blob: 770fdf9054c1bc4ddaf113e90234e6465125b0a7 [file] [log] [blame]
//===- InstructionSimplify.cpp - Fold instruction operands ----------------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
//
// This file implements routines for folding instructions into simpler forms
// that do not require creating new instructions. This does constant folding
// ("add i32 1, 1" -> "2") but can also handle non-constant operands, either
// returning a constant ("and i32 %x, 0" -> "0") or an already existing value
// ("and i32 %x, %x" -> "%x"). All operands are assumed to have already been
// simplified: This is usually true and assuming it simplifies the logic (if
// they have not been simplified then results are correct but maybe suboptimal).
//
//===----------------------------------------------------------------------===//
#include "llvm/Analysis/InstructionSimplify.h"
#include "llvm/ADT/SetVector.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/Analysis/AliasAnalysis.h"
#include "llvm/Analysis/AssumptionCache.h"
#include "llvm/Analysis/CaptureTracking.h"
#include "llvm/Analysis/CmpInstAnalysis.h"
#include "llvm/Analysis/ConstantFolding.h"
#include "llvm/Analysis/LoopAnalysisManager.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/GetElementPtrTypeIterator.h"
#include "llvm/IR/GlobalAlias.h"
#include "llvm/IR/InstrTypes.h"
#include "llvm/IR/Instructions.h"
#include "llvm/IR/Operator.h"
#include "llvm/IR/PatternMatch.h"
#include "llvm/IR/ValueHandle.h"
#include "llvm/Support/KnownBits.h"
#include <algorithm>
using namespace llvm;
using namespace llvm::PatternMatch;
#define DEBUG_TYPE "instsimplify"
enum { RecursionLimit = 3 };
STATISTIC(NumExpand, "Number of expansions");
STATISTIC(NumReassoc, "Number of reassociations");
static Value *SimplifyAndInst(Value *, Value *, const SimplifyQuery &, unsigned);
static Value *simplifyUnOp(unsigned, Value *, const SimplifyQuery &, unsigned);
static Value *simplifyFPUnOp(unsigned, Value *, const FastMathFlags &,
const SimplifyQuery &, unsigned);
static Value *SimplifyBinOp(unsigned, Value *, Value *, const SimplifyQuery &,
unsigned);
static Value *SimplifyFPBinOp(unsigned, Value *, Value *, const FastMathFlags &,
const SimplifyQuery &, unsigned);
static Value *SimplifyCmpInst(unsigned, Value *, Value *, const SimplifyQuery &,
unsigned);
static Value *SimplifyICmpInst(unsigned Predicate, Value *LHS, Value *RHS,
const SimplifyQuery &Q, unsigned MaxRecurse);
static Value *SimplifyOrInst(Value *, Value *, const SimplifyQuery &, unsigned);
static Value *SimplifyXorInst(Value *, Value *, const SimplifyQuery &, unsigned);
static Value *SimplifyCastInst(unsigned, Value *, Type *,
const SimplifyQuery &, unsigned);
static Value *SimplifyGEPInst(Type *, ArrayRef<Value *>, const SimplifyQuery &,
unsigned);
static Value *foldSelectWithBinaryOp(Value *Cond, Value *TrueVal,
Value *FalseVal) {
BinaryOperator::BinaryOps BinOpCode;
if (auto *BO = dyn_cast<BinaryOperator>(Cond))
BinOpCode = BO->getOpcode();
else
return nullptr;
CmpInst::Predicate ExpectedPred, Pred1, Pred2;
if (BinOpCode == BinaryOperator::Or) {
ExpectedPred = ICmpInst::ICMP_NE;
} else if (BinOpCode == BinaryOperator::And) {
ExpectedPred = ICmpInst::ICMP_EQ;
} else
return nullptr;
// %A = icmp eq %TV, %FV
// %B = icmp eq %X, %Y (and one of these is a select operand)
// %C = and %A, %B
// %D = select %C, %TV, %FV
// -->
// %FV
// %A = icmp ne %TV, %FV
// %B = icmp ne %X, %Y (and one of these is a select operand)
// %C = or %A, %B
// %D = select %C, %TV, %FV
// -->
// %TV
Value *X, *Y;
if (!match(Cond, m_c_BinOp(m_c_ICmp(Pred1, m_Specific(TrueVal),
m_Specific(FalseVal)),
m_ICmp(Pred2, m_Value(X), m_Value(Y)))) ||
Pred1 != Pred2 || Pred1 != ExpectedPred)
return nullptr;
if (X == TrueVal || X == FalseVal || Y == TrueVal || Y == FalseVal)
return BinOpCode == BinaryOperator::Or ? TrueVal : FalseVal;
return nullptr;
}
/// For a boolean type or a vector of boolean type, return false or a vector
/// with every element false.
static Constant *getFalse(Type *Ty) {
return ConstantInt::getFalse(Ty);
}
/// For a boolean type or a vector of boolean type, return true or a vector
/// with every element true.
static Constant *getTrue(Type *Ty) {
return ConstantInt::getTrue(Ty);
}
/// isSameCompare - Is V equivalent to the comparison "LHS Pred RHS"?
static bool isSameCompare(Value *V, CmpInst::Predicate Pred, Value *LHS,
Value *RHS) {
CmpInst *Cmp = dyn_cast<CmpInst>(V);
if (!Cmp)
return false;
CmpInst::Predicate CPred = Cmp->getPredicate();
Value *CLHS = Cmp->getOperand(0), *CRHS = Cmp->getOperand(1);
if (CPred == Pred && CLHS == LHS && CRHS == RHS)
return true;
return CPred == CmpInst::getSwappedPredicate(Pred) && CLHS == RHS &&
CRHS == LHS;
}
/// Does the given value dominate the specified phi node?
static bool valueDominatesPHI(Value *V, PHINode *P, const DominatorTree *DT) {
Instruction *I = dyn_cast<Instruction>(V);
if (!I)
// Arguments and constants dominate all instructions.
return true;
// If we are processing instructions (and/or basic blocks) that have not been
// fully added to a function, the parent nodes may still be null. Simply
// return the conservative answer in these cases.
if (!I->getParent() || !P->getParent() || !I->getFunction())
return false;
// If we have a DominatorTree then do a precise test.
if (DT)
return DT->dominates(I, P);
// Otherwise, if the instruction is in the entry block and is not an invoke,
// then it obviously dominates all phi nodes.
if (I->getParent() == &I->getFunction()->getEntryBlock() &&
!isa<InvokeInst>(I))
return true;
return false;
}
/// Simplify "A op (B op' C)" by distributing op over op', turning it into
/// "(A op B) op' (A op C)". Here "op" is given by Opcode and "op'" is
/// given by OpcodeToExpand, while "A" corresponds to LHS and "B op' C" to RHS.
/// Also performs the transform "(A op' B) op C" -> "(A op C) op' (B op C)".
/// Returns the simplified value, or null if no simplification was performed.
static Value *ExpandBinOp(Instruction::BinaryOps Opcode, Value *LHS, Value *RHS,
Instruction::BinaryOps OpcodeToExpand,
const SimplifyQuery &Q, unsigned MaxRecurse) {
// Recursion is always used, so bail out at once if we already hit the limit.
if (!MaxRecurse--)
return nullptr;
// Check whether the expression has the form "(A op' B) op C".
if (BinaryOperator *Op0 = dyn_cast<BinaryOperator>(LHS))
if (Op0->getOpcode() == OpcodeToExpand) {
// It does! Try turning it into "(A op C) op' (B op C)".
Value *A = Op0->getOperand(0), *B = Op0->getOperand(1), *C = RHS;
// Do "A op C" and "B op C" both simplify?
if (Value *L = SimplifyBinOp(Opcode, A, C, Q, MaxRecurse))
if (Value *R = SimplifyBinOp(Opcode, B, C, Q, MaxRecurse)) {
// They do! Return "L op' R" if it simplifies or is already available.
// If "L op' R" equals "A op' B" then "L op' R" is just the LHS.
if ((L == A && R == B) || (Instruction::isCommutative(OpcodeToExpand)
&& L == B && R == A)) {
++NumExpand;
return LHS;
}
// Otherwise return "L op' R" if it simplifies.
if (Value *V = SimplifyBinOp(OpcodeToExpand, L, R, Q, MaxRecurse)) {
++NumExpand;
return V;
}
}
}
// Check whether the expression has the form "A op (B op' C)".
if (BinaryOperator *Op1 = dyn_cast<BinaryOperator>(RHS))
if (Op1->getOpcode() == OpcodeToExpand) {
// It does! Try turning it into "(A op B) op' (A op C)".
Value *A = LHS, *B = Op1->getOperand(0), *C = Op1->getOperand(1);
// Do "A op B" and "A op C" both simplify?
if (Value *L = SimplifyBinOp(Opcode, A, B, Q, MaxRecurse))
if (Value *R = SimplifyBinOp(Opcode, A, C, Q, MaxRecurse)) {
// They do! Return "L op' R" if it simplifies or is already available.
// If "L op' R" equals "B op' C" then "L op' R" is just the RHS.
if ((L == B && R == C) || (Instruction::isCommutative(OpcodeToExpand)
&& L == C && R == B)) {
++NumExpand;
return RHS;
}
// Otherwise return "L op' R" if it simplifies.
if (Value *V = SimplifyBinOp(OpcodeToExpand, L, R, Q, MaxRecurse)) {
++NumExpand;
return V;
}
}
}
return nullptr;
}
/// Generic simplifications for associative binary operations.
/// Returns the simpler value, or null if none was found.
static Value *SimplifyAssociativeBinOp(Instruction::BinaryOps Opcode,
Value *LHS, Value *RHS,
const SimplifyQuery &Q,
unsigned MaxRecurse) {
assert(Instruction::isAssociative(Opcode) && "Not an associative operation!");
// Recursion is always used, so bail out at once if we already hit the limit.
if (!MaxRecurse--)
return nullptr;
BinaryOperator *Op0 = dyn_cast<BinaryOperator>(LHS);
BinaryOperator *Op1 = dyn_cast<BinaryOperator>(RHS);
// Transform: "(A op B) op C" ==> "A op (B op C)" if it simplifies completely.
if (Op0 && Op0->getOpcode() == Opcode) {
Value *A = Op0->getOperand(0);
Value *B = Op0->getOperand(1);
Value *C = RHS;
// Does "B op C" simplify?
if (Value *V = SimplifyBinOp(Opcode, B, C, Q, MaxRecurse)) {
// It does! Return "A op V" if it simplifies or is already available.
// If V equals B then "A op V" is just the LHS.
if (V == B) return LHS;
// Otherwise return "A op V" if it simplifies.
if (Value *W = SimplifyBinOp(Opcode, A, V, Q, MaxRecurse)) {
++NumReassoc;
return W;
}
}
}
// Transform: "A op (B op C)" ==> "(A op B) op C" if it simplifies completely.
if (Op1 && Op1->getOpcode() == Opcode) {
Value *A = LHS;
Value *B = Op1->getOperand(0);
Value *C = Op1->getOperand(1);
// Does "A op B" simplify?
if (Value *V = SimplifyBinOp(Opcode, A, B, Q, MaxRecurse)) {
// It does! Return "V op C" if it simplifies or is already available.
