X-Git-Url: http://plrg.eecs.uci.edu/git/?p=oota-llvm.git;a=blobdiff_plain;f=lib%2FTransforms%2FScalar%2FDeadStoreElimination.cpp;h=f157cbfe8bd758c6df2961e04b09a94cd3c7a7ec;hp=921bec8f4dec59089e630c01246a40a75e158f44;hb=609d95290a4d913139ba0b46f5b5b982dc4b1d96;hpb=2d32b86543aa857a07371a93766652510f1cc0d6 diff --git a/lib/Transforms/Scalar/DeadStoreElimination.cpp b/lib/Transforms/Scalar/DeadStoreElimination.cpp index 921bec8f4de..f157cbfe8bd 100644 --- a/lib/Transforms/Scalar/DeadStoreElimination.cpp +++ b/lib/Transforms/Scalar/DeadStoreElimination.cpp @@ -15,27 +15,31 @@ // //===----------------------------------------------------------------------===// -#define DEBUG_TYPE "dse" #include "llvm/Transforms/Scalar.h" -#include "llvm/Constants.h" -#include "llvm/Function.h" -#include "llvm/GlobalVariable.h" -#include "llvm/Instructions.h" -#include "llvm/IntrinsicInst.h" -#include "llvm/Pass.h" +#include "llvm/ADT/STLExtras.h" +#include "llvm/ADT/SetVector.h" +#include "llvm/ADT/Statistic.h" #include "llvm/Analysis/AliasAnalysis.h" #include "llvm/Analysis/CaptureTracking.h" -#include "llvm/Analysis/Dominators.h" #include "llvm/Analysis/MemoryBuiltins.h" #include "llvm/Analysis/MemoryDependenceAnalysis.h" +#include "llvm/Analysis/TargetLibraryInfo.h" #include "llvm/Analysis/ValueTracking.h" -#include "llvm/Target/TargetData.h" -#include "llvm/Transforms/Utils/Local.h" +#include "llvm/IR/Constants.h" +#include "llvm/IR/DataLayout.h" +#include "llvm/IR/Dominators.h" +#include "llvm/IR/Function.h" +#include "llvm/IR/GlobalVariable.h" +#include "llvm/IR/Instructions.h" +#include "llvm/IR/IntrinsicInst.h" +#include "llvm/Pass.h" #include "llvm/Support/Debug.h" -#include "llvm/ADT/SmallPtrSet.h" -#include "llvm/ADT/Statistic.h" +#include "llvm/Support/raw_ostream.h" +#include "llvm/Transforms/Utils/Local.h" using namespace llvm; +#define DEBUG_TYPE "dse" + STATISTIC(NumFastStores, "Number of stores deleted"); STATISTIC(NumFastOther , "Number of other instrs removed"); @@ -43,41 +47,49 @@ namespace { struct DSE : public FunctionPass { AliasAnalysis *AA; MemoryDependenceAnalysis *MD; + DominatorTree *DT; + const TargetLibraryInfo *TLI; static char ID; // Pass identification, replacement for typeid - DSE() : FunctionPass(ID), AA(0), MD(0) { + DSE() : FunctionPass(ID), AA(nullptr), MD(nullptr), DT(nullptr) { initializeDSEPass(*PassRegistry::getPassRegistry()); } - virtual bool runOnFunction(Function &F) { + bool runOnFunction(Function &F) override { + if (skipOptnoneFunction(F)) + return false; + AA = &getAnalysis(); MD = &getAnalysis(); - DominatorTree &DT = getAnalysis(); + DT = &getAnalysis().getDomTree(); + TLI = &getAnalysis().getTLI(); bool Changed = false; for (Function::iterator I = F.begin(), E = F.end(); I != E; ++I) // Only check non-dead blocks. Dead blocks may have strange pointer // cycles that will confuse alias analysis. - if (DT.isReachableFromEntry(I)) + if (DT->isReachableFromEntry(I)) Changed |= runOnBasicBlock(*I); - AA = 0; MD = 0; + AA = nullptr; MD = nullptr; DT = nullptr; return Changed; } bool runOnBasicBlock(BasicBlock &BB); bool HandleFree(CallInst *F); bool handleEndBlock(BasicBlock &BB); - void RemoveAccessedObjects(const AliasAnalysis::Location &LoadedLoc, - SmallPtrSet &DeadStackObjects); + void RemoveAccessedObjects(const MemoryLocation &LoadedLoc, + SmallSetVector &DeadStackObjects, + const DataLayout &DL); - virtual void getAnalysisUsage(AnalysisUsage &AU) const { + void getAnalysisUsage(AnalysisUsage &AU) const override { AU.setPreservesCFG(); - AU.addRequired(); + AU.addRequired(); AU.addRequired(); AU.addRequired(); + AU.addRequired(); AU.addPreserved(); - AU.addPreserved(); + AU.addPreserved(); AU.addPreserved(); } }; @@ -85,9 +97,10 @@ namespace { char DSE::ID = 0; INITIALIZE_PASS_BEGIN(DSE, "dse", "Dead Store Elimination", false, false) -INITIALIZE_PASS_DEPENDENCY(DominatorTree) -INITIALIZE_PASS_DEPENDENCY(MemoryDependenceAnalysis) INITIALIZE_AG_DEPENDENCY(AliasAnalysis) +INITIALIZE_PASS_DEPENDENCY(DominatorTreeWrapperPass) +INITIALIZE_PASS_DEPENDENCY(MemoryDependenceAnalysis) +INITIALIZE_PASS_DEPENDENCY(TargetLibraryInfoWrapperPass) INITIALIZE_PASS_END(DSE, "dse", "Dead Store Elimination", false, false) FunctionPass *llvm::createDeadStoreEliminationPass() { return new DSE(); } @@ -103,8 +116,9 @@ FunctionPass *llvm::createDeadStoreEliminationPass() { return new DSE(); } /// If ValueSet is non-null, remove any deleted instructions from it as well. /// static void DeleteDeadInstruction(Instruction *I, - MemoryDependenceAnalysis &MD, - SmallPtrSet *ValueSet = 0) { + MemoryDependenceAnalysis &MD, + const TargetLibraryInfo &TLI, + SmallSetVector *ValueSet = nullptr) { SmallVector NowDeadInsts; NowDeadInsts.push_back(I); @@ -122,26 +136,26 @@ static void DeleteDeadInstruction(Instruction *I, for (unsigned op = 0, e = DeadInst->getNumOperands(); op != e; ++op) { Value *Op = DeadInst->getOperand(op); - DeadInst->setOperand(op, 0); + DeadInst->setOperand(op, nullptr); // If this operand just became dead, add it to the NowDeadInsts list. if (!Op->use_empty()) continue; if (Instruction *OpI = dyn_cast(Op)) - if (isInstructionTriviallyDead(OpI)) + if (isInstructionTriviallyDead(OpI, &TLI)) NowDeadInsts.push_back(OpI); } DeadInst->eraseFromParent(); - if (ValueSet) ValueSet->erase(DeadInst); + if (ValueSet) ValueSet->remove(DeadInst); } while (!NowDeadInsts.empty()); } /// hasMemoryWrite - Does this instruction write some memory? This only returns /// true for things that we can analyze with other helpers below. -static bool hasMemoryWrite(Instruction *I) { +static bool hasMemoryWrite(Instruction *I, const TargetLibraryInfo &TLI) { if (isa(I)) return true; if (IntrinsicInst *II = dyn_cast(I)) { @@ -156,60 +170,71 @@ static bool hasMemoryWrite(Instruction *I) { return true; } } + if (auto CS = CallSite(I)) { + if (Function *F = CS.getCalledFunction()) { + if (TLI.has(LibFunc::strcpy) && + F->getName() == TLI.getName(LibFunc::strcpy)) { + return true; + } + if (TLI.has(LibFunc::strncpy) && + F->getName() == TLI.getName(LibFunc::strncpy)) { + return true; + } + if (TLI.has(LibFunc::strcat) && + F->getName() == TLI.getName(LibFunc::strcat)) { + return true; + } + if (TLI.has(LibFunc::strncat) && + F->getName() == TLI.getName(LibFunc::strncat)) { + return true; + } + } + } return false; } /// getLocForWrite - Return a Location stored to by the specified instruction. /// If isRemovable returns true, this function and getLocForRead completely /// describe the memory operations for this instruction. -static AliasAnalysis::Location -getLocForWrite(Instruction *Inst, AliasAnalysis &AA) { +static MemoryLocation getLocForWrite(Instruction *Inst, AliasAnalysis &AA) { if (StoreInst *SI = dyn_cast(Inst)) - return AA.getLocation(SI); + return MemoryLocation::get(SI); if (MemIntrinsic *MI = dyn_cast(Inst)) { // memcpy/memmove/memset. - AliasAnalysis::Location Loc = AA.getLocationForDest(MI); - // If we don't have target data around, an unknown size in Location means - // that we should use the size of the pointee type. This isn't valid for - // memset/memcpy, which writes more than an i8. - if (Loc.Size == AliasAnalysis::UnknownSize && AA.getTargetData() == 0) - return AliasAnalysis::Location(); + MemoryLocation Loc = MemoryLocation::getForDest(MI); return Loc; } IntrinsicInst *II = dyn_cast(Inst); - if (II == 0) return AliasAnalysis::Location(); + if (!II) + return MemoryLocation(); switch (II->getIntrinsicID()) { - default: return AliasAnalysis::Location(); // Unhandled intrinsic. + default: + return MemoryLocation(); // Unhandled intrinsic. case Intrinsic::init_trampoline: - // If we don't have target data around, an unknown size in Location means - // that we should use the size of the pointee type. This isn't valid for - // init.trampoline, which writes more than an i8. - if (AA.getTargetData() == 0) return AliasAnalysis::Location(); - // FIXME: We don't know the size of the trampoline, so we can't really // handle it here. - return AliasAnalysis::Location(II->getArgOperand(0)); + return MemoryLocation(II->getArgOperand(0)); case Intrinsic::lifetime_end: { uint64_t Len = cast(II->getArgOperand(0))->getZExtValue(); - return AliasAnalysis::Location(II->getArgOperand(1), Len); + return MemoryLocation(II->getArgOperand(1), Len); } } } /// getLocForRead - Return the location read by the specified "hasMemoryWrite" /// instruction if any. -static AliasAnalysis::Location -getLocForRead(Instruction *Inst, AliasAnalysis &AA) { - assert(hasMemoryWrite(Inst) && "Unknown instruction case"); +static MemoryLocation getLocForRead(Instruction *Inst, + const TargetLibraryInfo &TLI) { + assert(hasMemoryWrite(Inst, TLI) && "Unknown instruction case"); // The only instructions that both read and write are the mem transfer // instructions (memcpy/memmove). if (MemTransferInst *MTI = dyn_cast(Inst)) - return AA.getLocationForSource(MTI); - return AliasAnalysis::Location(); + return MemoryLocation::getForSource(MTI); + return MemoryLocation(); } @@ -220,23 +245,29 @@ static bool isRemovable(Instruction *I) { if (StoreInst *SI = dyn_cast(I)) return SI->isUnordered(); - IntrinsicInst *II = cast(I); - switch (II->getIntrinsicID()) { - default: assert(0 && "doesn't pass 'hasMemoryWrite' predicate"); - case Intrinsic::lifetime_end: - // Never remove dead lifetime_end's, e.g. because it is followed by a - // free. - return false; - case Intrinsic::init_trampoline: - // Always safe to remove init_trampoline. - return true; + if (IntrinsicInst *II = dyn_cast(I)) { + switch (II->getIntrinsicID()) { + default: llvm_unreachable("doesn't pass 'hasMemoryWrite' predicate"); + case Intrinsic::lifetime_end: + // Never remove dead lifetime_end's, e.g. because it is followed by a + // free. + return false; + case Intrinsic::init_trampoline: + // Always safe to remove init_trampoline. + return true; - case Intrinsic::memset: - case Intrinsic::memmove: - case Intrinsic::memcpy: - // Don't remove volatile memory intrinsics. - return !cast(II)->isVolatile(); + case Intrinsic::memset: + case Intrinsic::memmove: + case Intrinsic::memcpy: + // Don't remove volatile memory intrinsics. + return !cast(II)->isVolatile(); + } } + + if (auto CS = CallSite(I)) + return CS.getInstruction()->use_empty(); + + return false; } @@ -246,15 +277,20 @@ static bool isShortenable(Instruction *I) { // Don't shorten stores for now if (isa(I)) return false; - - IntrinsicInst *II = cast(I); - switch (II->getIntrinsicID()) { - default: return false; - case Intrinsic::memset: - case Intrinsic::memcpy: - // Do shorten memory intrinsics. - return true; + + if (IntrinsicInst *II = dyn_cast(I)) { + switch (II->getIntrinsicID()) { + default: return false; + case Intrinsic::memset: + case Intrinsic::memcpy: + // Do shorten memory intrinsics. + return true; + } } + + // Don't shorten libcalls calls for now. + + return false; } /// getStoredPointerOperand - Return the pointer that is being written to. @@ -264,51 +300,26 @@ static Value *getStoredPointerOperand(Instruction *I) { if (MemIntrinsic *MI = dyn_cast(I)) return MI->getDest(); - IntrinsicInst *II = cast(I); - switch (II->getIntrinsicID()) { - default: assert(false && "Unexpected intrinsic!"); - case Intrinsic::init_trampoline: - return II->getArgOperand(0); - } -} - -static uint64_t getPointerSize(Value *V, AliasAnalysis &AA) { - const TargetData *TD = AA.getTargetData(); - - if (CallInst *CI = dyn_cast(V)) { - assert(isMalloc(CI) && "Expected Malloc call!"); - if (ConstantInt *C = dyn_cast(CI->getArgOperand(0))) - return C->getZExtValue(); - return AliasAnalysis::UnknownSize; - } - - if (TD == 0) - return AliasAnalysis::UnknownSize; - - if (AllocaInst *A = dyn_cast(V)) { - // Get size information for the alloca - if (ConstantInt *C = dyn_cast(A->getArraySize())) - return C->getZExtValue() * TD->getTypeAllocSize(A->getAllocatedType()); - return AliasAnalysis::UnknownSize; + if (IntrinsicInst *II = dyn_cast(I)) { + switch (II->getIntrinsicID()) { + default: llvm_unreachable("Unexpected intrinsic!"); + case Intrinsic::init_trampoline: + return II->getArgOperand(0); + } } - assert(isa(V) && "Expected AllocaInst, malloc call or Argument!"); - PointerType *PT = cast(V->getType()); - return TD->getTypeAllocSize(PT->getElementType()); + CallSite CS(I); + // All the supported functions so far happen to have dest as their first + // argument. + return CS.getArgument(0); } -/// isObjectPointerWithTrustworthySize - Return true if the specified Value* is -/// pointing to an object with a pointer size we can trust. -static bool isObjectPointerWithTrustworthySize(const Value *V) { - if (const AllocaInst *AI = dyn_cast(V)) - return !AI->isArrayAllocation(); - if (const GlobalVariable *GV = dyn_cast(V)) - return !GV->mayBeOverridden(); - if (const Argument *A = dyn_cast(V)) - return A->hasByValAttr(); - if (isMalloc(V)) - return true; - return false; +static uint64_t getPointerSize(const Value *V, const DataLayout &DL, + const TargetLibraryInfo &TLI) { + uint64_t Size; + if (getObjectSize(V, Size, DL, &TLI)) + return Size; + return MemoryLocation::UnknownSize; } namespace { @@ -322,13 +333,13 @@ namespace { /// isOverwrite - Return 'OverwriteComplete' if a store to the 'Later' location /// completely overwrites a store to the 'Earlier' location. -/// 'OverwriteEnd' if the end of the 'Earlier' location is completely +/// 'OverwriteEnd' if the end of the 'Earlier' location is completely /// overwritten by 'Later', or 'OverwriteUnknown' if nothing can be determined -static OverwriteResult isOverwrite(const AliasAnalysis::Location &Later, - const AliasAnalysis::Location &Earlier, - AliasAnalysis &AA, - int64_t& EarlierOff, - int64_t& LaterOff) { +static OverwriteResult isOverwrite(const MemoryLocation &Later, + const MemoryLocation &Earlier, + const DataLayout &DL, + const TargetLibraryInfo &TLI, + int64_t &EarlierOff, int64_t &LaterOff) { const Value *P1 = Earlier.Ptr->stripPointerCasts(); const Value *P2 = Later.Ptr->stripPointerCasts(); @@ -337,17 +348,9 @@ static OverwriteResult isOverwrite(const AliasAnalysis::Location &Later, if (P1 == P2) { // If we don't know the sizes of either access, then we can't do a // comparison. - if (Later.Size == AliasAnalysis::UnknownSize || - Earlier.Size == AliasAnalysis::UnknownSize) { - // If we have no TargetData information around, then the size of the store - // is inferrable from the pointee type. If they are the same type, then - // we know that the store is safe. - if (AA.getTargetData() == 0 && - Later.Ptr->getType() == Earlier.Ptr->getType()) - return OverwriteComplete; - + if (Later.Size == MemoryLocation::UnknownSize || + Earlier.Size == MemoryLocation::UnknownSize) return OverwriteUnknown; - } // Make sure that the Later size is >= the Earlier size. if (Later.Size >= Earlier.Size) @@ -356,18 +359,15 @@ static OverwriteResult isOverwrite(const AliasAnalysis::Location &Later, // Otherwise, we have to have size information, and the later store has to be // larger than the earlier one. - if (Later.Size == AliasAnalysis::UnknownSize || - Earlier.Size == AliasAnalysis::UnknownSize || - AA.getTargetData() == 0) + if (Later.Size == MemoryLocation::UnknownSize || + Earlier.Size == MemoryLocation::UnknownSize) return OverwriteUnknown; // Check to see if the later store is to the entire object (either a global, - // an alloca, or a byval argument). If so, then it clearly overwrites any - // other store to the same object. - const TargetData &TD = *AA.getTargetData(); - - const Value *UO1 = GetUnderlyingObject(P1, &TD), - *UO2 = GetUnderlyingObject(P2, &TD); + // an alloca, or a byval/inalloca argument). If so, then it clearly + // overwrites any other store to the same object. + const Value *UO1 = GetUnderlyingObject(P1, DL), + *UO2 = GetUnderlyingObject(P2, DL); // If we can't resolve the same pointers to the same object, then we can't // analyze them at all. @@ -375,20 +375,18 @@ static OverwriteResult isOverwrite(const AliasAnalysis::Location &Later, return OverwriteUnknown; // If the "Later" store is to a recognizable object, get its size. - if (isObjectPointerWithTrustworthySize(UO2)) { - uint64_t ObjectSize = - TD.getTypeAllocSize(cast(UO2->getType())->getElementType()); - if (ObjectSize == Later.Size) + uint64_t ObjectSize = getPointerSize(UO2, DL, TLI); + if (ObjectSize != MemoryLocation::UnknownSize) + if (ObjectSize == Later.Size && ObjectSize >= Earlier.Size) return OverwriteComplete; - } // Okay, we have stores to two completely different pointers. Try to // decompose the pointer into a "base + constant_offset" form. If the base // pointers are equal, then we can reason about the two stores. EarlierOff = 0; LaterOff = 0; - const Value *BP1 = GetPointerBaseWithConstantOffset(P1, EarlierOff, TD); - const Value *BP2 = GetPointerBaseWithConstantOffset(P2, LaterOff, TD); + const Value *BP1 = GetPointerBaseWithConstantOffset(P1, EarlierOff, DL); + const Value *BP2 = GetPointerBaseWithConstantOffset(P2, LaterOff, DL); // If the base pointers still differ, we have two completely different stores. if (BP1 != BP2) @@ -410,10 +408,10 @@ static OverwriteResult isOverwrite(const AliasAnalysis::Location &Later, // // We have to be careful here as *Off is signed while *.Size is unsigned. if (EarlierOff >= LaterOff && - Later.Size > Earlier.Size && + Later.Size >= Earlier.Size && uint64_t(EarlierOff - LaterOff) + Earlier.Size <= Later.Size) return OverwriteComplete; - + // The other interesting case is if the later store overwrites the end of // the earlier store // @@ -423,7 +421,8 @@ static OverwriteResult isOverwrite(const AliasAnalysis::Location &Later, // In this case we may want to trim the size of earlier to avoid generating // writes to addresses which will definitely be overwritten later if (LaterOff > EarlierOff && - LaterOff + Later.Size >= EarlierOff + Earlier.Size) + LaterOff < int64_t(EarlierOff + Earlier.Size) && + int64_t(LaterOff + Later.Size) >= int64_t(EarlierOff + Earlier.Size)) return OverwriteEnd; // Otherwise, they don't completely overlap. @@ -444,12 +443,14 @@ static OverwriteResult isOverwrite(const AliasAnalysis::Location &Later, /// This function detects when it is unsafe to remove a dependent instruction /// because the DSE inducing instruction may be a self-read. static bool isPossibleSelfRead(Instruction *Inst, - const AliasAnalysis::Location &InstStoreLoc, - Instruction *DepWrite, AliasAnalysis &AA) { + const MemoryLocation &InstStoreLoc, + Instruction *DepWrite, + const TargetLibraryInfo &TLI, + AliasAnalysis &AA) { // Self reads can only happen for instructions that read memory. Get the // location read. - AliasAnalysis::Location InstReadLoc = getLocForRead(Inst, AA); - if (InstReadLoc.Ptr == 0) return false; // Not a reading instruction. + MemoryLocation InstReadLoc = getLocForRead(Inst, TLI); + if (!InstReadLoc.Ptr) return false; // Not a reading instruction. // If the read and written loc obviously don't alias, it isn't a read. if (AA.isNoAlias(InstReadLoc, InstStoreLoc)) return false; @@ -462,7 +463,7 @@ static bool isPossibleSelfRead(Instruction *Inst, // Here we don't know if A/B may alias, but we do know that B/B are must // aliases, so removing the first memcpy is safe (assuming it writes <= # // bytes as the second one. - AliasAnalysis::Location DepReadLoc = getLocForRead(DepWrite, AA); + MemoryLocation DepReadLoc = getLocForRead(DepWrite, TLI); if (DepReadLoc.Ptr && AA.isMustAlias(InstReadLoc.Ptr, DepReadLoc.Ptr)) return false; @@ -485,13 +486,13 @@ bool DSE::runOnBasicBlock(BasicBlock &BB) { Instruction *Inst = BBI++; // Handle 'free' calls specially. - if (CallInst *F = isFreeCall(Inst)) { + if (CallInst *F = isFreeCall(Inst, TLI)) { MadeChange |= HandleFree(F); continue; } // If we find something that writes memory, get its memory dependence. - if (!hasMemoryWrite(Inst)) + if (!hasMemoryWrite(Inst, *TLI)) continue; MemDepResult InstDep = MD->getDependency(Inst); @@ -514,9 +515,9 @@ bool DSE::runOnBasicBlock(BasicBlock &BB) { // in case we need it. WeakVH NextInst(BBI); - DeleteDeadInstruction(SI, *MD); + DeleteDeadInstruction(SI, *MD, *TLI); - if (NextInst == 0) // Next instruction deleted. + if (!NextInst) // Next instruction deleted. BBI = BB.begin(); else if (BBI != BB.begin()) // Revisit this instruction if possible. --BBI; @@ -528,10 +529,10 @@ bool DSE::runOnBasicBlock(BasicBlock &BB) { } // Figure out what location is being stored to. - AliasAnalysis::Location Loc = getLocForWrite(Inst, *AA); + MemoryLocation Loc = getLocForWrite(Inst, *AA); // If we didn't get a useful location, fail. - if (Loc.Ptr == 0) + if (!Loc.Ptr) continue; while (InstDep.isDef() || InstDep.isClobber()) { @@ -543,28 +544,29 @@ bool DSE::runOnBasicBlock(BasicBlock &BB) { // // Find out what memory location the dependent instruction stores. Instruction *DepWrite = InstDep.getInst(); - AliasAnalysis::Location DepLoc = getLocForWrite(DepWrite, *AA); + MemoryLocation DepLoc = getLocForWrite(DepWrite, *AA); // If we didn't get a useful location, or if it isn't a size, bail out. - if (DepLoc.Ptr == 0) + if (!DepLoc.Ptr) break; // If we find a