aboutsummaryrefslogtreecommitdiffstats
path: root/contrib/libs/llvm14/lib/Transforms/InstCombine/InstCombineShifts.cpp
blob: 17f0c5c4cff0ee428096c3844013f844f6e20474 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
//===- InstCombineShifts.cpp ----------------------------------------------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
//
// This file implements the visitShl, visitLShr, and visitAShr functions.
//
//===----------------------------------------------------------------------===//

#include "InstCombineInternal.h"
#include "llvm/Analysis/ConstantFolding.h"
#include "llvm/Analysis/InstructionSimplify.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/IR/PatternMatch.h"
#include "llvm/Transforms/InstCombine/InstCombiner.h"
using namespace llvm;
using namespace PatternMatch;

#define DEBUG_TYPE "instcombine"

bool canTryToConstantAddTwoShiftAmounts(Value *Sh0, Value *ShAmt0, Value *Sh1,
                                        Value *ShAmt1) {
  // We have two shift amounts from two different shifts. The types of those
  // shift amounts may not match. If that's the case let's bailout now..
  if (ShAmt0->getType() != ShAmt1->getType())
    return false;

  // As input, we have the following pattern:
  //   Sh0 (Sh1 X, Q), K
  // We want to rewrite that as:
  //   Sh x, (Q+K)  iff (Q+K) u< bitwidth(x)
  // While we know that originally (Q+K) would not overflow
  // (because  2 * (N-1) u<= iN -1), we have looked past extensions of
  // shift amounts. so it may now overflow in smaller bitwidth.
  // To ensure that does not happen, we need to ensure that the total maximal
  // shift amount is still representable in that smaller bit width.
  unsigned MaximalPossibleTotalShiftAmount =
      (Sh0->getType()->getScalarSizeInBits() - 1) +
      (Sh1->getType()->getScalarSizeInBits() - 1);
  APInt MaximalRepresentableShiftAmount =
      APInt::getAllOnes(ShAmt0->getType()->getScalarSizeInBits());
  return MaximalRepresentableShiftAmount.uge(MaximalPossibleTotalShiftAmount);
}

// Given pattern:
//   (x shiftopcode Q) shiftopcode K
// we should rewrite it as
//   x shiftopcode (Q+K)  iff (Q+K) u< bitwidth(x) and
//
// This is valid for any shift, but they must be identical, and we must be
// careful in case we have (zext(Q)+zext(K)) and look past extensions,
// (Q+K) must not overflow or else (Q+K) u< bitwidth(x) is bogus.
//
// AnalyzeForSignBitExtraction indicates that we will only analyze whether this
// pattern has any 2 right-shifts that sum to 1 less than original bit width.
Value *InstCombinerImpl::reassociateShiftAmtsOfTwoSameDirectionShifts(
    BinaryOperator *Sh0, const SimplifyQuery &SQ,
    bool AnalyzeForSignBitExtraction) {
  // Look for a shift of some instruction, ignore zext of shift amount if any.
  Instruction *Sh0Op0;
  Value *ShAmt0;
  if (!match(Sh0,
             m_Shift(m_Instruction(Sh0Op0), m_ZExtOrSelf(m_Value(ShAmt0)))))
    return nullptr;

  // If there is a truncation between the two shifts, we must make note of it
  // and look through it. The truncation imposes additional constraints on the
  // transform.
  Instruction *Sh1;
  Value *Trunc = nullptr;
  match(Sh0Op0,
        m_CombineOr(m_CombineAnd(m_Trunc(m_Instruction(Sh1)), m_Value(Trunc)),
                    m_Instruction(Sh1)));

  // Inner shift: (x shiftopcode ShAmt1)
  // Like with other shift, ignore zext of shift amount if any.
  Value *X, *ShAmt1;
  if (!match(Sh1, m_Shift(m_Value(X), m_ZExtOrSelf(m_Value(ShAmt1)))))
    return nullptr;

  // Verify that it would be safe to try to add those two shift amounts.
  if (!canTryToConstantAddTwoShiftAmounts(Sh0, ShAmt0, Sh1, ShAmt1))
    return nullptr;

  // We are only looking for signbit extraction if we have two right shifts.
  bool HadTwoRightShifts = match(Sh0, m_Shr(m_Value(), m_Value())) &&
                           match(Sh1, m_Shr(m_Value(), m_Value()));
  // ... and if it's not two right-shifts, we know the answer already.
  if (AnalyzeForSignBitExtraction && !HadTwoRightShifts)
    return nullptr;

  // The shift opcodes must be identical, unless we are just checking whether
  // this pattern can be interpreted as a sign-bit-extraction.
  Instruction::BinaryOps ShiftOpcode = Sh0->getOpcode();
  bool IdenticalShOpcodes = Sh0->getOpcode() == Sh1->getOpcode();
  if (!IdenticalShOpcodes && !AnalyzeForSignBitExtraction)
    return nullptr;

  // If we saw truncation, we'll need to produce extra instruction,
  // and for that one of the operands of the shift must be one-use,
  // unless of course we don't actually plan to produce any instructions here.
  if (Trunc && !AnalyzeForSignBitExtraction &&
      !match(Sh0, m_c_BinOp(m_OneUse(m_Value()), m_Value())))
    return nullptr;

  // Can we fold (ShAmt0+ShAmt1) ?
  auto *NewShAmt = dyn_cast_or_null<Constant>(
      SimplifyAddInst(ShAmt0, ShAmt1, /*isNSW=*/false, /*isNUW=*/false,
                      SQ.getWithInstruction(Sh0)));
  if (!NewShAmt)
    return nullptr; // Did not simplify.
  unsigned NewShAmtBitWidth = NewShAmt->getType()->getScalarSizeInBits();
  unsigned XBitWidth = X->getType()->getScalarSizeInBits();
  // Is the new shift amount smaller than the bit width of inner/new shift?
  if (!match(NewShAmt, m_SpecificInt_ICMP(ICmpInst::Predicate::ICMP_ULT,
                                          APInt(NewShAmtBitWidth, XBitWidth))))
    return nullptr; // FIXME: could perform constant-folding.

  // If there was a truncation, and we have a right-shift, we can only fold if
  // we are left with the original sign bit. Likewise, if we were just checking
  // that this is a sighbit extraction, this is the place to check it.
  // FIXME: zero shift amount is also legal here, but we can't *easily* check
  // more than one predicate so it's not really worth it.
  if (HadTwoRightShifts && (Trunc || AnalyzeForSignBitExtraction)) {
    // If it's not a sign bit extraction, then we're done.
    if (!match(NewShAmt,
               m_SpecificInt_ICMP(ICmpInst::Predicate::ICMP_EQ,
                                  APInt(NewShAmtBitWidth, XBitWidth - 1))))
      return nullptr;
    // If it is, and that was the question, return the base value.
    if (AnalyzeForSignBitExtraction)
      return X;
  }

  assert(IdenticalShOpcodes && "Should not get here with different shifts.");

  // All good, we can do this fold.
  NewShAmt = ConstantExpr::getZExtOrBitCast(NewShAmt, X->getType());

  BinaryOperator *NewShift = BinaryOperator::Create(ShiftOpcode, X, NewShAmt);

  // The flags can only be propagated if there wasn't a trunc.
  if (!Trunc) {
    // If the pattern did not involve trunc, and both of the original shifts
    // had the same flag set, preserve the flag.
    if (ShiftOpcode == Instruction::BinaryOps::Shl) {
      NewShift->setHasNoUnsignedWrap(Sh0->hasNoUnsignedWrap() &&
                                     Sh1->hasNoUnsignedWrap());
      NewShift->setHasNoSignedWrap(Sh0->hasNoSignedWrap() &&
                                   Sh1->hasNoSignedWrap());
    } else {
      NewShift->setIsExact(Sh0->isExact() && Sh1->isExact());
    }
  }