// If V equals B then "V op C" is just the RHS.
if (V == B) return RHS;
// Otherwise return "V op C" if it simplifies.
if (Value *W = SimplifyBinOp(Opcode, V, C, Q, MaxRecurse)) {
++NumReassoc;
return W;
}
}
}
// The remaining transforms require commutativity as well as associativity.
if (!Instruction::isCommutative(Opcode))
return nullptr;
// Transform: "(A op B) op C" ==> "(C op A) op B" if it simplifies completely.
if (Op0 && Op0->getOpcode() == Opcode) {
Value *A = Op0->getOperand(0);
Value *B = Op0->getOperand(1);
Value *C = RHS;
// Does "C op A" simplify?
if (Value *V = SimplifyBinOp(Opcode, C, A, Q, MaxRecurse)) {
// It does! Return "V op B" if it simplifies or is already available.
// If V equals A then "V op B" is just the LHS.
if (V == A) return LHS;
// Otherwise return "V op B" if it simplifies.
if (Value *W = SimplifyBinOp(Opcode, V, B, Q, MaxRecurse)) {
++NumReassoc;
return W;
}
}
}
// Transform: "A op (B op C)" ==> "B op (C op A)" if it simplifies completely.
if (Op1 && Op1->getOpcode() == Opcode) {
Value *A = LHS;
Value *B = Op1->getOperand(0);
Value *C = Op1->getOperand(1);
// Does "C op A" simplify?
if (Value *V = SimplifyBinOp(Opcode, C, A, Q, MaxRecurse)) {
// It does! Return "B op V" if it simplifies or is already available.
// If V equals C then "B op V" is just the RHS.
if (V == C) return RHS;
// Otherwise return "B op V" if it simplifies.
if (Value *W = SimplifyBinOp(Opcode, B, V, Q, MaxRecurse)) {
++NumReassoc;
return W;
}
}
}
return nullptr;
}
/// In the case of a binary operation with a select instruction as an operand,
/// try to simplify the binop by seeing whether evaluating it on both branches
/// of the select results in the same value. Returns the common value if so,
/// otherwise returns null.
static Value *ThreadBinOpOverSelect(Instruction::BinaryOps Opcode, Value *LHS,
Value *RHS, const SimplifyQuery &Q,
unsigned MaxRecurse) {
// Recursion is always used, so bail out at once if we already hit the limit.
if (!MaxRecurse--)
return nullptr;
SelectInst *SI;
if (isa<SelectInst>(LHS)) {
SI = cast<SelectInst>(LHS);
} else {
assert(isa<SelectInst>(RHS) && "No select instruction operand!");
SI = cast<SelectInst>(RHS);
}
// Evaluate the BinOp on the true and false branches of the select.
Value *TV;
Value *FV;
if (SI == LHS) {
TV = SimplifyBinOp(Opcode, SI->getTrueValue(), RHS, Q, MaxRecurse);
FV = SimplifyBinOp(Opcode, SI->getFalseValue(), RHS, Q, MaxRecurse);
} else {
TV = SimplifyBinOp(Opcode, LHS, SI->getTrueValue(), Q, MaxRecurse);
FV = SimplifyBinOp(Opcode, LHS, SI->getFalseValue(), Q, MaxRecurse);
}
// If they simplified to the same value, then return the common value.
// If they both failed to simplify then return null.
if (TV == FV)
return TV;
// If one branch simplified to undef, return the other one.
if (TV && isa<UndefValue>(TV))
return FV;
if (FV && isa<UndefValue>(FV))
return TV;
// If applying the operation did not change the true and false select values,
// then the result of the binop is the select itself.
if (TV == SI->getTrueValue() && FV == SI->getFalseValue())
return SI;
// If one branch simplified and the other did not, and the simplified
// value is equal to the unsimplified one, return the simplified value.
// For example, select (cond, X, X & Z) & Z -> X & Z.
if ((FV && !TV) || (TV && !FV)) {
// Check that the simplified value has the form "X op Y" where "op" is the
// same as the original operation.
Instruction *Simplified = dyn_cast<Instruction>(FV ? FV : TV);
if (Simplified && Simplified->getOpcode() == unsigned(Opcode)) {
// The value that didn't simplify is "UnsimplifiedLHS op UnsimplifiedRHS".
// We already know that "op" is the same as for the simplified value. See
// if the operands match too. If so, return the simplified value.
Value *UnsimplifiedBranch = FV ? SI->getTrueValue() : SI->getFalseValue();
Value *UnsimplifiedLHS = SI == LHS ? UnsimplifiedBranch : LHS;
Value *UnsimplifiedRHS = SI == LHS ? RHS : UnsimplifiedBranch;
if (Simplified->getOperand(0) == UnsimplifiedLHS &&
Simplified->getOperand(1) == UnsimplifiedRHS)
return Simplified;
if (Simplified->isCommutative() &&
Simplified->getOperand(1) == UnsimplifiedLHS &&
Simplified->getOperand(0) == UnsimplifiedRHS)
return Simplified;
}
}
return nullptr;
}
/// In the case of a comparison with a select instruction, try to simplify the
/// comparison by seeing whether both branches of the select result in the same
/// value. Returns the common value if so, otherwise returns null.
static Value *ThreadCmpOverSelect(CmpInst::Predicate Pred, Value *LHS,
Value *RHS, const SimplifyQuery &Q,
unsigned MaxRecurse) {
// Recursion is always used, so bail out at once if we already hit the limit.
if (!MaxRecurse--)
return nullptr;
// Make sure the select is on the LHS.
if (!isa<SelectInst>(LHS)) {
std::swap(LHS, RHS);
Pred = CmpInst::getSwappedPredicate(Pred);
}
assert(isa<SelectInst>(LHS) && "Not comparing with a select instruction!");
SelectInst *SI = cast<SelectInst>(LHS);
Value *Cond = SI->getCondition();
Value *TV = SI->getTrueValue();
Value *FV = SI->getFalseValue();
// Now that we have "cmp select(Cond, TV, FV), RHS", analyse it.
// Does "cmp TV, RHS" simplify?
Value *TCmp = SimplifyCmpInst(Pred, TV, RHS, Q, MaxRecurse);
if (TCmp == Cond) {
// It not only simplified, it simplified to the select condition. Replace
// it with 'true'.
TCmp = getTrue(Cond->getType());
} else if (!TCmp) {
// It didn't simplify. However if "cmp TV, RHS" is equal to the select
// condition then we can replace it with 'true'. Otherwise give up.
if (!isSameCompare(Cond, Pred, TV, RHS))
return nullptr;
TCmp = getTrue(Cond->getType());
}
// Does "cmp FV, RHS" simplify?
Value *FCmp = SimplifyCmpInst(Pred, FV, RHS, Q, MaxRecurse);
if (FCmp == Cond) {
// It not only simplified, it simplified to the select condition. Replace
// it with 'false'.
FCmp = getFalse(Cond->getType());
} else if (!FCmp) {
// It didn't simplify. However if "cmp FV, RHS" is equal to the select
// condition then we can replace it with 'false'. Otherwise give up.
if (!isSameCompare(Cond, Pred, FV, RHS))
return nullptr;
FCmp = getFalse(Cond->getType());
}
// If both sides simplified to the same value, then use it as the result of
// the original comparison.
if (TCmp == FCmp)
return TCmp;
// The remaining cases only make sense if the select condition has the same
// type as the result of the comparison, so bail out if this is not so.
if (Cond->getType()->isVectorTy() != RHS->getType()->isVectorTy())
return nullptr;
// If the false value simplified to false, then the result of the compare
// is equal to "Cond && TCmp". This also catches the case when the false
// value simplified to false and the true value to true, returning "Cond".
if (match(FCmp, m_Zero()))
if (Value *V = SimplifyAndInst(Cond, TCmp, Q, MaxRecurse))
return V;
// If the true value simplified to true, then the result of the compare
// is equal to "Cond || FCmp".
if (match(TCmp, m_One()))
if (Value *V = SimplifyOrInst(Cond, FCmp, Q, MaxRecurse))
return V;
// Finally, if the false value simplified to true and the true value to
// false, then the result of the compare is equal to "!Cond".
if (match(FCmp, m_One()) && match(TCmp, m_Zero()))
if (Value *V =
SimplifyXorInst(Cond, Constant::getAllOnesValue(Cond->getType()),
Q, MaxRecurse))
return V;
return nullptr;
}
/// In the case of a binary operation with an operand that is a PHI instruction,
/// try to simplify the binop by seeing whether evaluating it on the incoming
/// phi values yields the same result for every value. If so returns the common
/// value, otherwise returns null.
static Value *ThreadBinOpOverPHI(Instruction::BinaryOps Opcode, Value *LHS,
Value *RHS, const SimplifyQuery &Q,
unsigned MaxRecurse) {
// Recursion is always used, so bail out at once if we already hit the limit.
if (!MaxRecurse--)
return nullptr;
PHINode *PI;
if (isa<PHINode>(LHS)) {
PI = cast<PHINode>(LHS);
// Bail out if RHS and the phi may be mutually interdependent due to a loop.
if (!valueDominatesPHI(RHS, PI, Q.DT))
return nullptr;
} else {
assert(isa<PHINode>(RHS) && "No PHI instruction operand!");
PI = cast<PHINode>(RHS);
// Bail out if LHS and the phi may be mutually interdependent due to a loop.
if (!valueDominatesPHI(LHS, PI, Q.DT))
return nullptr;
}
// Evaluate the BinOp on the incoming phi values.
Value *CommonValue = nullptr;
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 ?
SimplifyBinOp(Opcode, Incoming, RHS, Q, MaxRecurse) :
SimplifyBinOp(Opcode, LHS, Incoming, Q, MaxRecurse);
// If the operation failed to simplify, or simplified to a different value
// to previously, then give up.
if (!V || (CommonValue && V != CommonValue))
return nullptr;
CommonValue = V;
}
return CommonValue;
}
/// In the case of a comparison with a PHI instruction, try to simplify the
/// comparison by seeing whether comparing with all of the incoming phi values
/// yields the same result every time. If so returns the common result,
/// otherwise returns null.
static Value *ThreadCmpOverPHI(CmpInst::Predicate Pred, Value *LHS, Value *RHS,
const SimplifyQuery &Q, unsigned MaxRecurse) {
// Recursion is always used, so bail out at once if we already hit the limit.
if (!MaxRecurse--)
return nullptr;
// Make sure the phi is on the LHS.
if (!isa<PHINode>(LHS)) {
std::swap(LHS, RHS);
Pred = CmpInst::getSwappedPredicate(Pred);
}
assert(isa<PHINode>(LHS) && "Not comparing with a phi instruction!");
PHINode *PI = cast<PHINode>(LHS);
// Bail out if RHS and the phi may be mutually interdependent due to a loop.
if (!valueDominatesPHI(RHS, PI, Q.DT))
return nullptr;
// Evaluate the BinOp on the incoming phi values.
Value *CommonValue = nullptr;
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);
// If the operation failed to simplify, or simplified to a different value
// to previously, then give up.
if (!V || (CommonValue && V != CommonValue))
return nullptr;
CommonValue = V;
}
return CommonValue;
}
static Constant *foldOrCommuteConstant(Instruction::BinaryOps Opcode,
Value *&Op0, Value *&Op1,
const SimplifyQuery &Q) {
if (auto *CLHS = dyn_cast<Constant>(Op0)) {
if (auto *CRHS = dyn_cast<Constant>(Op1))
return ConstantFoldBinaryOpOperands(Opcode, CLHS, CRHS, Q.DL);
// Canonicalize the constant to the RHS if this is a commutative operation.
if (Instruction::isCommutative(Opcode))
std::swap(Op0, Op1);
}
return nullptr;
}
/// Given operands for an Add, see if we can fold the result.
