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
|
//===- InstCombineCalls.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 visitCall and visitInvoke functions.
//
//===----------------------------------------------------------------------===//
#include "InstCombine.h"
#include "llvm/IntrinsicInst.h"
#include "llvm/Support/CallSite.h"
#include "llvm/Target/TargetData.h"
#include "llvm/Analysis/MemoryBuiltins.h"
#include "llvm/Transforms/Utils/BuildLibCalls.h"
using namespace llvm;
/// getPromotedType - Return the specified type promoted as it would be to pass
/// though a va_arg area.
static const Type *getPromotedType(const Type *Ty) {
if (const IntegerType* ITy = dyn_cast<IntegerType>(Ty)) {
if (ITy->getBitWidth() < 32)
return Type::getInt32Ty(Ty->getContext());
}
return Ty;
}
/// EnforceKnownAlignment - If the specified pointer points to an object that
/// we control, modify the object's alignment to PrefAlign. This isn't
/// often possible though. If alignment is important, a more reliable approach
/// is to simply align all global variables and allocation instructions to
/// their preferred alignment from the beginning.
///
static unsigned EnforceKnownAlignment(Value *V,
unsigned Align, unsigned PrefAlign) {
User *U = dyn_cast<User>(V);
if (!U) return Align;
switch (Operator::getOpcode(U)) {
default: break;
case Instruction::BitCast:
return EnforceKnownAlignment(U->getOperand(0), Align, PrefAlign);
case Instruction::GetElementPtr: {
// If all indexes are zero, it is just the alignment of the base pointer.
bool AllZeroOperands = true;
for (User::op_iterator i = U->op_begin() + 1, e = U->op_end(); i != e; ++i)
if (!isa<Constant>(*i) ||
!cast<Constant>(*i)->isNullValue()) {
AllZeroOperands = false;
break;
}
if (AllZeroOperands) {
// Treat this like a bitcast.
return EnforceKnownAlignment(U->getOperand(0), Align, PrefAlign);
}
return Align;
}
case Instruction::Alloca: {
AllocaInst *AI = cast<AllocaInst>(V);
// If there is a requested alignment and if this is an alloca, round up.
if (AI->getAlignment() >= PrefAlign)
return AI->getAlignment();
AI->setAlignment(PrefAlign);
return PrefAlign;
}
}
if (GlobalValue *GV = dyn_cast<GlobalValue>(V)) {
// If there is a large requested alignment and we can, bump up the alignment
// of the global.
if (GV->isDeclaration()) return Align;
if (GV->getAlignment() >= PrefAlign)
return GV->getAlignment();
// We can only increase the alignment of the global if it has no alignment
// specified or if it is not assigned a section. If it is assigned a
// section, the global could be densely packed with other objects in the
// section, increasing the alignment could cause padding issues.
if (!GV->hasSection() || GV->getAlignment() == 0)
GV->setAlignment(PrefAlign);
return GV->getAlignment();
}
return Align;
}
/// GetOrEnforceKnownAlignment - If the specified pointer has an alignment that
/// we can determine, return it, otherwise return 0. If PrefAlign is specified,
/// and it is more than the alignment of the ultimate object, see if we can
/// increase the alignment of the ultimate object, making this check succeed.
unsigned InstCombiner::GetOrEnforceKnownAlignment(Value *V,
unsigned PrefAlign) {
unsigned BitWidth = TD ? TD->getTypeSizeInBits(V->getType()) :
sizeof(PrefAlign) * CHAR_BIT;
APInt Mask = APInt::getAllOnesValue(BitWidth);
APInt KnownZero(BitWidth, 0), KnownOne(BitWidth, 0);
ComputeMaskedBits(V, Mask, KnownZero, KnownOne);
unsigned TrailZ = KnownZero.countTrailingOnes();
unsigned Align = 1u << std::min(BitWidth - 1, TrailZ);
if (PrefAlign > Align)
Align = EnforceKnownAlignment(V, Align, PrefAlign);
// We don't need to make any adjustment.
return Align;
}
Instruction *InstCombiner::SimplifyMemTransfer(MemIntrinsic *MI) {
unsigned DstAlign = GetOrEnforceKnownAlignment(MI->getOperand(1));
unsigned SrcAlign = GetOrEnforceKnownAlignment(MI->getOperand(2));
unsigned MinAlign = std::min(DstAlign, SrcAlign);
unsigned CopyAlign = MI->getAlignment();
if (CopyAlign < MinAlign) {
MI->setAlignment(ConstantInt::get(MI->getAlignmentType(),
MinAlign, false));
return MI;
}
// If MemCpyInst length is 1/2/4/8 bytes then replace memcpy with
// load/store.
ConstantInt *MemOpLength = dyn_cast<ConstantInt>(MI->getOperand(3));
if (MemOpLength == 0) return 0;
// Source and destination pointer types are always "i8*" for intrinsic. See
// if the size is something we can handle with a single primitive load/store.
// A single load+store correctly handles overlapping memory in the memmove
// case.
unsigned Size = MemOpLength->getZExtValue();
if (Size == 0) return MI; // Delete this mem transfer.
if (Size > 8 || (Size&(Size-1)))
return 0; // If not 1/2/4/8 bytes, exit.
// Use an integer load+store unless we can find something better.
unsigned SrcAddrSp =
cast<PointerType>(MI->getOperand(2)->getType())->getAddressSpace();
unsigned DstAddrSp =
cast<PointerType>(MI->getOperand(1)->getType())->getAddressSpace();
const IntegerType* IntType = IntegerType::get(MI->getContext(), Size<<3);
Type *NewSrcPtrTy = PointerType::get(IntType, SrcAddrSp);
Type *NewDstPtrTy = PointerType::get(IntType, DstAddrSp);
// Memcpy forces the use of i8* for the source and destination. That means
// that if you're using memcpy to move one double around, you'll get a cast
// from double* to i8*. We'd much rather use a double load+store rather than
// an i64 load+store, here because this improves the odds that the source or
// dest address will be promotable. See if we can find a better type than the
// integer datatype.
Value *StrippedDest = MI->getOperand(1)->stripPointerCasts();
if (StrippedDest != MI->getOperand(1)) {
const Type *SrcETy = cast<PointerType>(StrippedDest->getType())
->getElementType();
if (TD && SrcETy->isSized() && TD->getTypeStoreSize(SrcETy) == Size) {
// The SrcETy might be something like {{{double}}} or [1 x double]. Rip
// down through these levels if so.
while (!SrcETy->isSingleValueType()) {
if (const StructType *STy = dyn_cast<StructType>(SrcETy)) {
if (STy->getNumElements() == 1)
SrcETy = STy->getElementType(0);
else
break;
} else if (const ArrayType *ATy = dyn_cast<ArrayType>(SrcETy)) {
if (ATy->getNumElements() == 1)
SrcETy = ATy->getElementType();
else
break;
} else
break;
}
if (SrcETy->isSingleValueType()) {
NewSrcPtrTy = PointerType::get(SrcETy, SrcAddrSp);
NewDstPtrTy = PointerType::get(SrcETy, DstAddrSp);
}
}
}
// If the memcpy/memmove provides better alignment info than we can
// infer, use it.