  Instruction *Ret = NewShift;
  if (Trunc) {
    Builder.Insert(NewShift);
    Ret = CastInst::Create(Instruction::Trunc, NewShift, Sh0->getType());
  }

  return Ret;
}

// If we have some pattern that leaves only some low bits set, and then performs
// left-shift of those bits, if none of the bits that are left after the final
// shift are modified by the mask, we can omit the mask.
//
// There are many variants to this pattern:
//   a)  (x & ((1 << MaskShAmt) - 1)) << ShiftShAmt
//   b)  (x & (~(-1 << MaskShAmt))) << ShiftShAmt
//   c)  (x & (-1 l>> MaskShAmt)) << ShiftShAmt
//   d)  (x & ((-1 << MaskShAmt) l>> MaskShAmt)) << ShiftShAmt
//   e)  ((x << MaskShAmt) l>> MaskShAmt) << ShiftShAmt
//   f)  ((x << MaskShAmt) a>> MaskShAmt) << ShiftShAmt
// All these patterns can be simplified to just:
//   x << ShiftShAmt
// iff:
//   a,b)     (MaskShAmt+ShiftShAmt) u>= bitwidth(x)
//   c,d,e,f) (ShiftShAmt-MaskShAmt) s>= 0 (i.e. ShiftShAmt u>= MaskShAmt)
static Instruction *
dropRedundantMaskingOfLeftShiftInput(BinaryOperator *OuterShift,
                                     const SimplifyQuery &Q,
                                     InstCombiner::BuilderTy &Builder) {
  assert(OuterShift->getOpcode() == Instruction::BinaryOps::Shl &&
         "The input must be 'shl'!");

  Value *Masked, *ShiftShAmt;
  match(OuterShift,
        m_Shift(m_Value(Masked), m_ZExtOrSelf(m_Value(ShiftShAmt))));

  // *If* there is a truncation between an outer shift and a possibly-mask,
  // then said truncation *must* be one-use, else we can't perform the fold.
  Value *Trunc;
  if (match(Masked, m_CombineAnd(m_Trunc(m_Value(Masked)), m_Value(Trunc))) &&
      !Trunc->hasOneUse())
    return nullptr;

  Type *NarrowestTy = OuterShift->getType();
  Type *WidestTy = Masked->getType();
  bool HadTrunc = WidestTy != NarrowestTy;

  // The mask must be computed in a type twice as wide to ensure
  // that no bits are lost if the sum-of-shifts is wider than the base type.
  Type *ExtendedTy = WidestTy->getExtendedType();

  Value *MaskShAmt;

  // ((1 << MaskShAmt) - 1)
  auto MaskA = m_Add(m_Shl(m_One(), m_Value(MaskShAmt)), m_AllOnes());
  // (~(-1 << maskNbits))
  auto MaskB = m_Xor(m_Shl(m_AllOnes(), m_Value(MaskShAmt)), m_AllOnes());
  // (-1 l>> MaskShAmt)
  auto MaskC = m_LShr(m_AllOnes(), m_Value(MaskShAmt));
  // ((-1 << MaskShAmt) l>> MaskShAmt)
  auto MaskD =
      m_LShr(m_Shl(m_AllOnes(), m_Value(MaskShAmt)), m_Deferred(MaskShAmt));

  Value *X;
  Constant *NewMask;

  if (match(Masked, m_c_And(m_CombineOr(MaskA, MaskB), m_Value(X)))) {
    // Peek through an optional zext of the shift amount.
    match(MaskShAmt, m_ZExtOrSelf(m_Value(MaskShAmt)));

    // Verify that it would be safe to try to add those two shift amounts.
    if (!canTryToConstantAddTwoShiftAmounts(OuterShift, ShiftShAmt, Masked,
                                            MaskShAmt))
      return nullptr;

    // Can we simplify (MaskShAmt+ShiftShAmt) ?
    auto *SumOfShAmts = dyn_cast_or_null<Constant>(SimplifyAddInst(
        MaskShAmt, ShiftShAmt, /*IsNSW=*/false, /*IsNUW=*/false, Q));
    if (!SumOfShAmts)
      return nullptr; // Did not simplify.
    // In this pattern SumOfShAmts correlates with the number of low bits
    // that shall remain in the root value (OuterShift).

    // An extend of an undef value becomes zero because the high bits are never
    // completely unknown. Replace the `undef` shift amounts with final
    // shift bitwidth to ensure that the value remains undef when creating the
    // subsequent shift op.
    SumOfShAmts = Constant::replaceUndefsWith(
        SumOfShAmts, ConstantInt::get(SumOfShAmts->getType()->getScalarType(),
                                      ExtendedTy->getScalarSizeInBits()));
    auto *ExtendedSumOfShAmts = ConstantExpr::getZExt(SumOfShAmts, ExtendedTy);
    // And compute the mask as usual: ~(-1 << (SumOfShAmts))
    auto *ExtendedAllOnes = ConstantExpr::getAllOnesValue(ExtendedTy);
    auto *ExtendedInvertedMask =
        ConstantExpr::getShl(ExtendedAllOnes, ExtendedSumOfShAmts);
    NewMask = ConstantExpr::getNot(ExtendedInvertedMask);
  } else if (match(Masked, m_c_And(m_CombineOr(MaskC, MaskD), m_Value(X))) ||
             match(Masked, m_Shr(m_Shl(m_Value(X), m_Value(MaskShAmt)),
                                 m_Deferred(MaskShAmt)))) {
    // Peek through an optional zext of the shift amount.
    match(MaskShAmt, m_ZExtOrSelf(m_Value(MaskShAmt)));

    // Verify that it would be safe to try to add those two shift amounts.
    if (!canTryToConstantAddTwoShiftAmounts(OuterShift, ShiftShAmt, Masked,
                                            MaskShAmt))
      return nullptr;

    // Can we simplify (ShiftShAmt-MaskShAmt) ?
    auto *ShAmtsDiff = dyn_cast_or_null<Constant>(SimplifySubInst(
        ShiftShAmt, MaskShAmt, /*IsNSW=*/false, /*IsNUW=*/false, Q));
    if (!ShAmtsDiff)
      return nullptr; // Did not simplify.
    // In this pattern ShAmtsDiff correlates with the number of high bits that
    // shall be unset in the root value (OuterShift).

    // An extend of an undef value becomes zero because the high bits are never
    // completely unknown. Replace the `undef` shift amounts with negated
    // bitwidth of innermost shift to ensure that the value remains undef when
    // creating the subsequent shift op.
    unsigned WidestTyBitWidth = WidestTy->getScalarSizeInBits();
    ShAmtsDiff = Constant::replaceUndefsWith(
        ShAmtsDiff, ConstantInt::get(ShAmtsDiff->getType()->getScalarType(),
                                     -WidestTyBitWidth));
    auto *ExtendedNumHighBitsToClear = ConstantExpr::getZExt(
        ConstantExpr::getSub(ConstantInt::get(ShAmtsDiff->getType(),
                                              WidestTyBitWidth,
                                              /*isSigned=*/false),
                             ShAmtsDiff),
        ExtendedTy);
    // And compute the mask as usual: (-1 l>> (NumHighBitsToClear))
    auto *ExtendedAllOnes = ConstantExpr::getAllOnesValue(ExtendedTy);
    NewMask =
        ConstantExpr::getLShr(ExtendedAllOnes, ExtendedNumHighBitsToClear);
  } else
    return nullptr; // Don't know anything about this pattern.

  NewMask = ConstantExpr::getTrunc(NewMask, NarrowestTy);

  // Does this mask has any unset bits? If not then we can just not apply it.
  bool NeedMask = !match(NewMask, m_AllOnes());

  // If we need to apply a mask, there are several more restrictions we have.
  if (NeedMask) {
    // The old masking instruction must go away.
    if (!Masked->hasOneUse())
      return nullptr;
    // The original "masking" instruction must not have been`ashr`.
    if (match(Masked, m_AShr(m_Value(), m_Value())))
      return nullptr;
  }

  // If we need to apply truncation, let's do it first, since we can.
  // We have already ensured that the old truncation will go away.
  if (HadTrunc)
    X = Builder.CreateTrunc(X, NarrowestTy);

  // No 'NUW'/'NSW'! We no longer know that we won't shift-out non-0 bits.
  // We didn't change the Type of this outermost shift, so we can just do it.
  auto *NewShift = BinaryOperator::Create(OuterShift->getOpcode(), X,
                                          OuterShift->getOperand(1));
  if (!NeedMask)
    return NewShift;

  Builder.Insert(NewShift);
  return BinaryOperator::Create(Instruction::And, NewShift, NewMask);
}

/// If we have a shift-by-constant of a bitwise logic op that itself has a
/// shift-by-constant operand with identical opcode, we may be able to convert
/// that into 2 independent shifts followed by the logic op. This eliminates a
/// a use of an intermediate value (reduces dependency chain).
static Instruction *foldShiftOfShiftedLogic(BinaryOperator &I,
                                            InstCombiner::BuilderTy &Builder) {
  assert(I.isShift() && "Expected a shift as input");
  auto *LogicInst = dyn_cast<BinaryOperator>(I.getOperand(0));
  if (!LogicInst || !LogicInst->isBitwiseLogicOp() || !LogicInst->hasOneUse())
    return nullptr;