write that is a) removable (i.e., non-volatile), b) is // completely obliterated by the store to 'Loc', and c) which we know that // 'Inst' doesn't load from, then we can remove it. - if (isRemovable(DepWrite) && - !isPossibleSelfRead(Inst, Loc, DepWrite, *AA)) { - int64_t InstWriteOffset, DepWriteOffset; - OverwriteResult OR = isOverwrite(Loc, DepLoc, *AA, - DepWriteOffset, InstWriteOffset); + if (isRemovable(DepWrite) && + !isPossibleSelfRead(Inst, Loc, DepWrite, *TLI, *AA)) { + int64_t InstWriteOffset, DepWriteOffset; + const DataLayout &DL = BB.getModule()->getDataLayout(); + OverwriteResult OR = + isOverwrite(Loc, DepLoc, DL, *TLI, DepWriteOffset, InstWriteOffset); if (OR == OverwriteComplete) { DEBUG(dbgs() << "DSE: Remove Dead Store:\n DEAD: " << *DepWrite << "\n KILLER: " << *Inst << '\n'); // Delete the store and now-dead instructions that feed it. - DeleteDeadInstruction(DepWrite, *MD); + DeleteDeadInstruction(DepWrite, *MD, *TLI); ++NumFastStores; MadeChange = true; - + // DeleteDeadInstruction can delete the current instruction in loop // cases, reset BBI. BBI = Inst; @@ -582,16 +584,16 @@ bool DSE::runOnBasicBlock(BasicBlock &BB) { unsigned DepWriteAlign = DepIntrinsic->getAlignment(); if (llvm::isPowerOf2_64(InstWriteOffset) || ((DepWriteAlign != 0) && InstWriteOffset % DepWriteAlign == 0)) { - + DEBUG(dbgs() << "DSE: Remove Dead Store:\n OW END: " - << *DepWrite << "\n KILLER (offset " - << InstWriteOffset << ", " + << *DepWrite << "\n KILLER (offset " + << InstWriteOffset << ", " << DepLoc.Size << ")" << *Inst << '\n'); - + Value* DepWriteLength = DepIntrinsic->getLength(); Value* TrimmedLength = ConstantInt::get(DepWriteLength->getType(), - InstWriteOffset - + InstWriteOffset - DepWriteOffset); DepIntrinsic->setLength(TrimmedLength); MadeChange = true; @@ -610,7 +612,7 @@ bool DSE::runOnBasicBlock(BasicBlock &BB) { if (DepWrite == &BB.front()) break; // Can't look past this instruction if it might read 'Loc'. - if (AA->getModRefInfo(DepWrite, Loc) & AliasAnalysis::Ref) + if (AA->getModRefInfo(DepWrite, Loc) & MRI_Ref) break; InstDep = MD->getPointerDependencyFrom(Loc, false, DepWrite, &BB); @@ -625,37 +627,68 @@ bool DSE::runOnBasicBlock(BasicBlock &BB) { return MadeChange; } +/// Find all blocks that will unconditionally lead to the block BB and append +/// them to F. +static void FindUnconditionalPreds(SmallVectorImpl &Blocks, + BasicBlock *BB, DominatorTree *DT) { + for (pred_iterator I = pred_begin(BB), E = pred_end(BB); I != E; ++I) { + BasicBlock *Pred = *I; + if (Pred == BB) continue; + TerminatorInst *PredTI = Pred->getTerminator(); + if (PredTI->getNumSuccessors() != 1) + continue; + + if (DT->isReachableFromEntry(Pred)) + Blocks.push_back(Pred); + } +} + /// HandleFree - Handle frees of entire structures whose dependency is a store /// to a field of that structure. bool DSE::HandleFree(CallInst *F) { bool MadeChange = false; - MemDepResult Dep = MD->getDependency(F); + MemoryLocation Loc = MemoryLocation(F->getOperand(0)); + SmallVector Blocks; + Blocks.push_back(F->getParent()); + const DataLayout &DL = F->getModule()->getDataLayout(); + + while (!Blocks.empty()) { + BasicBlock *BB = Blocks.pop_back_val(); + Instruction *InstPt = BB->getTerminator(); + if (BB == F->getParent()) InstPt = F; - while (Dep.isDef() || Dep.isClobber()) { - Instruction *Dependency = Dep.getInst(); - if (!hasMemoryWrite(Dependency) || !isRemovable(Dependency)) - return MadeChange; + MemDepResult Dep = MD->getPointerDependencyFrom(Loc, false, InstPt, BB); + while (Dep.isDef() || Dep.isClobber()) { + Instruction *Dependency = Dep.getInst(); + if (!hasMemoryWrite(Dependency, *TLI) || !isRemovable(Dependency)) + break; - Value *DepPointer = - GetUnderlyingObject(getStoredPointerOperand(Dependency)); + Value *DepPointer = + GetUnderlyingObject(getStoredPointerOperand(Dependency), DL); - // Check for aliasing. - if (!AA->isMustAlias(F->getArgOperand(0), DepPointer)) - return MadeChange; + // Check for aliasing. + if (!AA->isMustAlias(F->getArgOperand(0), DepPointer)) + break; - // DCE instructions only used to calculate that store - DeleteDeadInstruction(Dependency, *MD); - ++NumFastStores; - MadeChange = true; + Instruction *Next = std::next(BasicBlock::iterator(Dependency)); - // Inst's old Dependency is now deleted. Compute the next dependency, - // which may also be dead, as in - // s[0] = 0; - // s[1] = 0; // This has just been deleted. - // free(s); - Dep = MD->getDependency(F); - }; + // DCE instructions only used to calculate that store + DeleteDeadInstruction(Dependency, *MD, *TLI); + ++NumFastStores; + MadeChange = true; + + // Inst's old Dependency is now deleted. Compute the next dependency, + // which may also be dead, as in + // s[0] = 0; + // s[1] = 0; // This has just been deleted. + // free(s); + Dep = MD->getPointerDependencyFrom(Loc, false, Next, BB); + } + + if (Dep.isNonLocal()) + FindUnconditionalPreds(Blocks, BB, DT); + } return MadeChange; } @@ -671,46 +704,63 @@ bool DSE::handleEndBlock(BasicBlock &BB) { // Keep track of all of the stack objects that are dead at the end of the // function. - SmallPtrSet DeadStackObjects; + SmallSetVector DeadStackObjects; // Find all of the alloca'd pointers in the entry block. BasicBlock *Entry = BB.getParent()->begin(); for (BasicBlock::iterator I = Entry->begin(), E = Entry->end(); I != E; ++I) { - if (AllocaInst *AI = dyn_cast(I)) - DeadStackObjects.insert(AI); + if (isa(I)) + DeadStackObjects.insert(I); // Okay, so these are dead heap objects, but if the pointer never escapes // then it's leaked by this function anyways. - if (CallInst *CI = extractMallocCall(I)) - if (!PointerMayBeCaptured(CI, true, true)) - DeadStackObjects.insert(CI); + else if (isAllocLikeFn(I, TLI) && !PointerMayBeCaptured(I, true, true)) + DeadStackObjects.insert(I); } - // Treat byval arguments the same, stores to them are dead at the end of the - // function. + // Treat byval or inalloca arguments the same, stores to them are dead at the + // end of the function. for (Function::arg_iterator AI = BB.getParent()->arg_begin(), AE = BB.getParent()->arg_end(); AI != AE; ++AI) - if (AI->hasByValAttr()) + if (AI->hasByValOrInAllocaAttr()) DeadStackObjects.insert(AI); + const DataLayout &DL = BB.getModule()->getDataLayout(); + // Scan the basic block backwards for (BasicBlock::iterator BBI = BB.end(); BBI != BB.begin(); ){ --BBI; // If we find a store, check to see if it points into a dead stack value. - if (hasMemoryWrite(BBI) && isRemovable(BBI)) { + if (hasMemoryWrite(BBI, *TLI) && isRemovable(BBI)) { // See through pointer-to-pointer bitcasts - Value *Pointer = GetUnderlyingObject(getStoredPointerOperand(BBI)); + SmallVector Pointers; + GetUnderlyingObjects(getStoredPointerOperand(BBI), Pointers, DL); // Stores to stack values are valid candidates for removal. - if (DeadStackObjects.count(Pointer)) { + bool AllDead = true; + for (SmallVectorImpl::iterator I = Pointers.begin(), + E = Pointers.end(); I != E; ++I) + if (!DeadStackObjects.count(*I)) { + AllDead = false; + break; + } + + if (AllDead) { Instruction *Dead = BBI++; DEBUG(dbgs() << "DSE: Dead Store at End of Block:\n DEAD: " - << *Dead << "\n Object: " << *Pointer << '\n'); + << *Dead << "\n Objects: "; + for (SmallVectorImpl::iterator I = Pointers.begin(), + E = Pointers.end(); I != E; ++I) { + dbgs() << **I; + if (std::next(I) != E) + dbgs() << ", "; + } + dbgs() << '\n'); // DCE instructions only used to calculate that store. - DeleteDeadInstruction(Dead, *MD, &DeadStackObjects); + DeleteDeadInstruction(Dead, *MD, *TLI, &DeadStackObjects); ++NumFastStores; MadeChange = true; continue; @@ -718,25 +768,27 @@ bool DSE::handleEndBlock(BasicBlock &BB) { } // Remove any dead non-memory-mutating instructions. - if (isInstructionTriviallyDead(BBI)) { + if (isInstructionTriviallyDead(BBI, TLI)) { Instruction *Inst = BBI++; - DeleteDeadInstruction(Inst, *MD, &DeadStackObjects); + DeleteDeadInstruction(Inst, *MD, *TLI, &DeadStackObjects); ++NumFastOther; MadeChange = true; continue; } - if (AllocaInst *A = dyn_cast(BBI)) { - DeadStackObjects.erase(A); + if (isa(BBI)) { + // Remove allocas from the list of dead stack objects; there can't be + // any references before the definition. + DeadStackObjects.remove(BBI); continue; } - if (CallInst *CI = extractMallocCall(BBI)) { - DeadStackObjects.erase(CI); - continue; - } + if (auto CS = CallSite(BBI)) { + // Remove allocation function calls