1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
|
//===- ScalarEvolutionExpander.cpp - Scalar Evolution Analysis --*- C++ -*-===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file contains the implementation of the scalar evolution expander,
// which is used to generate the code corresponding to a given scalar evolution
// expression.
//
//===----------------------------------------------------------------------===//
#include "llvm/Analysis/ScalarEvolutionExpander.h"
#include "llvm/Analysis/LoopInfo.h"
#include "llvm/IntrinsicInst.h"
#include "llvm/LLVMContext.h"
#include "llvm/Target/TargetData.h"
#include "llvm/ADT/STLExtras.h"
using namespace llvm;
/// InsertNoopCastOfTo - Insert a cast of V to the specified type,
/// which must be possible with a noop cast, doing what we can to share
/// the casts.
Value *SCEVExpander::InsertNoopCastOfTo(Value *V, const Type *Ty) {
Instruction::CastOps Op = CastInst::getCastOpcode(V, false, Ty, false);
assert((Op == Instruction::BitCast ||
Op == Instruction::PtrToInt ||
Op == Instruction::IntToPtr) &&
"InsertNoopCastOfTo cannot perform non-noop casts!");
assert(SE.getTypeSizeInBits(V->getType()) == SE.getTypeSizeInBits(Ty) &&
"InsertNoopCastOfTo cannot change sizes!");
// Short-circuit unnecessary bitcasts.
if (Op == Instruction::BitCast && V->getType() == Ty)
return V;
// Short-circuit unnecessary inttoptr<->ptrtoint casts.
if ((Op == Instruction::PtrToInt || Op == Instruction::IntToPtr) &&
SE.getTypeSizeInBits(Ty) == SE.getTypeSizeInBits(V->getType())) {
if (CastInst *CI = dyn_cast<CastInst>(V))
if ((CI->getOpcode() == Instruction::PtrToInt ||
CI->getOpcode() == Instruction::IntToPtr) &&
SE.getTypeSizeInBits(CI->getType()) ==
SE.getTypeSizeInBits(CI->getOperand(0)->getType()))
return CI->getOperand(0);
if (ConstantExpr *CE = dyn_cast<ConstantExpr>(V))
if ((CE->getOpcode() == Instruction::PtrToInt ||
CE->getOpcode() == Instruction::IntToPtr) &&
SE.getTypeSizeInBits(CE->getType()) ==
SE.getTypeSizeInBits(CE->getOperand(0)->getType()))
return CE->getOperand(0);
}
if (Constant *C = dyn_cast<Constant>(V))
return ConstantExpr::getCast(Op, C, Ty);
if (Argument *A = dyn_cast<Argument>(V)) {
// Check to see if there is already a cast!
for (Value::use_iterator UI = A->use_begin(), E = A->use_end();
UI != E; ++UI)
if ((*UI)->getType() == Ty)
if (CastInst *CI = dyn_cast<CastInst>(cast<Instruction>(*UI)))
if (CI->getOpcode() == Op) {
// If the cast isn't the first instruction of the function, move it.
if (BasicBlock::iterator(CI) !=
A->getParent()->getEntryBlock().begin()) {
// Recreate the cast at the beginning of the entry block.
// The old cast is left in place in case it is being used
// as an insert point.
Instruction *NewCI =
CastInst::Create(Op, V, Ty, "",
A->getParent()->getEntryBlock().begin());
NewCI->takeName(CI);
CI->replaceAllUsesWith(NewCI);
return NewCI;
}
return CI;
}
Instruction *I = CastInst::Create(Op, V, Ty, V->getName(),
A->getParent()->getEntryBlock().begin());
rememberInstruction(I);
return I;
}
Instruction *I = cast<Instruction>(V);
// Check to see if there is already a cast. If there is, use it.
for (Value::use_iterator UI = I->use_begin(), E = I->use_end();
UI != E; ++UI) {
if ((*UI)->getType() == Ty)
if (CastInst *CI = dyn_cast<CastInst>(cast<Instruction>(*UI)))
if (CI->getOpcode() == Op) {
BasicBlock::iterator It = I; ++It;
if (isa<InvokeInst>(I))
It = cast<InvokeInst>(I)->getNormalDest()->begin();
while (isa<PHINode>(It)) ++It;
if (It != BasicBlock::iterator(CI)) {
// Recreate the cast after the user.
// The old cast is left in place in case it is being used
// as an insert point.
Instruction *NewCI = CastInst::Create(Op, V, Ty, "", It);
NewCI->takeName(CI);
CI->replaceAllUsesWith(NewCI);
rememberInstruction(NewCI);
return NewCI;
}
rememberInstruction(CI);
return CI;
}
}
BasicBlock::iterator IP = I; ++IP;
if (InvokeInst *II = dyn_cast<InvokeInst>(I))
IP = II->getNormalDest()->begin();
while (isa<PHINode>(IP)) ++IP;
Instruction *CI = CastInst::Create(Op, V, Ty, V->getName(), IP);
rememberInstruction(CI);
return CI;
}
/// InsertBinop - Insert the specified binary operator, doing a small amount
/// of work to avoid inserting an obviously redundant operation.
Value *SCEVExpander::InsertBinop(Instruction::BinaryOps Opcode,
Value *LHS, Value *RHS) {
// Fold a binop with constant operands.
if (Constant *CLHS = dyn_cast<Constant>(LHS))
if (Constant *CRHS = dyn_cast<Constant>(RHS))
return ConstantExpr::get(Opcode, CLHS, CRHS);
// Do a quick scan to see if we have this binop nearby. If so, reuse it.
unsigned ScanLimit = 6;
BasicBlock::iterator BlockBegin = Builder.GetInsertBlock()->begin();
// Scanning starts from the last instruction before the insertion point.
BasicBlock::iterator IP = Builder.GetInsertPoint();
if (IP != BlockBegin) {
--IP;
for (; ScanLimit; --IP, --ScanLimit) {
// Don't count dbg.value against the ScanLimit, to avoid perturbing the
// generated code.
if (isa<DbgInfoIntrinsic>(IP))
ScanLimit++;
if (IP->getOpcode() == (unsigned)Opcode && IP->getOperand(0) == LHS &&
IP->getOperand(1) == RHS)
return IP;
if (IP == BlockBegin) break;
}
}
// Save the original insertion point so we can restore it when we're done.
BasicBlock *SaveInsertBB = Builder.GetInsertBlock();
BasicBlock::iterator SaveInsertPt = Builder.GetInsertPoint();
// Move the insertion point out of as many loops as we can.
while (const Loop *L = SE.LI->getLoopFor(Builder.GetInsertBlock())) {
if (!L->isLoopInvariant(LHS) || !L->isLoopInvariant(RHS)) break;
BasicBlock *Preheader = L->getLoopPreheader();
if (!Preheader) break;
// Ok, move up a level.
Builder.SetInsertPoint(Preheader, Preheader->getTerminator());
}
// If we haven't found this binop, insert it.
Value *BO = Builder.CreateBinOp(Opcode, LHS, RHS, "tmp");
rememberInstruction(BO);
// Restore the original insert point.
if (SaveInsertBB)
restoreInsertPoint(SaveInsertBB, SaveInsertPt);
return BO;
}
/// FactorOutConstant - Test if S is divisible by Factor, using signed
/// division. If so, update S with Factor divided out and return true.
