#include "llvm/Analysis/AssumptionCache.h"
#include "llvm/Analysis/CFG.h"
#include "llvm/Analysis/ConstantFolding.h"
+#include "llvm/Analysis/EHPersonalities.h"
+#include "llvm/Analysis/GlobalsModRef.h"
#include "llvm/Analysis/InstructionSimplify.h"
-#include "llvm/Analysis/LibCallSemantics.h"
#include "llvm/Analysis/LoopInfo.h"
#include "llvm/Analysis/MemoryBuiltins.h"
#include "llvm/Analysis/TargetLibraryInfo.h"
#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>
return llvm::EmitGEPOffset(Builder, DL, GEP);
}
-/// ShouldChangeType - Return true if it is desirable to convert a computation
-/// from 'From' to 'To'. We don't want to convert from a legal to an illegal
-/// type for example, or from a smaller to a larger illegal type.
-bool InstCombiner::ShouldChangeType(Type *From, Type *To) const {
- assert(From->isIntegerTy() && To->isIntegerTy());
-
- unsigned FromWidth = From->getPrimitiveSizeInBits();
- unsigned ToWidth = To->getPrimitiveSizeInBits();
+/// Return true if it is desirable to convert an integer computation from a
+/// given bit width to a new bit width.
+/// We don't want to convert from a legal to an illegal type for example or from
+/// a smaller to a larger illegal type.
+bool InstCombiner::ShouldChangeType(unsigned FromWidth,
+ unsigned ToWidth) const {
bool FromLegal = DL.isLegalInteger(FromWidth);
bool ToLegal = DL.isLegalInteger(ToWidth);
return true;
}
+/// Return true if it is desirable to convert a computation from 'From' to 'To'.
+/// We don't want to convert from a legal to an illegal type for example or from
+/// a smaller to a larger illegal type.
+bool InstCombiner::ShouldChangeType(Type *From, Type *To) const {
+ assert(From->isIntegerTy() && To->isIntegerTy());
+
+ unsigned FromWidth = From->getPrimitiveSizeInBits();
+ unsigned ToWidth = To->getPrimitiveSizeInBits();
+ return ShouldChangeType(FromWidth, ToWidth);
+}
+
// Return true, if No Signed Wrap should be maintained for I.
// The No Signed Wrap flag can be kept if the operation "B (I.getOpcode) C",
// where both B and C should be ConstantInts, results in a constant that does
I.setFastMathFlags(FMF);
}
-/// SimplifyAssociativeOrCommutative - This performs a few simplifications for
-/// operators which are associative or commutative:
-//
-// Commutative operators:
-//
-// 1. Order operands such that they are listed from right (least complex) to
-// left (most complex). This puts constants before unary operators before
-// binary operators.
-//
-// Associative operators:
-//
-// 2. Transform: "(A op B) op C" ==> "A op (B op C)" if "B op C" simplifies.
-// 3. Transform: "A op (B op C)" ==> "(A op B) op C" if "A op B" simplifies.
-//
-// Associative and commutative operators:
-//
-// 4. Transform: "(A op B) op C" ==> "(C op A) op B" if "C op A" simplifies.
-// 5. Transform: "A op (B op C)" ==> "B op (C op A)" if "C op A" simplifies.
-// 6. Transform: "(A op C1) op (B op C2)" ==> "(A op B) op (C1 op C2)"
-// if C1 and C2 are constants.
-//
+/// This performs a few simplifications for operators that are associative or
+/// commutative:
+///
+/// Commutative operators:
+///
+/// 1. Order operands such that they are listed from right (least complex) to
+/// left (most complex). This puts constants before unary operators before
+/// binary operators.
+///
+/// Associative operators:
+///
+/// 2. Transform: "(A op B) op C" ==> "A op (B op C)" if "B op C" simplifies.
+/// 3. Transform: "A op (B op C)" ==> "(A op B) op C" if "A op B" simplifies.
