHSUnknownLeadingOnes != BitWidth)
LeadZ = std::min(BitWidth,
LeadZ + BitWidth - RHSUnknownLeadingOnes - 1);
KnownZero = APInt::getHighBitsSet(BitWidth, LeadZ) & Mask;
return;
}
case Instruction::Select:
ComputeMaskedBits(I->getOperand(2), Mask, KnownZero, KnownOne, TD, Depth+1);
ComputeMaskedBits(I->getOperand(1), Mask, KnownZero2, KnownOne2, TD,
Depth+1);
assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?");
assert((KnownZero2 & KnownOne2) == 0 && "Bits known to be one AND zero?");
// Only known if known in both the LHS and RHS.
KnownOne &= KnownOne2;
KnownZero &= KnownZero2;
return;
case Instruction::FPTrunc:
case Instruction::FPExt:
case Instruction::FPToUI:
case Instruction::FPToSI:
case Instruction::SIToFP:
case Instruction::UIToFP:
return; // Can't work with floating point.
case Instruction::PtrToInt:
case Instruction::IntToPtr:
// We can't handle these if we don't know the pointer size.
if (!TD) return;
// FALL THROUGH and handle them the same as zext/trunc.
case Instruction::ZExt:
case Instruction::Trunc: {
const Type *SrcTy = I->getOperand(0)->getType();
unsigned SrcBitWidth;
// Note that we handle pointer operands here because of inttoptr/ptrtoint
// which fall through here.
if (isa<PointerType>(SrcTy))
SrcBitWidth = TD->getTypeSizeInBits(SrcTy);
else
SrcBitWidth = SrcTy->getScalarSizeInBits();
APInt MaskIn(Mask);
MaskIn.zextOrTrunc(SrcBitWidth);
KnownZero.zextOrTrunc(SrcBitWidth);
KnownOne.zextOrTrunc(SrcBitWidth);
ComputeMaskedBits(I->getOperand(0), MaskIn, KnownZero, KnownOne, TD,
Depth+1);
KnownZero.zextOrTrunc(BitWidth);
KnownOne.zextOrTrunc(BitWidth);
// Any top bits are known to be zero.
if (BitWidth > SrcBitWidth)
KnownZero |= APInt::getHighBitsSet(BitWidth, BitWidth - SrcBitWidth);
return;
}
case Instruction::BitCast: {
const Type *SrcTy = I->getOperand(0)->getType();
if ((SrcTy->isInteger() || isa<PointerType>(SrcTy)) &&
// TODO: For now, not handling conversions like:
// (bitcast i64 %x to <2 x i32>)
!isa<VectorType>(I->getType())) {
ComputeMaskedBits(I->getOperand(0), Mask, KnownZero, KnownOne, TD,
Depth+1);
return;
}
break;
}
case Instruction::SExt: {
// Compute the bits in the result that are not present in the input.
unsigned SrcBitWidth = I->getOperand(0)->getType()->getScalarSizeInBits();
APInt MaskIn(Mask);
MaskIn.trunc(SrcBitWidth);
KnownZero.trunc(SrcBitWidth);
KnownOne.trunc(SrcBitWidth);
ComputeMaskedBits(I->getOperand(0), MaskIn, KnownZero, KnownOne, TD,
Depth+1);
assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?");
KnownZero.zext(BitWidth);
KnownOne.zext(BitWidth);
// If the sign bit of the input is known set or clear, then we know the
// top bits of the result.
if (KnownZero[SrcBitWidth-1]) // Input sign bit known zero
KnownZero |= APInt::getHighBitsSet(BitWidth, BitWidth - SrcBitWidth);
else if (KnownOne[SrcBitWidth-1]) // Input sign bit known set
KnownOne |= APInt::getHighBitsSet(BitWidth, BitWidth - SrcBitWidth);
return;
}
case Instruction::Shl:
// (shl X, C1) & C2 == 0 iff (X & C2 >>u C1) == 0
if (ConstantInt *SA = dyn_cast<ConstantInt>(I->getOperand(1))) {
uint64_t ShiftAmt = SA->getLimitedValue(BitWidth);
APInt Mask2(Mask.lshr(ShiftAmt));
ComputeMaskedBits(I->getOperand(0), Mask2, KnownZero, KnownOne, TD,
Depth+1);
assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?");
KnownZero <<= ShiftAmt;
KnownOne <<= ShiftAmt;
KnownZero |= APInt::getLowBitsSet(BitWidth, ShiftAmt); // low bits known 0
return;
}
break;
case Instruction::LShr:
// (ushr X, C1) & C2 == 0 iff (-1 >> C1) & C2 == 0
if (ConstantInt *SA = dyn_cast<ConstantInt>(I->getOperand(1))) {
// Compute the new bits that are at the top now.
uint64_t ShiftAmt = SA->getLimitedValue(BitWidth);
// Unsigned shift right.
