MemCpyOptimizer.cpp [plain text]
#include "llvm/Transforms/Scalar.h"
#include "llvm/ADT/SmallVector.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/Analysis/AliasAnalysis.h"
#include "llvm/Analysis/MemoryDependenceAnalysis.h"
#include "llvm/Analysis/ValueTracking.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/IR/Dominators.h"
#include "llvm/IR/GetElementPtrTypeIterator.h"
#include "llvm/IR/GlobalVariable.h"
#include "llvm/IR/IRBuilder.h"
#include "llvm/IR/Instructions.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Target/TargetLibraryInfo.h"
#include "llvm/Transforms/Utils/Local.h"
#include <list>
using namespace llvm;
#define DEBUG_TYPE "memcpyopt"
STATISTIC(NumMemCpyInstr, "Number of memcpy instructions deleted");
STATISTIC(NumMemSetInfer, "Number of memsets inferred");
STATISTIC(NumMoveToCpy, "Number of memmoves converted to memcpy");
STATISTIC(NumCpyToSet, "Number of memcpys converted to memset");
static int64_t GetOffsetFromIndex(const GEPOperator *GEP, unsigned Idx,
bool &VariableIdxFound, const DataLayout &TD){
gep_type_iterator GTI = gep_type_begin(GEP);
for (unsigned i = 1; i != Idx; ++i, ++GTI)
;
int64_t Offset = 0;
for (unsigned i = Idx, e = GEP->getNumOperands(); i != e; ++i, ++GTI) {
ConstantInt *OpC = dyn_cast<ConstantInt>(GEP->getOperand(i));
if (!OpC)
return VariableIdxFound = true;
if (OpC->isZero()) continue;
if (StructType *STy = dyn_cast<StructType>(*GTI)) {
Offset += TD.getStructLayout(STy)->getElementOffset(OpC->getZExtValue());
continue;
}
uint64_t Size = TD.getTypeAllocSize(GTI.getIndexedType());
Offset += Size*OpC->getSExtValue();
}
return Offset;
}
static bool IsPointerOffset(Value *Ptr1, Value *Ptr2, int64_t &Offset,
const DataLayout &TD) {
Ptr1 = Ptr1->stripPointerCasts();
Ptr2 = Ptr2->stripPointerCasts();
if (Ptr1 == Ptr2) {
Offset = 0;
return true;
}
GEPOperator *GEP1 = dyn_cast<GEPOperator>(Ptr1);
GEPOperator *GEP2 = dyn_cast<GEPOperator>(Ptr2);
bool VariableIdxFound = false;
if (GEP1 && !GEP2 && GEP1->getOperand(0)->stripPointerCasts() == Ptr2) {
Offset = -GetOffsetFromIndex(GEP1, 1, VariableIdxFound, TD);
return !VariableIdxFound;
}
if (GEP2 && !GEP1 && GEP2->getOperand(0)->stripPointerCasts() == Ptr1) {
Offset = GetOffsetFromIndex(GEP2, 1, VariableIdxFound, TD);
return !VariableIdxFound;
}
if (!GEP1 || !GEP2 || GEP1->getOperand(0) != GEP2->getOperand(0))
return false;
unsigned Idx = 1;
for (; Idx != GEP1->getNumOperands() && Idx != GEP2->getNumOperands(); ++Idx)
if (GEP1->getOperand(Idx) != GEP2->getOperand(Idx))
break;
int64_t Offset1 = GetOffsetFromIndex(GEP1, Idx, VariableIdxFound, TD);
int64_t Offset2 = GetOffsetFromIndex(GEP2, Idx, VariableIdxFound, TD);
if (VariableIdxFound) return false;
Offset = Offset2-Offset1;
return true;
}
namespace {
struct MemsetRange {
int64_t Start, End;
Value *StartPtr;
unsigned Alignment;
SmallVector<Instruction*, 16> TheStores;
bool isProfitableToUseMemset(const DataLayout &TD) const;
};
}
bool MemsetRange::isProfitableToUseMemset(const DataLayout &TD) const {
