diff --git a/lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp b/lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp
index 1db55fea008deebd2d75cfb627c20ff4b1fda7ec..54b1af8ea79199e223bc4bbb449bc0662b3747dc 100644 (file)
#include "llvm/ADT/Statistic.h"
#include "llvm/Analysis/Loads.h"
#include "llvm/IR/DataLayout.h"
+#include "llvm/IR/LLVMContext.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/Transforms/Utils/BasicBlockUtils.h"
#include "llvm/Transforms/Utils/Local.h"
using namespace llvm;
+#define DEBUG_TYPE "instcombine"
+
STATISTIC(NumDeadStore, "Number of dead stores eliminated");
STATISTIC(NumGlobalCopies, "Number of allocas copied from constant global");
static bool pointsToConstantGlobal(Value *V) {
if (GlobalVariable *GV = dyn_cast<GlobalVariable>(V))
return GV->isConstant();
- if (ConstantExpr *CE = dyn_cast<ConstantExpr>(V))
+
+ if (ConstantExpr *CE = dyn_cast<ConstantExpr>(V)) {
if (CE->getOpcode() == Instruction::BitCast ||
+ CE->getOpcode() == Instruction::AddrSpaceCast ||
CE->getOpcode() == Instruction::GetElementPtr)
return pointsToConstantGlobal(CE->getOperand(0));
+ }
return false;
}
/// can optimize this.
static bool
isOnlyCopiedFromConstantGlobal(Value *V, MemTransferInst *&TheCopy,
- SmallVectorImpl<Instruction *> &ToDelete,
- bool IsOffset = false) {
+ SmallVectorImpl<Instruction *> &ToDelete) {
// We track lifetime intrinsics as we encounter them. If we decide to go
// ahead and replace the value with the global, this lets the caller quickly
// eliminate the markers.
- for (Use &U : V->uses()) {
- Instruction *I = cast<Instruction>(U.getUser());
-
- if (LoadInst *LI = dyn_cast<LoadInst>(I)) {
- // Ignore non-volatile loads, they are always ok.
- if (!LI->isSimple()) return false;
- continue;
- }
-
- if (BitCastInst *BCI = dyn_cast<BitCastInst>(I)) {
- // If uses of the bitcast are ok, we are ok.
- if (!isOnlyCopiedFromConstantGlobal(BCI, TheCopy, ToDelete, IsOffset))
- return false;
- continue;
- }
- if (GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(I)) {
- // If the GEP has all zero indices, it doesn't offset the pointer. If it
- // doesn't, it does.
- if (!isOnlyCopiedFromConstantGlobal(
- GEP, TheCopy, ToDelete, IsOffset || !GEP->hasAllZeroIndices()))
- return false;
- continue;
- }
+ SmallVector<std::pair<Value *, bool>, 35> ValuesToInspect;
+ ValuesToInspect.push_back(std::make_pair(V, false));
+ while (!ValuesToInspect.empty()) {
+ auto ValuePair = ValuesToInspect.pop_back_val();
+ const bool IsOffset = ValuePair.second;
+ for (auto &U : ValuePair.first->uses()) {
+ Instruction *I = cast<Instruction>(U.getUser());
+
+ if (LoadInst *LI = dyn_cast<LoadInst>(I)) {
+ // Ignore non-volatile loads, they are always ok.
+ if (!LI->isSimple()) return false;
+ continue;
+ }
- if (CallSite CS = I) {
- // If this is the function being called then we treat it like a load and
- // ignore it.
- if (CS.isCallee(&U))
+ if (isa<BitCastInst>(I) || isa<AddrSpaceCastInst>(I)) {
+ // If uses of the bitcast are ok, we are ok.
+ ValuesToInspect.push_back(std::make_pair(I, IsOffset));
continue;
+ }
+ if (GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(I)) {
+ // If the GEP has all zero indices, it doesn't offset the pointer. If it
+ // doesn't, it does.
+ ValuesToInspect.push_back(
+ std::make_pair(I, IsOffset || !GEP->hasAllZeroIndices()));
+ continue;
+ }
- // Inalloca arguments are clobbered by the call.