APInt Mask2(Mask.shl(ShiftAmt));
ComputeMaskedBits(I->getOperand(0), Mask2, KnownZero,KnownOne, TD,
Depth+1);
assert((KnownZero & KnownOne) == 0&&"Bits known to be one AND zero?");
KnownZero = APIntOps::lshr(KnownZero, ShiftAmt);
KnownOne = APIntOps::lshr(KnownOne, ShiftAmt);
// high bits known zero.
KnownZero |= APInt::getHighBitsSet(BitWidth, ShiftAmt);
return;
}
break;
case Instruction::AShr:
// (ashr X, C1) & C2 == 0 iff (-1 >> C1) & C2 == 0
if (ConstantInt *SA = dyn_cast<ConstantInt>(I->getOperand(1))) {
// Compute the new bits that are at the top now.
uint64_t ShiftAmt = SA->getLimitedValue(BitWidth);
// Signed shift right.
APInt Mask2(Mask.shl(ShiftAmt));
ComputeMaskedBits(I->getOperand(0), Mask2, KnownZero, KnownOne, TD,
Depth+1);
assert((KnownZero & KnownOne) == 0&&"Bits known to be one AND zero?");
KnownZero = APIntOps::lshr(KnownZero, ShiftAmt);
KnownOne = APIntOps::lshr(KnownOne, ShiftAmt);
APInt HighBits(APInt::getHighBitsSet(BitWidth, ShiftAmt));
if (KnownZero[BitWidth-ShiftAmt-1]) // New bits are known zero.
KnownZero |= HighBits;
else if (KnownOne[BitWidth-ShiftAmt-1]) // New bits are known one.
KnownOne |= HighBits;
return;
}
break;
case Instruction::Sub: {
if (ConstantInt *CLHS = dyn_cast<ConstantInt>(I->getOperand(0))) {
// We know that the top bits of C-X are clear if X contains less bits
// than C (i.e. no wrap-around can happen). For example, 20-X is
// positive if we can prove that X is >= 0 and < 16.
if (!CLHS->getValue().isNegative()) {
unsigned NLZ = (CLHS->getValue()+1).countLeadingZeros();
// NLZ can't be BitWidth with no sign bit
APInt MaskV = APInt::getHighBitsSet(BitWidth, NLZ+1);
ComputeMaskedBits(I->getOperand(1), MaskV, KnownZero2, KnownOne2,
TD, Depth+1);
// If all of the MaskV bits are known to be zero, then we know the
// output top bits are zero, because we now know that the output is
// from [0-C].
if ((KnownZero2 & MaskV) == MaskV) {
unsigned NLZ2 = CLHS->getValue().countLeadingZeros();
// Top bits known zero.
KnownZero = APInt::getHighBitsSet(BitWidth, NLZ2) & Mask;
}
}
}
}
// fall through
case Instruction::Add: {
// If one of the operands has trailing zeros, than the bits that the
// other operand has in those bit positions will be preserved in the
// result. For an add, this works with either operand. For a subtract,
// this only works if the known zeros are in the right operand.
APInt LHSKnownZero(BitWidth, 0), LHSKnownOne(BitWidth, 0);
APInt Mask2 = APInt::getLowBitsSet(BitWidth,
BitWidth - Mask.countLeadingZeros());
ComputeMaskedBits(I->getOperand(0), Mask2, LHSKnownZero, LHSKnownOne, TD,
Depth+1);
assert((LHSKnownZero & LHSKnownOne) == 0 &&
"Bits known to be one AND zero?");
unsigned LHSKnownZeroOut = LHSKnownZero.countTrailingOnes();
ComputeMaskedBits(I->getOperand(1), Mask2, KnownZero2, KnownOne2, TD,
Depth+1);
assert((KnownZero2 & KnownOne2) == 0 && "Bits known to be one AND zero?");
unsigned RHSKnownZeroOut = KnownZero2.countTrailingOnes();
// Determine which operand has more trailing zeros, and use that
// many bits from the other operand.
if (LHSKnownZeroOut > RHSKnownZeroOut) {
if (I->getOpcode() == Instruction::Add) {
APInt Mask = APInt::getLowBitsSet(BitWidth, LHSKnownZeroOut);
KnownZero |= KnownZero2 & Mask;
KnownOne |= KnownOne2 & Mask;
} else {
// If the known zeros are in the left operand for a subtract,
// fall back to the minimum known zeros in both operands.
