llvm.org GIT mirror llvm / release_32 lib / Transforms / IPO / MergeFunctions.cpp
release_32

Tree @release_32 (Download .tar.gz)

MergeFunctions.cpp @release_32raw · history · 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
//===- MergeFunctions.cpp - Merge identical functions ---------------------===//
//
//                     The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This pass looks for equivalent functions that are mergable and folds them.
//
// A hash is computed from the function, based on its type and number of
// basic blocks.
//
// Once all hashes are computed, we perform an expensive equality comparison
// on each function pair. This takes n^2/2 comparisons per bucket, so it's
// important that the hash function be high quality. The equality comparison
// iterates through each instruction in each basic block.
//
// When a match is found the functions are folded. If both functions are
// overridable, we move the functionality into a new internal function and
// leave two overridable thunks to it.
//
//===----------------------------------------------------------------------===//
//
// Future work:
//
// * virtual functions.
//
// Many functions have their address taken by the virtual function table for
// the object they belong to. However, as long as it's only used for a lookup
// and call, this is irrelevant, and we'd like to fold such functions.
//
// * switch from n^2 pair-wise comparisons to an n-way comparison for each
// bucket.
//
// * be smarter about bitcasts.
//
// In order to fold functions, we will sometimes add either bitcast instructions
// or bitcast constant expressions. Unfortunately, this can confound further
// analysis since the two functions differ where one has a bitcast and the
// other doesn't. We should learn to look through bitcasts.
//
//===----------------------------------------------------------------------===//

#define DEBUG_TYPE "mergefunc"
#include "llvm/Transforms/IPO.h"
#include "llvm/Constants.h"
#include "llvm/IRBuilder.h"
#include "llvm/InlineAsm.h"
#include "llvm/Instructions.h"
#include "llvm/LLVMContext.h"
#include "llvm/Module.h"
#include "llvm/Operator.h"
#include "llvm/Pass.h"
#include "llvm/ADT/DenseSet.h"
#include "llvm/ADT/FoldingSet.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/ADT/SmallSet.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/Support/CallSite.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/ValueHandle.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/DataLayout.h"
#include <vector>
using namespace llvm;

STATISTIC(NumFunctionsMerged, "Number of functions merged");
STATISTIC(NumThunksWritten, "Number of thunks generated");
STATISTIC(NumAliasesWritten, "Number of aliases generated");
STATISTIC(NumDoubleWeak, "Number of new functions created");

/// Creates a hash-code for the function which is the same for any two
/// functions that will compare equal, without looking at the instructions
/// inside the function.
static unsigned profileFunction(const Function *F) {
  FunctionType *FTy = F->getFunctionType();

  FoldingSetNodeID ID;
  ID.AddInteger(F->size());
  ID.AddInteger(F->getCallingConv());
  ID.AddBoolean(F->hasGC());
  ID.AddBoolean(FTy->isVarArg());
  ID.AddInteger(FTy->getReturnType()->getTypeID());
  for (unsigned i = 0, e = FTy->getNumParams(); i != e; ++i)
    ID.AddInteger(FTy->getParamType(i)->getTypeID());
  return ID.ComputeHash();
}

namespace {

/// ComparableFunction - A struct that pairs together functions with a
/// DataLayout so that we can keep them together as elements in the DenseSet.
class ComparableFunction {
public:
  static const ComparableFunction EmptyKey;
  static const ComparableFunction TombstoneKey;
  static DataLayout * const LookupOnly;

  ComparableFunction(Function *Func, DataLayout *TD)
    : Func(Func), Hash(profileFunction(Func)), TD(TD) {}

  Function *getFunc() const { return Func; }
  unsigned getHash() const { return Hash; }
  DataLayout *getTD() const { return TD; }

  // Drops AssertingVH reference to the function. Outside of debug mode, this
  // does nothing.
  void release() {
    assert(Func &&
           "Attempted to release function twice, or release empty/tombstone!");
    Func = NULL;
  }

private:
  explicit ComparableFunction(unsigned Hash)
    : Func(NULL), Hash(Hash), TD(NULL) {}

  AssertingVH<Function> Func;
  unsigned Hash;
  DataLayout *TD;
};

const ComparableFunction ComparableFunction::EmptyKey = ComparableFunction(0);
const ComparableFunction ComparableFunction::TombstoneKey =
    ComparableFunction(1);
DataLayout *const ComparableFunction::LookupOnly = (DataLayout*)(-1);

