#include "llvm/Analysis/InstructionSimplify.h"
#include "llvm/ADT/SetVector.h"
#include "llvm/ADT/Statistic.h"
+#include "llvm/Analysis/AliasAnalysis.h"
#include "llvm/Analysis/ConstantFolding.h"
#include "llvm/Analysis/MemoryBuiltins.h"
#include "llvm/Analysis/ValueTracking.h"
+#include "llvm/Analysis/VectorUtils.h"
#include "llvm/IR/ConstantRange.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/IR/Dominators.h"
#include "llvm/IR/Operator.h"
#include "llvm/IR/PatternMatch.h"
#include "llvm/IR/ValueHandle.h"
+#include <algorithm>
using namespace llvm;
using namespace llvm::PatternMatch;
STATISTIC(NumExpand, "Number of expansions");
STATISTIC(NumReassoc, "Number of reassociations");
+namespace {
struct Query {
- const DataLayout *DL;
+ const DataLayout &DL;
const TargetLibraryInfo *TLI;
const DominatorTree *DT;
+ AssumptionCache *AC;
+ const Instruction *CxtI;
- Query(const DataLayout *DL, const TargetLibraryInfo *tli,
- const DominatorTree *dt) : DL(DL), TLI(tli), DT(dt) {}
+ Query(const DataLayout &DL, const TargetLibraryInfo *tli,
+ const DominatorTree *dt, AssumptionCache *ac = nullptr,
+ const Instruction *cxti = nullptr)
+ : DL(DL), TLI(tli), DT(dt), AC(ac), CxtI(cxti) {}
};
+} // end anonymous namespace
static Value *SimplifyAndInst(Value *, Value *, const Query &, unsigned);
static Value *SimplifyBinOp(unsigned, Value *, Value *, const Query &,
unsigned);
+static Value *SimplifyFPBinOp(unsigned, Value *, Value *, const FastMathFlags &,
+ const Query &, unsigned);
static Value *SimplifyCmpInst(unsigned, Value *, Value *, const Query &,
unsigned);
static Value *SimplifyOrInst(Value *, Value *, const Query &, unsigned);
// Evaluate the BinOp on the incoming phi values.
Value *CommonValue = nullptr;
- for (unsigned i = 0, e = PI->getNumIncomingValues(); i != e; ++i) {
- Value *Incoming = PI->getIncomingValue(i);
+ for (Value *Incoming : PI->incoming_values()) {
// If the incoming value is the phi node itself, it can safely be skipped.
if (Incoming == PI) continue;
Value *V = PI == LHS ?
// Evaluate the BinOp on the incoming phi values.
Value *CommonValue = nullptr;
- for (unsigned i = 0, e = PI->getNumIncomingValues(); i != e; ++i) {
- Value *Incoming = PI->getIncomingValue(i);
+ for (Value *Incoming : PI->incoming_values()) {
// If the incoming value is the phi node itself, it can safely be skipped.
if (Incoming == PI) continue;
Value *V = SimplifyCmpInst(Pred, Incoming, RHS, Q, MaxRecurse);
}
Value *llvm::SimplifyAddInst(Value *Op0, Value *Op1, bool isNSW, bool isNUW,
- const DataLayout *DL, const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifyAddInst(Op0, Op1, isNSW, isNUW, Query (DL, TLI, DT),
+ const DataLayout &DL, const TargetLibraryInfo *TLI,
+ const DominatorTree *DT, AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifyAddInst(Op0, Op1, isNSW, isNUW, Query(DL, TLI, DT, AC, CxtI),
RecursionLimit);
}
/// This is very similar to GetPointerBaseWithConstantOffset except it doesn't
/// follow non-inbounds geps. This allows it to remain usable for icmp ult/etc.
/// folding.
-static Constant *stripAndComputeConstantOffsets(const DataLayout *DL,
- Value *&V,
+static Constant *stripAndComputeConstantOffsets(const DataLayout &DL, Value *&V,
bool AllowNonInbounds = false) {
assert(V->getType()->getScalarType()->isPointerTy());
- // Without DataLayout, just be conservative for now. Theoretically, more could
- // be done in this case.
- if (!DL)
- return ConstantInt::get(IntegerType::get(V->getContext(), 64), 0);
-
- Type *IntPtrTy = DL->getIntPtrType(V->getType())->getScalarType();
+ Type *IntPtrTy = DL.getIntPtrType(V->getType())->getScalarType();
APInt Offset = APInt::getNullValue(IntPtrTy->getIntegerBitWidth());
// Even though we don't look through PHI nodes, we could be called on an
do {
if (GEPOperator *GEP = dyn_cast<GEPOperator>(V)) {
if ((!AllowNonInbounds && !GEP->isInBounds()) ||
- !GEP->accumulateConstantOffset(*DL, Offset))
+ !GEP->accumulateConstantOffset(DL, Offset))
break;
V = GEP->getPointerOperand();
} else if (Operator::getOpcode(V) == Instruction::BitCast) {
}
assert(V->getType()->getScalarType()->isPointerTy() &&
"Unexpected operand type!");
- } while (Visited.insert(V));
+ } while (Visited.insert(V).second);
Constant *OffsetIntPtr = ConstantInt::get(IntPtrTy, Offset);
if (V->getType()->isVectorTy())
/// \brief Compute the constant difference between two pointer values.
/// If the difference is not a constant, returns zero.
-static Constant *computePointerDifference(const DataLayout *DL,
- Value *LHS, Value *RHS) {
+static Constant *computePointerDifference(const DataLayout &DL, Value *LHS,
+ Value *RHS) {
Constant *LHSOffset = stripAndComputeConstantOffsets(DL, LHS);
Constant *RHSOffset = stripAndComputeConstantOffsets(DL, RHS);
if (Op0 == Op1)
return Constant::getNullValue(Op0->getType());
+ // 0 - X -> 0 if the sub is NUW.
+ if (isNUW && match(Op0, m_Zero()))
+ return Op0;
+
// (X + Y) - Z -> X + (Y - Z) or Y + (X - Z) if everything simplifies.
// For example, (X + Y) - Y -> X; (Y + X) - Y -> X
Value *X = nullptr, *Y = nullptr, *Z = Op1;
}
Value *llvm::SimplifySubInst(Value *Op0, Value *Op1, bool isNSW, bool isNUW,
- const DataLayout *DL, const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifySubInst(Op0, Op1, isNSW, isNUW, Query (DL, TLI, DT),
+ const DataLayout &DL, const TargetLibraryInfo *TLI,
+ const DominatorTree *DT, AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifySubInst(Op0, Op1, isNSW, isNUW, Query(DL, TLI, DT, AC, CxtI),
RecursionLimit);
}
return X;
}
- // fsub nnan ninf x, x ==> 0.0
- if (FMF.noNaNs() && FMF.noInfs() && Op0 == Op1)
+ // fsub nnan x, x ==> 0.0
+ if (FMF.noNaNs() && Op0 == Op1)
return Constant::getNullValue(Op0->getType());
return nullptr;
}
Value *llvm::SimplifyFAddInst(Value *Op0, Value *Op1, FastMathFlags FMF,
- const DataLayout *DL, const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifyFAddInst(Op0, Op1, FMF, Query (DL, TLI, DT), RecursionLimit);
+ const DataLayout &DL,
+ const TargetLibraryInfo *TLI,
+ const DominatorTree *DT, AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifyFAddInst(Op0, Op1, FMF, Query(DL, TLI, DT, AC, CxtI),
+ RecursionLimit);
}
Value *llvm::SimplifyFSubInst(Value *Op0, Value *Op1, FastMathFlags FMF,
- const DataLayout *DL, const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifyFSubInst(Op0, Op1, FMF, Query (DL, TLI, DT), RecursionLimit);
+ const DataLayout &DL,
+ const TargetLibraryInfo *TLI,
+ const DominatorTree *DT, AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifyFSubInst(Op0, Op1, FMF, Query(DL, TLI, DT, AC, CxtI),
+ RecursionLimit);
}
-Value *llvm::SimplifyFMulInst(Value *Op0, Value *Op1,
- FastMathFlags FMF,
- const DataLayout *DL,
+Value *llvm::SimplifyFMulInst(Value *Op0, Value *Op1, FastMathFlags FMF,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifyFMulInst(Op0, Op1, FMF, Query (DL, TLI, DT), RecursionLimit);
+ const DominatorTree *DT, AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifyFMulInst(Op0, Op1, FMF, Query(DL, TLI, DT, AC, CxtI),
+ RecursionLimit);
}
-Value *llvm::SimplifyMulInst(Value *Op0, Value *Op1, const DataLayout *DL,
+Value *llvm::SimplifyMulInst(Value *Op0, Value *Op1, const DataLayout &DL,
const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifyMulInst(Op0, Op1, Query (DL, TLI, DT), RecursionLimit);
+ const DominatorTree *DT, AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifyMulInst(Op0, Op1, Query(DL, TLI, DT, AC, CxtI),
+ RecursionLimit);
}
/// SimplifyDiv - Given operands for an SDiv or UDiv, see if we can
if (match(Op1, m_Undef()))
return Op1;
+ // X / 0 -> undef, we don't need to preserve faults!
+ if (match(Op1, m_Zero()))
+ return UndefValue::get(Op1->getType());
+
// undef / X -> 0
if (match(Op0, m_Undef()))
return Constant::getNullValue(Op0->getType());
(!isSigned && match(Op0, m_URem(m_Value(), m_Specific(Op1)))))
return Constant::getNullValue(Op0->getType());
+ // (X /u C1) /u C2 -> 0 if C1 * C2 overflow
+ ConstantInt *C1, *C2;
+ if (!isSigned && match(Op0, m_UDiv(m_Value(X), m_ConstantInt(C1))) &&
+ match(Op1, m_ConstantInt(C2))) {
+ bool Overflow;
+ C1->getValue().umul_ov(C2->getValue(), Overflow);
+ if (Overflow)
+ return Constant::getNullValue(Op0->getType());
+ }
+
// If the operation is with the result of a select instruction, check whether
// operating on either branch of the select always yields the same value.
if (isa<SelectInst>(Op0) || isa<SelectInst>(Op1))
return nullptr;
}
-Value *llvm::SimplifySDivInst(Value *Op0, Value *Op1, const DataLayout *DL,
+Value *llvm::SimplifySDivInst(Value *Op0, Value *Op1, const DataLayout &DL,
const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifySDivInst(Op0, Op1, Query (DL, TLI, DT), RecursionLimit);
+ const DominatorTree *DT, AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifySDivInst(Op0, Op1, Query(DL, TLI, DT, AC, CxtI),
+ RecursionLimit);
}
/// SimplifyUDivInst - Given operands for a UDiv, see if we can
return nullptr;
}
-Value *llvm::SimplifyUDivInst(Value *Op0, Value *Op1, const DataLayout *DL,
+Value *llvm::SimplifyUDivInst(Value *Op0, Value *Op1, const DataLayout &DL,
const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifyUDivInst(Op0, Op1, Query (DL, TLI, DT), RecursionLimit);
+ const DominatorTree *DT, AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifyUDivInst(Op0, Op1, Query(DL, TLI, DT, AC, CxtI),
+ RecursionLimit);
}
-static Value *SimplifyFDivInst(Value *Op0, Value *Op1, const Query &Q,
- unsigned) {
+static Value *SimplifyFDivInst(Value *Op0, Value *Op1, FastMathFlags FMF,
+ const Query &Q, unsigned) {
// undef / X -> undef (the undef could be a snan).
if (match(Op0, m_Undef()))
return Op0;
if (match(Op1, m_Undef()))
return Op1;
+ // 0 / X -> 0
+ // Requires that NaNs are off (X could be zero) and signed zeroes are
+ // ignored (X could be positive or negative, so the output sign is unknown).