/// S need not be evenly divisible if a reasonable remainder can be
/// computed.
/// TODO: When ScalarEvolution gets a SCEVSDivExpr, this can be made
/// unnecessary; in its place, just signed-divide Ops[i] by the scale and
/// check to see if the divide was folded.
static bool FactorOutConstant(const SCEV *&S,
const SCEV *&Remainder,
const SCEV *Factor,
ScalarEvolution &SE,
const TargetData *TD) {
// Everything is divisible by one.
if (Factor->isOne())
return true;
// x/x == 1.
if (S == Factor) {
S = SE.getIntegerSCEV(1, S->getType());
return true;
}
// For a Constant, check for a multiple of the given factor.
if (const SCEVConstant *C = dyn_cast<SCEVConstant>(S)) {
// 0/x == 0.
if (C->isZero())
return true;
// Check for divisibility.
if (const SCEVConstant *FC = dyn_cast<SCEVConstant>(Factor)) {
ConstantInt *CI =
ConstantInt::get(SE.getContext(),
C->getValue()->getValue().sdiv(
FC->getValue()->getValue()));
// If the quotient is zero and the remainder is non-zero, reject
// the value at this scale. It will be considered for subsequent
// smaller scales.
if (!CI->isZero()) {
const SCEV *Div = SE.getConstant(CI);
S = Div;
Remainder =
SE.getAddExpr(Remainder,
SE.getConstant(C->getValue()->getValue().srem(
FC->getValue()->getValue())));
return true;
}
}
}
// In a Mul, check if there is a constant operand which is a multiple
// of the given factor.
if (const SCEVMulExpr *M = dyn_cast<SCEVMulExpr>(S)) {
if (TD) {
// With TargetData, the size is known. Check if there is a constant
// operand which is a multiple of the given factor. If so, we can
// factor it.
const SCEVConstant *FC = cast<SCEVConstant>(Factor);
if (const SCEVConstant *C = dyn_cast<SCEVConstant>(M->getOperand(0)))
if (!C->getValue()->getValue().srem(FC->getValue()->getValue())) {
SmallVector<const SCEV *, 4> NewMulOps(M->op_begin(), M->op_end());
NewMulOps[0] =
SE.getConstant(C->getValue()->getValue().sdiv(
FC->getValue()->getValue()));
S = SE.getMulExpr(NewMulOps);
return true;
}
} else {
// Without TargetData, check if Factor can be factored out of any of the
// Mul's operands. If so, we can just remove it.
for (unsigned i = 0, e = M->getNumOperands(); i != e; ++i) {
const SCEV *SOp = M->getOperand(i);
const SCEV *Remainder = SE.getIntegerSCEV(0, SOp->getType());
if (FactorOutConstant(SOp, Remainder, Factor, SE, TD) &&
Remainder->isZero()) {
SmallVector<const SCEV *, 4> NewMulOps(M->op_begin(), M->op_end());
NewMulOps[i] = SOp;
S = SE.getMulExpr(NewMulOps);
return true;
}
}
}
}
// In an AddRec, check if both start and step are divisible.
if (const SCEVAddRecExpr *A = dyn_cast<SCEVAddRecExpr>(S)) {
const SCEV *Step = A->getStepRecurrence(SE);
const SCEV *StepRem = SE.getIntegerSCEV(0, Step->getType());
if (!FactorOutConstant(Step, StepRem, Factor, SE, TD))
return false;
if (!StepRem->isZero())
return false;
const SCEV *Start = A->getStart();
if (!FactorOutConstant(Start, Remainder, Factor, SE, TD))
return false;
S = SE.getAddRecExpr(Start, Step, A->getLoop());
return true;
}
return false;
}
/// SimplifyAddOperands - Sort and simplify a list of add operands. NumAddRecs
/// is the number of SCEVAddRecExprs present, which are kept at the end of
/// the list.
///
static void SimplifyAddOperands(SmallVectorImpl<const SCEV *> &Ops,
const Type *Ty,
ScalarEvolution &SE) {
unsigned NumAddRecs = 0;
for (unsigned i = Ops.size(); i > 0 && isa<SCEVAddRecExpr>(Ops[i-1]); --i)
++NumAddRecs;
// Group Ops into non-addrecs and addrecs.
SmallVector<const SCEV *, 8> NoAddRecs(Ops.begin(), Ops.end() - NumAddRecs);
SmallVector<const SCEV *, 8> AddRecs(Ops.end() - NumAddRecs, Ops.end());
// Let ScalarEvolution sort and simplify the non-addrecs list.
const SCEV *Sum = NoAddRecs.empty() ?
SE.getIntegerSCEV(0, Ty) :
SE.getAddExpr(NoAddRecs);
// If it returned an add, use the operands. Otherwise it simplified
// the sum into a single value, so just use that.
Ops.clear();
if (const SCEVAddExpr *Add = dyn_cast<SCEVAddExpr>(Sum))
Ops.insert(Ops.end(), Add->op_begin(), Add->op_end());
else if (!Sum->isZero())
Ops.push_back(Sum);
// Then append the addrecs.
Ops.insert(Ops.end(), AddRecs.begin(), AddRecs.end());
}
/// SplitAddRecs - Flatten a list of add operands, moving addrec start values
/// out to the top level. For example, convert {a + b,+,c} to a, b, {0,+,d}.
/// This helps expose more opportunities for folding parts of the expressions
/// into GEP indices.
///
static void SplitAddRecs(SmallVectorImpl<const SCEV *> &Ops,
const Type *Ty,
ScalarEvolution &SE) {
// Find the addrecs.
SmallVector<const SCEV *, 8> AddRecs;
for (unsigned i = 0, e = Ops.size(); i != e; ++i)
while (const SCEVAddRecExpr *A = dyn_cast<SCEVAddRecExpr>(Ops[i])) {
const SCEV *Start = A->getStart();
if (Start->isZero()) break;
const SCEV *Zero = SE.getIntegerSCEV(0, Ty);
AddRecs.push_back(SE.getAddRecExpr(Zero,
A->getStepRecurrence(SE),
A->getLoop()));
if (const SCEVAddExpr *Add = dyn_cast<SCEVAddExpr>(Start)) {
Ops[i] = Zero;
Ops.insert(Ops.end(), Add->op_begin(), Add->op_end());
e += Add->getNumOperands();
} else {
Ops[i] = Start;
}
}
if (!AddRecs.empty()) {
// Add the addrecs onto the end of the list.
Ops.insert(Ops.end(), AddRecs.begin(), AddRecs.end());
// Resort the operand list, moving any constants to the front.
SimplifyAddOperands(Ops, Ty, SE);
}
}
/// expandAddToGEP - Expand an addition expression with a pointer type into
/// a GEP instead of using ptrtoint+arithmetic+inttoptr. This helps
/// BasicAliasAnalysis and other passes analyze the result. See the rules
/// for getelementptr vs. inttoptr in
/// http://llvm.org/docs/LangRef.html#pointeraliasing
/// for details.