+///
+/// Associative and commutative operators:
+///
+/// 4. Transform: "(A op B) op C" ==> "(C op A) op B" if "C op A" simplifies.
+/// 5. Transform: "A op (B op C)" ==> "B op (C op A)" if "C op A" simplifies.
+/// 6. Transform: "(A op C1) op (B op C2)" ==> "(A op B) op (C1 op C2)"
+/// if C1 and C2 are constants.
bool InstCombiner::SimplifyAssociativeOrCommutative(BinaryOperator &I) {
Instruction::BinaryOps Opcode = I.getOpcode();
bool Changed = false;
} while (1);
}
-/// LeftDistributesOverRight - Whether "X LOp (Y ROp Z)" is always equal to
+/// Return whether "X LOp (Y ROp Z)" is always equal to
/// "(X LOp Y) ROp (X LOp Z)".
static bool LeftDistributesOverRight(Instruction::BinaryOps LOp,
Instruction::BinaryOps ROp) {
}
}
-/// RightDistributesOverLeft - Whether "(X LOp Y) ROp Z" is always equal to
+/// Return whether "(X LOp Y) ROp Z" is always equal to
/// "(X ROp Z) LOp (Y ROp Z)".
static bool RightDistributesOverLeft(Instruction::BinaryOps LOp,
Instruction::BinaryOps ROp) {
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 propagate '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);
}
}
}
return SimplifiedInst;
}
-/// SimplifyUsingDistributiveLaws - This tries to simplify binary operations
-/// which some other binary operation distributes over either by factorizing
-/// out common terms (eg "(A*B)+(A*C)" -> "A*(B+C)") or expanding out if this
-/// results in simplifications (eg: "A & (B | C) -> (A&B) | (A&C)" if this is
-/// a win). Returns the simplified value, or null if it didn't simplify.
+/// This tries to simplify binary operations which some other binary operation
+/// distributes over either by factorizing out common terms
+/// (eg "(A*B)+(A*C)" -> "A*(B+C)") or expanding out if this results in
+/// simplifications (eg: "A & (B | C) -> (A&B) | (A&C)" if this is a win).
+/// Returns the simplified value, or null if it didn't simplify.
Value *InstCombiner::SimplifyUsingDistributiveLaws(BinaryOperator &I) {
Value *LHS = I.getOperand(0), *RHS = I.getOperand(1);
BinaryOperator *Op0 = dyn_cast<BinaryOperator>(LHS);
}
}
+ // (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;
}
-// dyn_castNegVal - Given a 'sub' instruction, return the RHS of the instruction
-// if the LHS is a constant zero (which is the 'negate' form).
-//
+/// Given a 'sub' instruction, return the RHS of the instruction if the LHS is a
+/// constant zero (which is the 'negate' form).
Value *InstCombiner::dyn_castNegVal(Value *V) const {
if (BinaryOperator::isNeg(V))
return BinaryOperator::getNegArgument(V);
return nullptr;
}
-// dyn_castFNegVal - Given a 'fsub' instruction, return the RHS of the
-// instruction if the LHS is a constant negative zero (which is the 'negate'
-// form).
-//
+/// Given a 'fsub' instruction, return the RHS of the instruction if the LHS is
+/// a constant negative zero (which is the 'negate' form).
Value *InstCombiner::dyn_castFNegVal(Value *V, bool IgnoreZeroSign) const {
if (BinaryOperator::isFNeg(V, IgnoreZeroSign))
return BinaryOperator::getFNegArgument(V);
llvm_unreachable("Unknown binary instruction type!");
}
-// FoldOpIntoSelect - Given an instruction with a select as one operand and a
-// constant as the other operand, try to fold the binary operator into the
-// select arguments. This also works for Cast instructions, which obviously do
-// not have a second operand.
+/// Given an instruction with a select as one operand and a constant as the
+/// other operand, try to fold the binary operator into the select arguments.
+/// This also works for Cast instructions, which obviously do not have a second
+/// operand.