KnownZero |= APInt::getLowBitsSet(BitWidth,
std::min(LHSKnownZeroOut,
RHSKnownZeroOut));
}
} else if (RHSKnownZeroOut >= LHSKnownZeroOut) {
APInt Mask = APInt::getLowBitsSet(BitWidth, RHSKnownZeroOut);
KnownZero |= LHSKnownZero & Mask;
KnownOne |= LHSKnownOne & Mask;
}
return;
}
case Instruction::SRem:
if (ConstantInt *Rem = dyn_cast<ConstantInt>(I->getOperand(1))) {
APInt RA = Rem->getValue();
if (RA.isPowerOf2() || (-RA).isPowerOf2()) {
APInt LowBits = RA.isStrictlyPositive() ? (RA - 1) : ~RA;
APInt Mask2 = LowBits | APInt::getSignBit(BitWidth);
ComputeMaskedBits(I->getOperand(0), Mask2, KnownZero2, KnownOne2, TD,
Depth+1);
// If the sign bit of the first operand is zero, the sign bit of
// the result is zero. If the first operand has no one bits below
// the second operand's single 1 bit, its sign will be zero.
if (KnownZero2[BitWidth-1] || ((KnownZero2 & LowBits) == LowBits))
KnownZero2 |= ~LowBits;
KnownZero |= KnownZero2 & Mask;
assert((KnownZero & KnownOne) == 0&&"Bits known to be one AND zero?");
}
}
break;
case Instruction::URem: {
if (ConstantInt *Rem = dyn_cast<ConstantInt>(I->getOperand(1))) {
APInt RA = Rem->getValue();
if (RA.isPowerOf2()) {
APInt LowBits = (RA - 1);
APInt Mask2 = LowBits & Mask;
KnownZero |= ~LowBits & Mask;
ComputeMaskedBits(I->getOperand(0), Mask2, KnownZero, KnownOne, TD,
Depth+1);
assert((KnownZero & KnownOne) == 0&&"Bits known to be one AND zero?");
break;
}
}
// Since the result is less than or equal to either operand, any leading
// zero bits in either operand must also exist in the result.
APInt AllOnes = APInt::getAllOnesValue(BitWidth);
ComputeMaskedBits(I->getOperand(0), AllOnes, KnownZero, KnownOne,
TD, Depth+1);
ComputeMaskedBits(I->getOperand(1), AllOnes, KnownZero2, KnownOne2,
TD, Depth+1);
unsigned Leaders = std::max(KnownZero.countLeadingOnes(),
KnownZero2.countLeadingOnes());
KnownOne.clear();
KnownZero = APInt::getHighBitsSet(BitWidth, Leaders) & Mask;
break;
}
case Instruction::Alloca: {
AllocaInst *AI = cast<AllocaInst>(V);
unsigned Align = AI->getAlignment();
if (Align == 0 && TD)
Align = TD->getABITypeAlignment(AI->getType()->getElementType());
if (Align > 0)
KnownZero = Mask & APInt::getLowBitsSet(BitWidth,
CountTrailingZeros_32(Align));
break;
}
case Instruction::GetElementPtr: {
// Analyze all of the subscripts of this getelementptr instruction
// to determine if we can prove known low zero bits.
APInt LocalMask = APInt::getAllOnesValue(BitWidth);
APInt LocalKnownZero(BitWidth, 0), LocalKnownOne(BitWidth, 0);
ComputeMaskedBits(I->getOperand(0), LocalMask,
LocalKnownZero, LocalKnownOne, TD, Depth+1);
unsigned TrailZ = LocalKnownZero.countTrailingOnes();
gep_type_iterator GTI = gep_type_begin(I);
for (unsigned i = 1, e = I->getNumOperands(); i != e; ++i, ++GTI) {
Value *Index = I->getOperand(i);
if (const StructType *STy = dyn_cast<StructType>(*GTI)) {
// Handle struct member offset arithmetic.
if (!TD) return;
const StructLayout *SL = TD->getStructLayout(STy);
unsigned Idx = cast<ConstantInt>(Index)->getZExtValue();
uint64_t Offset = SL->getElementOffset(Idx);
TrailZ = std::min(TrailZ,
CountTrailingZeros_64(Offset));
} else {
// Handle array index arithmetic.
const Type *IndexedTy = GTI.getIndexedType();
if (!IndexedTy->isSized()) return;
unsigned GEPOpiBits = Index->getType()->getScalarSizeInBits();
uint64_t TypeSize = TD ? TD->getTypeAllocSize(IndexedTy) : 1;
LocalMask = APInt::getAllOnesValue(GEPOpiBits);
LocalKnownZero = LocalKnownOne = APInt(GEPOpiBits, 0);
ComputeMaskedBits(Index, LocalMask,
LocalKnownZero, LocalKnownOne, TD, Depth+1);
TrailZ = std::min(TrailZ,
unsigned(CountTrailingZeros_64(TypeSize) +
LocalKnownZero.countTrailingOnes()));
}
}
KnownZero = APInt::getLowBitsSet(BitWidth, TrailZ) & Mask;
break;
}
case Instruction::PHI: {
PHINode *P = cast<PHINode>(I);
// Handle the case of a simple two-predecessor recurrence PHI.
