|
LLVM 23.0.0git
|
#include "llvm/Support/KnownFPClass.h"
Public Types | |
| enum class | MinMaxKind { minimum , maximum , minimumnum , maximumnum , minnum , maxnum } |
Public Member Functions | |
| KnownFPClass (FPClassTest Known=fcAllFlags, std::optional< bool > Sign={}) | |
| LLVM_ABI | KnownFPClass (const APFloat &C) |
| bool | operator== (KnownFPClass Other) const |
| bool | isKnownNever (FPClassTest Mask) const |
| Return true if it's known this can never be one of the mask entries. | |
| bool | isKnownAlways (FPClassTest Mask) const |
| bool | isUnknown () const |
| bool | isKnownNeverNaN () const |
| Return true if it's known this can never be a nan. | |
| bool | isKnownAlwaysNaN () const |
| Return true if it's known this must always be a nan. | |
| bool | isKnownNeverInfinity () const |
| Return true if it's known this can never be an infinity. | |
| bool | isKnownNeverInfOrNaN () const |
| Return true if it's known this can never be an infinity or nan. | |
| bool | isKnownNeverPosInfinity () const |
| Return true if it's known this can never be +infinity. | |
| bool | isKnownNeverNegInfinity () const |
| Return true if it's known this can never be -infinity. | |
| bool | isKnownNeverSubnormal () const |
| Return true if it's known this can never be a subnormal. | |
| bool | isKnownNeverPosSubnormal () const |
| Return true if it's known this can never be a positive subnormal. | |
| bool | isKnownNeverNegSubnormal () const |
| Return true if it's known this can never be a negative subnormal. | |
| bool | isKnownNeverZero () const |
| Return true if it's known this can never be a zero. | |
| bool | isKnownNeverPosZero () const |
| Return true if it's known this can never be a literal positive zero. | |
| bool | isKnownNeverNegZero () const |
| Return true if it's known this can never be a negative zero. | |
| LLVM_ABI bool | isKnownNeverLogicalZero (DenormalMode Mode) const |
| Return true if it's known this can never be interpreted as a zero. | |
| LLVM_ABI bool | isKnownNeverLogicalNegZero (DenormalMode Mode) const |
| Return true if it's known this can never be interpreted as a negative zero. | |
| LLVM_ABI bool | isKnownNeverLogicalPosZero (DenormalMode Mode) const |
| Return true if it's known this can never be interpreted as a positive zero. | |
| bool | cannotBeOrderedLessThanZero () const |
| Return true if we can prove that the analyzed floating-point value is either NaN or never less than -0.0. | |
| bool | cannotBeOrderedGreaterThanZero () const |
| Return true if we can prove that the analyzed floating-point value is either NaN or never greater than -0.0. | |
| bool | cannotBeOrderedGreaterEqZero (DenormalMode Mode) const |
| Return true if it's know this can never be a negative value or a logical 0. | |
| KnownFPClass | intersectWith (const KnownFPClass &RHS) const |
| KnownFPClass | unionWith (const KnownFPClass &RHS) const |
| KnownFPClass & | operator|= (const KnownFPClass &RHS) |
| void | knownNot (FPClassTest RuleOut) |
| void | fneg () |
| void | fabs () |
| bool | signBitIsZeroOrNaN () const |
| Return true if the sign bit must be 0, ignoring the sign of nans. | |
| void | signBitMustBeZero () |
| Assume the sign bit is zero. | |
| void | signBitMustBeOne () |
| Assume the sign bit is one. | |
| void | copysign (const KnownFPClass &Sign) |
| void | propagateNaN (const KnownFPClass &Src, bool PreserveSign=false) |
| LLVM_ABI void | propagateDenormal (const KnownFPClass &Src, DenormalMode Mode) |
| Propagate knowledge from a source value that could be a denormal or zero. | |
| LLVM_ABI void | propagateCanonicalizingSrc (const KnownFPClass &Src, DenormalMode Mode) |
Report known classes if Src is evaluated through a potentially canonicalizing operation. | |
| void | resetAll () |
Public Attributes | |
| FPClassTest | KnownFPClasses = fcAllFlags |
| Floating-point classes the value could be one of. | |
| std::optional< bool > | SignBit |
| std::nullopt if the sign bit is unknown, true if the sign bit is definitely set or false if the sign bit is definitely unset. | |
Static Public Attributes | |
| static constexpr FPClassTest | OrderedLessThanZeroMask |
| static constexpr FPClassTest | OrderedGreaterThanZeroMask |
Definition at line 25 of file KnownFPClass.h.
