// Returns a clone of `I` with its operands converted to those specified in
// ValueWithNewAddrSpace. Due to potential cycles in the data flow graph, an
// operand whose address space needs to be modified might not exist in
// ValueWithNewAddrSpace. In that case, uses undef as a placeholder operand and
// adds that operand use to UndefUsesToFix so that caller can fix them later.
//
// Note that we do not necessarily clone `I`, e.g., if it is an addrspacecast
// from a pointer whose type already matches. Therefore, this function returns a
// Value* instead of an Instruction*.
static Value *cloneInstructionWithNewAddressSpace(
Instruction *I, unsigned NewAddrSpace,
const ValueToValueMapTy &ValueWithNewAddrSpace,
SmallVectorImpl<const Use *> *UndefUsesToFix) {
Type *NewPtrType =
I->getType()->getPointerElementType()->getPointerTo(NewAddrSpace);
if (I->getOpcode() == Instruction::AddrSpaceCast) {
Value *Src = I->getOperand(0);
// Because `I` is flat, the source address space must be specific.
// Therefore, the inferred address space must be the source space, according
// to our algorithm.
assert(Src->getType()->getPointerAddressSpace() == NewAddrSpace);
if (Src->getType() != NewPtrType)
return new BitCastInst(Src, NewPtrType);
return Src;
}
// Computes the converted pointer operands.
SmallVector<Value *, 4> NewPointerOperands;
for (const Use &OperandUse : I->operands()) {
if (!OperandUse.get()->getType()->isPointerTy())
NewPointerOperands.push_back(nullptr);
else
NewPointerOperands.push_back(operandWithNewAddressSpaceOrCreateUndef(
OperandUse, NewAddrSpace, ValueWithNewAddrSpace, UndefUsesToFix));
}
switch (I->getOpcode()) {
case Instruction::BitCast:
return new BitCastInst(NewPointerOperands[0], NewPtrType);
case Instruction::PHI: {
assert(I->getType()->isPointerTy());
PHINode *PHI = cast<PHINode>(I);
PHINode *NewPHI = PHINode::Create(NewPtrType, PHI->getNumIncomingValues());
for (unsigned Index = 0; Index < PHI->getNumIncomingValues(); ++Index) {
unsigned OperandNo = PHINode::getOperandNumForIncomingValue(Index);
NewPHI->addIncoming(NewPointerOperands[OperandNo],
PHI->getIncomingBlock(Index));
}
return NewPHI;
}
case Instruction::GetElementPtr: {
GetElementPtrInst *GEP = cast<GetElementPtrInst>(I);
GetElementPtrInst *NewGEP = GetElementPtrInst::Create(
GEP->getSourceElementType(), NewPointerOperands[0],
SmallVector<Value *, 4>(GEP->idx_begin(), GEP->idx_end()));
NewGEP->setIsInBounds(GEP->isInBounds());
return NewGEP;
}
case Instruction::Select: {
assert(I->getType()->isPointerTy());
return SelectInst::Create(I->getOperand(0), NewPointerOperands[1],
NewPointerOperands[2], "", nullptr, I);
}
default:
llvm_unreachable("Unexpected opcode");
}
}
TEST(CloneInstruction, Inbounds) {
LLVMContext context;
Value *V = new Argument(Type::getInt32PtrTy(context));
Constant *Z = Constant::getNullValue(Type::getInt32Ty(context));
std::vector<Value *> ops;
ops.push_back(Z);
GetElementPtrInst *GEP = GetElementPtrInst::Create(V, ops.begin(), ops.end());
EXPECT_FALSE(cast<GetElementPtrInst>(GEP->clone())->isInBounds());
GEP->setIsInBounds();
EXPECT_TRUE(cast<GetElementPtrInst>(GEP->clone())->isInBounds());
}
Value *NVPTXFavorNonGenericAddrSpaces::hoistAddrSpaceCastFromGEP(
GEPOperator *GEP, int Depth) {
Value *NewOperand =
hoistAddrSpaceCastFrom(GEP->getPointerOperand(), Depth + 1);
if (NewOperand == nullptr)
return nullptr;
// hoistAddrSpaceCastFrom returns an eliminable addrspacecast or nullptr.
assert(isEliminableAddrSpaceCast(NewOperand));
Operator *Cast = cast<Operator>(NewOperand);
SmallVector<Value *, 8> Indices(GEP->idx_begin(), GEP->idx_end());
Value *NewASC;
if (Instruction *GEPI = dyn_cast<Instruction>(GEP)) {
// GEP = gep (addrspacecast X), indices
// =>
// NewGEP = gep X, indices
// NewASC = addrspacecast NewGEP
GetElementPtrInst *NewGEP = GetElementPtrInst::Create(
GEP->getSourceElementType(), Cast->getOperand(0), Indices,
"", GEPI);
NewGEP->setIsInBounds(GEP->isInBounds());
NewASC = new AddrSpaceCastInst(NewGEP, GEP->getType(), "", GEPI);
NewASC->takeName(GEP);
// Without RAUWing GEP, the compiler would visit GEP again and emit
// redundant instructions. This is exercised in test @rauw in
// access-non-generic.ll.
GEP->replaceAllUsesWith(NewASC);
} else {
// GEP is a constant expression.
Constant *NewGEP = ConstantExpr::getGetElementPtr(
GEP->getSourceElementType(), cast<Constant>(Cast->getOperand(0)),
Indices, GEP->isInBounds());
NewASC = ConstantExpr::getAddrSpaceCast(NewGEP, GEP->getType());
}
return NewASC;
}
/// InsertPHITranslatedPointer - Insert a computation of the PHI translated
/// version of 'V' for the edge PredBB->CurBB into the end of the PredBB
/// block. All newly created instructions are added to the NewInsts list.
/// This returns null on failure.
///
Value *PHITransAddr::
InsertPHITranslatedSubExpr(Value *InVal, BasicBlock *CurBB,
BasicBlock *PredBB, const DominatorTree &DT,
SmallVectorImpl<Instruction*> &NewInsts) {
// See if we have a version of this value already available and dominating
// PredBB. If so, there is no need to insert a new instance of it.
PHITransAddr Tmp(InVal, DL, AC);
if (!Tmp.PHITranslateValue(CurBB, PredBB, &DT, /*MustDominate=*/true))
return Tmp.getAddr();
// We don't need to PHI translate values which aren't instructions.
auto *Inst = dyn_cast<Instruction>(InVal);
if (!Inst)
return nullptr;
// Handle cast of PHI translatable value.
if (CastInst *Cast = dyn_cast<CastInst>(Inst)) {
if (!isSafeToSpeculativelyExecute(Cast)) return nullptr;
Value *OpVal = InsertPHITranslatedSubExpr(Cast->getOperand(0),
CurBB, PredBB, DT, NewInsts);
if (!OpVal) return nullptr;
// Otherwise insert a cast at the end of PredBB.
