#include "llvm/IR/ValueHandle.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Debug.h"
+#include "llvm/Support/raw_ostream.h"
#include "llvm/Transforms/Scalar.h"
#include "llvm/Transforms/Utils/Local.h"
#include <algorithm>
STATISTIC(NumReassoc , "Number of reassociations");
Value *InstCombiner::EmitGEPOffset(User *GEP) {
- return llvm::EmitGEPOffset(Builder, *getDataLayout(), GEP);
+ return llvm::EmitGEPOffset(Builder, DL, GEP);
}
/// ShouldChangeType - Return true if it is desirable to convert a computation
bool InstCombiner::ShouldChangeType(Type *From, Type *To) const {
assert(From->isIntegerTy() && To->isIntegerTy());
- // If we don't have DL, we don't know if the source/dest are legal.
- if (!DL) return false;
-
unsigned FromWidth = From->getPrimitiveSizeInBits();
unsigned ToWidth = To->getPrimitiveSizeInBits();
- bool FromLegal = DL->isLegalInteger(FromWidth);
- bool ToLegal = DL->isLegalInteger(ToWidth);
+ bool FromLegal = DL.isLegalInteger(FromWidth);
+ bool ToLegal = DL.isLegalInteger(ToWidth);
// If this is a legal integer from type, and the result would be an illegal
// type, don't do the transformation.
/// This tries to simplify binary operations by factorizing out common terms
/// (e. g. "(A*B)+(A*C)" -> "A*(B+C)").
static Value *tryFactorization(InstCombiner::BuilderTy *Builder,
- const DataLayout *DL, BinaryOperator &I,
+ const DataLayout &DL, BinaryOperator &I,
Instruction::BinaryOps InnerOpcode, Value *A,
Value *B, Value *C, Value *D) {
if (!A || !C || !B || !D)
return nullptr;
+ Value *V = nullptr;
Value *SimplifiedInst = nullptr;
Value *LHS = I.getOperand(0), *RHS = I.getOperand(1);
Instruction::BinaryOps TopLevelOpcode = I.getOpcode();
std::swap(C, D);
// Consider forming "A op' (B op D)".
// If "B op D" simplifies then it can be formed with no cost.
- Value *V = SimplifyBinOp(TopLevelOpcode, B, D, DL);
+ V = SimplifyBinOp(TopLevelOpcode, B, D, DL);
// If "B op D" doesn't simplify then only go on if both of the existing
// operations "A op' B" and "C op' D" will be zapped as no longer used.
if (!V && LHS->hasOneUse() && RHS->hasOneUse())
std::swap(C, D);
// Consider forming "(A op C) op' B".
// If "A op C" simplifies then it can be formed with no cost.
- Value *V = SimplifyBinOp(TopLevelOpcode, A, C, DL);
+ V = SimplifyBinOp(TopLevelOpcode, A, C, DL);
// If "A op C" doesn't simplify then only go on if both of the existing
// operations "A op' B" and "C op' D" will be zapped as no longer used.
if (BinaryOperator *Op1 = dyn_cast<BinaryOperator>(RHS))
if (isa<OverflowingBinaryOperator>(Op1))
HasNSW &= Op1->hasNoSignedWrap();
- BO->setHasNoSignedWrap(HasNSW);
+
+ // We can propogate 'nsw' if we know that
+ // %Y = mul nsw i16 %X, C
+ // %Z = add nsw i16 %Y, %X
+ // =>
+ // %Z = mul nsw i16 %X, C+1
+ //
+ // iff C+1 isn't INT_MIN
+ const APInt *CInt;
+ if (TopLevelOpcode == Instruction::Add &&
+ InnerOpcode == Instruction::Mul)
+ if (match(V, m_APInt(CInt)) && !CInt->isMinSignedValue())
+ BO->setHasNoSignedWrap(HasNSW);
}
}
}
}
}
+ // (op (select (a, c, b)), (select (a, d, b))) -> (select (a, (op c, d), 0))
+ // (op (select (a, b, c)), (select (a, b, d))) -> (select (a, 0, (op c, d)))
+ if (auto *SI0 = dyn_cast<SelectInst>(LHS)) {
+ if (auto *SI1 = dyn_cast<SelectInst>(RHS)) {
+ if (SI0->getCondition() == SI1->getCondition()) {
+ Value *SI = nullptr;
+ if (Value *V = SimplifyBinOp(TopLevelOpcode, SI0->getFalseValue(),
+ SI1->getFalseValue(), DL, TLI, DT, AC))
+ SI = Builder->CreateSelect(SI0->getCondition(),
+ Builder->CreateBinOp(TopLevelOpcode,
+ SI0->getTrueValue(),
+ SI1->getTrueValue()),
+ V);
+ if (Value *V = SimplifyBinOp(TopLevelOpcode, SI0->getTrueValue(),
+ SI1->getTrueValue(), DL, TLI, DT, AC))
+ SI = Builder->CreateSelect(
+ SI0->getCondition(), V,
+ Builder->CreateBinOp(TopLevelOpcode, SI0->getFalseValue(),
+ SI1->getFalseValue()));
+ if (SI) {
+ SI->takeName(&I);
+ return SI;
+ }
+ }
+ }
+ }
+
return nullptr;
}
return nullptr;
}
+ // Test if a CmpInst instruction is used exclusively by a select as
+ // part of a minimum or maximum operation. If so, refrain from doing
+ // any other folding. This helps out other analyses which understand
+ // non-obfuscated minimum and maximum idioms, such as ScalarEvolution
+ // and CodeGen. And in this case, at least one of the comparison
+ // operands has at least one user besides the compare (the select),
+ // which would often largely negate the benefit of folding anyway.