- unsigned ArgNo = CS.getArgumentNo(&U);
- if (CS.isInAllocaArgument(ArgNo))
- return false;
+ if (CallSite CS = I) {
+ // If this is the function being called then we treat it like a load and
+ // ignore it.
+ if (CS.isCallee(&U))
+ continue;
- // If this is a readonly/readnone call site, then we know it is just a
- // load (but one that potentially returns the value itself), so we can
- // ignore it if we know that the value isn't captured.
- if (CS.onlyReadsMemory() &&
- (CS.getInstruction()->use_empty() || CS.doesNotCapture(ArgNo)))
- continue;
+ // Inalloca arguments are clobbered by the call.
+ unsigned ArgNo = CS.getArgumentNo(&U);
+ if (CS.isInAllocaArgument(ArgNo))
+ return false;
- // If this is being passed as a byval argument, the caller is making a
- // copy, so it is only a read of the alloca.
- if (CS.isByValArgument(ArgNo))
- continue;
- }
+ // If this is a readonly/readnone call site, then we know it is just a
+ // load (but one that potentially returns the value itself), so we can
+ // ignore it if we know that the value isn't captured.
+ if (CS.onlyReadsMemory() &&
+ (CS.getInstruction()->use_empty() || CS.doesNotCapture(ArgNo)))
+ continue;
+
+ // If this is being passed as a byval argument, the caller is making a
+ // copy, so it is only a read of the alloca.
+ if (CS.isByValArgument(ArgNo))
+ continue;
+ }
- // Lifetime intrinsics can be handled by the caller.
- if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(I)) {
- if (II->getIntrinsicID() == Intrinsic::lifetime_start ||
- II->getIntrinsicID() == Intrinsic::lifetime_end) {
- assert(II->use_empty() && "Lifetime markers have no result to use!");
- ToDelete.push_back(II);
- continue;
+ // Lifetime intrinsics can be handled by the caller.
+ if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(I)) {
+ if (II->getIntrinsicID() == Intrinsic::lifetime_start ||
+ II->getIntrinsicID() == Intrinsic::lifetime_end) {
+ assert(II->use_empty() && "Lifetime markers have no result to use!");
+ ToDelete.push_back(II);
+ continue;
+ }
}
- }
- // If this is isn't our memcpy/memmove, reject it as something we can't
- // handle.
- MemTransferInst *MI = dyn_cast<MemTransferInst>(I);
- if (MI == 0)
- return false;
+ // If this is isn't our memcpy/memmove, reject it as something we can't
+ // handle.
+ MemTransferInst *MI = dyn_cast<MemTransferInst>(I);
+ if (!MI)
+ return false;
- // If the transfer is using the alloca as a source of the transfer, then
- // ignore it since it is a load (unless the transfer is volatile).
- if (U.getOperandNo() == 1) {
- if (MI->isVolatile()) return false;
- continue;
- }
+ // If the transfer is using the alloca as a source of the transfer, then
+ // ignore it since it is a load (unless the transfer is volatile).
+ if (U.getOperandNo() == 1) {
+ if (MI->isVolatile()) return false;
+ continue;
+ }
- // If we already have seen a copy, reject the second one.
- if (TheCopy) return false;
+ // If we already have seen a copy, reject the second one.
+ if (TheCopy) return false;
- // If the pointer has been offset from the start of the alloca, we can't
- // safely handle this.
- if (IsOffset) return false;
+ // If the pointer has been offset from the start of the alloca, we can't
+ // safely handle this.
+ if (IsOffset) return false;
- // If the memintrinsic isn't using the alloca as the dest, reject it.
- if (U.getOperandNo() != 0) return false;
+ // If the memintrinsic isn't using the alloca as the dest, reject it.
+ if (U.getOperandNo() != 0) return false;
- // If the source of the memcpy/move is not a constant global, reject it.
- if (!pointsToConstantGlobal(MI->getSource()))
- return false;
+ // If the source of the memcpy/move is not a constant global, reject it.
+ if (!pointsToConstantGlobal(MI->getSource()))
+ return false;
- // Otherwise, the transform is safe. Remember the copy instruction.
- TheCopy = MI;
+ // Otherwise, the transform is safe. Remember the copy instruction.