CastInst *New = CastInst::Create(Cast->getOpcode(), OpVal, InVal->getType(),
InVal->getName() + ".phi.trans.insert",
PredBB->getTerminator());
New->setDebugLoc(Inst->getDebugLoc());
NewInsts.push_back(New);
return New;
}
// Handle getelementptr with at least one PHI operand.
if (GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(Inst)) {
SmallVector<Value*, 8> GEPOps;
BasicBlock *CurBB = GEP->getParent();
for (unsigned i = 0, e = GEP->getNumOperands(); i != e; ++i) {
Value *OpVal = InsertPHITranslatedSubExpr(GEP->getOperand(i),
CurBB, PredBB, DT, NewInsts);
if (!OpVal) return nullptr;
GEPOps.push_back(OpVal);
}
GetElementPtrInst *Result = GetElementPtrInst::Create(
GEP->getSourceElementType(), GEPOps[0], makeArrayRef(GEPOps).slice(1),
InVal->getName() + ".phi.trans.insert", PredBB->getTerminator());
Result->setDebugLoc(Inst->getDebugLoc());
Result->setIsInBounds(GEP->isInBounds());
NewInsts.push_back(Result);
return Result;
}
#if 0
// FIXME: This code works, but it is unclear that we actually want to insert
// a big chain of computation in order to make a value available in a block.
// This needs to be evaluated carefully to consider its cost trade offs.
// Handle add with a constant RHS.
if (Inst->getOpcode() == Instruction::Add &&
isa<ConstantInt>(Inst->getOperand(1))) {
// PHI translate the LHS.
Value *OpVal = InsertPHITranslatedSubExpr(Inst->getOperand(0),
CurBB, PredBB, DT, NewInsts);
if (OpVal == 0) return 0;
BinaryOperator *Res = BinaryOperator::CreateAdd(OpVal, Inst->getOperand(1),
InVal->getName()+".phi.trans.insert",
PredBB->getTerminator());
Res->setHasNoSignedWrap(cast<BinaryOperator>(Inst)->hasNoSignedWrap());
Res->setHasNoUnsignedWrap(cast<BinaryOperator>(Inst)->hasNoUnsignedWrap());
NewInsts.push_back(Res);
return Res;
}
#endif
return nullptr;
}
GetElementPtrInst *
NaryReassociate::tryReassociateGEPAtIndex(GetElementPtrInst *GEP, unsigned I,
Value *LHS, Value *RHS,
Type *IndexedType) {
// Look for GEP's closest dominator that has the same SCEV as GEP except that
// the I-th index is replaced with LHS.
SmallVector<const SCEV *, 4> IndexExprs;
for (auto Index = GEP->idx_begin(); Index != GEP->idx_end(); ++Index)
IndexExprs.push_back(SE->getSCEV(*Index));
// Replace the I-th index with LHS.
IndexExprs[I] = SE->getSCEV(LHS);
const SCEV *CandidateExpr = SE->getGEPExpr(
GEP->getSourceElementType(), SE->getSCEV(GEP->getPointerOperand()),
IndexExprs, GEP->isInBounds());
auto *Candidate = findClosestMatchingDominator(CandidateExpr, GEP);
if (Candidate == nullptr)
return nullptr;
PointerType *TypeOfCandidate = dyn_cast<PointerType>(Candidate->getType());
// Pretty rare but theoretically possible when a numeric value happens to
// share CandidateExpr.
if (TypeOfCandidate == nullptr)
return nullptr;
// NewGEP = (char *)Candidate + RHS * sizeof(IndexedType)
uint64_t IndexedSize = DL->getTypeAllocSize(IndexedType);
Type *ElementType = TypeOfCandidate->getElementType();
uint64_t ElementSize = DL->getTypeAllocSize(ElementType);
// Another less rare case: because I is not necessarily the last index of the
// GEP, the size of the type at the I-th index (IndexedSize) is not
// necessarily divisible by ElementSize. For example,
//
// #pragma pack(1)
// struct S {
// int a[3];
// int64 b[8];
// };
// #pragma pack()
//
// sizeof(S) = 100 is indivisible by sizeof(int64) = 8.
//
// TODO: bail out on this case for now. We could emit uglygep.
if (IndexedSize % ElementSize != 0)
return nullptr;
// NewGEP = &Candidate[RHS * (sizeof(IndexedType) / sizeof(Candidate[0])));
IRBuilder<> Builder(GEP);
Type *IntPtrTy = DL->getIntPtrType(TypeOfCandidate);
if (RHS->getType() != IntPtrTy)
RHS = Builder.CreateSExtOrTrunc(RHS, IntPtrTy);
if (IndexedSize != ElementSize) {
RHS = Builder.CreateMul(
RHS, ConstantInt::get(IntPtrTy, IndexedSize / ElementSize));
}
GetElementPtrInst *NewGEP =
cast<GetElementPtrInst>(Builder.CreateGEP(Candidate, RHS));
NewGEP->setIsInBounds(GEP->isInBounds());
NewGEP->takeName(GEP);
return NewGEP;
}
//
// Method: runOnModule()
//
// Description:
// Entry point for this LLVM pass.
// Find all GEPs, and simplify them.
//
// Inputs:
// M - A reference to the LLVM module to transform
//
// Outputs:
// M - The transformed LLVM module.
//
// Return value:
// true - The module was modified.
// false - The module was not modified.
//
bool SimplifyGEP::runOnModule(Module& M) {
TD = &getAnalysis<TargetData>();
preprocess(M);
for (Module::iterator F = M.begin(); F != M.end(); ++F){
for (Function::iterator B = F->begin(), FE = F->end(); B != FE; ++B) {
for (BasicBlock::iterator I = B->begin(), BE = B->end(); I != BE; I++) {
if(!(isa<GetElementPtrInst>(I)))
continue;
GetElementPtrInst *GEP = cast<GetElementPtrInst>(I);
Value *PtrOp = GEP->getOperand(0);
Value *StrippedPtr = PtrOp->stripPointerCasts();
// Check if the GEP base pointer is enclosed in a cast
if (StrippedPtr != PtrOp) {
const PointerType *StrippedPtrTy =cast<PointerType>(StrippedPtr->getType());
bool HasZeroPointerIndex = false;
if (ConstantInt *C = dyn_cast<ConstantInt>(GEP->getOperand(1)))
HasZeroPointerIndex = C->isZero();
// Transform: GEP (bitcast [10 x i8]* X to [0 x i8]*), i32 0, ...
// into : GEP [10 x i8]* X, i32 0, ...
//
// Likewise, transform: GEP (bitcast i8* X to [0 x i8]*), i32 0, ...
// into : GEP i8* X, ...