+ if (auto *CI = dyn_cast<CmpInst>(SI->getCondition())) {
+ if (CI->hasOneUse()) {
+ Value *Op0 = CI->getOperand(0), *Op1 = CI->getOperand(1);
+ if ((SI->getOperand(1) == Op0 && SI->getOperand(2) == Op1) ||
+ (SI->getOperand(2) == Op0 && SI->getOperand(1) == Op1))
+ return nullptr;
+ }
+ }
+
Value *SelectTrueVal = FoldOperationIntoSelectOperand(Op, TV, this);
Value *SelectFalseVal = FoldOperationIntoSelectOperand(Op, FV, this);
return nullptr;
}
-
/// FoldOpIntoPhi - Given a binary operator, cast instruction, or select which
/// has a PHI node as operand #0, see if we can fold the instruction into the
/// PHI (which is only possible if all operands to the PHI are constants).
/// whether or not there is a sequence of GEP indices into the pointed type that
/// will land us at the specified offset. If so, fill them into NewIndices and
/// return the resultant element type, otherwise return null.
-Type *InstCombiner::FindElementAtOffset(Type *PtrTy, int64_t Offset,
- SmallVectorImpl<Value*> &NewIndices) {
- assert(PtrTy->isPtrOrPtrVectorTy());
-
- if (!DL)
- return nullptr;
-
- Type *Ty = PtrTy->getPointerElementType();
+Type *InstCombiner::FindElementAtOffset(PointerType *PtrTy, int64_t Offset,
+ SmallVectorImpl<Value *> &NewIndices) {
+ Type *Ty = PtrTy->getElementType();
if (!Ty->isSized())
return nullptr;
// Start with the index over the outer type. Note that the type size
// might be zero (even if the offset isn't zero) if the indexed type
// is something like [0 x {int, int}]
- Type *IntPtrTy = DL->getIntPtrType(PtrTy);
+ Type *IntPtrTy = DL.getIntPtrType(PtrTy);
int64_t FirstIdx = 0;
- if (int64_t TySize = DL->getTypeAllocSize(Ty)) {
+ if (int64_t TySize = DL.getTypeAllocSize(Ty)) {
FirstIdx = Offset/TySize;
Offset -= FirstIdx*TySize;
// Index into the types. If we fail, set OrigBase to null.
while (Offset) {
// Indexing into tail padding between struct/array elements.
- if (uint64_t(Offset*8) >= DL->getTypeSizeInBits(Ty))
+ if (uint64_t(Offset * 8) >= DL.getTypeSizeInBits(Ty))
return nullptr;
if (StructType *STy = dyn_cast<StructType>(Ty)) {
- const StructLayout *SL = DL->getStructLayout(STy);
+ const StructLayout *SL = DL.getStructLayout(STy);
assert(Offset < (int64_t)SL->getSizeInBytes() &&
"Offset must stay within the indexed type");
Offset -= SL->getElementOffset(Elt);
Ty = STy->getElementType(Elt);
} else if (ArrayType *AT = dyn_cast<ArrayType>(Ty)) {
- uint64_t EltSize = DL->getTypeAllocSize(AT->getElementType());
+ uint64_t EltSize = DL.getTypeAllocSize(AT->getElementType());
assert(EltSize && "Cannot index into a zero-sized array");
NewIndices.push_back(ConstantInt::get(IntPtrTy,Offset/EltSize));
Offset %= EltSize;
// It may not be safe to reorder shuffles and things like div, urem, etc.
// because we may trap when executing those ops on unknown vector elements.
// See PR20059.