///
/// Design note: The correctness of using getelementptr here depends on
/// ScalarEvolution not recognizing inttoptr and ptrtoint operators, as
/// they may introduce pointer arithmetic which may not be safely converted
/// into getelementptr.
///
/// Design note: It might seem desirable for this function to be more
/// loop-aware. If some of the indices are loop-invariant while others
/// aren't, it might seem desirable to emit multiple GEPs, keeping the
/// loop-invariant portions of the overall computation outside the loop.
/// However, there are a few reasons this is not done here. Hoisting simple
/// arithmetic is a low-level optimization that often isn't very
/// important until late in the optimization process. In fact, passes
/// like InstructionCombining will combine GEPs, even if it means
/// pushing loop-invariant computation down into loops, so even if the
/// GEPs were split here, the work would quickly be undone. The
/// LoopStrengthReduction pass, which is usually run quite late (and
/// after the last InstructionCombining pass), takes care of hoisting
/// loop-invariant portions of expressions, after considering what
/// can be folded using target addressing modes.
///
Value *SCEVExpander::expandAddToGEP(const SCEV *const *op_begin,
const SCEV *const *op_end,
const PointerType *PTy,
const Type *Ty,
Value *V) {
const Type *ElTy = PTy->getElementType();
SmallVector<Value *, 4> GepIndices;
SmallVector<const SCEV *, 8> Ops(op_begin, op_end);
bool AnyNonZeroIndices = false;
// Split AddRecs up into parts as either of the parts may be usable
// without the other.
SplitAddRecs(Ops, Ty, SE);
// Descend down the pointer's type and attempt to convert the other
// operands into GEP indices, at each level. The first index in a GEP
// indexes into the array implied by the pointer operand; the rest of
// the indices index into the element or field type selected by the
// preceding index.
for (;;) {
// If the scale size is not 0, attempt to factor out a scale for
// array indexing.
SmallVector<const SCEV *, 8> ScaledOps;
if (ElTy->isSized()) {
const SCEV *ElSize = SE.getSizeOfExpr(ElTy);
if (!ElSize->isZero()) {
SmallVector<const SCEV *, 8> NewOps;
for (unsigned i = 0, e = Ops.size(); i != e; ++i) {
const SCEV *Op = Ops[i];
const SCEV *Remainder = SE.getIntegerSCEV(0, Ty);
if (FactorOutConstant(Op, Remainder, ElSize, SE, SE.TD)) {
// Op now has ElSize factored out.
ScaledOps.push_back(Op);
if (!Remainder->isZero())
NewOps.push_back(Remainder);
AnyNonZeroIndices = true;
} else {
// The operand was not divisible, so add it to the list of operands
// we'll scan next iteration.
NewOps.push_back(Ops[i]);
}
}
// If we made any changes, update Ops.
if (!ScaledOps.empty()) {
Ops = NewOps;
SimplifyAddOperands(Ops, Ty, SE);
}
}
}
// Record the scaled array index for this level of the type. If
// we didn't find any operands that could be factored, tentatively
// assume that element zero was selected (since the zero offset
// would obviously be folded away).
Value *Scaled = ScaledOps.empty() ?
Constant::getNullValue(Ty) :
expandCodeFor(SE.getAddExpr(ScaledOps), Ty);
GepIndices.push_back(Scaled);
// Collect struct field index operands.
while (const StructType *STy = dyn_cast<StructType>(ElTy)) {
bool FoundFieldNo = false;
// An empty struct has no fields.
if (STy->getNumElements() == 0) break;
if (SE.TD) {
// With TargetData, field offsets are known. See if a constant offset
// falls within any of the struct fields.
if (Ops.empty()) break;
if (const SCEVConstant *C = dyn_cast<SCEVConstant>(Ops[0]))
if (SE.getTypeSizeInBits(C->getType()) <= 64) {
const StructLayout &SL = *SE.TD->getStructLayout(STy);
uint64_t FullOffset = C->getValue()->getZExtValue();
if (FullOffset < SL.getSizeInBytes()) {
unsigned ElIdx = SL.getElementContainingOffset(FullOffset);
GepIndices.push_back(
ConstantInt::get(Type::getInt32Ty(Ty->getContext()), ElIdx));
ElTy = STy->getTypeAtIndex(ElIdx);
Ops[0] =
SE.getConstant(Ty, FullOffset - SL.getElementOffset(ElIdx));
AnyNonZeroIndices = true;
FoundFieldNo = true;
}
}
} else {
// Without TargetData, just check for an offsetof expression of the
// appropriate struct type.
for (unsigned i = 0, e = Ops.size(); i != e; ++i)
if (const SCEVUnknown *U = dyn_cast<SCEVUnknown>(Ops[i])) {
const Type *CTy;
Constant *FieldNo;
if (U->isOffsetOf(CTy, FieldNo) && CTy == STy) {
GepIndices.push_back(FieldNo);
ElTy =
STy->getTypeAtIndex(cast<ConstantInt>(FieldNo)->getZExtValue());
Ops[i] = SE.getConstant(Ty, 0);
AnyNonZeroIndices = true;
FoundFieldNo = true;
break;
}
}
}
// If no struct field offsets were found, tentatively assume that
// field zero was selected (since the zero offset would obviously
// be folded away).
if (!FoundFieldNo) {
ElTy = STy->getTypeAtIndex(0u);
GepIndices.push_back(
Constant::getNullValue(Type::getInt32Ty(Ty->getContext())));
}
}
if (const ArrayType *ATy = dyn_cast<ArrayType>(ElTy))
ElTy = ATy->getElementType();
else
break;
}
// If none of the operands were convertible to proper GEP indices, cast
// the base to i8* and do an ugly getelementptr with that. It's still
// better than ptrtoint+arithmetic+inttoptr at least.
if (!AnyNonZeroIndices) {
// Cast the base to i8*.
V = InsertNoopCastOfTo(V,
Type::getInt8PtrTy(Ty->getContext(), PTy->getAddressSpace()));
// Expand the operands for a plain byte offset.
Value *Idx = expandCodeFor(SE.getAddExpr(Ops), Ty);
// Fold a GEP with constant operands.
if (Constant *CLHS = dyn_cast<Constant>(V))
if (Constant *CRHS = dyn_cast<Constant>(Idx))
return ConstantExpr::getGetElementPtr(CLHS, &CRHS, 1);
// Do a quick scan to see if we have this GEP nearby. If so, reuse it.
unsigned ScanLimit = 6;
BasicBlock::iterator BlockBegin = Builder.GetInsertBlock()->begin();
// Scanning starts from the last instruction before the insertion point.
BasicBlock::iterator IP = Builder.GetInsertPoint();
if (IP != BlockBegin) {
--IP;
for (; ScanLimit; --IP, --ScanLimit) {
// Don't count dbg.value against the ScanLimit, to avoid perturbing the
// generated code.
if (isa<DbgInfoIntrinsic>(IP))
ScanLimit++;
if (IP->getOpcode() == Instruction::GetElementPtr &&
IP->getOperand(0) == V && IP->getOperand(1) == Idx)
return IP;
if (IP == BlockBegin) break;
}
}
// Save the original insertion point so we can restore it when we're done.