//
// This occurs when the program declares an array extern like "int X[];"
if (HasZeroPointerIndex) {
const PointerType *CPTy = cast<PointerType>(PtrOp->getType());
if (const ArrayType *CATy =
dyn_cast<ArrayType>(CPTy->getElementType())) {
// GEP (bitcast i8* X to [0 x i8]*), i32 0, ... ?
if (CATy->getElementType() == StrippedPtrTy->getElementType()) {
// -> GEP i8* X, ...
SmallVector<Value*, 8> Idx(GEP->idx_begin()+1, GEP->idx_end());
GetElementPtrInst *Res =
GetElementPtrInst::Create(StrippedPtr, Idx, GEP->getName(), GEP);
Res->setIsInBounds(GEP->isInBounds());
GEP->replaceAllUsesWith(Res);
continue;
}
if (const ArrayType *XATy =
dyn_cast<ArrayType>(StrippedPtrTy->getElementType())){
// GEP (bitcast [10 x i8]* X to [0 x i8]*), i32 0, ... ?
if (CATy->getElementType() == XATy->getElementType()) {
// -> GEP [10 x i8]* X, i32 0, ...
// At this point, we know that the cast source type is a pointer
// to an array of the same type as the destination pointer
// array. Because the array type is never stepped over (there
// is a leading zero) we can fold the cast into this GEP.
GEP->setOperand(0, StrippedPtr);
continue;
}
}
}
} else if (GEP->getNumOperands() == 2) {
// Transform things like:
// %t = getelementptr i32* bitcast ([2 x i32]* %str to i32*), i32 %V
// into: %t1 = getelementptr [2 x i32]* %str, i32 0, i32 %V; bitcast
Type *SrcElTy = StrippedPtrTy->getElementType();
Type *ResElTy=cast<PointerType>(PtrOp->getType())->getElementType();
if (TD && SrcElTy->isArrayTy() &&
TD->getTypeAllocSize(cast<ArrayType>(SrcElTy)->getElementType()) ==
TD->getTypeAllocSize(ResElTy)) {
Value *Idx[2];
Idx[0] = Constant::getNullValue(Type::getInt32Ty(GEP->getContext()));
Idx[1] = GEP->getOperand(1);
Value *NewGEP = GetElementPtrInst::Create(StrippedPtr, Idx,
GEP->getName(), GEP);
// V and GEP are both pointer types --> BitCast
GEP->replaceAllUsesWith(new BitCastInst(NewGEP, GEP->getType(), GEP->getName(), GEP));
continue;
}
// Transform things like:
// getelementptr i8* bitcast ([100 x double]* X to i8*), i32 %tmp
// (where tmp = 8*tmp2) into:
// getelementptr [100 x double]* %arr, i32 0, i32 %tmp2; bitcast
if (TD && SrcElTy->isArrayTy() && ResElTy->isIntegerTy(8)) {
uint64_t ArrayEltSize =
TD->getTypeAllocSize(cast<ArrayType>(SrcElTy)->getElementType());
// Check to see if "tmp" is a scale by a multiple of ArrayEltSize. We
// allow either a mul, shift, or constant here.
Value *NewIdx = 0;
ConstantInt *Scale = 0;
if (ArrayEltSize == 1) {
NewIdx = GEP->getOperand(1);
Scale = ConstantInt::get(cast<IntegerType>(NewIdx->getType()), 1);
} else if (ConstantInt *CI = dyn_cast<ConstantInt>(GEP->getOperand(1))) {
NewIdx = ConstantInt::get(CI->getType(), 1);
Scale = CI;
} else if (Instruction *Inst =dyn_cast<Instruction>(GEP->getOperand(1))){
if (Inst->getOpcode() == Instruction::Shl &&
isa<ConstantInt>(Inst->getOperand(1))) {
ConstantInt *ShAmt = cast<ConstantInt>(Inst->getOperand(1));
uint32_t ShAmtVal = ShAmt->getLimitedValue(64);
Scale = ConstantInt::get(cast<IntegerType>(Inst->getType()),
1ULL << ShAmtVal);
NewIdx = Inst->getOperand(0);
} else if (Inst->getOpcode() == Instruction::Mul &&
isa<ConstantInt>(Inst->getOperand(1))) {
Scale = cast<ConstantInt>(Inst->getOperand(1));
NewIdx = Inst->getOperand(0);
}
}
// If the index will be to exactly the right offset with the scale taken
// out, perform the transformation. Note, we don't know whether Scale is
// signed or not. We'll use unsigned version of division/modulo
// operation after making sure Scale doesn't have the sign bit set.
if (ArrayEltSize && Scale && Scale->getSExtValue() >= 0LL &&
Scale->getZExtValue() % ArrayEltSize == 0) {
Scale = ConstantInt::get(Scale->getType(),
Scale->getZExtValue() / ArrayEltSize);
if (Scale->getZExtValue() != 1) {
Constant *C = ConstantExpr::getIntegerCast(Scale, NewIdx->getType(),
false /*ZExt*/);
NewIdx = BinaryOperator::Create(BinaryOperator::Mul, NewIdx, C, "idxscale");
}
// Insert the new GEP instruction.
Value *Idx[2];
Idx[0] = Constant::getNullValue(Type::getInt32Ty(GEP->getContext()));
Idx[1] = NewIdx;
Value *NewGEP = GetElementPtrInst::Create(StrippedPtr, Idx,
GEP->getName(), GEP);
GEP->replaceAllUsesWith(new BitCastInst(NewGEP, GEP->getType(), GEP->getName(), GEP));
continue;
}
}
}
}
}
}
}
return true;
}
bool PPCLoopPreIncPrep::runOnLoop(Loop *L) {
bool MadeChange = false;
// Only prep. the inner-most loop
if (!L->empty())
return MadeChange;
DEBUG(dbgs() << "PIP: Examining: " << *L << "\n");
BasicBlock *Header = L->getHeader();
const PPCSubtarget *ST =
TM ? TM->getSubtargetImpl(*Header->getParent()) : nullptr;
unsigned HeaderLoopPredCount =
std::distance(pred_begin(Header), pred_end(Header));
// Collect buckets of comparable addresses used by loads and stores.