+ TheCopy = MI;
+ }
}
return true;
}
static MemTransferInst *
isOnlyCopiedFromConstantGlobal(AllocaInst *AI,
SmallVectorImpl<Instruction *> &ToDelete) {
- MemTransferInst *TheCopy = 0;
+ MemTransferInst *TheCopy = nullptr;
if (isOnlyCopiedFromConstantGlobal(AI, TheCopy, ToDelete))
return TheCopy;
- return 0;
+ return nullptr;
}
Instruction *InstCombiner::visitAllocaInst(AllocaInst &AI) {
if (const ConstantInt *C = dyn_cast<ConstantInt>(AI.getArraySize())) {
Type *NewTy =
ArrayType::get(AI.getAllocatedType(), C->getZExtValue());
- AllocaInst *New = Builder->CreateAlloca(NewTy, 0, AI.getName());
+ AllocaInst *New = Builder->CreateAlloca(NewTy, nullptr, AI.getName());
New->setAlignment(AI.getAlignment());
// Scan to the end of the allocation instructions, to skip over a block of
// is only subsequently read.
SmallVector<Instruction *, 4> ToDelete;
if (MemTransferInst *Copy = isOnlyCopiedFromConstantGlobal(&AI, ToDelete)) {
- unsigned SourceAlign = getOrEnforceKnownAlignment(Copy->getSource(),
- AI.getAlignment(), DL);
+ unsigned SourceAlign = getOrEnforceKnownAlignment(
+ Copy->getSource(), AI.getAlignment(), DL, AC, &AI, DT);
if (AI.getAlignment() <= SourceAlign) {
DEBUG(dbgs() << "Found alloca equal to global: " << AI << '\n');
DEBUG(dbgs() << " memcpy = " << *Copy << '\n');
return visitAllocSite(AI);
}
+/// \brief Helper to combine a load to a new type.
+///
+/// This just does the work of combining a load to a new type. It handles
+/// metadata, etc., and returns the new instruction. The \c NewTy should be the
+/// loaded *value* type. This will convert it to a pointer, cast the operand to
+/// that pointer type, load it, etc.
+///
+/// Note that this will create all of the instructions with whatever insert
+/// point the \c InstCombiner currently is using.
+static LoadInst *combineLoadToNewType(InstCombiner &IC, LoadInst &LI, Type *NewTy) {
+ Value *Ptr = LI.getPointerOperand();
+ unsigned AS = LI.getPointerAddressSpace();
+ SmallVector<std::pair<unsigned, MDNode *>, 8> MD;
+ LI.getAllMetadata(MD);
+
+ LoadInst *NewLoad = IC.Builder->CreateAlignedLoad(
+ IC.Builder->CreateBitCast(Ptr, NewTy->getPointerTo(AS)),
+ LI.getAlignment(), LI.getName());
+ for (const auto &MDPair : MD) {
+ unsigned ID = MDPair.first;
+ MDNode *N = MDPair.second;
+ // Note, essentially every kind of metadata should be preserved here! This
+ // routine is supposed to clone a load instruction changing *only its type*.
+ // The only metadata it makes sense to drop is metadata which is invalidated
+ // when the pointer type changes. This should essentially never be the case
+ // in LLVM, but we explicitly switch over only known metadata to be
+ // conservatively correct. If you are adding metadata to LLVM which pertains
+ // to loads, you almost certainly want to add it here.
+ switch (ID) {
+ case LLVMContext::MD_dbg:
+ case LLVMContext::MD_tbaa:
+ case LLVMContext::MD_prof:
+ case LLVMContext::MD_fpmath:
+ case LLVMContext::MD_tbaa_struct:
+ case LLVMContext::MD_invariant_load:
+ case LLVMContext::MD_alias_scope:
+ case LLVMContext::MD_noalias:
+ case LLVMContext::MD_nontemporal:
+ case LLVMContext::MD_mem_parallel_loop_access:
+ case LLVMContext::MD_nonnull:
+ // All of these directly apply.
+ NewLoad->setMetadata(ID, N);
+ break;
+
+ case LLVMContext::MD_range:
+ // FIXME: It would be nice to propagate this in some way, but the type
+ // conversions make it hard.
+ break;
+ }
+ }
+ return NewLoad;
+}
+
+/// \brief Combine a store to a new type.