}

namespace llvm {
  template <>
  struct DenseMapInfo<ComparableFunction> {
    static ComparableFunction getEmptyKey() {
      return ComparableFunction::EmptyKey;
    }
    static ComparableFunction getTombstoneKey() {
      return ComparableFunction::TombstoneKey;
    }
    static unsigned getHashValue(const ComparableFunction &CF) {
      return CF.getHash();
    }
    static bool isEqual(const ComparableFunction &LHS,
                        const ComparableFunction &RHS);
  };
}

namespace {

/// FunctionComparator - Compares two functions to determine whether or not
/// they will generate machine code with the same behaviour. DataLayout is
/// used if available. The comparator always fails conservatively (erring on the
/// side of claiming that two functions are different).
class FunctionComparator {
public:
  FunctionComparator(const DataLayout *TD, const Function *F1,
                     const Function *F2)
    : F1(F1), F2(F2), TD(TD) {}

  /// Test whether the two functions have equivalent behaviour.
  bool compare();

private:
  /// Test whether two basic blocks have equivalent behaviour.
  bool compare(const BasicBlock *BB1, const BasicBlock *BB2);

  /// Assign or look up previously assigned numbers for the two values, and
  /// return whether the numbers are equal. Numbers are assigned in the order
  /// visited.
  bool enumerate(const Value *V1, const Value *V2);

  /// Compare two Instructions for equivalence, similar to
  /// Instruction::isSameOperationAs but with modifications to the type
  /// comparison.
  bool isEquivalentOperation(const Instruction *I1,
                             const Instruction *I2) const;

  /// Compare two GEPs for equivalent pointer arithmetic.
  bool isEquivalentGEP(const GEPOperator *GEP1, const GEPOperator *GEP2);
  bool isEquivalentGEP(const GetElementPtrInst *GEP1,
                       const GetElementPtrInst *GEP2) {
    return isEquivalentGEP(cast<GEPOperator>(GEP1), cast<GEPOperator>(GEP2));
  }

  /// Compare two Types, treating all pointer types as equal.
  bool isEquivalentType(Type *Ty1, Type *Ty2) const;

  // The two functions undergoing comparison.
  const Function *F1, *F2;

  const DataLayout *TD;

  DenseMap<const Value *, const Value *> id_map;
  DenseSet<const Value *> seen_values;
};

}

// Any two pointers in the same address space are equivalent, intptr_t and
// pointers are equivalent. Otherwise, standard type equivalence rules apply.
bool FunctionComparator::isEquivalentType(Type *Ty1,
                                          Type *Ty2) const {
  if (Ty1 == Ty2)
    return true;
  if (Ty1->getTypeID() != Ty2->getTypeID()) {
    if (TD) {
      LLVMContext &Ctx = Ty1->getContext();
      if (isa<PointerType>(Ty1) && Ty2 == TD->getIntPtrType(Ctx)) return true;
      if (isa<PointerType>(Ty2) && Ty1 == TD->getIntPtrType(Ctx)) return true;
    }
    return false;
  }

  switch (Ty1->getTypeID()) {
  default:
    llvm_unreachable("Unknown type!");
    // Fall through in Release mode.
  case Type::IntegerTyID:
  case Type::VectorTyID:
    // Ty1 == Ty2 would have returned true earlier.
    return false;

  case Type::VoidTyID:
  case Type::FloatTyID:
  case Type::DoubleTyID:
  case Type::X86_FP80TyID:
  case Type::FP128TyID:
  case Type::PPC_FP128TyID:
  case Type::LabelTyID:
  case Type::MetadataTyID:
    return true;

  case Type::PointerTyID: {
    PointerType *PTy1 = cast<PointerType>(Ty1);
    PointerType *PTy2 = cast<PointerType>(Ty2);
    return PTy1->getAddressSpace() == PTy2->getAddressSpace();
  }

  case Type::StructTyID: {
    StructType *STy1 = cast<StructType>(Ty1);
    StructType *STy2 = cast<StructType>(Ty2);
    if (STy1->getNumElements() != STy2->getNumElements())
      return false;