+ if (FMF.noNaNs() && FMF.noSignedZeros() && match(Op0, m_AnyZero()))
+ return Op0;
+
+ if (FMF.noNaNs()) {
+ // X / X -> 1.0 is legal when NaNs are ignored.
+ if (Op0 == Op1)
+ return ConstantFP::get(Op0->getType(), 1.0);
+
+ // -X / X -> -1.0 and
+ // X / -X -> -1.0 are legal when NaNs are ignored.
+ // We can ignore signed zeros because +-0.0/+-0.0 is NaN and ignored.
+ if ((BinaryOperator::isFNeg(Op0, /*IgnoreZeroSign=*/true) &&
+ BinaryOperator::getFNegArgument(Op0) == Op1) ||
+ (BinaryOperator::isFNeg(Op1, /*IgnoreZeroSign=*/true) &&
+ BinaryOperator::getFNegArgument(Op1) == Op0))
+ return ConstantFP::get(Op0->getType(), -1.0);
+ }
+
return nullptr;
}
-Value *llvm::SimplifyFDivInst(Value *Op0, Value *Op1, const DataLayout *DL,
+Value *llvm::SimplifyFDivInst(Value *Op0, Value *Op1, FastMathFlags FMF,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifyFDivInst(Op0, Op1, Query (DL, TLI, DT), RecursionLimit);
+ const DominatorTree *DT, AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifyFDivInst(Op0, Op1, FMF, Query(DL, TLI, DT, AC, CxtI),
+ RecursionLimit);
}
/// SimplifyRem - Given operands for an SRem or URem, see if we can
if (Op0 == Op1)
return Constant::getNullValue(Op0->getType());
+ // (X % Y) % Y -> X % Y
+ if ((Opcode == Instruction::SRem &&
+ match(Op0, m_SRem(m_Value(), m_Specific(Op1)))) ||
+ (Opcode == Instruction::URem &&
+ match(Op0, m_URem(m_Value(), m_Specific(Op1)))))
+ return Op0;
+
// If the operation is with the result of a select instruction, check whether
// operating on either branch of the select always yields the same value.
if (isa<SelectInst>(Op0) || isa<SelectInst>(Op1))
return nullptr;
}
-Value *llvm::SimplifySRemInst(Value *Op0, Value *Op1, const DataLayout *DL,
+Value *llvm::SimplifySRemInst(Value *Op0, Value *Op1, const DataLayout &DL,
const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifySRemInst(Op0, Op1, Query (DL, TLI, DT), RecursionLimit);
+ const DominatorTree *DT, AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifySRemInst(Op0, Op1, Query(DL, TLI, DT, AC, CxtI),
+ RecursionLimit);
}
/// SimplifyURemInst - Given operands for a URem, see if we can
return nullptr;
}
-Value *llvm::SimplifyURemInst(Value *Op0, Value *Op1, const DataLayout *DL,
+Value *llvm::SimplifyURemInst(Value *Op0, Value *Op1, const DataLayout &DL,
const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifyURemInst(Op0, Op1, Query (DL, TLI, DT), RecursionLimit);
+ const DominatorTree *DT, AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifyURemInst(Op0, Op1, Query(DL, TLI, DT, AC, CxtI),
+ RecursionLimit);
}
-static Value *SimplifyFRemInst(Value *Op0, Value *Op1, const Query &,
- unsigned) {
+static Value *SimplifyFRemInst(Value *Op0, Value *Op1, FastMathFlags FMF,
+ const Query &, unsigned) {
// undef % X -> undef (the undef could be a snan).
if (match(Op0, m_Undef()))
return Op0;
if (match(Op1, m_Undef()))
return Op1;
+ // 0 % X -> 0
+ // Requires that NaNs are off (X could be zero) and signed zeroes are
+ // ignored (X could be positive or negative, so the output sign is unknown).
+ if (FMF.noNaNs() && FMF.noSignedZeros() && match(Op0, m_AnyZero()))
+ return Op0;
+
return nullptr;
}
-Value *llvm::SimplifyFRemInst(Value *Op0, Value *Op1, const DataLayout *DL,
+Value *llvm::SimplifyFRemInst(Value *Op0, Value *Op1, FastMathFlags FMF,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifyFRemInst(Op0, Op1, Query (DL, TLI, DT), RecursionLimit);
+ const DominatorTree *DT, AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifyFRemInst(Op0, Op1, FMF, Query(DL, TLI, DT, AC, CxtI),
+ RecursionLimit);
}
/// isUndefShift - Returns true if a shift by \c Amount always yields undef.
return nullptr;
}
+/// \brief Given operands for an Shl, LShr or AShr, see if we can
+/// fold the result. If not, this returns null.
+static Value *SimplifyRightShift(unsigned Opcode, Value *Op0, Value *Op1,
+ bool isExact, const Query &Q,
+ unsigned MaxRecurse) {
+ if (Value *V = SimplifyShift(Opcode, Op0, Op1, Q, MaxRecurse))
+ return V;
+
+ // X >> X -> 0
+ if (Op0 == Op1)
+ return Constant::getNullValue(Op0->getType());
+
+ // undef >> X -> 0
+ // undef >> X -> undef (if it's exact)
+ if (match(Op0, m_Undef()))
+ return isExact ? Op0 : Constant::getNullValue(Op0->getType());
+
+ // The low bit cannot be shifted out of an exact shift if it is set.
+ if (isExact) {
+ unsigned BitWidth = Op0->getType()->getScalarSizeInBits();
+ APInt Op0KnownZero(BitWidth, 0);
+ APInt Op0KnownOne(BitWidth, 0);
+ computeKnownBits(Op0, Op0KnownZero, Op0KnownOne, Q.DL, /*Depth=*/0, Q.AC,
+ Q.CxtI, Q.DT);
+ if (Op0KnownOne[0])
+ return Op0;
+ }
+
+ return nullptr;
+}
+
/// SimplifyShlInst - Given operands for an Shl, see if we can
/// fold the result. If not, this returns null.
static Value *SimplifyShlInst(Value *Op0, Value *Op1, bool isNSW, bool isNUW,
return V;
// undef << X -> 0
+ // undef << X -> undef if (if it's NSW/NUW)
if (match(Op0, m_Undef()))
- return Constant::getNullValue(Op0->getType());
+ return isNSW || isNUW ? Op0 : Constant::getNullValue(Op0->getType());
// (X >> A) << A -> X
Value *X;
}
Value *llvm::SimplifyShlInst(Value *Op0, Value *Op1, bool isNSW, bool isNUW,
- const DataLayout *DL, const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifyShlInst(Op0, Op1, isNSW, isNUW, Query (DL, TLI, DT),
+ const DataLayout &DL, const TargetLibraryInfo *TLI,
+ const DominatorTree *DT, AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifyShlInst(Op0, Op1, isNSW, isNUW, Query(DL, TLI, DT, AC, CxtI),
RecursionLimit);
}
/// fold the result. If not, this returns null.
static Value *SimplifyLShrInst(Value *Op0, Value *Op1, bool isExact,
const Query &Q, unsigned MaxRecurse) {
- if (Value *V = SimplifyShift(Instruction::LShr, Op0, Op1, Q, MaxRecurse))
- return V;
-
- // X >> X -> 0
- if (Op0 == Op1)
- return Constant::getNullValue(Op0->getType());
-
- // undef >>l X -> 0
- if (match(Op0, m_Undef()))
- return Constant::getNullValue(Op0->getType());
+ if (Value *V = SimplifyRightShift(Instruction::LShr, Op0, Op1, isExact, Q,
+ MaxRecurse))
+ return V;
// (X << A) >> A -> X
Value *X;
- if (match(Op0, m_Shl(m_Value(X), m_Specific(Op1))) &&
- cast<OverflowingBinaryOperator>(Op0)->hasNoUnsignedWrap())
+ if (match(Op0, m_NUWShl(m_Value(X), m_Specific(Op1))))
return X;
return nullptr;
}
Value *llvm::SimplifyLShrInst(Value *Op0, Value *Op1, bool isExact,
- const DataLayout *DL,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifyLShrInst(Op0, Op1, isExact, Query (DL, TLI, DT),
+ const DominatorTree *DT, AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifyLShrInst(Op0, Op1, isExact, Query(DL, TLI, DT, AC, CxtI),
RecursionLimit);
}
/// fold the result. If not, this returns null.
static Value *SimplifyAShrInst(Value *Op0, Value *Op1, bool isExact,
const Query &Q, unsigned MaxRecurse) {
- if (Value *V = SimplifyShift(Instruction::AShr, Op0, Op1, Q, MaxRecurse))
+ if (Value *V = SimplifyRightShift(Instruction::AShr, Op0, Op1, isExact, Q,
+ MaxRecurse))
return V;
- // X >> X -> 0
- if (Op0 == Op1)
- return Constant::getNullValue(Op0->getType());
-
// all ones >>a X -> all ones
if (match(Op0, m_AllOnes()))
return Op0;
- // undef >>a X -> all ones
- if (match(Op0, m_Undef()))
- return Constant::getAllOnesValue(Op0->getType());
-
// (X << A) >> A -> X
Value *X;
- if (match(Op0, m_Shl(m_Value(X), m_Specific(Op1))) &&
- cast<OverflowingBinaryOperator>(Op0)->hasNoSignedWrap())
+ if (match(Op0, m_NSWShl(m_Value(X), m_Specific(Op1))))
return X;
// Arithmetic shifting an all-sign-bit value is a no-op.
- unsigned NumSignBits = ComputeNumSignBits(Op0);
+ unsigned NumSignBits = ComputeNumSignBits(Op0, Q.DL, 0, Q.AC, Q.CxtI, Q.DT);
if (NumSignBits == Op0->getType()->getScalarSizeInBits())
return Op0;
}
Value *llvm::SimplifyAShrInst(Value *Op0, Value *Op1, bool isExact,
- const DataLayout *DL,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifyAShrInst(Op0, Op1, isExact, Query (DL, TLI, DT),
+ const DominatorTree *DT, AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifyAShrInst(Op0, Op1, isExact, Query(DL, TLI, DT, AC, CxtI),
RecursionLimit);
}
+static Value *simplifyUnsignedRangeCheck(ICmpInst *ZeroICmp,
+ ICmpInst *UnsignedICmp, bool IsAnd) {
+ Value *X, *Y;
+
+ ICmpInst::Predicate EqPred;
+ if (!match(ZeroICmp, m_ICmp(EqPred, m_Value(Y), m_Zero())) ||
+ !ICmpInst::isEquality(EqPred))
+ return nullptr;
+
+ ICmpInst::Predicate UnsignedPred;
+ if (match(UnsignedICmp, m_ICmp(UnsignedPred, m_Value(X), m_Specific(Y))) &&
+ ICmpInst::isUnsigned(UnsignedPred))
+ ;
+ else if (match(UnsignedICmp,
+ m_ICmp(UnsignedPred, m_Value(Y), m_Specific(X))) &&
+ ICmpInst::isUnsigned(UnsignedPred))
+ UnsignedPred = ICmpInst::getSwappedPredicate(UnsignedPred);
+ else
+ return nullptr;
+
+ // X < Y && Y != 0 --> X < Y
+ // X < Y || Y != 0 --> Y != 0
+ if (UnsignedPred == ICmpInst::ICMP_ULT && EqPred == ICmpInst::ICMP_NE)
+ return IsAnd ? UnsignedICmp : ZeroICmp;
+
+ // X >= Y || Y != 0 --> true
+ // X >= Y || Y == 0 --> X >= Y
+ if (UnsignedPred == ICmpInst::ICMP_UGE && !IsAnd) {
+ if (EqPred == ICmpInst::ICMP_NE)
+ return getTrue(UnsignedICmp->getType());
+ return UnsignedICmp;
+ }
+
+ // X < Y && Y == 0 --> false
+ if (UnsignedPred == ICmpInst::ICMP_ULT && EqPred == ICmpInst::ICMP_EQ &&
+ IsAnd)
+ return getFalse(UnsignedICmp->getType());
+
+ return nullptr;
+}
+
+// Simplify (and (icmp ...) (icmp ...)) to true when we can tell that the range
+// of possible values cannot be satisfied.