if (TheStores.size() >= 4 || End-Start >= 16) return true;
if (TheStores.size() < 2) return false;
for (unsigned i = 0, e = TheStores.size(); i != e; ++i)
if (!isa<StoreInst>(TheStores[i]))
return true;
if (TheStores.size() == 2) return false;
unsigned Bytes = unsigned(End-Start);
unsigned MaxIntSize = TD.getLargestLegalIntTypeSize();
if (MaxIntSize == 0)
MaxIntSize = 1;
unsigned NumPointerStores = Bytes / MaxIntSize;
unsigned NumByteStores = Bytes - NumPointerStores * MaxIntSize;
return TheStores.size() > NumPointerStores+NumByteStores;
}
namespace {
class MemsetRanges {
std::list<MemsetRange> Ranges;
typedef std::list<MemsetRange>::iterator range_iterator;
const DataLayout &DL;
public:
MemsetRanges(const DataLayout &DL) : DL(DL) {}
typedef std::list<MemsetRange>::const_iterator const_iterator;
const_iterator begin() const { return Ranges.begin(); }
const_iterator end() const { return Ranges.end(); }
bool empty() const { return Ranges.empty(); }
void addInst(int64_t OffsetFromFirst, Instruction *Inst) {
if (StoreInst *SI = dyn_cast<StoreInst>(Inst))
addStore(OffsetFromFirst, SI);
else
addMemSet(OffsetFromFirst, cast<MemSetInst>(Inst));
}
void addStore(int64_t OffsetFromFirst, StoreInst *SI) {
int64_t StoreSize = DL.getTypeStoreSize(SI->getOperand(0)->getType());
addRange(OffsetFromFirst, StoreSize,
SI->getPointerOperand(), SI->getAlignment(), SI);
}
void addMemSet(int64_t OffsetFromFirst, MemSetInst *MSI) {
int64_t Size = cast<ConstantInt>(MSI->getLength())->getZExtValue();
addRange(OffsetFromFirst, Size, MSI->getDest(), MSI->getAlignment(), MSI);
}
void addRange(int64_t Start, int64_t Size, Value *Ptr,
unsigned Alignment, Instruction *Inst);
};
}
void MemsetRanges::addRange(int64_t Start, int64_t Size, Value *Ptr,
unsigned Alignment, Instruction *Inst) {
int64_t End = Start+Size;
range_iterator I = Ranges.begin(), E = Ranges.end();
while (I != E && Start > I->End)
++I;
if (I == E || End < I->Start) {
MemsetRange &R = *Ranges.insert(I, MemsetRange());
R.Start = Start;
R.End = End;
R.StartPtr = Ptr;
R.Alignment = Alignment;
R.TheStores.push_back(Inst);
return;
}
I->TheStores.push_back(Inst);
if (I->Start <= Start && I->End >= End)
return;
if (Start < I->Start) {
I->Start = Start;
I->StartPtr = Ptr;
I->Alignment = Alignment;
}
if (End > I->End) {
I->End = End;
range_iterator NextI = I;
while (++NextI != E && End >= NextI->Start) {
I->TheStores.append(NextI->TheStores.begin(), NextI->TheStores.end());
if (NextI->End > I->End)
I->End = NextI->End;
Ranges.erase(NextI);
NextI = I;
}
}
}
namespace {
class MemCpyOpt : public FunctionPass {
MemoryDependenceAnalysis *MD;
TargetLibraryInfo *TLI;
const DataLayout *DL;
public:
static char ID; MemCpyOpt() : FunctionPass(ID) {
initializeMemCpyOptPass(*PassRegistry::getPassRegistry());
MD = nullptr;
TLI = nullptr;
DL = nullptr;
}
bool runOnFunction(Function &F) override;
private:
void getAnalysisUsage(AnalysisUsage &AU) const override {
AU.setPreservesCFG();