    if (STy1->isPacked() != STy2->isPacked())
      return false;

    for (unsigned i = 0, e = STy1->getNumElements(); i != e; ++i) {
      if (!isEquivalentType(STy1->getElementType(i), STy2->getElementType(i)))
        return false;
    }
    return true;
  }

  case Type::FunctionTyID: {
    FunctionType *FTy1 = cast<FunctionType>(Ty1);
    FunctionType *FTy2 = cast<FunctionType>(Ty2);
    if (FTy1->getNumParams() != FTy2->getNumParams() ||
        FTy1->isVarArg() != FTy2->isVarArg())
      return false;

    if (!isEquivalentType(FTy1->getReturnType(), FTy2->getReturnType()))
      return false;

    for (unsigned i = 0, e = FTy1->getNumParams(); i != e; ++i) {
      if (!isEquivalentType(FTy1->getParamType(i), FTy2->getParamType(i)))
        return false;
    }
    return true;
  }

  case Type::ArrayTyID: {
    ArrayType *ATy1 = cast<ArrayType>(Ty1);
    ArrayType *ATy2 = cast<ArrayType>(Ty2);
    return ATy1->getNumElements() == ATy2->getNumElements() &&
           isEquivalentType(ATy1->getElementType(), ATy2->getElementType());
  }
  }
}

// Determine whether the two operations are the same except that pointer-to-A
// and pointer-to-B are equivalent. This should be kept in sync with
// Instruction::isSameOperationAs.
bool FunctionComparator::isEquivalentOperation(const Instruction *I1,
                                               const Instruction *I2) const {
  // Differences from Instruction::isSameOperationAs:
  //  * replace type comparison with calls to isEquivalentType.
  //  * we test for I->hasSameSubclassOptionalData (nuw/nsw/tail) at the top
  //  * because of the above, we don't test for the tail bit on calls later on
  if (I1->getOpcode() != I2->getOpcode() ||
      I1->getNumOperands() != I2->getNumOperands() ||
      !isEquivalentType(I1->getType(), I2->getType()) ||
      !I1->hasSameSubclassOptionalData(I2))
    return false;

  // We have two instructions of identical opcode and #operands.  Check to see
  // if all operands are the same type
  for (unsigned i = 0, e = I1->getNumOperands(); i != e; ++i)
    if (!isEquivalentType(I1->getOperand(i)->getType(),
                          I2->getOperand(i)->getType()))
      return false;

  // Check special state that is a part of some instructions.
  if (const LoadInst *LI = dyn_cast<LoadInst>(I1))
    return LI->isVolatile() == cast<LoadInst>(I2)->isVolatile() &&
           LI->getAlignment() == cast<LoadInst>(I2)->getAlignment() &&
           LI->getOrdering() == cast<LoadInst>(I2)->getOrdering() &&
           LI->getSynchScope() == cast<LoadInst>(I2)->getSynchScope();
  if (const StoreInst *SI = dyn_cast<StoreInst>(I1))
    return SI->isVolatile() == cast<StoreInst>(I2)->isVolatile() &&
           SI->getAlignment() == cast<StoreInst>(I2)->getAlignment() &&
           SI->getOrdering() == cast<StoreInst>(I2)->getOrdering() &&
           SI->getSynchScope() == cast<StoreInst>(I2)->getSynchScope();
  if (const CmpInst *CI = dyn_cast<CmpInst>(I1))
    return CI->getPredicate() == cast<CmpInst>(I2)->getPredicate();
  if (const CallInst *CI = dyn_cast<CallInst>(I1))
    return CI->getCallingConv() == cast<CallInst>(I2)->getCallingConv() &&
           CI->getAttributes() == cast<CallInst>(I2)->getAttributes();
  if (const InvokeInst *CI = dyn_cast<InvokeInst>(I1))
    return CI->getCallingConv() == cast<InvokeInst>(I2)->getCallingConv() &&
           CI->getAttributes() == cast<InvokeInst>(I2)->getAttributes();
  if (const InsertValueInst *IVI = dyn_cast<InsertValueInst>(I1))
    return IVI->getIndices() == cast<InsertValueInst>(I2)->getIndices();
  if (const ExtractValueInst *EVI = dyn_cast<ExtractValueInst>(I1))
    return EVI->getIndices() == cast<ExtractValueInst>(I2)->getIndices();
  if (const FenceInst *FI = dyn_cast<FenceInst>(I1))
    return FI->getOrdering() == cast<FenceInst>(I2)->getOrdering() &&
           FI->getSynchScope() == cast<FenceInst>(I2)->getSynchScope();
  if (const AtomicCmpXchgInst *CXI = dyn_cast<AtomicCmpXchgInst>(I1))
    return CXI->isVolatile() == cast<AtomicCmpXchgInst>(I2)->isVolatile() &&
           CXI->getOrdering() == cast<AtomicCmpXchgInst>(I2)->getOrdering() &&
           CXI->getSynchScope() == cast<AtomicCmpXchgInst>(I2)->getSynchScope();
  if (const AtomicRMWInst *RMWI = dyn_cast<AtomicRMWInst>(I1))
    return RMWI->getOperation() == cast<AtomicRMWInst>(I2)->getOperation() &&
           RMWI->isVolatile() == cast<AtomicRMWInst>(I2)->isVolatile() &&
           RMWI->getOrdering() == cast<AtomicRMWInst>(I2)->getOrdering() &&
           RMWI->getSynchScope() == cast<AtomicRMWInst>(I2)->getSynchScope();