// There's a lot more that could theoretically be done here, but
// this is sufficient to catch some interesting cases.
if (P->getNumIncomingValues() == 2) {
for (unsigned i = 0; i != 2; ++i) {
Value *L = P->getIncomingValue(i);
Value *R = P->getIncomingValue(!i);
Operator *LU = dyn_cast<Operator>(L);
if (!LU)
continue;
unsigned Opcode = LU->getOpcode();
// Check for operations that have the property that if
// both their operands have low zero bits, the result
// will have low zero bits.
if (Opcode == Instruction::Add ||
Opcode == Instruction::Sub ||
Opcode == Instruction::And ||
Opcode == Instruction::Or ||
Opcode == Instruction::Mul) {
Value *LL = LU->getOperand(0);
Value *LR = LU->getOperand(1);
// Find a recurrence.
if (LL == I)
L = LR;
else if (LR == I)
L = LL;
else
break;
// Ok, we have a PHI of the form L op= R. Check for low
// zero bits.
APInt Mask2 = APInt::getAllOnesValue(BitWidth);
ComputeMaskedBits(R, Mask2, KnownZero2, KnownOne2, TD, Depth+1);
Mask2 = APInt::getLowBitsSet(BitWidth,
KnownZero2.countTrailingOnes());
// We need to take the minimum number of known bits
APInt KnownZero3(KnownZero), KnownOne3(KnownOne);
ComputeMaskedBits(L, Mask2, KnownZero3, KnownOne3, TD, Depth+1);
KnownZero = Mask &
APInt::getLowBitsSet(BitWidth,
std::min(KnownZero2.countTrailingOnes(),
KnownZero3.countTrailingOnes()));
break;
}
}
}
// Otherwise take the unions of the known bit sets of the operands,
// taking conservative care to avoid excessive recursion.
if (Depth < MaxDepth - 1 && !KnownZero && !KnownOne) {
KnownZero = APInt::getAllOnesValue(BitWidth);
KnownOne = APInt::getAllOnesValue(BitWidth);
for (unsigned i = 0, e = P->getNumIncomingValues(); i != e; ++i) {
// Skip direct self references.
if (P->getIncomingValue(i) == P) continue;
KnownZero2 = APInt(BitWidth, 0);
KnownOne2 = APInt(BitWidth, 0);
// Recurse, but cap the recursion to one level, because we don't
// want to waste time spinning around in loops.
ComputeMaskedBits(P->getIncomingValue(i), KnownZero | KnownOne,
KnownZero2, KnownOne2, TD, MaxDepth-1);
KnownZero &= KnownZero2;
KnownOne &= KnownOne2;
// If all bits have been ruled out, there's no need to check
// more operands.
if (!KnownZero && !KnownOne)
break;
}
}
break;
}
case Instruction::Call:
if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(I)) {
switch (II->getIntrinsicID()) {
default: break;
case Intrinsic::ctpop:
case Intrinsic::ctlz:
case Intrinsic::cttz: {
unsigned LowBits = Log2_32(BitWidth)+1;
KnownZero = APInt::getHighBitsSet(BitWidth, BitWidth - LowBits);
break;
}
}
}
break;
}
}
/// MaskedValueIsZero - Return true if 'V & Mask' is known to be zero. We use
/// this predicate to simplify operations downstream. Mask is known to be zero
/// for bits that V cannot have.
///
/// This function is defined on values with integer type, values with pointer
/// type (but only if TD is non-null), and vectors of integers. In the case
/// where V is a vector, the mask, known zero, and known one values are the
/// same width as the vector element, and the bit is set only if it is true
/// for all of the elements in the vector.
bool llvm::MaskedValueIsZero(Value *V, const APInt &Mask,
const TargetData *TD, unsigned Depth) {
APInt KnownZero(Mask.getBitWidth(), 0), KnownOne(Mask.getBitWidth(), 0);
ComputeMaskedBits(V, Mask, KnownZero, KnownOne, TD, Depth);
assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?");
return (KnownZero & Mask) == Mask;
}
/// ComputeNumSignBits - Return the number of times the sign bit of the
/// register is replicated into the other bits. We know that at least 1 bit
/// is always equal to the sign bit (itself), but other cases can give us
/// information. For example, immediately after an "ashr X, 2", we know that
/// the top 3 bits are all equal to each other, so we return 3.
///
/// 'Op' must have a scalar integer type.
///
unsigned llvm::ComputeNumSignBits(Value *V, const TargetData *TD,
unsigned Depth) {
assert((TD || V->getType()->isIntOrIntVector()) &&
"ComputeNumSignBits requires a TargetData object to operate "
"on non-integer values!");
const Type *Ty = V->getType();
unsigned TyBits = TD ? TD->getTypeSizeInBits(V->getType()->getScalarType()) :
Ty->getScalarSizeInBits();
unsigned Tmp, Tmp2;
unsigned FirstAnswer = 1;
// Note that ConstantInt is handled by the general ComputeMaskedBits case
// below.
if (Depth == 6)
return 1; // Limit search depth.
