diff options
Diffstat (limited to 'lib/Analysis/InstructionSimplify.cpp')
-rw-r--r-- | lib/Analysis/InstructionSimplify.cpp | 877 |
1 files changed, 692 insertions, 185 deletions
diff --git a/lib/Analysis/InstructionSimplify.cpp b/lib/Analysis/InstructionSimplify.cpp index 7a820a58f6f0..3fbbd7cbd22a 100644 --- a/lib/Analysis/InstructionSimplify.cpp +++ b/lib/Analysis/InstructionSimplify.cpp @@ -20,6 +20,7 @@ #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" @@ -31,6 +32,7 @@ #include "llvm/IR/Operator.h" #include "llvm/IR/PatternMatch.h" #include "llvm/IR/ValueHandle.h" +#include <algorithm> using namespace llvm; using namespace llvm::PatternMatch; @@ -41,14 +43,20 @@ enum { RecursionLimit = 3 }; STATISTIC(NumExpand, "Number of expansions"); STATISTIC(NumReassoc, "Number of reassociations"); +namespace { struct Query { 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) {} + 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 &, @@ -575,8 +583,9 @@ static Value *SimplifyAddInst(Value *Op0, Value *Op1, bool isNSW, bool isNUW, 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 DominatorTree *DT, AssumptionCache *AC, + const Instruction *CxtI) { + return ::SimplifyAddInst(Op0, Op1, isNSW, isNUW, Query(DL, TLI, DT, AC, CxtI), RecursionLimit); } @@ -624,7 +633,7 @@ static Constant *stripAndComputeConstantOffsets(const DataLayout *DL, } 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()) @@ -676,6 +685,10 @@ static Value *SimplifySubInst(Value *Op0, Value *Op1, bool isNSW, bool isNUW, 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; @@ -769,8 +782,9 @@ static Value *SimplifySubInst(Value *Op0, Value *Op1, bool isNSW, bool isNUW, 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 DominatorTree *DT, AssumptionCache *AC, + const Instruction *CxtI) { + return ::SimplifySubInst(Op0, Op1, isNSW, isNUW, Query(DL, TLI, DT, AC, CxtI), RecursionLimit); } @@ -946,29 +960,38 @@ static Value *SimplifyMulInst(Value *Op0, Value *Op1, const Query &Q, } 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, +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, 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 @@ -988,6 +1011,10 @@ static Value *SimplifyDiv(Instruction::BinaryOps Opcode, Value *Op0, Value *Op1, 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()); @@ -1028,6 +1055,16 @@ static Value *SimplifyDiv(Instruction::BinaryOps Opcode, Value *Op0, Value *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; + 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)) @@ -1055,8 +1092,10 @@ static Value *SimplifySDivInst(Value *Op0, Value *Op1, const Query &Q, 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 @@ -1071,8 +1110,10 @@ static Value *SimplifyUDivInst(Value *Op0, Value *Op1, const Query &Q, 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, @@ -1090,8 +1131,10 @@ static Value *SimplifyFDivInst(Value *Op0, Value *Op1, const Query &Q, Value *llvm::SimplifyFDivInst(Value *Op0, Value *Op1, 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, Query(DL, TLI, DT, AC, CxtI), + RecursionLimit); } /// SimplifyRem - Given operands for an SRem or URem, see if we can @@ -1133,6 +1176,13 @@ static Value *SimplifyRem(Instruction::BinaryOps Opcode, Value *Op0, Value *Op1, 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)) @@ -1160,8 +1210,10 @@ static Value *SimplifySRemInst(Value *Op0, Value *Op1, const Query &Q, 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 @@ -1176,8 +1228,10 @@ static Value *SimplifyURemInst(Value *Op0, Value *Op1, const Query &Q, 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 &, @@ -1195,8 +1249,10 @@ static Value *SimplifyFRemInst(Value *Op0, Value *Op1, const Query &, Value *llvm::SimplifyFRemInst(Value *Op0, Value *Op1, 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, Query(DL, TLI, DT, AC, CxtI), + RecursionLimit); } /// isUndefShift - Returns true if a shift by \c Amount always yields undef. @@ -1264,6 +1320,37 @@ static Value *SimplifyShift(unsigned Opcode, Value *Op0, Value *Op1, 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, @@ -1272,8 +1359,9 @@ 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; @@ -1284,8 +1372,9 @@ static Value *SimplifyShlInst(Value *Op0, Value *Op1, bool isNSW, bool isNUW, 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 DominatorTree *DT, AssumptionCache *AC, + const Instruction *CxtI) { + return ::SimplifyShlInst(Op0, Op1, isNSW, isNUW, Query(DL, TLI, DT, AC, CxtI), RecursionLimit); } @@ -1293,21 +1382,13 @@ Value *llvm::SimplifyShlInst(Value *Op0, Value *Op1, bool isNSW, bool isNUW, /// 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; @@ -1316,8 +1397,9 @@ static Value *SimplifyLShrInst(Value *Op0, Value *Op1, bool isExact, Value *llvm::SimplifyLShrInst(Value *Op0, Value *Op1, bool isExact, 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); } @@ -1325,29 +1407,21 @@ Value *llvm::SimplifyLShrInst(Value *Op0, Value *Op1, bool isExact, /// 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, Q.DL); + unsigned NumSignBits = ComputeNumSignBits(Op0, Q.DL, 0, Q.AC, Q.CxtI, Q.DT); if (NumSignBits == Op0->getType()->getScalarSizeInBits()) return Op0; @@ -1357,11 +1431,105 @@ static Value *SimplifyAShrInst(Value *Op0, Value *Op1, bool isExact, Value *llvm::SimplifyAShrInst(Value *Op0, Value *Op1, bool isExact, 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, @@ -1412,12 +1580,21 @@ 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, /*OrZero*/ true, 0, Q.AC, Q.CxtI, Q.DT)) return Op0; - if (isKnownToBeAPowerOfTwo(Op1, /*OrZero*/true)) + if (isKnownToBeAPowerOfTwo(Op1, /*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)) @@ -1452,8 +1629,62 @@ static Value *SimplifyAndInst(Value *Op0, Value *Op1, const Query &Q, 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 @@ -1513,6 +1744,15 @@ static Value *SimplifyOrInst(Value *Op0, Value *Op1, const Query &Q, (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)) @@ -1545,18 +1785,22 @@ static Value *SimplifyOrInst(Value *Op0, Value *Op1, const Query &Q, 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; } } @@ -1573,8 +1817,10 @@ static Value *SimplifyOrInst(Value *Op0, Value *Op1, const Query &Q, 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 @@ -1628,8 +1874,10 @@ static Value *SimplifyXorInst(Value *Op0, Value *Op1, const Query &Q, 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) { @@ -1804,6 +2052,50 @@ static Constant *computePointerICmp(const DataLayout *DL, 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. @@ -1883,40 +2175,46 @@ static Value *SimplifyICmpInst(unsigned Predicate, Value *LHS, Value *RHS, 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; } @@ -1981,6 +2279,22 @@ static Value *SimplifyICmpInst(unsigned Predicate, Value *LHS, Value *RHS, 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); @@ -2189,25 +2503,6 @@ static Value *SimplifyICmpInst(unsigned Predicate, Value *LHS, Value *RHS, } } - // 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); @@ -2271,6 +2566,40 @@ static Value *SimplifyICmpInst(unsigned Predicate, Value *LHS, Value *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)) { @@ -2301,7 +2630,8 @@ static Value *SimplifyICmpInst(unsigned Predicate, Value *LHS, Value *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 @@ -2311,7 +2641,8 @@ static Value *SimplifyICmpInst(unsigned Predicate, Value *LHS, Value *RHS, 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 @@ -2330,7 +2661,8 @@ static Value *SimplifyICmpInst(unsigned Predicate, Value *LHS, Value *RHS, 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 @@ -2340,7 +2672,8 @@ static Value *SimplifyICmpInst(unsigned Predicate, Value *LHS, Value *RHS, 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 @@ -2360,6 +2693,41 @@ static Value *SimplifyICmpInst(unsigned Predicate, Value *LHS, Value *RHS, 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()) { @@ -2607,6 +2975,23 @@ static Value *SimplifyICmpInst(unsigned Predicate, Value *LHS, Value *RHS, } } + // 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)) @@ -2625,8 +3010,9 @@ static Value *SimplifyICmpInst(unsigned Predicate, Value *LHS, Value *RHS, Value *llvm::SimplifyICmpInst(unsigned Predicate, Value *LHS, Value *RHS, 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); } @@ -2722,8 +3108,9 @@ static Value *SimplifyFCmpInst(unsigned Predicate, Value *LHS, Value *RHS, Value *llvm::SimplifyFCmpInst(unsigned Predicate, Value *LHS, Value *RHS, const DataLayout *DL, const TargetLibraryInfo *TLI, - const DominatorTree *DT) { - return ::SimplifyFCmpInst(Predicate, LHS, RHS, Query (DL, TLI, DT), + const DominatorTree *DT, AssumptionCache *AC, + const Instruction *CxtI) { + return ::SimplifyFCmpInst(Predicate, LHS, RHS, Query(DL, TLI, DT, AC, CxtI), RecursionLimit); } @@ -2755,15 +3142,71 @@ static Value *SimplifySelectInst(Value *CondVal, Value *TrueVal, if (isa<UndefValue>(FalseVal)) // select C, X, undef -> X return TrueVal; + const auto *ICI = dyn_cast<ICmpInst>(CondVal); + unsigned BitWidth = TrueVal->getType()->getScalarSizeInBits(); + if (ICI && BitWidth) { + ICmpInst::Predicate Pred = ICI->getPredicate(); + APInt MinSignedValue = APInt::getSignBit(BitWidth); + Value *X; + const APInt *Y; + bool TrueWhenUnset; + bool IsBitTest = false; + if (ICmpInst::isEquality(Pred) && + match(ICI->getOperand(0), m_And(m_Value(X), m_APInt(Y))) && + match(ICI->getOperand(1), m_Zero())) { + IsBitTest = true; + TrueWhenUnset = Pred == ICmpInst::ICMP_EQ; + } else if (Pred == ICmpInst::ICMP_SLT && + match(ICI->getOperand(1), m_Zero())) { + X = ICI->getOperand(0); + Y = &MinSignedValue; + IsBitTest = true; + TrueWhenUnset = false; + } else if (Pred == ICmpInst::ICMP_SGT && + match(ICI->getOperand(1), m_AllOnes())) { + X = ICI->getOperand(0); + 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; + } + } + } + return nullptr; } Value *llvm::SimplifySelectInst(Value *Cond, Value *TrueVal, Value *FalseVal, 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 @@ -2771,29 +3214,72 @@ Value *llvm::SimplifySelectInst(Value *Cond, Value *TrueVal, Value *FalseVal, static Value *SimplifyGEPInst(ArrayRef<Value *> Ops, const Query &Q, unsigned) { // The type of the GEP pointer operand. PointerType *PtrTy = cast<PointerType>(Ops[0]->getType()->getScalarType()); + unsigned AS = PtrTy->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(PtrTy, 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 = PtrTy->getElementType(); + if (Q.DL && 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; + } } } @@ -2807,8 +3293,9 @@ static Value *SimplifyGEPInst(ArrayRef<Value *> Ops, const Query &Q, unsigned) { 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(Ops, Query(DL, TLI, DT, AC, CxtI), RecursionLimit); } /// SimplifyInsertValueInst - Given operands for an InsertValueInst, see if we @@ -2840,12 +3327,11 @@ static Value *SimplifyInsertValueInst(Value *Agg, Value *Val, 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); } @@ -2892,8 +3378,10 @@ static Value *SimplifyTruncInst(Value *Op, Type *Ty, const Query &Q, unsigned) { 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. @@ -2965,8 +3453,10 @@ static Value *SimplifyBinOp(unsigned Opcode, Value *LHS, Value *RHS, 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 DominatorTree *DT, AssumptionCache *AC, + const Instruction *CxtI) { + return ::SimplifyBinOp(Opcode, LHS, RHS, Query(DL, TLI, DT, AC, CxtI), + RecursionLimit); } /// SimplifyCmpInst - Given operands for a CmpInst, see if we can @@ -2980,8 +3470,9 @@ static Value *SimplifyCmpInst(unsigned Predicate, Value *LHS, Value *RHS, 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 DominatorTree *DT, AssumptionCache *AC, + const Instruction *CxtI) { + return ::SimplifyCmpInst(Predicate, LHS, RHS, Query(DL, TLI, DT, AC, CxtI), RecursionLimit); } @@ -3055,24 +3546,25 @@ static Value *SimplifyCall(Value *V, IterTy ArgBegin, IterTy ArgEnd, 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), + 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 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, const TargetLibraryInfo *TLI, - const DominatorTree *DT) { + const DominatorTree *DT, AssumptionCache *AC) { Value *Result; switch (I->getOpcode()) { @@ -3081,109 +3573,122 @@ Value *llvm::SimplifyInstruction(Instruction *I, const DataLayout *DL, 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), 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), 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); + Result = + SimplifyFCmpInst(cast<FCmpInst>(I)->getPredicate(), I->getOperand(0), + I->getOperand(1), 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::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; } @@ -3207,7 +3712,8 @@ Value *llvm::SimplifyInstruction(Instruction *I, const DataLayout *DL, 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; @@ -3234,7 +3740,7 @@ static bool replaceAndRecursivelySimplifyImpl(Instruction *I, Value *SimpleV, I = Worklist[Idx]; // See if this instruction simplifies. - SimpleV = SimplifyInstruction(I, DL, TLI, DT); + SimpleV = SimplifyInstruction(I, DL, TLI, DT, AC); if (!SimpleV) continue; @@ -3257,18 +3763,19 @@ static bool replaceAndRecursivelySimplifyImpl(Instruction *I, Value *SimpleV, return Simplified; } -bool llvm::recursivelySimplifyInstruction(Instruction *I, - const DataLayout *DL, +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, DL, 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, DL, TLI, DT, AC); } |