SrcAlign = std::max(SrcAlign, CopyAlign);
DstAlign = std::max(DstAlign, CopyAlign);
Value *Src = Builder->CreateBitCast(MI->getOperand(2), NewSrcPtrTy);
Value *Dest = Builder->CreateBitCast(MI->getOperand(1), NewDstPtrTy);
Instruction *L = new LoadInst(Src, "tmp", MI->isVolatile(), SrcAlign);
InsertNewInstBefore(L, *MI);
InsertNewInstBefore(new StoreInst(L, Dest, MI->isVolatile(), DstAlign),
*MI);
// Set the size of the copy to 0, it will be deleted on the next iteration.
MI->setOperand(3, Constant::getNullValue(MemOpLength->getType()));
return MI;
}
Instruction *InstCombiner::SimplifyMemSet(MemSetInst *MI) {
unsigned Alignment = GetOrEnforceKnownAlignment(MI->getDest());
if (MI->getAlignment() < Alignment) {
MI->setAlignment(ConstantInt::get(MI->getAlignmentType(),
Alignment, false));
return MI;
}
// Extract the length and alignment and fill if they are constant.
ConstantInt *LenC = dyn_cast<ConstantInt>(MI->getLength());
ConstantInt *FillC = dyn_cast<ConstantInt>(MI->getValue());
if (!LenC || !FillC || !FillC->getType()->isIntegerTy(8))
return 0;
uint64_t Len = LenC->getZExtValue();
Alignment = MI->getAlignment();
// If the length is zero, this is a no-op
if (Len == 0) return MI; // memset(d,c,0,a) -> noop
// memset(s,c,n) -> store s, c (for n=1,2,4,8)
if (Len <= 8 && isPowerOf2_32((uint32_t)Len)) {
const Type *ITy = IntegerType::get(MI->getContext(), Len*8); // n=1 -> i8.
Value *Dest = MI->getDest();
Dest = Builder->CreateBitCast(Dest, PointerType::getUnqual(ITy));
// Alignment 0 is identity for alignment 1 for memset, but not store.
if (Alignment == 0) Alignment = 1;
// Extract the fill value and store.
uint64_t Fill = FillC->getZExtValue()*0x0101010101010101ULL;
InsertNewInstBefore(new StoreInst(ConstantInt::get(ITy, Fill),
Dest, false, Alignment), *MI);
// Set the size of the copy to 0, it will be deleted on the next iteration.
MI->setLength(Constant::getNullValue(LenC->getType()));
return MI;
}
return 0;
}
/// visitCallInst - CallInst simplification. This mostly only handles folding
/// of intrinsic instructions. For normal calls, it allows visitCallSite to do
/// the heavy lifting.
///
Instruction *InstCombiner::visitCallInst(CallInst &CI) {
if (isFreeCall(&CI))
return visitFree(CI);
if (isMalloc(&CI))
return visitMalloc(CI);
// If the caller function is nounwind, mark the call as nounwind, even if the
// callee isn't.
if (CI.getParent()->getParent()->doesNotThrow() &&
!CI.doesNotThrow()) {
CI.setDoesNotThrow();
return &CI;
}
IntrinsicInst *II = dyn_cast<IntrinsicInst>(&CI);
if (!II) return visitCallSite(&CI);
// Intrinsics cannot occur in an invoke, so handle them here instead of in
// visitCallSite.
if (MemIntrinsic *MI = dyn_cast<MemIntrinsic>(II)) {
bool Changed = false;
// memmove/cpy/set of zero bytes is a noop.
if (Constant *NumBytes = dyn_cast<Constant>(MI->getLength())) {
if (NumBytes->isNullValue()) return EraseInstFromFunction(CI);
if (ConstantInt *CI = dyn_cast<ConstantInt>(NumBytes))
if (CI->getZExtValue() == 1) {
// Replace the instruction with just byte operations. We would
// transform other cases to loads/stores, but we don't know if
// alignment is sufficient.
}
}
// If we have a memmove and the source operation is a constant global,
// then the source and dest pointers can't alias, so we can change this
// into a call to memcpy.
if (MemMoveInst *MMI = dyn_cast<MemMoveInst>(MI)) {
if (GlobalVariable *GVSrc = dyn_cast<GlobalVariable>(MMI->getSource()))
if (GVSrc->isConstant()) {
Module *M = CI.getParent()->getParent()->getParent();
Intrinsic::ID MemCpyID = Intrinsic::memcpy;
const Type *Tys[3] = { CI.getOperand(1)->getType(),
CI.getOperand(2)->getType(),
CI.getOperand(3)->getType() };
CI.setCalledFunction(
Intrinsic::getDeclaration(M, MemCpyID, Tys, 3));
Changed = true;
}
}
if (MemTransferInst *MTI = dyn_cast<MemTransferInst>(MI)) {
// memmove(x,x,size) -> noop.
if (MTI->getSource() == MTI->getDest())
return EraseInstFromFunction(CI);
}
// If we can determine a pointer alignment that is bigger than currently
// set, update the alignment.
if (isa<MemTransferInst>(MI)) {
if (Instruction *I = SimplifyMemTransfer(MI))
return I;
} else if (MemSetInst *MSI = dyn_cast<MemSetInst>(MI)) {
if (Instruction *I = SimplifyMemSet(MSI))
return I;
}
if (Changed) return II;
}
switch (II->getIntrinsicID()) {
default: break;
case Intrinsic::objectsize: {
// We need target data for just about everything so depend on it.
if (!TD) break;
const Type *ReturnTy = CI.getType();
bool Min = (cast<ConstantInt>(II->getOperand(2))->getZExtValue() == 1);
// Get to the real allocated thing and offset as fast as possible.
Value *Op1 = II->getOperand(1)->stripPointerCasts();
// If we've stripped down to a single global variable that we
// can know the size of then just return that.
if (GlobalVariable *GV = dyn_cast<GlobalVariable>(Op1)) {
if (GV->hasDefinitiveInitializer()) {
Constant *C = GV->getInitializer();
uint64_t GlobalSize = TD->getTypeAllocSize(C->getType());
return ReplaceInstUsesWith(CI, ConstantInt::get(ReturnTy, GlobalSize));
} else {
// Can't determine size of the GV.