/// If not, this returns null.
static Value *SimplifyAddInst(Value *Op0, Value *Op1, bool IsNSW, bool IsNUW,
const SimplifyQuery &Q, unsigned MaxRecurse) {
if (Constant *C = foldOrCommuteConstant(Instruction::Add, Op0, Op1, Q))
return C;
// X + undef -> undef
if (match(Op1, m_Undef()))
return Op1;
// X + 0 -> X
if (match(Op1, m_Zero()))
return Op0;
// If two operands are negative, return 0.
if (isKnownNegation(Op0, Op1))
return Constant::getNullValue(Op0->getType());
// X + (Y - X) -> Y
// (Y - X) + X -> Y
// Eg: X + -X -> 0
Value *Y = nullptr;
if (match(Op1, m_Sub(m_Value(Y), m_Specific(Op0))) ||
match(Op0, m_Sub(m_Value(Y), m_Specific(Op1))))
return Y;
// X + ~X -> -1 since ~X = -X-1
Type *Ty = Op0->getType();
if (match(Op0, m_Not(m_Specific(Op1))) ||
match(Op1, m_Not(m_Specific(Op0))))
return Constant::getAllOnesValue(Ty);
// add nsw/nuw (xor Y, signmask), signmask --> Y
// The no-wrapping add guarantees that the top bit will be set by the add.
// Therefore, the xor must be clearing the already set sign bit of Y.
if ((IsNSW || IsNUW) && match(Op1, m_SignMask()) &&
match(Op0, m_Xor(m_Value(Y), m_SignMask())))
return Y;
// add nuw %x, -1 -> -1, because %x can only be 0.
if (IsNUW && match(Op1, m_AllOnes()))
return Op1; // Which is -1.
/// i1 add -> xor.
if (MaxRecurse && Op0->getType()->isIntOrIntVectorTy(1))
if (Value *V = SimplifyXorInst(Op0, Op1, Q, MaxRecurse-1))
return V;
// Try some generic simplifications for associative operations.
if (Value *V = SimplifyAssociativeBinOp(Instruction::Add, Op0, Op1, Q,
MaxRecurse))
return V;
// Threading Add over selects and phi nodes is pointless, so don't bother.
// Threading over the select in "A + select(cond, B, C)" means evaluating
// "A+B" and "A+C" and seeing if they are equal; but they are equal if and
// only if B and C are equal. If B and C are equal then (since we assume
// that operands have already been simplified) "select(cond, B, C)" should
// have been simplified to the common value of B and C already. Analysing
// "A+B" and "A+C" thus gains nothing, but costs compile time. Similarly
// for threading over phi nodes.
return nullptr;
}
Value *llvm::SimplifyAddInst(Value *Op0, Value *Op1, bool IsNSW, bool IsNUW,
const SimplifyQuery &Query) {
return ::SimplifyAddInst(Op0, Op1, IsNSW, IsNUW, Query, RecursionLimit);
}
/// Compute the base pointer and cumulative constant offsets for V.
///
/// This strips all constant offsets off of V, leaving it the base pointer, and
/// accumulates the total constant offset applied in the returned constant. It
/// returns 0 if V is not a pointer, and returns the constant '0' if there are
/// no constant offsets applied.
///
/// 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,
bool AllowNonInbounds = false) {
assert(V->getType()->isPtrOrPtrVectorTy());
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
// instruction in an unreachable block, which may be on a cycle.
SmallPtrSet<Value *, 4> Visited;
Visited.insert(V);
do {
if (GEPOperator *GEP = dyn_cast<GEPOperator>(V)) {
if ((!AllowNonInbounds && !GEP->isInBounds()) ||
!GEP->accumulateConstantOffset(DL, Offset))
break;
V = GEP->getPointerOperand();
} else if (Operator::getOpcode(V) == Instruction::BitCast) {
V = cast<Operator>(V)->getOperand(0);
} else if (GlobalAlias *GA = dyn_cast<GlobalAlias>(V)) {
if (GA->isInterposable())
break;
V = GA->getAliasee();
} else {
if (auto *Call = dyn_cast<CallBase>(V))
if (Value *RV = Call->getReturnedArgOperand()) {
V = RV;
continue;
}
break;
}
assert(V->getType()->isPtrOrPtrVectorTy() && "Unexpected operand type!");
} while (Visited.insert(V).second);
Constant *OffsetIntPtr = ConstantInt::get(IntPtrTy, Offset);
if (V->getType()->isVectorTy())
return ConstantVector::getSplat(V->getType()->getVectorNumElements(),
OffsetIntPtr);
return OffsetIntPtr;
}
/// 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) {
Constant *LHSOffset = stripAndComputeConstantOffsets(DL, LHS);
Constant *RHSOffset = stripAndComputeConstantOffsets(DL, RHS);
// If LHS and RHS are not related via constant offsets to the same base
// value, there is nothing we can do here.
if (LHS != RHS)
return nullptr;
// Otherwise, the difference of LHS - RHS can be computed as:
// LHS - RHS
// = (LHSOffset + Base) - (RHSOffset + Base)
// = LHSOffset - RHSOffset
return ConstantExpr::getSub(LHSOffset, RHSOffset);
}
/// Given operands for a Sub, see if we can fold the result.
/// If not, this returns null.
static Value *SimplifySubInst(Value *Op0, Value *Op1, bool isNSW, bool isNUW,
const SimplifyQuery &Q, unsigned MaxRecurse) {
if (Constant *C = foldOrCommuteConstant(Instruction::Sub, Op0, Op1, Q))
return C;
// X - undef -> undef
// undef - X -> undef
if (match(Op0, m_Undef()) || match(Op1, m_Undef()))
return UndefValue::get(Op0->getType());
// X - 0 -> X
if (match(Op1, m_Zero()))
return Op0;
// X - X -> 0
if (Op0 == Op1)
return Constant::getNullValue(Op0->getType());
// Is this a negation?
if (match(Op0, m_Zero())) {
// 0 - X -> 0 if the sub is NUW.
if (isNUW)
return Constant::getNullValue(Op0->getType());
KnownBits Known = computeKnownBits(Op1, Q.DL, 0, Q.AC, Q.CxtI, Q.DT);
if (Known.Zero.isMaxSignedValue()) {
// Op1 is either 0 or the minimum signed value. If the sub is NSW, then
// Op1 must be 0 because negating the minimum signed value is undefined.
if (isNSW)
return Constant::getNullValue(Op0->getType());
// 0 - X -> X if X is 0 or the minimum signed value.
return Op1;
}
}
// (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;
if (MaxRecurse && match(Op0, m_Add(m_Value(X), m_Value(Y)))) { // (X + Y) - Z
// See if "V === Y - Z" simplifies.
if (Value *V = SimplifyBinOp(Instruction::Sub, Y, Z, Q, MaxRecurse-1))
// It does! Now see if "X + V" simplifies.
if (Value *W = SimplifyBinOp(Instruction::Add, X, V, Q, MaxRecurse-1)) {
// It does, we successfully reassociated!
++NumReassoc;
return W;
}
// See if "V === X - Z" simplifies.
if (Value *V = SimplifyBinOp(Instruction::Sub, X, Z, Q, MaxRecurse-1))
// It does! Now see if "Y + V" simplifies.
if (Value *W = SimplifyBinOp(Instruction::Add, Y, V, Q, MaxRecurse-1)) {
// It does, we successfully reassociated!
++NumReassoc;
return W;
}
}
// X - (Y + Z) -> (X - Y) - Z or (X - Z) - Y if everything simplifies.
// For example, X - (X + 1) -> -1
X = Op0;
if (MaxRecurse && match(Op1, m_Add(m_Value(Y), m_Value(Z)))) { // X - (Y + Z)
// See if "V === X - Y" simplifies.
if (Value *V = SimplifyBinOp(Instruction::Sub, X, Y, Q, MaxRecurse-1))
// It does! Now see if "V - Z" simplifies.
if (Value *W = SimplifyBinOp(Instruction::Sub, V, Z, Q, MaxRecurse-1)) {
// It does, we successfully reassociated!
++NumReassoc;
return W;
}
// See if "V === X - Z" simplifies.
if (Value *V = SimplifyBinOp(Instruction::Sub, X, Z, Q, MaxRecurse-1))
// It does! Now see if "V - Y" simplifies.
if (Value *W = SimplifyBinOp(Instruction::Sub, V, Y, Q, MaxRecurse-1)) {
// It does, we successfully reassociated!
++NumReassoc;
return W;
}
}
// Z - (X - Y) -> (Z - X) + Y if everything simplifies.
// For example, X - (X - Y) -> Y.
Z = Op0;
if (MaxRecurse && match(Op1, m_Sub(m_Value(X), m_Value(Y)))) // Z - (X - Y)
// See if "V === Z - X" simplifies.
if (Value *V = SimplifyBinOp(Instruction::Sub, Z, X, Q, MaxRecurse-1))
// It does! Now see if "V + Y" simplifies.
if (Value *W = SimplifyBinOp(Instruction::Add, V, Y, Q, MaxRecurse-1)) {
// It does, we successfully reassociated!
++NumReassoc;
return W;
}
// trunc(X) - trunc(Y) -> trunc(X - Y) if everything simplifies.
if (MaxRecurse && match(Op0, m_Trunc(m_Value(X))) &&
match(Op1, m_Trunc(m_Value(Y))))
if (X->getType() == Y->getType())
// See if "V === X - Y" simplifies.
if (Value *V = SimplifyBinOp(Instruction::Sub, X, Y, Q, MaxRecurse-1))
// It does! Now see if "trunc V" simplifies.
if (Value *W = SimplifyCastInst(Instruction::Trunc, V, Op0->getType(),
Q, MaxRecurse - 1))
// It does, return the simplified "trunc V".
return W;
// Variations on GEP(base, I, ...) - GEP(base, i, ...) -> GEP(null, I-i, ...).
if (match(Op0, m_PtrToInt(m_Value(X))) &&
match(Op1, m_PtrToInt(m_Value(Y))))
if (Constant *Result = computePointerDifference(Q.DL, X, Y))
return ConstantExpr::getIntegerCast(Result, Op0->getType(), true);
// i1 sub -> xor.
if (MaxRecurse && Op0->getType()->isIntOrIntVectorTy(1))
if (Value *V = SimplifyXorInst(Op0, Op1, Q, MaxRecurse-1))
return V;
// Threading Sub over selects and phi nodes is pointless, so don't bother.
// Threading over the select in "A - select(cond, B, C)" means evaluating
// "A-B" and "A-C" and seeing if they are equal; but they are equal if and
// only if B and C are equal. If B and C are equal then (since we assume
// that operands have already been simplified) "select(cond, B, C)" should
// have been simplified to the common value of B and C already. Analysing
// "A-B" and "A-C" thus gains nothing, but costs compile time. Similarly
// for threading over phi nodes.
return nullptr;
}
Value *llvm::SimplifySubInst(Value *Op0, Value *Op1, bool isNSW, bool isNUW,
const SimplifyQuery &Q) {
return ::SimplifySubInst(Op0, Op1, isNSW, isNUW, Q, RecursionLimit);
}
/// Given operands for a Mul, see if we can fold the result.
/// If not, this returns null.
static Value *SimplifyMulInst(Value *Op0, Value *Op1, const SimplifyQuery &Q,
unsigned MaxRecurse) {
if (Constant *C = foldOrCommuteConstant(Instruction::Mul, Op0, Op1, Q))
return C;
// X * undef -> 0
// X * 0 -> 0
if (match(Op1, m_CombineOr(m_Undef(), m_Zero())))
return Constant::getNullValue(Op0->getType());
// X * 1 -> X
if (match(Op1, m_One()))
return Op0;
// (X / Y) * Y -> X if the division is exact.