BasicBlock *SaveInsertBB = Builder.GetInsertBlock();
BasicBlock::iterator SaveInsertPt = Builder.GetInsertPoint();
// Move the insertion point out of as many loops as we can.
while (const Loop *L = SE.LI->getLoopFor(Builder.GetInsertBlock())) {
if (!L->isLoopInvariant(V) || !L->isLoopInvariant(Idx)) break;
BasicBlock *Preheader = L->getLoopPreheader();
if (!Preheader) break;
// Ok, move up a level.
Builder.SetInsertPoint(Preheader, Preheader->getTerminator());
}
// Emit a GEP.
Value *GEP = Builder.CreateGEP(V, Idx, "uglygep");
rememberInstruction(GEP);
// Restore the original insert point.
if (SaveInsertBB)
restoreInsertPoint(SaveInsertBB, SaveInsertPt);
return GEP;
}
// Save the original insertion point so we can restore it when we're done.
BasicBlock *SaveInsertBB = Builder.GetInsertBlock();
BasicBlock::iterator SaveInsertPt = Builder.GetInsertPoint();
// Move the insertion point out of as many loops as we can.
while (const Loop *L = SE.LI->getLoopFor(Builder.GetInsertBlock())) {
if (!L->isLoopInvariant(V)) break;
bool AnyIndexNotLoopInvariant = false;
for (SmallVectorImpl<Value *>::const_iterator I = GepIndices.begin(),
E = GepIndices.end(); I != E; ++I)
if (!L->isLoopInvariant(*I)) {
AnyIndexNotLoopInvariant = true;
break;
}
if (AnyIndexNotLoopInvariant)
break;
BasicBlock *Preheader = L->getLoopPreheader();
if (!Preheader) break;
// Ok, move up a level.
Builder.SetInsertPoint(Preheader, Preheader->getTerminator());
}
// Insert a pretty getelementptr. Note that this GEP is not marked inbounds,
// because ScalarEvolution may have changed the address arithmetic to
// compute a value which is beyond the end of the allocated object.
Value *Casted = V;
if (V->getType() != PTy)
Casted = InsertNoopCastOfTo(Casted, PTy);
Value *GEP = Builder.CreateGEP(Casted,
GepIndices.begin(),
GepIndices.end(),
"scevgep");
Ops.push_back(SE.getUnknown(GEP));
rememberInstruction(GEP);
// Restore the original insert point.
if (SaveInsertBB)
restoreInsertPoint(SaveInsertBB, SaveInsertPt);
return expand(SE.getAddExpr(Ops));
}
/// isNonConstantNegative - Return true if the specified scev is negated, but
/// not a constant.
static bool isNonConstantNegative(const SCEV *F) {
const SCEVMulExpr *Mul = dyn_cast<SCEVMulExpr>(F);
if (!Mul) return false;
// If there is a constant factor, it will be first.
const SCEVConstant *SC = dyn_cast<SCEVConstant>(Mul->getOperand(0));
if (!SC) return false;
// Return true if the value is negative, this matches things like (-42 * V).
return SC->getValue()->getValue().isNegative();
}
/// PickMostRelevantLoop - Given two loops pick the one that's most relevant for
/// SCEV expansion. If they are nested, this is the most nested. If they are
/// neighboring, pick the later.
static const Loop *PickMostRelevantLoop(const Loop *A, const Loop *B,
DominatorTree &DT) {
if (!A) return B;
if (!B) return A;
if (A->contains(B)) return B;
if (B->contains(A)) return A;
if (DT.dominates(A->getHeader(), B->getHeader())) return B;
if (DT.dominates(B->getHeader(), A->getHeader())) return A;
return A; // Arbitrarily break the tie.
}
/// GetRelevantLoop - Get the most relevant loop associated with the given
/// expression, according to PickMostRelevantLoop.
static const Loop *GetRelevantLoop(const SCEV *S, LoopInfo &LI,
DominatorTree &DT) {
if (isa<SCEVConstant>(S))
return 0;
if (const SCEVUnknown *U = dyn_cast<SCEVUnknown>(S)) {
if (const Instruction *I = dyn_cast<Instruction>(U->getValue()))
return LI.getLoopFor(I->getParent());
return 0;
}
if (const SCEVNAryExpr *N = dyn_cast<SCEVNAryExpr>(S)) {
const Loop *L = 0;
if (const SCEVAddRecExpr *AR = dyn_cast<SCEVAddRecExpr>(S))
L = AR->getLoop();
for (SCEVNAryExpr::op_iterator I = N->op_begin(), E = N->op_end();
I != E; ++I)
L = PickMostRelevantLoop(L, GetRelevantLoop(*I, LI, DT), DT);
return L;
}
if (const SCEVCastExpr *C = dyn_cast<SCEVCastExpr>(S))
return GetRelevantLoop(C->getOperand(), LI, DT);
if (const SCEVUDivExpr *D = dyn_cast<SCEVUDivExpr>(S))
return PickMostRelevantLoop(GetRelevantLoop(D->getLHS(), LI, DT),
GetRelevantLoop(D->getRHS(), LI, DT),
DT);
llvm_unreachable("Unexpected SCEV type!");
}
/// LoopCompare - Compare loops by PickMostRelevantLoop.
class LoopCompare {
DominatorTree &DT;
public:
explicit LoopCompare(DominatorTree &dt) : DT(dt) {}
bool operator()(std::pair<const Loop *, const SCEV *> LHS,
std::pair<const Loop *, const SCEV *> RHS) const {
// Compare loops with PickMostRelevantLoop.
if (LHS.first != RHS.first)
return PickMostRelevantLoop(LHS.first, RHS.first, DT) != LHS.first;
// If one operand is a non-constant negative and the other is not,
// put the non-constant negative on the right so that a sub can
// be used instead of a negate and add.
if (isNonConstantNegative(LHS.second)) {
if (!isNonConstantNegative(RHS.second))
return false;
} else if (isNonConstantNegative(RHS.second))
return true;
// Otherwise they are equivalent according to this comparison.
return false;
}
};
Value *SCEVExpander::visitAddExpr(const SCEVAddExpr *S) {
const Type *Ty = SE.getEffectiveSCEVType(S->getType());
// Collect all the add operands in a loop, along with their associated loops.
// Iterate in reverse so that constants are emitted last, all else equal, and
// so that pointer operands are inserted first, which the code below relies on
// to form more involved GEPs.
SmallVector<std::pair<const Loop *, const SCEV *>, 8> OpsAndLoops;
for (std::reverse_iterator<SCEVAddExpr::op_iterator> I(S->op_end()),
E(S->op_begin()); I != E; ++I)
OpsAndLoops.push_back(std::make_pair(GetRelevantLoop(*I, *SE.LI, *SE.DT),
*I));
// Sort by loop. Use a stable sort so that constants follow non-constants and
// pointer operands precede non-pointer operands.
std::stable_sort(OpsAndLoops.begin(), OpsAndLoops.end(), LoopCompare(*SE.DT));
// Emit instructions to add all the operands. Hoist as much as possible
// out of loops, and form meaningful getelementptrs where possible.