  return true;
}

// Determine whether two GEP operations perform the same underlying arithmetic.
bool FunctionComparator::isEquivalentGEP(const GEPOperator *GEP1,
                                         const GEPOperator *GEP2) {
  // When we have target data, we can reduce the GEP down to the value in bytes
  // added to the address.
  if (TD && GEP1->hasAllConstantIndices() && GEP2->hasAllConstantIndices()) {
    SmallVector<Value *, 8> Indices1(GEP1->idx_begin(), GEP1->idx_end());
    SmallVector<Value *, 8> Indices2(GEP2->idx_begin(), GEP2->idx_end());
    uint64_t Offset1 = TD->getIndexedOffset(GEP1->getPointerOperandType(),
                                            Indices1);
    uint64_t Offset2 = TD->getIndexedOffset(GEP2->getPointerOperandType(),
                                            Indices2);
    return Offset1 == Offset2;
  }

  if (GEP1->getPointerOperand()->getType() !=
      GEP2->getPointerOperand()->getType())
    return false;

  if (GEP1->getNumOperands() != GEP2->getNumOperands())
    return false;

  for (unsigned i = 0, e = GEP1->getNumOperands(); i != e; ++i) {
    if (!enumerate(GEP1->getOperand(i), GEP2->getOperand(i)))
      return false;
  }

  return true;
}

// Compare two values used by the two functions under pair-wise comparison. If
// this is the first time the values are seen, they're added to the mapping so
// that we will detect mismatches on next use.
bool FunctionComparator::enumerate(const Value *V1, const Value *V2) {
  // Check for function @f1 referring to itself and function @f2 referring to
  // itself, or referring to each other, or both referring to either of them.
  // They're all equivalent if the two functions are otherwise equivalent.
  if (V1 == F1 && V2 == F2)
    return true;
  if (V1 == F2 && V2 == F1)
    return true;

  if (const Constant *C1 = dyn_cast<Constant>(V1)) {
    if (V1 == V2) return true;
    const Constant *C2 = dyn_cast<Constant>(V2);
    if (!C2) return false;
    // TODO: constant expressions with GEP or references to F1 or F2.
    if (C1->isNullValue() && C2->isNullValue() &&
        isEquivalentType(C1->getType(), C2->getType()))
      return true;
    // Try bitcasting C2 to C1's type. If the bitcast is legal and returns C1
    // then they must have equal bit patterns.
    return C1->getType()->canLosslesslyBitCastTo(C2->getType()) &&
      C1 == ConstantExpr::getBitCast(const_cast<Constant*>(C2), C1->getType());
  }

  if (isa<InlineAsm>(V1) || isa<InlineAsm>(V2))
    return V1 == V2;

  // Check that V1 maps to V2. If we find a value that V1 maps to then we simply
  // check whether it's equal to V2. When there is no mapping then we need to
  // ensure that V2 isn't already equivalent to something else. For this
  // purpose, we track the V2 values in a set.

