//===-- APFloat.cpp - Implement APFloat class -----------------------------===// // // 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 a class to represent arbitrary precision floating // point values and provide a variety of arithmetic operations on them. // //===----------------------------------------------------------------------===// #include "llvm/ADT/APFloat.h" #include "llvm/ADT/APSInt.h" #include "llvm/ADT/ArrayRef.h" #include "llvm/ADT/FloatingPointMode.h" #include "llvm/ADT/FoldingSet.h" #include "llvm/ADT/Hashing.h" #include "llvm/ADT/STLExtras.h" #include "llvm/ADT/StringExtras.h" #include "llvm/ADT/StringRef.h" #include "llvm/Config/llvm-config.h" #include "llvm/Support/Debug.h" #include "llvm/Support/Error.h" #include "llvm/Support/MathExtras.h" #include "llvm/Support/raw_ostream.h" #include #include #define APFLOAT_DISPATCH_ON_SEMANTICS(METHOD_CALL) \ do { \ if (usesLayout(getSemantics())) \ return U.IEEE.METHOD_CALL; \ if (usesLayout(getSemantics())) \ return U.Double.METHOD_CALL; \ llvm_unreachable("Unexpected semantics"); \ } while (false) using namespace llvm; /// A macro used to combine two fcCategory enums into one key which can be used /// in a switch statement to classify how the interaction of two APFloat's /// categories affects an operation. /// /// TODO: If clang source code is ever allowed to use constexpr in its own /// codebase, change this into a static inline function. #define PackCategoriesIntoKey(_lhs, _rhs) ((_lhs) * 4 + (_rhs)) /* Assumed in hexadecimal significand parsing, and conversion to hexadecimal strings. */ static_assert(APFloatBase::integerPartWidth % 4 == 0, "Part width must be divisible by 4!"); namespace llvm { // How the nonfinite values Inf and NaN are represented. enum class fltNonfiniteBehavior { // Represents standard IEEE 754 behavior. A value is nonfinite if the // exponent field is all 1s. In such cases, a value is Inf if the // significand bits are all zero, and NaN otherwise IEEE754, // This behavior is present in the Float8ExMyFN* types (Float8E4M3FN, // Float8E5M2FNUZ, Float8E4M3FNUZ, and Float8E4M3B11FNUZ). There is no // representation for Inf, and operations that would ordinarily produce Inf // produce NaN instead. // The details of the NaN representation(s) in this form are determined by the // `fltNanEncoding` enum. We treat all NaNs as quiet, as the available // encodings do not distinguish between signalling and quiet NaN. NanOnly, // This behavior is present in Float6E3M2FN, Float6E2M3FN, and // Float4E2M1FN types, which do not support Inf or NaN values. FiniteOnly, }; // How NaN values are represented. This is curently only used in combination // with fltNonfiniteBehavior::NanOnly, and using a variant other than IEEE // while having IEEE non-finite behavior is liable to lead to unexpected // results. enum class fltNanEncoding { // Represents the standard IEEE behavior where a value is NaN if its // exponent is all 1s and the significand is non-zero. IEEE, // Represents the behavior in the Float8E4M3FN floating point type where NaN // is represented by having the exponent and mantissa set to all 1s. // This behavior matches the FP8 E4M3 type described in // https://arxiv.org/abs/2209.05433. We treat both signed and unsigned NaNs // as non-signalling, although the paper does not state whether the NaN // values are signalling or not. AllOnes, // Represents the behavior in Float8E{5,4}E{2,3}FNUZ floating point types // where NaN is represented by a sign bit of 1 and all 0s in the exponent // and mantissa (i.e. the negative zero encoding in a IEEE float). Since // there is only one NaN value, it is treated as quiet NaN. This matches the // behavior described in https://arxiv.org/abs/2206.02915 . NegativeZero, }; /* Represents floating point arithmetic semantics. */ struct fltSemantics { /* The largest E such that 2^E is representable; this matches the definition of IEEE 754. */ APFloatBase::ExponentType maxExponent; /* The smallest E such that 2^E is a normalized number; this matches the definition of IEEE 754. */ APFloatBase::ExponentType minExponent; /* Number of bits in the significand. This includes the integer bit. */ unsigned int precision; /* Number of bits actually used in the semantics. */ unsigned int sizeInBits; fltNonfiniteBehavior nonFiniteBehavior = fltNonfiniteBehavior::IEEE754; fltNanEncoding nanEncoding = fltNanEncoding::IEEE; // Returns true if any number described by this semantics can be precisely // represented by the specified semantics. Does not take into account // the value of fltNonfiniteBehavior. bool isRepresentableBy(const fltSemantics &S) const { return maxExponent <= S.maxExponent && minExponent >= S.minExponent && precision <= S.precision; } }; static constexpr fltSemantics semIEEEhalf = {15, -14, 11, 16}; static constexpr fltSemantics semBFloat = {127, -126, 8, 16}; static constexpr fltSemantics semIEEEsingle = {127, -126, 24, 32}; static constexpr fltSemantics semIEEEdouble = {1023, -1022, 53, 64}; static constexpr fltSemantics semIEEEquad = {16383, -16382, 113, 128}; static constexpr fltSemantics semFloat8E5M2 = {15, -14, 3, 8}; static constexpr fltSemantics semFloat8E5M2FNUZ = { 15, -15, 3, 8, fltNonfiniteBehavior::NanOnly, fltNanEncoding::NegativeZero}; static constexpr fltSemantics semFloat8E4M3 = {7, -6, 4, 8}; static constexpr fltSemantics semFloat8E4M3FN = { 8, -6, 4, 8, fltNonfiniteBehavior::NanOnly, fltNanEncoding::AllOnes}; static constexpr fltSemantics semFloat8E4M3FNUZ = { 7, -7, 4, 8, fltNonfiniteBehavior::NanOnly, fltNanEncoding::NegativeZero}; static constexpr fltSemantics semFloat8E4M3B11FNUZ = { 4, -10, 4, 8, fltNonfiniteBehavior::NanOnly, fltNanEncoding::NegativeZero}; static constexpr fltSemantics semFloatTF32 = {127, -126, 11, 19}; static constexpr fltSemantics semFloat6E3M2FN = { 4, -2, 3, 6, fltNonfiniteBehavior::FiniteOnly}; static constexpr fltSemantics semFloat6E2M3FN = { 2, 0, 4, 6, fltNonfiniteBehavior::FiniteOnly}; static constexpr fltSemantics semFloat4E2M1FN = { 2, 0, 2, 4, fltNonfiniteBehavior::FiniteOnly}; static constexpr fltSemantics semX87DoubleExtended = {16383, -16382, 64, 80}; static constexpr fltSemantics semBogus = {0, 0, 0, 0}; /* The IBM double-double semantics. Such a number consists of a pair of IEEE 64-bit doubles (Hi, Lo), where |Hi| > |Lo|, and if normal, (double)(Hi + Lo) == Hi. The numeric value it's modeling is Hi + Lo. Therefore it has two 53-bit mantissa parts that aren't necessarily adjacent to each other, and two 11-bit exponents. Note: we need to make the value different from semBogus as otherwise an unsafe optimization may collapse both values to a single address, and we heavily rely on them having distinct addresses. */ static constexpr fltSemantics semPPCDoubleDouble = {-1, 0, 0, 128}; /* These are legacy semantics for the fallback, inaccrurate