SmallVector<Bucket, 16> Buckets;
for (Loop::block_iterator I = L->block_begin(), IE = L->block_end();
I != IE; ++I) {
for (BasicBlock::iterator J = (*I)->begin(), JE = (*I)->end();
J != JE; ++J) {
Value *PtrValue;
Instruction *MemI;
if (LoadInst *LMemI = dyn_cast<LoadInst>(J)) {
MemI = LMemI;
PtrValue = LMemI->getPointerOperand();
} else if (StoreInst *SMemI = dyn_cast<StoreInst>(J)) {
MemI = SMemI;
PtrValue = SMemI->getPointerOperand();
} else if (IntrinsicInst *IMemI = dyn_cast<IntrinsicInst>(J)) {
if (IMemI->getIntrinsicID() == Intrinsic::prefetch) {
MemI = IMemI;
PtrValue = IMemI->getArgOperand(0);
} else continue;
} else continue;
unsigned PtrAddrSpace = PtrValue->getType()->getPointerAddressSpace();
if (PtrAddrSpace)
continue;
// There are no update forms for Altivec vector load/stores.
if (ST && ST->hasAltivec() &&
PtrValue->getType()->getPointerElementType()->isVectorTy())
continue;
if (L->isLoopInvariant(PtrValue))
continue;
const SCEV *LSCEV = SE->getSCEVAtScope(PtrValue, L);
if (const SCEVAddRecExpr *LARSCEV = dyn_cast<SCEVAddRecExpr>(LSCEV)) {
if (LARSCEV->getLoop() != L)
continue;
} else {
continue;
}
bool FoundBucket = false;
for (auto &B : Buckets) {
const SCEV *Diff = SE->getMinusSCEV(LSCEV, B.BaseSCEV);
if (const auto *CDiff = dyn_cast<SCEVConstant>(Diff)) {
B.Elements.push_back(BucketElement(CDiff, MemI));
FoundBucket = true;
break;
}
}
if (!FoundBucket) {
if (Buckets.size() == MaxVars)
return MadeChange;
Buckets.push_back(Bucket(LSCEV, MemI));
}
}
}
if (Buckets.empty())
return MadeChange;
BasicBlock *LoopPredecessor = L->getLoopPredecessor();
// If there is no loop predecessor, or the loop predecessor's terminator
// returns a value (which might contribute to determining the loop's
// iteration space), insert a new preheader for the loop.
if (!LoopPredecessor ||
!LoopPredecessor->getTerminator()->getType()->isVoidTy()) {
LoopPredecessor = InsertPreheaderForLoop(L, DT, LI, PreserveLCSSA);
if (LoopPredecessor)
MadeChange = true;
}
if (!LoopPredecessor)
return MadeChange;
DEBUG(dbgs() << "PIP: Found " << Buckets.size() << " buckets\n");
SmallSet<BasicBlock *, 16> BBChanged;
for (unsigned i = 0, e = Buckets.size(); i != e; ++i) {
// The base address of each bucket is transformed into a phi and the others
// are rewritten as offsets of that variable.
// We have a choice now of which instruction's memory operand we use as the
// base for the generated PHI. Always picking the first instruction in each
// bucket does not work well, specifically because that instruction might
// be a prefetch (and there are no pre-increment dcbt variants). Otherwise,
// the choice is somewhat arbitrary, because the backend will happily
// generate direct offsets from both the pre-incremented and
// post-incremented pointer values. Thus, we'll pick the first non-prefetch
// instruction in each bucket, and adjust the recurrence and other offsets
// accordingly.
for (int j = 0, je = Buckets[i].Elements.size(); j != je; ++j) {
if (auto *II = dyn_cast<IntrinsicInst>(Buckets[i].Elements[j].Instr))
if (II->getIntrinsicID() == Intrinsic::prefetch)
continue;
// If we'd otherwise pick the first element anyway, there's nothing to do.
if (j == 0)
break;
// If our chosen element has no offset from the base pointer, there's
// nothing to do.
if (!Buckets[i].Elements[j].Offset ||
Buckets[i].Elements[j].Offset->isZero())
break;
const SCEV *Offset = Buckets[i].Elements[j].Offset;
Buckets[i].BaseSCEV = SE->getAddExpr(Buckets[i].BaseSCEV, Offset);
for (auto &E : Buckets[i].Elements) {
if (E.Offset)
E.Offset = cast<SCEVConstant>(SE->getMinusSCEV(E.Offset, Offset));
else
E.Offset = cast<SCEVConstant>(SE->getNegativeSCEV(Offset));
}
std::swap(Buckets[i].Elements[j], Buckets[i].Elements[0]);
break;
}
const SCEVAddRecExpr *BasePtrSCEV =
cast<SCEVAddRecExpr>(Buckets[i].BaseSCEV);
if (!BasePtrSCEV->isAffine())
continue;
DEBUG(dbgs() << "PIP: Transforming: " << *BasePtrSCEV << "\n");
assert(BasePtrSCEV->getLoop() == L &&
"AddRec for the wrong loop?");
// The instruction corresponding to the Bucket's BaseSCEV must be the first
// in the vector of elements.
Instruction *MemI = Buckets[i].Elements.begin()->Instr;
Value *BasePtr = GetPointerOperand(MemI);
assert(BasePtr && "No pointer operand");
Type *I8Ty = Type::getInt8Ty(MemI->getParent()->getContext());
Type *I8PtrTy = Type::getInt8PtrTy(MemI->getParent()->getContext(),
BasePtr->getType()->getPointerAddressSpace());
const SCEV *BasePtrStartSCEV = BasePtrSCEV->getStart();
if (!SE->isLoopInvariant(BasePtrStartSCEV, L))
continue;
const SCEVConstant *BasePtrIncSCEV =
dyn_cast<SCEVConstant>(BasePtrSCEV->getStepRecurrence(*SE));
if (!BasePtrIncSCEV)
continue;
BasePtrStartSCEV = SE->getMinusSCEV(BasePtrStartSCEV, BasePtrIncSCEV);
if (!isSafeToExpand(BasePtrStartSCEV, *SE))
continue;
DEBUG(dbgs() << "PIP: New start is: " << *BasePtrStartSCEV << "\n");
PHINode *NewPHI = PHINode::Create(I8PtrTy, HeaderLoopPredCount,
MemI->hasName() ? MemI->getName() + ".phi" : "",
Header->getFirstNonPHI());
SCEVExpander SCEVE(*SE, Header->getModule()->getDataLayout(), "pistart");
Value *BasePtrStart = SCEVE.expandCodeFor(BasePtrStartSCEV, I8PtrTy,
LoopPredecessor->getTerminator());
// Note that LoopPredecessor might occur in the predecessor list multiple
// times, and we need to add it the right number of times.
for (pred_iterator PI = pred_begin(Header), PE = pred_end(Header);
PI != PE; ++PI) {
if (*PI != LoopPredecessor)
continue;
NewPHI->addIncoming(BasePtrStart, LoopPredecessor);
}
Instruction *InsPoint = &*Header->getFirstInsertionPt();
GetElementPtrInst *PtrInc = GetElementPtrInst::Create(
I8Ty, NewPHI, BasePtrIncSCEV->getValue(),
MemI->hasName() ? MemI->getName() + ".inc" : "", InsPoint);
PtrInc->setIsInBounds(IsPtrInBounds(BasePtr));
for (pred_iterator PI = pred_begin(Header), PE = pred_end(Header);
PI != PE; ++PI) {
if (*PI == LoopPredecessor)
continue;
NewPHI->addIncoming(PtrInc, *PI);
}
Instruction *NewBasePtr;
if (PtrInc->getType() != BasePtr->getType())
NewBasePtr = new BitCastInst(PtrInc, BasePtr->getType(),
PtrInc->hasName() ? PtrInc->getName() + ".cast" : "", InsPoint);
else
NewBasePtr = PtrInc;
if (Instruction *IDel = dyn_cast<Instruction>(BasePtr))
BBChanged.insert(IDel->getParent());
BasePtr->replaceAllUsesWith(NewBasePtr);
RecursivelyDeleteTriviallyDeadInstructions(BasePtr);
// Keep track of the replacement pointer values we've inserted so that we
// don't generate more pointer values than necessary.