  Constant *C0, *C1;
  if (!match(I.getOperand(1), m_Constant(C1)))
    return nullptr;

  Instruction::BinaryOps ShiftOpcode = I.getOpcode();
  Type *Ty = I.getType();

  // Find a matching one-use shift by constant. The fold is not valid if the sum
  // of the shift values equals or exceeds bitwidth.
  // TODO: Remove the one-use check if the other logic operand (Y) is constant.
  Value *X, *Y;
  auto matchFirstShift = [&](Value *V) {
    APInt Threshold(Ty->getScalarSizeInBits(), Ty->getScalarSizeInBits());
    return match(V, m_BinOp(ShiftOpcode, m_Value(), m_Value())) &&
           match(V, m_OneUse(m_Shift(m_Value(X), m_Constant(C0)))) &&
           match(ConstantExpr::getAdd(C0, C1),
                 m_SpecificInt_ICMP(ICmpInst::ICMP_ULT, Threshold));
  };

  // Logic ops are commutative, so check each operand for a match.
  if (matchFirstShift(LogicInst->getOperand(0)))
    Y = LogicInst->getOperand(1);
  else if (matchFirstShift(LogicInst->getOperand(1)))
    Y = LogicInst->getOperand(0);
  else
    return nullptr;

  // shift (logic (shift X, C0), Y), C1 -> logic (shift X, C0+C1), (shift Y, C1)
  Constant *ShiftSumC = ConstantExpr::getAdd(C0, C1);
  Value *NewShift1 = Builder.CreateBinOp(ShiftOpcode, X, ShiftSumC);
  Value *NewShift2 = Builder.CreateBinOp(ShiftOpcode, Y, I.getOperand(1));
  return BinaryOperator::Create(LogicInst->getOpcode(), NewShift1, NewShift2);
}

Instruction *InstCombinerImpl::commonShiftTransforms(BinaryOperator &I) {
  if (Instruction *Phi = foldBinopWithPhiOperands(I))
    return Phi;

  Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
  assert(Op0->getType() == Op1->getType());

  // If the shift amount is a one-use `sext`, we can demote it to `zext`.
  Value *Y;
  if (match(Op1, m_OneUse(m_SExt(m_Value(Y))))) {
    Value *NewExt = Builder.CreateZExt(Y, I.getType(), Op1->getName());
    return BinaryOperator::Create(I.getOpcode(), Op0, NewExt);
  }

  // See if we can fold away this shift.
  if (SimplifyDemandedInstructionBits(I))
    return &I;

  // Try to fold constant and into select arguments.
  if (isa<Constant>(Op0))
    if (SelectInst *SI = dyn_cast<SelectInst>(Op1))
      if (Instruction *R = FoldOpIntoSelect(I, SI))
        return R;

  if (Constant *CUI = dyn_cast<Constant>(Op1))
    if (Instruction *Res = FoldShiftByConstant(Op0, CUI, I))
      return Res;

  if (auto *NewShift = cast_or_null<Instruction>(
          reassociateShiftAmtsOfTwoSameDirectionShifts(&I, SQ)))
    return NewShift;

  // (C1 shift (A add C2)) -> (C1 shift C2) shift A)
  // iff A and C2 are both positive.
  Value *A;
  Constant *C;
  if (match(Op0, m_Constant()) && match(Op1, m_Add(m_Value(A), m_Constant(C))))
    if (isKnownNonNegative(A, DL, 0, &AC, &I, &DT) &&
        isKnownNonNegative(C, DL, 0, &AC, &I, &DT))
      return BinaryOperator::Create(
          I.getOpcode(), Builder.CreateBinOp(I.getOpcode(), Op0, C), A);

  // X shift (A srem C) -> X shift (A and (C - 1)) iff C is a power of 2.
  // Because shifts by negative values (which could occur if A were negative)
  // are undefined.
  if (Op1->hasOneUse() && match(Op1, m_SRem(m_Value(A), m_Constant(C))) &&
      match(C, m_Power2())) {
    // FIXME: Should this get moved into SimplifyDemandedBits by saying we don't
    // demand the sign bit (and many others) here??
    Constant *Mask = ConstantExpr::getSub(C, ConstantInt::get(I.getType(), 1));
    Value *Rem = Builder.CreateAnd(A, Mask, Op1->getName());
    return replaceOperand(I, 1, Rem);
  }

  if (Instruction *Logic = foldShiftOfShiftedLogic(I, Builder))
    return Logic;

  return nullptr;
}

/// Return true if we can simplify two logical (either left or right) shifts
/// that have constant shift amounts: OuterShift (InnerShift X, C1), C2.
static bool canEvaluateShiftedShift(unsigned OuterShAmt, bool IsOuterShl,
                                    Instruction *InnerShift,
                                    InstCombinerImpl &IC, Instruction *CxtI) {
  assert(InnerShift->isLogicalShift() && "Unexpected instruction type");

  // We need constant scalar or constant splat shifts.
  const APInt *InnerShiftConst;
  if (!match(InnerShift->getOperand(1), m_APInt(InnerShiftConst)))
    return false;

  // Two logical shifts in the same direction:
  // shl (shl X, C1), C2 -->  shl X, C1 + C2
  // lshr (lshr X, C1), C2 --> lshr X, C1 + C2
  bool IsInnerShl = InnerShift->getOpcode() == Instruction::Shl;
  if (IsInnerShl == IsOuterShl)
    return true;

  // Equal shift amounts in opposite directions become bitwise 'and':
  // lshr (shl X, C), C --> and X, C'
  // shl (lshr X, C), C --> and X, C'
  if (*InnerShiftConst == OuterShAmt)
    return true;

  // If the 2nd shift is bigger than the 1st, we can fold:
  // lshr (shl X, C1), C2 -->  and (shl X, C1 - C2), C3
  // shl (lshr X, C1), C2 --> and (lshr X, C1 - C2), C3
  // but it isn't profitable unless we know the and'd out bits are already zero.
  // Also, check that the inner shift is valid (less than the type width) or
  // we'll crash trying to produce the bit mask for the 'and'.
  unsigned TypeWidth = InnerShift->getType()->getScalarSizeInBits();
  if (InnerShiftConst->ugt(OuterShAmt) && InnerShiftConst->ult(TypeWidth)) {
    unsigned InnerShAmt = InnerShiftConst->getZExtValue();
    unsigned MaskShift =
        IsInnerShl ? TypeWidth - InnerShAmt : InnerShAmt - OuterShAmt;
    APInt Mask = APInt::getLowBitsSet(TypeWidth, OuterShAmt) << MaskShift;
    if (IC.MaskedValueIsZero(InnerShift->getOperand(0), Mask, 0, CxtI))
      return true;
  }

  return false;
}

/// See if we can compute the specified value, but shifted logically to the left
/// or right by some number of bits. This should return true if the expression
/// can be computed for the same cost as the current expression tree. This is
/// used to eliminate extraneous shifting from things like:
///      %C = shl i128 %A, 64
///      %D = shl i128 %B, 96
///      %E = or i128 %C, %D
///      %F = lshr i128 %E, 64
/// where the client will ask if E can be computed shifted right by 64-bits. If
/// this succeeds, getShiftedValue() will be called to produce the value.
static bool canEvaluateShifted(Value *V, unsigned NumBits, bool IsLeftShift,
                               InstCombinerImpl &IC, Instruction *CxtI) {
  // We can always evaluate constants shifted.
  if (isa<Constant>(V))
    return true;

  Instruction *I = dyn_cast<Instruction>(V);
  if (!I) return false;

  // We can't mutate something that has multiple uses: doing so would
  // require duplicating the instruction in general, which isn't profitable.
  if (!I->hasOneUse()) return false;

  switch (I->getOpcode()) {
  default: return false;
  case Instruction::And:
  case Instruction::Or:
  case Instruction::Xor:
    // Bitwise operators can all arbitrarily be arbitrarily evaluated shifted.
    return canEvaluateShifted(I->getOperand(0), NumBits, IsLeftShift, IC, I) &&
           canEvaluateShifted(I->getOperand(1), NumBits, IsLeftShift, IC, I);

  case Instruction::Shl:
  case Instruction::LShr:
    return canEvaluateShiftedShift(NumBits, IsLeftShift, I, IC, CxtI);

  case Instruction::Select: {
    SelectInst *SI = cast<SelectInst>(I);
    Value *TrueVal = SI->getTrueValue();
    Value *FalseVal = SI->getFalseValue();
    return canEvaluateShifted(TrueVal, NumBits, IsLeftShift, IC, SI) &&
           canEvaluateShifted(FalseVal, NumBits, IsLeftShift, IC, SI);
  }
  case Instruction::PHI: {
    // We can change a phi if we can change all operands.  Note that we never
    // get into trouble with cyclic PHIs here because we only consider
    // instructions with a single use.
    PHINode *PN = cast<PHINode>(I);
    for (Value *IncValue : PN->incoming_values())
      if (!canEvaluateShifted(IncValue, NumBits, IsLeftShift, IC, PN))
        return false;
    return true;
  }
  }
}