  const Value *&map_elem = id_map[V1];
  if (map_elem)
    return map_elem == V2;
  if (!seen_values.insert(V2).second)
    return false;
  map_elem = V2;
  return true;
}

// Test whether two basic blocks have equivalent behaviour.
bool FunctionComparator::compare(const BasicBlock *BB1, const BasicBlock *BB2) {
  BasicBlock::const_iterator F1I = BB1->begin(), F1E = BB1->end();
  BasicBlock::const_iterator F2I = BB2->begin(), F2E = BB2->end();

  do {
    if (!enumerate(F1I, F2I))
      return false;

    if (const GetElementPtrInst *GEP1 = dyn_cast<GetElementPtrInst>(F1I)) {
      const GetElementPtrInst *GEP2 = dyn_cast<GetElementPtrInst>(F2I);
      if (!GEP2)
        return false;

      if (!enumerate(GEP1->getPointerOperand(), GEP2->getPointerOperand()))
        return false;

      if (!isEquivalentGEP(GEP1, GEP2))
        return false;
    } else {
      if (!isEquivalentOperation(F1I, F2I))
        return false;

      assert(F1I->getNumOperands() == F2I->getNumOperands());
      for (unsigned i = 0, e = F1I->getNumOperands(); i != e; ++i) {
        Value *OpF1 = F1I->getOperand(i);
        Value *OpF2 = F2I->getOperand(i);

        if (!enumerate(OpF1, OpF2))
          return false;

        if (OpF1->getValueID() != OpF2->getValueID() ||
            !isEquivalentType(OpF1->getType(), OpF2->getType()))
          return false;
      }
    }

    ++F1I, ++F2I;
  } while (F1I != F1E && F2I != F2E);

  return F1I == F1E && F2I == F2E;
}

// Test whether the two functions have equivalent behaviour.
bool FunctionComparator::compare() {
  // We need to recheck everything, but check the things that weren't included
  // in the hash first.

  if (F1->getAttributes() != F2->getAttributes())
    return false;

  if (F1->hasGC() != F2->hasGC())
    return false;

  if (F1->hasGC() && F1->getGC() != F2->getGC())
    return false;

  if (F1->hasSection() != F2->hasSection())
    return false;

  if (F1->hasSection() && F1->getSection() != F2->getSection())
    return false;

  if (F1->isVarArg() != F2->isVarArg())
    return false;

  // TODO: if it's internal and only used in direct calls, we could handle this
  // case too.
  if (F1->getCallingConv() != F2->getCallingConv())
    return false;

  if (!isEquivalentType(F1->getFunctionType(), F2->getFunctionType()))
    return false;

  assert(F1->arg_size() == F2->arg_size() &&
         "Identically typed functions have different numbers of args!");

  // Visit the arguments so that they get enumerated in the order they're
  // passed in.
  for (Function::const_arg_iterator f1i = F1->arg_begin(),
         f2i = F2->arg_begin(), f1e = F1->arg_end(); f1i != f1e; ++f1i, ++f2i) {
    if (!enumerate(f1i, f2i))
      llvm_unreachable("Arguments repeat!");
  }

  // We do a CFG-ordered walk since the actual ordering of the blocks in the
  // linked list is immaterial. Our walk starts at the entry block for both
  // functions, then takes each block from each terminator in order. As an
  // artifact, this also means that unreachable blocks are ignored.
  SmallVector<const BasicBlock *, 8> F1BBs, F2BBs;
  SmallSet<const BasicBlock *, 128> VisitedBBs; // in terms of F1.

  F1BBs.push_back(&F1->getEntryBlock());
  F2BBs.push_back(&F2->getEntryBlock());

  VisitedBBs.insert(F1BBs[0]);
  while (!F1BBs.empty()) {
    const BasicBlock *F1BB = F1BBs.pop_back_val();
    const BasicBlock *F2BB = F2BBs.pop_back_val();

    if (!enumerate(F1BB, F2BB) || !compare(F1BB, F2BB))
      return false;

    const TerminatorInst *F1TI = F1BB->getTerminator();
    const TerminatorInst *F2TI = F2BB->getTerminator();

    assert(F1TI->getNumSuccessors() == F2TI->getNumSuccessors());
    for (unsigned i = 0, e = F1TI->getNumSuccessors(); i != e; ++i) {
      if (!VisitedBBs.insert(F1TI->getSuccessor(i)))
        continue;