Constant *RetVal = ConstantInt::get(ReturnTy, Min ? 0 : -1ULL);
return ReplaceInstUsesWith(CI, RetVal);
}
} else if (AllocaInst *AI = dyn_cast<AllocaInst>(Op1)) {
// Get alloca size.
if (AI->getAllocatedType()->isSized()) {
uint64_t AllocaSize = TD->getTypeAllocSize(AI->getAllocatedType());
if (AI->isArrayAllocation()) {
const ConstantInt *C = dyn_cast<ConstantInt>(AI->getArraySize());
if (!C) break;
AllocaSize *= C->getZExtValue();
}
return ReplaceInstUsesWith(CI, ConstantInt::get(ReturnTy, AllocaSize));
}
} else if (CallInst *MI = extractMallocCall(Op1)) {
const Type* MallocType = getMallocAllocatedType(MI);
// Get alloca size.
if (MallocType && MallocType->isSized()) {
if (Value *NElems = getMallocArraySize(MI, TD, true)) {
if (ConstantInt *NElements = dyn_cast<ConstantInt>(NElems))
return ReplaceInstUsesWith(CI, ConstantInt::get(ReturnTy,
(NElements->getZExtValue() * TD->getTypeAllocSize(MallocType))));
}
}
} else if (ConstantExpr *CE = dyn_cast<ConstantExpr>(Op1)) {
// Only handle constant GEPs here.
if (CE->getOpcode() != Instruction::GetElementPtr) break;
GEPOperator *GEP = cast<GEPOperator>(CE);
// Make sure we're not a constant offset from an external
// global.
Value *Operand = GEP->getPointerOperand();
Operand = Operand->stripPointerCasts();
if (GlobalVariable *GV = dyn_cast<GlobalVariable>(Operand))
if (!GV->hasDefinitiveInitializer()) break;
// Get what we're pointing to and its size.
const PointerType *BaseType =
cast<PointerType>(Operand->getType());
uint64_t Size = TD->getTypeAllocSize(BaseType->getElementType());
// Get the current byte offset into the thing. Use the original
// operand in case we're looking through a bitcast.
SmallVector<Value*, 8> Ops(CE->op_begin()+1, CE->op_end());
const PointerType *OffsetType =
cast<PointerType>(GEP->getPointerOperand()->getType());
uint64_t Offset = TD->getIndexedOffset(OffsetType, &Ops[0], Ops.size());
if (Size < Offset) {
// Out of bound reference? Negative index normalized to large
// index? Just return "I don't know".
Constant *RetVal = ConstantInt::get(ReturnTy, Min ? 0 : -1ULL);
return ReplaceInstUsesWith(CI, RetVal);
}
Constant *RetVal = ConstantInt::get(ReturnTy, Size-Offset);
return ReplaceInstUsesWith(CI, RetVal);
}
// Do not return "I don't know" here. Later optimization passes could
// make it possible to evaluate objectsize to a constant.
break;
}
case Intrinsic::bswap:
// bswap(bswap(x)) -> x
if (IntrinsicInst *Operand = dyn_cast<IntrinsicInst>(II->getOperand(1)))
if (Operand->getIntrinsicID() == Intrinsic::bswap)
return ReplaceInstUsesWith(CI, Operand->getOperand(1));
// bswap(trunc(bswap(x))) -> trunc(lshr(x, c))
if (TruncInst *TI = dyn_cast<TruncInst>(II->getOperand(1))) {
if (IntrinsicInst *Operand = dyn_cast<IntrinsicInst>(TI->getOperand(0)))
if (Operand->getIntrinsicID() == Intrinsic::bswap) {
unsigned C = Operand->getType()->getPrimitiveSizeInBits() -
TI->getType()->getPrimitiveSizeInBits();
Value *CV = ConstantInt::get(Operand->getType(), C);
Value *V = Builder->CreateLShr(Operand->getOperand(1), CV);
return new TruncInst(V, TI->getType());
}
}
break;
case Intrinsic::powi:
if (ConstantInt *Power = dyn_cast<ConstantInt>(II->getOperand(2))) {
// powi(x, 0) -> 1.0
if (Power->isZero())
return ReplaceInstUsesWith(CI, ConstantFP::get(CI.getType(), 1.0));
// powi(x, 1) -> x
if (Power->isOne())
return ReplaceInstUsesWith(CI, II->getOperand(1));
// powi(x, -1) -> 1/x
if (Power->isAllOnesValue())
return BinaryOperator::CreateFDiv(ConstantFP::get(CI.getType(), 1.0),
II->getOperand(1));
}
break;
case Intrinsic::cttz: {
// If all bits below the first known one are known zero,
// this value is constant.
const IntegerType *IT = cast<IntegerType>(II->getOperand(1)->getType());
uint32_t BitWidth = IT->getBitWidth();
APInt KnownZero(BitWidth, 0);
APInt KnownOne(BitWidth, 0);
ComputeMaskedBits(II->getOperand(1), APInt::getAllOnesValue(BitWidth),
KnownZero, KnownOne);
unsigned TrailingZeros = KnownOne.countTrailingZeros();
APInt Mask(APInt::getLowBitsSet(BitWidth, TrailingZeros));
if ((Mask & KnownZero) == Mask)
return ReplaceInstUsesWith(CI, ConstantInt::get(IT,
APInt(BitWidth, TrailingZeros)));
}
break;
case Intrinsic::ctlz: {
// If all bits above the first known one are known zero,
// this value is constant.
const IntegerType *IT = cast<IntegerType>(II->getOperand(1)->getType());
uint32_t BitWidth = IT->getBitWidth();
APInt KnownZero(BitWidth, 0);
APInt KnownOne(BitWidth, 0);
ComputeMaskedBits(II->getOperand(1), APInt::getAllOnesValue(BitWidth),
KnownZero, KnownOne);
unsigned LeadingZeros = KnownOne.countLeadingZeros();
APInt Mask(APInt::getHighBitsSet(BitWidth, LeadingZeros));
if ((Mask & KnownZero) == Mask)
return ReplaceInstUsesWith(CI, ConstantInt::get(IT,
APInt(BitWidth, LeadingZeros)));
}
break;
case Intrinsic::uadd_with_overflow: {
Value *LHS = II->getOperand(1), *RHS = II->getOperand(2);
const IntegerType *IT = cast<IntegerType>(II->getOperand(1)->getType());
uint32_t BitWidth = IT->getBitWidth();
APInt Mask = APInt::getSignBit(BitWidth);
APInt LHSKnownZero(BitWidth, 0);
APInt LHSKnownOne(BitWidth, 0);
ComputeMaskedBits(LHS, Mask, LHSKnownZero, LHSKnownOne);
bool LHSKnownNegative = LHSKnownOne[BitWidth - 1];
bool LHSKnownPositive = LHSKnownZero[BitWidth - 1];
if (LHSKnownNegative || LHSKnownPositive) {
APInt RHSKnownZero(BitWidth, 0);
APInt RHSKnownOne(BitWidth, 0);
ComputeMaskedBits(RHS, Mask, RHSKnownZero, RHSKnownOne);
bool RHSKnownNegative = RHSKnownOne[BitWidth - 1];
bool RHSKnownPositive = RHSKnownZero[BitWidth - 1];
if (LHSKnownNegative && RHSKnownNegative) {
// The sign bit is set in both cases: this MUST overflow.