Operator *U = dyn_cast<Operator>(V);
switch (Operator::getOpcode(V)) {
default: break;
case Instruction::SExt:
Tmp = TyBits-cast<IntegerType>(U->getOperand(0)->getType())->getBitWidth();
return ComputeNumSignBits(U->getOperand(0), TD, Depth+1) + Tmp;
case Instruction::AShr:
Tmp = ComputeNumSignBits(U->getOperand(0), TD, Depth+1);
// ashr X, C -> adds C sign bits.
if (ConstantInt *C = dyn_cast<ConstantInt>(U->getOperand(1))) {
Tmp += C->getZExtValue();
if (Tmp > TyBits) Tmp = TyBits;
}
return Tmp;
case Instruction::Shl:
if (ConstantInt *C = dyn_cast<ConstantInt>(U->getOperand(1))) {
// shl destroys sign bits.
Tmp = ComputeNumSignBits(U->getOperand(0), TD, Depth+1);
if (C->getZExtValue() >= TyBits || // Bad shift.
C->getZExtValue() >= Tmp) break; // Shifted all sign bits out.
return Tmp - C->getZExtValue();
}
break;
case Instruction::And:
case Instruction::Or:
case Instruction::Xor: // NOT is handled here.
// Logical binary ops preserve the number of sign bits at the worst.
Tmp = ComputeNumSignBits(U->getOperand(0), TD, Depth+1);
if (Tmp != 1) {
Tmp2 = ComputeNumSignBits(U->getOperand(1), TD, Depth+1);
FirstAnswer = std::min(Tmp, Tmp2);
// We computed what we know about the sign bits as our first
// answer. Now proceed to the generic code that uses
// ComputeMaskedBits, and pick whichever answer is better.
}
break;
case Instruction::Select:
Tmp = ComputeNumSignBits(U->getOperand(1), TD, Depth+1);
if (Tmp == 1) return 1; // Early out.
Tmp2 = ComputeNumSignBits(U->getOperand(2), TD, Depth+1);
return std::min(Tmp, Tmp2);
case Instruction::Add:
// Add can have at most one carry bit. Thus we know that the output
// is, at worst, one more bit than the inputs.
Tmp = ComputeNumSignBits(U->getOperand(0), TD, Depth+1);
if (Tmp == 1) return 1; // Early out.
// Special case decrementing a value (ADD X, -1):
if (ConstantInt *CRHS = dyn_cast<ConstantInt>(U->getOperand(1)))
if (CRHS->isAllOnesValue()) {
APInt KnownZero(TyBits, 0), KnownOne(TyBits, 0);
APInt Mask = APInt::getAllOnesValue(TyBits);
ComputeMaskedBits(U->getOperand(0), Mask, KnownZero, KnownOne, TD,
Depth+1);
// If the input is known to be 0 or 1, the output is 0/-1, which is all
// sign bits set.
if ((KnownZero | APInt(TyBits, 1)) == Mask)
return TyBits;
// If we are subtracting one from a positive number, there is no carry
// out of the result.
if (KnownZero.isNegative())
return Tmp;
}
Tmp2 = ComputeNumSignBits(U->getOperand(1), TD, Depth+1);
if (Tmp2 == 1) return 1;
return std::min(Tmp, Tmp2)-1;
break;
case Instruction::Sub:
Tmp2 = ComputeNumSignBits(U->getOperand(1), TD, Depth+1);
if (Tmp2 == 1) return 1;
// Handle NEG.
if (ConstantInt *CLHS = dyn_cast<ConstantInt>(U->getOperand(0)))
if (CLHS->isNullValue()) {
APInt KnownZero(TyBits, 0), KnownOne(TyBits, 0);
APInt Mask = APInt::getAllOnesValue(TyBits);
ComputeMaskedBits(U->getOperand(1), Mask, KnownZero, KnownOne,
TD, Depth+1);
// If the input is known to be 0 or 1, the output is 0/-1, which is all
// sign bits set.
if ((KnownZero | APInt(TyBits, 1)) == Mask)
return TyBits;
// If the input is known to be positive (the sign bit is known clear),
// the output of the NEG has the same number of sign bits as the input.
if (KnownZero.isNegative())
return Tmp2;
// Otherwise, we treat this like a SUB.
}
// Sub can have at most one carry bit. Thus we know that the output
// is, at worst, one more bit than the inputs.
Tmp = ComputeNumSignBits(U->getOperand(0), TD, Depth+1);
if (Tmp == 1) return 1; // Early out.
return std::min(Tmp, Tmp2)-1;
break;
case Instruction::Trunc:
// FIXME: it's tricky to do anything useful for this, but it is an important
// case for targets like X86.
break;
}
// Finally, if we can prove that the top bits of the result are 0's or 1's,
// use this information.