/// Fold OuterShift (InnerShift X, C1), C2.
/// See canEvaluateShiftedShift() for the constraints on these instructions.
static Value *foldShiftedShift(BinaryOperator *InnerShift, unsigned OuterShAmt,
                               bool IsOuterShl,
                               InstCombiner::BuilderTy &Builder) {
  bool IsInnerShl = InnerShift->getOpcode() == Instruction::Shl;
  Type *ShType = InnerShift->getType();
  unsigned TypeWidth = ShType->getScalarSizeInBits();

  // We only accept shifts-by-a-constant in canEvaluateShifted().
  const APInt *C1;
  match(InnerShift->getOperand(1), m_APInt(C1));
  unsigned InnerShAmt = C1->getZExtValue();

  // Change the shift amount and clear the appropriate IR flags.
  auto NewInnerShift = [&](unsigned ShAmt) {
    InnerShift->setOperand(1, ConstantInt::get(ShType, ShAmt));
    if (IsInnerShl) {
      InnerShift->setHasNoUnsignedWrap(false);
      InnerShift->setHasNoSignedWrap(false);
    } else {
      InnerShift->setIsExact(false);
    }
    return InnerShift;
  };

  // Two logical shifts in the same direction:
  // shl (shl X, C1), C2 -->  shl X, C1 + C2
  // lshr (lshr X, C1), C2 --> lshr X, C1 + C2
  if (IsInnerShl == IsOuterShl) {
    // If this is an oversized composite shift, then unsigned shifts get 0.
    if (InnerShAmt + OuterShAmt >= TypeWidth)
      return Constant::getNullValue(ShType);

    return NewInnerShift(InnerShAmt + OuterShAmt);
  }

  // Equal shift amounts in opposite directions become bitwise 'and':
  // lshr (shl X, C), C --> and X, C'
  // shl (lshr X, C), C --> and X, C'
  if (InnerShAmt == OuterShAmt) {
    APInt Mask = IsInnerShl
                     ? APInt::getLowBitsSet(TypeWidth, TypeWidth - OuterShAmt)
                     : APInt::getHighBitsSet(TypeWidth, TypeWidth - OuterShAmt);
    Value *And = Builder.CreateAnd(InnerShift->getOperand(0),
                                   ConstantInt::get(ShType, Mask));
    if (auto *AndI = dyn_cast<Instruction>(And)) {
      AndI->moveBefore(InnerShift);
      AndI->takeName(InnerShift);
    }
    return And;
  }

  assert(InnerShAmt > OuterShAmt &&
         "Unexpected opposite direction logical shift pair");

  // In general, we would need an 'and' for this transform, but
  // canEvaluateShiftedShift() guarantees that the masked-off bits are not used.
  // lshr (shl X, C1), C2 -->  shl X, C1 - C2
  // shl (lshr X, C1), C2 --> lshr X, C1 - C2
  return NewInnerShift(InnerShAmt - OuterShAmt);
}

/// When canEvaluateShifted() returns true for an expression, this function
/// inserts the new computation that produces the shifted value.
static Value *getShiftedValue(Value *V, unsigned NumBits, bool isLeftShift,
                              InstCombinerImpl &IC, const DataLayout &DL) {
  // We can always evaluate constants shifted.
  if (Constant *C = dyn_cast<Constant>(V)) {
    if (isLeftShift)
      return IC.Builder.CreateShl(C, NumBits);
    else
      return IC.Builder.CreateLShr(C, NumBits);
  }

  Instruction *I = cast<Instruction>(V);
  IC.addToWorklist(I);

  switch (I->getOpcode()) {
  default: llvm_unreachable("Inconsistency with CanEvaluateShifted");
  case Instruction::And:
  case Instruction::Or:
  case Instruction::Xor:
    // Bitwise operators can all arbitrarily be arbitrarily evaluated shifted.
    I->setOperand(
        0, getShiftedValue(I->getOperand(0), NumBits, isLeftShift, IC, DL));
    I->setOperand(
        1, getShiftedValue(I->getOperand(1), NumBits, isLeftShift, IC, DL));
    return I;

  case Instruction::Shl:
  case Instruction::LShr:
    return foldShiftedShift(cast<BinaryOperator>(I), NumBits, isLeftShift,
                            IC.Builder);

  case Instruction::Select:
    I->setOperand(
        1, getShiftedValue(I->getOperand(1), NumBits, isLeftShift, IC, DL));
    I->setOperand(
        2, getShiftedValue(I->getOperand(2), NumBits, isLeftShift, IC, DL));
    return I;
  case Instruction::PHI: {
    // We can change a phi if we can change all operands.  Note that we never
    // get into trouble with cyclic PHIs here because we only consider
    // instructions with a single use.
    PHINode *PN = cast<PHINode>(I);
    for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i)
      PN->setIncomingValue(i, getShiftedValue(PN->getIncomingValue(i), NumBits,
                                              isLeftShift, IC, DL));
    return PN;
  }
  }
}

// If this is a bitwise operator or add with a constant RHS we might be able
// to pull it through a shift.
static bool canShiftBinOpWithConstantRHS(BinaryOperator &Shift,
                                         BinaryOperator *BO) {
  switch (BO->getOpcode()) {
  default:
    return false; // Do not perform transform!
  case Instruction::Add:
    return Shift.getOpcode() == Instruction::Shl;
  case Instruction::Or:
  case Instruction::And:
    return true;
  case Instruction::Xor:
    // Do not change a 'not' of logical shift because that would create a normal
    // 'xor'. The 'not' is likely better for analysis, SCEV, and codegen.
    return !(Shift.isLogicalShift() && match(BO, m_Not(m_Value())));
  }
}

Instruction *InstCombinerImpl::FoldShiftByConstant(Value *Op0, Constant *Op1,
                                                   BinaryOperator &I) {
  const APInt *Op1C;
  if (!match(Op1, m_APInt(Op1C)))
    return nullptr;

  // See if we can propagate this shift into the input, this covers the trivial
  // cast of lshr(shl(x,c1),c2) as well as other more complex cases.
  bool IsLeftShift = I.getOpcode() == Instruction::Shl;
  if (I.getOpcode() != Instruction::AShr &&
      canEvaluateShifted(Op0, Op1C->getZExtValue(), IsLeftShift, *this, &I)) {
    LLVM_DEBUG(
        dbgs() << "ICE: GetShiftedValue propagating shift through expression"
                  " to eliminate shift:\n  IN: "
               << *Op0 << "\n  SH: " << I << "\n");

    return replaceInstUsesWith(
        I, getShiftedValue(Op0, Op1C->getZExtValue(), IsLeftShift, *this, DL));
  }

  // See if we can simplify any instructions used by the instruction whose sole
  // purpose is to compute bits we don't care about.
  Type *Ty = I.getType();
  unsigned TypeBits = Ty->getScalarSizeInBits();
  assert(!Op1C->uge(TypeBits) &&
         "Shift over the type width should have been removed already");
  (void)TypeBits;

  if (Instruction *FoldedShift = foldBinOpIntoSelectOrPhi(I))
    return FoldedShift;

  if (!Op0->hasOneUse())
    return nullptr;

  if (auto *Op0BO = dyn_cast<BinaryOperator>(Op0)) {
    // If the operand is a bitwise operator with a constant RHS, and the
    // shift is the only use, we can pull it out of the shift.
    const APInt *Op0C;
    if (match(Op0BO->getOperand(1), m_APInt(Op0C))) {
      if (canShiftBinOpWithConstantRHS(I, Op0BO)) {
        Constant *NewRHS = ConstantExpr::get(
            I.getOpcode(), cast<Constant>(Op0BO->getOperand(1)), Op1);

        Value *NewShift =
            Builder.CreateBinOp(I.getOpcode(), Op0BO->getOperand(0), Op1);
        NewShift->takeName(Op0BO);

        return BinaryOperator::Create(Op0BO->getOpcode(), NewShift, NewRHS);
      }
    }
  }