// Create a simple add instruction, and insert it into the struct.
Instruction *Add = BinaryOperator::CreateAdd(LHS, RHS, "", &CI);
Worklist.Add(Add);
Constant *V[] = {
UndefValue::get(LHS->getType()),ConstantInt::getTrue(II->getContext())
};
Constant *Struct = ConstantStruct::get(II->getContext(), V, 2, false);
return InsertValueInst::Create(Struct, Add, 0);
}
if (LHSKnownPositive && RHSKnownPositive) {
// The sign bit is clear in both cases: this CANNOT overflow.
// Create a simple add instruction, and insert it into the struct.
Instruction *Add = BinaryOperator::CreateNUWAdd(LHS, RHS, "", &CI);
Worklist.Add(Add);
Constant *V[] = {
UndefValue::get(LHS->getType()),
ConstantInt::getFalse(II->getContext())
};
Constant *Struct = ConstantStruct::get(II->getContext(), V, 2, false);
return InsertValueInst::Create(Struct, Add, 0);
}
}
}
// FALL THROUGH uadd into sadd
case Intrinsic::sadd_with_overflow:
// Canonicalize constants into the RHS.
if (isa<Constant>(II->getOperand(1)) &&
!isa<Constant>(II->getOperand(2))) {
Value *LHS = II->getOperand(1);
II->setOperand(1, II->getOperand(2));
II->setOperand(2, LHS);
return II;
}
// X + undef -> undef
if (isa<UndefValue>(II->getOperand(2)))
return ReplaceInstUsesWith(CI, UndefValue::get(II->getType()));
if (ConstantInt *RHS = dyn_cast<ConstantInt>(II->getOperand(2))) {
// X + 0 -> {X, false}
if (RHS->isZero()) {
Constant *V[] = {
UndefValue::get(II->getCalledValue()->getType()),
ConstantInt::getFalse(II->getContext())
};
Constant *Struct = ConstantStruct::get(II->getContext(), V, 2, false);
return InsertValueInst::Create(Struct, II->getOperand(1), 0);
}
}
break;
case Intrinsic::usub_with_overflow:
case Intrinsic::ssub_with_overflow:
// undef - X -> undef
// X - undef -> undef
if (isa<UndefValue>(II->getOperand(1)) ||
isa<UndefValue>(II->getOperand(2)))
return ReplaceInstUsesWith(CI, UndefValue::get(II->getType()));
if (ConstantInt *RHS = dyn_cast<ConstantInt>(II->getOperand(2))) {
// X - 0 -> {X, false}
if (RHS->isZero()) {
Constant *V[] = {
UndefValue::get(II->getOperand(1)->getType()),
ConstantInt::getFalse(II->getContext())
};
Constant *Struct = ConstantStruct::get(II->getContext(), V, 2, false);
return InsertValueInst::Create(Struct, II->getOperand(1), 0);
}
}
break;
case Intrinsic::umul_with_overflow:
case Intrinsic::smul_with_overflow:
// Canonicalize constants into the RHS.
if (isa<Constant>(II->getOperand(1)) &&
!isa<Constant>(II->getOperand(2))) {
Value *LHS = II->getOperand(1);
II->setOperand(1, II->getOperand(2));
II->setOperand(2, LHS);
return II;
}
// X * undef -> undef
if (isa<UndefValue>(II->getOperand(2)))
return ReplaceInstUsesWith(CI, UndefValue::get(II->getType()));
if (ConstantInt *RHSI = dyn_cast<ConstantInt>(II->getOperand(2))) {
// X*0 -> {0, false}
if (RHSI->isZero())
return ReplaceInstUsesWith(CI, Constant::getNullValue(II->getType()));
// X * 1 -> {X, false}
if (RHSI->equalsInt(1)) {
Constant *V[] = {
UndefValue::get(II->getOperand(1)->getType()),
ConstantInt::getFalse(II->getContext())
};
Constant *Struct = ConstantStruct::get(II->getContext(), V, 2, false);
return InsertValueInst::Create(Struct, II->getOperand(1), 0);
}
}
break;
case Intrinsic::ppc_altivec_lvx:
case Intrinsic::ppc_altivec_lvxl:
case Intrinsic::x86_sse_loadu_ps:
case Intrinsic::x86_sse2_loadu_pd:
case Intrinsic::x86_sse2_loadu_dq:
// Turn PPC lvx -> load if the pointer is known aligned.
// Turn X86 loadups -> load if the pointer is known aligned.
if (GetOrEnforceKnownAlignment(II->getOperand(1), 16) >= 16) {
Value *Ptr = Builder->CreateBitCast(II->getOperand(1),
PointerType::getUnqual(II->getType()));
return new LoadInst(Ptr);
}
break;
case Intrinsic::ppc_altivec_stvx:
case Intrinsic::ppc_altivec_stvxl:
// Turn stvx -> store if the pointer is known aligned.
if (GetOrEnforceKnownAlignment(II->getOperand(2), 16) >= 16) {
const Type *OpPtrTy =
PointerType::getUnqual(II->getOperand(1)->getType());
Value *Ptr = Builder->CreateBitCast(II->getOperand(2), OpPtrTy);
return new StoreInst(II->getOperand(1), Ptr);
}
break;
case Intrinsic::x86_sse_storeu_ps:
case Intrinsic::x86_sse2_storeu_pd:
case Intrinsic::x86_sse2_storeu_dq:
// Turn X86 storeu -> store if the pointer is known aligned.
if (GetOrEnforceKnownAlignment(II->getOperand(1), 16) >= 16) {
const Type *OpPtrTy =
PointerType::getUnqual(II->getOperand(2)->getType());
Value *Ptr = Builder->CreateBitCast(II->getOperand(1), OpPtrTy);
return new StoreInst(II->getOperand(2), Ptr);
}
break;
case Intrinsic::x86_sse_cvttss2si: {
// These intrinsics only demands the 0th element of its input vector. If
// we can simplify the input based on that, do so now.