Instruction *InstCombiner::FoldOpIntoSelect(Instruction &Op, SelectInst *SI) {
// Don't modify shared select instructions
if (!SI->hasOneUse()) 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).
-///
+/// 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).
Instruction *InstCombiner::FoldOpIntoPhi(Instruction &I) {
PHINode *PN = cast<PHINode>(I.getOperand(0));
unsigned NumPHIValues = PN->getNumIncomingValues();
NewPN->takeName(PN);
// If we are going to have to insert a new computation, do so right before the
- // predecessors terminator.
+ // predecessor's terminator.
if (NonConstBB)
Builder->SetInsertPoint(NonConstBB->getTerminator());
return ReplaceInstUsesWith(I, NewPN);
}
-/// FindElementAtOffset - Given a pointer type and a constant offset, determine
-/// 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,
+/// Given a pointer type and a constant offset, determine 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(PointerType *PtrTy, int64_t Offset,
SmallVectorImpl<Value *> &NewIndices) {
- assert(PtrTy->isPtrOrPtrVectorTy());
-
- Type *Ty = PtrTy->getPointerElementType();
+ Type *Ty = PtrTy->getElementType();
if (!Ty->isSized())
return nullptr;
return true;
}
-/// Descale - Return a value X such that Val = X * Scale, or null if none. If
-/// the multiplication is known not to overflow then NoSignedWrap is set.
+/// Return a value X such that Val = X * Scale, or null if none.
+/// If the multiplication is known not to overflow, then NoSignedWrap is set.
Value *InstCombiner::Descale(Value *Val, APInt Scale, bool &NoSignedWrap) {
assert(isa<IntegerType>(Val->getType()) && "Can only descale integers!");
assert(cast<IntegerType>(Val->getType())->getBitWidth() ==
// 0'th operand of Val.
std::pair<Instruction*, unsigned> Parent;
- // RequireNoSignedWrap - Set if the transform requires a descaling at deeper
- // levels that doesn't overflow.
+ // Set if the transform requires a descaling at deeper levels that doesn't
+ // overflow.
bool RequireNoSignedWrap = false;
- // logScale - log base 2 of the scale. Negative if not a power of 2.
+ // Log base 2 of the scale. Negative if not a power of 2.
int32_t logScale = Scale.exactLogBase2();
for (;; Op = Parent.first->getOperand(Parent.second)) { // Drill down
/// specified one but with other operands.
static Value *CreateBinOpAsGiven(BinaryOperator &Inst, Value *LHS, Value *RHS,
InstCombiner::BuilderTy *B) {
- Value *BORes = B->CreateBinOp(Inst.getOpcode(), LHS, RHS);
- if (BinaryOperator *NewBO = dyn_cast<BinaryOperator>(BORes)) {
- if (isa<OverflowingBinaryOperator>(NewBO)) {
- NewBO->setHasNoSignedWrap(Inst.hasNoSignedWrap());
- NewBO->setHasNoUnsignedWrap(Inst.hasNoUnsignedWrap());
- }
- if (isa<PossiblyExactOperator>(NewBO))
- NewBO->setIsExact(Inst.isExact());
- }
- return BORes;
+ Value *BO = B->CreateBinOp(Inst.getOpcode(), LHS, RHS);
+ // If LHS and RHS are constant, BO won't be a binary operator.
+ if (BinaryOperator *NewBO = dyn_cast<BinaryOperator>(BO))
+ NewBO->copyIRFlags(&Inst);
+ return BO;
}
/// \brief Makes transformation of binary operation specific for vector types.