+static Value *SimplifyAndOfICmps(ICmpInst *Op0, ICmpInst *Op1) {
+ ICmpInst::Predicate Pred0, Pred1;
+ ConstantInt *CI1, *CI2;
+ Value *V;
+
+ if (Value *X = simplifyUnsignedRangeCheck(Op0, Op1, /*IsAnd=*/true))
+ return X;
+
+ if (!match(Op0, m_ICmp(Pred0, m_Add(m_Value(V), m_ConstantInt(CI1)),
+ m_ConstantInt(CI2))))
+ return nullptr;
+
+ if (!match(Op1, m_ICmp(Pred1, m_Specific(V), m_Specific(CI1))))
+ return nullptr;
+
+ Type *ITy = Op0->getType();
+
+ auto *AddInst = cast<BinaryOperator>(Op0->getOperand(0));
+ bool isNSW = AddInst->hasNoSignedWrap();
+ bool isNUW = AddInst->hasNoUnsignedWrap();
+
+ const APInt &CI1V = CI1->getValue();
+ const APInt &CI2V = CI2->getValue();
+ const APInt Delta = CI2V - CI1V;
+ if (CI1V.isStrictlyPositive()) {
+ if (Delta == 2) {
+ if (Pred0 == ICmpInst::ICMP_ULT && Pred1 == ICmpInst::ICMP_SGT)
+ return getFalse(ITy);
+ if (Pred0 == ICmpInst::ICMP_SLT && Pred1 == ICmpInst::ICMP_SGT && isNSW)
+ return getFalse(ITy);
+ }
+ if (Delta == 1) {
+ if (Pred0 == ICmpInst::ICMP_ULE && Pred1 == ICmpInst::ICMP_SGT)
+ return getFalse(ITy);
+ if (Pred0 == ICmpInst::ICMP_SLE && Pred1 == ICmpInst::ICMP_SGT && isNSW)
+ return getFalse(ITy);
+ }
+ }
+ if (CI1V.getBoolValue() && isNUW) {
+ if (Delta == 2)
+ if (Pred0 == ICmpInst::ICMP_ULT && Pred1 == ICmpInst::ICMP_UGT)
+ return getFalse(ITy);
+ if (Delta == 1)
+ if (Pred0 == ICmpInst::ICMP_ULE && Pred1 == ICmpInst::ICMP_UGT)
+ return getFalse(ITy);
+ }
+
+ return nullptr;
+}
+
/// SimplifyAndInst - Given operands for an And, see if we can
/// fold the result. If not, this returns null.
static Value *SimplifyAndInst(Value *Op0, Value *Op1, const Query &Q,
// A & (-A) = A if A is a power of two or zero.
if (match(Op0, m_Neg(m_Specific(Op1))) ||
match(Op1, m_Neg(m_Specific(Op0)))) {
- if (isKnownToBeAPowerOfTwo(Op0, /*OrZero*/true))
+ if (isKnownToBeAPowerOfTwo(Op0, Q.DL, /*OrZero*/ true, 0, Q.AC, Q.CxtI,
+ Q.DT))
return Op0;
- if (isKnownToBeAPowerOfTwo(Op1, /*OrZero*/true))
+ if (isKnownToBeAPowerOfTwo(Op1, Q.DL, /*OrZero*/ true, 0, Q.AC, Q.CxtI,
+ Q.DT))
return Op1;
}
+ if (auto *ICILHS = dyn_cast<ICmpInst>(Op0)) {
+ if (auto *ICIRHS = dyn_cast<ICmpInst>(Op1)) {
+ if (Value *V = SimplifyAndOfICmps(ICILHS, ICIRHS))
+ return V;
+ if (Value *V = SimplifyAndOfICmps(ICIRHS, ICILHS))
+ return V;
+ }
+ }
+
// Try some generic simplifications for associative operations.
if (Value *V = SimplifyAssociativeBinOp(Instruction::And, Op0, Op1, Q,
MaxRecurse))
return nullptr;
}
-Value *llvm::SimplifyAndInst(Value *Op0, Value *Op1, const DataLayout *DL,
+Value *llvm::SimplifyAndInst(Value *Op0, Value *Op1, const DataLayout &DL,
const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifyAndInst(Op0, Op1, Query (DL, TLI, DT), RecursionLimit);
+ const DominatorTree *DT, AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifyAndInst(Op0, Op1, Query(DL, TLI, DT, AC, CxtI),
+ RecursionLimit);
+}
+
+// Simplify (or (icmp ...) (icmp ...)) to true when we can tell that the union
+// contains all possible values.
+static Value *SimplifyOrOfICmps(ICmpInst *Op0, ICmpInst *Op1) {
+ ICmpInst::Predicate Pred0, Pred1;
+ ConstantInt *CI1, *CI2;
+ Value *V;
+
+ if (Value *X = simplifyUnsignedRangeCheck(Op0, Op1, /*IsAnd=*/false))
+ return X;
+
+ if (!match(Op0, m_ICmp(Pred0, m_Add(m_Value(V), m_ConstantInt(CI1)),
+ m_ConstantInt(CI2))))
+ return nullptr;
+
+ if (!match(Op1, m_ICmp(Pred1, m_Specific(V), m_Specific(CI1))))
+ return nullptr;
+
+ Type *ITy = Op0->getType();
+
+ auto *AddInst = cast<BinaryOperator>(Op0->getOperand(0));
+ bool isNSW = AddInst->hasNoSignedWrap();
+ bool isNUW = AddInst->hasNoUnsignedWrap();
+
+ const APInt &CI1V = CI1->getValue();
+ const APInt &CI2V = CI2->getValue();
+ const APInt Delta = CI2V - CI1V;
+ if (CI1V.isStrictlyPositive()) {
+ if (Delta == 2) {
+ if (Pred0 == ICmpInst::ICMP_UGE && Pred1 == ICmpInst::ICMP_SLE)
+ return getTrue(ITy);
+ if (Pred0 == ICmpInst::ICMP_SGE && Pred1 == ICmpInst::ICMP_SLE && isNSW)
+ return getTrue(ITy);
+ }
+ if (Delta == 1) {
+ if (Pred0 == ICmpInst::ICMP_UGT && Pred1 == ICmpInst::ICMP_SLE)
+ return getTrue(ITy);
+ if (Pred0 == ICmpInst::ICMP_SGT && Pred1 == ICmpInst::ICMP_SLE && isNSW)
+ return getTrue(ITy);
+ }
+ }
+ if (CI1V.getBoolValue() && isNUW) {
+ if (Delta == 2)
+ if (Pred0 == ICmpInst::ICMP_UGE && Pred1 == ICmpInst::ICMP_ULE)
+ return getTrue(ITy);
+ if (Delta == 1)
+ if (Pred0 == ICmpInst::ICMP_UGT && Pred1 == ICmpInst::ICMP_ULE)
+ return getTrue(ITy);
+ }
+
+ return nullptr;
}
/// SimplifyOrInst - Given operands for an Or, see if we can
(A == Op0 || B == Op0))
return Constant::getAllOnesValue(Op0->getType());
+ if (auto *ICILHS = dyn_cast<ICmpInst>(Op0)) {
+ if (auto *ICIRHS = dyn_cast<ICmpInst>(Op1)) {
+ if (Value *V = SimplifyOrOfICmps(ICILHS, ICIRHS))
+ return V;
+ if (Value *V = SimplifyOrOfICmps(ICIRHS, ICILHS))
+ return V;
+ }
+ }
+
// Try some generic simplifications for associative operations.
if (Value *V = SimplifyAssociativeBinOp(Instruction::Or, Op0, Op1, Q,
MaxRecurse))
if ((C2->getValue() & (C2->getValue() + 1)) == 0 && // C2 == 0+1+
match(A, m_Add(m_Value(V1), m_Value(V2)))) {
// Add commutes, try both ways.
- if (V1 == B && MaskedValueIsZero(V2, C2->getValue()))
+ if (V1 == B &&
+ MaskedValueIsZero(V2, C2->getValue(), Q.DL, 0, Q.AC, Q.CxtI, Q.DT))
return A;
- if (V2 == B && MaskedValueIsZero(V1, C2->getValue()))
+ if (V2 == B &&
+ MaskedValueIsZero(V1, C2->getValue(), Q.DL, 0, Q.AC, Q.CxtI, Q.DT))
return A;
}
// Or commutes, try both ways.
if ((C1->getValue() & (C1->getValue() + 1)) == 0 &&
match(B, m_Add(m_Value(V1), m_Value(V2)))) {
// Add commutes, try both ways.
- if (V1 == A && MaskedValueIsZero(V2, C1->getValue()))
+ if (V1 == A &&
+ MaskedValueIsZero(V2, C1->getValue(), Q.DL, 0, Q.AC, Q.CxtI, Q.DT))
return B;
- if (V2 == A && MaskedValueIsZero(V1, C1->getValue()))
+ if (V2 == A &&
+ MaskedValueIsZero(V1, C1->getValue(), Q.DL, 0, Q.AC, Q.CxtI, Q.DT))
return B;
}
}
return nullptr;
}
-Value *llvm::SimplifyOrInst(Value *Op0, Value *Op1, const DataLayout *DL,
+Value *llvm::SimplifyOrInst(Value *Op0, Value *Op1, const DataLayout &DL,
const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifyOrInst(Op0, Op1, Query (DL, TLI, DT), RecursionLimit);
+ const DominatorTree *DT, AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifyOrInst(Op0, Op1, Query(DL, TLI, DT, AC, CxtI),
+ RecursionLimit);
}
/// SimplifyXorInst - Given operands for a Xor, see if we can
return nullptr;
}
-Value *llvm::SimplifyXorInst(Value *Op0, Value *Op1, const DataLayout *DL,
+Value *llvm::SimplifyXorInst(Value *Op0, Value *Op1, const DataLayout &DL,
const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifyXorInst(Op0, Op1, Query (DL, TLI, DT), RecursionLimit);
+ const DominatorTree *DT, AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifyXorInst(Op0, Op1, Query(DL, TLI, DT, AC, CxtI),
+ RecursionLimit);
}
static Type *GetCompareTy(Value *Op) {
// If the C and C++ standards are ever made sufficiently restrictive in this
// area, it may be possible to update LLVM's semantics accordingly and reinstate
// this optimization.
-static Constant *computePointerICmp(const DataLayout *DL,
+static Constant *computePointerICmp(const DataLayout &DL,
const TargetLibraryInfo *TLI,
- CmpInst::Predicate Pred,
- Value *LHS, Value *RHS) {
+ CmpInst::Predicate Pred, Value *LHS,
+ Value *RHS) {
// First, skip past any trivial no-ops.
LHS = LHS->stripPointerCasts();
RHS = RHS->stripPointerCasts();
return ConstantExpr::getICmp(Pred,
ConstantExpr::getAdd(LHSOffset, LHSNoBound),
ConstantExpr::getAdd(RHSOffset, RHSNoBound));
+
+ // If one side of the equality comparison must come from a noalias call
+ // (meaning a system memory allocation function), and the other side must
+ // come from a pointer that cannot overlap with dynamically-allocated
+ // memory within the lifetime of the current function (allocas, byval
+ // arguments, globals), then determine the comparison result here.
+ SmallVector<Value *, 8> LHSUObjs, RHSUObjs;
+ GetUnderlyingObjects(LHS, LHSUObjs, DL);
+ GetUnderlyingObjects(RHS, RHSUObjs, DL);
+
+ // Is the set of underlying objects all noalias calls?