SmallPtrSet<Value *, 16> NewPtrs;
NewPtrs.insert( NewBasePtr);
for (auto I = std::next(Buckets[i].Elements.begin()),
IE = Buckets[i].Elements.end(); I != IE; ++I) {
Value *Ptr = GetPointerOperand(I->Instr);
assert(Ptr && "No pointer operand");
if (NewPtrs.count(Ptr))
continue;
Instruction *RealNewPtr;
if (!I->Offset || I->Offset->getValue()->isZero()) {
RealNewPtr = NewBasePtr;
} else {
Instruction *PtrIP = dyn_cast<Instruction>(Ptr);
if (PtrIP && isa<Instruction>(NewBasePtr) &&
cast<Instruction>(NewBasePtr)->getParent() == PtrIP->getParent())
PtrIP = nullptr;
else if (isa<PHINode>(PtrIP))
PtrIP = &*PtrIP->getParent()->getFirstInsertionPt();
else if (!PtrIP)
PtrIP = I->Instr;
GetElementPtrInst *NewPtr = GetElementPtrInst::Create(
I8Ty, PtrInc, I->Offset->getValue(),
I->Instr->hasName() ? I->Instr->getName() + ".off" : "", PtrIP);
if (!PtrIP)
NewPtr->insertAfter(cast<Instruction>(PtrInc));
NewPtr->setIsInBounds(IsPtrInBounds(Ptr));
RealNewPtr = NewPtr;
}
if (Instruction *IDel = dyn_cast<Instruction>(Ptr))
BBChanged.insert(IDel->getParent());
Instruction *ReplNewPtr;
if (Ptr->getType() != RealNewPtr->getType()) {
ReplNewPtr = new BitCastInst(RealNewPtr, Ptr->getType(),
Ptr->hasName() ? Ptr->getName() + ".cast" : "");
ReplNewPtr->insertAfter(RealNewPtr);
} else
ReplNewPtr = RealNewPtr;
Ptr->replaceAllUsesWith(ReplNewPtr);
RecursivelyDeleteTriviallyDeadInstructions(Ptr);
NewPtrs.insert(RealNewPtr);
}
MadeChange = true;
}
for (Loop::block_iterator I = L->block_begin(), IE = L->block_end();
I != IE; ++I) {
if (BBChanged.count(*I))
DeleteDeadPHIs(*I);
}
return MadeChange;
}
/// Rebuild a new instruction just like 'I' but with the new operands given.
/// In the event of type mismatch, the type of the operands is correct.
static Value *BuildNew(Instruction *I, ArrayRef<Value*> NewOps) {
// We don't want to use the IRBuilder here because we want the replacement
// instructions to appear next to 'I', not the builder's insertion point.
switch (I->getOpcode()) {
case Instruction::Add:
case Instruction::FAdd:
case Instruction::Sub:
case Instruction::FSub:
case Instruction::Mul:
case Instruction::FMul:
case Instruction::UDiv:
case Instruction::SDiv:
case Instruction::FDiv:
case Instruction::URem:
case Instruction::SRem:
case Instruction::FRem:
case Instruction::Shl:
case Instruction::LShr:
case Instruction::AShr:
case Instruction::And:
case Instruction::Or:
case Instruction::Xor: {
BinaryOperator *BO = cast<BinaryOperator>(I);
assert(NewOps.size() == 2 && "binary operator with #ops != 2");
BinaryOperator *New =
BinaryOperator::Create(cast<BinaryOperator>(I)->getOpcode(),
NewOps[0], NewOps[1], "", BO);
if (isa<OverflowingBinaryOperator>(BO)) {
New->setHasNoUnsignedWrap(BO->hasNoUnsignedWrap());
New->setHasNoSignedWrap(BO->hasNoSignedWrap());
}
if (isa<PossiblyExactOperator>(BO)) {
New->setIsExact(BO->isExact());
}
return New;
}
case Instruction::ICmp:
assert(NewOps.size() == 2 && "icmp with #ops != 2");
return new ICmpInst(I, cast<ICmpInst>(I)->getPredicate(),
NewOps[0], NewOps[1]);
case Instruction::FCmp:
assert(NewOps.size() == 2 && "fcmp with #ops != 2");
return new FCmpInst(I, cast<FCmpInst>(I)->getPredicate(),
NewOps[0], NewOps[1]);
case Instruction::Trunc:
case Instruction::ZExt:
case Instruction::SExt:
case Instruction::FPToUI:
case Instruction::FPToSI:
case Instruction::UIToFP:
case Instruction::SIToFP:
case Instruction::FPTrunc:
case Instruction::FPExt: {
// It's possible that the mask has a different number of elements from
// the original cast. We recompute the destination type to match the mask.
Type *DestTy =
VectorType::get(I->getType()->getScalarType(),
NewOps[0]->getType()->getVectorNumElements());
assert(NewOps.size() == 1 && "cast with #ops != 1");
return CastInst::Create(cast<CastInst>(I)->getOpcode(), NewOps[0], DestTy,
"", I);
}
case Instruction::GetElementPtr: {
Value *Ptr = NewOps[0];
ArrayRef<Value*> Idx = NewOps.slice(1);
GetElementPtrInst *GEP = GetElementPtrInst::Create(Ptr, Idx, "", I);
GEP->setIsInBounds(cast<GetElementPtrInst>(I)->isInBounds());
return GEP;
}
}
llvm_unreachable("failed to rebuild vector instructions");
}
bool PPCLoopPreIncPrep::runOnLoop(Loop *L) {
bool MadeChange = false;
if (!DL)
return MadeChange;
// Only prep. the inner-most loop
if (!L->empty())
return MadeChange;
BasicBlock *Header = L->getHeader();
const PPCSubtarget *ST =
TM ? TM->getSubtargetImpl(*Header->getParent()) : nullptr;
unsigned HeaderLoopPredCount = 0;
for (pred_iterator PI = pred_begin(Header), PE = pred_end(Header);
PI != PE; ++PI) {
++HeaderLoopPredCount;
}
// Collect buckets of comparable addresses used by loads and stores.