AU.addRequired<DominatorTreeWrapperPass>();
AU.addRequired<MemoryDependenceAnalysis>();
AU.addRequired<AliasAnalysis>();
AU.addRequired<TargetLibraryInfo>();
AU.addPreserved<AliasAnalysis>();
AU.addPreserved<MemoryDependenceAnalysis>();
}
bool processStore(StoreInst *SI, BasicBlock::iterator &BBI);
bool processMemSet(MemSetInst *SI, BasicBlock::iterator &BBI);
bool processMemCpy(MemCpyInst *M);
bool processMemMove(MemMoveInst *M);
bool performCallSlotOptzn(Instruction *cpy, Value *cpyDst, Value *cpySrc,
uint64_t cpyLen, unsigned cpyAlign, CallInst *C);
bool processMemCpyMemCpyDependence(MemCpyInst *M, MemCpyInst *MDep,
uint64_t MSize);
bool processByValArgument(CallSite CS, unsigned ArgNo);
Instruction *tryMergingIntoMemset(Instruction *I, Value *StartPtr,
Value *ByteVal);
bool iterateOnFunction(Function &F);
};
char MemCpyOpt::ID = 0;
}
FunctionPass *llvm::createMemCpyOptPass() { return new MemCpyOpt(); }
INITIALIZE_PASS_BEGIN(MemCpyOpt, "memcpyopt", "MemCpy Optimization",
false, false)
INITIALIZE_PASS_DEPENDENCY(DominatorTreeWrapperPass)
INITIALIZE_PASS_DEPENDENCY(MemoryDependenceAnalysis)
INITIALIZE_PASS_DEPENDENCY(TargetLibraryInfo)
INITIALIZE_AG_DEPENDENCY(AliasAnalysis)
INITIALIZE_PASS_END(MemCpyOpt, "memcpyopt", "MemCpy Optimization",
false, false)
Instruction *MemCpyOpt::tryMergingIntoMemset(Instruction *StartInst,
Value *StartPtr, Value *ByteVal) {
if (!DL) return nullptr;
MemsetRanges Ranges(*DL);
BasicBlock::iterator BI = StartInst;
for (++BI; !isa<TerminatorInst>(BI); ++BI) {
if (!isa<StoreInst>(BI) && !isa<MemSetInst>(BI)) {
if (BI->mayWriteToMemory() || BI->mayReadFromMemory())
break;
continue;
}
if (StoreInst *NextStore = dyn_cast<StoreInst>(BI)) {
if (!NextStore->isSimple()) break;
if (ByteVal != isBytewiseValue(NextStore->getOperand(0)))
break;
int64_t Offset;
if (!IsPointerOffset(StartPtr, NextStore->getPointerOperand(),
Offset, *DL))
break;
Ranges.addStore(Offset, NextStore);
} else {
MemSetInst *MSI = cast<MemSetInst>(BI);
if (MSI->isVolatile() || ByteVal != MSI->getValue() ||
!isa<ConstantInt>(MSI->getLength()))
break;
int64_t Offset;
if (!IsPointerOffset(StartPtr, MSI->getDest(), Offset, *DL))
break;
Ranges.addMemSet(Offset, MSI);
}
}
if (Ranges.empty())
return nullptr;
Ranges.addInst(0, StartInst);
IRBuilder<> Builder(BI);
Instruction *AMemSet = nullptr;
for (MemsetRanges::const_iterator I = Ranges.begin(), E = Ranges.end();
I != E; ++I) {
const MemsetRange &Range = *I;
if (Range.TheStores.size() == 1) continue;
if (!Range.isProfitableToUseMemset(*DL))
continue;
StartPtr = Range.StartPtr;
unsigned Alignment = Range.Alignment;
if (Alignment == 0) {
Type *EltType =
cast<PointerType>(StartPtr->getType())->getElementType();
Alignment = DL->getABITypeAlignment(EltType);
}
AMemSet =
Builder.CreateMemSet(StartPtr, ByteVal, Range.End-Range.Start, Alignment);
DEBUG(dbgs() << "Replace stores:\n";
for (unsigned i = 0, e = Range.TheStores.size(); i != e; ++i)
dbgs() << *Range.TheStores[i] << '\n';