Value *X = nullptr;
if (Q.IIQ.UseInstrInfo &&
(match(Op0,
m_Exact(m_IDiv(m_Value(X), m_Specific(Op1)))) || // (X / Y) * Y
match(Op1, m_Exact(m_IDiv(m_Value(X), m_Specific(Op0)))))) // Y * (X / Y)
return X;
// i1 mul -> and.
if (MaxRecurse && Op0->getType()->isIntOrIntVectorTy(1))
if (Value *V = SimplifyAndInst(Op0, Op1, Q, MaxRecurse-1))
return V;
// Try some generic simplifications for associative operations.
if (Value *V = SimplifyAssociativeBinOp(Instruction::Mul, Op0, Op1, Q,
MaxRecurse))
return V;
// Mul distributes over Add. Try some generic simplifications based on this.
if (Value *V = ExpandBinOp(Instruction::Mul, Op0, Op1, Instruction::Add,
Q, MaxRecurse))
return V;
// 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))
if (Value *V = ThreadBinOpOverSelect(Instruction::Mul, Op0, Op1, Q,
MaxRecurse))
return V;
// If the operation is with the result of a phi instruction, check whether
// operating on all incoming values of the phi always yields the same value.
if (isa<PHINode>(Op0) || isa<PHINode>(Op1))
if (Value *V = ThreadBinOpOverPHI(Instruction::Mul, Op0, Op1, Q,
MaxRecurse))
return V;
return nullptr;
}
Value *llvm::SimplifyMulInst(Value *Op0, Value *Op1, const SimplifyQuery &Q) {
return ::SimplifyMulInst(Op0, Op1, Q, RecursionLimit);
}
/// Check for common or similar folds of integer division or integer remainder.
/// This applies to all 4 opcodes (sdiv/udiv/srem/urem).
static Value *simplifyDivRem(Value *Op0, Value *Op1, bool IsDiv) {
Type *Ty = Op0->getType();
// X / undef -> undef
// X % undef -> undef
if (match(Op1, m_Undef()))
return Op1;
// X / 0 -> undef
// X % 0 -> undef
// We don't need to preserve faults!
if (match(Op1, m_Zero()))
return UndefValue::get(Ty);
// If any element of a constant divisor vector is zero or undef, the whole op
// is undef.
auto *Op1C = dyn_cast<Constant>(Op1);
if (Op1C && Ty->isVectorTy()) {
unsigned NumElts = Ty->getVectorNumElements();
for (unsigned i = 0; i != NumElts; ++i) {
Constant *Elt = Op1C->getAggregateElement(i);
if (Elt && (Elt->isNullValue() || isa<UndefValue>(Elt)))
return UndefValue::get(Ty);
}
}
// undef / X -> 0
// undef % X -> 0
if (match(Op0, m_Undef()))
return Constant::getNullValue(Ty);
// 0 / X -> 0
// 0 % X -> 0
if (match(Op0, m_Zero()))
return Constant::getNullValue(Op0->getType());
// X / X -> 1
// X % X -> 0
if (Op0 == Op1)
return IsDiv ? ConstantInt::get(Ty, 1) : Constant::getNullValue(Ty);
// X / 1 -> X
// X % 1 -> 0
// If this is a boolean op (single-bit element type), we can't have
// division-by-zero or remainder-by-zero, so assume the divisor is 1.
// Similarly, if we're zero-extending a boolean divisor, then assume it's a 1.
Value *X;
if (match(Op1, m_One()) || Ty->isIntOrIntVectorTy(1) ||
(match(Op1, m_ZExt(m_Value(X))) && X->getType()->isIntOrIntVectorTy(1)))
return IsDiv ? Op0 : Constant::getNullValue(Ty);
return nullptr;
}
/// Given a predicate and two operands, return true if the comparison is true.
/// This is a helper for div/rem simplification where we return some other value
/// when we can prove a relationship between the operands.
static bool isICmpTrue(ICmpInst::Predicate Pred, Value *LHS, Value *RHS,
const SimplifyQuery &Q, unsigned MaxRecurse) {
Value *V = SimplifyICmpInst(Pred, LHS, RHS, Q, MaxRecurse);
Constant *C = dyn_cast_or_null<Constant>(V);
return (C && C->isAllOnesValue());
}
/// Return true if we can simplify X / Y to 0. Remainder can adapt that answer
/// to simplify X % Y to X.
static bool isDivZero(Value *X, Value *Y, const SimplifyQuery &Q,
unsigned MaxRecurse, bool IsSigned) {
// Recursion is always used, so bail out at once if we already hit the limit.
if (!MaxRecurse--)
return false;
if (IsSigned) {
// |X| / |Y| --> 0
//
// We require that 1 operand is a simple constant. That could be extended to
// 2 variables if we computed the sign bit for each.
//
// Make sure that a constant is not the minimum signed value because taking
// the abs() of that is undefined.
Type *Ty = X->getType();
const APInt *C;
if (match(X, m_APInt(C)) && !C->isMinSignedValue()) {
// Is the variable divisor magnitude always greater than the constant
// dividend magnitude?
// |Y| > |C| --> Y < -abs(C) or Y > abs(C)
Constant *PosDividendC = ConstantInt::get(Ty, C->abs());
Constant *NegDividendC = ConstantInt::get(Ty, -C->abs());
if (isICmpTrue(CmpInst::ICMP_SLT, Y, NegDividendC, Q, MaxRecurse) ||
isICmpTrue(CmpInst::ICMP_SGT, Y, PosDividendC, Q, MaxRecurse))
return true;
}
if (match(Y, m_APInt(C))) {
// Special-case: we can't take the abs() of a minimum signed value. If
// that's the divisor, then all we have to do is prove that the dividend
// is also not the minimum signed value.
if (C->isMinSignedValue())
return isICmpTrue(CmpInst::ICMP_NE, X, Y, Q, MaxRecurse);
// Is the variable dividend magnitude always less than the constant
// divisor magnitude?
// |X| < |C| --> X > -abs(C) and X < abs(C)
Constant *PosDivisorC = ConstantInt::get(Ty, C->abs());
Constant *NegDivisorC = ConstantInt::get(Ty, -C->abs());
if (isICmpTrue(CmpInst::ICMP_SGT, X, NegDivisorC, Q, MaxRecurse) &&
isICmpTrue(CmpInst::ICMP_SLT, X, PosDivisorC, Q, MaxRecurse))
return true;
}
return false;
}
// IsSigned == false.
// Is the dividend unsigned less than the divisor?
return isICmpTrue(ICmpInst::ICMP_ULT, X, Y, Q, MaxRecurse);
}
/// These are simplifications common to SDiv and UDiv.
static Value *simplifyDiv(Instruction::BinaryOps Opcode, Value *Op0, Value *Op1,
const SimplifyQuery &Q, unsigned MaxRecurse) {
if (Constant *C = foldOrCommuteConstant(Opcode, Op0, Op1, Q))
return C;
if (Value *V = simplifyDivRem(Op0, Op1, true))
return V;
bool IsSigned = Opcode == Instruction::SDiv;
// (X * Y) / Y -> X if the multiplication does not overflow.
Value *X;
if (match(Op0, m_c_Mul(m_Value(X), m_Specific(Op1)))) {
auto *Mul = cast<OverflowingBinaryOperator>(Op0);
// If the Mul does not overflow, then we are good to go.
if ((IsSigned && Q.IIQ.hasNoSignedWrap(Mul)) ||
(!IsSigned && Q.IIQ.hasNoUnsignedWrap(Mul)))
return X;
// If X has the form X = A / Y, then X * Y cannot overflow.
if ((IsSigned && match(X, m_SDiv(m_Value(), m_Specific(Op1)))) ||
(!IsSigned && match(X, m_UDiv(m_Value(), m_Specific(Op1)))))
return X;
}
// (X rem Y) / Y -> 0
if ((IsSigned && match(Op0, m_SRem(m_Value(), m_Specific(Op1)))) ||
(!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;
(void)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))
if (Value *V = ThreadBinOpOverSelect(Opcode, Op0, Op1, Q, MaxRecurse))
return V;
// If the operation is with the result of a phi instruction, check whether
// operating on all incoming values of the phi always yields the same value.
if (isa<PHINode>(Op0) || isa<PHINode>(Op1))
if (Value *V = ThreadBinOpOverPHI(Opcode, Op0, Op1, Q, MaxRecurse))
return V;
if (isDivZero(Op0, Op1, Q, MaxRecurse, IsSigned))
return Constant::getNullValue(Op0->getType());
return nullptr;
}
/// These are simplifications common to SRem and URem.
static Value *simplifyRem(Instruction::BinaryOps Opcode, Value *Op0, Value *Op1,
const SimplifyQuery &Q, unsigned MaxRecurse) {
if (Constant *C = foldOrCommuteConstant(Opcode, Op0, Op1, Q))
return C;
if (Value *V = simplifyDivRem(Op0, Op1, false))
return V;
// (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;
// (X << Y) % X -> 0
if (Q.IIQ.UseInstrInfo &&
((Opcode == Instruction::SRem &&
match(Op0, m_NSWShl(m_Specific(Op1), m_Value()))) ||
(Opcode == Instruction::URem &&
match(Op0, m_NUWShl(m_Specific(Op1), m_Value())))))
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))
if (Value *V = ThreadBinOpOverSelect(Opcode, Op0, Op1, Q, MaxRecurse))
return V;
// If the operation is with the result of a phi instruction, check whether
// operating on all incoming values of the phi always yields the same value.
if (isa<PHINode>(Op0) || isa<PHINode>(Op1))
if (Value *V = ThreadBinOpOverPHI(Opcode, Op0, Op1, Q, MaxRecurse))
return V;
// If X / Y == 0, then X % Y == X.
if (isDivZero(Op0, Op1, Q, MaxRecurse, Opcode == Instruction::SRem))
return Op0;
return nullptr;
}
/// Given operands for an SDiv, see if we can fold the result.
/// If not, this returns null.
static Value *SimplifySDivInst(Value *Op0, Value *Op1, const SimplifyQuery &Q,
unsigned MaxRecurse) {
// If two operands are negated and no signed overflow, return -1.
if (isKnownNegation(Op0, Op1, /*NeedNSW=*/true))
return Constant::getAllOnesValue(Op0->getType());
return simplifyDiv(Instruction::SDiv, Op0, Op1, Q, MaxRecurse);
}
Value *llvm::SimplifySDivInst(Value *Op0, Value *Op1, const SimplifyQuery &Q) {
return ::SimplifySDivInst(Op0, Op1, Q, RecursionLimit);
}
/// Given operands for a UDiv, see if we can fold the result.
/// If not, this returns null.
static Value *SimplifyUDivInst(Value *Op0, Value *Op1, const SimplifyQuery &Q,
unsigned MaxRecurse) {
return simplifyDiv(Instruction::UDiv, Op0, Op1, Q, MaxRecurse);
}
Value *llvm::SimplifyUDivInst(Value *Op0, Value *Op1, const SimplifyQuery &Q) {
return ::SimplifyUDivInst(Op0, Op1, Q, RecursionLimit);
}
/// Given operands for an SRem, see if we can fold the result.
/// If not, this returns null.
static Value *SimplifySRemInst(Value *Op0, Value *Op1, const SimplifyQuery &Q,
unsigned MaxRecurse) {
// If the divisor is 0, the result is undefined, so assume the divisor is -1.