unsigned VWidth =
cast<VectorType>(II->getOperand(1)->getType())->getNumElements();
APInt DemandedElts(VWidth, 1);
APInt UndefElts(VWidth, 0);
if (Value *V = SimplifyDemandedVectorElts(II->getOperand(1), DemandedElts,
UndefElts)) {
II->setOperand(1, V);
return II;
}
break;
}
case Intrinsic::ppc_altivec_vperm:
// Turn vperm(V1,V2,mask) -> shuffle(V1,V2,mask) if mask is a constant.
if (ConstantVector *Mask = dyn_cast<ConstantVector>(II->getOperand(3))) {
assert(Mask->getNumOperands() == 16 && "Bad type for intrinsic!");
// Check that all of the elements are integer constants or undefs.
bool AllEltsOk = true;
for (unsigned i = 0; i != 16; ++i) {
if (!isa<ConstantInt>(Mask->getOperand(i)) &&
!isa<UndefValue>(Mask->getOperand(i))) {
AllEltsOk = false;
break;
}
}
if (AllEltsOk) {
// Cast the input vectors to byte vectors.
Value *Op0 = Builder->CreateBitCast(II->getOperand(1), Mask->getType());
Value *Op1 = Builder->CreateBitCast(II->getOperand(2), Mask->getType());
Value *Result = UndefValue::get(Op0->getType());
// Only extract each element once.
Value *ExtractedElts[32];
memset(ExtractedElts, 0, sizeof(ExtractedElts));
for (unsigned i = 0; i != 16; ++i) {
if (isa<UndefValue>(Mask->getOperand(i)))
continue;
unsigned Idx=cast<ConstantInt>(Mask->getOperand(i))->getZExtValue();
Idx &= 31; // Match the hardware behavior.
if (ExtractedElts[Idx] == 0) {
ExtractedElts[Idx] =
Builder->CreateExtractElement(Idx < 16 ? Op0 : Op1,
ConstantInt::get(Type::getInt32Ty(II->getContext()),
Idx&15, false), "tmp");
}
// Insert this value into the result vector.
Result = Builder->CreateInsertElement(Result, ExtractedElts[Idx],
ConstantInt::get(Type::getInt32Ty(II->getContext()),
i, false), "tmp");
}
return CastInst::Create(Instruction::BitCast, Result, CI.getType());
}
}
break;
case Intrinsic::stackrestore: {
// If the save is right next to the restore, remove the restore. This can
// happen when variable allocas are DCE'd.
if (IntrinsicInst *SS = dyn_cast<IntrinsicInst>(II->getOperand(1))) {
if (SS->getIntrinsicID() == Intrinsic::stacksave) {
BasicBlock::iterator BI = SS;
if (&*++BI == II)
return EraseInstFromFunction(CI);
}
}
// Scan down this block to see if there is another stack restore in the
// same block without an intervening call/alloca.
BasicBlock::iterator BI = II;
TerminatorInst *TI = II->getParent()->getTerminator();
bool CannotRemove = false;
for (++BI; &*BI != TI; ++BI) {
if (isa<AllocaInst>(BI) || isMalloc(BI)) {
CannotRemove = true;
break;
}
if (CallInst *BCI = dyn_cast<CallInst>(BI)) {
if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(BCI)) {
// If there is a stackrestore below this one, remove this one.
if (II->getIntrinsicID() == Intrinsic::stackrestore)
return EraseInstFromFunction(CI);
// Otherwise, ignore the intrinsic.
} else {
// If we found a non-intrinsic call, we can't remove the stack
// restore.
CannotRemove = true;
break;
}
}
}
// If the stack restore is in a return/unwind block and if there are no
// allocas or calls between the restore and the return, nuke the restore.
if (!CannotRemove && (isa<ReturnInst>(TI) || isa<UnwindInst>(TI)))
return EraseInstFromFunction(CI);
break;
}
}
return visitCallSite(II);
}
// InvokeInst simplification
//
Instruction *InstCombiner::visitInvokeInst(InvokeInst &II) {
return visitCallSite(&II);
}
/// isSafeToEliminateVarargsCast - If this cast does not affect the value
/// passed through the varargs area, we can eliminate the use of the cast.
static bool isSafeToEliminateVarargsCast(const CallSite CS,
const CastInst * const CI,
const TargetData * const TD,
const int ix) {
if (!CI->isLosslessCast())
return false;
// The size of ByVal arguments is derived from the type, so we
// can't change to a type with a different size. If the size were
// passed explicitly we could avoid this check.
if (!CS.paramHasAttr(ix, Attribute::ByVal))
return true;
const Type* SrcTy =
cast<PointerType>(CI->getOperand(0)->getType())->getElementType();
const Type* DstTy = cast<PointerType>(CI->getType())->getElementType();
if (!SrcTy->isSized() || !DstTy->isSized())
return false;
if (!TD || TD->getTypeAllocSize(SrcTy) != TD->getTypeAllocSize(DstTy))
return false;
return true;
}
namespace {
class InstCombineFortifiedLibCalls : public SimplifyFortifiedLibCalls {
InstCombiner *IC;
protected:
void replaceCall(Value *With) {
NewInstruction = IC->ReplaceInstUsesWith(*CI, With);
}
bool isFoldable(unsigned SizeCIOp, unsigned SizeArgOp, bool isString) const {
if (ConstantInt *SizeCI = dyn_cast<ConstantInt>(CI->getOperand(SizeCIOp))) {
if (SizeCI->isAllOnesValue())
return true;
if (isString)
return SizeCI->getZExtValue() >=
GetStringLength(CI->getOperand(SizeArgOp));
if (ConstantInt *Arg = dyn_cast<ConstantInt>(CI->getOperand(SizeArgOp)))
return SizeCI->getZExtValue() >= Arg->getZExtValue();
}
return false;
}
public:
InstCombineFortifiedLibCalls(InstCombiner *IC) : IC(IC), NewInstruction(0) { }
Instruction *NewInstruction;
};
} // end anonymous namespace
// Try to fold some different type of calls here.
// Currently we're only working with the checking functions, memcpy_chk,
// mempcpy_chk, memmove_chk, memset_chk, strcpy_chk, stpcpy_chk, strncpy_chk,
// strcat_chk and strncat_chk.