dbgs() << "With: " << *AMemSet << '\n');
if (!Range.TheStores.empty())
AMemSet->setDebugLoc(Range.TheStores[0]->getDebugLoc());
for (SmallVectorImpl<Instruction *>::const_iterator
SI = Range.TheStores.begin(),
SE = Range.TheStores.end(); SI != SE; ++SI) {
MD->removeInstruction(*SI);
(*SI)->eraseFromParent();
}
++NumMemSetInfer;
}
return AMemSet;
}
bool MemCpyOpt::processStore(StoreInst *SI, BasicBlock::iterator &BBI) {
if (!SI->isSimple()) return false;
if (!DL) return false;
if (LoadInst *LI = dyn_cast<LoadInst>(SI->getOperand(0))) {
if (LI->isSimple() && LI->hasOneUse() &&
LI->getParent() == SI->getParent()) {
MemDepResult ldep = MD->getDependency(LI);
CallInst *C = nullptr;
if (ldep.isClobber() && !isa<MemCpyInst>(ldep.getInst()))
C = dyn_cast<CallInst>(ldep.getInst());
if (C) {
AliasAnalysis &AA = getAnalysis<AliasAnalysis>();
AliasAnalysis::Location StoreLoc = AA.getLocation(SI);
for (BasicBlock::iterator I = --BasicBlock::iterator(SI),
E = C; I != E; --I) {
if (AA.getModRefInfo(&*I, StoreLoc) != AliasAnalysis::NoModRef) {
C = nullptr;
break;
}
}
}
if (C) {
unsigned storeAlign = SI->getAlignment();
if (!storeAlign)
storeAlign = DL->getABITypeAlignment(SI->getOperand(0)->getType());
unsigned loadAlign = LI->getAlignment();
if (!loadAlign)
loadAlign = DL->getABITypeAlignment(LI->getType());
bool changed = performCallSlotOptzn(LI,
SI->getPointerOperand()->stripPointerCasts(),
LI->getPointerOperand()->stripPointerCasts(),
DL->getTypeStoreSize(SI->getOperand(0)->getType()),
std::min(storeAlign, loadAlign), C);
if (changed) {
MD->removeInstruction(SI);
SI->eraseFromParent();
MD->removeInstruction(LI);
LI->eraseFromParent();
++NumMemCpyInstr;
return true;
}
}
}
}
if (Value *ByteVal = isBytewiseValue(SI->getOperand(0)))
if (Instruction *I = tryMergingIntoMemset(SI, SI->getPointerOperand(),
ByteVal)) {
BBI = I; return true;
}
return false;
}
bool MemCpyOpt::processMemSet(MemSetInst *MSI, BasicBlock::iterator &BBI) {
if (isa<ConstantInt>(MSI->getLength()) && !MSI->isVolatile())
if (Instruction *I = tryMergingIntoMemset(MSI, MSI->getDest(),
MSI->getValue())) {
BBI = I; return true;
}
return false;
}
bool MemCpyOpt::performCallSlotOptzn(Instruction *cpy,
Value *cpyDest, Value *cpySrc,
uint64_t cpyLen, unsigned cpyAlign,
CallInst *C) {
CallSite CS(C);
AllocaInst *srcAlloca = dyn_cast<AllocaInst>(cpySrc);
if (!srcAlloca)
return false;
if (!DL) return false;
ConstantInt *srcArraySize = dyn_cast<ConstantInt>(srcAlloca->getArraySize());
if (!srcArraySize)
return false;
uint64_t srcSize = DL->getTypeAllocSize(srcAlloca->getAllocatedType()) *
srcArraySize->getZExtValue();
if (cpyLen < srcSize)
return false;
if (AllocaInst *A = dyn_cast<AllocaInst>(cpyDest)) {
ConstantInt *destArraySize = dyn_cast<ConstantInt>(A->getArraySize());
if (!destArraySize)
return false;
uint64_t destSize = DL->getTypeAllocSize(A->getAllocatedType()) *
destArraySize->getZExtValue();
if (destSize < srcSize)
return false;
} else if (Argument *A = dyn_cast<Argument>(cpyDest)) {