// srem Op0, (sext i1 X) --> srem Op0, -1 --> 0
Value *X;
if (match(Op1, m_SExt(m_Value(X))) && X->getType()->isIntOrIntVectorTy(1))
return ConstantInt::getNullValue(Op0->getType());
// If the two operands are negated, return 0.
if (isKnownNegation(Op0, Op1))
return ConstantInt::getNullValue(Op0->getType());
return simplifyRem(Instruction::SRem, Op0, Op1, Q, MaxRecurse);
}
Value *llvm::SimplifySRemInst(Value *Op0, Value *Op1, const SimplifyQuery &Q) {
return ::SimplifySRemInst(Op0, Op1, Q, RecursionLimit);
}
/// Given operands for a URem, see if we can fold the result.
/// If not, this returns null.
static Value *SimplifyURemInst(Value *Op0, Value *Op1, const SimplifyQuery &Q,
unsigned MaxRecurse) {
return simplifyRem(Instruction::URem, Op0, Op1, Q, MaxRecurse);
}
Value *llvm::SimplifyURemInst(Value *Op0, Value *Op1, const SimplifyQuery &Q) {
return ::SimplifyURemInst(Op0, Op1, Q, RecursionLimit);
}
/// Returns true if a shift by \c Amount always yields undef.
static bool isUndefShift(Value *Amount) {
Constant *C = dyn_cast<Constant>(Amount);
if (!C)
return false;
// X shift by undef -> undef because it may shift by the bitwidth.
if (isa<UndefValue>(C))
return true;
// Shifting by the bitwidth or more is undefined.
if (ConstantInt *CI = dyn_cast<ConstantInt>(C))
if (CI->getValue().getLimitedValue() >=
CI->getType()->getScalarSizeInBits())
return true;
// If all lanes of a vector shift are undefined the whole shift is.
if (isa<ConstantVector>(C) || isa<ConstantDataVector>(C)) {
for (unsigned I = 0, E = C->getType()->getVectorNumElements(); I != E; ++I)
if (!isUndefShift(C->getAggregateElement(I)))
return false;
return true;
}
return false;
}
/// Given operands for an Shl, LShr or AShr, see if we can fold the result.
/// If not, this returns null.
static Value *SimplifyShift(Instruction::BinaryOps Opcode, Value *Op0,
Value *Op1, const SimplifyQuery &Q, unsigned MaxRecurse) {
if (Constant *C = foldOrCommuteConstant(Opcode, Op0, Op1, Q))
return C;
// 0 shift by X -> 0
if (match(Op0, m_Zero()))
return Constant::getNullValue(Op0->getType());
// X shift by 0 -> X
// Shift-by-sign-extended bool must be shift-by-0 because shift-by-all-ones
// would be poison.
Value *X;
if (match(Op1, m_Zero()) ||
(match(Op1, m_SExt(m_Value(X))) && X->getType()->isIntOrIntVectorTy(1)))
return Op0;
// Fold undefined shifts.
if (isUndefShift(Op1))
return UndefValue::get(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))
if (Value *V = ThreadBinOpOverSelect(Opcode, Op0, Op1, Q, MaxRecurse))
return V;
// If the operation is with the result of a phi instruction, check whether
// operating on all incoming values of the phi always yields the same value.
if (isa<PHINode>(Op0) || isa<PHINode>(Op1))
if (Value *V = ThreadBinOpOverPHI(Opcode, Op0, Op1, Q, MaxRecurse))
return V;
// If any bits in the shift amount make that value greater than or equal to
// the number of bits in the type, the shift is undefined.
KnownBits Known = computeKnownBits(Op1, Q.DL, 0, Q.AC, Q.CxtI, Q.DT);
if (Known.One.getLimitedValue() >= Known.getBitWidth())
return UndefValue::get(Op0->getType());
// If all valid bits in the shift amount are known zero, the first operand is
// unchanged.
unsigned NumValidShiftBits = Log2_32_Ceil(Known.getBitWidth());
if (Known.countMinTrailingZeros() >= NumValidShiftBits)
return Op0;
return nullptr;
}
/// Given operands for an Shl, LShr or AShr, see if we can
/// fold the result. If not, this returns null.
static Value *SimplifyRightShift(Instruction::BinaryOps Opcode, Value *Op0,
Value *Op1, bool isExact, const SimplifyQuery &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) {
KnownBits Op0Known = computeKnownBits(Op0, Q.DL, /*Depth=*/0, Q.AC, Q.CxtI, Q.DT);
if (Op0Known.One[0])
return Op0;
}
return nullptr;
}
/// 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,
const SimplifyQuery &Q, unsigned MaxRecurse) {
if (Value *V = SimplifyShift(Instruction::Shl, Op0, Op1, Q, MaxRecurse))
return V;
// undef << X -> 0
// undef << X -> undef if (if it's NSW/NUW)
if (match(Op0, m_Undef()))
return isNSW || isNUW ? Op0 : Constant::getNullValue(Op0->getType());
// (X >> A) << A -> X
Value *X;
if (Q.IIQ.UseInstrInfo &&
match(Op0, m_Exact(m_Shr(m_Value(X), m_Specific(Op1)))))
return X;
// shl nuw i8 C, %x -> C iff C has sign bit set.
if (isNUW && match(Op0, m_Negative()))
return Op0;
// NOTE: could use computeKnownBits() / LazyValueInfo,
// but the cost-benefit analysis suggests it isn't worth it.
return nullptr;
}
Value *llvm::SimplifyShlInst(Value *Op0, Value *Op1, bool isNSW, bool isNUW,
const SimplifyQuery &Q) {
return ::SimplifyShlInst(Op0, Op1, isNSW, isNUW, Q, RecursionLimit);
}
/// Given operands for an LShr, see if we can fold the result.
/// If not, this returns null.
static Value *SimplifyLShrInst(Value *Op0, Value *Op1, bool isExact,
const SimplifyQuery &Q, unsigned MaxRecurse) {
if (Value *V = SimplifyRightShift(Instruction::LShr, Op0, Op1, isExact, Q,
MaxRecurse))
return V;
// (X << A) >> A -> X
Value *X;
if (match(Op0, m_NUWShl(m_Value(X), m_Specific(Op1))))
return X;
// ((X << A) | Y) >> A -> X if effective width of Y is not larger than A.
// We can return X as we do in the above case since OR alters no bits in X.
// SimplifyDemandedBits in InstCombine can do more general optimization for
// bit manipulation. This pattern aims to provide opportunities for other
// optimizers by supporting a simple but common case in InstSimplify.
Value *Y;
const APInt *ShRAmt, *ShLAmt;
if (match(Op1, m_APInt(ShRAmt)) &&
match(Op0, m_c_Or(m_NUWShl(m_Value(X), m_APInt(ShLAmt)), m_Value(Y))) &&
*ShRAmt == *ShLAmt) {
const KnownBits YKnown = computeKnownBits(Y, Q.DL, 0, Q.AC, Q.CxtI, Q.DT);
const unsigned Width = Op0->getType()->getScalarSizeInBits();
const unsigned EffWidthY = Width - YKnown.countMinLeadingZeros();
if (ShRAmt->uge(EffWidthY))
return X;
}
return nullptr;
}
Value *llvm::SimplifyLShrInst(Value *Op0, Value *Op1, bool isExact,
const SimplifyQuery &Q) {
return ::SimplifyLShrInst(Op0, Op1, isExact, Q, RecursionLimit);
}
/// Given operands for an AShr, see if we can fold the result.
/// If not, this returns null.
static Value *SimplifyAShrInst(Value *Op0, Value *Op1, bool isExact,
const SimplifyQuery &Q, unsigned MaxRecurse) {
if (Value *V = SimplifyRightShift(Instruction::AShr, Op0, Op1, isExact, Q,
MaxRecurse))
return V;
// all ones >>a X -> -1
// Do not return Op0 because it may contain undef elements if it's a vector.
if (match(Op0, m_AllOnes()))
return Constant::getAllOnesValue(Op0->getType());
// (X << A) >> A -> X
Value *X;
if (Q.IIQ.UseInstrInfo && 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, Q.DL, 0, Q.AC, Q.CxtI, Q.DT);
if (NumSignBits == Op0->getType()->getScalarSizeInBits())
return Op0;
return nullptr;
}
Value *llvm::SimplifyAShrInst(Value *Op0, Value *Op1, bool isExact,
const SimplifyQuery &Q) {
return ::SimplifyAShrInst(Op0, Op1, isExact, Q, RecursionLimit);
}
/// Commuted variants are assumed to be handled by calling this function again
/// with the parameters swapped.
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_Specific(Y), m_Value(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;
}
/// Commuted variants are assumed to be handled by calling this function again
/// with the parameters swapped.
static Value *simplifyAndOfICmpsWithSameOperands(ICmpInst *Op0, ICmpInst *Op1) {
ICmpInst::Predicate Pred0, Pred1;
Value *A ,*B;
if (!match(Op0, m_ICmp(Pred0, m_Value(A), m_Value(B))) ||
!match(Op1, m_ICmp(Pred1, m_Specific(A), m_Specific(B))))
return nullptr;
// We have (icmp Pred0, A, B) & (icmp Pred1, A, B).
// If Op1 is always implied true by Op0, then Op0 is a subset of Op1, and we
// can eliminate Op1 from this 'and'.
if (ICmpInst::isImpliedTrueByMatchingCmp(Pred0, Pred1))
return Op0;
// Check for any combination of predicates that are guaranteed to be disjoint.
if ((Pred0 == ICmpInst::getInversePredicate(Pred1)) ||
(Pred0 == ICmpInst::ICMP_EQ && ICmpInst::isFalseWhenEqual(Pred1)) ||
(Pred0 == ICmpInst::ICMP_SLT && Pred1 == ICmpInst::ICMP_SGT) ||
(Pred0 == ICmpInst::ICMP_ULT && Pred1 == ICmpInst::ICMP_UGT))
return getFalse(Op0->getType());
return nullptr;
}
/// Commuted variants are assumed to be handled by calling this function again
/// with the parameters swapped.
static Value *simplifyOrOfICmpsWithSameOperands(ICmpInst *Op0, ICmpInst *Op1) {
ICmpInst::Predicate Pred0, Pred1;
Value *A ,*B;
if (!match(Op0, m_ICmp(Pred0, m_Value(A), m_Value(B))) ||
!match(Op1, m_ICmp(Pred1, m_Specific(A), m_Specific(B))))
return nullptr;
// We have (icmp Pred0, A, B) | (icmp Pred1, A, B).
// If Op1 is always implied true by Op0, then Op0 is a subset of Op1, and we
// can eliminate Op0 from this 'or'.
if (ICmpInst::isImpliedTrueByMatchingCmp(Pred0, Pred1))
return Op1;
// Check for any combination of predicates that cover the entire range of
// possibilities.
if ((Pred0 == ICmpInst::getInversePredicate(Pred1)) ||
(Pred0 == ICmpInst::ICMP_NE && ICmpInst::isTrueWhenEqual(Pred1)) ||
(Pred0 == ICmpInst::ICMP_SLE && Pred1 == ICmpInst::ICMP_SGE) ||
(Pred0 == ICmpInst::ICMP_ULE && Pred1 == ICmpInst::ICMP_UGE))
return getTrue(Op0->getType());
return nullptr;
}
/// Test if a pair of compares with a shared operand and 2 constants has an
/// empty set intersection, full set union, or if one compare is a superset of
/// the other.
static Value *simplifyAndOrOfICmpsWithConstants(ICmpInst *Cmp0, ICmpInst *Cmp1,
bool IsAnd) {
// Look for this pattern: {and/or} (icmp X, C0), (icmp X, C1)).
if (Cmp0->getOperand(0) != Cmp1->getOperand(0))
return nullptr;
const APInt *C0, *C1;
if (!match(Cmp0->getOperand(1), m_APInt(C0)) ||
!match(Cmp1->getOperand(1), m_APInt(C1)))
return nullptr;
auto Range0 = ConstantRange::makeExactICmpRegion(Cmp0->getPredicate(), *C0);
auto Range1 = ConstantRange::makeExactICmpRegion(Cmp1->getPredicate(), *C1);
// For and-of-compares, check if the intersection is empty:
// (icmp X, C0) && (icmp X, C1) --> empty set --> false
if (IsAnd && Range0.intersectWith(Range1).isEmptySet())
return getFalse(Cmp0->getType());
// For or-of-compares, check if the union is full:
// (icmp X, C0) || (icmp X, C1) --> full set --> true
if (!IsAnd && Range0.unionWith(Range1).isFullSet())
return getTrue(Cmp0->getType());
// Is one range a superset of the other?