+ auto IsNAC = [](SmallVectorImpl<Value *> &Objects) {
+ return std::all_of(Objects.begin(), Objects.end(),
+ [](Value *V){ return isNoAliasCall(V); });
+ };
+
+ // Is the set of underlying objects all things which must be disjoint from
+ // noalias calls. For allocas, we consider only static ones (dynamic
+ // allocas might be transformed into calls to malloc not simultaneously
+ // live with the compared-to allocation). For globals, we exclude symbols
+ // that might be resolve lazily to symbols in another dynamically-loaded
+ // library (and, thus, could be malloc'ed by the implementation).
+ auto IsAllocDisjoint = [](SmallVectorImpl<Value *> &Objects) {
+ return std::all_of(Objects.begin(), Objects.end(),
+ [](Value *V){
+ if (const AllocaInst *AI = dyn_cast<AllocaInst>(V))
+ return AI->getParent() && AI->getParent()->getParent() &&
+ AI->isStaticAlloca();
+ if (const GlobalValue *GV = dyn_cast<GlobalValue>(V))
+ return (GV->hasLocalLinkage() ||
+ GV->hasHiddenVisibility() ||
+ GV->hasProtectedVisibility() ||
+ GV->hasUnnamedAddr()) &&
+ !GV->isThreadLocal();
+ if (const Argument *A = dyn_cast<Argument>(V))
+ return A->hasByValAttr();
+ return false;
+ });
+ };
+
+ if ((IsNAC(LHSUObjs) && IsAllocDisjoint(RHSUObjs)) ||
+ (IsNAC(RHSUObjs) && IsAllocDisjoint(LHSUObjs)))
+ return ConstantInt::get(GetCompareTy(LHS),
+ !CmpInst::isTrueWhenEqual(Pred));
}
// Otherwise, fail.
return getTrue(ITy);
case ICmpInst::ICMP_EQ:
case ICmpInst::ICMP_ULE:
- if (isKnownNonZero(LHS, Q.DL))
+ if (isKnownNonZero(LHS, Q.DL, 0, Q.AC, Q.CxtI, Q.DT))
return getFalse(ITy);
break;
case ICmpInst::ICMP_NE:
case ICmpInst::ICMP_UGT:
- if (isKnownNonZero(LHS, Q.DL))
+ if (isKnownNonZero(LHS, Q.DL, 0, Q.AC, Q.CxtI, Q.DT))
return getTrue(ITy);
break;
case ICmpInst::ICMP_SLT:
- ComputeSignBit(LHS, LHSKnownNonNegative, LHSKnownNegative, Q.DL);
+ ComputeSignBit(LHS, LHSKnownNonNegative, LHSKnownNegative, Q.DL, 0, Q.AC,
+ Q.CxtI, Q.DT);
if (LHSKnownNegative)
return getTrue(ITy);
if (LHSKnownNonNegative)
return getFalse(ITy);
break;
case ICmpInst::ICMP_SLE:
- ComputeSignBit(LHS, LHSKnownNonNegative, LHSKnownNegative, Q.DL);
+ ComputeSignBit(LHS, LHSKnownNonNegative, LHSKnownNegative, Q.DL, 0, Q.AC,
+ Q.CxtI, Q.DT);
if (LHSKnownNegative)
return getTrue(ITy);
- if (LHSKnownNonNegative && isKnownNonZero(LHS, Q.DL))
+ if (LHSKnownNonNegative &&
+ isKnownNonZero(LHS, Q.DL, 0, Q.AC, Q.CxtI, Q.DT))
return getFalse(ITy);
break;
case ICmpInst::ICMP_SGE:
- ComputeSignBit(LHS, LHSKnownNonNegative, LHSKnownNegative, Q.DL);
+ ComputeSignBit(LHS, LHSKnownNonNegative, LHSKnownNegative, Q.DL, 0, Q.AC,
+ Q.CxtI, Q.DT);
if (LHSKnownNegative)
return getFalse(ITy);
if (LHSKnownNonNegative)
return getTrue(ITy);
break;
case ICmpInst::ICMP_SGT:
- ComputeSignBit(LHS, LHSKnownNonNegative, LHSKnownNegative, Q.DL);
+ ComputeSignBit(LHS, LHSKnownNonNegative, LHSKnownNegative, Q.DL, 0, Q.AC,
+ Q.CxtI, Q.DT);
if (LHSKnownNegative)
return getFalse(ITy);
- if (LHSKnownNonNegative && isKnownNonZero(LHS, Q.DL))
+ if (LHSKnownNonNegative &&
+ isKnownNonZero(LHS, Q.DL, 0, Q.AC, Q.CxtI, Q.DT))
return getTrue(ITy);
break;
}
Upper = Upper + 1;
assert(Upper != Lower && "Upper part of range has wrapped!");
}
+ } else if (match(LHS, m_NUWShl(m_ConstantInt(CI2), m_Value()))) {
+ // 'shl nuw CI2, x' produces [CI2, CI2 << CLZ(CI2)]
+ Lower = CI2->getValue();
+ Upper = Lower.shl(Lower.countLeadingZeros()) + 1;
+ } else if (match(LHS, m_NSWShl(m_ConstantInt(CI2), m_Value()))) {
+ if (CI2->isNegative()) {
+ // 'shl nsw CI2, x' produces [CI2 << CLO(CI2)-1, CI2]
+ unsigned ShiftAmount = CI2->getValue().countLeadingOnes() - 1;
+ Lower = CI2->getValue().shl(ShiftAmount);
+ Upper = CI2->getValue() + 1;
+ } else {
+ // 'shl nsw CI2, x' produces [CI2, CI2 << CLZ(CI2)-1]
+ unsigned ShiftAmount = CI2->getValue().countLeadingZeros() - 1;
+ Lower = CI2->getValue();
+ Upper = CI2->getValue().shl(ShiftAmount) + 1;
+ }
} else if (match(LHS, m_LShr(m_Value(), m_ConstantInt(CI2)))) {
// 'lshr x, CI2' produces [0, UINT_MAX >> CI2].
APInt NegOne = APInt::getAllOnesValue(Width);
// Turn icmp (ptrtoint x), (ptrtoint/constant) into a compare of the input
// if the integer type is the same size as the pointer type.
- if (MaxRecurse && Q.DL && isa<PtrToIntInst>(LI) &&
- Q.DL->getTypeSizeInBits(SrcTy) == DstTy->getPrimitiveSizeInBits()) {
+ if (MaxRecurse && isa<PtrToIntInst>(LI) &&
+ Q.DL.getTypeSizeInBits(SrcTy) == DstTy->getPrimitiveSizeInBits()) {
if (Constant *RHSC = dyn_cast<Constant>(RHS)) {
// Transfer the cast to the constant.
if (Value *V = SimplifyICmpInst(Pred, SrcOp,
}
}
- // If a bit is known to be zero for A and known to be one for B,
- // then A and B cannot be equal.
- if (ICmpInst::isEquality(Pred)) {
- if (ConstantInt *CI = dyn_cast<ConstantInt>(RHS)) {
- uint32_t BitWidth = CI->getBitWidth();
- APInt LHSKnownZero(BitWidth, 0);
- APInt LHSKnownOne(BitWidth, 0);
- computeKnownBits(LHS, LHSKnownZero, LHSKnownOne);
- APInt RHSKnownZero(BitWidth, 0);
- APInt RHSKnownOne(BitWidth, 0);
- computeKnownBits(RHS, RHSKnownZero, RHSKnownOne);
- if (((LHSKnownOne & RHSKnownZero) != 0) ||
- ((LHSKnownZero & RHSKnownOne) != 0))
- return (Pred == ICmpInst::ICMP_EQ)
- ? ConstantInt::getFalse(CI->getContext())
- : ConstantInt::getTrue(CI->getContext());
- }
- }
-
// Special logic for binary operators.
BinaryOperator *LBO = dyn_cast<BinaryOperator>(LHS);
BinaryOperator *RBO = dyn_cast<BinaryOperator>(RHS);
}
}
+ // icmp pred (or X, Y), X
+ if (LBO && match(LBO, m_CombineOr(m_Or(m_Value(), m_Specific(RHS)),
+ m_Or(m_Specific(RHS), m_Value())))) {
+ if (Pred == ICmpInst::ICMP_ULT)
+ return getFalse(ITy);
+ if (Pred == ICmpInst::ICMP_UGE)
+ return getTrue(ITy);
+ }
+ // icmp pred X, (or X, Y)
+ if (RBO && match(RBO, m_CombineOr(m_Or(m_Value(), m_Specific(LHS)),
+ m_Or(m_Specific(LHS), m_Value())))) {
+ if (Pred == ICmpInst::ICMP_ULE)
+ return getTrue(ITy);
+ if (Pred == ICmpInst::ICMP_UGT)
+ return getFalse(ITy);
+ }
+
+ // icmp pred (and X, Y), X
+ if (LBO && match(LBO, m_CombineOr(m_And(m_Value(), m_Specific(RHS)),
+ m_And(m_Specific(RHS), m_Value())))) {
+ if (Pred == ICmpInst::ICMP_UGT)
+ return getFalse(ITy);
+ if (Pred == ICmpInst::ICMP_ULE)
+ return getTrue(ITy);
+ }
+ // icmp pred X, (and X, Y)
+ if (RBO && match(RBO, m_CombineOr(m_And(m_Value(), m_Specific(LHS)),
+ m_And(m_Specific(LHS), m_Value())))) {
+ if (Pred == ICmpInst::ICMP_UGE)
+ return getTrue(ITy);
+ if (Pred == ICmpInst::ICMP_ULT)
+ return getFalse(ITy);
+ }
+
// 0 - (zext X) pred C
if (!CmpInst::isUnsigned(Pred) && match(LHS, m_Neg(m_ZExt(m_Value())))) {
if (ConstantInt *RHSC = dyn_cast<ConstantInt>(RHS)) {
break;
case ICmpInst::ICMP_SGT:
case ICmpInst::ICMP_SGE:
- ComputeSignBit(RHS, KnownNonNegative, KnownNegative, Q.DL);
+ ComputeSignBit(RHS, KnownNonNegative, KnownNegative, Q.DL, 0, Q.AC,
+ Q.CxtI, Q.DT);
if (!KnownNonNegative)
break;
// fall-through
return getFalse(ITy);
case ICmpInst::ICMP_SLT:
case ICmpInst::ICMP_SLE:
- ComputeSignBit(RHS, KnownNonNegative, KnownNegative, Q.DL);
+ ComputeSignBit(RHS, KnownNonNegative, KnownNegative, Q.DL, 0, Q.AC,
+ Q.CxtI, Q.DT);
if (!KnownNonNegative)
break;
// fall-through
break;
case ICmpInst::ICMP_SGT:
case ICmpInst::ICMP_SGE:
- ComputeSignBit(LHS, KnownNonNegative, KnownNegative, Q.DL);
+ ComputeSignBit(LHS, KnownNonNegative, KnownNegative, Q.DL, 0, Q.AC,
+ Q.CxtI, Q.DT);
if (!KnownNonNegative)
break;
// fall-through
return getTrue(ITy);
case ICmpInst::ICMP_SLT:
case ICmpInst::ICMP_SLE:
- ComputeSignBit(LHS, KnownNonNegative, KnownNegative, Q.DL);
+ ComputeSignBit(LHS, KnownNonNegative, KnownNegative, Q.DL, 0, Q.AC,
+ Q.CxtI, Q.DT);
if (!KnownNonNegative)
break;
// fall-through
return getTrue(ITy);
}
+ // handle:
+ // CI2 << X == CI
+ // CI2 << X != CI
+ //
+ // where CI2 is a power of 2 and CI isn't
+ if (auto *CI = dyn_cast<ConstantInt>(RHS)) {
+ const APInt *CI2Val, *CIVal = &CI->getValue();
+ if (LBO && match(LBO, m_Shl(m_APInt(CI2Val), m_Value())) &&
+ CI2Val->isPowerOf2()) {
+ if (!CIVal->isPowerOf2()) {
+ // CI2 << X can equal zero in some circumstances,
+ // this simplification is unsafe if CI is zero.