+///
+/// Returns the newly created store instruction.
+static StoreInst *combineStoreToNewValue(InstCombiner &IC, StoreInst &SI, Value *V) {
+ Value *Ptr = SI.getPointerOperand();
+ unsigned AS = SI.getPointerAddressSpace();
+ SmallVector<std::pair<unsigned, MDNode *>, 8> MD;
+ SI.getAllMetadata(MD);
+
+ StoreInst *NewStore = IC.Builder->CreateAlignedStore(
+ V, IC.Builder->CreateBitCast(Ptr, V->getType()->getPointerTo(AS)),
+ SI.getAlignment());
+ for (const auto &MDPair : MD) {
+ unsigned ID = MDPair.first;
+ MDNode *N = MDPair.second;
+ // Note, essentially every kind of metadata should be preserved here! This
+ // routine is supposed to clone a store instruction changing *only its
+ // type*. The only metadata it makes sense to drop is metadata which is
+ // invalidated when the pointer type changes. This should essentially
+ // never be the case in LLVM, but we explicitly switch over only known
+ // metadata to be conservatively correct. If you are adding metadata to
+ // LLVM which pertains to stores, you almost certainly want to add it
+ // here.
+ switch (ID) {
+ case LLVMContext::MD_dbg:
+ case LLVMContext::MD_tbaa:
+ case LLVMContext::MD_prof:
+ case LLVMContext::MD_fpmath:
+ case LLVMContext::MD_tbaa_struct:
+ case LLVMContext::MD_alias_scope:
+ case LLVMContext::MD_noalias:
+ case LLVMContext::MD_nontemporal:
+ case LLVMContext::MD_mem_parallel_loop_access:
+ case LLVMContext::MD_nonnull:
+ // All of these directly apply.
+ NewStore->setMetadata(ID, N);
+ break;
-/// InstCombineLoadCast - Fold 'load (cast P)' -> cast (load P)' when possible.
-static Instruction *InstCombineLoadCast(InstCombiner &IC, LoadInst &LI,
- const DataLayout *DL) {
- User *CI = cast<User>(LI.getOperand(0));
- Value *CastOp = CI->getOperand(0);
-
- PointerType *DestTy = cast<PointerType>(CI->getType());
- Type *DestPTy = DestTy->getElementType();
- if (PointerType *SrcTy = dyn_cast<PointerType>(CastOp->getType())) {
-
- // If the address spaces don't match, don't eliminate the cast.
- if (DestTy->getAddressSpace() != SrcTy->getAddressSpace())
- return 0;
-
- Type *SrcPTy = SrcTy->getElementType();
-
- if (DestPTy->isIntegerTy() || DestPTy->isPointerTy() ||
- DestPTy->isVectorTy()) {
- // If the source is an array, the code below will not succeed. Check to
- // see if a trivial 'gep P, 0, 0' will help matters. Only do this for
- // constants.
- if (ArrayType *ASrcTy = dyn_cast<ArrayType>(SrcPTy))
- if (Constant *CSrc = dyn_cast<Constant>(CastOp))
- if (ASrcTy->getNumElements() != 0) {
- Type *IdxTy = DL
- ? DL->getIntPtrType(SrcTy)
- : Type::getInt64Ty(SrcTy->getContext());
- Value *Idx = Constant::getNullValue(IdxTy);
- Value *Idxs[2] = { Idx, Idx };
- CastOp = ConstantExpr::getGetElementPtr(CSrc, Idxs);
- SrcTy = cast<PointerType>(CastOp->getType());
- SrcPTy = SrcTy->getElementType();
- }
-
- if (IC.getDataLayout() &&
- (SrcPTy->isIntegerTy() || SrcPTy->isPointerTy() ||
- SrcPTy->isVectorTy()) &&
- // Do not allow turning this into a load of an integer, which is then
- // casted to a pointer, this pessimizes pointer analysis a lot.
- (SrcPTy->isPtrOrPtrVectorTy() ==
- LI.getType()->isPtrOrPtrVectorTy()) &&
- IC.getDataLayout()->getTypeSizeInBits(SrcPTy) ==
- IC.getDataLayout()->getTypeSizeInBits(DestPTy)) {
-
- // Okay, we are casting from one integer or pointer type to another of
- // the same size. Instead of casting the pointer before the load, cast
- // the result of the loaded value.