LShuf->getMask() == RShuf->getMask()) {
Value *NewBO = CreateBinOpAsGiven(Inst, LShuf->getOperand(0),
RShuf->getOperand(0), Builder);
- Value *Res = Builder->CreateShuffleVector(NewBO,
+ return Builder->CreateShuffleVector(NewBO,
UndefValue::get(NewBO->getType()), LShuf->getMask());
- return Res;
}
}
}
if (MayChange) {
Constant *C2 = ConstantVector::get(C2M);
- Value *NewLHS, *NewRHS;
- if (isa<Constant>(LHS)) {
- NewLHS = C2;
- NewRHS = Shuffle->getOperand(0);
- } else {
- NewLHS = Shuffle->getOperand(0);
- NewRHS = C2;
- }
+ Value *NewLHS = isa<Constant>(LHS) ? C2 : Shuffle->getOperand(0);
+ Value *NewRHS = isa<Constant>(LHS) ? Shuffle->getOperand(0) : C2;
Value *NewBO = CreateBinOpAsGiven(Inst, NewLHS, NewRHS, Builder);
- Value *Res = Builder->CreateShuffleVector(NewBO,
+ return Builder->CreateShuffleVector(NewBO,
UndefValue::get(Inst.getType()), Shuffle->getMask());
- return Res;
}
}
// Eliminate unneeded casts for indices, and replace indices which displace
// by multiples of a zero size type with zero.
bool MadeChange = false;
- Type *IntPtrTy = DL.getIntPtrType(GEP.getPointerOperandType());
+ Type *IntPtrTy =
+ DL.getIntPtrType(GEP.getPointerOperandType()->getScalarType());
gep_type_iterator GTI = gep_type_begin(GEP);
for (User::op_iterator I = GEP.op_begin() + 1, E = GEP.op_end(); I != E;
if (!SeqTy)
continue;
+ // Index type should have the same width as IntPtr
+ Type *IndexTy = (*I)->getType();
+ Type *NewIndexType = IndexTy->isVectorTy() ?
+ VectorType::get(IntPtrTy, IndexTy->getVectorNumElements()) : IntPtrTy;
+
// 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);
+ *I = Constant::getNullValue(NewIndexType);
MadeChange = true;
}
- Type *IndexTy = (*I)->getType();
- if (IndexTy != IntPtrTy) {
+ if (IndexTy != NewIndexType) {
// 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);
+ *I = Builder->CreateIntCast(*I, NewIndexType, true);
MadeChange = true;
}
}
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();
}
}
- GetElementPtrInst *NewGEP = cast<GetElementPtrInst>(Op1->clone());
+ // If not all GEPs are identical we'll have to create a new PHI node.
+ // Check that the old PHI node has only one use so that it will get
+ // removed.
+ if (DI != -1 && !PN->hasOneUse())
+ return nullptr;
+ GetElementPtrInst *NewGEP = cast<GetElementPtrInst>(Op1->clone());
if (DI == -1) {
// All the GEPs feeding the PHI are identical. Clone one down into our
// BB so that it can be merged with the current GEP.
// All the GEPs feeding the PHI differ at a single offset. Clone a GEP
// into the current block so it can be merged, and create a new PHI to
// set that index.
- Instruction *InsertPt = Builder->GetInsertPoint();
- Builder->SetInsertPoint(PN);
- PHINode *NewPN = Builder->CreatePHI(Op1->getOperand(DI)->getType(),
- PN->getNumOperands());
- Builder->SetInsertPoint(InsertPt);
+ PHINode *NewPN;
+ {
+ IRBuilderBase::InsertPointGuard Guard(*Builder);
+ Builder->SetInsertPoint(PN);
+ NewPN = Builder->CreatePHI(Op1->getOperand(DI)->getType(),
+ PN->getNumOperands());
+ }
for (auto &I : PN->operands())
NewPN->addIncoming(cast<GEPOperator>(I)->getOperand(DI),
// 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 (GEP.getNumIndices() == 1) {
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());
}
}
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,
// 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,
Constant::getNullValue(DL.getIntPtrType(GEP.getType())),
NewIdx};
- Value *NewGEP = GEP.isInBounds() && NSW ?