|
strong |
| Enumerator | |
|---|---|
| minimum | |
| maximum | |
| minimumnum | |
| maximumnum | |
| minnum | |
| maxnum | |
Definition at line 196 of file KnownFPClass.h.
|
inline |
Definition at line 33 of file KnownFPClass.h.
References llvm::fcAllFlags.
Referenced by canonicalize(), copysign(), exp(), fadd(), fadd_self(), fmul(), fpext(), intersectWith(), log(), minMaxLike(), operator==(), operator|=(), propagateCanonicalizingSrc(), propagateDenormal(), propagateNaN(), resetAll(), roundToIntegral(), sqrt(), square(), and unionWith().
Definition at line 20 of file KnownFPClass.cpp.
References llvm::CallingConv::C, KnownFPClasses, and SignBit.
|
inline |
Return true if it's know this can never be a negative value or a logical 0.
NaN --> true
x >= -0 --> false nsub --> true if mode is ieee, false otherwise. x < -0 --> true
Definition at line 136 of file KnownFPClass.h.
References llvm::fcPositive, isKnownNever(), isKnownNeverLogicalNegZero(), and Mode.
Referenced by llvm::InstCombinerImpl::SimplifyDemandedUseFPClass().
|
inline |
Return true if we can prove that the analyzed floating-point value is either NaN or never greater than -0.0.
NaN --> true +0 --> true -0 --> true x > +0 --> false x < -0 --> true
Definition at line 125 of file KnownFPClass.h.
References isKnownNever(), and OrderedGreaterThanZeroMask.
Referenced by computeKnownFPClass(), exp(), fadd(), and minMaxLike().
|
inline |
Return true if we can prove that the analyzed floating-point value is either NaN or never less than -0.0.
NaN --> true +0 --> true -0 --> true
x > +0 --> true x < -0 --> false
Definition at line 114 of file KnownFPClass.h.
References isKnownNever(), and OrderedLessThanZeroMask.
Referenced by llvm::cannotBeOrderedLessThanZero(), computeKnownFPClass(), computeKnownFPClassForFPTrunc(), exp(), fadd(), log(), minMaxLike(), simplifyFCmpInst(), and sqrt().
|
static |
Apply the canonicalize intrinsic to this value.
This is essentially a stronger form of propagateCanonicalizingSrc.
Definition at line 186 of file KnownFPClass.cpp.
References llvm::fcNan, llvm::fcNegSubnormal, llvm::fcNegZero, llvm::fcPosSubnormal, llvm::fcPosZero, llvm::fcQNan, llvm::fcSNan, llvm::fcSubnormal, llvm::fcZero, llvm::DenormalMode::getIEEE(), llvm::DenormalMode::getPreserveSign(), llvm::DenormalMode::IEEE, llvm::DenormalMode::Input, llvm::DenormalMode::inputsAreZero(), isKnownNever(), isKnownNeverNaN(), KnownFPClass(), KnownFPClasses, knownNot(), llvm::DenormalMode::Output, llvm::DenormalMode::outputsAreZero(), and llvm::DenormalMode::PositiveZero.
Referenced by computeKnownFPClass(), and llvm::InstCombinerImpl::SimplifyDemandedUseFPClass().
|
inline |
Definition at line 258 of file KnownFPClass.h.
References llvm::fcInf, llvm::fcNan, llvm::fcNegative, llvm::fcNormal, llvm::fcPositive, llvm::fcSubnormal, llvm::fcZero, isKnownNever(), KnownFPClass(), KnownFPClasses, and SignBit.
