#include "InstCombine.h"
#include "llvm/Support/CallSite.h"
-#include "llvm/Target/TargetData.h"
+#include "llvm/DataLayout.h"
#include "llvm/Analysis/MemoryBuiltins.h"
#include "llvm/Transforms/Utils/BuildLibCalls.h"
#include "llvm/Transforms/Utils/Local.h"
return Ty;
}
+/// reduceToSingleValueType - Given an aggregate type which ultimately holds a
+/// single scalar element, like {{{type}}} or [1 x type], return type.
+static Type *reduceToSingleValueType(Type *T) {
+ while (!T->isSingleValueType()) {
+ if (StructType *STy = dyn_cast<StructType>(T)) {
+ if (STy->getNumElements() == 1)
+ T = STy->getElementType(0);
+ else
+ break;
+ } else if (ArrayType *ATy = dyn_cast<ArrayType>(T)) {
+ if (ATy->getNumElements() == 1)
+ T = ATy->getElementType();
+ else
+ break;
+ } else
+ break;
+ }
+
+ return T;
+}
Instruction *InstCombiner::SimplifyMemTransfer(MemIntrinsic *MI) {
unsigned DstAlign = getKnownAlignment(MI->getArgOperand(0), TD);
// if the size is something we can handle with a single primitive load/store.
// A single load+store correctly handles overlapping memory in the memmove
// case.
- unsigned Size = MemOpLength->getZExtValue();
- if (Size == 0) return MI; // Delete this mem transfer.
+ uint64_t Size = MemOpLength->getLimitedValue();
+ assert(Size && "0-sized memory transfering should be removed already.");
if (Size > 8 || (Size&(Size-1)))
return 0; // If not 1/2/4/8 bytes, exit.
// dest address will be promotable. See if we can find a better type than the
// integer datatype.
Value *StrippedDest = MI->getArgOperand(0)->stripPointerCasts();
+ MDNode *CopyMD = 0;
if (StrippedDest != MI->getArgOperand(0)) {
Type *SrcETy = cast<PointerType>(StrippedDest->getType())
->getElementType();
if (TD && SrcETy->isSized() && TD->getTypeStoreSize(SrcETy) == Size) {
// The SrcETy might be something like {{{double}}} or [1 x double]. Rip
// down through these levels if so.
- while (!SrcETy->isSingleValueType()) {
- if (StructType *STy = dyn_cast<StructType>(SrcETy)) {
- if (STy->getNumElements() == 1)
- SrcETy = STy->getElementType(0);
- else
- break;
- } else if (ArrayType *ATy = dyn_cast<ArrayType>(SrcETy)) {
- if (ATy->getNumElements() == 1)
- SrcETy = ATy->getElementType();
- else
- break;
- } else
- break;
- }
+ SrcETy = reduceToSingleValueType(SrcETy);
if (SrcETy->isSingleValueType()) {
NewSrcPtrTy = PointerType::get(SrcETy, SrcAddrSp);
NewDstPtrTy = PointerType::get(SrcETy, DstAddrSp);
+
+ // If the memcpy has metadata describing the members, see if we can
+ // get the TBAA tag describing our copy.
+ if (MDNode *M = MI->getMetadata(LLVMContext::MD_tbaa_struct)) {
+ if (M->getNumOperands() == 3 &&
+ isa<ConstantInt>(M->getOperand(0)) &&
+ cast<ConstantInt>(M->getOperand(0))->isNullValue() &&
+ isa<ConstantInt>(M->getOperand(1)) &&
+ cast<ConstantInt>(M->getOperand(1))->getValue() == Size &&
+ isa<MDNode>(M->getOperand(2)))
+ CopyMD = cast<MDNode>(M->getOperand(2));
+ }
}
}
}
-
// If the memcpy/memmove provides better alignment info than we can
// infer, use it.
SrcAlign = std::max(SrcAlign, CopyAlign);
Value *Dest = Builder->CreateBitCast(MI->getArgOperand(0), NewDstPtrTy);
LoadInst *L = Builder->CreateLoad(Src, MI->isVolatile());
L->setAlignment(SrcAlign);
+ if (CopyMD)
+ L->setMetadata(LLVMContext::MD_tbaa, CopyMD);
StoreInst *S = Builder->CreateStore(L, Dest, MI->isVolatile());
S->setAlignment(DstAlign);
+ if (CopyMD)
+ S->setMetadata(LLVMContext::MD_tbaa, CopyMD);
// Set the size of the copy to 0, it will be deleted on the next iteration.
MI->setArgOperand(2, Constant::getNullValue(MemOpLength->getType()));
ConstantInt *FillC = dyn_cast<ConstantInt>(MI->getValue());
if (!LenC || !FillC || !FillC->getType()->isIntegerTy(8))
return 0;
- uint64_t Len = LenC->getZExtValue();
+ uint64_t Len = LenC->getLimitedValue();
Alignment = MI->getAlignment();
-
- // If the length is zero, this is a no-op
- if (Len == 0) return MI; // memset(d,c,0,a) -> noop
+ assert(Len && "0-sized memory setting should be removed already.");
// memset(s,c,n) -> store s, c (for n=1,2,4,8)
if (Len <= 8 && isPowerOf2_32((uint32_t)Len)) {
/// the heavy lifting.