// If this is and-of-compares, take the smaller set:
// (icmp sgt X, 4) && (icmp sgt X, 42) --> icmp sgt X, 42
// If this is or-of-compares, take the larger set:
// (icmp sgt X, 4) || (icmp sgt X, 42) --> icmp sgt X, 4
if (Range0.contains(Range1))
return IsAnd ? Cmp1 : Cmp0;
if (Range1.contains(Range0))
return IsAnd ? Cmp0 : Cmp1;
return nullptr;
}
static Value *simplifyAndOrOfICmpsWithZero(ICmpInst *Cmp0, ICmpInst *Cmp1,
bool IsAnd) {
ICmpInst::Predicate P0 = Cmp0->getPredicate(), P1 = Cmp1->getPredicate();
if (!match(Cmp0->getOperand(1), m_Zero()) ||
!match(Cmp1->getOperand(1), m_Zero()) || P0 != P1)
return nullptr;
if ((IsAnd && P0 != ICmpInst::ICMP_NE) || (!IsAnd && P1 != ICmpInst::ICMP_EQ))
return nullptr;
// We have either "(X == 0 || Y == 0)" or "(X != 0 && Y != 0)".
Value *X = Cmp0->getOperand(0);
Value *Y = Cmp1->getOperand(0);
// If one of the compares is a masked version of a (not) null check, then
// that compare implies the other, so we eliminate the other. Optionally, look
// through a pointer-to-int cast to match a null check of a pointer type.
// (X == 0) || (([ptrtoint] X & ?) == 0) --> ([ptrtoint] X & ?) == 0
// (X == 0) || ((? & [ptrtoint] X) == 0) --> (? & [ptrtoint] X) == 0
// (X != 0) && (([ptrtoint] X & ?) != 0) --> ([ptrtoint] X & ?) != 0
// (X != 0) && ((? & [ptrtoint] X) != 0) --> (? & [ptrtoint] X) != 0
if (match(Y, m_c_And(m_Specific(X), m_Value())) ||
match(Y, m_c_And(m_PtrToInt(m_Specific(X)), m_Value())))
return Cmp1;
// (([ptrtoint] Y & ?) == 0) || (Y == 0) --> ([ptrtoint] Y & ?) == 0
// ((? & [ptrtoint] Y) == 0) || (Y == 0) --> (? & [ptrtoint] Y) == 0
// (([ptrtoint] Y & ?) != 0) && (Y != 0) --> ([ptrtoint] Y & ?) != 0
// ((? & [ptrtoint] Y) != 0) && (Y != 0) --> (? & [ptrtoint] Y) != 0
if (match(X, m_c_And(m_Specific(Y), m_Value())) ||
match(X, m_c_And(m_PtrToInt(m_Specific(Y)), m_Value())))
return Cmp0;
return nullptr;
}
static Value *simplifyAndOfICmpsWithAdd(ICmpInst *Op0, ICmpInst *Op1,
const InstrInfoQuery &IIQ) {
// (icmp (add V, C0), C1) & (icmp V, C0)
ICmpInst::Predicate Pred0, Pred1;
const APInt *C0, *C1;
Value *V;
if (!match(Op0, m_ICmp(Pred0, m_Add(m_Value(V), m_APInt(C0)), m_APInt(C1))))
return nullptr;
if (!match(Op1, m_ICmp(Pred1, m_Specific(V), m_Value())))
return nullptr;
auto *AddInst = cast<OverflowingBinaryOperator>(Op0->getOperand(0));
if (AddInst->getOperand(1) != Op1->getOperand(1))
return nullptr;
Type *ITy = Op0->getType();
bool isNSW = IIQ.hasNoSignedWrap(AddInst);
bool isNUW = IIQ.hasNoUnsignedWrap(AddInst);
const APInt Delta = *C1 - *C0;
if (C0->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 (C0->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;
}
static Value *simplifyAndOfICmps(ICmpInst *Op0, ICmpInst *Op1,
const InstrInfoQuery &IIQ) {
if (Value *X = simplifyUnsignedRangeCheck(Op0, Op1, /*IsAnd=*/true))
return X;
if (Value *X = simplifyUnsignedRangeCheck(Op1, Op0, /*IsAnd=*/true))
return X;
if (Value *X = simplifyAndOfICmpsWithSameOperands(Op0, Op1))
return X;
if (Value *X = simplifyAndOfICmpsWithSameOperands(Op1, Op0))
return X;
if (Value *X = simplifyAndOrOfICmpsWithConstants(Op0, Op1, true))
return X;
if (Value *X = simplifyAndOrOfICmpsWithZero(Op0, Op1, true))
return X;
if (Value *X = simplifyAndOfICmpsWithAdd(Op0, Op1, IIQ))
return X;
if (Value *X = simplifyAndOfICmpsWithAdd(Op1, Op0, IIQ))
return X;
return nullptr;
}
static Value *simplifyOrOfICmpsWithAdd(ICmpInst *Op0, ICmpInst *Op1,
const InstrInfoQuery &IIQ) {
// (icmp (add V, C0), C1) | (icmp V, C0)
ICmpInst::Predicate Pred0, Pred1;
const APInt *C0, *C1;
Value *V;
if (!match(Op0, m_ICmp(Pred0, m_Add(m_Value(V), m_APInt(C0)), m_APInt(C1))))
return nullptr;
if (!match(Op1, m_ICmp(Pred1, m_Specific(V), m_Value())))
return nullptr;
auto *AddInst = cast<BinaryOperator>(Op0->getOperand(0));
if (AddInst->getOperand(1) != Op1->getOperand(1))
return nullptr;
Type *ITy = Op0->getType();
bool isNSW = IIQ.hasNoSignedWrap(AddInst);
bool isNUW = IIQ.hasNoUnsignedWrap(AddInst);
const APInt Delta = *C1 - *C0;
if (C0->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 (C0->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;
}
static Value *simplifyOrOfICmps(ICmpInst *Op0, ICmpInst *Op1,
const InstrInfoQuery &IIQ) {
if (Value *X = simplifyUnsignedRangeCheck(Op0, Op1, /*IsAnd=*/false))
return X;
if (Value *X = simplifyUnsignedRangeCheck(Op1, Op0, /*IsAnd=*/false))
return X;
if (Value *X = simplifyOrOfICmpsWithSameOperands(Op0, Op1))
return X;
if (Value *X = simplifyOrOfICmpsWithSameOperands(Op1, Op0))
return X;
if (Value *X = simplifyAndOrOfICmpsWithConstants(Op0, Op1, false))
return X;
if (Value *X = simplifyAndOrOfICmpsWithZero(Op0, Op1, false))
return X;
if (Value *X = simplifyOrOfICmpsWithAdd(Op0, Op1, IIQ))
return X;
if (Value *X = simplifyOrOfICmpsWithAdd(Op1, Op0, IIQ))
return X;
return nullptr;
}
static Value *simplifyAndOrOfFCmps(const TargetLibraryInfo *TLI,
FCmpInst *LHS, FCmpInst *RHS, bool IsAnd) {
Value *LHS0 = LHS->getOperand(0), *LHS1 = LHS->getOperand(1);
Value *RHS0 = RHS->getOperand(0), *RHS1 = RHS->getOperand(1);
if (LHS0->getType() != RHS0->getType())
return nullptr;
FCmpInst::Predicate PredL = LHS->getPredicate(), PredR = RHS->getPredicate();
if ((PredL == FCmpInst::FCMP_ORD && PredR == FCmpInst::FCMP_ORD && IsAnd) ||
(PredL == FCmpInst::FCMP_UNO && PredR == FCmpInst::FCMP_UNO && !IsAnd)) {
// (fcmp ord NNAN, X) & (fcmp ord X, Y) --> fcmp ord X, Y
// (fcmp ord NNAN, X) & (fcmp ord Y, X) --> fcmp ord Y, X
// (fcmp ord X, NNAN) & (fcmp ord X, Y) --> fcmp ord X, Y
// (fcmp ord X, NNAN) & (fcmp ord Y, X) --> fcmp ord Y, X
// (fcmp uno NNAN, X) | (fcmp uno X, Y) --> fcmp uno X, Y
// (fcmp uno NNAN, X) | (fcmp uno Y, X) --> fcmp uno Y, X
// (fcmp uno X, NNAN) | (fcmp uno X, Y) --> fcmp uno X, Y
// (fcmp uno X, NNAN) | (fcmp uno Y, X) --> fcmp uno Y, X
if ((isKnownNeverNaN(LHS0, TLI) && (LHS1 == RHS0 || LHS1 == RHS1)) ||
(isKnownNeverNaN(LHS1, TLI) && (LHS0 == RHS0 || LHS0 == RHS1)))
return RHS;
// (fcmp ord X, Y) & (fcmp ord NNAN, X) --> fcmp ord X, Y
// (fcmp ord Y, X) & (fcmp ord NNAN, X) --> fcmp ord Y, X
// (fcmp ord X, Y) & (fcmp ord X, NNAN) --> fcmp ord X, Y
// (fcmp ord Y, X) & (fcmp ord X, NNAN) --> fcmp ord Y, X
// (fcmp uno X, Y) | (fcmp uno NNAN, X) --> fcmp uno X, Y
// (fcmp uno Y, X) | (fcmp uno NNAN, X) --> fcmp uno Y, X
// (fcmp uno X, Y) | (fcmp uno X, NNAN) --> fcmp uno X, Y
// (fcmp uno Y, X) | (fcmp uno X, NNAN) --> fcmp uno Y, X
if ((isKnownNeverNaN(RHS0, TLI) && (RHS1 == LHS0 || RHS1 == LHS1)) ||
(isKnownNeverNaN(RHS1, TLI) && (RHS0 == LHS0 || RHS0 == LHS1)))
return LHS;
}
return nullptr;
}
static Value *simplifyAndOrOfCmps(const SimplifyQuery &Q,
Value *Op0, Value *Op1, bool IsAnd) {
// Look through casts of the 'and' operands to find compares.
auto *Cast0 = dyn_cast<CastInst>(Op0);
auto *Cast1 = dyn_cast<CastInst>(Op1);
if (Cast0 && Cast1 && Cast0->getOpcode() == Cast1->getOpcode() &&
Cast0->getSrcTy() == Cast1->getSrcTy()) {
Op0 = Cast0->getOperand(0);
Op1 = Cast1->getOperand(0);
}
Value *V = nullptr;
auto *ICmp0 = dyn_cast<ICmpInst>(Op0);
auto *ICmp1 = dyn_cast<ICmpInst>(Op1);
if (ICmp0 && ICmp1)
V = IsAnd ? simplifyAndOfICmps(ICmp0, ICmp1, Q.IIQ)
: simplifyOrOfICmps(ICmp0, ICmp1, Q.IIQ);
auto *FCmp0 = dyn_cast<FCmpInst>(Op0);
auto *FCmp1 = dyn_cast<FCmpInst>(Op1);
if (FCmp0 && FCmp1)
V = simplifyAndOrOfFCmps(Q.TLI, FCmp0, FCmp1, IsAnd);
if (!V)
return nullptr;
if (!Cast0)
return V;
// If we looked through casts, we can only handle a constant simplification
// because we are not allowed to create a cast instruction here.
if (auto *C = dyn_cast<Constant>(V))
return ConstantExpr::getCast(Cast0->getOpcode(), C, Cast0->getType());
return nullptr;
}
/// 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 SimplifyQuery &Q,
unsigned MaxRecurse) {
if (Constant *C = foldOrCommuteConstant(Instruction::And, Op0, Op1, Q))
return C;
// X & undef -> 0
if (match(Op1, m_Undef()))
return Constant::getNullValue(Op0->getType());
// X & X = X
if (Op0 == Op1)
return Op0;
// X & 0 = 0
if (match(Op1, m_Zero()))
return Constant::getNullValue(Op0->getType());
// X & -1 = X
if (match(Op1, m_AllOnes()))
return Op0;
// A & ~A = ~A & A = 0
if (match(Op0, m_Not(m_Specific(Op1))) ||
match(Op1, m_Not(m_Specific(Op0))))
return Constant::getNullValue(Op0->getType());
// (A | ?) & A = A
if (match(Op0, m_c_Or(m_Specific(Op1), m_Value())))
return Op1;
// A & (A | ?) = A
if (match(Op1, m_c_Or(m_Specific(Op0), m_Value())))
return Op0;
// A mask that only clears known zeros of a shifted value is a no-op.