- LoadInst *NewLoad =
- IC.Builder->CreateLoad(CastOp, LI.isVolatile(), CI->getName());
- NewLoad->setAlignment(LI.getAlignment());
- NewLoad->setAtomic(LI.getOrdering(), LI.getSynchScope());
- // Now cast the result of the load.
- return new BitCastInst(NewLoad, LI.getType());
+ case LLVMContext::MD_invariant_load:
+ case LLVMContext::MD_range:
+ break;
+ }
+ }
+
+ return NewStore;
+}
+
+/// \brief Combine loads to match the type of value their uses after looking
+/// through intervening bitcasts.
+///
+/// The core idea here is that if the result of a load is used in an operation,
+/// we should load the type most conducive to that operation. For example, when
+/// loading an integer and converting that immediately to a pointer, we should
+/// instead directly load a pointer.
+///
+/// However, this routine must never change the width of a load or the number of
+/// loads as that would introduce a semantic change. This combine is expected to
+/// be a semantic no-op which just allows loads to more closely model the types
+/// of their consuming operations.
+///
+/// Currently, we also refuse to change the precise type used for an atomic load
+/// or a volatile load. This is debatable, and might be reasonable to change
+/// later. However, it is risky in case some backend or other part of LLVM is
+/// relying on the exact type loaded to select appropriate atomic operations.
+static Instruction *combineLoadToOperationType(InstCombiner &IC, LoadInst &LI) {
+ // FIXME: We could probably with some care handle both volatile and atomic
+ // loads here but it isn't clear that this is important.
+ if (!LI.isSimple())
+ return nullptr;
+
+ if (LI.use_empty())
+ return nullptr;
+
+ Type *Ty = LI.getType();
+
+ // Try to canonicalize loads which are only ever stored to operate over
+ // integers instead of any other type. We only do this when the loaded type
+ // is sized and has a size exactly the same as its store size and the store
+ // size is a legal integer type.
+ const DataLayout *DL = IC.getDataLayout();
+ if (!Ty->isIntegerTy() && Ty->isSized() && DL &&
+ DL->isLegalInteger(DL->getTypeStoreSizeInBits(Ty)) &&
+ DL->getTypeStoreSizeInBits(Ty) == DL->getTypeSizeInBits(Ty)) {
+ if (std::all_of(LI.user_begin(), LI.user_end(), [&LI](User *U) {
+ auto *SI = dyn_cast<StoreInst>(U);
+ return SI && SI->getPointerOperand() != &LI;
+ })) {
+ LoadInst *NewLoad = combineLoadToNewType(
+ IC, LI,
+ Type::getIntNTy(LI.getContext(), DL->getTypeStoreSizeInBits(Ty)));
+ // Replace all the stores with stores of the newly loaded value.
+ for (auto UI = LI.user_begin(), UE = LI.user_end(); UI != UE;) {
+ auto *SI = cast<StoreInst>(*UI++);
+ IC.Builder->SetInsertPoint(SI);
+ combineStoreToNewValue(IC, *SI, NewLoad);
+ IC.EraseInstFromFunction(*SI);
}
+ assert(LI.use_empty() && "Failed to remove all users of the load!");
+ // Return the old load so the combiner can delete it safely.
+ return &LI;
}
}
- return 0;
+
+ // Fold away bit casts of the loaded value by loading the desired type.
+ if (LI.hasOneUse())
+ if (auto *BC = dyn_cast<BitCastInst>(LI.user_back())) {
+ LoadInst *NewLoad = combineLoadToNewType(IC, LI, BC->getDestTy());
+ BC->replaceAllUsesWith(NewLoad);
+ IC.EraseInstFromFunction(*BC);
+ return &LI;
+ }
+
+ // FIXME: We should also canonicalize loads of vectors when their elements are
+ // cast to other types.
+ return nullptr;
}
Instruction *InstCombiner::visitLoadInst(LoadInst &LI) {
Value *Op = LI.getOperand(0);
+ // Try to canonicalize the loaded type.