- Builder->CreateInBoundsGEP(StrippedPtr, Off, GEP.getName()) :
- Builder->CreateGEP(StrippedPtr, Off, GEP.getName());
+ 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 (Instruction *I = visitBitCast(*BCI)) {
if (I != BCI) {
I->takeName(BCI);
- BCI->getParent()->getInstList().insert(BCI, I);
+ BCI->getParent()->getInstList().insert(BCI->getIterator(), I);
ReplaceInstUsesWith(*BCI, I);
}
return &GEP;
// 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;
}
}
if (InvokeInst *II = dyn_cast<InvokeInst>(&MI)) {
// Replace invoke with a NOP intrinsic to maintain the original CFG
- Module *M = II->getParent()->getParent()->getParent();
+ Module *M = II->getModule();
Function *F = Intrinsic::getDeclaration(M, Intrinsic::donothing);
InvokeInst::Create(F, II->getNormalDest(), II->getUnwindDest(),
None, "", II->getParent());
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)),
// 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 (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
}
if (LoadInst *L = dyn_cast<LoadInst>(Agg))
// If the (non-volatile) load only has one use, we can rewrite this to a
- // load from a GEP. This reduces the size of the load.
- // FIXME: If a load is used only by extractvalue instructions then this
- // could be done regardless of having multiple uses.
+ // load from a GEP. This reduces the size of the load. If a load is used
+ // only by extractvalue instructions then this either must have been
+ // optimized before, or it is a struct with padding, in which case we
+ // don't want to do the transformation as it loses padding knowledge.
if (L->isSimple() && L->hasOneUse()) {
// extractvalue has integer indices, getelementptr has Value*s. Convert.
SmallVector<Value*, 4> 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);
+ Builder->SetInsertPoint(L);
+ 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));
return nullptr;
}
-/// isCatchAll - Return 'true' if the given typeinfo will match anything.
+/// Return 'true' if the given typeinfo will match anything.
static bool isCatchAll(EHPersonality Personality, Constant *TypeInfo) {
switch (Personality) {
case EHPersonality::GNU_C:
case EHPersonality::MSVC_X86SEH:
case EHPersonality::MSVC_Win64SEH:
case EHPersonality::MSVC_CXX:
+ case EHPersonality::CoreCLR:
return TypeInfo->isNullValue();
}
llvm_unreachable("invalid enum");
// 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).
SawCatchAll = true;
break;
}
- if (AlreadyCaught.count(TypeInfo))
- // Already caught by an earlier clause, so having it in the filter
- // is pointless.
- continue;
+
+ // Even if we've seen a type in a catch clause, we don't want to
+ // remove it from the filter. An unexpected type handler may be
+ // set up for a call site which throws an exception of the same
+ // type caught. In order for the exception thrown by the unexpected
+ // handler to propogate correctly, the filter must be correctly
+ // described for the call site.
+ //
+ // Example:
+ //
+ // void unexpected() { throw 1;}
+ // void foo() throw (int) {
+ // std::set_unexpected(unexpected);
+ // try {
+ // throw 2.0;
+ // } catch (int i) {}
+ // }
+
// There is no point in having multiple copies of the same typeinfo in
// a filter, so only add it if we didn't already.
if (SeenInFilter.insert(TypeInfo).second)
// 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]);
return nullptr;
}
-/// TryToSinkInstruction - Try to move the specified instruction from its
-/// current block into the beginning of DestBlock, which can only happen if it's
-/// safe to move the instruction past all of the instructions between it and the
-/// end of its block.
+/// Try to move the specified instruction from its current block into the
+/// beginning of DestBlock, which can only happen if it's safe to move the
+/// instruction past all of the instructions between it and the end of its
+/// block.
static bool TryToSinkInstruction(Instruction *I, BasicBlock *DestBlock) {
assert(I->hasOneUse() && "Invariants didn't hold!");
// Cannot move control-flow-involving, volatile loads, vaarg, etc.
- if (isa<PHINode>(I) || isa<LandingPadInst>(I) || I->mayHaveSideEffects() ||
+ if (isa<PHINode>(I) || I->isEHPad() || I->mayHaveSideEffects() ||
isa<TerminatorInst>(I))
return false;
&DestBlock->getParent()->getEntryBlock())
return false;
+ // Do not sink convergent call instructions.
