lib/Transforms/IPO/MergeFunctions.cpp

//===- 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/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/IR/Constants.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/IR/IRBuilder.h"
#include "llvm/IR/InlineAsm.h"
#include "llvm/IR/Instructions.h"
#include "llvm/IR/LLVMContext.h"
#include "llvm/IR/Module.h"
#include "llvm/IR/Operator.h"
#include "llvm/Pass.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 <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");

/// Returns the type id for a type to be hashed. We turn pointer types into
/// integers here because the actual compare logic below considers pointers and
/// integers of the same size as equal.
static Type::TypeID getTypeIDForHash(Type *Ty) {
  if (Ty->isPointerTy())
    return Type::IntegerTyID;
  return Ty->getTypeID();
}

/// 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(getTypeIDForHash(FTy->getReturnType()));
  for (unsigned i = 0, e = FTy->getNumParams(); i != e; ++i)
    ID.AddInteger(getTypeIDForHash(FTy->getParamType(i)));
  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 {

  PointerType *PTy1 = dyn_cast<PointerType>(Ty1);
  PointerType *PTy2 = dyn_cast<PointerType>(Ty2);

  if (TD) {
    if (PTy1 && PTy1->getAddressSpace() == 0) Ty1 = TD->getIntPtrType(Ty1);
    if (PTy2 && PTy2->getAddressSpace() == 0) Ty2 = TD->getIntPtrType(Ty2);
  }

  if (Ty1 == Ty2)
    return true;

  if (Ty1->getTypeID() != Ty2->getTypeID())
    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: {
    assert(PTy1 && PTy2 && "Both types must be pointers here.");
    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) {
  unsigned AS = GEP1->getPointerAddressSpace();
  if (AS != GEP2->getPointerAddressSpace())
    return false;

  if (TD) {
    // When we have target data, we can reduce the GEP down to the value in bytes
    // added to the address.
    unsigned BitWidth = TD ? TD->getPointerSizeInBits(AS) : 1;
    APInt Offset1(BitWidth, 0), Offset2(BitWidth, 0);
    if (GEP1->accumulateConstantOffset(*TD, Offset1) &&
        GEP2->accumulateConstantOffset(*TD, Offset2)) {
      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);
}

// Helper for writeThunk,
// Selects proper bitcast operation,
// but a bit simplier then CastInst::getCastOpcode.
static Value* createCast(IRBuilder<false> &Builder, Value *V, Type *DestTy) {
  Type *SrcTy = V->getType();
  if (SrcTy->isIntegerTy() && DestTy->isPointerTy())
    return Builder.CreateIntToPtr(V, DestTy);
  else if (SrcTy->isPointerTy() && DestTy->isIntegerTy())
    return Builder.CreatePtrToInt(V, DestTy);
  else
    return Builder.CreateBitCast(V, DestTy);
}

// 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(createCast(Builder, (Value*)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(createCast(Builder, 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;

  // Don't merge tiny functions, since it can just end up making the function
  // larger.
  // FIXME: Should still merge them if they are unnamed_addr and produce an
  // alias.
  if (NewF.getFunc()->size() == 1) {
    if (NewF.getFunc()->front().size() <= 2) {
      DEBUG(dbgs() << NewF.getFunc()->getName()
            << " is to small to bother merging\n");
      return false;
    }
  }

  // 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);
      }
    }
  }
}