+ if (Instruction *Res = combineLoadToOperationType(*this, LI))
+ return Res;
+
// Attempt to improve the alignment.
if (DL) {
- unsigned KnownAlign =
- getOrEnforceKnownAlignment(Op, DL->getPrefTypeAlignment(LI.getType()),DL);
+ unsigned KnownAlign = getOrEnforceKnownAlignment(
+ Op, DL->getPrefTypeAlignment(LI.getType()), DL, AC, &LI, DT);
unsigned LoadAlign = LI.getAlignment();
unsigned EffectiveLoadAlign = LoadAlign != 0 ? LoadAlign :
DL->getABITypeAlignment(LI.getType());
LI.setAlignment(EffectiveLoadAlign);
}
- // load (cast X) --> cast (load X) iff safe.
- if (isa<CastInst>(Op))
- if (Instruction *Res = InstCombineLoadCast(*this, LI, DL))
- return Res;
-
// None of the following transforms are legal for volatile/atomic loads.
// FIXME: Some of it is okay for atomic loads; needs refactoring.
- if (!LI.isSimple()) return 0;
+ if (!LI.isSimple()) return nullptr;
// Do really simple store-to-load forwarding and load CSE, to catch cases
// where there are several consecutive memory accesses to the same location,
// separated by a few arithmetic operations.
BasicBlock::iterator BBI = &LI;
if (Value *AvailableVal = FindAvailableLoadedValue(Op, LI.getParent(), BBI,6))
- return ReplaceInstUsesWith(LI, AvailableVal);
+ return ReplaceInstUsesWith(
+ LI, Builder->CreateBitOrPointerCast(AvailableVal, LI.getType(),
+ LI.getName() + ".cast"));
// load(gep null, ...) -> unreachable
if (GetElementPtrInst *GEPI = dyn_cast<GetElementPtrInst>(Op)) {
return ReplaceInstUsesWith(LI, UndefValue::get(LI.getType()));
}
- // Instcombine load (constantexpr_cast global) -> cast (load global)
- if (ConstantExpr *CE = dyn_cast<ConstantExpr>(Op))
- if (CE->isCast())
- if (Instruction *Res = InstCombineLoadCast(*this, LI, DL))
- return Res;
-
if (Op->hasOneUse()) {
// Change select and PHI nodes to select values instead of addresses: this
// helps alias analysis out a lot, allows many others simplifications, and
}
// load (select (cond, null, P)) -> load P
- if (Constant *C = dyn_cast<Constant>(SI->getOperand(1)))
- if (C->isNullValue()) {
- LI.setOperand(0, SI->getOperand(2));
- return &LI;
- }
+ if (isa<ConstantPointerNull>(SI->getOperand(1)) &&
+ LI.getPointerAddressSpace() == 0) {
+ LI.setOperand(0, SI->getOperand(2));
+ return &LI;
+ }
// load (select (cond, P, null)) -> load P
- if (Constant *C = dyn_cast<Constant>(SI->getOperand(2)))
- if (C->isNullValue()) {
- LI.setOperand(0, SI->getOperand(1));
- return &LI;
- }
+ if (isa<ConstantPointerNull>(SI->getOperand(2)) &&
+ LI.getPointerAddressSpace() == 0) {
+ LI.setOperand(0, SI->getOperand(1));
+ return &LI;
+ }
}
}
- return 0;
+ return nullptr;
}
-/// InstCombineStoreToCast - Fold store V, (cast P) -> store (cast V), P
-/// when possible. This makes it generally easy to do alias analysis and/or
-/// SROA/mem2reg of the memory object.
-static Instruction *InstCombineStoreToCast(InstCombiner &IC, StoreInst &SI) {
- User *CI = cast<User>(SI.getOperand(1));
- Value *CastOp = CI->getOperand(0);
-
- Type *DestPTy = cast<PointerType>(CI->getType())->getElementType();
- PointerType *SrcTy = dyn_cast<PointerType>(CastOp->getType());
- if (SrcTy == 0) return 0;
-
- Type *SrcPTy = SrcTy->getElementType();
-
- if (!DestPTy->isIntegerTy() && !DestPTy->isPointerTy())
- return 0;
-
- /// NewGEPIndices - If SrcPTy is an aggregate type, we can emit a "noop gep"
- /// to its first element. This allows us to handle things like:
- /// store i32 xxx, (bitcast {foo*, float}* %P to i32*)
- /// on 32-bit hosts.