- if (!isSafeToSpeculativelyExecute(&Inst, DL)) return nullptr;
+ if (!isSafeToSpeculativelyExecute(&Inst))
+ return nullptr;
unsigned VWidth = cast<VectorType>(Inst.getType())->getNumElements();
Value *LHS = Inst.getOperand(0), *RHS = Inst.getOperand(1);
// Eliminate unneeded casts for indices, and replace indices which displace
// by multiples of a zero size type with zero.
- if (DL) {
- bool MadeChange = false;
- Type *IntPtrTy = DL->getIntPtrType(GEP.getPointerOperandType());
-
- gep_type_iterator GTI = gep_type_begin(GEP);
- for (User::op_iterator I = GEP.op_begin() + 1, E = GEP.op_end();
- I != E; ++I, ++GTI) {
- // Skip indices into struct types.
- SequentialType *SeqTy = dyn_cast<SequentialType>(*GTI);
- if (!SeqTy) continue;
-
- // If the element type has zero size then any index over it is equivalent
- // to an index of zero, so replace it with zero if it is not zero already.
- if (SeqTy->getElementType()->isSized() &&
- DL->getTypeAllocSize(SeqTy->getElementType()) == 0)
- if (!isa<Constant>(*I) || !cast<Constant>(*I)->isNullValue()) {
- *I = Constant::getNullValue(IntPtrTy);
- MadeChange = true;
- }
+ bool MadeChange = false;
+ Type *IntPtrTy = DL.getIntPtrType(GEP.getPointerOperandType());
+
+ gep_type_iterator GTI = gep_type_begin(GEP);
+ for (User::op_iterator I = GEP.op_begin() + 1, E = GEP.op_end(); I != E;
+ ++I, ++GTI) {
+ // Skip indices into struct types.
+ SequentialType *SeqTy = dyn_cast<SequentialType>(*GTI);
+ if (!SeqTy)
+ continue;
- Type *IndexTy = (*I)->getType();
- if (IndexTy != IntPtrTy) {
- // If we are using a wider index than needed for this platform, shrink
- // it to what we need. If narrower, sign-extend it to what we need.
- // This explicit cast can make subsequent optimizations more obvious.
- *I = Builder->CreateIntCast(*I, IntPtrTy, true);
+ // If the element type has zero size then any index over it is equivalent
+ // to an index of zero, so replace it with zero if it is not zero already.
+ if (SeqTy->getElementType()->isSized() &&
+ DL.getTypeAllocSize(SeqTy->getElementType()) == 0)
+ if (!isa<Constant>(*I) || !cast<Constant>(*I)->isNullValue()) {
+ *I = Constant::getNullValue(IntPtrTy);
MadeChange = true;
}
+
+ Type *IndexTy = (*I)->getType();
+ if (IndexTy != IntPtrTy) {
+ // If we are using a wider index than needed for this platform, shrink
+ // it to what we need. If narrower, sign-extend it to what we need.
+ // This explicit cast can make subsequent optimizations more obvious.
+ *I = Builder->CreateIntCast(*I, IntPtrTy, true);
+ MadeChange = true;
}
- if (MadeChange) return &GEP;
}
+ if (MadeChange)
+ return &GEP;
// Check to see if the inputs to the PHI node are getelementptr instructions.
if (PHINode *PN = dyn_cast<PHINode>(PtrOp)) {
if (!Op1)
return nullptr;
+ // Don't fold a GEP into itself through a PHI node. This can only happen
+ // through the back-edge of a loop. Folding a GEP into itself means that
+ // the value of the previous iteration needs to be stored in the meantime,
+ // thus requiring an additional register variable to be live, but not
+ // actually achieving anything (the GEP still needs to be executed once per
+ // loop iteration).
+ if (Op1 == &GEP)
+ return nullptr;
+
signed DI = -1;
for (auto I = PN->op_begin()+1, E = PN->op_end(); I !=E; ++I) {
if (!Op2 || Op1->getNumOperands() != Op2->getNumOperands())
return nullptr;
+ // As for Op1 above, don't try to fold a GEP into itself.
+ if (Op2 == &GEP)
+ return nullptr;
+
// Keep track of the type as we walk the GEP.
Type *CurTy = Op1->getOperand(0)->getType()->getScalarType();
// normalized.
if (SO1->getType() != GO1->getType())
return nullptr;
+ // Only do the combine when GO1 and SO1 are both constants. Only in
+ // this case, we are sure the cost after the merge is never more than
+ // that before the merge.
+ if (!isa<Constant>(GO1) || !isa<Constant>(SO1))
+ return nullptr;
Sum = Builder->CreateAdd(SO1, GO1, PtrOp->getName()+".sum");
}
}
if (!Indices.empty())
- return (GEP.isInBounds() && Src->isInBounds()) ?