  // If we have a select that conditionally executes some binary operator,
  // see if we can pull it the select and operator through the shift.
  //
  // For example, turning:
  //   shl (select C, (add X, C1), X), C2
  // Into:
  //   Y = shl X, C2
  //   select C, (add Y, C1 << C2), Y
  Value *Cond;
  BinaryOperator *TBO;
  Value *FalseVal;
  if (match(Op0, m_Select(m_Value(Cond), m_OneUse(m_BinOp(TBO)),
                          m_Value(FalseVal)))) {
    const APInt *C;
    if (!isa<Constant>(FalseVal) && TBO->getOperand(0) == FalseVal &&
        match(TBO->getOperand(1), m_APInt(C)) &&
        canShiftBinOpWithConstantRHS(I, TBO)) {
      Constant *NewRHS = ConstantExpr::get(
          I.getOpcode(), cast<Constant>(TBO->getOperand(1)), Op1);

      Value *NewShift = Builder.CreateBinOp(I.getOpcode(), FalseVal, Op1);
      Value *NewOp = Builder.CreateBinOp(TBO->getOpcode(), NewShift, NewRHS);
      return SelectInst::Create(Cond, NewOp, NewShift);
    }
  }

  BinaryOperator *FBO;
  Value *TrueVal;
  if (match(Op0, m_Select(m_Value(Cond), m_Value(TrueVal),
                          m_OneUse(m_BinOp(FBO))))) {
    const APInt *C;
    if (!isa<Constant>(TrueVal) && FBO->getOperand(0) == TrueVal &&
        match(FBO->getOperand(1), m_APInt(C)) &&
        canShiftBinOpWithConstantRHS(I, FBO)) {
      Constant *NewRHS = ConstantExpr::get(
          I.getOpcode(), cast<Constant>(FBO->getOperand(1)), Op1);

      Value *NewShift = Builder.CreateBinOp(I.getOpcode(), TrueVal, Op1);
      Value *NewOp = Builder.CreateBinOp(FBO->getOpcode(), NewShift, NewRHS);
      return SelectInst::Create(Cond, NewShift, NewOp);
    }
  }

  return nullptr;
}

Instruction *InstCombinerImpl::visitShl(BinaryOperator &I) {
  const SimplifyQuery Q = SQ.getWithInstruction(&I);

  if (Value *V = SimplifyShlInst(I.getOperand(0), I.getOperand(1),
                                 I.hasNoSignedWrap(), I.hasNoUnsignedWrap(), Q))
    return replaceInstUsesWith(I, V);

  if (Instruction *X = foldVectorBinop(I))
    return X;

  if (Instruction *V = commonShiftTransforms(I))
    return V;

  if (Instruction *V = dropRedundantMaskingOfLeftShiftInput(&I, Q, Builder))
    return V;

  Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
  Type *Ty = I.getType();
  unsigned BitWidth = Ty->getScalarSizeInBits();

  const APInt *C;
  if (match(Op1, m_APInt(C))) {
    unsigned ShAmtC = C->getZExtValue();

    // shl (zext X), C --> zext (shl X, C)
    // This is only valid if X would have zeros shifted out.
    Value *X;
    if (match(Op0, m_OneUse(m_ZExt(m_Value(X))))) {
      unsigned SrcWidth = X->getType()->getScalarSizeInBits();
      if (ShAmtC < SrcWidth &&
          MaskedValueIsZero(X, APInt::getHighBitsSet(SrcWidth, ShAmtC), 0, &I))
        return new ZExtInst(Builder.CreateShl(X, ShAmtC), Ty);
    }

    // (X >> C) << C --> X & (-1 << C)
    if (match(Op0, m_Shr(m_Value(X), m_Specific(Op1)))) {
      APInt Mask(APInt::getHighBitsSet(BitWidth, BitWidth - ShAmtC));
      return BinaryOperator::CreateAnd(X, ConstantInt::get(Ty, Mask));
    }

    const APInt *C1;
    if (match(Op0, m_Exact(m_Shr(m_Value(X), m_APInt(C1)))) &&
        C1->ult(BitWidth)) {
      unsigned ShrAmt = C1->getZExtValue();
      if (ShrAmt < ShAmtC) {
        // If C1 < C: (X >>?,exact C1) << C --> X << (C - C1)
        Constant *ShiftDiff = ConstantInt::get(Ty, ShAmtC - ShrAmt);
        auto *NewShl = BinaryOperator::CreateShl(X, ShiftDiff);
        NewShl->setHasNoUnsignedWrap(I.hasNoUnsignedWrap());
        NewShl->setHasNoSignedWrap(I.hasNoSignedWrap());
        return NewShl;
      }
      if (ShrAmt > ShAmtC) {
        // If C1 > C: (X >>?exact C1) << C --> X >>?exact (C1 - C)
        Constant *ShiftDiff = ConstantInt::get(Ty, ShrAmt - ShAmtC);
        auto *NewShr = BinaryOperator::Create(
            cast<BinaryOperator>(Op0)->getOpcode(), X, ShiftDiff);
        NewShr->setIsExact(true);
        return NewShr;
      }
    }

    if (match(Op0, m_OneUse(m_Shr(m_Value(X), m_APInt(C1)))) &&
        C1->ult(BitWidth)) {
      unsigned ShrAmt = C1->getZExtValue();
      if (ShrAmt < ShAmtC) {
        // If C1 < C: (X >>? C1) << C --> (X << (C - C1)) & (-1 << C)
        Constant *ShiftDiff = ConstantInt::get(Ty, ShAmtC - ShrAmt);
        auto *NewShl = BinaryOperator::CreateShl(X, ShiftDiff);
        NewShl->setHasNoUnsignedWrap(I.hasNoUnsignedWrap());
        NewShl->setHasNoSignedWrap(I.hasNoSignedWrap());
        Builder.Insert(NewShl);
        APInt Mask(APInt::getHighBitsSet(BitWidth, BitWidth - ShAmtC));
        return BinaryOperator::CreateAnd(NewShl, ConstantInt::get(Ty, Mask));
      }
      if (ShrAmt > ShAmtC) {
        // If C1 > C: (X >>? C1) << C --> (X >>? (C1 - C)) & (-1 << C)
        Constant *ShiftDiff = ConstantInt::get(Ty, ShrAmt - ShAmtC);
        auto *OldShr = cast<BinaryOperator>(Op0);
        auto *NewShr =
            BinaryOperator::Create(OldShr->getOpcode(), X, ShiftDiff);
        NewShr->setIsExact(OldShr->isExact());
        Builder.Insert(NewShr);
        APInt Mask(APInt::getHighBitsSet(BitWidth, BitWidth - ShAmtC));
        return BinaryOperator::CreateAnd(NewShr, ConstantInt::get(Ty, Mask));
      }
    }

    // Similar to above, but look through an intermediate trunc instruction.
    BinaryOperator *Shr;
    if (match(Op0, m_OneUse(m_Trunc(m_OneUse(m_BinOp(Shr))))) &&
        match(Shr, m_Shr(m_Value(X), m_APInt(C1)))) {
      // The larger shift direction survives through the transform.
      unsigned ShrAmtC = C1->getZExtValue();
      unsigned ShDiff = ShrAmtC > ShAmtC ? ShrAmtC - ShAmtC : ShAmtC - ShrAmtC;
      Constant *ShiftDiffC = ConstantInt::get(X->getType(), ShDiff);
      auto ShiftOpc = ShrAmtC > ShAmtC ? Shr->getOpcode() : Instruction::Shl;

      // If C1 > C:
      // (trunc (X >> C1)) << C --> (trunc (X >> (C1 - C))) && (-1 << C)
      // If C > C1:
      // (trunc (X >> C1)) << C --> (trunc (X << (C - C1))) && (-1 << C)
      Value *NewShift = Builder.CreateBinOp(ShiftOpc, X, ShiftDiffC, "sh.diff");
      Value *Trunc = Builder.CreateTrunc(NewShift, Ty, "tr.sh.diff");
      APInt Mask(APInt::getHighBitsSet(BitWidth, BitWidth - ShAmtC));
      return BinaryOperator::CreateAnd(Trunc, ConstantInt::get(Ty, Mask));
    }

    if (match(Op0, m_Shl(m_Value(X), m_APInt(C1))) && C1->ult(BitWidth)) {
      unsigned AmtSum = ShAmtC + C1->getZExtValue();
      // Oversized shifts are simplified to zero in InstSimplify.
      if (AmtSum < BitWidth)
        // (X << C1) << C2 --> X << (C1 + C2)
        return BinaryOperator::CreateShl(X, ConstantInt::get(Ty, AmtSum));
    }