+ //
+ // We know it is safe if:
+ // - The shift is nsw, we can't shift out the one bit.
+ // - The shift is nuw, we can't shift out the one bit.
+ // - CI2 is one
+ // - CI isn't zero
+ if (LBO->hasNoSignedWrap() || LBO->hasNoUnsignedWrap() ||
+ *CI2Val == 1 || !CI->isZero()) {
+ if (Pred == ICmpInst::ICMP_EQ)
+ return ConstantInt::getFalse(RHS->getContext());
+ if (Pred == ICmpInst::ICMP_NE)
+ return ConstantInt::getTrue(RHS->getContext());
+ }
+ }
+ if (CIVal->isSignBit() && *CI2Val == 1) {
+ if (Pred == ICmpInst::ICMP_UGT)
+ return ConstantInt::getFalse(RHS->getContext());
+ if (Pred == ICmpInst::ICMP_ULE)
+ return ConstantInt::getTrue(RHS->getContext());
+ }
+ }
+ }
+
if (MaxRecurse && LBO && RBO && LBO->getOpcode() == RBO->getOpcode() &&
LBO->getOperand(1) == RBO->getOperand(1)) {
switch (LBO->getOpcode()) {
// what constant folding can make out of it.
Constant *Null = Constant::getNullValue(GLHS->getPointerOperandType());
SmallVector<Value *, 4> IndicesLHS(GLHS->idx_begin(), GLHS->idx_end());
- Constant *NewLHS = ConstantExpr::getGetElementPtr(Null, IndicesLHS);
+ Constant *NewLHS = ConstantExpr::getGetElementPtr(
+ GLHS->getSourceElementType(), Null, IndicesLHS);
SmallVector<Value *, 4> IndicesRHS(GRHS->idx_begin(), GRHS->idx_end());
- Constant *NewRHS = ConstantExpr::getGetElementPtr(Null, IndicesRHS);
+ Constant *NewRHS = ConstantExpr::getGetElementPtr(
+ GLHS->getSourceElementType(), Null, IndicesRHS);
return ConstantExpr::getICmp(Pred, NewLHS, NewRHS);
}
}
}
+ // If a bit is known to be zero for A and known to be one for B,
+ // then A and B cannot be equal.
+ if (ICmpInst::isEquality(Pred)) {
+ if (ConstantInt *CI = dyn_cast<ConstantInt>(RHS)) {
+ uint32_t BitWidth = CI->getBitWidth();
+ APInt LHSKnownZero(BitWidth, 0);
+ APInt LHSKnownOne(BitWidth, 0);
+ computeKnownBits(LHS, LHSKnownZero, LHSKnownOne, Q.DL, /*Depth=*/0, Q.AC,
+ Q.CxtI, Q.DT);
+ const APInt &RHSVal = CI->getValue();
+ if (((LHSKnownZero & RHSVal) != 0) || ((LHSKnownOne & ~RHSVal) != 0))
+ return Pred == ICmpInst::ICMP_EQ
+ ? ConstantInt::getFalse(CI->getContext())
+ : ConstantInt::getTrue(CI->getContext());
+ }
+ }
+
// If the comparison is with the result of a select instruction, check whether
// comparing with either branch of the select always yields the same value.
if (isa<SelectInst>(LHS) || isa<SelectInst>(RHS))
}
Value *llvm::SimplifyICmpInst(unsigned Predicate, Value *LHS, Value *RHS,
- const DataLayout *DL,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifyICmpInst(Predicate, LHS, RHS, Query (DL, TLI, DT),
+ const DominatorTree *DT, AssumptionCache *AC,
+ Instruction *CxtI) {
+ return ::SimplifyICmpInst(Predicate, LHS, RHS, Query(DL, TLI, DT, AC, CxtI),
RecursionLimit);
}
/// SimplifyFCmpInst - Given operands for an FCmpInst, see if we can
/// fold the result. If not, this returns null.
static Value *SimplifyFCmpInst(unsigned Predicate, Value *LHS, Value *RHS,
- const Query &Q, unsigned MaxRecurse) {
+ FastMathFlags FMF, const Query &Q,
+ unsigned MaxRecurse) {
CmpInst::Predicate Pred = (CmpInst::Predicate)Predicate;
assert(CmpInst::isFPPredicate(Pred) && "Not an FP compare!");
if (Pred == FCmpInst::FCMP_TRUE)
return ConstantInt::get(GetCompareTy(LHS), 1);
- if (isa<UndefValue>(RHS)) // fcmp pred X, undef -> undef
- return UndefValue::get(GetCompareTy(LHS));
+ // UNO/ORD predicates can be trivially folded if NaNs are ignored.
+ if (FMF.noNaNs()) {
+ if (Pred == FCmpInst::FCMP_UNO)
+ return ConstantInt::get(GetCompareTy(LHS), 0);
+ if (Pred == FCmpInst::FCMP_ORD)
+ return ConstantInt::get(GetCompareTy(LHS), 1);
+ }
+
+ // fcmp pred x, undef and fcmp pred undef, x
+ // fold to true if unordered, false if ordered
+ if (isa<UndefValue>(LHS) || isa<UndefValue>(RHS)) {
+ // Choosing NaN for the undef will always make unordered comparison succeed
+ // and ordered comparison fail.
+ return ConstantInt::get(GetCompareTy(LHS), CmpInst::isUnordered(Pred));
+ }
// fcmp x,x -> true/false. Not all compares are foldable.
if (LHS == RHS) {
}
// Handle fcmp with constant RHS
- if (Constant *RHSC = dyn_cast<Constant>(RHS)) {
+ if (ConstantFP *CFP = dyn_cast<ConstantFP>(RHS)) {
// If the constant is a nan, see if we can fold the comparison based on it.
- if (ConstantFP *CFP = dyn_cast<ConstantFP>(RHSC)) {
- if (CFP->getValueAPF().isNaN()) {
- if (FCmpInst::isOrdered(Pred)) // True "if ordered and foo"
+ if (CFP->getValueAPF().isNaN()) {
+ if (FCmpInst::isOrdered(Pred)) // True "if ordered and foo"
+ return ConstantInt::getFalse(CFP->getContext());
+ assert(FCmpInst::isUnordered(Pred) &&
+ "Comparison must be either ordered or unordered!");
+ // True if unordered.
+ return ConstantInt::getTrue(CFP->getContext());
+ }
+ // Check whether the constant is an infinity.
+ if (CFP->getValueAPF().isInfinity()) {
+ if (CFP->getValueAPF().isNegative()) {
+ switch (Pred) {
+ case FCmpInst::FCMP_OLT:
+ // No value is ordered and less than negative infinity.
return ConstantInt::getFalse(CFP->getContext());
- assert(FCmpInst::isUnordered(Pred) &&
- "Comparison must be either ordered or unordered!");
- // True if unordered.
- return ConstantInt::getTrue(CFP->getContext());
- }
- // Check whether the constant is an infinity.
- if (CFP->getValueAPF().isInfinity()) {
- if (CFP->getValueAPF().isNegative()) {
- switch (Pred) {
- case FCmpInst::FCMP_OLT:
- // No value is ordered and less than negative infinity.
- return ConstantInt::getFalse(CFP->getContext());
- case FCmpInst::FCMP_UGE:
- // All values are unordered with or at least negative infinity.
- return ConstantInt::getTrue(CFP->getContext());
- default:
- break;
- }
- } else {
- switch (Pred) {
- case FCmpInst::FCMP_OGT:
- // No value is ordered and greater than infinity.
- return ConstantInt::getFalse(CFP->getContext());
- case FCmpInst::FCMP_ULE:
- // All values are unordered with and at most infinity.
- return ConstantInt::getTrue(CFP->getContext());
- default:
- break;
- }
+ case FCmpInst::FCMP_UGE:
+ // All values are unordered with or at least negative infinity.
+ return ConstantInt::getTrue(CFP->getContext());
+ default:
+ break;
+ }
+ } else {
+ switch (Pred) {
+ case FCmpInst::FCMP_OGT:
+ // No value is ordered and greater than infinity.
+ return ConstantInt::getFalse(CFP->getContext());
+ case FCmpInst::FCMP_ULE:
+ // All values are unordered with and at most infinity.
+ return ConstantInt::getTrue(CFP->getContext());
+ default:
+ break;
}
}
}
+ if (CFP->getValueAPF().isZero()) {
+ switch (Pred) {
+ case FCmpInst::FCMP_UGE:
+ if (CannotBeOrderedLessThanZero(LHS))
+ return ConstantInt::getTrue(CFP->getContext());
+ break;
+ case FCmpInst::FCMP_OLT:
+ // X < 0
+ if (CannotBeOrderedLessThanZero(LHS))
+ return ConstantInt::getFalse(CFP->getContext());
+ break;
+ default:
+ break;
+ }
+ }
}
// If the comparison is with the result of a select instruction, check whether
}
Value *llvm::SimplifyFCmpInst(unsigned Predicate, Value *LHS, Value *RHS,
- const DataLayout *DL,
+ FastMathFlags FMF, const DataLayout &DL,
const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifyFCmpInst(Predicate, LHS, RHS, Query (DL, TLI, DT),
- RecursionLimit);
+ const DominatorTree *DT, AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifyFCmpInst(Predicate, LHS, RHS, FMF,
+ Query(DL, TLI, DT, AC, CxtI), RecursionLimit);
+}
+
+/// SimplifyWithOpReplaced - See if V simplifies when its operand Op is
+/// replaced with RepOp.
+static const Value *SimplifyWithOpReplaced(Value *V, Value *Op, Value *RepOp,
+ const Query &Q,
+ unsigned MaxRecurse) {
+ // Trivial replacement.
+ if (V == Op)
+ return RepOp;
+
+ auto *I = dyn_cast<Instruction>(V);
+ if (!I)
+ return nullptr;
+
+ // If this is a binary operator, try to simplify it with the replaced op.
+ if (auto *B = dyn_cast<BinaryOperator>(I)) {
+ // Consider:
+ // %cmp = icmp eq i32 %x, 2147483647
+ // %add = add nsw i32 %x, 1
+ // %sel = select i1 %cmp, i32 -2147483648, i32 %add
+ //
+ // We can't replace %sel with %add unless we strip away the flags.
+ if (isa<OverflowingBinaryOperator>(B))
+ if (B->hasNoSignedWrap() || B->hasNoUnsignedWrap())
+ return nullptr;
+ if (isa<PossiblyExactOperator>(B))
+ if (B->isExact())
+ return nullptr;
+
+ if (MaxRecurse) {
+ if (B->getOperand(0) == Op)
+ return SimplifyBinOp(B->getOpcode(), RepOp, B->getOperand(1), Q,
+ MaxRecurse - 1);
+ if (B->getOperand(1) == Op)
+ return SimplifyBinOp(B->getOpcode(), B->getOperand(0), RepOp, Q,
+ MaxRecurse - 1);
+ }
+ }
+
+ // Same for CmpInsts.
+ if (CmpInst *C = dyn_cast<CmpInst>(I)) {
+ if (MaxRecurse) {
+ if (C->getOperand(0) == Op)
+ return SimplifyCmpInst(C->getPredicate(), RepOp, C->getOperand(1), Q,
+ MaxRecurse - 1);
+ if (C->getOperand(1) == Op)
+ return SimplifyCmpInst(C->getPredicate(), C->getOperand(0), RepOp, Q,
+ MaxRecurse - 1);
+ }
+ }
+
+ // TODO: We could hand off more cases to instsimplify here.
+
+ // If all operands are constant after substituting Op for RepOp then we can
+ // constant fold the instruction.
+ if (Constant *CRepOp = dyn_cast<Constant>(RepOp)) {
+ // Build a list of all constant operands.