Referenced by computeKnownFPClass(), and llvm::InstCombinerImpl::SimplifyDemandedUseFPClass().
|
static |
Report known values for exp, exp2 and exp10.
Definition at line 335 of file KnownFPClass.cpp.
References cannotBeOrderedGreaterThanZero(), cannotBeOrderedLessThanZero(), llvm::fcNegative, llvm::fcPosInf, llvm::fcPosSubnormal, llvm::fcPosZero, KnownFPClass(), knownNot(), and propagateNaN().
Referenced by computeKnownFPClass(), and llvm::InstCombinerImpl::SimplifyDemandedUseFPClass().
|
inline |
Definition at line 179 of file KnownFPClass.h.
References llvm::fcNegInf, llvm::fcNegNormal, llvm::fcNegSubnormal, llvm::fcNegZero, llvm::fcPosInf, llvm::fcPosNormal, llvm::fcPosSubnormal, llvm::fcPosZero, KnownFPClasses, and signBitMustBeZero().
Referenced by computeKnownFPClass(), and llvm::InstCombinerImpl::SimplifyDemandedUseFPClass().
|
static |
Report known values for fadd.
Definition at line 236 of file KnownFPClass.cpp.
References cannotBeOrderedGreaterThanZero(), cannotBeOrderedLessThanZero(), llvm::fcNan, llvm::fcNegZero, llvm::DenormalMode::IEEE, isKnownNeverInfinity(), isKnownNeverLogicalNegZero(), isKnownNeverNaN(), KnownFPClass(), knownNot(), OrderedGreaterThanZeroMask, OrderedLessThanZeroMask, and llvm::DenormalMode::PositiveZero.
Referenced by fadd_self(), and llvm::InstCombinerImpl::SimplifyDemandedUseFPClass().
|
static |
Report known values for fadd x, x.
Definition at line 266 of file KnownFPClass.cpp.
References fadd(), llvm::fcPosZero, llvm::DenormalMode::IEEE, isKnownNeverLogicalPosZero(), isKnownNeverPosSubnormal(), isKnownNeverSubnormal(), KnownFPClass(), knownNot(), llvm::DenormalMode::PositiveZero, and llvm::DenormalMode::PreserveSign.
Referenced by llvm::InstCombinerImpl::SimplifyDemandedUseFPClass().
|
static |
Report known values for fmul.
Definition at line 282 of file KnownFPClass.cpp.
References llvm::fcInf, llvm::fcNan, llvm::fcNegative, llvm::fcNormal, llvm::fcPositive, llvm::fcSubnormal, llvm::fcZero, isKnownAlways(), isKnownNever(), isKnownNeverInfinity(), isKnownNeverLogicalZero(), isKnownNeverNaN(), KnownFPClass(), knownNot(), SignBit, signBitMustBeOne(), and signBitMustBeZero().
Referenced by computeKnownFPClass(), llvm::InstCombinerImpl::SimplifyDemandedUseFPClass(), and square().
|
inline |
Definition at line 173 of file KnownFPClass.h.
References llvm::fneg(), KnownFPClasses, and SignBit.
Referenced by computeKnownFPClass(), and llvm::InstCombinerImpl::SimplifyDemandedUseFPClass().
|
static |
Propagate known class for fpext.
Definition at line 404 of file KnownFPClass.cpp.
References llvm::fcNegNormal, llvm::fcNegSubnormal, llvm::fcPosNormal, llvm::fcPosSubnormal, llvm::fcSubnormal, isKnownNeverNaN(), llvm::APFloatBase::isRepresentableAsNormalIn(), KnownFPClass(), KnownFPClasses, knownNot(), and SignBit.
Referenced by computeKnownFPClass(), and llvm::InstCombinerImpl::SimplifyDemandedUseFPClass().
|
inline |
Definition at line 140 of file KnownFPClass.h.
References KnownFPClass(), KnownFPClasses, RHS, and SignBit.