if (!A->hasStructRetAttr())
return false;
Type *StructTy = cast<PointerType>(A->getType())->getElementType();
if (!StructTy->isSized()) {
return false;
}
uint64_t destSize = DL->getTypeAllocSize(StructTy);
if (destSize < srcSize)
return false;
} else {
return false;
}
unsigned srcAlign = srcAlloca->getAlignment();
if (!srcAlign)
srcAlign = DL->getABITypeAlignment(srcAlloca->getAllocatedType());
bool isDestSufficientlyAligned = srcAlign <= cpyAlign;
if (!isDestSufficientlyAligned && !isa<AllocaInst>(cpyDest))
return false;
SmallVector<User*, 8> srcUseList(srcAlloca->user_begin(),
srcAlloca->user_end());
while (!srcUseList.empty()) {
User *U = srcUseList.pop_back_val();
if (isa<BitCastInst>(U) || isa<AddrSpaceCastInst>(U)) {
for (User *UU : U->users())
srcUseList.push_back(UU);
} else if (GetElementPtrInst *G = dyn_cast<GetElementPtrInst>(U)) {
if (G->hasAllZeroIndices())
for (User *UU : U->users())
srcUseList.push_back(UU);
else
return false;
} else if (U != C && U != cpy) {
return false;
}
}
for (unsigned i = 0, e = CS.arg_size(); i != e; ++i)
if (CS.getArgument(i) == cpySrc && !CS.doesNotCapture(i))
return false;
DominatorTree &DT = getAnalysis<DominatorTreeWrapperPass>().getDomTree();
if (Instruction *cpyDestInst = dyn_cast<Instruction>(cpyDest))
if (!DT.dominates(cpyDestInst, C))
return false;
AliasAnalysis &AA = getAnalysis<AliasAnalysis>();
AliasAnalysis::ModRefResult MR = AA.getModRefInfo(C, cpyDest, srcSize);
if (MR != AliasAnalysis::NoModRef)
MR = AA.callCapturesBefore(C, cpyDest, srcSize, &DT);
if (MR != AliasAnalysis::NoModRef)
return false;
bool changedArgument = false;
for (unsigned i = 0; i < CS.arg_size(); ++i)
if (CS.getArgument(i)->stripPointerCasts() == cpySrc) {
Value *Dest = cpySrc->getType() == cpyDest->getType() ? cpyDest
: CastInst::CreatePointerCast(cpyDest, cpySrc->getType(),
cpyDest->getName(), C);
changedArgument = true;
if (CS.getArgument(i)->getType() == Dest->getType())
CS.setArgument(i, Dest);
else
CS.setArgument(i, CastInst::CreatePointerCast(Dest,
CS.getArgument(i)->getType(), Dest->getName(), C));
}
if (!changedArgument)
return false;
if (!isDestSufficientlyAligned) {
assert(isa<AllocaInst>(cpyDest) && "Can only increase alloca alignment!");
cast<AllocaInst>(cpyDest)->setAlignment(srcAlign);
}
MD->removeInstruction(C);
MD->removeInstruction(cpy);
++NumMemCpyInstr;
return true;
}
bool MemCpyOpt::processMemCpyMemCpyDependence(MemCpyInst *M, MemCpyInst *MDep,
uint64_t MSize) {
if (M->getSource() != MDep->getDest() || MDep->isVolatile())
return false;
if (M->getSource() == MDep->getSource())
return false;
ConstantInt *MDepLen = dyn_cast<ConstantInt>(MDep->getLength());
ConstantInt *MLen = dyn_cast<ConstantInt>(M->getLength());
if (!MDepLen || !MLen || MDepLen->getZExtValue() < MLen->getZExtValue())
return false;
AliasAnalysis &AA = getAnalysis<AliasAnalysis>();
MemDepResult SourceDep =
MD->getPointerDependencyFrom(AA.getLocationForSource(MDep),
false, M, M->getParent());