+ SmallVector<Constant *, 8> ConstOps;
+ for (unsigned i = 0, e = I->getNumOperands(); i != e; ++i) {
+ if (I->getOperand(i) == Op)
+ ConstOps.push_back(CRepOp);
+ else if (Constant *COp = dyn_cast<Constant>(I->getOperand(i)))
+ ConstOps.push_back(COp);
+ else
+ break;
+ }
+
+ // All operands were constants, fold it.
+ if (ConstOps.size() == I->getNumOperands()) {
+ if (CmpInst *C = dyn_cast<CmpInst>(I))
+ return ConstantFoldCompareInstOperands(C->getPredicate(), ConstOps[0],
+ ConstOps[1], Q.DL, Q.TLI);
+
+ if (LoadInst *LI = dyn_cast<LoadInst>(I))
+ if (!LI->isVolatile())
+ return ConstantFoldLoadFromConstPtr(ConstOps[0], Q.DL);
+
+ return ConstantFoldInstOperands(I->getOpcode(), I->getType(), ConstOps,
+ Q.DL, Q.TLI);
+ }
+ }
+
+ return nullptr;
}
/// SimplifySelectInst - Given operands for a SelectInst, see if we can fold
if (isa<UndefValue>(FalseVal)) // select C, X, undef -> X
return TrueVal;
+ if (const auto *ICI = dyn_cast<ICmpInst>(CondVal)) {
+ unsigned BitWidth = Q.DL.getTypeSizeInBits(TrueVal->getType());
+ ICmpInst::Predicate Pred = ICI->getPredicate();
+ Value *CmpLHS = ICI->getOperand(0);
+ Value *CmpRHS = ICI->getOperand(1);
+ APInt MinSignedValue = APInt::getSignBit(BitWidth);
+ Value *X;
+ const APInt *Y;
+ bool TrueWhenUnset;
+ bool IsBitTest = false;
+ if (ICmpInst::isEquality(Pred) &&
+ match(CmpLHS, m_And(m_Value(X), m_APInt(Y))) &&
+ match(CmpRHS, m_Zero())) {
+ IsBitTest = true;
+ TrueWhenUnset = Pred == ICmpInst::ICMP_EQ;
+ } else if (Pred == ICmpInst::ICMP_SLT && match(CmpRHS, m_Zero())) {
+ X = CmpLHS;
+ Y = &MinSignedValue;
+ IsBitTest = true;
+ TrueWhenUnset = false;
+ } else if (Pred == ICmpInst::ICMP_SGT && match(CmpRHS, m_AllOnes())) {
+ X = CmpLHS;
+ Y = &MinSignedValue;
+ IsBitTest = true;
+ TrueWhenUnset = true;
+ }
+ if (IsBitTest) {
+ const APInt *C;
+ // (X & Y) == 0 ? X & ~Y : X --> X
+ // (X & Y) != 0 ? X & ~Y : X --> X & ~Y
+ if (FalseVal == X && match(TrueVal, m_And(m_Specific(X), m_APInt(C))) &&
+ *Y == ~*C)
+ return TrueWhenUnset ? FalseVal : TrueVal;
+ // (X & Y) == 0 ? X : X & ~Y --> X & ~Y
+ // (X & Y) != 0 ? X : X & ~Y --> X
+ if (TrueVal == X && match(FalseVal, m_And(m_Specific(X), m_APInt(C))) &&
+ *Y == ~*C)
+ return TrueWhenUnset ? FalseVal : TrueVal;
+
+ if (Y->isPowerOf2()) {
+ // (X & Y) == 0 ? X | Y : X --> X | Y
+ // (X & Y) != 0 ? X | Y : X --> X
+ if (FalseVal == X && match(TrueVal, m_Or(m_Specific(X), m_APInt(C))) &&
+ *Y == *C)
+ return TrueWhenUnset ? TrueVal : FalseVal;
+ // (X & Y) == 0 ? X : X | Y --> X
+ // (X & Y) != 0 ? X : X | Y --> X | Y
+ if (TrueVal == X && match(FalseVal, m_Or(m_Specific(X), m_APInt(C))) &&
+ *Y == *C)
+ return TrueWhenUnset ? TrueVal : FalseVal;
+ }
+ }
+ if (ICI->hasOneUse()) {
+ const APInt *C;
+ if (match(CmpRHS, m_APInt(C))) {
+ // X < MIN ? T : F --> F
+ if (Pred == ICmpInst::ICMP_SLT && C->isMinSignedValue())
+ return FalseVal;
+ // X < MIN ? T : F --> F
+ if (Pred == ICmpInst::ICMP_ULT && C->isMinValue())
+ return FalseVal;
+ // X > MAX ? T : F --> F
+ if (Pred == ICmpInst::ICMP_SGT && C->isMaxSignedValue())
+ return FalseVal;
+ // X > MAX ? T : F --> F
+ if (Pred == ICmpInst::ICMP_UGT && C->isMaxValue())
+ return FalseVal;
+ }
+ }
+
+ // If we have an equality comparison then we know the value in one of the
+ // arms of the select. See if substituting this value into the arm and
+ // simplifying the result yields the same value as the other arm.
+ if (Pred == ICmpInst::ICMP_EQ) {
+ if (SimplifyWithOpReplaced(FalseVal, CmpLHS, CmpRHS, Q, MaxRecurse) ==
+ TrueVal ||
+ SimplifyWithOpReplaced(FalseVal, CmpRHS, CmpLHS, Q, MaxRecurse) ==
+ TrueVal)
+ return FalseVal;
+ if (SimplifyWithOpReplaced(TrueVal, CmpLHS, CmpRHS, Q, MaxRecurse) ==
+ FalseVal ||
+ SimplifyWithOpReplaced(TrueVal, CmpRHS, CmpLHS, Q, MaxRecurse) ==
+ FalseVal)
+ return FalseVal;
+ } else if (Pred == ICmpInst::ICMP_NE) {
+ if (SimplifyWithOpReplaced(TrueVal, CmpLHS, CmpRHS, Q, MaxRecurse) ==
+ FalseVal ||
+ SimplifyWithOpReplaced(TrueVal, CmpRHS, CmpLHS, Q, MaxRecurse) ==
+ FalseVal)
+ return TrueVal;
+ if (SimplifyWithOpReplaced(FalseVal, CmpLHS, CmpRHS, Q, MaxRecurse) ==
+ TrueVal ||
+ SimplifyWithOpReplaced(FalseVal, CmpRHS, CmpLHS, Q, MaxRecurse) ==
+ TrueVal)
+ return TrueVal;
+ }
+ }
+
return nullptr;
}
Value *llvm::SimplifySelectInst(Value *Cond, Value *TrueVal, Value *FalseVal,
- const DataLayout *DL,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifySelectInst(Cond, TrueVal, FalseVal, Query (DL, TLI, DT),
- RecursionLimit);
+ const DominatorTree *DT, AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifySelectInst(Cond, TrueVal, FalseVal,
+ Query(DL, TLI, DT, AC, CxtI), RecursionLimit);
}
/// SimplifyGEPInst - Given operands for an GetElementPtrInst, see if we can
/// fold the result. If not, this returns null.
-static Value *SimplifyGEPInst(ArrayRef<Value *> Ops, const Query &Q, unsigned) {
+static Value *SimplifyGEPInst(Type *SrcTy, ArrayRef<Value *> Ops,
+ const Query &Q, unsigned) {
// The type of the GEP pointer operand.
- PointerType *PtrTy = cast<PointerType>(Ops[0]->getType()->getScalarType());
+ unsigned AS =
+ cast<PointerType>(Ops[0]->getType()->getScalarType())->getAddressSpace();
// getelementptr P -> P.
if (Ops.size() == 1)
return Ops[0];
- if (isa<UndefValue>(Ops[0])) {
- // Compute the (pointer) type returned by the GEP instruction.
- Type *LastType = GetElementPtrInst::getIndexedType(PtrTy, Ops.slice(1));
- Type *GEPTy = PointerType::get(LastType, PtrTy->getAddressSpace());
- if (VectorType *VT = dyn_cast<VectorType>(Ops[0]->getType()))
- GEPTy = VectorType::get(GEPTy, VT->getNumElements());
+ // Compute the (pointer) type returned by the GEP instruction.
+ Type *LastType = GetElementPtrInst::getIndexedType(SrcTy, Ops.slice(1));
+ Type *GEPTy = PointerType::get(LastType, AS);
+ if (VectorType *VT = dyn_cast<VectorType>(Ops[0]->getType()))
+ GEPTy = VectorType::get(GEPTy, VT->getNumElements());
+
+ if (isa<UndefValue>(Ops[0]))
return UndefValue::get(GEPTy);
- }
if (Ops.size() == 2) {
// getelementptr P, 0 -> P.
if (match(Ops[1], m_Zero()))
return Ops[0];
- // getelementptr P, N -> P if P points to a type of zero size.
- if (Q.DL) {
- Type *Ty = PtrTy->getElementType();
- if (Ty->isSized() && Q.DL->getTypeAllocSize(Ty) == 0)
+
+ Type *Ty = SrcTy;
+ if (Ty->isSized()) {
+ Value *P;
+ uint64_t C;
+ uint64_t TyAllocSize = Q.DL.getTypeAllocSize(Ty);
+ // getelementptr P, N -> P if P points to a type of zero size.
+ if (TyAllocSize == 0)
return Ops[0];
+
+ // The following transforms are only safe if the ptrtoint cast
+ // doesn't truncate the pointers.
+ if (Ops[1]->getType()->getScalarSizeInBits() ==
+ Q.DL.getPointerSizeInBits(AS)) {
+ auto PtrToIntOrZero = [GEPTy](Value *P) -> Value * {
+ if (match(P, m_Zero()))
+ return Constant::getNullValue(GEPTy);
+ Value *Temp;
+ if (match(P, m_PtrToInt(m_Value(Temp))))
+ if (Temp->getType() == GEPTy)
+ return Temp;
+ return nullptr;
+ };
+
+ // getelementptr V, (sub P, V) -> P if P points to a type of size 1.
+ if (TyAllocSize == 1 &&
+ match(Ops[1], m_Sub(m_Value(P), m_PtrToInt(m_Specific(Ops[0])))))
+ if (Value *R = PtrToIntOrZero(P))
+ return R;
+
+ // getelementptr V, (ashr (sub P, V), C) -> Q
+ // if P points to a type of size 1 << C.
+ if (match(Ops[1],
+ m_AShr(m_Sub(m_Value(P), m_PtrToInt(m_Specific(Ops[0]))),
+ m_ConstantInt(C))) &&
+ TyAllocSize == 1ULL << C)
+ if (Value *R = PtrToIntOrZero(P))
+ return R;
+
+ // getelementptr V, (sdiv (sub P, V), C) -> Q
+ // if P points to a type of size C.
+ if (match(Ops[1],
+ m_SDiv(m_Sub(m_Value(P), m_PtrToInt(m_Specific(Ops[0]))),
+ m_SpecificInt(TyAllocSize))))
+ if (Value *R = PtrToIntOrZero(P))
+ return R;
+ }
}
}
if (!isa<Constant>(Ops[i]))
return nullptr;
- return ConstantExpr::getGetElementPtr(cast<Constant>(Ops[0]), Ops.slice(1));
+ return ConstantExpr::getGetElementPtr(SrcTy, cast<Constant>(Ops[0]),
+ Ops.slice(1));
}
-Value *llvm::SimplifyGEPInst(ArrayRef<Value *> Ops, const DataLayout *DL,
+Value *llvm::SimplifyGEPInst(ArrayRef<Value *> Ops, const DataLayout &DL,
const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifyGEPInst(Ops, Query (DL, TLI, DT), RecursionLimit);
+ const DominatorTree *DT, AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifyGEPInst(
+ cast<PointerType>(Ops[0]->getType()->getScalarType())->getElementType(),
+ Ops, Query(DL, TLI, DT, AC, CxtI), RecursionLimit);
}
/// SimplifyInsertValueInst - Given operands for an InsertValueInst, see if we
return nullptr;
}
-Value *llvm::SimplifyInsertValueInst(Value *Agg, Value *Val,
- ArrayRef<unsigned> Idxs,
- const DataLayout *DL,
- const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifyInsertValueInst(Agg, Val, Idxs, Query (DL, TLI, DT),
+Value *llvm::SimplifyInsertValueInst(
+ Value *Agg, Value *Val, ArrayRef<unsigned> Idxs, const DataLayout &DL,
+ const TargetLibraryInfo *TLI, const DominatorTree *DT, AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifyInsertValueInst(Agg, Val, Idxs, Query(DL, TLI, DT, AC, CxtI),
RecursionLimit);
}
+/// SimplifyExtractValueInst - Given operands for an ExtractValueInst, see if we
+/// can fold the result. If not, this returns null.