implementation of IBM double-double, if the accurate semPPCDoubleDouble doesn't handle the operation. It's equivalent to having an IEEE number with consecutive 106 bits of mantissa and 11 bits of exponent. It's not equivalent to IBM double-double. For example, a legit IBM double-double, 1 + epsilon: 1 + epsilon = 1 + (1 >> 1076) is not representable by a consecutive 106 bits of mantissa. Currently, these semantics are used in the following way: semPPCDoubleDouble -> (IEEEdouble, IEEEdouble) -> (64-bit APInt, 64-bit APInt) -> (128-bit APInt) -> semPPCDoubleDoubleLegacy -> IEEE operations We use bitcastToAPInt() to get the bit representation (in APInt) of the underlying IEEEdouble, then use the APInt constructor to construct the legacy IEEE float. TODO: Implement all operations in semPPCDoubleDouble, and delete these semantics. */ static constexpr fltSemantics semPPCDoubleDoubleLegacy = {1023, -1022 + 53, 53 + 53, 128}; const llvm::fltSemantics &APFloatBase::EnumToSemantics(Semantics S) { switch (S) { case S_IEEEhalf: return IEEEhalf(); case S_BFloat: return BFloat(); case S_IEEEsingle: return IEEEsingle(); case S_IEEEdouble: return IEEEdouble(); case S_IEEEquad: return IEEEquad(); case S_PPCDoubleDouble: return PPCDoubleDouble(); case S_Float8E5M2: return Float8E5M2(); case S_Float8E5M2FNUZ: return Float8E5M2FNUZ(); case S_Float8E4M3: return Float8E4M3(); case S_Float8E4M3FN: return Float8E4M3FN(); case S_Float8E4M3FNUZ: return Float8E4M3FNUZ(); case S_Float8E4M3B11FNUZ: return Float8E4M3B11FNUZ(); case S_FloatTF32: return FloatTF32(); case S_Float6E3M2FN: return Float6E3M2FN(); case S_Float6E2M3FN: return Float6E2M3FN(); case S_Float4E2M1FN: return Float4E2M1FN(); case S_x87DoubleExtended: return x87DoubleExtended(); } llvm_unreachable("Unrecognised floating semantics"); } APFloatBase::Semantics APFloatBase::SemanticsToEnum(const llvm::fltSemantics &Sem) { if (&Sem == &llvm::APFloat::IEEEhalf()) return S_IEEEhalf; else if (&Sem == &llvm::APFloat::BFloat()) return S_BFloat; else if (&Sem == &llvm::APFloat::IEEEsingle()) return S_IEEEsingle; else if (&Sem == &llvm::APFloat::IEEEdouble()) return S_IEEEdouble; else if (&Sem == &llvm::APFloat::IEEEquad()) return S_IEEEquad; else if (&Sem == &llvm::APFloat::PPCDoubleDouble()) return S_PPCDoubleDouble; else if (&Sem == &llvm::APFloat::Float8E5M2()) return S_Float8E5M2; else if (&Sem == &llvm::APFloat::Float8E5M2FNUZ()) return S_Float8E5M2FNUZ; else if (&Sem == &llvm::APFloat::Float8E4M3()) return S_Float8E4M3; else if (&Sem == &llvm::APFloat::Float8E4M3FN()) return S_Float8E4M3FN; else if (&Sem == &llvm::APFloat::Float8E4M3FNUZ()) return S_Float8E4M3FNUZ; else if (&Sem == &llvm::APFloat::Float8E4M3B11FNUZ()) return S_Float8E4M3B11FNUZ; else if (&Sem == &llvm::APFloat::FloatTF32()) return S_FloatTF32; else if (&Sem == &llvm::APFloat::Float6E3M2FN()) return S_Float6E3M2FN; else if (&Sem == &llvm::APFloat::Float6E2M3FN()) return S_Float6E2M3FN; else if (&Sem == &llvm::APFloat::Float4E2M1FN()) return S_Float4E2M1FN; else if (&Sem == &llvm::APFloat::x87DoubleExtended()) return S_x87DoubleExtended; else llvm_unreachable("Unknown floating semantics"); } const fltSemantics &APFloatBase::IEEEhalf() { return semIEEEhalf; } const fltSemantics &APFloatBase::BFloat() { return semBFloat; } const fltSemantics &APFloatBase::IEEEsingle() { return semIEEEsingle; } const fltSemantics &APFloatBase::IEEEdouble() { return semIEEEdouble; } const fltSemantics &APFloatBase::IEEEquad() { return semIEEEquad; } const fltSemantics &APFloatBase::PPCDoubleDouble() { return semPPCDoubleDouble; } const fltSemantics &APFloatBase::Float8E5M2() { return semFloat8E5M2; } const fltSemantics &APFloatBase::Float8E5M2FNUZ() { return semFloat8E5M2FNUZ; } const fltSemantics &APFloatBase::Float8E4M3() { return semFloat8E4M3; } const fltSemantics &APFloatBase::Float8E4M3FN() { return semFloat8E4M3FN; } const fltSemantics &APFloatBase::Float8E4M3FNUZ() { return semFloat8E4M3FNUZ; } const fltSemantics &APFloatBase::Float8E4M3B11FNUZ() { return semFloat8E4M3B11FNUZ; } const fltSemantics &APFloatBase::FloatTF32() { return semFloatTF32; } const fltSemantics &APFloatBase::Float6E3M2FN() { return semFloat6E3M2FN; } const fltSemantics &APFloatBase::Float6E2M3FN() { return semFloat6E2M3FN; } const fltSemantics &APFloatBase::Float4E2M1FN() { return semFloat4E2M1FN; } const fltSemantics &APFloatBase::x87DoubleExtended() { return semX87DoubleExtended; } const fltSemantics &APFloatBase::Bogus() { return semBogus; } constexpr RoundingMode APFloatBase::rmNearestTiesToEven; constexpr RoundingMode APFloatBase::rmTowardPositive; constexpr RoundingMode APFloatBase::rmTowardNegative; constexpr RoundingMode APFloatBase::rmTowardZero; constexpr RoundingMode APFloatBase::rmNearestTiesToAway; /* A tight upper bound on number of parts required to hold the value pow(5, power) is power * 815 / (351 * integerPartWidth) + 1 However, whilst the result may require only this many parts, because we are multiplying two values to get it, the multiplication may require an extra part with the excess part being zero (consider the trivial case of 1 * 1, tcFullMultiply requires two parts to hold the single-part result). So we add an extra one to guarantee enough space whilst multiplying. */ const unsigned int maxExponent = 16383; const unsigned int maxPrecision = 113; const unsigned int maxPowerOfFiveExponent = maxExponent + maxPrecision - 1; const unsigned int maxPowerOfFiveParts = 2 + ((maxPowerOfFiveExponent * 815) / (351 * APFloatBase::integerPartWidth)); unsigned int APFloatBase::semanticsPrecision(const fltSemantics &semantics) { return semantics.precision; } APFloatBase::ExponentType APFloatBase::semanticsMaxExponent(const fltSemantics &semantics) { return semantics.maxExponent; } APFloatBase::ExponentType APFloatBase::semanticsMinExponent(const fltSemantics &semantics) { return semantics.minExponent; } unsigned int APFloatBase::semanticsSizeInBits(const fltSemantics &semantics) { return semantics.sizeInBits; } unsigned int APFloatBase::semanticsIntSizeInBits(const fltSemantics &semantics, bool isSigned) { // The max FP value is pow(2, MaxExponent) * (1 + MaxFraction), so we need // at least one more bit than the MaxExponent to hold the max FP value. unsigned int MinBitWidth = semanticsMaxExponent(semantics) + 1; // Extra sign bit needed. if (isSigned) ++MinBitWidth; return MinBitWidth; } bool APFloatBase::isRepresentableAsNormalIn(const fltSemantics &Src, const fltSemantics &Dst) { // Exponent range must be larger. if (Src.maxExponent >= Dst.maxExponent || Src.minExponent <= Dst.minExponent) return false; // If the mantissa is long enough, the result value could still be denormal // with a larger exponent range. // // FIXME: This condition is probably not accurate but also shouldn't be a // practical concern with existing