typedef std::multimap<const SCEV *, Instruction *, SCEVLess> Bucket;
SmallVector<Bucket, 16> Buckets;
for (Loop::block_iterator I = L->block_begin(), IE = L->block_end();
I != IE; ++I) {
for (BasicBlock::iterator J = (*I)->begin(), JE = (*I)->end();
J != JE; ++J) {
Value *PtrValue;
Instruction *MemI;
if (LoadInst *LMemI = dyn_cast<LoadInst>(J)) {
MemI = LMemI;
PtrValue = LMemI->getPointerOperand();
} else if (StoreInst *SMemI = dyn_cast<StoreInst>(J)) {
MemI = SMemI;
PtrValue = SMemI->getPointerOperand();
} else if (IntrinsicInst *IMemI = dyn_cast<IntrinsicInst>(J)) {
if (IMemI->getIntrinsicID() == Intrinsic::prefetch) {
MemI = IMemI;
PtrValue = IMemI->getArgOperand(0);
} else continue;
} else continue;
unsigned PtrAddrSpace = PtrValue->getType()->getPointerAddressSpace();
if (PtrAddrSpace)
continue;
// There are no update forms for Altivec vector load/stores.
if (ST && ST->hasAltivec() &&
PtrValue->getType()->getPointerElementType()->isVectorTy())
continue;
if (L->isLoopInvariant(PtrValue))
continue;
const SCEV *LSCEV = SE->getSCEV(PtrValue);
if (!isa<SCEVAddRecExpr>(LSCEV))
continue;
bool FoundBucket = false;
for (unsigned i = 0, e = Buckets.size(); i != e; ++i)
for (Bucket::iterator K = Buckets[i].begin(), KE = Buckets[i].end();
K != KE; ++K) {
const SCEV *Diff = SE->getMinusSCEV(K->first, LSCEV);
if (isa<SCEVConstant>(Diff)) {
Buckets[i].insert(std::make_pair(LSCEV, MemI));
FoundBucket = true;
break;
}
}
if (!FoundBucket) {
Buckets.push_back(Bucket(SCEVLess(SE)));
Buckets[Buckets.size()-1].insert(std::make_pair(LSCEV, MemI));
}
}
}
if (Buckets.empty() || Buckets.size() > MaxVars)
return MadeChange;
BasicBlock *LoopPredecessor = L->getLoopPredecessor();
// If there is no loop predecessor, or the loop predecessor's terminator
// returns a value (which might contribute to determining the loop's
// iteration space), insert a new preheader for the loop.
if (!LoopPredecessor ||
!LoopPredecessor->getTerminator()->getType()->isVoidTy())
LoopPredecessor = InsertPreheaderForLoop(L, this);
if (!LoopPredecessor)
return MadeChange;
SmallSet<BasicBlock *, 16> BBChanged;
for (unsigned i = 0, e = Buckets.size(); i != e; ++i) {
// The base address of each bucket is transformed into a phi and the others
// are rewritten as offsets of that variable.
const SCEVAddRecExpr *BasePtrSCEV =
cast<SCEVAddRecExpr>(Buckets[i].begin()->first);
if (!BasePtrSCEV->isAffine())
continue;
Instruction *MemI = Buckets[i].begin()->second;
Value *BasePtr = GetPointerOperand(MemI);
assert(BasePtr && "No pointer operand");
Type *I8PtrTy = Type::getInt8PtrTy(MemI->getParent()->getContext(),
BasePtr->getType()->getPointerAddressSpace());
const SCEV *BasePtrStartSCEV = BasePtrSCEV->getStart();
if (!SE->isLoopInvariant(BasePtrStartSCEV, L))
continue;
const SCEVConstant *BasePtrIncSCEV =
dyn_cast<SCEVConstant>(BasePtrSCEV->getStepRecurrence(*SE));
if (!BasePtrIncSCEV)
continue;
BasePtrStartSCEV = SE->getMinusSCEV(BasePtrStartSCEV, BasePtrIncSCEV);
if (!isSafeToExpand(BasePtrStartSCEV, *SE))
continue;
PHINode *NewPHI = PHINode::Create(I8PtrTy, HeaderLoopPredCount,
MemI->hasName() ? MemI->getName() + ".phi" : "",
Header->getFirstNonPHI());
SCEVExpander SCEVE(*SE, "pistart");
Value *BasePtrStart = SCEVE.expandCodeFor(BasePtrStartSCEV, I8PtrTy,
LoopPredecessor->getTerminator());
// Note that LoopPredecessor might occur in the predecessor list multiple
// times, and we need to add it the right number of times.
for (pred_iterator PI = pred_begin(Header), PE = pred_end(Header);
PI != PE; ++PI) {
if (*PI != LoopPredecessor)
continue;
NewPHI->addIncoming(BasePtrStart, LoopPredecessor);
}
Instruction *InsPoint = Header->getFirstInsertionPt();
GetElementPtrInst *PtrInc =
GetElementPtrInst::Create(NewPHI, BasePtrIncSCEV->getValue(),
MemI->hasName() ? MemI->getName() + ".inc" : "", InsPoint);
PtrInc->setIsInBounds(IsPtrInBounds(BasePtr));
for (pred_iterator PI = pred_begin(Header), PE = pred_end(Header);
PI != PE; ++PI) {
if (*PI == LoopPredecessor)
continue;
NewPHI->addIncoming(PtrInc, *PI);
}
Instruction *NewBasePtr;
if (PtrInc->getType() != BasePtr->getType())
NewBasePtr = new BitCastInst(PtrInc, BasePtr->getType(),
PtrInc->hasName() ? PtrInc->getName() + ".cast" : "", InsPoint);
else
NewBasePtr = PtrInc;
if (Instruction *IDel = dyn_cast<Instruction>(BasePtr))
BBChanged.insert(IDel->getParent());
BasePtr->replaceAllUsesWith(NewBasePtr);
RecursivelyDeleteTriviallyDeadInstructions(BasePtr);
Value *LastNewPtr = NewBasePtr;
for (Bucket::iterator I = std::next(Buckets[i].begin()),
IE = Buckets[i].end(); I != IE; ++I) {
Value *Ptr = GetPointerOperand(I->second);
assert(Ptr && "No pointer operand");
if (Ptr == LastNewPtr)
continue;
Instruction *RealNewPtr;
const SCEVConstant *Diff =
cast<SCEVConstant>(SE->getMinusSCEV(I->first, BasePtrSCEV));
if (Diff->isZero()) {
RealNewPtr = NewBasePtr;
} else {
Instruction *PtrIP = dyn_cast<Instruction>(Ptr);
if (PtrIP && isa<Instruction>(NewBasePtr) &&
cast<Instruction>(NewBasePtr)->getParent() == PtrIP->getParent())
PtrIP = 0;
else if (isa<PHINode>(PtrIP))
PtrIP = PtrIP->getParent()->getFirstInsertionPt();
else if (!PtrIP)
PtrIP = I->second;
GetElementPtrInst *NewPtr =
GetElementPtrInst::Create(PtrInc, Diff->getValue(),
I->second->hasName() ? I->second->getName() + ".off" : "", PtrIP);
if (!PtrIP)
NewPtr->insertAfter(cast<Instruction>(PtrInc));
NewPtr->setIsInBounds(IsPtrInBounds(Ptr));
RealNewPtr = NewPtr;
}
if (Instruction *IDel = dyn_cast<Instruction>(Ptr))
BBChanged.insert(IDel->getParent());