Referenced by computeKnownFPClass(), llvm::InstCombinerImpl::SimplifyDemandedUseFPClass(), and llvm::InstCombinerImpl::SimplifyMultipleUseDemandedFPClass().
|
inline |
Definition at line 46 of file KnownFPClass.h.
References isKnownNever().
Referenced by computeKnownFPClass(), fmul(), isKnownAlwaysNaN(), and llvm::InstCombinerImpl::SimplifyDemandedUseFPClass().
|
inline |
Return true if it's known this must always be a nan.
Definition at line 54 of file KnownFPClass.h.
References llvm::fcNan, and isKnownAlways().
Referenced by simplifyFCmpInst().
|
inline |
Return true if it's known this can never be one of the mask entries.
Definition at line 42 of file KnownFPClass.h.
References llvm::fcNone, and KnownFPClasses.
Referenced by cannotBeOrderedGreaterEqZero(), cannotBeOrderedGreaterThanZero(), cannotBeOrderedLessThanZero(), canonicalize(), computeKnownFPClass(), copysign(), fmul(), foldFPtoI(), isKnownAlways(), isKnownNeverInfinity(), isKnownNeverInfOrNaN(), isKnownNeverNaN(), isKnownNeverNegInfinity(), isKnownNeverNegSubnormal(), isKnownNeverNegZero(), isKnownNeverPosInfinity(), isKnownNeverPosSubnormal(), isKnownNeverPosZero(), isKnownNeverSubnormal(), isKnownNeverZero(), knownNot(), roundToIntegral(), signBitIsZeroOrNaN(), llvm::InstCombinerImpl::SimplifyDemandedUseFPClass(), simplifyFMAFMul(), llvm::InstCombinerImpl::SimplifyMultipleUseDemandedFPClass(), simplifyX86FPMaxMin(), and sqrt().
|
inline |
Return true if it's known this can never be an infinity.
Definition at line 57 of file KnownFPClass.h.
References llvm::fcInf, and isKnownNever().
Referenced by computeKnownFPClass(), fadd(), fmul(), inferFastMathValueFlagsBinOp(), llvm::isKnownNeverInfinity(), and llvm::isKnownNeverInfOrNaN().
|
inline |
Return true if it's known this can never be an infinity or nan.
Definition at line 60 of file KnownFPClass.h.
References llvm::fcInf, llvm::fcNan, and isKnownNever().
Referenced by computeKnownFPClass(), and llvm::InstCombinerImpl::SimplifyDemandedUseFPClass().
| bool KnownFPClass::isKnownNeverLogicalNegZero | ( | DenormalMode | Mode | ) | const |
Return true if it's known this can never be interpreted as a negative zero.
Definition at line 39 of file KnownFPClass.cpp.
References inputDenormalIsIEEEOrPosZero(), isKnownNeverNegSubnormal(), and isKnownNeverNegZero().
Referenced by cannotBeOrderedGreaterEqZero(), computeKnownFPClass(), fadd(), and sqrt().
| bool KnownFPClass::isKnownNeverLogicalPosZero | ( | DenormalMode | Mode | ) | const |
Return true if it's known this can never be interpreted as a positive zero.
Definition at line 44 of file KnownFPClass.cpp.
References llvm::DenormalMode::IEEE, isKnownNeverPosSubnormal(), isKnownNeverPosZero(), isKnownNeverSubnormal(), llvm_unreachable, llvm::DenormalMode::PositiveZero, and llvm::DenormalMode::PreserveSign.
Referenced by computeKnownFPClass(), and fadd_self().
| bool KnownFPClass::isKnownNeverLogicalZero | ( | DenormalMode | Mode | ) | const |
Return true if it's known this can never be interpreted as a zero.
This extends isKnownNeverZero to cover the case where the assumed floating-point mode for the function interprets denormals as zero.
Definition at line 34 of file KnownFPClass.cpp.
References inputDenormalIsIEEE(), isKnownNeverSubnormal(), and isKnownNeverZero().