- GetElementPtrInst::CreateInBounds(Src->getOperand(0), Indices,
- GEP.getName()) :
- GetElementPtrInst::Create(Src->getOperand(0), Indices, GEP.getName());
+ return GEP.isInBounds() && Src->isInBounds()
+ ? GetElementPtrInst::CreateInBounds(
+ Src->getSourceElementType(), Src->getOperand(0), Indices,
+ GEP.getName())
+ : GetElementPtrInst::Create(Src->getSourceElementType(),
+ Src->getOperand(0), Indices,
+ GEP.getName());
}
- if (DL && GEP.getNumIndices() == 1) {
+ if (GEP.getNumIndices() == 1) {
unsigned AS = GEP.getPointerAddressSpace();
if (GEP.getOperand(1)->getType()->getScalarSizeInBits() ==
- DL->getPointerSizeInBits(AS)) {
+ DL.getPointerSizeInBits(AS)) {
Type *PtrTy = GEP.getPointerOperandType();
Type *Ty = PtrTy->getPointerElementType();
- uint64_t TyAllocSize = DL->getTypeAllocSize(Ty);
+ uint64_t TyAllocSize = DL.getTypeAllocSize(Ty);
bool Matched = false;
uint64_t C;
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());
+ GetElementPtrInst *Res = GetElementPtrInst::Create(
+ StrippedPtrTy->getElementType(), StrippedPtr, Idx, GEP.getName());
Res->setIsInBounds(GEP.isInBounds());
if (StrippedPtrTy->getAddressSpace() == GEP.getAddressSpace())
return Res;
// is a leading zero) we can fold the cast into this GEP.
if (StrippedPtrTy->getAddressSpace() == GEP.getAddressSpace()) {
GEP.setOperand(0, StrippedPtr);
+ GEP.setSourceElementType(XATy);
return &GEP;
}
// Cannot replace the base pointer directly because StrippedPtr's
// %0 = GEP [10 x i8] addrspace(1)* X, ...
// addrspacecast i8 addrspace(1)* %0 to i8*
SmallVector<Value*, 8> Idx(GEP.idx_begin(), GEP.idx_end());
- Value *NewGEP = GEP.isInBounds() ?
- Builder->CreateInBoundsGEP(StrippedPtr, Idx, GEP.getName()) :
- Builder->CreateGEP(StrippedPtr, Idx, GEP.getName());
+ Value *NewGEP = GEP.isInBounds()
+ ? Builder->CreateInBoundsGEP(
+ nullptr, StrippedPtr, Idx, GEP.getName())
+ : Builder->CreateGEP(nullptr, StrippedPtr, Idx,
+ GEP.getName());
return new AddrSpaceCastInst(NewGEP, GEP.getType());
}
}
// into: %t1 = getelementptr [2 x i32]* %str, i32 0, i32 %V; bitcast
Type *SrcElTy = StrippedPtrTy->getElementType();
Type *ResElTy = PtrOp->getType()->getPointerElementType();
- if (DL && SrcElTy->isArrayTy() &&
- DL->getTypeAllocSize(SrcElTy->getArrayElementType()) ==
- DL->getTypeAllocSize(ResElTy)) {
- Type *IdxType = DL->getIntPtrType(GEP.getType());
+ if (SrcElTy->isArrayTy() &&
+ DL.getTypeAllocSize(SrcElTy->getArrayElementType()) ==
+ DL.getTypeAllocSize(ResElTy)) {
+ Type *IdxType = DL.getIntPtrType(GEP.getType());
Value *Idx[2] = { Constant::getNullValue(IdxType), GEP.getOperand(1) };
- Value *NewGEP = GEP.isInBounds() ?
- Builder->CreateInBoundsGEP(StrippedPtr, Idx, GEP.getName()) :
- Builder->CreateGEP(StrippedPtr, Idx, GEP.getName());
+ Value *NewGEP =
+ GEP.isInBounds()
+ ? Builder->CreateInBoundsGEP(nullptr, StrippedPtr, Idx,
+ GEP.getName())
+ : Builder->CreateGEP(nullptr, StrippedPtr, Idx, GEP.getName());
// V and GEP are both pointer types --> BitCast
return CastInst::CreatePointerBitCastOrAddrSpaceCast(NewGEP,
// %V = mul i64 %N, 4
// %t = getelementptr i8* bitcast (i32* %arr to i8*), i32 %V
// into: %t1 = getelementptr i32* %arr, i32 %N; bitcast
- if (DL && ResElTy->isSized() && SrcElTy->isSized()) {
+ if (ResElTy->isSized() && SrcElTy->isSized()) {
// Check that changing the type amounts to dividing the index by a scale
// factor.