Instruction *InstCombiner::tryOptimizeCall(CallInst *CI, const TargetData *TD) {
if (CI->getCalledFunction() == 0) return 0;
InstCombineFortifiedLibCalls Simplifier(this);
Simplifier.fold(CI, TD);
return Simplifier.NewInstruction;
}
// visitCallSite - Improvements for call and invoke instructions.
//
Instruction *InstCombiner::visitCallSite(CallSite CS) {
bool Changed = false;
// If the callee is a constexpr cast of a function, attempt to move the cast
// to the arguments of the call/invoke.
if (transformConstExprCastCall(CS)) return 0;
Value *Callee = CS.getCalledValue();
if (Function *CalleeF = dyn_cast<Function>(Callee))
// If the call and callee calling conventions don't match, this call must
// be unreachable, as the call is undefined.
if (CalleeF->getCallingConv() != CS.getCallingConv() &&
// Only do this for calls to a function with a body. A prototype may
// not actually end up matching the implementation's calling conv for a
// variety of reasons (e.g. it may be written in assembly).
!CalleeF->isDeclaration()) {
Instruction *OldCall = CS.getInstruction();
new StoreInst(ConstantInt::getTrue(Callee->getContext()),
UndefValue::get(Type::getInt1PtrTy(Callee->getContext())),
OldCall);
// If OldCall dues not return void then replaceAllUsesWith undef.
// This allows ValueHandlers and custom metadata to adjust itself.
if (!OldCall->getType()->isVoidTy())
OldCall->replaceAllUsesWith(UndefValue::get(OldCall->getType()));
if (isa<CallInst>(OldCall))
return EraseInstFromFunction(*OldCall);
// We cannot remove an invoke, because it would change the CFG, just
// change the callee to a null pointer.
cast<InvokeInst>(OldCall)->setCalledFunction(
Constant::getNullValue(CalleeF->getType()));
return 0;
}
if (isa<ConstantPointerNull>(Callee) || isa<UndefValue>(Callee)) {
// This instruction is not reachable, just remove it. We insert a store to
// undef so that we know that this code is not reachable, despite the fact
// that we can't modify the CFG here.
new StoreInst(ConstantInt::getTrue(Callee->getContext()),
UndefValue::get(Type::getInt1PtrTy(Callee->getContext())),
CS.getInstruction());
// If CS dues not return void then replaceAllUsesWith undef.
// This allows ValueHandlers and custom metadata to adjust itself.
if (!CS.getInstruction()->getType()->isVoidTy())
CS.getInstruction()->
replaceAllUsesWith(UndefValue::get(CS.getInstruction()->getType()));
if (InvokeInst *II = dyn_cast<InvokeInst>(CS.getInstruction())) {
// Don't break the CFG, insert a dummy cond branch.
BranchInst::Create(II->getNormalDest(), II->getUnwindDest(),
ConstantInt::getTrue(Callee->getContext()), II);
}
return EraseInstFromFunction(*CS.getInstruction());
}
if (BitCastInst *BC = dyn_cast<BitCastInst>(Callee))
if (IntrinsicInst *In = dyn_cast<IntrinsicInst>(BC->getOperand(0)))
if (In->getIntrinsicID() == Intrinsic::init_trampoline)
return transformCallThroughTrampoline(CS);
const PointerType *PTy = cast<PointerType>(Callee->getType());
const FunctionType *FTy = cast<FunctionType>(PTy->getElementType());
if (FTy->isVarArg()) {
int ix = FTy->getNumParams() + (isa<InvokeInst>(Callee) ? 3 : 1);
// See if we can optimize any arguments passed through the varargs area of
// the call.
for (CallSite::arg_iterator I = CS.arg_begin()+FTy->getNumParams(),
E = CS.arg_end(); I != E; ++I, ++ix) {
CastInst *CI = dyn_cast<CastInst>(*I);
if (CI && isSafeToEliminateVarargsCast(CS, CI, TD, ix)) {
*I = CI->getOperand(0);
Changed = true;
}
}
}
if (isa<InlineAsm>(Callee) && !CS.doesNotThrow()) {
// Inline asm calls cannot throw - mark them 'nounwind'.
CS.setDoesNotThrow();
Changed = true;
}
// Try to optimize the call if possible, we require TargetData for most of
// this. None of these calls are seen as possibly dead so go ahead and
// delete the instruction now.
if (CallInst *CI = dyn_cast<CallInst>(CS.getInstruction())) {
Instruction *I = tryOptimizeCall(CI, TD);
// If we changed something return the result, etc. Otherwise let
// the fallthrough check.
if (I) return EraseInstFromFunction(*I);
}
return Changed ? CS.getInstruction() : 0;
}
// transformConstExprCastCall - If the callee is a constexpr cast of a function,
// attempt to move the cast to the arguments of the call/invoke.
//
bool InstCombiner::transformConstExprCastCall(CallSite CS) {
if (!isa<ConstantExpr>(CS.getCalledValue())) return false;
ConstantExpr *CE = cast<ConstantExpr>(CS.getCalledValue());
if (CE->getOpcode() != Instruction::BitCast ||
!isa<Function>(CE->getOperand(0)))
return false;
Function *Callee = cast<Function>(CE->getOperand(0));
Instruction *Caller = CS.getInstruction();
const AttrListPtr &CallerPAL = CS.getAttributes();
// Okay, this is a cast from a function to a different type. Unless doing so
// would cause a type conversion of one of our arguments, change this call to
// be a direct call with arguments casted to the appropriate types.
//
const FunctionType *FT = Callee->getFunctionType();
const Type *OldRetTy = Caller->getType();
const Type *NewRetTy = FT->getReturnType();
if (NewRetTy->isStructTy())
return false; // TODO: Handle multiple return values.
// Check to see if we are changing the return type...
if (OldRetTy != NewRetTy) {
if (Callee->isDeclaration() &&
// Conversion is ok if changing from one pointer type to another or from
// a pointer to an integer of the same size.
!((OldRetTy->isPointerTy() || !TD ||
OldRetTy == TD->getIntPtrType(Caller->getContext())) &&
(NewRetTy->isPointerTy() || !TD ||
NewRetTy == TD->getIntPtrType(Caller->getContext()))))
return false; // Cannot transform this return value.
if (!Caller->use_empty() &&
// void -> non-void is handled specially
!NewRetTy->isVoidTy() && !CastInst::isCastable(NewRetTy, OldRetTy))
return false; // Cannot transform this return value.
if (!CallerPAL.isEmpty() && !Caller->use_empty()) {
Attributes RAttrs = CallerPAL.getRetAttributes();
if (RAttrs & Attribute::typeIncompatible(NewRetTy))
return false; // Attribute not compatible with transformed value.