- SmallVector<Value*, 4> NewGEPIndices;
-
- // If the source is an array, the code below will not succeed. Check to
- // see if a trivial 'gep P, 0, 0' will help matters. Only do this for
- // constants.
- if (SrcPTy->isArrayTy() || SrcPTy->isStructTy()) {
- // Index through pointer.
- Constant *Zero = Constant::getNullValue(Type::getInt32Ty(SI.getContext()));
- NewGEPIndices.push_back(Zero);
-
- while (1) {
- if (StructType *STy = dyn_cast<StructType>(SrcPTy)) {
- if (!STy->getNumElements()) /* Struct can be empty {} */
- break;
- NewGEPIndices.push_back(Zero);
- SrcPTy = STy->getElementType(0);
- } else if (ArrayType *ATy = dyn_cast<ArrayType>(SrcPTy)) {
- NewGEPIndices.push_back(Zero);
- SrcPTy = ATy->getElementType();
- } else {
- break;
- }
- }
+/// \brief Combine stores to match the type of value being stored.
+///
+/// The core idea here is that the memory does not have any intrinsic type and
+/// where we can we should match the type of a store to the type of value being
+/// stored.
+///
+/// However, this routine must never change the width of a store or the number of
+/// stores as that would introduce a semantic change. This combine is expected to
+/// be a semantic no-op which just allows stores to more closely model the types
+/// of their incoming values.
+///
+/// Currently, we also refuse to change the precise type used for an atomic or
+/// volatile store. This is debatable, and might be reasonable to change later.
+/// However, it is risky in case some backend or other part of LLVM is relying
+/// on the exact type stored to select appropriate atomic operations.
+///
+/// \returns true if the store was successfully combined away. This indicates
+/// the caller must erase the store instruction. We have to let the caller erase
+/// the store instruction sas otherwise there is no way to signal whether it was
+/// combined or not: IC.EraseInstFromFunction returns a null pointer.
+static bool combineStoreToValueType(InstCombiner &IC, StoreInst &SI) {
+ // FIXME: We could probably with some care handle both volatile and atomic
+ // stores here but it isn't clear that this is important.
+ if (!SI.isSimple())
+ return false;
- SrcTy = PointerType::get(SrcPTy, SrcTy->getAddressSpace());
- }
+ Value *V = SI.getValueOperand();
- if (!SrcPTy->isIntegerTy() && !SrcPTy->isPointerTy())
- return 0;
-
- // If the pointers point into different address spaces or if they point to
- // values with different sizes, we can't do the transformation.
- if (!IC.getDataLayout() ||
- SrcTy->getAddressSpace() !=
- cast<PointerType>(CI->getType())->getAddressSpace() ||
- IC.getDataLayout()->getTypeSizeInBits(SrcPTy) !=
- IC.getDataLayout()->getTypeSizeInBits(DestPTy))
- return 0;
-
- // Okay, we are casting from one integer or pointer type to another of
- // the same size. Instead of casting the pointer before
- // the store, cast the value to be stored.
- Value *NewCast;
- Value *SIOp0 = SI.getOperand(0);
- Instruction::CastOps opcode = Instruction::BitCast;
- Type* CastSrcTy = SIOp0->getType();
- Type* CastDstTy = SrcPTy;
- if (CastDstTy->isPointerTy()) {
- if (CastSrcTy->isIntegerTy())
- opcode = Instruction::IntToPtr;
- } else if (CastDstTy->isIntegerTy()) {
- if (SIOp0->getType()->isPointerTy())
- opcode = Instruction::PtrToInt;
+ // Fold away bit casts of the stored value by storing the original type.
+ if (auto *BC = dyn_cast<BitCastInst>(V)) {
+ V = BC->getOperand(0);
+ combineStoreToNewValue(IC, SI, V);
+ return true;
}
- // SIOp0 is a pointer to aggregate and this is a store to the first field,
- // emit a GEP to index into its first field.