- uint64_t ResSize = DL->getTypeAllocSize(ResElTy);
- uint64_t SrcSize = DL->getTypeAllocSize(SrcElTy);
+ uint64_t ResSize = DL.getTypeAllocSize(ResElTy);
+ uint64_t SrcSize = DL.getTypeAllocSize(SrcElTy);
if (ResSize && SrcSize % ResSize == 0) {
Value *Idx = GEP.getOperand(1);
unsigned BitWidth = Idx->getType()->getPrimitiveSizeInBits();
// Earlier transforms ensure that the index has type IntPtrType, which
// considerably simplifies the logic by eliminating implicit casts.
- assert(Idx->getType() == DL->getIntPtrType(GEP.getType()) &&
+ assert(Idx->getType() == DL.getIntPtrType(GEP.getType()) &&
"Index not cast to pointer width?");
bool NSW;
// Successfully decomposed Idx as NewIdx * Scale, form a new GEP.
// If the multiplication NewIdx * Scale may overflow then the new
// GEP may not be "inbounds".
- Value *NewGEP = GEP.isInBounds() && NSW ?
- Builder->CreateInBoundsGEP(StrippedPtr, NewIdx, GEP.getName()) :
- Builder->CreateGEP(StrippedPtr, NewIdx, GEP.getName());
+ Value *NewGEP =
+ GEP.isInBounds() && NSW
+ ? Builder->CreateInBoundsGEP(nullptr, StrippedPtr, NewIdx,
+ GEP.getName())
+ : Builder->CreateGEP(nullptr, StrippedPtr, NewIdx,
+ GEP.getName());
// The NewGEP must be pointer typed, so must the old one -> BitCast
return CastInst::CreatePointerBitCastOrAddrSpaceCast(NewGEP,
// 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 (DL && ResElTy->isSized() && SrcElTy->isSized() &&
- SrcElTy->isArrayTy()) {
+ if (ResElTy->isSized() && SrcElTy->isSized() && SrcElTy->isArrayTy()) {
// Check that changing to the array element type amounts to dividing the
// index by a scale factor.
- uint64_t ResSize = DL->getTypeAllocSize(ResElTy);
- uint64_t ArrayEltSize
- = DL->getTypeAllocSize(SrcElTy->getArrayElementType());
+ uint64_t ResSize = DL.getTypeAllocSize(ResElTy);
+ uint64_t ArrayEltSize =
+ DL.getTypeAllocSize(SrcElTy->getArrayElementType());
if (ResSize && ArrayEltSize % ResSize == 0) {
Value *Idx = GEP.getOperand(1);
unsigned BitWidth = Idx->getType()->getPrimitiveSizeInBits();
// Earlier transforms ensure that the index has type IntPtrType, which
// considerably simplifies the logic by eliminating implicit casts.
- assert(Idx->getType() == DL->getIntPtrType(GEP.getType()) &&
+ assert(Idx->getType() == DL.getIntPtrType(GEP.getType()) &&
"Index not cast to pointer width?");
bool NSW;
// If the multiplication NewIdx * Scale may overflow then the new
// GEP may not be "inbounds".
Value *Off[2] = {
- Constant::getNullValue(DL->getIntPtrType(GEP.getType())),
- NewIdx
- };
-
- Value *NewGEP = GEP.isInBounds() && NSW ?
- Builder->CreateInBoundsGEP(StrippedPtr, Off, GEP.getName()) :
- Builder->CreateGEP(StrippedPtr, Off, GEP.getName());
+ Constant::getNullValue(DL.getIntPtrType(GEP.getType())),
+ NewIdx};
+
+ Value *NewGEP = GEP.isInBounds() && NSW
+ ? Builder->CreateInBoundsGEP(
+ SrcElTy, StrippedPtr, Off, GEP.getName())
+ : Builder->CreateGEP(SrcElTy, StrippedPtr, Off,
+ GEP.getName());
// The NewGEP must be pointer typed, so must the old one -> BitCast
return CastInst::CreatePointerBitCastOrAddrSpaceCast(NewGEP,
GEP.getType());
}
}
- if (!DL)
- return nullptr;
-
// addrspacecast between types is canonicalized as a bitcast, then an
// addrspacecast. To take advantage of the below bitcast + struct GEP, look
// through the addrspacecast.
if (BitCastInst *BCI = dyn_cast<BitCastInst>(PtrOp)) {
Value *Operand = BCI->getOperand(0);
PointerType *OpType = cast<PointerType>(Operand->getType());
- unsigned OffsetBits = DL->getPointerTypeSizeInBits(GEP.getType());
+ unsigned OffsetBits = DL.getPointerTypeSizeInBits(GEP.getType());
APInt Offset(OffsetBits, 0);
if (!isa<BitCastInst>(Operand) &&
- GEP.accumulateConstantOffset(*DL, Offset)) {
+ GEP.accumulateConstantOffset(DL, Offset)) {
// If this GEP instruction doesn't move the pointer, just replace the GEP
// with a bitcast of the real input to the dest type.