Referenced by computeKnownFPClass(), fmul(), log(), and llvm::InstCombinerImpl::SimplifyDemandedUseFPClass().
|
inline |
Return true if it's known this can never be a nan.
Definition at line 51 of file KnownFPClass.h.
References llvm::fcNan, and isKnownNever().
Referenced by canonicalize(), computeKnownFPClass(), fadd(), fmul(), fpext(), inferFastMathValueFlagsBinOp(), llvm::isKnownNeverInfOrNaN(), llvm::isKnownNeverNaN(), log(), minMaxLike(), llvm::InstCombinerImpl::SimplifyDemandedUseFPClass(), simplifyFCmpInst(), and sqrt().
|
inline |
Return true if it's known this can never be -infinity.
Definition at line 66 of file KnownFPClass.h.
References llvm::fcNegInf, and isKnownNever().
Referenced by computeKnownFPClass(), and roundToIntegral().
|
inline |
Return true if it's known this can never be a negative subnormal.
Definition at line 75 of file KnownFPClass.h.
References llvm::fcNegSubnormal, and isKnownNever().
Referenced by computeKnownFPClass(), and isKnownNeverLogicalNegZero().
|
inline |
Return true if it's known this can never be a negative zero.
This means a literal -0 and does not include denormal inputs implicitly treated as -0.
Definition at line 86 of file KnownFPClass.h.
References llvm::fcNegZero, and isKnownNever().
Referenced by llvm::cannotBeNegativeZero(), isKnownNeverLogicalNegZero(), and minMaxLike().
|
inline |
Return true if it's known this can never be +infinity.
Definition at line 63 of file KnownFPClass.h.
References llvm::fcPosInf, and isKnownNever().
Referenced by computeKnownFPClass(), log(), roundToIntegral(), and sqrt().
|
inline |
Return true if it's known this can never be a positive subnormal.
Definition at line 72 of file KnownFPClass.h.
References llvm::fcPosSubnormal, and isKnownNever().
Referenced by computeKnownFPClass(), fadd_self(), and isKnownNeverLogicalPosZero().
|
inline |
Return true if it's known this can never be a literal positive zero.
Definition at line 82 of file KnownFPClass.h.
References llvm::fcPosZero, and isKnownNever().
Referenced by isKnownNeverLogicalPosZero(), and minMaxLike().
|
inline |
Return true if it's known this can never be a subnormal.
Definition at line 69 of file KnownFPClass.h.
References llvm::fcSubnormal, and isKnownNever().
Referenced by computeKnownFPClass(), fadd_self(), isKnownNeverLogicalPosZero(), isKnownNeverLogicalZero(), minMaxLike(), and llvm::InstCombinerImpl::SimplifyDemandedUseFPClass().
|
inline |
Return true if it's known this can never be a zero.
This means a literal [+-]0, and does not include denormal inputs implicitly treated as [+-]0.
Definition at line 79 of file KnownFPClass.h.
References llvm::fcZero, and isKnownNever().
Referenced by isKnownNeverLogicalZero(), and llvm::LegalizerHelper::lowerFMinimumMaximum().
|
inline |
Definition at line 48 of file KnownFPClass.h.
References llvm::fcAllFlags, KnownFPClasses, and SignBit.
Referenced by llvm::adjustKnownFPClassForSelectArm(), and computeKnownFPClass().
|
inline |
Definition at line 163 of file KnownFPClass.h.
References llvm::fcNan, llvm::fcNegative, llvm::fcPositive, isKnownNever(), KnownFPClasses, and SignBit.
Referenced by canonicalize(), computeKnownFPClass(), computeKnownFPClassForFPTrunc(), computeKnownFPClassFromCond(), exp(), fadd(), fadd_self(), fmul(), fpext(), log(), minMaxLike(), propagateNaN(), roundToIntegral(), llvm::InstCombinerImpl::SimplifyDemandedUseFPClass(), sqrt(), and square().
|
static |
Propagate known class for log/log2/log10.
Definition at line 362 of file KnownFPClass.cpp.