- if (!NewGEPIndices.empty())
- CastOp = IC.Builder->CreateInBoundsGEP(CastOp, NewGEPIndices);
-
- NewCast = IC.Builder->CreateCast(opcode, SIOp0, CastDstTy,
- SIOp0->getName()+".c");
- SI.setOperand(0, NewCast);
- SI.setOperand(1, CastOp);
- return &SI;
+ // FIXME: We should also canonicalize loads of vectors when their elements are
+ // cast to other types.
+ return false;
}
/// equivalentAddressValues - Test if A and B will obviously have the same
Value *Val = SI.getOperand(0);
Value *Ptr = SI.getOperand(1);
+ // Try to canonicalize the stored type.
+ if (combineStoreToValueType(*this, SI))
+ return EraseInstFromFunction(SI);
+
// Attempt to improve the alignment.
if (DL) {
- unsigned KnownAlign =
- getOrEnforceKnownAlignment(Ptr, DL->getPrefTypeAlignment(Val->getType()),
- DL);
+ unsigned KnownAlign = getOrEnforceKnownAlignment(
+ Ptr, DL->getPrefTypeAlignment(Val->getType()), DL, AC, &SI, DT);
unsigned StoreAlign = SI.getAlignment();
unsigned EffectiveStoreAlign = StoreAlign != 0 ? StoreAlign :
DL->getABITypeAlignment(Val->getType());
// Don't hack volatile/atomic stores.
// FIXME: Some bits are legal for atomic stores; needs refactoring.
- if (!SI.isSimple()) return 0;
+ if (!SI.isSimple()) return nullptr;
// If the RHS is an alloca with a single use, zapify the store, making the
// alloca dead.
if (Instruction *U = dyn_cast<Instruction>(Val))
Worklist.Add(U); // Dropped a use.
}
- return 0; // Do not modify these!
+ return nullptr; // Do not modify these!
}
// store undef, Ptr -> noop
if (isa<UndefValue>(Val))
return EraseInstFromFunction(SI);
- // If the pointer destination is a cast, see if we can fold the cast into the
- // source instead.
- if (isa<CastInst>(Ptr))
- if (Instruction *Res = InstCombineStoreToCast(*this, SI))
- return Res;
- if (ConstantExpr *CE = dyn_cast<ConstantExpr>(Ptr))
- if (CE->isCast())
- if (Instruction *Res = InstCombineStoreToCast(*this, SI))
- return Res;
-
-
// If this store is the last instruction in the basic block (possibly
// excepting debug info instructions), and if the block ends with an
// unconditional branch, try to move it to the successor block.
if (BranchInst *BI = dyn_cast<BranchInst>(BBI))
if (BI->isUnconditional())
if (SimplifyStoreAtEndOfBlock(SI))
- return 0; // xform done!
+ return nullptr; // xform done!
- return 0;
+ return nullptr;
}
/// SimplifyStoreAtEndOfBlock - Turn things like:
// the other predecessor.
pred_iterator PI = pred_begin(DestBB);
BasicBlock *P = *PI;
- BasicBlock *OtherBB = 0;
+ BasicBlock *OtherBB = nullptr;
if (P != StoreBB)
OtherBB = P;
// If the other block ends in an unconditional branch, check for the 'if then
// else' case. there is an instruction before the branch.
- StoreInst *OtherStore = 0;
+ StoreInst *OtherStore = nullptr;
if (OtherBr->isUnconditional()) {
--BBI;
// Skip over debugging info.
InsertNewInstBefore(NewSI, *BBI);
NewSI->setDebugLoc(OtherStore->getDebugLoc());
- // If the two stores had the same TBAA tag, preserve it.
- if (MDNode *TBAATag = SI.getMetadata(LLVMContext::MD_tbaa))
- if ((TBAATag = MDNode::getMostGenericTBAA(TBAATag,
- OtherStore->getMetadata(LLVMContext::MD_tbaa))))
- NewSI->setMetadata(LLVMContext::MD_tbaa, TBAATag);
-
+ // If the two stores had AA tags, merge them.
+ AAMDNodes AATags;
+ SI.getAAMetadata(AATags);
+ if (AATags) {
+ OtherStore->getAAMetadata(AATags, /* Merge = */ true);
+ NewSI->setAAMetadata(AATags);
+ }
// Nuke the old stores.
EraseInstFromFunction(SI);