types. return Dst.precision >= Src.precision; } unsigned APFloatBase::getSizeInBits(const fltSemantics &Sem) { return Sem.sizeInBits; } static constexpr APFloatBase::ExponentType exponentZero(const fltSemantics &semantics) { return semantics.minExponent - 1; } static constexpr APFloatBase::ExponentType exponentInf(const fltSemantics &semantics) { return semantics.maxExponent + 1; } static constexpr APFloatBase::ExponentType exponentNaN(const fltSemantics &semantics) { if (semantics.nonFiniteBehavior == fltNonfiniteBehavior::NanOnly) { if (semantics.nanEncoding == fltNanEncoding::NegativeZero) return exponentZero(semantics); return semantics.maxExponent; } return semantics.maxExponent + 1; } /* A bunch of private, handy routines. */ static inline Error createError(const Twine &Err) { return make_error(Err, inconvertibleErrorCode()); } static constexpr inline unsigned int partCountForBits(unsigned int bits) { return ((bits) + APFloatBase::integerPartWidth - 1) / APFloatBase::integerPartWidth; } /* Returns 0U-9U. Return values >= 10U are not digits. */ static inline unsigned int decDigitValue(unsigned int c) { return c - '0'; } /* Return the value of a decimal exponent of the form [+-]ddddddd. If the exponent overflows, returns a large exponent with the appropriate sign. */ static Expected readExponent(StringRef::iterator begin, StringRef::iterator end) { bool isNegative; unsigned int absExponent; const unsigned int overlargeExponent = 24000; /* FIXME. */ StringRef::iterator p = begin; // Treat no exponent as 0 to match binutils if (p == end || ((*p == '-' || *p == '+') && (p + 1) == end)) { return 0; } isNegative = (*p == '-'); if (*p == '-' || *p == '+') { p++; if (p == end) return createError("Exponent has no digits"); } absExponent = decDigitValue(*p++); if (absExponent >= 10U) return createError("Invalid character in exponent"); for (; p != end; ++p) { unsigned int value; value = decDigitValue(*p); if (value >= 10U) return createError("Invalid character in exponent"); absExponent = absExponent * 10U + value; if (absExponent >= overlargeExponent) { absExponent = overlargeExponent; break; } } if (isNegative) return -(int) absExponent; else return (int) absExponent; } /* This is ugly and needs cleaning up, but I don't immediately see how whilst remaining safe. */ static Expected totalExponent(StringRef::iterator p, StringRef::iterator end, int exponentAdjustment) { int unsignedExponent; bool negative, overflow; int exponent = 0; if (p == end) return createError("Exponent has no digits"); negative = *p == '-'; if (*p == '-' || *p == '+') { p++; if (p == end) return createError("Exponent has no digits"); } unsignedExponent = 0; overflow = false; for (; p != end; ++p) { unsigned int value; value = decDigitValue(*p); if (value >= 10U) return createError("Invalid character in exponent"); unsignedExponent = unsignedExponent * 10 + value; if (unsignedExponent > 32767) { overflow = true; break; } } if (exponentAdjustment > 32767 || exponentAdjustment < -32768) overflow = true; if (!overflow) { exponent = unsignedExponent; if (negative) exponent = -exponent; exponent += exponentAdjustment; if (exponent > 32767 || exponent < -32768) overflow = true; } if (overflow) exponent = negative ? -32768: 32767; return exponent; } static Expected skipLeadingZeroesAndAnyDot(StringRef::iterator begin, StringRef::iterator end, StringRef::iterator *dot) { StringRef::iterator p = begin; *dot = end; while (p != end && *p == '0') p++; if (p != end && *p == '.') { *dot = p++; if (end - begin == 1) return createError("Significand has no digits"); while (p != end && *p == '0') p++; } return p; } /* Given a normal decimal floating point number of the form dddd.dddd[eE][+-]ddd where the decimal point and exponent are optional, fill out the structure D. Exponent is appropriate if the significand is treated as an integer, and normalizedExponent if the significand is taken to have the decimal point after a single leading non-zero digit. If the value is zero, V->firstSigDigit points to a non-digit, and the return exponent is zero. */ struct decimalInfo { const char *firstSigDigit; const char *lastSigDigit; int exponent; int normalizedExponent; }; static Error interpretDecimal(StringRef::iterator begin, StringRef::iterator end, decimalInfo *D) { StringRef::iterator dot = end; auto PtrOrErr = skipLeadingZeroesAndAnyDot(begin, end, &dot); if (!PtrOrErr) return PtrOrErr.takeError(); StringRef::iterator p = *PtrOrErr; D->firstSigDigit = p; D->exponent = 0; D->normalizedExponent = 0; for (; p != end; ++p) { if (*p == '.') { if (dot != end) return createError("String contains multiple dots"); dot = p++; if (p == end) break; } if (decDigitValue(*p) >= 10U) break; } if (p != end) { if (*p != 'e' && *p != 'E') return createError("Invalid character in significand"); if (p == begin) return createError("Significand has no digits"); if (dot != end && p - begin == 1) return createError("Significand has no digits"); /* p points to the first non-digit in the string */ auto ExpOrErr = readExponent(p + 1, end); if (!ExpOrErr) return ExpOrErr.takeError(); D->exponent = *ExpOrErr; /* Implied decimal point? */ if (dot == end) dot = p; } /* If number is all zeroes accept any exponent. */ if (p != D->firstSigDigit) { /* Drop insignificant trailing zeroes. */ if (p != begin) { do do p--; while (p != begin && *p == '0'); while (p != begin && *p == '.'); } /* Adjust the exponents for any decimal point. */ D->exponent += static_cast((dot - p) - (dot > p)); D->normalizedExponent = (D->exponent + static_cast((p - D->firstSigDigit) - (dot > D->firstSigDigit && dot < p))); } D->lastSigDigit = p; return Error::success(); } /* Return the trailing fraction of a hexadecimal number. DIGITVALUE is the first hex digit of the fraction, P points to the next digit. */ static Expected trailingHexadecimalFraction(StringRef::iterator p, StringRef::iterator end, unsigned int digitValue) { unsigned int hexDigit; /* If the first trailing digit isn't 0 or 8 we can work out the fraction immediately. */ if (digitValue > 8) return lfMoreThanHalf; else if (digitValue < 8 && digitValue > 0) return lfLessThanHalf; // Otherwise we need to find the first non-zero digit. while (p != end && (*p == '0' || *p == '.')) p++; if (p == end) return createError("Invalid trailing hexadecimal fraction!"); hexDigit = hexDigitValue(*p); /* If we ran off the end it is exactly zero or one-half, otherwise a little more. */ if (hexDigit == UINT_MAX) return digitValue == 0 ? lfExactlyZero: lfExactlyHalf; else return digitValue == 0 ? lfLessThanHalf: lfMoreThanHalf; } /* Return the fraction lost were a bignum truncated losing the least significant BITS bits. */ static lostFraction lostFractionThroughTruncation(const APFloatBase::integerPart *parts, unsigned int partCount, unsigned int bits) { unsigned int lsb; lsb = APInt::tcLSB(parts, partCount); /* Note this is guaranteed true if bits == 0, or LSB == UINT_MAX. */ if (bits <= lsb) return lfExactlyZero; if (bits == lsb + 1) return lfExactlyHalf; if (bits <= partCount * APFloatBase::integerPartWidth && APInt::tcExtractBit(parts, bits - 1)) return lfMoreThanHalf; return lfLessThanHalf; } /* Shift DST right BITS bits noting lost fraction. */ static lostFraction shiftRight(APFloatBase::integerPart *dst, unsigned int parts, unsigned int bits) { lostFraction lost_fraction; lost_fraction = lostFractionThroughTruncation(dst, parts, bits); APInt::tcShiftRight(dst, parts, bits); return lost_fraction; } /* Combine the effect of two lost fractions. */ static lostFraction combineLostFractions(lostFraction moreSignificant, lostFraction lessSignificant) { if (lessSignificant != lfExactlyZero) { if (moreSignificant == lfExactlyZero) moreSignificant = lfLessThanHalf; else if (moreSignificant == lfExactlyHalf) moreSignificant = lfMoreThanHalf; } return moreSignificant; } /* The error from the true value, in half-ulps, on multiplying two floating point numbers, which differ from the value they approximate by at most HUE1 and HUE2 half-ulps, is strictly less than the returned value. See "How to Read Floating Point Numbers Accurately" by William D Clinger. */ static unsigned int HUerrBound(bool inexactMultiply, unsigned int HUerr1, unsigned int HUerr2) { assert(HUerr1 < 2 || HUerr2 < 2 || (HUerr1 + HUerr2 < 8)); if (HUerr1 + HUerr2 == 0) return inexactMultiply * 2; /* <= inexactMultiply half-ulps. */ else return inexactMultiply + 2 * (HUerr1 + HUerr2); } /* The number of ulps from the boundary (zero, or half if ISNEAREST) when the least significant BITS are truncated. BITS cannot be zero. */ static APFloatBase::integerPart ulpsFromBoundary(const APFloatBase::integerPart *parts, unsigned int bits, bool isNearest) { unsigned int count, partBits; APFloatBase::integerPart part, boundary; assert(bits != 0); bits--; count = bits / APFloatBase::integerPartWidth; partBits = bits % APFloatBase::integerPartWidth + 1; part = parts[count] & (~(APFloatBase::integerPart) 0 >> (APFloatBase::integerPartWidth - partBits)); if (isNearest) boundary = (APFloatBase::integerPart) 1 << (partBits - 1); else boundary = 0; if (count == 0) { if (part - boundary <= boundary - part) return part - boundary; else return boundary - part; } if (part == boundary) { while (--count) if (parts[count]) return ~(APFloatBase::integerPart) 0; /* A lot. */ return parts[0]; } else if (part == boundary - 1) { while (--count) if (~parts[count]) return ~(APFloatBase::integerPart) 0; /* A lot. */ return -parts[0]; } return ~(APFloatBase::integerPart) 0; /* A lot. */ } /* Place pow(5, power) in DST, and return the number of parts used. DST must be at least one part larger than size of the answer. */ static unsigned int powerOf5(APFloatBase::integerPart *dst, unsigned int power) { static const APFloatBase::integerPart firstEightPowers[] = { 1, 5, 25, 125, 625, 3125, 15625, 78125 }; APFloatBase::integerPart pow5s[maxPowerOfFiveParts * 2 + 5]; pow5s[0] = 78125 * 5; unsigned int partsCount = 1; APFloatBase::integerPart scratch[maxPowerOfFiveParts], *p1, *p2, *pow5; unsigned int result; assert(power <= maxExponent); p1 = dst; p2 = scratch; *p1 = firstEightPowers[power & 7]; power >>= 3; result = 1; pow5 = pow5s; for (unsigned int n = 0; power; power >>= 1, n++) { /* Calculate pow(5,pow(2,n+3)) if we haven't yet. */ if (n != 0) { APInt::tcFullMultiply(pow5, pow5 - partsCount, pow5 - partsCount, partsCount, partsCount); partsCount *= 2; if (pow5[partsCount - 1] == 0) partsCount--; } if (power & 1) { APFloatBase::integerPart *tmp; APInt::tcFullMultiply(p2, p1, pow5, result, partsCount); result += partsCount; if (p2[result - 1] == 0) result--; /* Now result is in p1 with partsCount parts and p2 is scratch space. */ tmp = p1; p1 = p2; p2 = tmp; } pow5 += partsCount; } if (p1 != dst) APInt::tcAssign(dst, p1, result); return result; } /* Zero at the end to avoid modular arithmetic when adding one; used when rounding up during hexadecimal output. */ static const char hexDigitsLower[] = "0123456789abcdef0"; static const char hexDigitsUpper[] = "0123456789ABCDEF0"; static const char infinityL[] = "infinity"; static const char infinityU[] = "INFINITY"; static const char NaNL[] = "nan"; static const char NaNU[] = "NAN"; /* Write out an integerPart in hexadecimal, starting with the most significant nibble. Write out exactly COUNT hexdigits, return COUNT. */ static unsigned int partAsHex (char *dst, APFloatBase::integerPart part, unsigned int count, const char *hexDigitChars) { unsigned int result = count; assert(count != 0 && count <= APFloatBase::integerPartWidth / 4); part >>= (APFloatBase::integerPartWidth - 4 * count); while (count--) { dst[count] = hexDigitChars[part & 0xf]; part >>= 4; } return result; } /* Write out an unsigned decimal integer. */ static char * writeUnsignedDecimal (char *dst, unsigned int n) { char buff[40], *p; p = buff; do *p++ = '0' + n % 10; while (n /= 10); do *dst++ = *--p; while (p != buff); return dst; } /* Write out a signed decimal integer. */ static char * writeSignedDecimal (char *dst, int value) { if (value < 0) { *dst++ = '-'; dst = writeUnsignedDecimal(dst, -(unsigned) value); } else dst = writeUnsignedDecimal(dst, value); return dst; } namespace detail { /* Constructors. */ void IEEEFloat::initialize(const fltSemantics *ourSemantics) { unsigned int count; semantics = ourSemantics; count = partCount(); if (count > 1) significand.parts = new integerPart[count]; } void IEEEFloat::freeSignificand() { if (needsCleanup()) delete [] significand.parts; } void IEEEFloat::assign(const IEEEFloat &rhs) { assert(semantics == rhs.semantics); sign = rhs.sign; category = rhs.category; exponent = rhs.exponent; if (isFiniteNonZero() || category == fcNaN) copySignificand(rhs); } void IEEEFloat::copySignificand(const IEEEFloat &rhs) { assert(isFiniteNonZero() || category == fcNaN); assert(rhs.partCount() >= partCount()); APInt::tcAssign(significandParts(), rhs.significandParts(), partCount()); } /* Make this number a NaN, with an arbitrary but deterministic value for the significand. If double or longer, this is a signalling NaN, which may not be ideal. If float, this is QNaN(0). */ void IEEEFloat::makeNaN(bool SNaN, bool Negative, const APInt *fill) { if (semantics->nonFiniteBehavior == fltNonfiniteBehavior::FiniteOnly) llvm_unreachable("This floating point format does not support NaN"); category = fcNaN; sign = Negative; exponent = exponentNaN(); integerPart *significand = significandParts(); unsigned numParts = partCount(); APInt fill_storage; if (semantics->nonFiniteBehavior == fltNonfiniteBehavior::NanOnly) { // Finite-only types do not distinguish signalling and quiet NaN, so // make them all signalling. SNaN = false; if (semantics->nanEncoding == fltNanEncoding::NegativeZero) { sign = true; fill_storage = APInt::getZero(semantics->precision - 1); } else { fill_storage = APInt::getAllOnes(semantics->precision - 1); } fill = &fill_storage; } // Set the significand bits to the fill. if (!fill || fill->getNumWords() < numParts) APInt::tcSet(significand, 0, numParts); if (fill) { APInt::tcAssign(significand, fill->getRawData(), std::min(fill->getNumWords(), numParts)); // Zero out the excess bits of the significand. unsigned bitsToPreserve = semantics->precision - 1; unsigned part = bitsToPreserve / 64; bitsToPreserve %= 64; significand[part] &= ((1ULL << bitsToPreserve) - 1); for (part++; part != numParts; ++part) significand[part] = 0; } unsigned QNaNBit = semantics->precision - 2; if (SNaN) { // We always have to clear the QNaN bit to make it an SNaN. APInt::tcClearBit(significand, QNaNBit); // If there are no bits set in the payload, we have to set // *something* to make it a NaN instead of an infinity; // conventionally, this is the next bit down from the QNaN bit. if (APInt::tcIsZero(significand, numParts)) APInt::tcSetBit(significand, QNaNBit - 1); } else if (semantics->nanEncoding == fltNanEncoding::NegativeZero) { // The only NaN is a quiet NaN, and it has no bits sets in the significand. // Do nothing. } else { // We always have to set the QNaN bit to make it a QNaN. APInt::tcSetBit(significand, QNaNBit); } // For x87 extended precision, we want to make a NaN, not a // pseudo-NaN. Maybe we should expose the ability to make // pseudo-NaNs? if (semantics == &semX87DoubleExtended) APInt::tcSetBit(significand, QNaNBit + 1); } IEEEFloat &IEEEFloat::operator=(const IEEEFloat &rhs) { if (this != &rhs) { if (semantics != rhs.semantics) { freeSignificand(); initialize(rhs.semantics); } assign(rhs); } return *this; } IEEEFloat &IEEEFloat::operator=(IEEEFloat &&rhs) { freeSignificand(); semantics = rhs.semantics; significand = rhs.significand; exponent = rhs.exponent; category = rhs.category; sign = rhs.sign; rhs.semantics = &semBogus; return *this; } bool IEEEFloat::isDenormal() const { return isFiniteNonZero() && (exponent == semantics->minExponent) && (APInt::tcExtractBit(significandParts(), semantics->precision - 1) == 0); } bool IEEEFloat::isSmallest() const { // The smallest number by magnitude in our format will be the smallest // denormal, i.e. the floating point number with exponent being minimum // exponent and significand bitwise equal to 1 (i.e. with MSB equal to 0). return isFiniteNonZero() && exponent == semantics->minExponent && significandMSB() == 0; } bool IEEEFloat::isSmallestNormalized() const { return getCategory() == fcNormal && exponent == semantics->minExponent && isSignificandAllZerosExceptMSB(); } bool IEEEFloat::isSignificandAllOnes() const { // Test if the significand excluding the integral bit is all ones. This allows // us to test for binade boundaries. const integerPart *Parts = significandParts(); const unsigned PartCount = partCountForBits(semantics->precision); for (unsigned i = 0; i < PartCount - 1; i++) if (~Parts[i]) return false; // Set the unused high bits to all ones when we compare. const unsigned NumHighBits = PartCount*integerPartWidth - semantics->precision + 1; assert(NumHighBits <= integerPartWidth && NumHighBits > 0 && "Can not have more high bits to fill than integerPartWidth"); const integerPart HighBitFill = ~integerPart(0) << (integerPartWidth - NumHighBits); if (~(Parts[PartCount - 1] | HighBitFill)) return false; return true; } bool IEEEFloat::isSignificandAllOnesExceptLSB() const { // Test if the significand excluding the integral bit is all ones except for // the least significant bit. const integerPart *Parts = significandParts(); if (Parts[0] & 1) return false; const unsigned PartCount = partCountForBits(semantics->precision); for (unsigned i = 0; i < PartCount - 1; i++) { if (~Parts[i] & ~unsigned{!i}) return false; } // Set the unused high bits to all ones when we compare. const unsigned NumHighBits = PartCount * integerPartWidth - semantics->precision + 1; assert(NumHighBits <= integerPartWidth && NumHighBits > 0 && "Can not have more high bits to fill than integerPartWidth"); const integerPart HighBitFill = ~integerPart(0) << (integerPartWidth - NumHighBits); if (~(Parts[PartCount - 1] | HighBitFill | 0x1)) return false; return true; } bool IEEEFloat::isSignificandAllZeros() const { // Test if the significand excluding the integral bit is all zeros. This // allows us to test for binade boundaries. const integerPart *Parts = significandParts(); const unsigned PartCount = partCountForBits(semantics->precision); for (unsigned i = 0; i < PartCount - 1; i++) if (Parts[i]) return false; // Compute how many bits are used in the final word. const unsigned NumHighBits = PartCount*integerPartWidth - semantics->precision + 1; assert(NumHighBits < integerPartWidth && "Can not have more high bits to " "clear than integerPartWidth"); const integerPart HighBitMask = ~integerPart(0) >> NumHighBits; if (Parts[PartCount - 1] & HighBitMask) return false; return true; } bool IEEEFloat::isSignificandAllZerosExceptMSB() const { const integerPart *Parts = significandParts(); const unsigned PartCount = partCountForBits(semantics->precision); for (unsigned i = 0; i < PartCount - 1; i++) { if (Parts[i]) return false; } const unsigned NumHighBits = PartCount * integerPartWidth - semantics->precision + 1; return Parts[PartCount - 1] == integerPart(1) << (integerPartWidth - NumHighBits); } bool IEEEFloat::isLargest() const { if (semantics->nonFiniteBehavior == fltNonfiniteBehavior::NanOnly && semantics->nanEncoding == fltNanEncoding::AllOnes) { // The largest number by magnitude in our format will be the floating point // number with maximum exponent and with significand that is all ones except // the LSB. return isFiniteNonZero() && exponent == semantics->maxExponent && isSignificandAllOnesExceptLSB(); } else { // The largest number by magnitude in our format will be the floating point // number with maximum exponent and with significand that is all ones. return isFiniteNonZero() && exponent == semantics->maxExponent && isSignificandAllOnes(); } } bool IEEEFloat::isInteger() const { // This could be made more efficient; I'm going for obviously correct. if (!isFinite()) return false; IEEEFloat truncated = *this; truncated.roundToIntegral(rmTowardZero); return compare(truncated) == cmpEqual; } bool IEEEFloat::bitwiseIsEqual(const IEEEFloat &rhs) const { if (this == &rhs) return true; if (semantics != rhs.semantics || category != rhs.category || sign != rhs.sign) return false; if (category==fcZero || category==fcInfinity) return true; if (isFiniteNonZero() && exponent != rhs.exponent) return false; return std::equal(significandParts(), significandParts() + partCount(), rhs.significandParts()); } IEEEFloat::IEEEFloat(const