    // If we have an opposite shift by the same amount, we may be able to
    // reorder binops and shifts to eliminate math/logic.
    auto isSuitableBinOpcode = [](Instruction::BinaryOps BinOpcode) {
      switch (BinOpcode) {
      default:
        return false;
      case Instruction::Add:
      case Instruction::And:
      case Instruction::Or:
      case Instruction::Xor:
      case Instruction::Sub:
        // NOTE: Sub is not commutable and the tranforms below may not be valid
        //       when the shift-right is operand 1 (RHS) of the sub.
        return true;
      }
    };
    BinaryOperator *Op0BO;
    if (match(Op0, m_OneUse(m_BinOp(Op0BO))) &&
        isSuitableBinOpcode(Op0BO->getOpcode())) {
      // Commute so shift-right is on LHS of the binop.
      // (Y bop (X >> C)) << C         ->  ((X >> C) bop Y) << C
      // (Y bop ((X >> C) & CC)) << C  ->  (((X >> C) & CC) bop Y) << C
      Value *Shr = Op0BO->getOperand(0);
      Value *Y = Op0BO->getOperand(1);
      Value *X;
      const APInt *CC;
      if (Op0BO->isCommutative() && Y->hasOneUse() &&
          (match(Y, m_Shr(m_Value(), m_Specific(Op1))) ||
           match(Y, m_And(m_OneUse(m_Shr(m_Value(), m_Specific(Op1))),
                          m_APInt(CC)))))
        std::swap(Shr, Y);

      // ((X >> C) bop Y) << C  ->  (X bop (Y << C)) & (~0 << C)
      if (match(Shr, m_OneUse(m_Shr(m_Value(X), m_Specific(Op1))))) {
        // Y << C
        Value *YS = Builder.CreateShl(Y, Op1, Op0BO->getName());
        // (X bop (Y << C))
        Value *B =
            Builder.CreateBinOp(Op0BO->getOpcode(), X, YS, Shr->getName());
        unsigned Op1Val = C->getLimitedValue(BitWidth);
        APInt Bits = APInt::getHighBitsSet(BitWidth, BitWidth - Op1Val);
        Constant *Mask = ConstantInt::get(Ty, Bits);
        return BinaryOperator::CreateAnd(B, Mask);
      }

      // (((X >> C) & CC) bop Y) << C  ->  (X & (CC << C)) bop (Y << C)
      if (match(Shr,
                m_OneUse(m_And(m_OneUse(m_Shr(m_Value(X), m_Specific(Op1))),
                               m_APInt(CC))))) {
        // Y << C
        Value *YS = Builder.CreateShl(Y, Op1, Op0BO->getName());
        // X & (CC << C)
        Value *M = Builder.CreateAnd(X, ConstantInt::get(Ty, CC->shl(*C)),
                                     X->getName() + ".mask");
        return BinaryOperator::Create(Op0BO->getOpcode(), M, YS);
      }
    }

    // (C1 - X) << C --> (C1 << C) - (X << C)
    if (match(Op0, m_OneUse(m_Sub(m_APInt(C1), m_Value(X))))) {
      Constant *NewLHS = ConstantInt::get(Ty, C1->shl(*C));
      Value *NewShift = Builder.CreateShl(X, Op1);
      return BinaryOperator::CreateSub(NewLHS, NewShift);
    }

    // If the shifted-out value is known-zero, then this is a NUW shift.
    if (!I.hasNoUnsignedWrap() &&
        MaskedValueIsZero(Op0, APInt::getHighBitsSet(BitWidth, ShAmtC), 0,
                          &I)) {
      I.setHasNoUnsignedWrap();
      return &I;
    }

    // If the shifted-out value is all signbits, then this is a NSW shift.
    if (!I.hasNoSignedWrap() && ComputeNumSignBits(Op0, 0, &I) > ShAmtC) {
      I.setHasNoSignedWrap();
      return &I;
    }
  }

  // Transform  (x >> y) << y  to  x & (-1 << y)
  // Valid for any type of right-shift.
  Value *X;
  if (match(Op0, m_OneUse(m_Shr(m_Value(X), m_Specific(Op1))))) {
    Constant *AllOnes = ConstantInt::getAllOnesValue(Ty);
    Value *Mask = Builder.CreateShl(AllOnes, Op1);
    return BinaryOperator::CreateAnd(Mask, X);
  }

  Constant *C1;
  if (match(Op1, m_Constant(C1))) {
    Constant *C2;
    Value *X;
    // (C2 << X) << C1 --> (C2 << C1) << X
    if (match(Op0, m_OneUse(m_Shl(m_Constant(C2), m_Value(X)))))
      return BinaryOperator::CreateShl(ConstantExpr::getShl(C2, C1), X);

    // (X * C2) << C1 --> X * (C2 << C1)
    if (match(Op0, m_Mul(m_Value(X), m_Constant(C2))))
      return BinaryOperator::CreateMul(X, ConstantExpr::getShl(C2, C1));

    // shl (zext i1 X), C1 --> select (X, 1 << C1, 0)
    if (match(Op0, m_ZExt(m_Value(X))) && X->getType()->isIntOrIntVectorTy(1)) {
      auto *NewC = ConstantExpr::getShl(ConstantInt::get(Ty, 1), C1);
      return SelectInst::Create(X, NewC, ConstantInt::getNullValue(Ty));
    }
  }

  // (1 << (C - x)) -> ((1 << C) >> x) if C is bitwidth - 1
  if (match(Op0, m_One()) &&
      match(Op1, m_Sub(m_SpecificInt(BitWidth - 1), m_Value(X))))
    return BinaryOperator::CreateLShr(
        ConstantInt::get(Ty, APInt::getSignMask(BitWidth)), X);

  return nullptr;
}

Instruction *InstCombinerImpl::visitLShr(BinaryOperator &I) {
  if (Value *V = SimplifyLShrInst(I.getOperand(0), I.getOperand(1), I.isExact(),
                                  SQ.getWithInstruction(&I)))
    return replaceInstUsesWith(I, V);

  if (Instruction *X = foldVectorBinop(I))
    return X;

  if (Instruction *R = commonShiftTransforms(I))
    return R;

  Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
  Type *Ty = I.getType();
  const APInt *C;
  if (match(Op1, m_APInt(C))) {
    unsigned ShAmtC = C->getZExtValue();
    unsigned BitWidth = Ty->getScalarSizeInBits();
    auto *II = dyn_cast<IntrinsicInst>(Op0);
    if (II && isPowerOf2_32(BitWidth) && Log2_32(BitWidth) == ShAmtC &&
        (II->getIntrinsicID() == Intrinsic::ctlz ||
         II->getIntrinsicID() == Intrinsic::cttz ||
         II->getIntrinsicID() == Intrinsic::ctpop)) {
      // ctlz.i32(x)>>5  --> zext(x == 0)
      // cttz.i32(x)>>5  --> zext(x == 0)
      // ctpop.i32(x)>>5 --> zext(x == -1)
      bool IsPop = II->getIntrinsicID() == Intrinsic::ctpop;
      Constant *RHS = ConstantInt::getSigned(Ty, IsPop ? -1 : 0);
      Value *Cmp = Builder.CreateICmpEQ(II->getArgOperand(0), RHS);
      return new ZExtInst(Cmp, Ty);
    }

    Value *X;
    const APInt *C1;
    if (match(Op0, m_Shl(m_Value(X), m_APInt(C1))) && C1->ult(BitWidth)) {
      if (C1->ult(ShAmtC)) {
        unsigned ShlAmtC = C1->getZExtValue();
        Constant *ShiftDiff = ConstantInt::get(Ty, ShAmtC - ShlAmtC);
        if (cast<BinaryOperator>(Op0)->hasNoUnsignedWrap()) {
          // (X <<nuw C1) >>u C --> X >>u (C - C1)
          auto *NewLShr = BinaryOperator::CreateLShr(X, ShiftDiff);
          NewLShr->setIsExact(I.isExact());
          return NewLShr;
        }
        if (Op0->hasOneUse()) {
          // (X << C1) >>u C  --> (X >>u (C - C1)) & (-1 >> C)
          Value *NewLShr = Builder.CreateLShr(X, ShiftDiff, "", I.isExact());
          APInt Mask(APInt::getLowBitsSet(BitWidth, BitWidth - ShAmtC));
          return BinaryOperator::CreateAnd(NewLShr, ConstantInt::get(Ty, Mask));
        }
      } else if (C1->ugt(ShAmtC)) {
        unsigned ShlAmtC = C1->getZExtValue();
        Constant *ShiftDiff = ConstantInt::get(Ty, ShlAmtC - ShAmtC);
        if (cast<BinaryOperator>(Op0)->hasNoUnsignedWrap()) {
          // (X <<nuw C1) >>u C --> X <<nuw (C1 - C)
          auto *NewShl = BinaryOperator::CreateShl(X, ShiftDiff);
          NewShl->setHasNoUnsignedWrap(true);
          return NewShl;
        }
        if (Op0->hasOneUse()) {
          // (X << C1) >>u C  --> X << (C1 - C) & (-1 >> C)
          Value *NewShl = Builder.CreateShl(X, ShiftDiff);
          APInt Mask(APInt::getLowBitsSet(BitWidth, BitWidth - ShAmtC));
          return BinaryOperator::CreateAnd(NewShl, ConstantInt::get(Ty, Mask));
        }
      } else {
        assert(*C1 == ShAmtC);
        // (X << C) >>u C --> X & (-1 >>u C)
        APInt Mask(APInt::getLowBitsSet(BitWidth, BitWidth - ShAmtC));
        return BinaryOperator::CreateAnd(X, ConstantInt::get(Ty, Mask));
      }
    }