Value *Sum = 0;
for (SmallVectorImpl<std::pair<const Loop *, const SCEV *> >::iterator
I = OpsAndLoops.begin(), E = OpsAndLoops.end(); I != E; ) {
const Loop *CurLoop = I->first;
const SCEV *Op = I->second;
if (!Sum) {
// This is the first operand. Just expand it.
Sum = expand(Op);
++I;
} else if (const PointerType *PTy = dyn_cast<PointerType>(Sum->getType())) {
// The running sum expression is a pointer. Try to form a getelementptr
// at this level with that as the base.
SmallVector<const SCEV *, 4> NewOps;
for (; I != E && I->first == CurLoop; ++I)
NewOps.push_back(I->second);
Sum = expandAddToGEP(NewOps.begin(), NewOps.end(), PTy, Ty, Sum);
} else if (const PointerType *PTy = dyn_cast<PointerType>(Op->getType())) {
// The running sum is an integer, and there's a pointer at this level.
// Try to form a getelementptr. Use a SCEVUnknown so that we don't
// re-analyze the instructions that we just emitted.
SmallVector<const SCEV *, 4> NewOps;
NewOps.push_back(SE.getUnknown(Sum));
for (++I; I != E && I->first == CurLoop; ++I)
NewOps.push_back(I->second);
Sum = expandAddToGEP(NewOps.begin(), NewOps.end(), PTy, Ty, expand(Op));
} else if (isNonConstantNegative(Op)) {
// Instead of doing a negate and add, just do a subtract.
Value *W = expandCodeFor(SE.getNegativeSCEV(Op), Ty);
Sum = InsertNoopCastOfTo(Sum, Ty);
Sum = InsertBinop(Instruction::Sub, Sum, W);
++I;
} else {
// A simple add.
Value *W = expandCodeFor(Op, Ty);
Sum = InsertNoopCastOfTo(Sum, Ty);
// Canonicalize a constant to the RHS.
if (isa<Constant>(Sum)) std::swap(Sum, W);
Sum = InsertBinop(Instruction::Add, Sum, W);
++I;
}
}
return Sum;
}
Value *SCEVExpander::visitMulExpr(const SCEVMulExpr *S) {
const Type *Ty = SE.getEffectiveSCEVType(S->getType());
// Collect all the mul operands in a loop, along with their associated loops.
// Iterate in reverse so that constants are emitted last, all else equal.
SmallVector<std::pair<const Loop *, const SCEV *>, 8> OpsAndLoops;
for (std::reverse_iterator<SCEVMulExpr::op_iterator> I(S->op_end()),
E(S->op_begin()); I != E; ++I)
OpsAndLoops.push_back(std::make_pair(GetRelevantLoop(*I, *SE.LI, *SE.DT),
*I));
// Sort by loop. Use a stable sort so that constants follow non-constants.
std::stable_sort(OpsAndLoops.begin(), OpsAndLoops.end(), LoopCompare(*SE.DT));
// Emit instructions to mul all the operands. Hoist as much as possible
// out of loops.
Value *Prod = 0;
for (SmallVectorImpl<std::pair<const Loop *, const SCEV *> >::iterator
I = OpsAndLoops.begin(), E = OpsAndLoops.end(); I != E; ) {
const SCEV *Op = I->second;
if (!Prod) {
// This is the first operand. Just expand it.
Prod = expand(Op);
++I;
} else if (Op->isAllOnesValue()) {
// Instead of doing a multiply by negative one, just do a negate.
Prod = InsertNoopCastOfTo(Prod, Ty);
Prod = InsertBinop(Instruction::Sub, Constant::getNullValue(Ty), Prod);
++I;
} else {
// A simple mul.
Value *W = expandCodeFor(Op, Ty);
Prod = InsertNoopCastOfTo(Prod, Ty);
// Canonicalize a constant to the RHS.
if (isa<Constant>(Prod)) std::swap(Prod, W);
Prod = InsertBinop(Instruction::Mul, Prod, W);
++I;
}
}
return Prod;
}
Value *SCEVExpander::visitUDivExpr(const SCEVUDivExpr *S) {
const Type *Ty = SE.getEffectiveSCEVType(S->getType());
Value *LHS = expandCodeFor(S->getLHS(), Ty);
if (const SCEVConstant *SC = dyn_cast<SCEVConstant>(S->getRHS())) {
const APInt &RHS = SC->getValue()->getValue();
if (RHS.isPowerOf2())
return InsertBinop(Instruction::LShr, LHS,
ConstantInt::get(Ty, RHS.logBase2()));
}
Value *RHS = expandCodeFor(S->getRHS(), Ty);
return InsertBinop(Instruction::UDiv, LHS, RHS);
}
/// Move parts of Base into Rest to leave Base with the minimal
/// expression that provides a pointer operand suitable for a
/// GEP expansion.
static void ExposePointerBase(const SCEV *&Base, const SCEV *&Rest,
ScalarEvolution &SE) {
while (const SCEVAddRecExpr *A = dyn_cast<SCEVAddRecExpr>(Base)) {
Base = A->getStart();
Rest = SE.getAddExpr(Rest,
SE.getAddRecExpr(SE.getIntegerSCEV(0, A->getType()),
A->getStepRecurrence(SE),
A->getLoop()));
}
if (const SCEVAddExpr *A = dyn_cast<SCEVAddExpr>(Base)) {
Base = A->getOperand(A->getNumOperands()-1);
SmallVector<const SCEV *, 8> NewAddOps(A->op_begin(), A->op_end());
NewAddOps.back() = Rest;
Rest = SE.getAddExpr(NewAddOps);
ExposePointerBase(Base, Rest, SE);
}
}
/// getAddRecExprPHILiterally - Helper for expandAddRecExprLiterally. Expand
/// the base addrec, which is the addrec without any non-loop-dominating
/// values, and return the PHI.
PHINode *
SCEVExpander::getAddRecExprPHILiterally(const SCEVAddRecExpr *Normalized,
const Loop *L,
const Type *ExpandTy,
const Type *IntTy) {
// Reuse a previously-inserted PHI, if present.
for (BasicBlock::iterator I = L->getHeader()->begin();
PHINode *PN = dyn_cast<PHINode>(I); ++I)
if (SE.isSCEVable(PN->getType()) &&
(SE.getEffectiveSCEVType(PN->getType()) ==
SE.getEffectiveSCEVType(Normalized->getType())) &&
SE.getSCEV(PN) == Normalized)
if (BasicBlock *LatchBlock = L->getLoopLatch()) {
Instruction *IncV =
cast<Instruction>(PN->getIncomingValueForBlock(LatchBlock));
// Determine if this is a well-behaved chain of instructions leading
// back to the PHI. It probably will be, if we're scanning an inner
// loop already visited by LSR for example, but it wouldn't have
// to be.
do {
if (IncV->getNumOperands() == 0 || isa<PHINode>(IncV)) {
IncV = 0;
break;
}
// If any of the operands don't dominate the insert position, bail.