      F1BBs.push_back(F1TI->getSuccessor(i));
      F2BBs.push_back(F2TI->getSuccessor(i));
    }
  }
  return true;
}

namespace {

/// MergeFunctions finds functions which will generate identical machine code,
/// by considering all pointer types to be equivalent. Once identified,
/// MergeFunctions will fold them by replacing a call to one to a call to a
/// bitcast of the other.
///
class MergeFunctions : public ModulePass {
public:
  static char ID;
  MergeFunctions()
    : ModulePass(ID), HasGlobalAliases(false) {
    initializeMergeFunctionsPass(*PassRegistry::getPassRegistry());
  }

  bool runOnModule(Module &M);

private:
  typedef DenseSet<ComparableFunction> FnSetType;

  /// A work queue of functions that may have been modified and should be
  /// analyzed again.
  std::vector<WeakVH> Deferred;

  /// Insert a ComparableFunction into the FnSet, or merge it away if it's
  /// equal to one that's already present.
  bool insert(ComparableFunction &NewF);

  /// Remove a Function from the FnSet and queue it up for a second sweep of
  /// analysis.
  void remove(Function *F);

  /// Find the functions that use this Value and remove them from FnSet and
  /// queue the functions.
  void removeUsers(Value *V);

  /// Replace all direct calls of Old with calls of New. Will bitcast New if
  /// necessary to make types match.
  void replaceDirectCallers(Function *Old, Function *New);

  /// Merge two equivalent functions. Upon completion, G may be deleted, or may
  /// be converted into a thunk. In either case, it should never be visited
  /// again.
  void mergeTwoFunctions(Function *F, Function *G);

  /// Replace G with a thunk or an alias to F. Deletes G.
  void writeThunkOrAlias(Function *F, Function *G);

  /// Replace G with a simple tail call to bitcast(F). Also replace direct uses
  /// of G with bitcast(F). Deletes G.
  void writeThunk(Function *F, Function *G);

  /// Replace G with an alias to F. Deletes G.
  void writeAlias(Function *F, Function *G);

  /// The set of all distinct functions. Use the insert() and remove() methods
  /// to modify it.
  FnSetType FnSet;

  /// DataLayout for more accurate GEP comparisons. May be NULL.
  DataLayout *TD;

  /// Whether or not the target supports global aliases.
  bool HasGlobalAliases;
};

}  // end anonymous namespace

char MergeFunctions::ID = 0;
INITIALIZE_PASS(MergeFunctions, "mergefunc", "Merge Functions", false, false)

ModulePass *llvm::createMergeFunctionsPass() {
  return new MergeFunctions();
}

bool MergeFunctions::runOnModule(Module &M) {
  bool Changed = false;
  TD = getAnalysisIfAvailable<DataLayout>();

  for (Module::iterator I = M.begin(), E = M.end(); I != E; ++I) {
    if (!I->isDeclaration() && !I->hasAvailableExternallyLinkage())
      Deferred.push_back(WeakVH(I));
  }
  FnSet.resize(Deferred.size());

  do {
    std::vector<WeakVH> Worklist;
    Deferred.swap(Worklist);

    DEBUG(dbgs() << "size of module: " << M.size() << '\n');
    DEBUG(dbgs() << "size of worklist: " << Worklist.size() << '\n');

    // Insert only strong functions and merge them. Strong function merging
    // always deletes one of them.
    for (std::vector<WeakVH>::iterator I = Worklist.begin(),
           E = Worklist.end(); I != E; ++I) {
      if (!*I) continue;
      Function *F = cast<Function>(*I);
      if (!F->isDeclaration() && !F->hasAvailableExternallyLinkage() &&
          !F->mayBeOverridden()) {
        ComparableFunction CF = ComparableFunction(F, TD);
        Changed |= insert(CF);
      }
    }