Instruction *ReplNewPtr;
if (Ptr->getType() != RealNewPtr->getType()) {
ReplNewPtr = new BitCastInst(RealNewPtr, Ptr->getType(),
Ptr->hasName() ? Ptr->getName() + ".cast" : "");
ReplNewPtr->insertAfter(RealNewPtr);
} else
ReplNewPtr = RealNewPtr;
Ptr->replaceAllUsesWith(ReplNewPtr);
RecursivelyDeleteTriviallyDeadInstructions(Ptr);
LastNewPtr = RealNewPtr;
}
MadeChange = true;
}
for (Loop::block_iterator I = L->block_begin(), IE = L->block_end();
I != IE; ++I) {
if (BBChanged.count(*I))
DeleteDeadPHIs(*I);
}
return MadeChange;
}
void FaultInjectionPass::insertInjectionFuncCall(
std::map<Instruction*, std::list< int >* > *inst_regs_map, Module &M) {
for (std::map<Instruction*, std::list< int >* >::iterator inst_reg_it =
inst_regs_map->begin(); inst_reg_it != inst_regs_map->end();
++inst_reg_it) {
Instruction *fi_inst = inst_reg_it->first;
std::list<int> *fi_reg_pos_list = inst_reg_it->second;
unsigned reg_index = 0;
unsigned total_reg_num = fi_reg_pos_list->size();
for (std::list<int>::iterator reg_pos_it = fi_reg_pos_list->begin();
reg_pos_it != fi_reg_pos_list->end(); ++reg_pos_it, ++reg_index) {
if(isa<GetElementPtrInst>(fi_inst)){
GetElementPtrInst* gepi = dyn_cast<GetElementPtrInst>(fi_inst);
gepi->setIsInBounds(false);
}
if(isa<CallInst>(fi_inst)){
CallInst* ci = dyn_cast<CallInst>(fi_inst);
ci->setTailCall(false);
}
Value* fi_reg = NULL;
if(*reg_pos_it == DST_REG_POS) fi_reg = fi_inst;
else fi_reg = fi_inst->getOperand(*reg_pos_it);
//if(isa<Constant>(fi_reg)) continue;
Type *returntype = fi_reg->getType();
LLVMContext &context = M.getContext();
Type *i64type = Type::getInt64Ty(context);
Type *i32type = Type::getInt32Ty(context);
// function declaration
std::vector<Type*> paramtypes(7);
paramtypes[0] = i64type; // llfi index
paramtypes[1] = returntype; // the instruction to be injected
paramtypes[2] = i32type; // opcode
paramtypes[3] = i32type; // current fi reg index
paramtypes[4] = i32type; // total fi reg number
//======== Add reg_pos QINING @DEC 23rd ==========
paramtypes[5] = i32type;
//================================================
//======== Add opcode_str QINING @MAR 11th========
paramtypes[6] = PointerType::get(Type::getInt8Ty(context), 0);
//================================================
//LLVM 3.3 Upgrade
ArrayRef<Type*> paramtypes_array_ref(paramtypes);
FunctionType* injectfunctype = FunctionType::get(returntype, paramtypes_array_ref, false);
std::string funcname = getFIFuncNameforType(returntype);
Constant *injectfunc = M.getOrInsertFunction(funcname, injectfunctype);
// argument preparation for calling function
// since the source register is another way of simulating fault
// injection into "the instruction", use instruction's index instead
Value *indexval = ConstantInt::get(i64type, getLLFIIndexofInst(fi_inst));
std::vector<Value*> args(7);
args[0] = indexval; //llfi index
args[1] = fi_reg; // target register
args[2] = ConstantInt::get(i32type, fi_inst->getOpcode()); // opcode in i32
args[3] = ConstantInt::get(i32type, reg_index); // reg_index not reg_pos
args[4] = ConstantInt::get(i32type, total_reg_num); // total_reg_num
//======== Add reg_pos QINING @DEC 23rd ==========
args[5] = ConstantInt::get(i32type, *reg_pos_it+1); // dstreg->0, operand0->1, operand1->2 ...
//================================================
//======== Add opcode_str QINING @MAR 11th========
std::string opcode_str = fi_inst->getOpcodeName();
GlobalVariable* opcode_str_gv = findOrCreateGlobalNameString(M, opcode_str);
std::vector<Constant*> indices_for_gep(2);
indices_for_gep[0] = ConstantInt::get(Type::getInt32Ty(context),0);
indices_for_gep[1] = ConstantInt::get(Type::getInt32Ty(context),0);
ArrayRef<Constant*> indices_for_gep_array_ref(indices_for_gep);
Constant* gep_expr = ConstantExpr::getGetElementPtr(opcode_str_gv, indices_for_gep_array_ref, true);
args[6] = gep_expr; // opcode in string
//================================================
// LLVM 3.3 Upgrade
ArrayRef<Value*> args_array_ref(args);
Instruction *insertptr = getInsertPtrforRegsofInst(fi_reg, fi_inst);
Instruction* ficall = CallInst::Create(
injectfunc, args_array_ref, "fi", insertptr);
setInjectFaultInst(fi_reg, fi_inst, ficall); // sets the instruction metadata
// redirect the data dependencies
if (fi_reg == fi_inst) {
// inject into destination
std::list<User*> inst_uses;
for (Value::use_iterator use_it = fi_inst->use_begin();
use_it != fi_inst->use_end(); ++use_it) {
User *user = *use_it;
if (user != ficall) {
inst_uses.push_back(user);
}
}
for (std::list<User*>::iterator use_it = inst_uses.begin();
use_it != inst_uses.end(); ++use_it) {
User *user = *use_it;
user->replaceUsesOfWith(fi_inst, ficall);
// update the selected inst pool
/*if (Instruction *use_inst = dyn_cast<Instruction>(user)) {
if (inst_regs_map->find(use_inst) != inst_regs_map->end()) {
std::list<int> *reg_pos_list = (*inst_regs_map)[use_inst];
for (std::list<int>::iterator reg_pos_it = reg_pos_list->begin();
reg_pos_it != reg_pos_list->end(); ++reg_pos_it) {
if (use_inst->getOperand(*reg_pos_it) == fi_inst) {
use_inst->setOperand(*reg_pos_it, ficall);
}
}
}
}*/
}
} else {
// inject into source
//fi_inst->replaceUsesOfWith(fi_reg, ficall);
fi_inst->setOperand(*reg_pos_it, ficall);
}
}
}
}
/// InsertPHITranslatedPointer - Insert a computation of the PHI translated
/// version of 'V' for the edge PredBB->CurBB into the end of the PredBB
/// block. All newly created instructions are added to the NewInsts list.