// GEP.
SmallVector<Value*, 8> NewIndices;
if (FindElementAtOffset(OpType, Offset.getSExtValue(), NewIndices)) {
- Value *NGEP = GEP.isInBounds() ?
- Builder->CreateInBoundsGEP(Operand, NewIndices) :
- Builder->CreateGEP(Operand, NewIndices);
+ Value *NGEP =
+ GEP.isInBounds()
+ ? Builder->CreateInBoundsGEP(nullptr, Operand, NewIndices)
+ : Builder->CreateGEP(nullptr, Operand, NewIndices);
if (NGEP->getType() == GEP.getType())
return ReplaceInstUsesWith(GEP, NGEP);
case Instruction::BitCast:
case Instruction::GetElementPtr:
- Users.push_back(I);
+ Users.emplace_back(I);
Worklist.push_back(I);
continue;
// We can fold eq/ne comparisons with null to false/true, respectively.
if (!ICI->isEquality() || !isa<ConstantPointerNull>(ICI->getOperand(1)))
return false;
- Users.push_back(I);
+ Users.emplace_back(I);
continue;
}
case Intrinsic::lifetime_start:
case Intrinsic::lifetime_end:
case Intrinsic::objectsize:
- Users.push_back(I);
+ Users.emplace_back(I);
continue;
}
}
if (isFreeCall(I, TLI)) {
- Users.push_back(I);
+ Users.emplace_back(I);
continue;
}
return false;
StoreInst *SI = cast<StoreInst>(I);
if (SI->isVolatile() || SI->getPointerOperand() != PI)
return false;
- Users.push_back(I);
+ Users.emplace_back(I);
continue;
}
}
return &BI;
}
+ // If the condition is irrelevant, remove the use so that other
+ // transforms on the condition become more effective.
+ if (BI.isConditional() &&
+ BI.getSuccessor(0) == BI.getSuccessor(1) &&
+ !isa<UndefValue>(BI.getCondition())) {
+ BI.setCondition(UndefValue::get(BI.getCondition()->getType()));
+ return &BI;
+ }
+
// Canonicalize fcmp_one -> fcmp_oeq
FCmpInst::Predicate FPred; Value *Y;
if (match(&BI, m_Br(m_FCmp(FPred, m_Value(X), m_Value(Y)),
Value *Cond = SI.getCondition();
unsigned BitWidth = cast<IntegerType>(Cond->getType())->getBitWidth();
APInt KnownZero(BitWidth, 0), KnownOne(BitWidth, 0);
- computeKnownBits(Cond, KnownZero, KnownOne);
+ computeKnownBits(Cond, KnownZero, KnownOne, 0, &SI);
unsigned LeadingKnownZeros = KnownZero.countLeadingOnes();
unsigned LeadingKnownOnes = KnownOne.countLeadingOnes();
// Truncate the condition operand if the new type is equal to or larger than
// the largest legal integer type. We need to be conservative here since
- // x86 generates redundant zero-extenstion instructions if the operand is
+ // x86 generates redundant zero-extension instructions if the operand is
// truncated to i8 or i16.
bool TruncCond = false;
- if (DL && BitWidth > NewWidth &&
- NewWidth >= DL->getLargestLegalIntTypeSize()) {
+ if (NewWidth > 0 && BitWidth > NewWidth &&
+ NewWidth >= DL.getLargestLegalIntTypeSize()) {
TruncCond = true;
IntegerType *Ty = IntegerType::get(SI.getContext(), NewWidth);
Builder->SetInsertPoint(&SI);
if (!EV.hasIndices())
return ReplaceInstUsesWith(EV, Agg);
- if (Constant *C = dyn_cast<Constant>(Agg)) {
- if (Constant *C2 = C->getAggregateElement(*EV.idx_begin())) {
- if (EV.getNumIndices() == 0)
- return ReplaceInstUsesWith(EV, C2);
- // Extract the remaining indices out of the constant indexed by the
- // first index
- return ExtractValueInst::Create(C2, EV.getIndices().slice(1));
- }
- return nullptr; // Can't handle other constants
- }
+ if (Value *V =
+ SimplifyExtractValueInst(Agg, EV.getIndices(), DL, TLI, DT, AC))
+ return ReplaceInstUsesWith(EV, V);
if (InsertValueInst *IV = dyn_cast<InsertValueInst>(Agg)) {
// We're extracting from an insertvalue instruction, compare the indices
// We need to insert these at the location of the old load, not at that of
// the extractvalue.