///
Instruction *InstCombiner::visitCallInst(CallInst &CI) {
- if (isFreeCall(&CI))
+ if (isFreeCall(&CI, TLI))
return visitFree(CI);
- if (isMalloc(&CI))
- return visitMalloc(CI);
// If the caller function is nounwind, mark the call as nounwind, even if the
// callee isn't.
switch (II->getIntrinsicID()) {
default: break;
case Intrinsic::objectsize: {
- // We need target data for just about everything so depend on it.
- if (!TD) break;
-
- Type *ReturnTy = CI.getType();
- uint64_t DontKnow = II->getArgOperand(1) == Builder->getTrue() ? 0 : -1ULL;
-
- // Get to the real allocated thing and offset as fast as possible.
- Value *Op1 = II->getArgOperand(0)->stripPointerCasts();
-
- uint64_t Offset = 0;
- uint64_t Size = -1ULL;
-
- // Try to look through constant GEPs.
- if (GEPOperator *GEP = dyn_cast<GEPOperator>(Op1)) {
- if (!GEP->hasAllConstantIndices()) break;
-
- // Get the current byte offset into the thing. Use the original
- // operand in case we're looking through a bitcast.
- SmallVector<Value*, 8> Ops(GEP->idx_begin(), GEP->idx_end());
- if (!GEP->getPointerOperandType()->isPointerTy())
- return 0;
- Offset = TD->getIndexedOffset(GEP->getPointerOperandType(), Ops);
-
- Op1 = GEP->getPointerOperand()->stripPointerCasts();
-
- // Make sure we're not a constant offset from an external
- // global.
- if (GlobalVariable *GV = dyn_cast<GlobalVariable>(Op1))
- if (!GV->hasDefinitiveInitializer()) break;
- }
-
- // If we've stripped down to a single global variable that we
- // can know the size of then just return that.
- if (GlobalVariable *GV = dyn_cast<GlobalVariable>(Op1)) {
- if (GV->hasDefinitiveInitializer()) {
- Constant *C = GV->getInitializer();
- Size = TD->getTypeAllocSize(C->getType());
- } else {
- // Can't determine size of the GV.
- Constant *RetVal = ConstantInt::get(ReturnTy, DontKnow);
- return ReplaceInstUsesWith(CI, RetVal);
- }
- } else if (AllocaInst *AI = dyn_cast<AllocaInst>(Op1)) {
- // Get alloca size.
- if (AI->getAllocatedType()->isSized()) {
- Size = TD->getTypeAllocSize(AI->getAllocatedType());
- if (AI->isArrayAllocation()) {
- const ConstantInt *C = dyn_cast<ConstantInt>(AI->getArraySize());
- if (!C) break;
- Size *= C->getZExtValue();
- }
- }
- } else if (CallInst *MI = extractMallocCall(Op1)) {
- // Get allocation size.
- Type* MallocType = getMallocAllocatedType(MI);
- if (MallocType && MallocType->isSized())
- if (Value *NElems = getMallocArraySize(MI, TD, true))
- if (ConstantInt *NElements = dyn_cast<ConstantInt>(NElems))
- Size = NElements->getZExtValue() * TD->getTypeAllocSize(MallocType);
- }
-
- // Do not return "I don't know" here. Later optimization passes could
- // make it possible to evaluate objectsize to a constant.
- if (Size == -1ULL)
- break;
-
- if (Size < Offset) {
- // Out of bound reference? Negative index normalized to large
- // index? Just return "I don't know".
- return ReplaceInstUsesWith(CI, ConstantInt::get(ReturnTy, DontKnow));
- }
- return ReplaceInstUsesWith(CI, ConstantInt::get(ReturnTy, Size-Offset));
+ uint64_t Size;
+ if (getObjectSize(II->getArgOperand(0), Size, TD, TLI))
+ return ReplaceInstUsesWith(CI, ConstantInt::get(CI.getType(), Size));
+ return 0;
}
case Intrinsic::bswap:
// bswap(bswap(x)) -> x
TerminatorInst *TI = II->getParent()->getTerminator();
bool CannotRemove = false;
for (++BI; &*BI != TI; ++BI) {
- if (isa<AllocaInst>(BI) || isMalloc(BI)) {
+ if (isa<AllocaInst>(BI)) {
CannotRemove = true;
break;
}
/// passed through the varargs area, we can eliminate the use of the cast.
static bool isSafeToEliminateVarargsCast(const CallSite CS,
const CastInst * const CI,
- const TargetData * const TD,
+ const DataLayout * const TD,
const int ix) {
if (!CI->isLosslessCast())
return false;
// Currently we're only working with the checking functions, memcpy_chk,
// mempcpy_chk, memmove_chk, memset_chk, strcpy_chk, stpcpy_chk, strncpy_chk,
// strcat_chk and strncat_chk.