    // Insert only weak functions and merge them. By doing these second we
    // create thunks to the strong function when possible. When two weak
    // functions are identical, we create a new strong function with two weak
    // weak thunks to it which are identical but not mergable.
    for (std::vector<WeakVH>::iterator I = Worklist.begin(),
           E = Worklist.end(); I != E; ++I) {
      if (!*I) continue;
      Function *F = cast<Function>(*I);
      if (!F->isDeclaration() && !F->hasAvailableExternallyLinkage() &&
          F->mayBeOverridden()) {
        ComparableFunction CF = ComparableFunction(F, TD);
        Changed |= insert(CF);
      }
    }
    DEBUG(dbgs() << "size of FnSet: " << FnSet.size() << '\n');
  } while (!Deferred.empty());

  FnSet.clear();

  return Changed;
}

bool DenseMapInfo<ComparableFunction>::isEqual(const ComparableFunction &LHS,
                                               const ComparableFunction &RHS) {
  if (LHS.getFunc() == RHS.getFunc() &&
      LHS.getHash() == RHS.getHash())
    return true;
  if (!LHS.getFunc() || !RHS.getFunc())
    return false;

  // One of these is a special "underlying pointer comparison only" object.
  if (LHS.getTD() == ComparableFunction::LookupOnly ||
      RHS.getTD() == ComparableFunction::LookupOnly)
    return false;

  assert(LHS.getTD() == RHS.getTD() &&
         "Comparing functions for different targets");

  return FunctionComparator(LHS.getTD(), LHS.getFunc(),
                            RHS.getFunc()).compare();
}

// Replace direct callers of Old with New.
void MergeFunctions::replaceDirectCallers(Function *Old, Function *New) {
  Constant *BitcastNew = ConstantExpr::getBitCast(New, Old->getType());
  for (Value::use_iterator UI = Old->use_begin(), UE = Old->use_end();
       UI != UE;) {
    Value::use_iterator TheIter = UI;
    ++UI;
    CallSite CS(*TheIter);
    if (CS && CS.isCallee(TheIter)) {
      remove(CS.getInstruction()->getParent()->getParent());
      TheIter.getUse().set(BitcastNew);
    }
  }
}

// Replace G with an alias to F if possible, or else a thunk to F. Deletes G.
void MergeFunctions::writeThunkOrAlias(Function *F, Function *G) {
  if (HasGlobalAliases && G->hasUnnamedAddr()) {
    if (G->hasExternalLinkage() || G->hasLocalLinkage() ||
        G->hasWeakLinkage()) {
      writeAlias(F, G);
      return;
    }
  }

  writeThunk(F, G);
}

// Replace G with a simple tail call to bitcast(F). Also replace direct uses
// of G with bitcast(F). Deletes G.
void MergeFunctions::writeThunk(Function *F, Function *G) {
  if (!G->mayBeOverridden()) {
    // Redirect direct callers of G to F.
    replaceDirectCallers(G, F);
  }

  // If G was internal then we may have replaced all uses of G with F. If so,
  // stop here and delete G. There's no need for a thunk.
  if (G->hasLocalLinkage() && G->use_empty()) {
    G->eraseFromParent();
    return;
  }

  Function *NewG = Function::Create(G->getFunctionType(), G->getLinkage(), "",
                                    G->getParent());
  BasicBlock *BB = BasicBlock::Create(F->getContext(), "", NewG);
  IRBuilder<false> Builder(BB);

  SmallVector<Value *, 16> Args;
  unsigned i = 0;
  FunctionType *FFTy = F->getFunctionType();
  for (Function::arg_iterator AI = NewG->arg_begin(), AE = NewG->arg_end();
       AI != AE; ++AI) {
    Args.push_back(Builder.CreateBitCast(AI, FFTy->getParamType(i)));
    ++i;
  }

  CallInst *CI = Builder.CreateCall(F, Args);
  CI->setTailCall();
  CI->setCallingConv(F->getCallingConv());
  if (NewG->getReturnType()->isVoidTy()) {
    Builder.CreateRetVoid();
  } else {
    Builder.CreateRet(Builder.CreateBitCast(CI, NewG->getReturnType()));
  }

  NewG->copyAttributesFrom(G);
  NewG->takeName(G);
  removeUsers(G);
  G->replaceAllUsesWith(NewG);
  G->eraseFromParent();

  DEBUG(dbgs() << "writeThunk: " << NewG->getName() << '\n');
  ++NumThunksWritten;
}