}
// If the callsite is an invoke instruction, and the return value is used by
// a PHI node in a successor, we cannot change the return type of the call
// because there is no place to put the cast instruction (without breaking
// the critical edge). Bail out in this case.
if (!Caller->use_empty())
if (InvokeInst *II = dyn_cast<InvokeInst>(Caller))
for (Value::use_iterator UI = II->use_begin(), E = II->use_end();
UI != E; ++UI)
if (PHINode *PN = dyn_cast<PHINode>(*UI))
if (PN->getParent() == II->getNormalDest() ||
PN->getParent() == II->getUnwindDest())
return false;
}
unsigned NumActualArgs = unsigned(CS.arg_end()-CS.arg_begin());
unsigned NumCommonArgs = std::min(FT->getNumParams(), NumActualArgs);
CallSite::arg_iterator AI = CS.arg_begin();
for (unsigned i = 0, e = NumCommonArgs; i != e; ++i, ++AI) {
const Type *ParamTy = FT->getParamType(i);
const Type *ActTy = (*AI)->getType();
if (!CastInst::isCastable(ActTy, ParamTy))
return false; // Cannot transform this parameter value.
if (CallerPAL.getParamAttributes(i + 1)
& Attribute::typeIncompatible(ParamTy))
return false; // Attribute not compatible with transformed value.
// Converting from one pointer type to another or between a pointer and an
// integer of the same size is safe even if we do not have a body.
bool isConvertible = ActTy == ParamTy ||
(TD && ((ParamTy->isPointerTy() ||
ParamTy == TD->getIntPtrType(Caller->getContext())) &&
(ActTy->isPointerTy() ||
ActTy == TD->getIntPtrType(Caller->getContext()))));
if (Callee->isDeclaration() && !isConvertible) return false;
}
if (FT->getNumParams() < NumActualArgs && !FT->isVarArg() &&
Callee->isDeclaration())
return false; // Do not delete arguments unless we have a function body.
if (FT->getNumParams() < NumActualArgs && FT->isVarArg() &&
!CallerPAL.isEmpty())
// In this case we have more arguments than the new function type, but we
// won't be dropping them. Check that these extra arguments have attributes
// that are compatible with being a vararg call argument.
for (unsigned i = CallerPAL.getNumSlots(); i; --i) {
if (CallerPAL.getSlot(i - 1).Index <= FT->getNumParams())
break;
Attributes PAttrs = CallerPAL.getSlot(i - 1).Attrs;
if (PAttrs & Attribute::VarArgsIncompatible)
return false;
}
// Okay, we decided that this is a safe thing to do: go ahead and start
// inserting cast instructions as necessary...
std::vector<Value*> Args;
Args.reserve(NumActualArgs);
SmallVector<AttributeWithIndex, 8> attrVec;
attrVec.reserve(NumCommonArgs);
// Get any return attributes.
Attributes RAttrs = CallerPAL.getRetAttributes();
// If the return value is not being used, the type may not be compatible
// with the existing attributes. Wipe out any problematic attributes.
RAttrs &= ~Attribute::typeIncompatible(NewRetTy);
// Add the new return attributes.
if (RAttrs)
attrVec.push_back(AttributeWithIndex::get(0, RAttrs));
AI = CS.arg_begin();
for (unsigned i = 0; i != NumCommonArgs; ++i, ++AI) {
const Type *ParamTy = FT->getParamType(i);
if ((*AI)->getType() == ParamTy) {
Args.push_back(*AI);
} else {
Instruction::CastOps opcode = CastInst::getCastOpcode(*AI,
false, ParamTy, false);
Args.push_back(Builder->CreateCast(opcode, *AI, ParamTy, "tmp"));
}
// Add any parameter attributes.
if (Attributes PAttrs = CallerPAL.getParamAttributes(i + 1))
attrVec.push_back(AttributeWithIndex::get(i + 1, PAttrs));
}
// If the function takes more arguments than the call was taking, add them
// now.
for (unsigned i = NumCommonArgs; i != FT->getNumParams(); ++i)
Args.push_back(Constant::getNullValue(FT->getParamType(i)));
// If we are removing arguments to the function, emit an obnoxious warning.
if (FT->getNumParams() < NumActualArgs) {
if (!FT->isVarArg()) {
errs() << "WARNING: While resolving call to function '"
<< Callee->getName() << "' arguments were dropped!\n";
} else {
// Add all of the arguments in their promoted form to the arg list.
for (unsigned i = FT->getNumParams(); i != NumActualArgs; ++i, ++AI) {
const Type *PTy = getPromotedType((*AI)->getType());
if (PTy != (*AI)->getType()) {
// Must promote to pass through va_arg area!
Instruction::CastOps opcode =
CastInst::getCastOpcode(*AI, false, PTy, false);
Args.push_back(Builder->CreateCast(opcode, *AI, PTy, "tmp"));
} else {
Args.push_back(*AI);
}
// Add any parameter attributes.
if (Attributes PAttrs = CallerPAL.getParamAttributes(i + 1))
attrVec.push_back(AttributeWithIndex::get(i + 1, PAttrs));
}
}
}
if (Attributes FnAttrs = CallerPAL.getFnAttributes())
attrVec.push_back(AttributeWithIndex::get(~0, FnAttrs));
if (NewRetTy->isVoidTy())
Caller->setName(""); // Void type should not have a name.
const AttrListPtr &NewCallerPAL = AttrListPtr::get(attrVec.begin(),
attrVec.end());
Instruction *NC;
if (InvokeInst *II = dyn_cast<InvokeInst>(Caller)) {
NC = InvokeInst::Create(Callee, II->getNormalDest(), II->getUnwindDest(),
Args.begin(), Args.end(),
Caller->getName(), Caller);
cast<InvokeInst>(NC)->setCallingConv(II->getCallingConv());
cast<InvokeInst>(NC)->setAttributes(NewCallerPAL);
} else {
NC = CallInst::Create(Callee, Args.begin(), Args.end(),
Caller->getName(), Caller);
CallInst *CI = cast<CallInst>(Caller);
if (CI->isTailCall())
cast<CallInst>(NC)->setTailCall();
cast<CallInst>(NC)->setCallingConv(CI->getCallingConv());
cast<CallInst>(NC)->setAttributes(NewCallerPAL);
}
// Insert a cast of the return type as necessary.