+static Value *SimplifyExtractValueInst(Value *Agg, ArrayRef<unsigned> Idxs,
+ const Query &, unsigned) {
+ if (auto *CAgg = dyn_cast<Constant>(Agg))
+ return ConstantFoldExtractValueInstruction(CAgg, Idxs);
+
+ // extractvalue x, (insertvalue y, elt, n), n -> elt
+ unsigned NumIdxs = Idxs.size();
+ for (auto *IVI = dyn_cast<InsertValueInst>(Agg); IVI != nullptr;
+ IVI = dyn_cast<InsertValueInst>(IVI->getAggregateOperand())) {
+ ArrayRef<unsigned> InsertValueIdxs = IVI->getIndices();
+ unsigned NumInsertValueIdxs = InsertValueIdxs.size();
+ unsigned NumCommonIdxs = std::min(NumInsertValueIdxs, NumIdxs);
+ if (InsertValueIdxs.slice(0, NumCommonIdxs) ==
+ Idxs.slice(0, NumCommonIdxs)) {
+ if (NumIdxs == NumInsertValueIdxs)
+ return IVI->getInsertedValueOperand();
+ break;
+ }
+ }
+
+ return nullptr;
+}
+
+Value *llvm::SimplifyExtractValueInst(Value *Agg, ArrayRef<unsigned> Idxs,
+ const DataLayout &DL,
+ const TargetLibraryInfo *TLI,
+ const DominatorTree *DT,
+ AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifyExtractValueInst(Agg, Idxs, Query(DL, TLI, DT, AC, CxtI),
+ RecursionLimit);
+}
+
+/// SimplifyExtractElementInst - Given operands for an ExtractElementInst, see if we
+/// can fold the result. If not, this returns null.
+static Value *SimplifyExtractElementInst(Value *Vec, Value *Idx, const Query &,
+ unsigned) {
+ if (auto *CVec = dyn_cast<Constant>(Vec)) {
+ if (auto *CIdx = dyn_cast<Constant>(Idx))
+ return ConstantFoldExtractElementInstruction(CVec, CIdx);
+
+ // The index is not relevant if our vector is a splat.
+ if (auto *Splat = CVec->getSplatValue())
+ return Splat;
+
+ if (isa<UndefValue>(Vec))
+ return UndefValue::get(Vec->getType()->getVectorElementType());
+ }
+
+ // If extracting a specified index from the vector, see if we can recursively
+ // find a previously computed scalar that was inserted into the vector.
+ if (auto *IdxC = dyn_cast<ConstantInt>(Idx)) {
+ unsigned IndexVal = IdxC->getZExtValue();
+ unsigned VectorWidth = Vec->getType()->getVectorNumElements();
+
+ // If this is extracting an invalid index, turn this into undef, to avoid
+ // crashing the code below.
+ if (IndexVal >= VectorWidth)
+ return UndefValue::get(Vec->getType()->getVectorElementType());
+
+ if (Value *Elt = findScalarElement(Vec, IndexVal))
+ return Elt;
+ }
+
+ return nullptr;
+}
+
+Value *llvm::SimplifyExtractElementInst(
+ Value *Vec, Value *Idx, const DataLayout &DL, const TargetLibraryInfo *TLI,
+ const DominatorTree *DT, AssumptionCache *AC, const Instruction *CxtI) {
+ return ::SimplifyExtractElementInst(Vec, Idx, Query(DL, TLI, DT, AC, CxtI),
+ RecursionLimit);
+}
+
/// SimplifyPHINode - See if we can fold the given phi. If not, returns null.
static Value *SimplifyPHINode(PHINode *PN, const Query &Q) {
// If all of the PHI's incoming values are the same then replace the PHI node
// with the common value.
Value *CommonValue = nullptr;
bool HasUndefInput = false;
- for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i) {
- Value *Incoming = PN->getIncomingValue(i);
+ for (Value *Incoming : PN->incoming_values()) {
// If the incoming value is the phi node itself, it can safely be skipped.
if (Incoming == PN) continue;
if (isa<UndefValue>(Incoming)) {
return nullptr;
}
-Value *llvm::SimplifyTruncInst(Value *Op, Type *Ty, const DataLayout *DL,
+Value *llvm::SimplifyTruncInst(Value *Op, Type *Ty, const DataLayout &DL,
const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifyTruncInst(Op, Ty, Query (DL, TLI, DT), RecursionLimit);
+ const DominatorTree *DT, AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifyTruncInst(Op, Ty, Query(DL, TLI, DT, AC, CxtI),
+ RecursionLimit);
}
//=== Helper functions for higher up the class hierarchy.
return SimplifyFMulInst (LHS, RHS, FastMathFlags(), Q, MaxRecurse);
case Instruction::SDiv: return SimplifySDivInst(LHS, RHS, Q, MaxRecurse);
case Instruction::UDiv: return SimplifyUDivInst(LHS, RHS, Q, MaxRecurse);
- case Instruction::FDiv: return SimplifyFDivInst(LHS, RHS, Q, MaxRecurse);
+ case Instruction::FDiv:
+ return SimplifyFDivInst(LHS, RHS, FastMathFlags(), Q, MaxRecurse);
case Instruction::SRem: return SimplifySRemInst(LHS, RHS, Q, MaxRecurse);
case Instruction::URem: return SimplifyURemInst(LHS, RHS, Q, MaxRecurse);
- case Instruction::FRem: return SimplifyFRemInst(LHS, RHS, Q, MaxRecurse);
+ case Instruction::FRem:
+ return SimplifyFRemInst(LHS, RHS, FastMathFlags(), Q, MaxRecurse);
case Instruction::Shl:
return SimplifyShlInst(LHS, RHS, /*isNSW*/false, /*isNUW*/false,
Q, MaxRecurse);
}
}
+/// SimplifyFPBinOp - Given operands for a BinaryOperator, see if we can
+/// fold the result. If not, this returns null.
+/// In contrast to SimplifyBinOp, try to use FastMathFlag when folding the
+/// result. In case we don't need FastMathFlags, simply fall to SimplifyBinOp.
+static Value *SimplifyFPBinOp(unsigned Opcode, Value *LHS, Value *RHS,
+ const FastMathFlags &FMF, const Query &Q,
+ unsigned MaxRecurse) {
+ switch (Opcode) {
+ case Instruction::FAdd:
+ return SimplifyFAddInst(LHS, RHS, FMF, Q, MaxRecurse);
+ case Instruction::FSub:
+ return SimplifyFSubInst(LHS, RHS, FMF, Q, MaxRecurse);
+ case Instruction::FMul:
+ return SimplifyFMulInst(LHS, RHS, FMF, Q, MaxRecurse);
+ default:
+ return SimplifyBinOp(Opcode, LHS, RHS, Q, MaxRecurse);
+ }
+}
+
Value *llvm::SimplifyBinOp(unsigned Opcode, Value *LHS, Value *RHS,
- const DataLayout *DL, const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifyBinOp(Opcode, LHS, RHS, Query (DL, TLI, DT), RecursionLimit);
+ const DataLayout &DL, const TargetLibraryInfo *TLI,
+ const DominatorTree *DT, AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifyBinOp(Opcode, LHS, RHS, Query(DL, TLI, DT, AC, CxtI),
+ RecursionLimit);
+}
+
+Value *llvm::SimplifyFPBinOp(unsigned Opcode, Value *LHS, Value *RHS,
+ const FastMathFlags &FMF, const DataLayout &DL,
+ const TargetLibraryInfo *TLI,
+ const DominatorTree *DT, AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifyFPBinOp(Opcode, LHS, RHS, FMF, Query(DL, TLI, DT, AC, CxtI),
+ RecursionLimit);
}
/// SimplifyCmpInst - Given operands for a CmpInst, see if we can
const Query &Q, unsigned MaxRecurse) {
if (CmpInst::isIntPredicate((CmpInst::Predicate)Predicate))
return SimplifyICmpInst(Predicate, LHS, RHS, Q, MaxRecurse);
- return SimplifyFCmpInst(Predicate, LHS, RHS, Q, MaxRecurse);
+ return SimplifyFCmpInst(Predicate, LHS, RHS, FastMathFlags(), Q, MaxRecurse);
}
Value *llvm::SimplifyCmpInst(unsigned Predicate, Value *LHS, Value *RHS,
- const DataLayout *DL, const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifyCmpInst(Predicate, LHS, RHS, Query (DL, TLI, DT),
+ const DataLayout &DL, const TargetLibraryInfo *TLI,
+ const DominatorTree *DT, AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifyCmpInst(Predicate, LHS, RHS, Query(DL, TLI, DT, AC, CxtI),
RecursionLimit);
}
}
template <typename IterTy>
-static Value *SimplifyIntrinsic(Intrinsic::ID IID, IterTy ArgBegin, IterTy ArgEnd,
+static Value *SimplifyIntrinsic(Function *F, IterTy ArgBegin, IterTy ArgEnd,
const Query &Q, unsigned MaxRecurse) {
+ Intrinsic::ID IID = F->getIntrinsicID();
+ unsigned NumOperands = std::distance(ArgBegin, ArgEnd);
+ Type *ReturnType = F->getReturnType();
+
+ // Binary Ops
+ if (NumOperands == 2) {
+ Value *LHS = *ArgBegin;
+ Value *RHS = *(ArgBegin + 1);
+ if (IID == Intrinsic::usub_with_overflow ||
+ IID == Intrinsic::ssub_with_overflow) {
+ // X - X -> { 0, false }
+ if (LHS == RHS)
+ return Constant::getNullValue(ReturnType);
+
+ // X - undef -> undef
+ // undef - X -> undef
+ if (isa<UndefValue>(LHS) || isa<UndefValue>(RHS))
+ return UndefValue::get(ReturnType);
+ }
+
+ if (IID == Intrinsic::uadd_with_overflow ||
+ IID == Intrinsic::sadd_with_overflow) {
+ // X + undef -> undef
+ if (isa<UndefValue>(RHS))
+ return UndefValue::get(ReturnType);
+ }
+
+ if (IID == Intrinsic::umul_with_overflow ||
+ IID == Intrinsic::smul_with_overflow) {
+ // X * 0 -> { 0, false }
+ if (match(RHS, m_Zero()))
+ return Constant::getNullValue(ReturnType);
+
+ // X * undef -> { 0, false }
+ if (match(RHS, m_Undef()))
+ return Constant::getNullValue(ReturnType);
+ }
+ }
+
// Perform idempotent optimizations
if (!IsIdempotent(IID))
return nullptr;
// Unary Ops
- if (std::distance(ArgBegin, ArgEnd) == 1)
+ if (NumOperands == 1)
if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(*ArgBegin))
if (II->getIntrinsicID() == IID)
return II;
if (!F)
return nullptr;
- if (unsigned IID = F->getIntrinsicID())
- if (Value *Ret =
- SimplifyIntrinsic((Intrinsic::ID) IID, ArgBegin, ArgEnd, Q, MaxRecurse))
+ if (F->isIntrinsic())
+ if (Value *Ret = SimplifyIntrinsic(F, ArgBegin, ArgEnd, Q, MaxRecurse))
return Ret;
if (!canConstantFoldCallTo(F))
}
Value *llvm::SimplifyCall(Value *V, User::op_iterator ArgBegin,
- User::op_iterator ArgEnd, const DataLayout *DL,
- const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifyCall(V, ArgBegin, ArgEnd, Query(DL, TLI, DT),
+ User::op_iterator ArgEnd, const DataLayout &DL,
+ const TargetLibraryInfo *TLI, const DominatorTree *DT,
+ AssumptionCache *AC, const Instruction *CxtI) {
+ return ::SimplifyCall(V, ArgBegin, ArgEnd, Query(DL, TLI, DT, AC, CxtI),
RecursionLimit);
}
Value *llvm::SimplifyCall(Value *V, ArrayRef<Value *> Args,
- const DataLayout *DL, const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return ::SimplifyCall(V, Args.begin(), Args.end(), Query(DL, TLI, DT),
- RecursionLimit);
+ const DataLayout &DL, const TargetLibraryInfo *TLI,
+ const DominatorTree *DT, AssumptionCache *AC,
+ const Instruction *CxtI) {
+ return ::SimplifyCall(V, Args.begin(), Args.end(),
+ Query(DL, TLI, DT, AC, CxtI), RecursionLimit);
}
/// SimplifyInstruction - See if we can compute a simplified version of this
/// instruction. If not, this returns null.