APInt KnownZero(TyBits, 0), KnownOne(TyBits, 0);
APInt Mask = APInt::getAllOnesValue(TyBits);
ComputeMaskedBits(V, Mask, KnownZero, KnownOne, TD, Depth);
if (KnownZero.isNegative()) { // sign bit is 0
Mask = KnownZero;
} else if (KnownOne.isNegative()) { // sign bit is 1;
Mask = KnownOne;
} else {
// Nothing known.
return FirstAnswer;
}
// Okay, we know that the sign bit in Mask is set. Use CLZ to determine
// the number of identical bits in the top of the input value.
Mask = ~Mask;
Mask <<= Mask.getBitWidth()-TyBits;
// Return # leading zeros. We use 'min' here in case Val was zero before
// shifting. We don't want to return '64' as for an i32 "0".
return std::max(FirstAnswer, std::min(TyBits, Mask.countLeadingZeros()));
}
/// CannotBeNegativeZero - Return true if we can prove that the specified FP
/// value is never equal to -0.0.
///
/// NOTE: this function will need to be revisited when we support non-default
/// rounding modes!
///
bool llvm::CannotBeNegativeZero(const Value *V, unsigned Depth) {
if (const ConstantFP *CFP = dyn_cast<ConstantFP>(V))
return !CFP->getValueAPF().isNegZero();
if (Depth == 6)
return 1; // Limit search depth.
const Operator *I = dyn_cast<Operator>(V);
if (I == 0) return false;
// (add x, 0.0) is guaranteed to return +0.0, not -0.0.
if (I->getOpcode() == Instruction::FAdd &&
isa<ConstantFP>(I->getOperand(1)) &&
cast<ConstantFP>(I->getOperand(1))->isNullValue())
return true;
// sitofp and uitofp turn into +0.0 for zero.
if (isa<SIToFPInst>(I) || isa<UIToFPInst>(I))
return true;
if (const IntrinsicInst *II = dyn_cast<IntrinsicInst>(I))
// sqrt(-0.0) = -0.0, no other negative results are possible.
if (II->getIntrinsicID() == Intrinsic::sqrt)
return CannotBeNegativeZero(II->getOperand(1), Depth+1);
if (const CallInst *CI = dyn_cast<CallInst>(I))
if (const Function *F = CI->getCalledFunction()) {
if (F->isDeclaration()) {
// abs(x) != -0.0
if (F->getName() == "abs") return true;
// fabs[lf](x) != -0.0
if (F->getName() == "fabs") return true;
if (F->getName() == "fabsf") return true;
if (F->getName() == "fabsl") return true;
if (F->getName() == "sqrt" || F->getName() == "sqrtf" ||
F->getName() == "sqrtl")
return CannotBeNegativeZero(CI->getOperand(1), Depth+1);
}
}
return false;
}
// This is the recursive version of BuildSubAggregate. It takes a few different
// arguments. Idxs is the index within the nested struct From that we are
// looking at now (which is of type IndexedType). IdxSkip is the number of
// indices from Idxs that should be left out when inserting into the resulting
// struct. To is the result struct built so far, new insertvalue instructions
// build on that.
static Value *BuildSubAggregate(Value *From, Value* To, const Type *IndexedType,
SmallVector<unsigned, 10> &Idxs,
unsigned IdxSkip,
LLVMContext &Context,
Instruction *InsertBefore) {
const llvm::StructType *STy = llvm::dyn_cast<llvm::StructType>(IndexedType);
if (STy) {
// Save the original To argument so we can modify it
Value *OrigTo = To;
// General case, the type indexed by Idxs is a struct
for (unsigned i = 0, e = STy->getNumElements(); i != e; ++i) {
// Process each struct element recursively
Idxs.push_back(i);
Value *PrevTo = To;
To = BuildSubAggregate(From, To, STy->getElementType(i), Idxs, IdxSkip,
Context, InsertBefore);
Idxs.pop_back();
if (!To) {
// Couldn't find any inserted value for this index? Cleanup
while (PrevTo != OrigTo) {
InsertValueInst* Del = cast<InsertValueInst>(PrevTo);
PrevTo = Del->getAggregateOperand();
Del->eraseFromParent();
}
// Stop processing elements
break;
}
}
// If we succesfully found a value for each of our subaggregates
if (To)
return To;
}
// Base case, the type indexed by SourceIdxs is not a struct, or not all of
// the struct's elements had a value that was inserted directly. In the latter
// case, perhaps we can't determine each of the subelements individually, but
// we might be able to find the complete struct somewhere.