// Replace G with an alias to F and delete G.
void MergeFunctions::writeAlias(Function *F, Function *G) {
  Constant *BitcastF = ConstantExpr::getBitCast(F, G->getType());
  GlobalAlias *GA = new GlobalAlias(G->getType(), G->getLinkage(), "",
                                    BitcastF, G->getParent());
  F->setAlignment(std::max(F->getAlignment(), G->getAlignment()));
  GA->takeName(G);
  GA->setVisibility(G->getVisibility());
  removeUsers(G);
  G->replaceAllUsesWith(GA);
  G->eraseFromParent();

  DEBUG(dbgs() << "writeAlias: " << GA->getName() << '\n');
  ++NumAliasesWritten;
}

// Merge two equivalent functions. Upon completion, Function G is deleted.
void MergeFunctions::mergeTwoFunctions(Function *F, Function *G) {
  if (F->mayBeOverridden()) {
    assert(G->mayBeOverridden());

    if (HasGlobalAliases) {
      // Make them both thunks to the same internal function.
      Function *H = Function::Create(F->getFunctionType(), F->getLinkage(), "",
                                     F->getParent());
      H->copyAttributesFrom(F);
      H->takeName(F);
      removeUsers(F);
      F->replaceAllUsesWith(H);

      unsigned MaxAlignment = std::max(G->getAlignment(), H->getAlignment());

      writeAlias(F, G);
      writeAlias(F, H);

      F->setAlignment(MaxAlignment);
      F->setLinkage(GlobalValue::PrivateLinkage);
    } else {
      // We can't merge them. Instead, pick one and update all direct callers
      // to call it and hope that we improve the instruction cache hit rate.
      replaceDirectCallers(G, F);
    }

    ++NumDoubleWeak;
  } else {
    writeThunkOrAlias(F, G);
  }

  ++NumFunctionsMerged;
}

// Insert a ComparableFunction into the FnSet, or merge it away if equal to one
// that was already inserted.
bool MergeFunctions::insert(ComparableFunction &NewF) {
  std::pair<FnSetType::iterator, bool> Result = FnSet.insert(NewF);
  if (Result.second) {
    DEBUG(dbgs() << "Inserting as unique: " << NewF.getFunc()->getName() << '\n');
    return false;
  }

  const ComparableFunction &OldF = *Result.first;

  // Never thunk a strong function to a weak function.
  assert(!OldF.getFunc()->mayBeOverridden() ||
         NewF.getFunc()->mayBeOverridden());

  DEBUG(dbgs() << "  " << OldF.getFunc()->getName() << " == "
               << NewF.getFunc()->getName() << '\n');

  Function *DeleteF = NewF.getFunc();
  NewF.release();
  mergeTwoFunctions(OldF.getFunc(), DeleteF);
  return true;
}

// Remove a function from FnSet. If it was already in FnSet, add it to Deferred
// so that we'll look at it in the next round.
void MergeFunctions::remove(Function *F) {
  // We need to make sure we remove F, not a function "equal" to F per the
  // function equality comparator.
  //
  // The special "lookup only" ComparableFunction bypasses the expensive
  // function comparison in favour of a pointer comparison on the underlying
  // Function*'s.
  ComparableFunction CF = ComparableFunction(F, ComparableFunction::LookupOnly);
  if (FnSet.erase(CF)) {
    DEBUG(dbgs() << "Removed " << F->getName() << " from set and deferred it.\n");
    Deferred.push_back(F);
  }
}

// For each instruction used by the value, remove() the function that contains
// the instruction. This should happen right before a call to RAUW.
void MergeFunctions::removeUsers(Value *V) {
  std::vector<Value *> Worklist;
  Worklist.push_back(V);
  while (!Worklist.empty()) {
    Value *V = Worklist.back();
    Worklist.pop_back();

    for (Value::use_iterator UI = V->use_begin(), UE = V->use_end();
         UI != UE; ++UI) {
      Use &U = UI.getUse();
      if (Instruction *I = dyn_cast<Instruction>(U.getUser())) {
        remove(I->getParent()->getParent());
      } else if (isa<GlobalValue>(U.getUser())) {
        // do nothing
      } else if (Constant *C = dyn_cast<Constant>(U.getUser())) {
        for (Value::use_iterator CUI = C->use_begin(), CUE = C->use_end();
             CUI != CUE; ++CUI)
          Worklist.push_back(*CUI);
      }
    }
  }
}