fltSemantics &ourSemantics, integerPart value) { initialize(&ourSemantics); sign = 0; category = fcNormal; zeroSignificand(); exponent = ourSemantics.precision - 1; significandParts()[0] = value; normalize(rmNearestTiesToEven, lfExactlyZero); } IEEEFloat::IEEEFloat(const fltSemantics &ourSemantics) { initialize(&ourSemantics); makeZero(false); } // Delegate to the previous constructor, because later copy constructor may // actually inspects category, which can't be garbage. IEEEFloat::IEEEFloat(const fltSemantics &ourSemantics, uninitializedTag tag) : IEEEFloat(ourSemantics) {} IEEEFloat::IEEEFloat(const IEEEFloat &rhs) { initialize(rhs.semantics); assign(rhs); } IEEEFloat::IEEEFloat(IEEEFloat &&rhs) : semantics(&semBogus) { *this = std::move(rhs); } IEEEFloat::~IEEEFloat() { freeSignificand(); } unsigned int IEEEFloat::partCount() const { return partCountForBits(semantics->precision + 1); } const IEEEFloat::integerPart *IEEEFloat::significandParts() const { return const_cast(this)->significandParts(); } IEEEFloat::integerPart *IEEEFloat::significandParts() { if (partCount() > 1) return significand.parts; else return &significand.part; } void IEEEFloat::zeroSignificand() { APInt::tcSet(significandParts(), 0, partCount()); } /* Increment an fcNormal floating point number's significand. */ void IEEEFloat::incrementSignificand() { integerPart carry; carry = APInt::tcIncrement(significandParts(), partCount()); /* Our callers should never cause us to overflow. */ assert(carry == 0); (void)carry; } /* Add the significand of the RHS. Returns the carry flag. */ IEEEFloat::integerPart IEEEFloat::addSignificand(const IEEEFloat &rhs) { integerPart *parts; parts = significandParts(); assert(semantics == rhs.semantics); assert(exponent == rhs.exponent); return APInt::tcAdd(parts, rhs.significandParts(), 0, partCount()); } /* Subtract the significand of the RHS with a borrow flag. Returns the borrow flag. */ IEEEFloat::integerPart IEEEFloat::subtractSignificand(const IEEEFloat &rhs, integerPart borrow) { integerPart *parts; parts = significandParts(); assert(semantics == rhs.semantics); assert(exponent == rhs.exponent); return APInt::tcSubtract(parts, rhs.significandParts(), borrow, partCount()); } /* Multiply the significand of the RHS. If ADDEND is non-NULL, add it on to the full-precision result of the multiplication. Returns the lost fraction. */ lostFraction IEEEFloat::multiplySignificand(const IEEEFloat &rhs, IEEEFloat addend) { unsigned int omsb; // One, not zero, based MSB. unsigned int partsCount, newPartsCount, precision; integerPart *lhsSignificand; integerPart scratch[4]; integerPart *fullSignificand; lostFraction lost_fraction; bool ignored; assert(semantics == rhs.semantics); precision = semantics->precision; // Allocate space for twice as many bits as the original significand, plus one // extra bit for the addition to overflow into. newPartsCount = partCountForBits(precision * 2 + 1); if (newPartsCount > 4) fullSignificand = new integerPart[newPartsCount]; else fullSignificand = scratch; lhsSignificand = significandParts(); partsCount = partCount(); APInt::tcFullMultiply(fullSignificand, lhsSignificand, rhs.significandParts(), partsCount, partsCount); lost_fraction = lfExactlyZero; omsb = APInt::tcMSB(fullSignificand, newPartsCount) + 1; exponent += rhs.exponent; // Assume the operands involved in the multiplication are single-precision // FP, and the two multiplicants are: // *this = a23 . a22 ... a0 * 2^e1 // rhs = b23 . b22 ... b0 * 2^e2 // the result of multiplication is: // *this = c48 c47 c46 . c45 ... c0 * 2^(e1+e2) // Note that there are three significant bits at the left-hand side of the // radix point: two for the multiplication, and an overflow bit for the // addition (that will always be zero at this point). Move the radix point // toward left by two bits, and adjust exponent accordingly. exponent += 2; if (addend.isNonZero()) { // The intermediate result of the multiplication has "2 * precision" // signicant bit; adjust the addend to be consistent with mul result. // Significand savedSignificand = significand; const fltSemantics *savedSemantics = semantics; fltSemantics extendedSemantics; opStatus status; unsigned int extendedPrecision; // Normalize our MSB to one below the top bit to allow for overflow. extendedPrecision = 2 * precision + 1; if (omsb != extendedPrecision - 1) { assert(extendedPrecision > omsb); APInt::tcShiftLeft(fullSignificand, newPartsCount, (extendedPrecision - 1) - omsb); exponent -= (extendedPrecision - 1) - omsb; } /* Create new semantics. */ extendedSemantics = *semantics; extendedSemantics.precision = extendedPrecision; if (newPartsCount == 1) significand.part = fullSignificand[0]; else significand.parts = fullSignificand; semantics = &extendedSemantics; // Make a copy so we can convert it to the extended semantics. // Note that we cannot convert the addend directly, as the extendedSemantics // is a local variable (which we take a reference to). IEEEFloat extendedAddend(addend); status = extendedAddend.convert(extendedSemantics, rmTowardZero, &ignored); assert(status == opOK); (void)status; // Shift the significand of the addend right by one bit. This guarantees // that the high bit of the significand is zero (same as fullSignificand), // so the addition will overflow (if it does overflow at all) into the top bit. lost_fraction = extendedAddend.shiftSignificandRight(1); assert(lost_fraction == lfExactlyZero && "Lost precision while shifting addend for fused-multiply-add."); lost_fraction = addOrSubtractSignificand(extendedAddend, false); /* Restore our state. */ if (newPartsCount == 1) fullSignificand[0] = significand.part; significand = savedSignificand; semantics = savedSemantics; omsb = APInt::tcMSB(fullSignificand, newPartsCount) + 1; } // Convert the result having "2 * precision" significant-bits back to the one // having "precision" significant-bits. First, move the radix point from // poision "2*precision - 1" to "precision - 1". The exponent need to be // adjusted by "2*precision - 1" - "precision - 1" = "precision". exponent -= precision + 1; // In case MSB resides at the left-hand side of radix point, shift the // mantissa right by some amount to make sure the MSB reside right before // the radix point (i.e. "MSB . rest-significant-bits"). // // Note that the result is not normalized when "omsb < precision". So, the // caller needs to call IEEEFloat::normalize() if normalized value is // expected. if (omsb > precision) { unsigned int bits, significantParts; lostFraction lf; bits = omsb - precision; significantParts = partCountForBits(omsb); lf = shiftRight(fullSignificand, significantParts, bits); lost_fraction = combineLostFractions(lf, lost_fraction); exponent += bits; } APInt::tcAssign(lhsSignificand, fullSignificand, partsCount); if (newPartsCount > 4) delete [] fullSignificand; return lost_fraction; } lostFraction IEEEFloat::multiplySignificand(const IEEEFloat &rhs) { return multiplySignificand(rhs, IEEEFloat(*semantics)); } /* Multiply the significands of LHS and RHS to DST. */ lostFraction IEEEFloat::divideSignificand(const IEEEFloat &rhs) { unsigned int bit, i, partsCount; const integerPart *rhsSignificand; integerPart *lhsSignificand, *dividend, *divisor; integerPart scratch[4]; lostFraction lost_fraction; assert(semantics == rhs.semantics); lhsSignificand = significandParts(); rhsSignificand = rhs.significandParts(); partsCount = partCount(); if (partsCount > 2) dividend = new integerPart[partsCount * 2]; else dividend = scratch; divisor = dividend + partsCount; /* Copy the dividend and divisor as they will be modified in-place. */ for (i = 0; i < partsCount; i++) { dividend[i] = lhsSignificand[i]; divisor[i] = rhsSignificand[i]; lhsSignificand[i] = 0; } exponent -= rhs.exponent; unsigned int precision = semantics->precision; /* Normalize the divisor. */ bit = precision - APInt::tcMSB(divisor, partsCount) - 1; if (bit) { exponent += bit; APInt::tcShiftLeft(divisor, partsCount, bit); } /* Normalize the dividend. */ bit = precision - APInt::tcMSB(dividend, partsCount) - 1; if (bit) { exponent -= bit; APInt::tcShiftLeft(dividend, partsCount, bit); } /* Ensure the dividend >= divisor initially for the loop below. Incidentally, this means that the division loop below is guaranteed to set the integer bit to one. */ if (APInt::tcCompare(dividend, divisor, partsCount) < 0) { exponent--; APInt::tcShiftLeft(dividend, partsCount, 1); assert(APInt::tcCompare(dividend, divisor, partsCount) >= 0); } /* Long division. */ for (bit = precision; bit; bit -= 1) { if (APInt::tcCompare(dividend, divisor, partsCount) >= 0) { APInt::tcSubtract(dividend, divisor, 0, partsCount); APInt::tcSetBit(lhsSignificand, bit - 1); } APInt::tcShiftLeft(dividend, partsCount, 1); } /* Figure out the lost fraction. */ int cmp = APInt::tcCompare(dividend, divisor, partsCount); if (cmp > 0) lost_fraction = lfMoreThanHalf; else if (cmp == 0) lost_fraction = lfExactlyHalf; else if (APInt::tcIsZero(dividend, partsCount)) lost_fraction = lfExactlyZero; else lost_fraction = lfLessThanHalf; if (partsCount > 2) delete [] dividend; return lost_fraction; } unsigned int IEEEFloat::significandMSB() const { return APInt::tcMSB(significandParts(), partCount()); } unsigned int IEEEFloat::significandLSB() const { return APInt::tcLSB(significandParts(), partCount()); } /* Note that a zero result is NOT normalized to fcZero. */ lostFraction IEEEFloat::shiftSignificandRight(unsigned int bits) { /* Our exponent should not overflow. */ assert((ExponentType) (exponent + bits) >= exponent); exponent += bits; return shiftRight(significandParts(), partCount(), bits); } /* Shift the significand left BITS bits, subtract BITS from its exponent. */ void IEEEFloat::shiftSignificandLeft(unsigned int bits) { assert(bits < semantics->precision); if (bits) { unsigned int partsCount = partCount(); APInt::tcShiftLeft(significandParts(), partsCount, bits); exponent -= bits; assert(!APInt::tcIsZero(significandParts(), partsCount)); } } IEEEFloat::cmpResult IEEEFloat::compareAbsoluteValue(const IEEEFloat &rhs) const { int compare; assert(semantics == rhs.semantics); assert(isFiniteNonZero()); assert(rhs.isFiniteNonZero()); compare = exponent - rhs.exponent; /* If exponents are equal, do an unsigned bignum comparison of the significands. */ if (compare == 0) compare = APInt::tcCompare(significandParts(), rhs.significandParts(), partCount()); if (compare > 0) return cmpGreaterThan; else if (compare < 0) return cmpLessThan; else return cmpEqual; } /* Set the least significant BITS bits of a bignum, clear the rest. */ static void tcSetLeastSignificantBits(APInt::WordType *dst, unsigned parts, unsigned bits) { unsigned i = 0; while (bits > APInt::APINT_BITS_PER_WORD) { dst[i++] = ~(APInt::WordType)0; bits -= APInt::APINT_BITS_PER_WORD; } if (bits) dst[i++] = ~(APInt::WordType)0 >> (APInt::APINT_BITS_PER_WORD - bits); while (i < parts) dst[i++] = 0; } /* Handle overflow. Sign is preserved. We either become infinity or the largest finite number. */ IEEEFloat::opStatus IEEEFloat::handleOverflow(roundingMode rounding_mode) { if (semantics->nonFiniteBehavior != fltNonfiniteBehavior::FiniteOnly) { /* Infinity? */ if (rounding_mode == rmNearestTiesToEven || rounding_mode == rmNearestTiesToAway || (rounding_mode == rmTowardPositive && !sign) || (rounding_mode == rmTowardNegative && sign)) { if (semantics->nonFiniteBehavior == fltNonfiniteBehavior::NanOnly) makeNaN(false, sign); else category = fcInfinity; return static_cast(opOverflow | opInexact); } } /* Otherwise we become the largest finite number. */ category = fcNormal; exponent = semantics->maxExponent; tcSetLeastSignificantBits(significandParts(), partCount(), semantics->precision); if (semantics->nonFiniteBehavior == fltNonfiniteBehavior::NanOnly && semantics->nanEncoding == fltNanEncoding::AllOnes) APInt::tcClearBit(significandParts(), 0); return opInexact; } /* Returns TRUE if, when truncating the current number, with BIT the new LSB, with the given lost fraction and rounding mode, the result would need to be rounded away from zero (i.e., by increasing the signficand). This routine must work for fcZero of both signs, and fcNormal numbers. */ bool IEEEFloat::roundAwayFromZero(roundingMode rounding_mode, lostFraction lost_fraction, unsigned int bit) const { /* NaNs and infinities should not have lost fractions. */ assert(isFiniteNonZero() || category == fcZero); /* Current callers never pass this so we don't handle it. */ assert(lost_fraction != lfExactlyZero); switch (rounding_mode) { case rmNearestTiesToAway: return lost_fraction == lfExactlyHalf || lost_fraction == lfMoreThanHalf; case rmNearestTiesToEven: if (lost_fraction == lfMoreThanHalf) return true; /* Our zeroes don't have a significand to test. */ if (lost_fraction == lfExactlyHalf && category != fcZero) return APInt::tcExtractBit(significandParts(), bit); return false; case rmTowardZero: return false; case rmTowardPositive: return !sign; case rmTowardNegative: return sign; default: break; } llvm_unreachable("Invalid rounding mode found"); } IEEEFloat::opStatus IEEEFloat::normalize(roundingMode rounding_mode, lostFraction lost_fraction) { unsigned int omsb; /* One, not zero, based MSB. */ int exponentChange; if (!isFiniteNonZero()) return opOK; /* Before rounding normalize the exponent of fcNormal numbers. */ omsb = significandMSB() + 1; if (omsb) { /* OMSB is numbered from 1. We want to place it in the integer bit numbered PRECISION if possible, with a compensating change in the exponent. */ exponentChange = omsb - semantics->precision; /* If the resulting exponent is