+ if (auto *CI = dyn_cast<CallInst>(I)) {
+ if (CI->isConvergent())
+ return false;
+ }
+
// We can only sink load instructions if there is nothing between the load and
// the end of block that could change the value.
if (I->mayReadFromMemory()) {
- for (BasicBlock::iterator Scan = I, E = I->getParent()->end();
+ for (BasicBlock::iterator Scan = I->getIterator(),
+ E = I->getParent()->end();
Scan != E; ++Scan)
if (Scan->mayWriteToMemory())
return false;
}
BasicBlock::iterator InsertPos = DestBlock->getFirstInsertionPt();
- I->moveBefore(InsertPos);
+ I->moveBefore(&*InsertPos);
++NumSunkInst;
return true;
}
}
// 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');
}
}
+ // In general, it is possible for computeKnownBits to determine all bits in a
+ // value even when the operands are not all constants.
+ if (!I->use_empty() && I->getType()->isIntegerTy()) {
+ unsigned BitWidth = I->getType()->getScalarSizeInBits();
+ APInt KnownZero(BitWidth, 0);
+ APInt KnownOne(BitWidth, 0);
+ computeKnownBits(I, KnownZero, KnownOne, /*Depth*/0, I);
+ if ((KnownZero | KnownOne).isAllOnesValue()) {
+ Constant *C = ConstantInt::get(I->getContext(), KnownOne);
+ DEBUG(dbgs() << "IC: ConstFold (all bits known) to: " << *C <<
+ " from: " << *I << '\n');
+
+ // Add operands to the worklist.
+ ReplaceInstUsesWith(*I, C);
+ ++NumConstProp;
+ EraseInstFromFunction(*I);
+ MadeIRChange = true;
+ continue;
+ }
+ }
+
// See if we can trivially sink this instruction to a successor basic block.
if (I->hasOneUse()) {
BasicBlock *BB = I->getParent();
}
// Now that we have an instruction, try combining it to simplify it.
- Builder->SetInsertPoint(I->getParent(), I);
+ Builder->SetInsertPoint(I);
Builder->SetCurrentDebugLocation(I->getDebugLoc());
#ifndef NDEBUG
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);
// Insert the new instruction into the basic block...
BasicBlock *InstParent = I->getParent();
- BasicBlock::iterator InsertPos = I;
+ BasicBlock::iterator InsertPos = I->getIterator();
// If we replace a PHI with something that isn't a PHI, fix up the
// insertion point.
return MadeIRChange;
}
-/// AddReachableCodeToWorklist - Walk the function in depth-first order, adding
-/// all reachable code to the worklist.
+/// Walk the function in depth-first order, adding all reachable code to the
+/// worklist.
///
/// This has a couple of tricks to make the code faster and more powerful. In
/// particular, we constant fold and DCE instructions as we go, to avoid adding
continue;
for (BasicBlock::iterator BBI = BB->begin(), E = BB->end(); BBI != E; ) {
- Instruction *Inst = BBI++;
+ Instruction *Inst = &*BBI++;
// DCE instruction if trivially dead.
if (isInstructionTriviallyDead(Inst, TLI)) {
}
// 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');
}
}
- for (unsigned i = 0, e = TI->getNumSuccessors(); i != e; ++i)
- Worklist.push_back(TI->getSuccessor(i));
+ for (BasicBlock *SuccBB : TI->successors())
+ Worklist.push_back(SuccBB);
} while (!Worklist.empty());
// Once we've found all of the instructions to add to instcombine's worklist,
// of the function down. This jives well with the way that it adds all uses
// of instructions to the worklist after doing a transformation, thus avoiding
// some N^2 behavior in pathological cases.
- ICWorklist.AddInitialGroup(&InstrsForInstCombineWorklist[0],
- InstrsForInstCombineWorklist.size());
+ ICWorklist.AddInitialGroup(InstrsForInstCombineWorklist);
return MadeIRChange;
}
// track of which blocks we visit.