if (!SourceDep.isClobber() || SourceDep.getInst() != MDep)
return false;
bool UseMemMove = false;
if (!AA.isNoAlias(AA.getLocationForDest(M), AA.getLocationForSource(MDep)))
UseMemMove = true;
unsigned Align = std::min(MDep->getAlignment(), M->getAlignment());
IRBuilder<> Builder(M);
if (UseMemMove)
Builder.CreateMemMove(M->getRawDest(), MDep->getRawSource(), M->getLength(),
Align, M->isVolatile());
else
Builder.CreateMemCpy(M->getRawDest(), MDep->getRawSource(), M->getLength(),
Align, M->isVolatile());
MD->removeInstruction(M);
M->eraseFromParent();
++NumMemCpyInstr;
return true;
}
bool MemCpyOpt::processMemCpy(MemCpyInst *M) {
if (M->isVolatile()) return false;
if (M->getSource() == M->getDest()) {
MD->removeInstruction(M);
M->eraseFromParent();
return false;
}
if (GlobalVariable *GV = dyn_cast<GlobalVariable>(M->getSource()))
if (GV->isConstant() && GV->hasDefinitiveInitializer())
if (Value *ByteVal = isBytewiseValue(GV->getInitializer())) {
IRBuilder<> Builder(M);
Builder.CreateMemSet(M->getRawDest(), ByteVal, M->getLength(),
M->getAlignment(), false);
MD->removeInstruction(M);
M->eraseFromParent();
++NumCpyToSet;
return true;
}
ConstantInt *CopySize = dyn_cast<ConstantInt>(M->getLength());
if (!CopySize) return false;
MemDepResult DepInfo = MD->getDependency(M);
if (DepInfo.isClobber()) {
if (CallInst *C = dyn_cast<CallInst>(DepInfo.getInst())) {
if (performCallSlotOptzn(M, M->getDest(), M->getSource(),
CopySize->getZExtValue(), M->getAlignment(),
C)) {
MD->removeInstruction(M);
M->eraseFromParent();
return true;
}
}
}
AliasAnalysis::Location SrcLoc = AliasAnalysis::getLocationForSource(M);
MemDepResult SrcDepInfo = MD->getPointerDependencyFrom(SrcLoc, true,
M, M->getParent());
if (SrcDepInfo.isClobber()) {
if (MemCpyInst *MDep = dyn_cast<MemCpyInst>(SrcDepInfo.getInst()))
return processMemCpyMemCpyDependence(M, MDep, CopySize->getZExtValue());
} else if (SrcDepInfo.isDef()) {
Instruction *I = SrcDepInfo.getInst();
bool hasUndefContents = false;
if (isa<AllocaInst>(I)) {
hasUndefContents = true;
} else if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(I)) {
if (II->getIntrinsicID() == Intrinsic::lifetime_start)
if (ConstantInt *LTSize = dyn_cast<ConstantInt>(II->getArgOperand(0)))
if (LTSize->getZExtValue() >= CopySize->getZExtValue())
hasUndefContents = true;
}
if (hasUndefContents) {
MD->removeInstruction(M);
M->eraseFromParent();
++NumMemCpyInstr;
return true;
}
}
return false;
}
bool MemCpyOpt::processMemMove(MemMoveInst *M) {
AliasAnalysis &AA = getAnalysis<AliasAnalysis>();
if (!TLI->has(LibFunc::memmove))
return false;
if (!AA.isNoAlias(AA.getLocationForDest(M), AA.getLocationForSource(M)))
return false;
DEBUG(dbgs() << "MemCpyOpt: Optimizing memmove -> memcpy: " << *M << "\n");
Module *Mod = M->getParent()->getParent()->getParent();
Type *ArgTys[3] = { M->getRawDest()->getType(),
M->getRawSource()->getType(),
M->getLength()->getType() };
M->setCalledFunction(Intrinsic::getDeclaration(Mod, Intrinsic::memcpy,
ArgTys));
MD->removeInstruction(M);