Value *X;
const APInt *Mask;
const APInt *ShAmt;
if (match(Op1, m_APInt(Mask))) {
// If all bits in the inverted and shifted mask are clear:
// and (shl X, ShAmt), Mask --> shl X, ShAmt
if (match(Op0, m_Shl(m_Value(X), m_APInt(ShAmt))) &&
(~(*Mask)).lshr(*ShAmt).isNullValue())
return Op0;
// If all bits in the inverted and shifted mask are clear:
// and (lshr X, ShAmt), Mask --> lshr X, ShAmt
if (match(Op0, m_LShr(m_Value(X), m_APInt(ShAmt))) &&
(~(*Mask)).shl(*ShAmt).isNullValue())
return Op0;
}
// 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, Q.DL, /*OrZero*/ true, 0, Q.AC, Q.CxtI,
Q.DT))
return Op0;
if (isKnownToBeAPowerOfTwo(Op1, Q.DL, /*OrZero*/ true, 0, Q.AC, Q.CxtI,
Q.DT))
return Op1;
}
if (Value *V = simplifyAndOrOfCmps(Q, Op0, Op1, true))
return V;
// Try some generic simplifications for associative operations.
if (Value *V = SimplifyAssociativeBinOp(Instruction::And, Op0, Op1, Q,
MaxRecurse))
return V;
// And distributes over Or. Try some generic simplifications based on this.
if (Value *V = ExpandBinOp(Instruction::And, Op0, Op1, Instruction::Or,
Q, MaxRecurse))
return V;
// And distributes over Xor. Try some generic simplifications based on this.
if (Value *V = ExpandBinOp(Instruction::And, Op0, Op1, Instruction::Xor,
Q, MaxRecurse))
return V;
// 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))
if (Value *V = ThreadBinOpOverSelect(Instruction::And, Op0, Op1, Q,
MaxRecurse))
return V;
// If the operation is with the result of a phi instruction, check whether
// operating on all incoming values of the phi always yields the same value.
if (isa<PHINode>(Op0) || isa<PHINode>(Op1))
if (Value *V = ThreadBinOpOverPHI(Instruction::And, Op0, Op1, Q,
MaxRecurse))
return V;
// Assuming the effective width of Y is not larger than A, i.e. all bits
// from X and Y are disjoint in (X << A) | Y,
// if the mask of this AND op covers all bits of X or Y, while it covers
// no bits from the other, we can bypass this AND op. E.g.,
// ((X << A) | Y) & Mask -> Y,
// if Mask = ((1 << effective_width_of(Y)) - 1)
// ((X << A) | Y) & Mask -> X << A,
// if Mask = ((1 << effective_width_of(X)) - 1) << A
// SimplifyDemandedBits in InstCombine can optimize the general case.
// This pattern aims to help other passes for a common case.
Value *Y, *XShifted;
if (match(Op1, m_APInt(Mask)) &&
match(Op0, m_c_Or(m_CombineAnd(m_NUWShl(m_Value(X), m_APInt(ShAmt)),
m_Value(XShifted)),
m_Value(Y)))) {
const unsigned Width = Op0->getType()->getScalarSizeInBits();
const unsigned ShftCnt = ShAmt->getLimitedValue(Width);
const KnownBits YKnown = computeKnownBits(Y, Q.DL, 0, Q.AC, Q.CxtI, Q.DT);
const unsigned EffWidthY = Width - YKnown.countMinLeadingZeros();
if (EffWidthY <= ShftCnt) {
const KnownBits XKnown = computeKnownBits(X, Q.DL, 0, Q.AC, Q.CxtI,
Q.DT);
const unsigned EffWidthX = Width - XKnown.countMinLeadingZeros();
const APInt EffBitsY = APInt::getLowBitsSet(Width, EffWidthY);
const APInt EffBitsX = APInt::getLowBitsSet(Width, EffWidthX) << ShftCnt;
// If the mask is extracting all bits from X or Y as is, we can skip
// this AND op.
if (EffBitsY.isSubsetOf(*Mask) && !EffBitsX.intersects(*Mask))
return Y;
if (EffBitsX.isSubsetOf(*Mask) && !EffBitsY.intersects(*Mask))
return XShifted;
}
}
return nullptr;
}
Value *llvm::SimplifyAndInst(Value *Op0, Value *Op1, const SimplifyQuery &Q) {
return ::SimplifyAndInst(Op0, Op1, Q, RecursionLimit);
}
/// Given operands for an Or, see if we can fold the result.
/// If not, this returns null.
static Value *SimplifyOrInst(Value *Op0, Value *Op1, const SimplifyQuery &Q,
unsigned MaxRecurse) {
if (Constant *C = foldOrCommuteConstant(Instruction::Or, Op0, Op1, Q))
return C;
// X | undef -> -1
// X | -1 = -1
// Do not return Op1 because it may contain undef elements if it's a vector.
if (match(Op1, m_Undef()) || match(Op1, m_AllOnes()))
return Constant::getAllOnesValue(Op0->getType());
// X | X = X
// X | 0 = X
if (Op0 == Op1 || match(Op1, m_Zero()))
return Op0;
// A | ~A = ~A | A = -1
if (match(Op0, m_Not(m_Specific(Op1))) ||
match(Op1, m_Not(m_Specific(Op0))))
return Constant::getAllOnesValue(Op0->getType());
// (A & ?) | A = A
if (match(Op0, m_c_And(m_Specific(Op1), m_Value())))
return Op1;
// A | (A & ?) = A
if (match(Op1, m_c_And(m_Specific(Op0), m_Value())))
return Op0;
// ~(A & ?) | A = -1
if (match(Op0, m_Not(m_c_And(m_Specific(Op1), m_Value()))))
return Constant::getAllOnesValue(Op1->getType());
// A | ~(A & ?) = -1
if (match(Op1, m_Not(m_c_And(m_Specific(Op1), m_Value()))))
return Constant::getAllOnesValue(Op0->getType());
Value *A, *B;
// (A & ~B) | (A ^ B) -> (A ^ B)
// (~B & A) | (A ^ B) -> (A ^ B)
// (A & ~B) | (B ^ A) -> (B ^ A)
// (~B & A) | (B ^ A) -> (B ^ A)
if (match(Op1, m_Xor(m_Value(A), m_Value(B))) &&
(match(Op0, m_c_And(m_Specific(A), m_Not(m_Specific(B)))) ||
match(Op0, m_c_And(m_Not(m_Specific(A)), m_Specific(B)))))
return Op1;
// Commute the 'or' operands.
// (A ^ B) | (A & ~B) -> (A ^ B)
// (A ^ B) | (~B & A) -> (A ^ B)
// (B ^ A) | (A & ~B) -> (B ^ A)
// (B ^ A) | (~B & A) -> (B ^ A)
if (match(Op0, m_Xor(m_Value(A), m_Value(B))) &&
(match(Op1, m_c_And(m_Specific(A), m_Not(m_Specific(B)))) ||
match(Op1, m_c_And(m_Not(m_Specific(A)), m_Specific(B)))))
return Op0;
// (A & B) | (~A ^ B) -> (~A ^ B)
// (B & A) | (~A ^ B) -> (~A ^ B)
// (A & B) | (B ^ ~A) -> (B ^ ~A)
// (B & A) | (B ^ ~A) -> (B ^ ~A)
if (match(Op0, m_And(m_Value(A), m_Value(B))) &&
(match(Op1, m_c_Xor(m_Specific(A), m_Not(m_Specific(B)))) ||
match(Op1, m_c_Xor(m_Not(m_Specific(A)), m_Specific(B)))))
return Op1;
// (~A ^ B) | (A & B) -> (~A ^ B)
// (~A ^ B) | (B & A) -> (~A ^ B)
// (B ^ ~A) | (A & B) -> (B ^ ~A)
// (B ^ ~A) | (B & A) -> (B ^ ~A)
if (match(Op1, m_And(m_Value(A), m_Value(B))) &&
(match(Op0, m_c_Xor(m_Specific(A), m_Not(m_Specific(B)))) ||
match(Op0, m_c_Xor(m_Not(m_Specific(A)), m_Specific(B)))))
return Op0;
if (Value *V = simplifyAndOrOfCmps(Q, Op0, Op1, false))
return V;
// Try some generic simplifications for associative operations.
if (Value *V = SimplifyAssociativeBinOp(Instruction::Or, Op0, Op1, Q,
MaxRecurse))
return V;
// Or distributes over And. Try some generic simplifications based on this.
if (Value *V = ExpandBinOp(Instruction::Or, Op0, Op1, Instruction::And, Q,
MaxRecurse))
return V;
// 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))
if (Value *V = ThreadBinOpOverSelect(Instruction::Or, Op0, Op1, Q,
MaxRecurse))
return V;
// (A & C1)|(B & C2)
const APInt *C1, *C2;
if (match(Op0, m_And(m_Value(A), m_APInt(C1))) &&
match(Op1, m_And(m_Value(B), m_APInt(C2)))) {
if (*C1 == ~*C2) {
// (A & C1)|(B & C2)
// If we have: ((V + N) & C1) | (V & C2)
// .. and C2 = ~C1 and C2 is 0+1+ and (N & C2) == 0
// replace with V+N.
Value *N;
if (C2->isMask() && // C2 == 0+1+
match(A, m_c_Add(m_Specific(B), m_Value(N)))) {
// Add commutes, try both ways.
if (MaskedValueIsZero(N, *C2, Q.DL, 0, Q.AC, Q.CxtI, Q.DT))
return A;
}
// Or commutes, try both ways.
if (C1->isMask() &&
match(B, m_c_Add(m_Specific(A), m_Value(N)))) {
// Add commutes, try both ways.
if (MaskedValueIsZero(N, *C1, Q.DL, 0, Q.AC, Q.CxtI, Q.DT))
return B;
}
}
}
// If the operation is with the result of a phi instruction, check whether
// operating on all incoming values of the phi always yields the same value.
if (isa<PHINode>(Op0) || isa<PHINode>(Op1))
if (Value *V = ThreadBinOpOverPHI(Instruction::Or, Op0, Op1, Q, MaxRecurse))
return V;
return nullptr;
}
Value *llvm::SimplifyOrInst(Value *Op0, Value *Op1, const SimplifyQuery &Q) {
return ::SimplifyOrInst(Op0, Op1, Q, RecursionLimit);
}
/// Given operands for a Xor, see if we can fold the result.