// Addrec operands are always loop-invariant, so this can only happen
// if there are instructions which haven't been hoisted.
for (User::op_iterator OI = IncV->op_begin()+1,
OE = IncV->op_end(); OI != OE; ++OI)
if (Instruction *OInst = dyn_cast<Instruction>(OI))
if (!SE.DT->dominates(OInst, IVIncInsertPos)) {
IncV = 0;
break;
}
if (!IncV)
break;
// Advance to the next instruction.
IncV = dyn_cast<Instruction>(IncV->getOperand(0));
if (!IncV)
break;
if (IncV->mayHaveSideEffects()) {
IncV = 0;
break;
}
} while (IncV != PN);
if (IncV) {
// Ok, the add recurrence looks usable.
// Remember this PHI, even in post-inc mode.
InsertedValues.insert(PN);
// Remember the increment.
IncV = cast<Instruction>(PN->getIncomingValueForBlock(LatchBlock));
rememberInstruction(IncV);
if (L == IVIncInsertLoop)
do {
if (SE.DT->dominates(IncV, IVIncInsertPos))
break;
// Make sure the increment is where we want it. But don't move it
// down past a potential existing post-inc user.
IncV->moveBefore(IVIncInsertPos);
IVIncInsertPos = IncV;
IncV = cast<Instruction>(IncV->getOperand(0));
} while (IncV != PN);
return PN;
}
}
// Save the original insertion point so we can restore it when we're done.
BasicBlock *SaveInsertBB = Builder.GetInsertBlock();
BasicBlock::iterator SaveInsertPt = Builder.GetInsertPoint();
// Expand code for the start value.
Value *StartV = expandCodeFor(Normalized->getStart(), ExpandTy,
L->getHeader()->begin());
// Expand code for the step value. Insert instructions right before the
// terminator corresponding to the back-edge. Do this before creating the PHI
// so that PHI reuse code doesn't see an incomplete PHI. If the stride is
// negative, insert a sub instead of an add for the increment (unless it's a
// constant, because subtracts of constants are canonicalized to adds).
const SCEV *Step = Normalized->getStepRecurrence(SE);
bool isPointer = ExpandTy->isPointerTy();
bool isNegative = !isPointer && isNonConstantNegative(Step);
if (isNegative)
Step = SE.getNegativeSCEV(Step);
Value *StepV = expandCodeFor(Step, IntTy, L->getHeader()->begin());
// Create the PHI.
Builder.SetInsertPoint(L->getHeader(), L->getHeader()->begin());
PHINode *PN = Builder.CreatePHI(ExpandTy, "lsr.iv");
rememberInstruction(PN);
// Create the step instructions and populate the PHI.
BasicBlock *Header = L->getHeader();
for (pred_iterator HPI = pred_begin(Header), HPE = pred_end(Header);
HPI != HPE; ++HPI) {
BasicBlock *Pred = *HPI;
// Add a start value.
if (!L->contains(Pred)) {
PN->addIncoming(StartV, Pred);
continue;
}
// Create a step value and add it to the PHI. If IVIncInsertLoop is
// non-null and equal to the addrec's loop, insert the instructions
// at IVIncInsertPos.
Instruction *InsertPos = L == IVIncInsertLoop ?
IVIncInsertPos : Pred->getTerminator();
Builder.SetInsertPoint(InsertPos->getParent(), InsertPos);
Value *IncV;
// If the PHI is a pointer, use a GEP, otherwise use an add or sub.
if (isPointer) {
const PointerType *GEPPtrTy = cast<PointerType>(ExpandTy);
// If the step isn't constant, don't use an implicitly scaled GEP, because
// that would require a multiply inside the loop.
if (!isa<ConstantInt>(StepV))
GEPPtrTy = PointerType::get(Type::getInt1Ty(SE.getContext()),
GEPPtrTy->getAddressSpace());
const SCEV *const StepArray[1] = { SE.getSCEV(StepV) };
IncV = expandAddToGEP(StepArray, StepArray+1, GEPPtrTy, IntTy, PN);
if (IncV->getType() != PN->getType()) {
IncV = Builder.CreateBitCast(IncV, PN->getType(), "tmp");
rememberInstruction(IncV);
}
} else {
IncV = isNegative ?
Builder.CreateSub(PN, StepV, "lsr.iv.next") :
Builder.CreateAdd(PN, StepV, "lsr.iv.next");
rememberInstruction(IncV);
}
PN->addIncoming(IncV, Pred);
}
// Restore the original insert point.
if (SaveInsertBB)
restoreInsertPoint(SaveInsertBB, SaveInsertPt);
// Remember this PHI, even in post-inc mode.
InsertedValues.insert(PN);
return PN;
}
Value *SCEVExpander::expandAddRecExprLiterally(const SCEVAddRecExpr *S) {
const Type *STy = S->getType();
const Type *IntTy = SE.getEffectiveSCEVType(STy);
const Loop *L = S->getLoop();
// Determine a normalized form of this expression, which is the expression
// before any post-inc adjustment is made.
const SCEVAddRecExpr *Normalized = S;
if (PostIncLoops.count(L)) {
PostIncLoopSet Loops;
Loops.insert(L);
Normalized =
cast<SCEVAddRecExpr>(TransformForPostIncUse(Normalize, S, 0, 0,
Loops, SE, *SE.DT));
}
// Strip off any non-loop-dominating component from the addrec start.
const SCEV *Start = Normalized->getStart();
const SCEV *PostLoopOffset = 0;
if (!Start->properlyDominates(L->getHeader(), SE.DT)) {
PostLoopOffset = Start;
Start = SE.getIntegerSCEV(0, Normalized->getType());
Normalized =
cast<SCEVAddRecExpr>(SE.getAddRecExpr(Start,
Normalized->getStepRecurrence(SE),
Normalized->getLoop()));
}
// Strip off any non-loop-dominating component from the addrec step.
const SCEV *Step = Normalized->getStepRecurrence(SE);
const SCEV *PostLoopScale = 0;
if (!Step->hasComputableLoopEvolution(L) &&
!Step->dominates(L->getHeader(), SE.DT)) {
PostLoopScale = Step;
Step = SE.getIntegerSCEV(1, Normalized->getType());
Normalized =
cast<SCEVAddRecExpr>(SE.getAddRecExpr(Start, Step,
Normalized->getLoop()));
}
// Expand the core addrec. If we need post-loop scaling, force it to
// expand to an integer type to avoid the need for additional casting.
const Type *ExpandTy = PostLoopScale ? IntTy : STy;
PHINode *PN = getAddRecExprPHILiterally(Normalized, L, ExpandTy, IntTy);
// Accommodate post-inc mode, if necessary.
Value *Result;
if (!PostIncLoops.count(L))
Result = PN;
else {
// In PostInc mode, use the post-incremented value.