References cannotBeOrderedLessThanZero(), llvm::fcNan, llvm::fcNegInf, llvm::fcNegZero, llvm::fcPosInf, isKnownNeverLogicalZero(), isKnownNeverNaN(), isKnownNeverPosInfinity(), KnownFPClass(), and knownNot().
Referenced by computeKnownFPClass(), and llvm::InstCombinerImpl::SimplifyDemandedUseFPClass().
|
static |
Definition at line 94 of file KnownFPClass.cpp.
References cannotBeOrderedGreaterThanZero(), cannotBeOrderedLessThanZero(), llvm::fcNan, llvm::fcNone, llvm::fcZero, llvm::DenormalMode::getIEEE(), isKnownNeverNaN(), isKnownNeverNegZero(), isKnownNeverPosZero(), isKnownNeverSubnormal(), KnownFPClass(), KnownFPClasses, knownNot(), llvm_unreachable, maximum, maximumnum, maxnum, minimum, minimumnum, minnum, OrderedGreaterThanZeroMask, OrderedLessThanZeroMask, SignBit, signBitMustBeOne(), and signBitMustBeZero().
Referenced by computeKnownFPClass(), and llvm::InstCombinerImpl::SimplifyDemandedUseFPClass().
|
inline |
Definition at line 37 of file KnownFPClass.h.
References KnownFPClass(), KnownFPClasses, llvm::Other, and SignBit.
|
inline |
Definition at line 155 of file KnownFPClass.h.
References KnownFPClass(), KnownFPClasses, RHS, and SignBit.
| void KnownFPClass::propagateCanonicalizingSrc | ( | const KnownFPClass & | Src, |
| DenormalMode | Mode ) |
Report known classes if Src is evaluated through a potentially canonicalizing operation.
We can assume signaling nans will not be introduced, but cannot assume a denormal will be flushed under FTZ/DAZ.
This assumes a copy-like operation and will replace any currently known information.
Definition at line 356 of file KnownFPClass.cpp.
References KnownFPClass(), propagateDenormal(), and propagateNaN().
Referenced by computeKnownFPClass().
| void KnownFPClass::propagateDenormal | ( | const KnownFPClass & | Src, |
| DenormalMode | Mode ) |
Propagate knowledge from a source value that could be a denormal or zero.
We have to be conservative since output flushing is not guaranteed, so known-never-zero may not hold.
This assumes a copy-like operation and will replace any currently known information.
Definition at line 67 of file KnownFPClass.cpp.
References llvm::DenormalMode::Dynamic, llvm::fcNegZero, llvm::fcPosZero, llvm::DenormalMode::getIEEE(), llvm::DenormalMode::getPositiveZero(), KnownFPClass(), KnownFPClasses, and llvm::DenormalMode::PositiveZero.
Referenced by propagateCanonicalizingSrc().
|
inline |
Definition at line 283 of file KnownFPClass.h.
References llvm::fcNan, llvm::fcSNan, KnownFPClass(), knownNot(), and SignBit.
Referenced by computeKnownFPClass(), computeKnownFPClassForFPTrunc(), exp(), propagateCanonicalizingSrc(), and roundToIntegral().
|
inline |
Definition at line 329 of file KnownFPClass.h.
References KnownFPClass().
Referenced by computeKnownFPClass(), and llvm::TargetLowering::computeKnownFPClassForTargetInstr().
|
static |
Propagate known class for rounding intrinsics (trunc, floor, ceil, rint, nearbyint, round, roundeven).
This is trunc if IsTrunc. IsMultiUnitFPType if this is for a multi-unit floating-point type.
Definition at line 426 of file KnownFPClass.cpp.
References llvm::fcNegFinite, llvm::fcNegInf, llvm::fcPosFinite, llvm::fcPosInf, llvm::fcSubnormal, isKnownNever(), isKnownNeverNegInfinity(), isKnownNeverPosInfinity(), KnownFPClass(), knownNot(), and propagateNaN().
Referenced by computeKnownFPClass(), and llvm::InstCombinerImpl::SimplifyDemandedUseFPClass().
|
inline |
Return true if the sign bit must be 0, ignoring the sign of nans.
