#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/GetElementPtrTypeIterator.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Target/TargetLibraryInfo.h"
#include "llvm/Transforms/Utils/Local.h"
const DataLayout &TD) {
Ptr1 = Ptr1->stripPointerCasts();
Ptr2 = Ptr2->stripPointerCasts();
+
+ // Handle the trivial case first.
+ if (Ptr1 == Ptr2) {
+ Offset = 0;
+ return true;
+ }
+
GEPOperator *GEP1 = dyn_cast<GEPOperator>(Ptr1);
GEPOperator *GEP2 = dyn_cast<GEPOperator>(Ptr2);
/// because each element is relatively large and expensive to copy.
std::list<MemsetRange> Ranges;
typedef std::list<MemsetRange>::iterator range_iterator;
- const DataLayout &TD;
+ const DataLayout &DL;
public:
- MemsetRanges(const DataLayout &td) : TD(td) {}
+ MemsetRanges(const DataLayout &DL) : DL(DL) {}
typedef std::list<MemsetRange>::const_iterator const_iterator;
const_iterator begin() const { return Ranges.begin(); }
}
void addStore(int64_t OffsetFromFirst, StoreInst *SI) {
- int64_t StoreSize = TD.getTypeStoreSize(SI->getOperand(0)->getType());
+ int64_t StoreSize = DL.getTypeStoreSize(SI->getOperand(0)->getType());
addRange(OffsetFromFirst, StoreSize,
SI->getPointerOperand(), SI->getAlignment(), SI);
class MemCpyOpt : public FunctionPass {
MemoryDependenceAnalysis *MD;
TargetLibraryInfo *TLI;
- const DataLayout *TD;
+ const DataLayout *DL;
public:
static char ID; // Pass identification, replacement for typeid
MemCpyOpt() : FunctionPass(ID) {
initializeMemCpyOptPass(*PassRegistry::getPassRegistry());
MD = 0;
TLI = 0;
- TD = 0;
+ DL = 0;
}
- bool runOnFunction(Function &F);
+ bool runOnFunction(Function &F) override;
private:
// This transformation requires dominator postdominator info
- virtual void getAnalysisUsage(AnalysisUsage &AU) const {
+ void getAnalysisUsage(AnalysisUsage &AU) const override {
AU.setPreservesCFG();
AU.addRequired<DominatorTreeWrapperPass>();
AU.addRequired<MemoryDependenceAnalysis>();
/// attempts to merge them together into a memcpy/memset.
Instruction *MemCpyOpt::tryMergingIntoMemset(Instruction *StartInst,
Value *StartPtr, Value *ByteVal) {
- if (TD == 0) return 0;
+ if (DL == 0) return 0;
// Okay, so we now have a single store that can be splatable. Scan to find
// all subsequent stores of the same value to offset from the same pointer.
// Join these together into ranges, so we can decide whether contiguous blocks
// are stored.
- MemsetRanges Ranges(*TD);
+ MemsetRanges Ranges(*DL);
BasicBlock::iterator BI = StartInst;
for (++BI; !isa<TerminatorInst>(BI); ++BI) {
// Check to see if this store is to a constant offset from the start ptr.
int64_t Offset;
if (!IsPointerOffset(StartPtr, NextStore->getPointerOperand(),
- Offset, *TD))
+ Offset, *DL))
break;
Ranges.addStore(Offset, NextStore);
// Check to see if this store is to a constant offset from the start ptr.
int64_t Offset;
- if (!IsPointerOffset(StartPtr, MSI->getDest(), Offset, *TD))
+ if (!IsPointerOffset(StartPtr, MSI->getDest(), Offset, *DL))
break;
Ranges.addMemSet(Offset, MSI);
if (Range.TheStores.size() == 1) continue;
// If it is profitable to lower this range to memset, do so now.
- if (!Range.isProfitableToUseMemset(*TD))
+ if (!Range.isProfitableToUseMemset(*DL))
continue;
// Otherwise, we do want to transform this! Create a new memset.
if (Alignment == 0) {
Type *EltType =
cast<PointerType>(StartPtr->getType())->getElementType();
- Alignment = TD->getABITypeAlignment(EltType);
+ Alignment = DL->getABITypeAlignment(EltType);
}
AMemSet =
bool MemCpyOpt::processStore(StoreInst *SI, BasicBlock::iterator &BBI) {
if (!SI->isSimple()) return false;
- if (TD == 0) return false;
+ if (DL == 0) return false;
// Detect cases where we're performing call slot forwarding, but
// happen to be using a load-store pair to implement it, rather than
if (C) {
unsigned storeAlign = SI->getAlignment();
if (!storeAlign)
- storeAlign = TD->getABITypeAlignment(SI->getOperand(0)->getType());
+ storeAlign = DL->getABITypeAlignment(SI->getOperand(0)->getType());
unsigned loadAlign = LI->getAlignment();
if (!loadAlign)
- loadAlign = TD->getABITypeAlignment(LI->getType());
+ loadAlign = DL->getABITypeAlignment(LI->getType());
bool changed = performCallSlotOptzn(LI,
SI->getPointerOperand()->stripPointerCasts(),
LI->getPointerOperand()->stripPointerCasts(),
- TD->getTypeStoreSize(SI->getOperand(0)->getType()),
+ DL->getTypeStoreSize(SI->getOperand(0)->getType()),
std::min(storeAlign, loadAlign), C);
if (changed) {
MD->removeInstruction(SI);
return false;
// Check that all of src is copied to dest.