++NumMoveToCpy;
return true;
}
bool MemCpyOpt::processByValArgument(CallSite CS, unsigned ArgNo) {
if (!DL) return false;
Value *ByValArg = CS.getArgument(ArgNo);
Type *ByValTy = cast<PointerType>(ByValArg->getType())->getElementType();
uint64_t ByValSize = DL->getTypeAllocSize(ByValTy);
MemDepResult DepInfo =
MD->getPointerDependencyFrom(AliasAnalysis::Location(ByValArg, ByValSize),
true, CS.getInstruction(),
CS.getInstruction()->getParent());
if (!DepInfo.isClobber())
return false;
MemCpyInst *MDep = dyn_cast<MemCpyInst>(DepInfo.getInst());
if (!MDep || MDep->isVolatile() ||
ByValArg->stripPointerCasts() != MDep->getDest())
return false;
ConstantInt *C1 = dyn_cast<ConstantInt>(MDep->getLength());
if (!C1 || C1->getValue().getZExtValue() < ByValSize)
return false;
unsigned ByValAlign = CS.getParamAlignment(ArgNo+1);
if (ByValAlign == 0) return false;
if (MDep->getAlignment() < ByValAlign &&
getOrEnforceKnownAlignment(MDep->getSource(),ByValAlign, DL) < ByValAlign)
return false;
MemDepResult SourceDep =
MD->getPointerDependencyFrom(AliasAnalysis::getLocationForSource(MDep),
false, CS.getInstruction(), MDep->getParent());
if (!SourceDep.isClobber() || SourceDep.getInst() != MDep)
return false;
Value *TmpCast = MDep->getSource();
if (MDep->getSource()->getType() != ByValArg->getType())
TmpCast = new BitCastInst(MDep->getSource(), ByValArg->getType(),
"tmpcast", CS.getInstruction());
DEBUG(dbgs() << "MemCpyOpt: Forwarding memcpy to byval:\n"
<< " " << *MDep << "\n"
<< " " << *CS.getInstruction() << "\n");
CS.setArgument(ArgNo, TmpCast);
++NumMemCpyInstr;
return true;
}
bool MemCpyOpt::iterateOnFunction(Function &F) {
bool MadeChange = false;
for (Function::iterator BB = F.begin(), BBE = F.end(); BB != BBE; ++BB) {
for (BasicBlock::iterator BI = BB->begin(), BE = BB->end(); BI != BE;) {
Instruction *I = BI++;
bool RepeatInstruction = false;
if (StoreInst *SI = dyn_cast<StoreInst>(I))
MadeChange |= processStore(SI, BI);
else if (MemSetInst *M = dyn_cast<MemSetInst>(I))
RepeatInstruction = processMemSet(M, BI);
else if (MemCpyInst *M = dyn_cast<MemCpyInst>(I))
RepeatInstruction = processMemCpy(M);
else if (MemMoveInst *M = dyn_cast<MemMoveInst>(I))
RepeatInstruction = processMemMove(M);
else if (CallSite CS = (Value*)I) {
for (unsigned i = 0, e = CS.arg_size(); i != e; ++i)
if (CS.isByValArgument(i))
MadeChange |= processByValArgument(CS, i);
}
if (RepeatInstruction) {
if (BI != BB->begin()) --BI;
MadeChange = true;
}
}
}
return MadeChange;
}
bool MemCpyOpt::runOnFunction(Function &F) {
if (skipOptnoneFunction(F))
return false;
bool MadeChange = false;
MD = &getAnalysis<MemoryDependenceAnalysis>();
DataLayoutPass *DLP = getAnalysisIfAvailable<DataLayoutPass>();
DL = DLP ? &DLP->getDataLayout() : nullptr;
TLI = &getAnalysis<TargetLibraryInfo>();
if (!TLI->has(LibFunc::memset) || !TLI->has(LibFunc::memcpy))
return false;
while (1) {
if (!iterateOnFunction(F))
break;
MadeChange = true;
}
MD = nullptr;
return MadeChange;
}