Builder->SetInsertPoint(L->getParent(), L);
- Value *GEP = Builder->CreateInBoundsGEP(L->getPointerOperand(), Indices);
+ Value *GEP = Builder->CreateInBoundsGEP(L->getType(),
+ L->getPointerOperand(), Indices);
// Returning the load directly will cause the main loop to insert it in
// the wrong spot, so use ReplaceInstUsesWith().
return ReplaceInstUsesWith(EV, Builder->CreateLoad(GEP));
// The logic here should be correct for any real-world personality function.
// However if that turns out not to be true, the offending logic can always
// be conditioned on the personality function, like the catch-all logic is.
- EHPersonality Personality = classifyEHPersonality(LI.getPersonalityFn());
+ EHPersonality Personality =
+ classifyEHPersonality(LI.getParent()->getParent()->getPersonalityFn());
// Simplify the list of clauses, eg by removing repeated catch clauses
// (these are often created by inlining).
// with a new one.
if (MakeNewInstruction) {
LandingPadInst *NLI = LandingPadInst::Create(LI.getType(),
- LI.getPersonalityFn(),
NewClauses.size());
for (unsigned i = 0, e = NewClauses.size(); i != e; ++i)
NLI->addClause(NewClauses[i]);
}
// Instruction isn't dead, see if we can constant propagate it.
- if (!I->use_empty() && isa<Constant>(I->getOperand(0)))
+ if (!I->use_empty() &&
+ (I->getNumOperands() == 0 || isa<Constant>(I->getOperand(0)))) {
if (Constant *C = ConstantFoldInstruction(I, DL, TLI)) {
DEBUG(dbgs() << "IC: ConstFold to: " << *C << " from: " << *I << '\n');
MadeIRChange = true;
continue;
}
+ }
// See if we can trivially sink this instruction to a successor basic block.
if (I->hasOneUse()) {
DEBUG(dbgs() << "IC: Old = " << *I << '\n'
<< " New = " << *Result << '\n');
- if (!I->getDebugLoc().isUnknown())
+ if (I->getDebugLoc())
Result->setDebugLoc(I->getDebugLoc());
// Everything uses the new instruction now.
I->replaceAllUsesWith(Result);
/// many instructions are dead or constant). Additionally, if we find a branch
/// whose condition is a known constant, we only visit the reachable successors.
///
-static bool AddReachableCodeToWorklist(BasicBlock *BB,
- SmallPtrSetImpl<BasicBlock*> &Visited,
+static bool AddReachableCodeToWorklist(BasicBlock *BB, const DataLayout &DL,
+ SmallPtrSetImpl<BasicBlock *> &Visited,
InstCombineWorklist &ICWorklist,
- const DataLayout *DL,
const TargetLibraryInfo *TLI) {
bool MadeIRChange = false;
SmallVector<BasicBlock*, 256> Worklist;
}
// ConstantProp instruction if trivially constant.
- if (!Inst->use_empty() && isa<Constant>(Inst->getOperand(0)))
+ if (!Inst->use_empty() &&
+ (Inst->getNumOperands() == 0 || isa<Constant>(Inst->getOperand(0))))
if (Constant *C = ConstantFoldInstruction(Inst, DL, TLI)) {
DEBUG(dbgs() << "IC: ConstFold to: " << *C << " from: "
<< *Inst << '\n');
continue;
}
- if (DL) {
- // See if we can constant fold its operands.
- for (User::op_iterator i = Inst->op_begin(), e = Inst->op_end();
- i != e; ++i) {
- ConstantExpr *CE = dyn_cast<ConstantExpr>(i);
- if (CE == nullptr) continue;
+ // See if we can constant fold its operands.
+ for (User::op_iterator i = Inst->op_begin(), e = Inst->op_end(); i != e;
+ ++i) {
+ ConstantExpr *CE = dyn_cast<ConstantExpr>(i);
+ if (CE == nullptr)
+ continue;
- Constant*& FoldRes = FoldedConstants[CE];
- if (!FoldRes)
- FoldRes = ConstantFoldConstantExpression(CE, DL, TLI);
- if (!FoldRes)
- FoldRes = CE;
+ Constant *&FoldRes = FoldedConstants[CE];
+ if (!FoldRes)
+ FoldRes = ConstantFoldConstantExpression(CE, DL, TLI);
+ if (!FoldRes)
+ FoldRes = CE;
- if (FoldRes != CE) {
- *i = FoldRes;
- MadeIRChange = true;
- }
+ if (FoldRes != CE) {
+ *i = FoldRes;
+ MadeIRChange = true;
}
}
///
/// This also does basic constant propagation and other forward fixing to make
/// the combiner itself run much faster.