BasicBlock *LatchBlock = L->getLoopLatch();
assert(LatchBlock && "PostInc mode requires a unique loop latch!");
Result = PN->getIncomingValueForBlock(LatchBlock);
}
// Re-apply any non-loop-dominating scale.
if (PostLoopScale) {
Result = InsertNoopCastOfTo(Result, IntTy);
Result = Builder.CreateMul(Result,
expandCodeFor(PostLoopScale, IntTy));
rememberInstruction(Result);
}
// Re-apply any non-loop-dominating offset.
if (PostLoopOffset) {
if (const PointerType *PTy = dyn_cast<PointerType>(ExpandTy)) {
const SCEV *const OffsetArray[1] = { PostLoopOffset };
Result = expandAddToGEP(OffsetArray, OffsetArray+1, PTy, IntTy, Result);
} else {
Result = InsertNoopCastOfTo(Result, IntTy);
Result = Builder.CreateAdd(Result,
expandCodeFor(PostLoopOffset, IntTy));
rememberInstruction(Result);
}
}
return Result;
}
Value *SCEVExpander::visitAddRecExpr(const SCEVAddRecExpr *S) {
if (!CanonicalMode) return expandAddRecExprLiterally(S);
const Type *Ty = SE.getEffectiveSCEVType(S->getType());
const Loop *L = S->getLoop();
// First check for an existing canonical IV in a suitable type.
PHINode *CanonicalIV = 0;
if (PHINode *PN = L->getCanonicalInductionVariable())
if (SE.isSCEVable(PN->getType()) &&
SE.getEffectiveSCEVType(PN->getType())->isIntegerTy() &&
SE.getTypeSizeInBits(PN->getType()) >= SE.getTypeSizeInBits(Ty))
CanonicalIV = PN;
// Rewrite an AddRec in terms of the canonical induction variable, if
// its type is more narrow.
if (CanonicalIV &&
SE.getTypeSizeInBits(CanonicalIV->getType()) >
SE.getTypeSizeInBits(Ty)) {
SmallVector<const SCEV *, 4> NewOps(S->getNumOperands());
for (unsigned i = 0, e = S->getNumOperands(); i != e; ++i)
NewOps[i] = SE.getAnyExtendExpr(S->op_begin()[i], CanonicalIV->getType());
Value *V = expand(SE.getAddRecExpr(NewOps, S->getLoop()));
BasicBlock *SaveInsertBB = Builder.GetInsertBlock();
BasicBlock::iterator SaveInsertPt = Builder.GetInsertPoint();
BasicBlock::iterator NewInsertPt =
llvm::next(BasicBlock::iterator(cast<Instruction>(V)));
while (isa<PHINode>(NewInsertPt)) ++NewInsertPt;
V = expandCodeFor(SE.getTruncateExpr(SE.getUnknown(V), Ty), 0,
NewInsertPt);
restoreInsertPoint(SaveInsertBB, SaveInsertPt);
return V;
}
// {X,+,F} --> X + {0,+,F}
if (!S->getStart()->isZero()) {
SmallVector<const SCEV *, 4> NewOps(S->op_begin(), S->op_end());
NewOps[0] = SE.getIntegerSCEV(0, Ty);
const SCEV *Rest = SE.getAddRecExpr(NewOps, L);
// Turn things like ptrtoint+arithmetic+inttoptr into GEP. See the
// comments on expandAddToGEP for details.
const SCEV *Base = S->getStart();
const SCEV *RestArray[1] = { Rest };
// Dig into the expression to find the pointer base for a GEP.
ExposePointerBase(Base, RestArray[0], SE);
// If we found a pointer, expand the AddRec with a GEP.
if (const PointerType *PTy = dyn_cast<PointerType>(Base->getType())) {
// Make sure the Base isn't something exotic, such as a multiplied
// or divided pointer value. In those cases, the result type isn't
// actually a pointer type.
if (!isa<SCEVMulExpr>(Base) && !isa<SCEVUDivExpr>(Base)) {
Value *StartV = expand(Base);
assert(StartV->getType() == PTy && "Pointer type mismatch for GEP!");
return expandAddToGEP(RestArray, RestArray+1, PTy, Ty, StartV);
}
}
// Just do a normal add. Pre-expand the operands to suppress folding.
return expand(SE.getAddExpr(SE.getUnknown(expand(S->getStart())),
SE.getUnknown(expand(Rest))));
}
// {0,+,1} --> Insert a canonical induction variable into the loop!
if (S->isAffine() &&
S->getOperand(1) == SE.getIntegerSCEV(1, Ty)) {
// If there's a canonical IV, just use it.
if (CanonicalIV) {
assert(Ty == SE.getEffectiveSCEVType(CanonicalIV->getType()) &&
"IVs with types different from the canonical IV should "
"already have been handled!");
return CanonicalIV;
}
// Create and insert the PHI node for the induction variable in the
// specified loop.
BasicBlock *Header = L->getHeader();
PHINode *PN = PHINode::Create(Ty, "indvar", Header->begin());
rememberInstruction(PN);
Constant *One = ConstantInt::get(Ty, 1);
for (pred_iterator HPI = pred_begin(Header), HPE = pred_end(Header);
HPI != HPE; ++HPI)
if (L->contains(*HPI)) {
// Insert a unit add instruction right before the terminator
// corresponding to the back-edge.
Instruction *Add = BinaryOperator::CreateAdd(PN, One, "indvar.next",
(*HPI)->getTerminator());
rememberInstruction(Add);
PN->addIncoming(Add, *HPI);
} else {
PN->addIncoming(Constant::getNullValue(Ty), *HPI);
}
}
// {0,+,F} --> {0,+,1} * F
// Get the canonical induction variable I for this loop.
Value *I = CanonicalIV ?
CanonicalIV :
getOrInsertCanonicalInductionVariable(L, Ty);
// If this is a simple linear addrec, emit it now as a special case.
if (S->isAffine()) // {0,+,F} --> i*F
return
expand(SE.getTruncateOrNoop(
SE.getMulExpr(SE.getUnknown(I),
SE.getNoopOrAnyExtend(S->getOperand(1),
I->getType())),
Ty));
// If this is a chain of recurrences, turn it into a closed form, using the
// folders, then expandCodeFor the closed form. This allows the folders to
// simplify the expression without having to build a bunch of special code
// into this folder.
const SCEV *IH = SE.getUnknown(I); // Get I as a "symbolic" SCEV.
// Promote S up to the canonical IV type, if the cast is foldable.
const SCEV *NewS = S;
const SCEV *Ext = SE.getNoopOrAnyExtend(S, I->getType());
if (isa<SCEVAddRecExpr>(Ext))
NewS = Ext;
const SCEV *V = cast<SCEVAddRecExpr>(NewS)->evaluateAtIteration(IH, SE);
//cerr << "Evaluated: " << *this << "\n to: " << *V << "\n";
// Truncate the result down to the original type, if needed.