    // ((X << C) + Y) >>u C --> (X + (Y >>u C)) & (-1 >>u C)
    // TODO: Consolidate with the more general transform that starts from shl
    //       (the shifts are in the opposite order).
    Value *Y;
    if (match(Op0,
              m_OneUse(m_c_Add(m_OneUse(m_Shl(m_Value(X), m_Specific(Op1))),
                               m_Value(Y))))) {
      Value *NewLshr = Builder.CreateLShr(Y, Op1);
      Value *NewAdd = Builder.CreateAdd(NewLshr, X);
      unsigned Op1Val = C->getLimitedValue(BitWidth);
      APInt Bits = APInt::getLowBitsSet(BitWidth, BitWidth - Op1Val);
      Constant *Mask = ConstantInt::get(Ty, Bits);
      return BinaryOperator::CreateAnd(NewAdd, Mask);
    }

    if (match(Op0, m_OneUse(m_ZExt(m_Value(X)))) &&
        (!Ty->isIntegerTy() || shouldChangeType(Ty, X->getType()))) {
      assert(ShAmtC < X->getType()->getScalarSizeInBits() &&
             "Big shift not simplified to zero?");
      // lshr (zext iM X to iN), C --> zext (lshr X, C) to iN
      Value *NewLShr = Builder.CreateLShr(X, ShAmtC);
      return new ZExtInst(NewLShr, Ty);
    }

    if (match(Op0, m_SExt(m_Value(X)))) {
      unsigned SrcTyBitWidth = X->getType()->getScalarSizeInBits();
      // lshr (sext i1 X to iN), C --> select (X, -1 >> C, 0)
      if (SrcTyBitWidth == 1) {
        auto *NewC = ConstantInt::get(
            Ty, APInt::getLowBitsSet(BitWidth, BitWidth - ShAmtC));
        return SelectInst::Create(X, NewC, ConstantInt::getNullValue(Ty));
      }

      if ((!Ty->isIntegerTy() || shouldChangeType(Ty, X->getType())) &&
          Op0->hasOneUse()) {
        // Are we moving the sign bit to the low bit and widening with high
        // zeros? lshr (sext iM X to iN), N-1 --> zext (lshr X, M-1) to iN
        if (ShAmtC == BitWidth - 1) {
          Value *NewLShr = Builder.CreateLShr(X, SrcTyBitWidth - 1);
          return new ZExtInst(NewLShr, Ty);
        }

        // lshr (sext iM X to iN), N-M --> zext (ashr X, min(N-M, M-1)) to iN
        if (ShAmtC == BitWidth - SrcTyBitWidth) {
          // The new shift amount can't be more than the narrow source type.
          unsigned NewShAmt = std::min(ShAmtC, SrcTyBitWidth - 1);
          Value *AShr = Builder.CreateAShr(X, NewShAmt);
          return new ZExtInst(AShr, Ty);
        }
      }
    }

    if (ShAmtC == BitWidth - 1) {
      // lshr i32 or(X,-X), 31 --> zext (X != 0)
      if (match(Op0, m_OneUse(m_c_Or(m_Neg(m_Value(X)), m_Deferred(X)))))
        return new ZExtInst(Builder.CreateIsNotNull(X), Ty);

      // lshr i32 (X -nsw Y), 31 --> zext (X < Y)
      if (match(Op0, m_OneUse(m_NSWSub(m_Value(X), m_Value(Y)))))
        return new ZExtInst(Builder.CreateICmpSLT(X, Y), Ty);

      // Check if a number is negative and odd:
      // lshr i32 (srem X, 2), 31 --> and (X >> 31), X
      if (match(Op0, m_OneUse(m_SRem(m_Value(X), m_SpecificInt(2))))) {
        Value *Signbit = Builder.CreateLShr(X, ShAmtC);
        return BinaryOperator::CreateAnd(Signbit, X);
      }
    }

    // (X >>u C1) >>u C --> X >>u (C1 + C)
    if (match(Op0, m_LShr(m_Value(X), m_APInt(C1)))) {
      // Oversized shifts are simplified to zero in InstSimplify.
      unsigned AmtSum = ShAmtC + C1->getZExtValue();
      if (AmtSum < BitWidth)
        return BinaryOperator::CreateLShr(X, ConstantInt::get(Ty, AmtSum));
    }

    Instruction *TruncSrc;
    if (match(Op0, m_OneUse(m_Trunc(m_Instruction(TruncSrc)))) &&
        match(TruncSrc, m_LShr(m_Value(X), m_APInt(C1)))) {
      unsigned SrcWidth = X->getType()->getScalarSizeInBits();
      unsigned AmtSum = ShAmtC + C1->getZExtValue();

      // If the combined shift fits in the source width:
      // (trunc (X >>u C1)) >>u C --> and (trunc (X >>u (C1 + C)), MaskC
      //
      // If the first shift covers the number of bits truncated, then the
      // mask instruction is eliminated (and so the use check is relaxed).
      if (AmtSum < SrcWidth &&
          (TruncSrc->hasOneUse() || C1->uge(SrcWidth - BitWidth))) {
        Value *SumShift = Builder.CreateLShr(X, AmtSum, "sum.shift");
        Value *Trunc = Builder.CreateTrunc(SumShift, Ty, I.getName());

        // If the first shift does not cover the number of bits truncated, then
        // we require a mask to get rid of high bits in the result.
        APInt MaskC = APInt::getAllOnes(BitWidth).lshr(ShAmtC);
        return BinaryOperator::CreateAnd(Trunc, ConstantInt::get(Ty, MaskC));
      }
    }

    // Look for a "splat" mul pattern - it replicates bits across each half of
    // a value, so a right shift is just a mask of the low bits:
    // lshr i32 (mul nuw X, Pow2+1), 16 --> and X, Pow2-1
    // TODO: Generalize to allow more than just half-width shifts?
    const APInt *MulC;
    if (match(Op0, m_NUWMul(m_Value(X), m_APInt(MulC))) &&
        ShAmtC * 2 == BitWidth && (*MulC - 1).isPowerOf2() &&
        MulC->logBase2() == ShAmtC)
      return BinaryOperator::CreateAnd(X, ConstantInt::get(Ty, *MulC - 2));

    // If the shifted-out value is known-zero, then this is an exact shift.
    if (!I.isExact() &&
        MaskedValueIsZero(Op0, APInt::getLowBitsSet(BitWidth, ShAmtC), 0, &I)) {
      I.setIsExact();
      return &I;
    }
  }

  // Transform  (x << y) >> y  to  x & (-1 >> y)
  Value *X;
  if (match(Op0, m_OneUse(m_Shl(m_Value(X), m_Specific(Op1))))) {
    Constant *AllOnes = ConstantInt::getAllOnesValue(Ty);
    Value *Mask = Builder.CreateLShr(AllOnes, Op1);
    return BinaryOperator::CreateAnd(Mask, X);
  }

  return nullptr;
}

Instruction *
InstCombinerImpl::foldVariableSignZeroExtensionOfVariableHighBitExtract(
    BinaryOperator &OldAShr) {
  assert(OldAShr.getOpcode() == Instruction::AShr &&
         "Must be called with arithmetic right-shift instruction only.");

  // Check that constant C is a splat of the element-wise bitwidth of V.
  auto BitWidthSplat = [](Constant *C, Value *V) {
    return match(
        C, m_SpecificInt_ICMP(ICmpInst::Predicate::ICMP_EQ,
                              APInt(C->getType()->getScalarSizeInBits(),
                                    V->getType()->getScalarSizeInBits())));
  };