-static bool prepareICWorklistFromFunction(Function &F, const DataLayout *DL,
+static bool prepareICWorklistFromFunction(Function &F, const DataLayout &DL,
TargetLibraryInfo *TLI,
InstCombineWorklist &ICWorklist) {
bool MadeIRChange = false;
// track of which blocks we visit.
SmallPtrSet<BasicBlock *, 64> Visited;
MadeIRChange |=
- AddReachableCodeToWorklist(F.begin(), Visited, ICWorklist, DL, TLI);
+ AddReachableCodeToWorklist(F.begin(), DL, Visited, ICWorklist, TLI);
// Do a quick scan over the function. If we find any blocks that are
// unreachable, remove any instructions inside of them. This prevents
return MadeIRChange;
}
-static bool combineInstructionsOverFunction(
- Function &F, InstCombineWorklist &Worklist, AssumptionCache &AC,
- TargetLibraryInfo &TLI, DominatorTree &DT, const DataLayout *DL = nullptr,
- LoopInfo *LI = nullptr) {
+static bool
+combineInstructionsOverFunction(Function &F, InstCombineWorklist &Worklist,
+ AliasAnalysis *AA, AssumptionCache &AC,
+ TargetLibraryInfo &TLI, DominatorTree &DT,
+ LoopInfo *LI = nullptr) {
// Minimizing size?
bool MinimizeSize = F.hasFnAttribute(Attribute::MinSize);
+ auto &DL = F.getParent()->getDataLayout();
/// Builder - This is an IRBuilder that automatically inserts new
/// instructions into the worklist when they are created.
if (prepareICWorklistFromFunction(F, DL, &TLI, Worklist))
Changed = true;
- InstCombiner IC(Worklist, &Builder, MinimizeSize, &AC, &TLI, &DT, DL, LI);
+ InstCombiner IC(Worklist, &Builder, MinimizeSize,
+ AA, &AC, &TLI, &DT, DL, LI);
if (IC.run())
Changed = true;
PreservedAnalyses InstCombinePass::run(Function &F,
AnalysisManager<Function> *AM) {
- auto *DL = F.getParent()->getDataLayout();
-
auto &AC = AM->getResult<AssumptionAnalysis>(F);
auto &DT = AM->getResult<DominatorTreeAnalysis>(F);
auto &TLI = AM->getResult<TargetLibraryAnalysis>(F);
auto *LI = AM->getCachedResult<LoopAnalysis>(F);
- if (!combineInstructionsOverFunction(F, Worklist, AC, TLI, DT, DL, LI))
+ // FIXME: The AliasAnalysis is not yet supported in the new pass manager
+ if (!combineInstructionsOverFunction(F, Worklist, nullptr, AC, TLI, DT, LI))
// No changes, all analyses are preserved.
return PreservedAnalyses::all();
void InstructionCombiningPass::getAnalysisUsage(AnalysisUsage &AU) const {
AU.setPreservesCFG();
+ AU.addRequired<AliasAnalysis>();
AU.addRequired<AssumptionCacheTracker>();
AU.addRequired<TargetLibraryInfoWrapperPass>();
AU.addRequired<DominatorTreeWrapperPass>();
return false;
// Required analyses.
+ auto AA = &getAnalysis<AliasAnalysis>();
auto &AC = getAnalysis<AssumptionCacheTracker>().getAssumptionCache(F);
auto &TLI = getAnalysis<TargetLibraryInfoWrapperPass>().getTLI();
auto &DT = getAnalysis<DominatorTreeWrapperPass>().getDomTree();
// Optional analyses.
- auto *DLP = getAnalysisIfAvailable<DataLayoutPass>();
- auto *DL = DLP ? &DLP->getDataLayout() : nullptr;
auto *LIWP = getAnalysisIfAvailable<LoopInfoWrapperPass>();
auto *LI = LIWP ? &LIWP->getLoopInfo() : nullptr;
- return combineInstructionsOverFunction(F, Worklist, AC, TLI, DT, DL, LI);
+ return combineInstructionsOverFunction(F, Worklist, AA, AC, TLI, DT, LI);
}
char InstructionCombiningPass::ID = 0;
INITIALIZE_PASS_DEPENDENCY(AssumptionCacheTracker)
INITIALIZE_PASS_DEPENDENCY(TargetLibraryInfoWrapperPass)
INITIALIZE_PASS_DEPENDENCY(DominatorTreeWrapperPass)
+INITIALIZE_AG_DEPENDENCY(AliasAnalysis)
INITIALIZE_PASS_END(InstructionCombiningPass, "instcombine",
"Combine redundant instructions", false, false)