// Find the value that is at that particular spot
Value *V = FindInsertedValue(From, Idxs.begin(), Idxs.end(), Context);
if (!V)
return NULL;
// Insert the value in the new (sub) aggregrate
return llvm::InsertValueInst::Create(To, V, Idxs.begin() + IdxSkip,
Idxs.end(), "tmp", InsertBefore);
}
// This helper takes a nested struct and extracts a part of it (which is again a
// struct) into a new value. For example, given the struct:
// { a, { b, { c, d }, e } }
// and the indices "1, 1" this returns
// { c, d }.
//
// It does this by inserting an insertvalue for each element in the resulting
// struct, as opposed to just inserting a single struct. This will only work if
// each of the elements of the substruct are known (ie, inserted into From by an
// insertvalue instruction somewhere).
//
// All inserted insertvalue instructions are inserted before InsertBefore
static Value *BuildSubAggregate(Value *From, const unsigned *idx_begin,
const unsigned *idx_end, LLVMContext &Context,
Instruction *InsertBefore) {
assert(InsertBefore && "Must have someplace to insert!");
const Type *IndexedType = ExtractValueInst::getIndexedType(From->getType(),
idx_begin,
idx_end);
Value *To = UndefValue::get(IndexedType);
SmallVector<unsigned, 10> Idxs(idx_begin, idx_end);
unsigned IdxSkip = Idxs.size();
return BuildSubAggregate(From, To, IndexedType, Idxs, IdxSkip,
Context, InsertBefore);
}
/// FindInsertedValue - Given an aggregrate and an sequence of indices, see if
/// the scalar value indexed is already around as a register, for example if it
/// were inserted directly into the aggregrate.
///
/// If InsertBefore is not null, this function will duplicate (modified)
/// insertvalues when a part of a nested struct is extracted.
Value *llvm::FindInsertedValue(Value *V, const unsigned *idx_begin,
const unsigned *idx_end, LLVMContext &Context,
Instruction *InsertBefore) {
// Nothing to index? Just return V then (this is useful at the end of our
// recursion)
if (idx_begin == idx_end)
return V;
// We have indices, so V should have an indexable type
assert((isa<StructType>(V->getType()) || isa<ArrayType>(V->getType()))
&& "Not looking at a struct or array?");
assert(ExtractValueInst::getIndexedType(V->getType(), idx_begin, idx_end)
&& "Invalid indices for type?");
const CompositeType *PTy = cast<CompositeType>(V->getType());
if (isa<UndefValue>(V))
return UndefValue::get(ExtractValueInst::getIndexedType(PTy,
idx_begin,
idx_end));
else if (isa<ConstantAggregateZero>(V))
return Constant::getNullValue(ExtractValueInst::getIndexedType(PTy,
idx_begin,
idx_end));
else if (Constant *C = dyn_cast<Constant>(V)) {
if (isa<ConstantArray>(C) || isa<ConstantStruct>(C))
// Recursively process this constant
return FindInsertedValue(C->getOperand(*idx_begin), idx_begin + 1,
idx_end, Context, InsertBefore);
} else if (InsertValueInst *I = dyn_cast<InsertValueInst>(V)) {
// Loop the indices for the insertvalue instruction in parallel with the
// requested indices
const unsigned *req_idx = idx_begin;
for (const unsigned *i = I->idx_begin(), *e = I->idx_end();
i != e; ++i, ++req_idx) {
if (req_idx == idx_end) {
if (InsertBefore)
// The requested index identifies a part of a nested aggregate. Handle
// this specially. For example,
// %A = insertvalue { i32, {i32, i32 } } undef, i32 10, 1, 0
// %B = insertvalue { i32, {i32, i32 } } %A, i32 11, 1, 1
// %C = extractvalue {i32, { i32, i32 } } %B, 1
// This can be changed into
// %A = insertvalue {i32, i32 } undef, i32 10, 0
// %C = insertvalue {i32, i32 } %A, i32 11, 1
// which allows the unused 0,0 element from the nested struct to be
// removed.
return BuildSubAggregate(V, idx_begin, req_idx,
Context, InsertBefore);
else
// We can't handle this without inserting insertvalues
return 0;
}
// This insert value inserts something else than what we are looking for.
// See if the (aggregrate) value inserted into has the value we are
// looking for, then.
if (*req_idx != *i)
return FindInsertedValue(I->getAggregateOperand(), idx_begin, idx_end,
Context, InsertBefore);
}
// If we end up here, the indices of the insertvalue match with those
// requested (though possibly only partially). Now we recursively look at
// the inserted value, passing any remaining indices.
return FindInsertedValue(I->getInsertedValueOperand(), req_idx, idx_end,
Context, InsertBefore);
} else if (ExtractValueInst *I = dyn_cast<ExtractValueInst>(V)) {
// If we're extracting a value from an aggregrate that was extracted from
// something else, we can extract from that something else directly instead.
// However, we will need to chain I's indices with the requested indices.