/// If not, this returns null.
static Value *SimplifyXorInst(Value *Op0, Value *Op1, const SimplifyQuery &Q,
unsigned MaxRecurse) {
if (Constant *C = foldOrCommuteConstant(Instruction::Xor, Op0, Op1, Q))
return C;
// A ^ undef -> undef
if (match(Op1, m_Undef()))
return Op1;
// A ^ 0 = A
if (match(Op1, m_Zero()))
return Op0;
// A ^ A = 0
if (Op0 == Op1)
return Constant::getNullValue(Op0->getType());
// A ^ ~A = ~A ^ A = -1
if (match(Op0, m_Not(m_Specific(Op1))) ||
match(Op1, m_Not(m_Specific(Op0))))
return Constant::getAllOnesValue(Op0->getType());
// Try some generic simplifications for associative operations.
if (Value *V = SimplifyAssociativeBinOp(Instruction::Xor, Op0, Op1, Q,
MaxRecurse))
return V;
// Threading Xor over selects and phi nodes is pointless, so don't bother.
// Threading over the select in "A ^ select(cond, B, C)" means evaluating
// "A^B" and "A^C" and seeing if they are equal; but they are equal if and
// only if B and C are equal. If B and C are equal then (since we assume
// that operands have already been simplified) "select(cond, B, C)" should
// have been simplified to the common value of B and C already. Analysing
// "A^B" and "A^C" thus gains nothing, but costs compile time. Similarly
// for threading over phi nodes.
return nullptr;
}
Value *llvm::SimplifyXorInst(Value *Op0, Value *Op1, const SimplifyQuery &Q) {
return ::SimplifyXorInst(Op0, Op1, Q, RecursionLimit);
}
static Type *GetCompareTy(Value *Op) {
return CmpInst::makeCmpResultType(Op->getType());
}
/// Rummage around inside V looking for something equivalent to the comparison
/// "LHS Pred RHS". Return such a value if found, otherwise return null.
/// Helper function for analyzing max/min idioms.
static Value *ExtractEquivalentCondition(Value *V, CmpInst::Predicate Pred,
Value *LHS, Value *RHS) {
SelectInst *SI = dyn_cast<SelectInst>(V);
if (!SI)
return nullptr;
CmpInst *Cmp = dyn_cast<CmpInst>(SI->getCondition());
if (!Cmp)
return nullptr;
Value *CmpLHS = Cmp->getOperand(0), *CmpRHS = Cmp->getOperand(1);
if (Pred == Cmp->getPredicate() && LHS == CmpLHS && RHS == CmpRHS)
return Cmp;
if (Pred == CmpInst::getSwappedPredicate(Cmp->getPredicate()) &&
LHS == CmpRHS && RHS == CmpLHS)
return Cmp;
return nullptr;
}
// A significant optimization not implemented here is assuming that alloca
// addresses are not equal to incoming argument values. They don't *alias*,
// as we say, but that doesn't mean they aren't equal, so we take a
// conservative approach.
//
// This is inspired in part by C++11 5.10p1:
// "Two pointers of the same type compare equal if and only if they are both
// null, both point to the same function, or both represent the same
// address."
//
// This is pretty permissive.
//
// It's also partly due to C11 6.5.9p6:
// "Two pointers compare equal if and only if both are null pointers, both are
// pointers to the same object (including a pointer to an object and a
// subobject at its beginning) or function, both are pointers to one past the
// last element of the same array object, or one is a pointer to one past the
// end of one array object and the other is a pointer to the start of a
// different array object that happens to immediately follow the first array
// object in the address space.)
//
// C11's version is more restrictive, however there's no reason why an argument
// couldn't be a one-past-the-end value for a stack object in the caller and be
// equal to the beginning of a stack object in the callee.
//
// 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, const TargetLibraryInfo *TLI,
const DominatorTree *DT, CmpInst::Predicate Pred,
AssumptionCache *AC, const Instruction *CxtI,
const InstrInfoQuery &IIQ, Value *LHS, Value *RHS) {
// First, skip past any trivial no-ops.
LHS = LHS->stripPointerCasts();
RHS = RHS->stripPointerCasts();
// A non-null pointer is not equal to a null pointer.
if (llvm::isKnownNonZero(LHS, DL, 0, nullptr, nullptr, nullptr,
IIQ.UseInstrInfo) &&
isa<ConstantPointerNull>(RHS) &&
(Pred == CmpInst::ICMP_EQ || Pred == CmpInst::ICMP_NE))
return ConstantInt::get(GetCompareTy(LHS),
!CmpInst::isTrueWhenEqual(Pred));
// We can only fold certain predicates on pointer comparisons.
switch (Pred) {
default:
return nullptr;
// Equality comaprisons are easy to fold.
case CmpInst::ICMP_EQ:
case CmpInst::ICMP_NE:
break;
// We can only handle unsigned relational comparisons because 'inbounds' on
// a GEP only protects against unsigned wrapping.
case CmpInst::ICMP_UGT:
case CmpInst::ICMP_UGE:
case CmpInst::ICMP_ULT:
case CmpInst::ICMP_ULE:
// However, we have to switch them to their signed variants to handle
// negative indices from the base pointer.
Pred = ICmpInst::getSignedPredicate(Pred);
break;
}
// Strip off any constant offsets so that we can reason about them.
// It's tempting to use getUnderlyingObject or even just stripInBoundsOffsets
// here and compare base addresses like AliasAnalysis does, however there are
// numerous hazards. AliasAnalysis and its utilities rely on special rules
// governing loads and stores which don't apply to icmps. Also, AliasAnalysis
// doesn't need to guarantee pointer inequality when it says NoAlias.
Constant *LHSOffset = stripAndComputeConstantOffsets(DL, LHS);
Constant *RHSOffset = stripAndComputeConstantOffsets(DL, RHS);
// If LHS and RHS are related via constant offsets to the same base
// value, we can replace it with an icmp which just compares the offsets.
if (LHS == RHS)
return ConstantExpr::getICmp(Pred, LHSOffset, RHSOffset);
// Various optimizations for (in)equality comparisons.
if (Pred == CmpInst::ICMP_EQ || Pred == CmpInst::ICMP_NE) {
// Different non-empty allocations that exist at the same time have
// different addresses (if the program can tell). Global variables always
// exist, so they always exist during the lifetime of each other and all
// allocas. Two different allocas usually have different addresses...
//
// However, if there's an @llvm.stackrestore dynamically in between two
// allocas, they may have the same address. It's tempting to reduce the
// scope of the problem by only looking at *static* allocas here. That would
// cover the majority of allocas while significantly reducing the likelihood
// of having an @llvm.stackrestore pop up in the middle. However, it's not
// actually impossible for an @llvm.stackrestore to pop up in the middle of
// an entry block. Also, if we have a block that's not attached to a
// function, we can't tell if it's "static" under the current definition.
// Theoretically, this problem could be fixed by creating a new kind of
// instruction kind specifically for static allocas. Such a new instruction
// could be required to be at the top of the entry block, thus preventing it
// from being subject to a @llvm.stackrestore. Instcombine could even
// convert regular allocas into these special allocas. It'd be nifty.
// However, until then, this problem remains open.
//
// So, we'll assume that two non-empty allocas have different addresses
// for now.
//
// With all that, if the offsets are within the bounds of their allocations
// (and not one-past-the-end! so we can't use inbounds!), and their
// allocations aren't the same, the pointers are not equal.
//
// Note that it's not necessary to check for LHS being a global variable
// address, due to canonicalization and constant folding.
if (isa<AllocaInst>(LHS) &&
(isa<AllocaInst>(RHS) || isa<GlobalVariable>(RHS))) {
ConstantInt *LHSOffsetCI = dyn_cast<ConstantInt>(LHSOffset);
ConstantInt *RHSOffsetCI = dyn_cast<ConstantInt>(RHSOffset);
uint64_t LHSSize, RHSSize;
ObjectSizeOpts Opts;
Opts.NullIsUnknownSize =
NullPointerIsDefined(cast<AllocaInst>(LHS)->getFunction());
if (LHSOffsetCI && RHSOffsetCI &&
getObjectSize(LHS, LHSSize, DL, TLI, Opts) &&
getObjectSize(RHS, RHSSize, DL, TLI, Opts)) {
const APInt &LHSOffsetValue = LHSOffsetCI->getValue();
const APInt &RHSOffsetValue = RHSOffsetCI->getValue();
if (!LHSOffsetValue.isNegative() &&
!RHSOffsetValue.isNegative() &&
LHSOffsetValue.ult(LHSSize) &&
RHSOffsetValue.ult(RHSSize)) {
return ConstantInt::get(GetCompareTy(LHS),
!CmpInst::isTrueWhenEqual(Pred));
}
}
// Repeat the above check but this time without depending on DataLayout
// or being able to compute a precise size.
if (!cast<PointerType>(LHS->getType())->isEmptyTy() &&
!cast<PointerType>(RHS->getType())->isEmptyTy() &&
LHSOffset->isNullValue() &&
RHSOffset->isNullValue())
return ConstantInt::get(GetCompareTy(LHS),
!CmpInst::isTrueWhenEqual(Pred));
}
// Even if an non-inbounds GEP occurs along the path we can still optimize
// equality comparisons concerning the result. We avoid walking the whole
// chain again by starting where the last calls to
// stripAndComputeConstantOffsets left off and accumulate the offsets.
Constant *LHSNoBound = stripAndComputeConstantOffsets(DL, LHS, true);
Constant *RHSNoBound = stripAndComputeConstantOffsets(DL, RHS, true);
if (LHS == RHS)
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<const Value *, 8> LHSUObjs, RHSUObjs;
GetUnderlyingObjects(LHS, LHSUObjs, DL);
GetUnderlyingObjects(RHS, RHSUObjs, DL);
// Is the set of underlying objects all noalias calls?
auto IsNAC = [](ArrayRef<const Value *> Objects) {
return all_of(Objects, isNoAliasCall);
};
// 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 = [](ArrayRef<const Value *> Objects) {
return all_of(Objects, [](const Value *V) {
if (const AllocaInst *AI = dyn_cast<AllocaInst>(V))
return AI->getParent() && AI->getFunction() && AI->isStaticAlloca();
if (const GlobalValue *GV = dyn_cast<GlobalValue>(V))
return (GV->hasLocalLinkage() || GV->hasHiddenVisibility() ||
GV->hasProtectedVisibility() || GV->hasGlobalUnnamedAddr()) &&
!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));
// Fold comparisons for non-escaping pointer even if the allocation call
// cannot be elided. We cannot fold malloc comparison to null. Also, the
// dynamic allocation call could be either of the operands.
Value *MI = nullptr;
if (isAllocLikeFn(LHS, TLI) &&
llvm::isKnownNonZero(RHS, DL, 0, nullptr, CxtI, DT))
MI = LHS;
else if (isAllocLikeFn(RHS, TLI) &&
llvm::isKnownNonZero(LHS, DL, 0, nullptr, CxtI, DT))
MI = RHS;
// FIXME: We should also fold the compare when the pointer escapes, but the
// compare dominates the pointer escape
if (MI && !PointerMayBeCaptured(MI, true, true))
return ConstantInt::get(GetCompareTy(LHS),
CmpInst::isFalseWhenEqual(Pred));
}
// Otherwise, fail.
return