/// This returns null on failure.
///
Value *PHITransAddr::
InsertPHITranslatedSubExpr(Value *InVal, BasicBlock *CurBB,
BasicBlock *PredBB, const DominatorTree &DT,
SmallVectorImpl<Instruction*> &NewInsts) {
// See if we have a version of this value already available and dominating
// PredBB. If so, there is no need to insert a new instance of it.
PHITransAddr Tmp(InVal, TD);
if (!Tmp.PHITranslateValue(CurBB, PredBB, &DT))
return Tmp.getAddr();
// If we don't have an available version of this value, it must be an
// instruction.
Instruction *Inst = cast<Instruction>(InVal);
// Handle bitcast of PHI translatable value.
if (BitCastInst *BC = dyn_cast<BitCastInst>(Inst)) {
Value *OpVal = InsertPHITranslatedSubExpr(BC->getOperand(0),
CurBB, PredBB, DT, NewInsts);
if (OpVal == 0) return 0;
// Otherwise insert a bitcast at the end of PredBB.
BitCastInst *New = new BitCastInst(OpVal, InVal->getType(),
InVal->getName()+".phi.trans.insert",
PredBB->getTerminator());
NewInsts.push_back(New);
return New;
}
// Handle getelementptr with at least one PHI operand.
if (GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(Inst)) {
SmallVector<Value*, 8> GEPOps;
BasicBlock *CurBB = GEP->getParent();
for (unsigned i = 0, e = GEP->getNumOperands(); i != e; ++i) {
Value *OpVal = InsertPHITranslatedSubExpr(GEP->getOperand(i),
CurBB, PredBB, DT, NewInsts);
if (OpVal == 0) return 0;
GEPOps.push_back(OpVal);
}
GetElementPtrInst *Result =
GetElementPtrInst::Create(GEPOps[0], GEPOps.begin()+1, GEPOps.end(),
InVal->getName()+".phi.trans.insert",
PredBB->getTerminator());
Result->setIsInBounds(GEP->isInBounds());
NewInsts.push_back(Result);
return Result;
}
#if 0
// FIXME: This code works, but it is unclear that we actually want to insert
// a big chain of computation in order to make a value available in a block.
// This needs to be evaluated carefully to consider its cost trade offs.
// Handle add with a constant RHS.
if (Inst->getOpcode() == Instruction::Add &&
isa<ConstantInt>(Inst->getOperand(1))) {
// PHI translate the LHS.
Value *OpVal = InsertPHITranslatedSubExpr(Inst->getOperand(0),
CurBB, PredBB, DT, NewInsts);
if (OpVal == 0) return 0;
BinaryOperator *Res = BinaryOperator::CreateAdd(OpVal, Inst->getOperand(1),
InVal->getName()+".phi.trans.insert",
PredBB->getTerminator());
Res->setHasNoSignedWrap(cast<BinaryOperator>(Inst)->hasNoSignedWrap());
Res->setHasNoUnsignedWrap(cast<BinaryOperator>(Inst)->hasNoUnsignedWrap());
NewInsts.push_back(Res);
return Res;
}
#endif
return 0;
}
GetElementPtrInst *
NaryReassociatePass::tryReassociateGEPAtIndex(GetElementPtrInst *GEP,
unsigned I, Value *LHS,
Value *RHS, Type *IndexedType) {
// Look for GEP's closest dominator that has the same SCEV as GEP except that
// the I-th index is replaced with LHS.
SmallVector<const SCEV *, 4> IndexExprs;
for (auto Index = GEP->idx_begin(); Index != GEP->idx_end(); ++Index)
IndexExprs.push_back(SE->getSCEV(*Index));
// Replace the I-th index with LHS.
IndexExprs[I] = SE->getSCEV(LHS);
if (isKnownNonNegative(LHS, *DL, 0, AC, GEP, DT) &&
DL->getTypeSizeInBits(LHS->getType()) <
DL->getTypeSizeInBits(GEP->getOperand(I)->getType())) {
// Zero-extend LHS if it is non-negative. InstCombine canonicalizes sext to
// zext if the source operand is proved non-negative. We should do that
// consistently so that CandidateExpr more likely appears before. See
// @reassociate_gep_assume for an example of this canonicalization.
IndexExprs[I] =
SE->getZeroExtendExpr(IndexExprs[I], GEP->getOperand(I)->getType());
}
const SCEV *CandidateExpr = SE->getGEPExpr(cast<GEPOperator>(GEP),
IndexExprs);
Value *Candidate = findClosestMatchingDominator(CandidateExpr, GEP);
if (Candidate == nullptr)
return nullptr;
IRBuilder<> Builder(GEP);
// Candidate does not necessarily have the same pointer type as GEP. Use
// bitcast or pointer cast to make sure they have the same type, so that the
// later RAUW doesn't complain.
Candidate = Builder.CreateBitOrPointerCast(Candidate, GEP->getType());
assert(Candidate->getType() == GEP->getType());
// NewGEP = (char *)Candidate + RHS * sizeof(IndexedType)
uint64_t IndexedSize = DL->getTypeAllocSize(IndexedType);
Type *ElementType = GEP->getResultElementType();
uint64_t ElementSize = DL->getTypeAllocSize(ElementType);
// Another less rare case: because I is not necessarily the last index of the
// GEP, the size of the type at the I-th index (IndexedSize) is not
// necessarily divisible by ElementSize. For example,
//
// #pragma pack(1)
// struct S {
// int a[3];
// int64 b[8];
// };
// #pragma pack()
//
// sizeof(S) = 100 is indivisible by sizeof(int64) = 8.
//
// TODO: bail out on this case for now. We could emit uglygep.
if (IndexedSize % ElementSize != 0)
return nullptr;
// NewGEP = &Candidate[RHS * (sizeof(IndexedType) / sizeof(Candidate[0])));
Type *IntPtrTy = DL->getIntPtrType(GEP->getType());
if (RHS->getType() != IntPtrTy)
RHS = Builder.CreateSExtOrTrunc(RHS, IntPtrTy);
if (IndexedSize != ElementSize) {
RHS = Builder.CreateMul(
RHS, ConstantInt::get(IntPtrTy, IndexedSize / ElementSize));
}
GetElementPtrInst *NewGEP =
cast<GetElementPtrInst>(Builder.CreateGEP(Candidate, RHS));
NewGEP->setIsInBounds(GEP->isInBounds());
NewGEP->takeName(GEP);
return NewGEP;
}