- if (TD == 0) return false;
+ if (DL == 0) return false;
ConstantInt *srcArraySize = dyn_cast<ConstantInt>(srcAlloca->getArraySize());
if (!srcArraySize)
return false;
- uint64_t srcSize = TD->getTypeAllocSize(srcAlloca->getAllocatedType()) *
+ uint64_t srcSize = DL->getTypeAllocSize(srcAlloca->getAllocatedType()) *
srcArraySize->getZExtValue();
if (cpyLen < srcSize)
if (!destArraySize)
return false;
- uint64_t destSize = TD->getTypeAllocSize(A->getAllocatedType()) *
+ uint64_t destSize = DL->getTypeAllocSize(A->getAllocatedType()) *
destArraySize->getZExtValue();
if (destSize < srcSize)
return false;
}
- uint64_t destSize = TD->getTypeAllocSize(StructTy);
+ uint64_t destSize = DL->getTypeAllocSize(StructTy);
if (destSize < srcSize)
return false;
} else {
// Check that dest points to memory that is at least as aligned as src.
unsigned srcAlign = srcAlloca->getAlignment();
if (!srcAlign)
- srcAlign = TD->getABITypeAlignment(srcAlloca->getAllocatedType());
+ srcAlign = DL->getABITypeAlignment(srcAlloca->getAllocatedType());
bool isDestSufficientlyAligned = srcAlign <= cpyAlign;
// If dest is not aligned enough and we can't increase its alignment then
// bail out.
// guarantees that it holds only undefined values when passed in (so the final
// memcpy can be dropped), that it is not read or written between the call and
// the memcpy, and that writing beyond the end of it is undefined.
- SmallVector<User*, 8> srcUseList(srcAlloca->use_begin(),
- srcAlloca->use_end());
+ SmallVector<User*, 8> srcUseList(srcAlloca->user_begin(),
+ srcAlloca->user_end());
while (!srcUseList.empty()) {
- User *UI = srcUseList.pop_back_val();
+ User *U = srcUseList.pop_back_val();
- if (isa<BitCastInst>(UI) || isa<AddrSpaceCastInst>(UI)) {
- for (User::use_iterator I = UI->use_begin(), E = UI->use_end();
- I != E; ++I)
- srcUseList.push_back(*I);
- } else if (GetElementPtrInst *G = dyn_cast<GetElementPtrInst>(UI)) {
+ 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::use_iterator I = UI->use_begin(), E = UI->use_end();
- I != E; ++I)
- srcUseList.push_back(*I);
+ for (User *UU : U->users())
+ srcUseList.push_back(UU);
else
return false;
- } else if (UI != C && UI != cpy) {
+ } else if (U != C && U != cpy) {
return false;
}
}
ConstantInt *CopySize = dyn_cast<ConstantInt>(M->getLength());
if (CopySize == 0) return false;
- // The are two possible optimizations we can do for memcpy:
+ // The are three possible optimizations we can do for memcpy:
// a) memcpy-memcpy xform which exposes redundance for DSE.
// b) call-memcpy xform for return slot optimization.
+ // c) memcpy from freshly alloca'd space copies undefined data, and we can
+ // therefore eliminate the memcpy in favor of the data that was already
+ // at the destination.
MemDepResult DepInfo = MD->getDependency(M);
if (DepInfo.isClobber()) {
if (CallInst *C = dyn_cast<CallInst>(DepInfo.getInst())) {
if (SrcDepInfo.isClobber()) {
if (MemCpyInst *MDep = dyn_cast<MemCpyInst>(SrcDepInfo.getInst()))
return processMemCpyMemCpyDependence(M, MDep, CopySize->getZExtValue());
+ } else if (SrcDepInfo.isDef()) {
+ if (isa<AllocaInst>(SrcDepInfo.getInst())) {
+ MD->removeInstruction(M);
+ M->eraseFromParent();
+ ++NumMemCpyInstr;
+ return true;
+ }
}
return false;
/// processByValArgument - This is called on every byval argument in call sites.
bool MemCpyOpt::processByValArgument(CallSite CS, unsigned ArgNo) {
- if (TD == 0) return false;
+ if (DL == 0) return false;
// Find out what feeds this byval argument.
Value *ByValArg = CS.getArgument(ArgNo);
Type *ByValTy = cast<PointerType>(ByValArg->getType())->getElementType();
- uint64_t ByValSize = TD->getTypeAllocSize(ByValTy);
+ uint64_t ByValSize = DL->getTypeAllocSize(ByValTy);
MemDepResult DepInfo =
MD->getPointerDependencyFrom(AliasAnalysis::Location(ByValArg, ByValSize),
true, CS.getInstruction(),
// If it is greater than the memcpy, then we check to see if we can force the
// source of the memcpy to the alignment we need. If we fail, we bail out.
if (MDep->getAlignment() < ByValAlign &&
- getOrEnforceKnownAlignment(MDep->getSource(),ByValAlign, TD) < ByValAlign)
+ getOrEnforceKnownAlignment(MDep->getSource(),ByValAlign, DL) < ByValAlign)
return false;
// Verify that the copied-from memory doesn't change in between the memcpy and
bool MadeChange = false;
MD = &getAnalysis<MemoryDependenceAnalysis>();
- TD = getAnalysisIfAvailable<DataLayout>();
+ DataLayoutPass *DLP = getAnalysisIfAvailable<DataLayoutPass>();
+ DL = DLP ? &DLP->getDataLayout() : 0;
TLI = &getAnalysis<TargetLibraryInfo>();
// If we don't have at least memset and memcpy, there is little point of doing