Definition at line 244 of file KnownFPClass.h.
References llvm::fcNegative, and isKnownNever().
|
inline |
Assume the sign bit is one.
Definition at line 253 of file KnownFPClass.h.
References llvm::fcNan, llvm::fcNegative, KnownFPClasses, and SignBit.
Referenced by computeKnownFPClass(), computeKnownFPClassFromCond(), fmul(), and minMaxLike().
|
inline |
Assume the sign bit is zero.
Definition at line 247 of file KnownFPClass.h.
References llvm::fcNan, llvm::fcPositive, KnownFPClasses, and SignBit.
Referenced by computeKnownFPClass(), computeKnownFPClassFromCond(), fabs(), fmul(), and minMaxLike().
|
static |
Propagate known class for sqrt.
Definition at line 379 of file KnownFPClass.cpp.
References cannotBeOrderedLessThanZero(), llvm::fcNan, llvm::fcNegInf, llvm::fcNegNormal, llvm::fcNegSubnormal, llvm::fcNegZero, llvm::fcPosInf, llvm::fcPosSubnormal, llvm::fcSNan, isKnownNever(), isKnownNeverLogicalNegZero(), isKnownNeverNaN(), isKnownNeverPosInfinity(), KnownFPClass(), and knownNot().
Referenced by computeKnownFPClass(), and llvm::InstCombinerImpl::SimplifyDemandedUseFPClass().
|
inlinestatic |
Definition at line 231 of file KnownFPClass.h.
References llvm::fcNegative, fmul(), llvm::DenormalMode::getDynamic(), KnownFPClass(), knownNot(), and Mode.
Referenced by computeKnownFPClass(), and llvm::InstCombinerImpl::SimplifyDemandedUseFPClass().
|
inline |
Definition at line 145 of file KnownFPClass.h.
References KnownFPClass(), KnownFPClasses, RHS, and SignBit.
Referenced by llvm::adjustKnownFPClassForSelectArm().
| FPClassTest llvm::KnownFPClass::KnownFPClasses = fcAllFlags |
Floating-point classes the value could be one of.
Definition at line 27 of file KnownFPClass.h.
Referenced by canonicalize(), computeKnownFPClass(), copysign(), fabs(), fneg(), fpext(), intersectWith(), isKnownNever(), isUnknown(), KnownFPClass(), knownNot(), minMaxLike(), operator==(), operator|=(), propagateDenormal(), signBitMustBeOne(), signBitMustBeZero(), llvm::InstCombinerImpl::SimplifyDemandedFPClass(), llvm::InstCombinerImpl::SimplifyDemandedUseFPClass(), llvm::InstCombinerImpl::SimplifyMultipleUseDemandedFPClass(), and unionWith().
|
staticconstexpr |
Definition at line 103 of file KnownFPClass.h.
Referenced by cannotBeOrderedGreaterThanZero(), computeKnownFPClass(), fadd(), and minMaxLike().
|
staticconstexpr |
Definition at line 101 of file KnownFPClass.h.
Referenced by llvm::cannotBeOrderedLessThanZero(), cannotBeOrderedLessThanZero(), computeKnownFPClass(), computeKnownFPClassForFPTrunc(), fadd(), minMaxLike(), and simplifyFCmpInst().
| std::optional<bool> llvm::KnownFPClass::SignBit |
std::nullopt if the sign bit is unknown, true if the sign bit is definitely set or false if the sign bit is definitely unset.
Definition at line 31 of file KnownFPClass.h.
Referenced by computeKnownFPClass(), llvm::computeKnownFPSignBit(), copysign(), fmul(), fneg(), fpext(), intersectWith(), isUnknown(), KnownFPClass(), knownNot(), minMaxLike(), operator==(), operator|=(), propagateNaN(), signBitMustBeOne(), signBitMustBeZero(), llvm::InstCombinerImpl::SimplifyDemandedUseFPClass(), simplifyFMAFMul(), and unionWith().