-Value *llvm::SimplifyInstruction(Instruction *I, const DataLayout *DL,
+Value *llvm::SimplifyInstruction(Instruction *I, const DataLayout &DL,
const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
+ const DominatorTree *DT, AssumptionCache *AC) {
Value *Result;
switch (I->getOpcode()) {
break;
case Instruction::FAdd:
Result = SimplifyFAddInst(I->getOperand(0), I->getOperand(1),
- I->getFastMathFlags(), DL, TLI, DT);
+ I->getFastMathFlags(), DL, TLI, DT, AC, I);
break;
case Instruction::Add:
Result = SimplifyAddInst(I->getOperand(0), I->getOperand(1),
cast<BinaryOperator>(I)->hasNoSignedWrap(),
- cast<BinaryOperator>(I)->hasNoUnsignedWrap(),
- DL, TLI, DT);
+ cast<BinaryOperator>(I)->hasNoUnsignedWrap(), DL,
+ TLI, DT, AC, I);
break;
case Instruction::FSub:
Result = SimplifyFSubInst(I->getOperand(0), I->getOperand(1),
- I->getFastMathFlags(), DL, TLI, DT);
+ I->getFastMathFlags(), DL, TLI, DT, AC, I);
break;
case Instruction::Sub:
Result = SimplifySubInst(I->getOperand(0), I->getOperand(1),
cast<BinaryOperator>(I)->hasNoSignedWrap(),
- cast<BinaryOperator>(I)->hasNoUnsignedWrap(),
- DL, TLI, DT);
+ cast<BinaryOperator>(I)->hasNoUnsignedWrap(), DL,
+ TLI, DT, AC, I);
break;
case Instruction::FMul:
Result = SimplifyFMulInst(I->getOperand(0), I->getOperand(1),
- I->getFastMathFlags(), DL, TLI, DT);
+ I->getFastMathFlags(), DL, TLI, DT, AC, I);
break;
case Instruction::Mul:
- Result = SimplifyMulInst(I->getOperand(0), I->getOperand(1), DL, TLI, DT);
+ Result =
+ SimplifyMulInst(I->getOperand(0), I->getOperand(1), DL, TLI, DT, AC, I);
break;
case Instruction::SDiv:
- Result = SimplifySDivInst(I->getOperand(0), I->getOperand(1), DL, TLI, DT);
+ Result = SimplifySDivInst(I->getOperand(0), I->getOperand(1), DL, TLI, DT,
+ AC, I);
break;
case Instruction::UDiv:
- Result = SimplifyUDivInst(I->getOperand(0), I->getOperand(1), DL, TLI, DT);
+ Result = SimplifyUDivInst(I->getOperand(0), I->getOperand(1), DL, TLI, DT,
+ AC, I);
break;
case Instruction::FDiv:
- Result = SimplifyFDivInst(I->getOperand(0), I->getOperand(1), DL, TLI, DT);
+ Result = SimplifyFDivInst(I->getOperand(0), I->getOperand(1),
+ I->getFastMathFlags(), DL, TLI, DT, AC, I);
break;
case Instruction::SRem:
- Result = SimplifySRemInst(I->getOperand(0), I->getOperand(1), DL, TLI, DT);
+ Result = SimplifySRemInst(I->getOperand(0), I->getOperand(1), DL, TLI, DT,
+ AC, I);
break;
case Instruction::URem:
- Result = SimplifyURemInst(I->getOperand(0), I->getOperand(1), DL, TLI, DT);
+ Result = SimplifyURemInst(I->getOperand(0), I->getOperand(1), DL, TLI, DT,
+ AC, I);
break;
case Instruction::FRem:
- Result = SimplifyFRemInst(I->getOperand(0), I->getOperand(1), DL, TLI, DT);
+ Result = SimplifyFRemInst(I->getOperand(0), I->getOperand(1),
+ I->getFastMathFlags(), DL, TLI, DT, AC, I);
break;
case Instruction::Shl:
Result = SimplifyShlInst(I->getOperand(0), I->getOperand(1),
cast<BinaryOperator>(I)->hasNoSignedWrap(),
- cast<BinaryOperator>(I)->hasNoUnsignedWrap(),
- DL, TLI, DT);
+ cast<BinaryOperator>(I)->hasNoUnsignedWrap(), DL,
+ TLI, DT, AC, I);
break;
case Instruction::LShr:
Result = SimplifyLShrInst(I->getOperand(0), I->getOperand(1),
- cast<BinaryOperator>(I)->isExact(),
- DL, TLI, DT);
+ cast<BinaryOperator>(I)->isExact(), DL, TLI, DT,
+ AC, I);
break;
case Instruction::AShr:
Result = SimplifyAShrInst(I->getOperand(0), I->getOperand(1),
- cast<BinaryOperator>(I)->isExact(),
- DL, TLI, DT);
+ cast<BinaryOperator>(I)->isExact(), DL, TLI, DT,
+ AC, I);
break;
case Instruction::And:
- Result = SimplifyAndInst(I->getOperand(0), I->getOperand(1), DL, TLI, DT);
+ Result =
+ SimplifyAndInst(I->getOperand(0), I->getOperand(1), DL, TLI, DT, AC, I);
break;
case Instruction::Or:
- Result = SimplifyOrInst(I->getOperand(0), I->getOperand(1), DL, TLI, DT);
+ Result =
+ SimplifyOrInst(I->getOperand(0), I->getOperand(1), DL, TLI, DT, AC, I);
break;
case Instruction::Xor:
- Result = SimplifyXorInst(I->getOperand(0), I->getOperand(1), DL, TLI, DT);
+ Result =
+ SimplifyXorInst(I->getOperand(0), I->getOperand(1), DL, TLI, DT, AC, I);
break;
case Instruction::ICmp:
- Result = SimplifyICmpInst(cast<ICmpInst>(I)->getPredicate(),
- I->getOperand(0), I->getOperand(1), DL, TLI, DT);
+ Result =
+ SimplifyICmpInst(cast<ICmpInst>(I)->getPredicate(), I->getOperand(0),
+ I->getOperand(1), DL, TLI, DT, AC, I);
break;
case Instruction::FCmp:
Result = SimplifyFCmpInst(cast<FCmpInst>(I)->getPredicate(),
- I->getOperand(0), I->getOperand(1), DL, TLI, DT);
+ I->getOperand(0), I->getOperand(1),
+ I->getFastMathFlags(), DL, TLI, DT, AC, I);
break;
case Instruction::Select:
Result = SimplifySelectInst(I->getOperand(0), I->getOperand(1),
- I->getOperand(2), DL, TLI, DT);
+ I->getOperand(2), DL, TLI, DT, AC, I);
break;
case Instruction::GetElementPtr: {
SmallVector<Value*, 8> Ops(I->op_begin(), I->op_end());
- Result = SimplifyGEPInst(Ops, DL, TLI, DT);
+ Result = SimplifyGEPInst(Ops, DL, TLI, DT, AC, I);
break;
}
case Instruction::InsertValue: {
InsertValueInst *IV = cast<InsertValueInst>(I);
Result = SimplifyInsertValueInst(IV->getAggregateOperand(),
IV->getInsertedValueOperand(),
- IV->getIndices(), DL, TLI, DT);
+ IV->getIndices(), DL, TLI, DT, AC, I);
+ break;
+ }
+ case Instruction::ExtractValue: {
+ auto *EVI = cast<ExtractValueInst>(I);
+ Result = SimplifyExtractValueInst(EVI->getAggregateOperand(),
+ EVI->getIndices(), DL, TLI, DT, AC, I);
+ break;
+ }
+ case Instruction::ExtractElement: {
+ auto *EEI = cast<ExtractElementInst>(I);
+ Result = SimplifyExtractElementInst(
+ EEI->getVectorOperand(), EEI->getIndexOperand(), DL, TLI, DT, AC, I);
break;
}
case Instruction::PHI:
- Result = SimplifyPHINode(cast<PHINode>(I), Query (DL, TLI, DT));
+ Result = SimplifyPHINode(cast<PHINode>(I), Query(DL, TLI, DT, AC, I));
break;
case Instruction::Call: {
CallSite CS(cast<CallInst>(I));
- Result = SimplifyCall(CS.getCalledValue(), CS.arg_begin(), CS.arg_end(),
- DL, TLI, DT);
+ Result = SimplifyCall(CS.getCalledValue(), CS.arg_begin(), CS.arg_end(), DL,
+ TLI, DT, AC, I);
break;
}
case Instruction::Trunc:
- Result = SimplifyTruncInst(I->getOperand(0), I->getType(), DL, TLI, DT);
+ Result =
+ SimplifyTruncInst(I->getOperand(0), I->getType(), DL, TLI, DT, AC, I);
break;
}
/// This routine returns 'true' only when *it* simplifies something. The passed
/// in simplified value does not count toward this.
static bool replaceAndRecursivelySimplifyImpl(Instruction *I, Value *SimpleV,
- const DataLayout *DL,
const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
+ const DominatorTree *DT,
+ AssumptionCache *AC) {
bool Simplified = false;
SmallSetVector<Instruction *, 8> Worklist;
+ const DataLayout &DL = I->getModule()->getDataLayout();
// If we have an explicit value to collapse to, do that round of the
// simplification loop by hand initially.
I = Worklist[Idx];
// See if this instruction simplifies.
- SimpleV = SimplifyInstruction(I, DL, TLI, DT);
+ SimpleV = SimplifyInstruction(I, DL, TLI, DT, AC);
if (!SimpleV)
continue;
}
bool llvm::recursivelySimplifyInstruction(Instruction *I,
- const DataLayout *DL,
const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
- return replaceAndRecursivelySimplifyImpl(I, nullptr, DL, TLI, DT);
+ const DominatorTree *DT,
+ AssumptionCache *AC) {
+ return replaceAndRecursivelySimplifyImpl(I, nullptr, TLI, DT, AC);
}
bool llvm::replaceAndRecursivelySimplify(Instruction *I, Value *SimpleV,
- const DataLayout *DL,
const TargetLibraryInfo *TLI,
- const DominatorTree *DT) {
+ const DominatorTree *DT,
+ AssumptionCache *AC) {
assert(I != SimpleV && "replaceAndRecursivelySimplify(X,X) is not valid!");
assert(SimpleV && "Must provide a simplified value.");
- return replaceAndRecursivelySimplifyImpl(I, SimpleV, DL, TLI, DT);
+ return replaceAndRecursivelySimplifyImpl(I, SimpleV, TLI, DT, AC);
}