  // It should look like variable-length sign-extension on the outside:
  //   (Val << (bitwidth(Val)-Nbits)) a>> (bitwidth(Val)-Nbits)
  Value *NBits;
  Instruction *MaybeTrunc;
  Constant *C1, *C2;
  if (!match(&OldAShr,
             m_AShr(m_Shl(m_Instruction(MaybeTrunc),
                          m_ZExtOrSelf(m_Sub(m_Constant(C1),
                                             m_ZExtOrSelf(m_Value(NBits))))),
                    m_ZExtOrSelf(m_Sub(m_Constant(C2),
                                       m_ZExtOrSelf(m_Deferred(NBits)))))) ||
      !BitWidthSplat(C1, &OldAShr) || !BitWidthSplat(C2, &OldAShr))
    return nullptr;

  // There may or may not be a truncation after outer two shifts.
  Instruction *HighBitExtract;
  match(MaybeTrunc, m_TruncOrSelf(m_Instruction(HighBitExtract)));
  bool HadTrunc = MaybeTrunc != HighBitExtract;

  // And finally, the innermost part of the pattern must be a right-shift.
  Value *X, *NumLowBitsToSkip;
  if (!match(HighBitExtract, m_Shr(m_Value(X), m_Value(NumLowBitsToSkip))))
    return nullptr;

  // Said right-shift must extract high NBits bits - C0 must be it's bitwidth.
  Constant *C0;
  if (!match(NumLowBitsToSkip,
             m_ZExtOrSelf(
                 m_Sub(m_Constant(C0), m_ZExtOrSelf(m_Specific(NBits))))) ||
      !BitWidthSplat(C0, HighBitExtract))
    return nullptr;

  // Since the NBits is identical for all shifts, if the outermost and
  // innermost shifts are identical, then outermost shifts are redundant.
  // If we had truncation, do keep it though.
  if (HighBitExtract->getOpcode() == OldAShr.getOpcode())
    return replaceInstUsesWith(OldAShr, MaybeTrunc);

  // Else, if there was a truncation, then we need to ensure that one
  // instruction will go away.
  if (HadTrunc && !match(&OldAShr, m_c_BinOp(m_OneUse(m_Value()), m_Value())))
    return nullptr;

  // Finally, bypass two innermost shifts, and perform the outermost shift on
  // the operands of the innermost shift.
  Instruction *NewAShr =
      BinaryOperator::Create(OldAShr.getOpcode(), X, NumLowBitsToSkip);
  NewAShr->copyIRFlags(HighBitExtract); // We can preserve 'exact'-ness.
  if (!HadTrunc)
    return NewAShr;

  Builder.Insert(NewAShr);
  return TruncInst::CreateTruncOrBitCast(NewAShr, OldAShr.getType());
}

Instruction *InstCombinerImpl::visitAShr(BinaryOperator &I) {
  if (Value *V = SimplifyAShrInst(I.getOperand(0), I.getOperand(1), I.isExact(),
                                  SQ.getWithInstruction(&I)))
    return replaceInstUsesWith(I, V);

  if (Instruction *X = foldVectorBinop(I))
    return X;

  if (Instruction *R = commonShiftTransforms(I))
    return R;

  Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
  Type *Ty = I.getType();
  unsigned BitWidth = Ty->getScalarSizeInBits();
  const APInt *ShAmtAPInt;
  if (match(Op1, m_APInt(ShAmtAPInt)) && ShAmtAPInt->ult(BitWidth)) {
    unsigned ShAmt = ShAmtAPInt->getZExtValue();

    // If the shift amount equals the difference in width of the destination
    // and source scalar types:
    // ashr (shl (zext X), C), C --> sext X
    Value *X;
    if (match(Op0, m_Shl(m_ZExt(m_Value(X)), m_Specific(Op1))) &&
        ShAmt == BitWidth - X->getType()->getScalarSizeInBits())
      return new SExtInst(X, Ty);

    // We can't handle (X << C1) >>s C2. It shifts arbitrary bits in. However,
    // we can handle (X <<nsw C1) >>s C2 since it only shifts in sign bits.
    const APInt *ShOp1;
    if (match(Op0, m_NSWShl(m_Value(X), m_APInt(ShOp1))) &&
        ShOp1->ult(BitWidth)) {
      unsigned ShlAmt = ShOp1->getZExtValue();
      if (ShlAmt < ShAmt) {
        // (X <<nsw C1) >>s C2 --> X >>s (C2 - C1)
        Constant *ShiftDiff = ConstantInt::get(Ty, ShAmt - ShlAmt);
        auto *NewAShr = BinaryOperator::CreateAShr(X, ShiftDiff);
        NewAShr->setIsExact(I.isExact());
        return NewAShr;
      }
      if (ShlAmt > ShAmt) {
        // (X <<nsw C1) >>s C2 --> X <<nsw (C1 - C2)
        Constant *ShiftDiff = ConstantInt::get(Ty, ShlAmt - ShAmt);
        auto *NewShl = BinaryOperator::Create(Instruction::Shl, X, ShiftDiff);
        NewShl->setHasNoSignedWrap(true);
        return NewShl;
      }
    }

    if (match(Op0, m_AShr(m_Value(X), m_APInt(ShOp1))) &&
        ShOp1->ult(BitWidth)) {
      unsigned AmtSum = ShAmt + ShOp1->getZExtValue();
      // Oversized arithmetic shifts replicate the sign bit.
      AmtSum = std::min(AmtSum, BitWidth - 1);
      // (X >>s C1) >>s C2 --> X >>s (C1 + C2)
      return BinaryOperator::CreateAShr(X, ConstantInt::get(Ty, AmtSum));
    }

    if (match(Op0, m_OneUse(m_SExt(m_Value(X)))) &&
        (Ty->isVectorTy() || shouldChangeType(Ty, X->getType()))) {
      // ashr (sext X), C --> sext (ashr X, C')
      Type *SrcTy = X->getType();
      ShAmt = std::min(ShAmt, SrcTy->getScalarSizeInBits() - 1);
      Value *NewSh = Builder.CreateAShr(X, ConstantInt::get(SrcTy, ShAmt));
      return new SExtInst(NewSh, Ty);
    }

    if (ShAmt == BitWidth - 1) {
      // ashr i32 or(X,-X), 31 --> sext (X != 0)
      if (match(Op0, m_OneUse(m_c_Or(m_Neg(m_Value(X)), m_Deferred(X)))))
        return new SExtInst(Builder.CreateIsNotNull(X), Ty);

      // ashr i32 (X -nsw Y), 31 --> sext (X < Y)
      Value *Y;
      if (match(Op0, m_OneUse(m_NSWSub(m_Value(X), m_Value(Y)))))
        return new SExtInst(Builder.CreateICmpSLT(X, Y), Ty);
    }

    // If the shifted-out value is known-zero, then this is an exact shift.
    if (!I.isExact() &&
        MaskedValueIsZero(Op0, APInt::getLowBitsSet(BitWidth, ShAmt), 0, &I)) {
      I.setIsExact();
      return &I;
    }
  }

  // Prefer `-(x & 1)` over `(x << (bitwidth(x)-1)) a>> (bitwidth(x)-1)`
  // as the pattern to splat the lowest bit.
  // FIXME: iff X is already masked, we don't need the one-use check.
  Value *X;
  if (match(Op1, m_SpecificIntAllowUndef(BitWidth - 1)) &&
      match(Op0, m_OneUse(m_Shl(m_Value(X),
                                m_SpecificIntAllowUndef(BitWidth - 1))))) {
    Constant *Mask = ConstantInt::get(Ty, 1);
    // Retain the knowledge about the ignored lanes.
    Mask = Constant::mergeUndefsWith(
        Constant::mergeUndefsWith(Mask, cast<Constant>(Op1)),
        cast<Constant>(cast<Instruction>(Op0)->getOperand(1)));
    X = Builder.CreateAnd(X, Mask);
    return BinaryOperator::CreateNeg(X);
  }

  if (Instruction *R = foldVariableSignZeroExtensionOfVariableHighBitExtract(I))
    return R;

  // See if we can turn a signed shr into an unsigned shr.
  if (MaskedValueIsZero(Op0, APInt::getSignMask(BitWidth), 0, &I))
    return BinaryOperator::CreateLShr(Op0, Op1);

  // ashr (xor %x, -1), %y  -->  xor (ashr %x, %y), -1
  if (match(Op0, m_OneUse(m_Not(m_Value(X))))) {
    // Note that we must drop 'exact'-ness of the shift!
    // Note that we can't keep undef's in -1 vector constant!
    auto *NewAShr = Builder.CreateAShr(X, Op1, Op0->getName() + ".not");
    return BinaryOperator::CreateNot(NewAShr);
  }

  return nullptr;
}