// Calculate the number of indices required
unsigned size = I->getNumIndices() + (idx_end - idx_begin);
// Allocate some space to put the new indices in
SmallVector<unsigned, 5> Idxs;
Idxs.reserve(size);
// Add indices from the extract value instruction
for (const unsigned *i = I->idx_begin(), *e = I->idx_end();
i != e; ++i)
Idxs.push_back(*i);
// Add requested indices
for (const unsigned *i = idx_begin, *e = idx_end; i != e; ++i)
Idxs.push_back(*i);
assert(Idxs.size() == size
&& "Number of indices added not correct?");
return FindInsertedValue(I->getAggregateOperand(), Idxs.begin(), Idxs.end(),
Context, InsertBefore);
}
// Otherwise, we don't know (such as, extracting from a function return value
// or load instruction)
return 0;
}
/// GetConstantStringInfo - This function computes the length of a
/// null-terminated C string pointed to by V. If successful, it returns true
/// and returns the string in Str. If unsuccessful, it returns false.
bool llvm::GetConstantStringInfo(Value *V, std::string &Str, uint64_t Offset,
bool StopAtNul) {
// If V is NULL then return false;
if (V == NULL) return false;
// Look through bitcast instructions.
if (BitCastInst *BCI = dyn_cast<BitCastInst>(V))
return GetConstantStringInfo(BCI->getOperand(0), Str, Offset, StopAtNul);
// If the value is not a GEP instruction nor a constant expression with a
// GEP instruction, then return false because ConstantArray can't occur
// any other way
User *GEP = 0;
if (GetElementPtrInst *GEPI = dyn_cast<GetElementPtrInst>(V)) {
GEP = GEPI;
} else if (ConstantExpr *CE = dyn_cast<ConstantExpr>(V)) {
if (CE->getOpcode() == Instruction::BitCast)
return GetConstantStringInfo(CE->getOperand(0), Str, Offset, StopAtNul);
if (CE->getOpcode() != Instruction::GetElementPtr)
return false;
GEP = CE;
}
if (GEP) {
// Make sure the GEP has exactly three arguments.
if (GEP->getNumOperands() != 3)
return false;
// Make sure the index-ee is a pointer to array of i8.
const PointerType *PT = cast<PointerType>(GEP->getOperand(0)->getType());
const ArrayType *AT = dyn_cast<ArrayType>(PT->getElementType());
if (AT == 0 || AT->getElementType() != Type::getInt8Ty(V->getContext()))
return false;
// Check to make sure that the first operand of the GEP is an integer and
// has value 0 so that we are sure we're indexing into the initializer.
ConstantInt *FirstIdx = dyn_cast<ConstantInt>(GEP->getOperand(1));
if (FirstIdx == 0 || !FirstIdx->isZero())
return false;
// If the second index isn't a ConstantInt, then this is a variable index
// into the array. If this occurs, we can't say anything meaningful about
// the string.
uint64_t StartIdx = 0;
if (ConstantInt *CI = dyn_cast<ConstantInt>(GEP->getOperand(2)))
StartIdx = CI->getZExtValue();
else
return false;
return GetConstantStringInfo(GEP->getOperand(0), Str, StartIdx+Offset,
StopAtNul);
}
if (MDString *MDStr = dyn_cast<MDString>(V)) {
Str = MDStr->getString();
return true;
}
// The GEP instruction, constant or instruction, must reference a global
// variable that is a constant and is initialized. The referenced constant
// initializer is the array that we'll use for optimization.
GlobalVariable* GV = dyn_cast<GlobalVariable>(V);
if (!GV || !GV->isConstant() || !GV->hasDefinitiveInitializer())
return false;
Constant *GlobalInit = GV->getInitializer();
// Handle the ConstantAggregateZero case
if (isa<ConstantAggregateZero>(GlobalInit)) {
// This is a degenerate case. The initializer is constant zero so the
// length of the string must be zero.
Str.clear();
return true;
}
// Must be a Constant Array
ConstantArray *Array = dyn_cast<ConstantArray>(GlobalInit);
if (Array == 0 ||
Array->getType()->getElementType() != Type::getInt8Ty(V->getContext()))
return false;
// Get the number of elements in the array
uint64_t NumElts = Array->getType()->getNumElements();
if (Offset > NumElts)
return false;
// Traverse the constant array from 'Offset' which is the place the GEP refers
// to in the array.
Str.reserve(NumElts-Offset);
for (unsigned i = Offset; i != NumElts; ++i) {
Constant *Elt = Array->getOperand(i);
ConstantInt *CI = dyn_cast<ConstantInt>(Elt);
if (!CI) // This array isn't suitable, non-int initializer.
return false;
if (StopAtNul && CI->isZero())
return true; // we found end of string, success!
Str += (char)CI->getZExtValue();
}
// The array isn't null terminated, but maybe this is a memcpy, not a strcpy.
return true;
}
|