from the list of dead stack objects; + // there can't be any references before the definition. + if (isAllocLikeFn(BBI, TLI)) + DeadStackObjects.remove(BBI); - if (CallSite CS = cast(BBI)) { // If this call does not access memory, it can't be loading any of our // pointers. if (AA->doesNotAccessMemory(CS)) @@ -744,40 +796,32 @@ bool DSE::handleEndBlock(BasicBlock &BB) { // If the call might load from any of our allocas, then any store above // the call is live. - SmallVector LiveAllocas; - for (SmallPtrSet::iterator I = DeadStackObjects.begin(), - E = DeadStackObjects.end(); I != E; ++I) { - // See if the call site touches it. - AliasAnalysis::ModRefResult A = - AA->getModRefInfo(CS, *I, getPointerSize(*I, *AA)); - - if (A == AliasAnalysis::ModRef || A == AliasAnalysis::Ref) - LiveAllocas.push_back(*I); - } + DeadStackObjects.remove_if([&](Value *I) { + // See if the call site touches the value. + ModRefInfo A = AA->getModRefInfo(CS, I, getPointerSize(I, DL, *TLI)); - for (SmallVector::iterator I = LiveAllocas.begin(), - E = LiveAllocas.end(); I != E; ++I) - DeadStackObjects.erase(*I); + return A == MRI_ModRef || A == MRI_Ref; + }); // If all of the allocas were clobbered by the call then we're not going // to find anything else to process. if (DeadStackObjects.empty()) - return MadeChange; + break; continue; } - AliasAnalysis::Location LoadedLoc; + MemoryLocation LoadedLoc; // If we encounter a use of the pointer, it is no longer considered dead if (LoadInst *L = dyn_cast(BBI)) { if (!L->isUnordered()) // Be conservative with atomic/volatile load break; - LoadedLoc = AA->getLocation(L); + LoadedLoc = MemoryLocation::get(L); } else if (VAArgInst *V = dyn_cast(BBI)) { - LoadedLoc = AA->getLocation(V); + LoadedLoc = MemoryLocation::get(V); } else if (MemTransferInst *MTI = dyn_cast(BBI)) { - LoadedLoc = AA->getLocationForSource(MTI); + LoadedLoc = MemoryLocation::getForSource(MTI); } else if (!BBI->mayReadFromMemory()) { // Instruction doesn't read memory. Note that stores that weren't removed // above will hit this case. @@ -789,7 +833,7 @@ bool DSE::handleEndBlock(BasicBlock &BB) { // Remove any allocas from the DeadPointer set that are loaded, as this // makes any stores above the access live. - RemoveAccessedObjects(LoadedLoc, DeadStackObjects); + RemoveAccessedObjects(LoadedLoc, DeadStackObjects, DL); // If all of the allocas were clobbered by the access then we're not going // to find anything else to process. @@ -803,9 +847,10 @@ bool DSE::handleEndBlock(BasicBlock &BB) { /// RemoveAccessedObjects - Check to see if the specified location may alias any /// of the stack objects in the DeadStackObjects set. If so, they become live /// because the location is being loaded. -void DSE::RemoveAccessedObjects(const AliasAnalysis::Location &LoadedLoc, - SmallPtrSet &DeadStackObjects) { - const Value *UnderlyingPointer = GetUnderlyingObject(LoadedLoc.Ptr); +void DSE::RemoveAccessedObjects(const MemoryLocation &LoadedLoc, + SmallSetVector &DeadStackObjects, + const DataLayout &DL) { + const Value *UnderlyingPointer = GetUnderlyingObject(LoadedLoc.Ptr, DL); // A constant can't be in the dead pointer set. if (isa(UnderlyingPointer)) @@ -814,21 +859,14 @@ void DSE::RemoveAccessedObjects(const AliasAnalysis::Location &LoadedLoc, // If the kill pointer can be easily reduced to an alloca, don't bother doing // extraneous AA queries. if (isa(UnderlyingPointer) || isa(UnderlyingPointer)) { - DeadStackObjects.erase(const_cast(UnderlyingPointer)); + DeadStackObjects.remove(const_cast(UnderlyingPointer)); return; } - SmallVector NowLive; - for (SmallPtrSet::iterator I = DeadStackObjects.begin(), - E = DeadStackObjects.end(); I != E; ++I) { + // Remove objects that could alias LoadedLoc. + DeadStackObjects.remove_if([&](Value *I) { // See if the loaded location could alias the stack location. - AliasAnalysis::Location StackLoc(*I, getPointerSize(*I, *AA)); - if (!AA->isNoAlias(StackLoc, LoadedLoc)) - NowLive.push_back(*I); - } - - for (SmallVector::iterator I = NowLive.begin(), E = NowLive.end(); - I != E; ++I) - DeadStackObjects.erase(*I); + MemoryLocation StackLoc(I, getPointerSize(I, DL, *TLI)); + return !AA->isNoAlias(StackLoc, LoadedLoc); + }); } -