-Instruction *InstCombiner::tryOptimizeCall(CallInst *CI, const TargetData *TD) {
+Instruction *InstCombiner::tryOptimizeCall(CallInst *CI, const DataLayout *TD) {
if (CI->getCalledFunction() == 0) return 0;
InstCombineFortifiedLibCalls Simplifier(this);
- Simplifier.fold(CI, TD);
+ Simplifier.fold(CI, TD, TLI);
return Simplifier.NewInstruction;
}
// visitCallSite - Improvements for call and invoke instructions.
//
Instruction *InstCombiner::visitCallSite(CallSite CS) {
+ if (isAllocLikeFn(CS.getInstruction(), TLI))
+ return visitAllocSite(*CS.getInstruction());
+
bool Changed = false;
// If the callee is a pointer to a function, attempt to move any casts to the
}
if (isa<ConstantPointerNull>(Callee) || isa<UndefValue>(Callee)) {
- // This instruction is not reachable, just remove it. We insert a store to
- // undef so that we know that this code is not reachable, despite the fact
- // that we can't modify the CFG here.
- new StoreInst(ConstantInt::getTrue(Callee->getContext()),
- UndefValue::get(Type::getInt1PtrTy(Callee->getContext())),
- CS.getInstruction());
-
// If CS does not return void then replaceAllUsesWith undef.
// This allows ValueHandlers and custom metadata to adjust itself.
if (!CS.getInstruction()->getType()->isVoidTy())
ReplaceInstUsesWith(*CS.getInstruction(),
UndefValue::get(CS.getInstruction()->getType()));
- if (InvokeInst *II = dyn_cast<InvokeInst>(CS.getInstruction())) {
- // Don't break the CFG, insert a dummy cond branch.
- BranchInst::Create(II->getNormalDest(), II->getUnwindDest(),
- ConstantInt::getTrue(Callee->getContext()), II);
+ if (isa<InvokeInst>(CS.getInstruction())) {
+ // Can't remove an invoke because we cannot change the CFG.
+ return 0;
}
+
+ // This instruction is not reachable, just remove it. We insert a store to
+ // undef so that we know that this code is not reachable, despite the fact
+ // that we can't modify the CFG here.
+ new StoreInst(ConstantInt::getTrue(Callee->getContext()),
+ UndefValue::get(Type::getInt1PtrTy(Callee->getContext())),
+ CS.getInstruction());
+
return EraseInstFromFunction(*CS.getInstruction());
}
Changed = true;
}
- // Try to optimize the call if possible, we require TargetData for most of
+ // Try to optimize the call if possible, we require DataLayout for most of
// this. None of these calls are seen as possibly dead so go ahead and
// delete the instruction now.
if (CallInst *CI = dyn_cast<CallInst>(CS.getInstruction())) {
if (!CallerPAL.isEmpty() && !Caller->use_empty()) {
Attributes RAttrs = CallerPAL.getRetAttributes();
- if (RAttrs & Attribute::typeIncompatible(NewRetTy))
+ if (RAttrs & Attributes::typeIncompatible(NewRetTy))
return false; // Attribute not compatible with transformed value.
}
return false; // Cannot transform this parameter value.
Attributes Attrs = CallerPAL.getParamAttributes(i + 1);
- if (Attrs & Attribute::typeIncompatible(ParamTy))
+ if (Attrs & Attributes::typeIncompatible(ParamTy))
return false; // Attribute not compatible with transformed value.
// If the parameter is passed as a byval argument, then we have to have a
// If the return value is not being used, the type may not be compatible
// with the existing attributes. Wipe out any problematic attributes.
- RAttrs &= ~Attribute::typeIncompatible(NewRetTy);
+ RAttrs &= ~Attributes::typeIncompatible(NewRetTy);
// Add the new return attributes.
if (RAttrs)
if (NewRetTy->isVoidTy())
Caller->setName(""); // Void type should not have a name.
- const AttrListPtr &NewCallerPAL = AttrListPtr::get(attrVec.begin(),
- attrVec.end());
+ const AttrListPtr &NewCallerPAL = AttrListPtr::get(attrVec);
Instruction *NC;
if (InvokeInst *II = dyn_cast<InvokeInst>(Caller)) {
NestF->getType() == PointerType::getUnqual(NewFTy) ?
NestF : ConstantExpr::getBitCast(NestF,
PointerType::getUnqual(NewFTy));
- const AttrListPtr &NewPAL = AttrListPtr::get(NewAttrs.begin(),
- NewAttrs.end());
+ const AttrListPtr &NewPAL = AttrListPtr::get(NewAttrs);
Instruction *NewCaller;
if (InvokeInst *II = dyn_cast<InvokeInst>(Caller)) {