const SCEV *T = SE.getTruncateOrNoop(V, Ty);
return expand(T);
}
Value *SCEVExpander::visitTruncateExpr(const SCEVTruncateExpr *S) {
const Type *Ty = SE.getEffectiveSCEVType(S->getType());
Value *V = expandCodeFor(S->getOperand(),
SE.getEffectiveSCEVType(S->getOperand()->getType()));
Value *I = Builder.CreateTrunc(V, Ty, "tmp");
rememberInstruction(I);
return I;
}
Value *SCEVExpander::visitZeroExtendExpr(const SCEVZeroExtendExpr *S) {
const Type *Ty = SE.getEffectiveSCEVType(S->getType());
Value *V = expandCodeFor(S->getOperand(),
SE.getEffectiveSCEVType(S->getOperand()->getType()));
Value *I = Builder.CreateZExt(V, Ty, "tmp");
rememberInstruction(I);
return I;
}
Value *SCEVExpander::visitSignExtendExpr(const SCEVSignExtendExpr *S) {
const Type *Ty = SE.getEffectiveSCEVType(S->getType());
Value *V = expandCodeFor(S->getOperand(),
SE.getEffectiveSCEVType(S->getOperand()->getType()));
Value *I = Builder.CreateSExt(V, Ty, "tmp");
rememberInstruction(I);
return I;
}
Value *SCEVExpander::visitSMaxExpr(const SCEVSMaxExpr *S) {
Value *LHS = expand(S->getOperand(S->getNumOperands()-1));
const Type *Ty = LHS->getType();
for (int i = S->getNumOperands()-2; i >= 0; --i) {
// In the case of mixed integer and pointer types, do the
// rest of the comparisons as integer.
if (S->getOperand(i)->getType() != Ty) {
Ty = SE.getEffectiveSCEVType(Ty);
LHS = InsertNoopCastOfTo(LHS, Ty);
}
Value *RHS = expandCodeFor(S->getOperand(i), Ty);
Value *ICmp = Builder.CreateICmpSGT(LHS, RHS, "tmp");
rememberInstruction(ICmp);
Value *Sel = Builder.CreateSelect(ICmp, LHS, RHS, "smax");
rememberInstruction(Sel);
LHS = Sel;
}
// In the case of mixed integer and pointer types, cast the
// final result back to the pointer type.
if (LHS->getType() != S->getType())
LHS = InsertNoopCastOfTo(LHS, S->getType());
return LHS;
}
Value *SCEVExpander::visitUMaxExpr(const SCEVUMaxExpr *S) {
Value *LHS = expand(S->getOperand(S->getNumOperands()-1));
const Type *Ty = LHS->getType();
for (int i = S->getNumOperands()-2; i >= 0; --i) {
// In the case of mixed integer and pointer types, do the
// rest of the comparisons as integer.
if (S->getOperand(i)->getType() != Ty) {
Ty = SE.getEffectiveSCEVType(Ty);
LHS = InsertNoopCastOfTo(LHS, Ty);
}
Value *RHS = expandCodeFor(S->getOperand(i), Ty);
Value *ICmp = Builder.CreateICmpUGT(LHS, RHS, "tmp");
rememberInstruction(ICmp);
Value *Sel = Builder.CreateSelect(ICmp, LHS, RHS, "umax");
rememberInstruction(Sel);
LHS = Sel;
}
// In the case of mixed integer and pointer types, cast the
// final result back to the pointer type.
if (LHS->getType() != S->getType())
LHS = InsertNoopCastOfTo(LHS, S->getType());
return LHS;
}
Value *SCEVExpander::expandCodeFor(const SCEV *SH, const Type *Ty,
Instruction *I) {
BasicBlock::iterator IP = I;
while (isInsertedInstruction(IP) || isa<DbgInfoIntrinsic>(IP))
++IP;
Builder.SetInsertPoint(IP->getParent(), IP);
return expandCodeFor(SH, Ty);
}
Value *SCEVExpander::expandCodeFor(const SCEV *SH, const Type *Ty) {
// Expand the code for this SCEV.
Value *V = expand(SH);
if (Ty) {
assert(SE.getTypeSizeInBits(Ty) == SE.getTypeSizeInBits(SH->getType()) &&
"non-trivial casts should be done with the SCEVs directly!");
V = InsertNoopCastOfTo(V, Ty);
}
return V;
}
Value *SCEVExpander::expand(const SCEV *S) {
// Compute an insertion point for this SCEV object. Hoist the instructions
// as far out in the loop nest as possible.
Instruction *InsertPt = Builder.GetInsertPoint();
for (Loop *L = SE.LI->getLoopFor(Builder.GetInsertBlock()); ;
L = L->getParentLoop())
if (S->isLoopInvariant(L)) {
if (!L) break;
if (BasicBlock *Preheader = L->getLoopPreheader())
InsertPt = Preheader->getTerminator();
} else {
// If the SCEV is computable at this level, insert it into the header
// after the PHIs (and after any other instructions that we've inserted
// there) so that it is guaranteed to dominate any user inside the loop.
if (L && S->hasComputableLoopEvolution(L) && !PostIncLoops.count(L))
InsertPt = L->getHeader()->getFirstNonPHI();
while (isInsertedInstruction(InsertPt) || isa<DbgInfoIntrinsic>(InsertPt))
InsertPt = llvm::next(BasicBlock::iterator(InsertPt));
break;
}
// Check to see if we already expanded this here.
std::map<std::pair<const SCEV *, Instruction *>,
AssertingVH<Value> >::iterator I =
InsertedExpressions.find(std::make_pair(S, InsertPt));
if (I != InsertedExpressions.end())
return I->second;
BasicBlock *SaveInsertBB = Builder.GetInsertBlock();
BasicBlock::iterator SaveInsertPt = Builder.GetInsertPoint();
Builder.SetInsertPoint(InsertPt->getParent(), InsertPt);
// Expand the expression into instructions.
Value *V = visit(S);
// Remember the expanded value for this SCEV at this location.
if (PostIncLoops.empty())
InsertedExpressions[std::make_pair(S, InsertPt)] = V;
restoreInsertPoint(SaveInsertBB, SaveInsertPt);
return V;
}
void SCEVExpander::rememberInstruction(Value *I) {
if (PostIncLoops.empty())
InsertedValues.insert(I);
// If we just claimed an existing instruction and that instruction had
// been the insert point, adjust the insert point forward so that
// subsequently inserted code will be dominated.
if (Builder.GetInsertPoint() == I) {
BasicBlock::iterator It = cast<Instruction>(I);
do { ++It; } while (isInsertedInstruction(It) ||
isa<DbgInfoIntrinsic>(It));
Builder.SetInsertPoint(Builder.GetInsertBlock(), It);
}
}
void SCEVExpander::restoreInsertPoint(BasicBlock *BB, BasicBlock::iterator I) {
// If we acquired more instructions since the old insert point was saved,
// advance past them.
while (isInsertedInstruction(I) || isa<DbgInfoIntrinsic>(I)) ++I;
Builder.SetInsertPoint(BB, I);
}
/// getOrInsertCanonicalInductionVariable - This method returns the
/// canonical induction variable of the specified type for the specified
/// loop (inserting one if there is none). A canonical induction variable
/// starts at zero and steps by one on each iteration.
Value *
SCEVExpander::getOrInsertCanonicalInductionVariable(const Loop *L,
const Type *Ty) {
assert(Ty->isIntegerTy() && "Can only insert integer induction variables!");
const SCEV *H = SE.getAddRecExpr(SE.getIntegerSCEV(0, Ty),
SE.getIntegerSCEV(1, Ty), L);
BasicBlock *SaveInsertBB = Builder.GetInsertBlock();
BasicBlock::iterator SaveInsertPt = Builder.GetInsertPoint();
Value *V = expandCodeFor(H, 0, L->getHeader()->begin());
if (SaveInsertBB)
restoreInsertPoint(SaveInsertBB, SaveInsertPt);
return V;
}
|