Value *NV = NC;
if (OldRetTy != NV->getType() && !Caller->use_empty()) {
if (!NV->getType()->isVoidTy()) {
Instruction::CastOps opcode = CastInst::getCastOpcode(NC, false,
OldRetTy, false);
NV = NC = CastInst::Create(opcode, NC, OldRetTy, "tmp");
// If this is an invoke instruction, we should insert it after the first
// non-phi, instruction in the normal successor block.
if (InvokeInst *II = dyn_cast<InvokeInst>(Caller)) {
BasicBlock::iterator I = II->getNormalDest()->getFirstNonPHI();
InsertNewInstBefore(NC, *I);
} else {
// Otherwise, it's a call, just insert cast right after the call instr
InsertNewInstBefore(NC, *Caller);
}
Worklist.AddUsersToWorkList(*Caller);
} else {
NV = UndefValue::get(Caller->getType());
}
}
if (!Caller->use_empty())
Caller->replaceAllUsesWith(NV);
EraseInstFromFunction(*Caller);
return true;
}
// transformCallThroughTrampoline - Turn a call to a function created by the
// init_trampoline intrinsic into a direct call to the underlying function.
//
Instruction *InstCombiner::transformCallThroughTrampoline(CallSite CS) {
Value *Callee = CS.getCalledValue();
const PointerType *PTy = cast<PointerType>(Callee->getType());
const FunctionType *FTy = cast<FunctionType>(PTy->getElementType());
const AttrListPtr &Attrs = CS.getAttributes();
// If the call already has the 'nest' attribute somewhere then give up -
// otherwise 'nest' would occur twice after splicing in the chain.
if (Attrs.hasAttrSomewhere(Attribute::Nest))
return 0;
IntrinsicInst *Tramp =
cast<IntrinsicInst>(cast<BitCastInst>(Callee)->getOperand(0));
Function *NestF = cast<Function>(Tramp->getOperand(2)->stripPointerCasts());
const PointerType *NestFPTy = cast<PointerType>(NestF->getType());
const FunctionType *NestFTy = cast<FunctionType>(NestFPTy->getElementType());
const AttrListPtr &NestAttrs = NestF->getAttributes();
if (!NestAttrs.isEmpty()) {
unsigned NestIdx = 1;
const Type *NestTy = 0;
Attributes NestAttr = Attribute::None;
// Look for a parameter marked with the 'nest' attribute.
for (FunctionType::param_iterator I = NestFTy->param_begin(),
E = NestFTy->param_end(); I != E; ++NestIdx, ++I)
if (NestAttrs.paramHasAttr(NestIdx, Attribute::Nest)) {
// Record the parameter type and any other attributes.
NestTy = *I;
NestAttr = NestAttrs.getParamAttributes(NestIdx);
break;
}
if (NestTy) {
Instruction *Caller = CS.getInstruction();
std::vector<Value*> NewArgs;
NewArgs.reserve(unsigned(CS.arg_end()-CS.arg_begin())+1);
SmallVector<AttributeWithIndex, 8> NewAttrs;
NewAttrs.reserve(Attrs.getNumSlots() + 1);
// Insert the nest argument into the call argument list, which may
// mean appending it. Likewise for attributes.
// Add any result attributes.
if (Attributes Attr = Attrs.getRetAttributes())
NewAttrs.push_back(AttributeWithIndex::get(0, Attr));
{
unsigned Idx = 1;
CallSite::arg_iterator I = CS.arg_begin(), E = CS.arg_end();
do {
if (Idx == NestIdx) {
// Add the chain argument and attributes.
Value *NestVal = Tramp->getOperand(3);
if (NestVal->getType() != NestTy)
NestVal = new BitCastInst(NestVal, NestTy, "nest", Caller);
NewArgs.push_back(NestVal);
NewAttrs.push_back(AttributeWithIndex::get(NestIdx, NestAttr));
}
if (I == E)
break;
// Add the original argument and attributes.
NewArgs.push_back(*I);
if (Attributes Attr = Attrs.getParamAttributes(Idx))
NewAttrs.push_back
(AttributeWithIndex::get(Idx + (Idx >= NestIdx), Attr));
++Idx, ++I;
} while (1);
}
// Add any function attributes.
if (Attributes Attr = Attrs.getFnAttributes())
NewAttrs.push_back(AttributeWithIndex::get(~0, Attr));
// The trampoline may have been bitcast to a bogus type (FTy).
// Handle this by synthesizing a new function type, equal to FTy
// with the chain parameter inserted.
std::vector<const Type*> NewTypes;
NewTypes.reserve(FTy->getNumParams()+1);
// Insert the chain's type into the list of parameter types, which may
// mean appending it.
{
unsigned Idx = 1;
FunctionType::param_iterator I = FTy->param_begin(),
E = FTy->param_end();
do {
if (Idx == NestIdx)
// Add the chain's type.
NewTypes.push_back(NestTy);
if (I == E)
break;
// Add the original type.
NewTypes.push_back(*I);
++Idx, ++I;
} while (1);
}
// Replace the trampoline call with a direct call. Let the generic
// code sort out any function type mismatches.
FunctionType *NewFTy = FunctionType::get(FTy->getReturnType(), NewTypes,
FTy->isVarArg());
Constant *NewCallee =
NestF->getType() == PointerType::getUnqual(NewFTy) ?
NestF : ConstantExpr::getBitCast(NestF,
PointerType::getUnqual(NewFTy));
const AttrListPtr &NewPAL = AttrListPtr::get(NewAttrs.begin(),
NewAttrs.end());
Instruction *NewCaller;
if (InvokeInst *II = dyn_cast<InvokeInst>(Caller)) {
NewCaller = InvokeInst::Create(NewCallee,
II->getNormalDest(), II->getUnwindDest(),
NewArgs.begin(), NewArgs.end(),
Caller->getName(), Caller);
cast<InvokeInst>(NewCaller)->setCallingConv(II->getCallingConv());
cast<InvokeInst>(NewCaller)->setAttributes(NewPAL);
} else {
NewCaller = CallInst::Create(NewCallee, NewArgs.begin(), NewArgs.end(),
Caller->getName(), Caller);
if (cast<CallInst>(Caller)->isTailCall())
cast<CallInst>(NewCaller)->setTailCall();
cast<CallInst>(NewCaller)->
setCallingConv(cast<CallInst>(Caller)->getCallingConv());
cast<CallInst>(NewCaller)->setAttributes(NewPAL);
}
if (!Caller->getType()->isVoidTy())
Caller->replaceAllUsesWith(NewCaller);
Caller->eraseFromParent();
Worklist.Remove(Caller);
return 0;
}
}
// Replace the trampoline call with a direct call. Since there is no 'nest'
// parameter, there is no need to adjust the argument list. Let the generic
// code sort out any function type mismatches.
Constant *NewCallee =
NestF->getType() == PTy ? NestF :
ConstantExpr::getBitCast(NestF, PTy);
CS.setCalledFunction(NewCallee);
return CS.getInstruction();
}
|