SmallPtrSet<BasicBlock *, 64> Visited;
MadeIRChange |=
- AddReachableCodeToWorklist(F.begin(), DL, Visited, ICWorklist, TLI);
+ AddReachableCodeToWorklist(&F.front(), 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
// the instcombine code from having to deal with some bad special cases.
for (Function::iterator BB = F.begin(), E = F.end(); BB != E; ++BB) {
- if (Visited.count(BB))
+ if (Visited.count(&*BB))
continue;
// Delete the instructions backwards, as it has a reduced likelihood of
Instruction *EndInst = BB->getTerminator(); // Last not to be deleted.
while (EndInst != BB->begin()) {
// Delete the next to last instruction.
- BasicBlock::iterator I = EndInst;
- Instruction *Inst = --I;
- if (!Inst->use_empty())
+ Instruction *Inst = &*--EndInst->getIterator();
+ if (!Inst->use_empty() && !Inst->getType()->isTokenTy())
Inst->replaceAllUsesWith(UndefValue::get(Inst->getType()));
- if (isa<LandingPadInst>(Inst)) {
+ if (Inst->isEHPad()) {
EndInst = Inst;
continue;
}
++NumDeadInst;
MadeIRChange = true;
}
- Inst->eraseFromParent();
+ if (!Inst->getType()->isTokenTy())
+ Inst->eraseFromParent();
}
}
static bool
combineInstructionsOverFunction(Function &F, InstCombineWorklist &Worklist,
- AssumptionCache &AC, TargetLibraryInfo &TLI,
- DominatorTree &DT, LoopInfo *LI = nullptr) {
- // Minimizing size?
- bool MinimizeSize = F.hasFnAttribute(Attribute::MinSize);
+ AliasAnalysis *AA, AssumptionCache &AC,
+ TargetLibraryInfo &TLI, DominatorTree &DT,
+ LoopInfo *LI = nullptr) {
auto &DL = F.getParent()->getDataLayout();
/// Builder - This is an IRBuilder that automatically inserts new
if (prepareICWorklistFromFunction(F, DL, &TLI, Worklist))
Changed = true;
- InstCombiner IC(Worklist, &Builder, MinimizeSize, &AC, &TLI, &DT, DL, LI);
+ InstCombiner IC(Worklist, &Builder, F.optForMinSize(),
+ AA, &AC, &TLI, &DT, DL, LI);
if (IC.run())
Changed = true;
auto *LI = AM->getCachedResult<LoopAnalysis>(F);
- if (!combineInstructionsOverFunction(F, Worklist, AC, TLI, DT, 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<AAResultsWrapperPass>();
AU.addRequired<AssumptionCacheTracker>();
AU.addRequired<TargetLibraryInfoWrapperPass>();
AU.addRequired<DominatorTreeWrapperPass>();
AU.addPreserved<DominatorTreeWrapperPass>();
+ AU.addPreserved<GlobalsAAWrapperPass>();
}
bool InstructionCombiningPass::runOnFunction(Function &F) {
return false;
// Required analyses.
+ auto AA = &getAnalysis<AAResultsWrapperPass>().getAAResults();
auto &AC = getAnalysis<AssumptionCacheTracker>().getAssumptionCache(F);
auto &TLI = getAnalysis<TargetLibraryInfoWrapperPass>().getTLI();
auto &DT = getAnalysis<DominatorTreeWrapperPass>().getDomTree();
auto *LIWP = getAnalysisIfAvailable<LoopInfoWrapperPass>();
auto *LI = LIWP ? &LIWP->getLoopInfo() : nullptr;
- return combineInstructionsOverFunction(F, Worklist, AC, TLI, DT, 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_PASS_DEPENDENCY(AAResultsWrapperPass)
+INITIALIZE_PASS_DEPENDENCY(GlobalsAAWrapperPass)
INITIALIZE_PASS_END(InstructionCombiningPass, "instcombine",
"Combine redundant instructions", false, false)
projects / oota-llvm.git / blobdiff