too high, overflow according to the rounding mode. */ if (exponent + exponentChange > semantics->maxExponent) return handleOverflow(rounding_mode); /* Subnormal numbers have exponent minExponent, and their MSB is forced based on that. */ if (exponent + exponentChange < semantics->minExponent) exponentChange = semantics->minExponent - exponent; /* Shifting left is easy as we don't lose precision. */ if (exponentChange < 0) { assert(lost_fraction == lfExactlyZero); shiftSignificandLeft(-exponentChange); return opOK; } if (exponentChange > 0) { lostFraction lf; /* Shift right and capture any new lost fraction. */ lf = shiftSignificandRight(exponentChange); lost_fraction = combineLostFractions(lf, lost_fraction); /* Keep OMSB up-to-date. */ if (omsb > (unsigned) exponentChange) omsb -= exponentChange; else omsb = 0; } } // The all-ones values is an overflow if NaN is all ones. If NaN is // represented by negative zero, then it is a valid finite value. if (semantics->nonFiniteBehavior == fltNonfiniteBehavior::NanOnly && semantics->nanEncoding == fltNanEncoding::AllOnes && exponent == semantics->maxExponent && isSignificandAllOnes()) return handleOverflow(rounding_mode); /* Now round the number according to rounding_mode given the lost fraction. */ /* As specified in IEEE 754, since we do not trap we do not report underflow for exact results. */ if (lost_fraction == lfExactlyZero) { /* Canonicalize zeroes. */ if (omsb == 0) { category = fcZero; if (semantics->nanEncoding == fltNanEncoding::NegativeZero) sign = false; } return opOK; } /* Increment the significand if we're rounding away from zero. */ if (roundAwayFromZero(rounding_mode, lost_fraction, 0)) { if (omsb == 0) exponent = semantics->minExponent; incrementSignificand(); omsb = significandMSB() + 1; /* Did the significand increment overflow? */ if (omsb == (unsigned) semantics->precision + 1) { /* Renormalize by incrementing the exponent and shifting our significand right one. However if we already have the maximum exponent we overflow to infinity. */ if (exponent == semantics->maxExponent) // Invoke overflow handling with a rounding mode that will guarantee // that the result gets turned into the correct infinity representation. // This is needed instead of just setting the category to infinity to // account for 8-bit floating point types that have no inf, only NaN. return handleOverflow(sign ? rmTowardNegative : rmTowardPositive); shiftSignificandRight(1); return opInexact; } // The all-ones values is an overflow if NaN is all ones. If NaN is // represented by negative zero, then it is a valid finite value. if (semantics->nonFiniteBehavior == fltNonfiniteBehavior::NanOnly && semantics->nanEncoding == fltNanEncoding::AllOnes && exponent == semantics->maxExponent && isSignificandAllOnes()) return handleOverflow(rounding_mode); } /* The normal case - we were and are not denormal, and any significand increment above didn't overflow. */ if (omsb == semantics->precision) return opInexact; /* We have a non-zero denormal. */ assert(omsb < semantics->precision); /* Canonicalize zeroes. */ if (omsb == 0) { category = fcZero; if (semantics->nanEncoding == fltNanEncoding::NegativeZero) sign = false; } /* The fcZero case is a denormal that underflowed to zero. */ return (opStatus) (opUnderflow | opInexact); } IEEEFloat::opStatus IEEEFloat::addOrSubtractSpecials(const IEEEFloat &rhs, bool subtract) { switch (PackCategoriesIntoKey(category, rhs.category)) { default: llvm_unreachable(nullptr); case PackCategoriesIntoKey(fcZero, fcNaN): case PackCategoriesIntoKey(fcNormal, fcNaN): case PackCategoriesIntoKey(fcInfinity, fcNaN): assign(rhs); [[fallthrough]]; case PackCategoriesIntoKey(fcNaN, fcZero): case PackCategoriesIntoKey(fcNaN, fcNormal): case PackCategoriesIntoKey(fcNaN, fcInfinity): case PackCategoriesIntoKey(fcNaN, fcNaN): if (isSignaling()) { makeQuiet(); return opInvalidOp; } return rhs.isSignaling() ? opInvalidOp : opOK; case PackCategoriesIntoKey(fcNormal, fcZero): case PackCategoriesIntoKey(fcInfinity, fcNormal): case PackCategoriesIntoKey(fcInfinity, fcZero): return opOK; case PackCategoriesIntoKey(fcNormal, fcInfinity): case PackCategoriesIntoKey(fcZero, fcInfinity): category = fcInfinity; sign = rhs.sign ^ subtract; return opOK; case PackCategoriesIntoKey(fcZero, fcNormal): assign(rhs); sign = rhs.sign ^ subtract; return opOK; case PackCategoriesIntoKey(fcZero, fcZero): /* Sign depends on rounding mode; handled by caller. */ return opOK; case PackCategoriesIntoKey(fcInfinity, fcInfinity): /* Differently signed infinities can only be validly subtracted. */ if (((sign ^ rhs.sign)!=0) != subtract) { makeNaN(); return opInvalidOp; } return opOK; case PackCategoriesIntoKey(fcNormal, fcNormal): return opDivByZero; } } /* Add or subtract two normal numbers. */ lostFraction IEEEFloat::addOrSubtractSignificand(const IEEEFloat &rhs, bool subtract) { integerPart carry; lostFraction lost_fraction; int bits; /* Determine if the operation on the absolute values is effectively an addition or subtraction. */ subtract ^= static_cast(sign ^ rhs.sign); /* Are we bigger exponent-wise than the RHS? */ bits = exponent - rhs.exponent; /* Subtraction is more subtle than one might naively expect. */ if (subtract) { IEEEFloat temp_rhs(rhs); if (bits == 0) lost_fraction = lfExactlyZero; else if (bits > 0) { lost_fraction = temp_rhs.shiftSignificandRight(bits - 1); shiftSignificandLeft(1); } else { lost_fraction = shiftSignificandRight(-bits - 1); temp_rhs.shiftSignificandLeft(1); } // Should we reverse the subtraction. if (compareAbsoluteValue(temp_rhs) == cmpLessThan) { carry = temp_rhs.subtractSignificand (*this, lost_fraction != lfExactlyZero); copySignificand(temp_rhs); sign = !sign; } else { carry = subtractSignificand (temp_rhs, lost_fraction != lfExactlyZero); } /* Invert the lost fraction - it was on the RHS and subtracted. */ if (lost_fraction == lfLessThanHalf) lost_fraction = lfMoreThanHalf; else if (lost_fraction == lfMoreThanHalf) lost_fraction = lfLessThanHalf; /* The code above is intended to ensure that no borrow is necessary. */ assert(!carry); (void)carry; } else { if (bits > 0) { IEEEFloat temp_rhs(rhs); lost_fraction = temp_rhs.shiftSignificandRight(bits); carry = addSignificand(temp_rhs); } else { lost_fraction = shiftSignificandRight(-bits); carry = addSignificand(rhs); } /* We have a guard bit; generating a carry cannot happen. */ assert(!carry); (void)carry; } return lost_fraction; } IEEEFloat::opStatus IEEEFloat::multiplySpecials(const IEEEFloat &rhs) { switch (PackCategoriesIntoKey(category, rhs.category)) { default: llvm_unreachable(nullptr); case PackCategoriesIntoKey(fcZero, fcNaN): case PackCategoriesIntoKey(fcNormal, fcNaN): case PackCategoriesIntoKey(fcInfinity, fcNaN): assign(rhs); sign = false; [[fallthrough]]; case PackCategoriesIntoKey(fcNaN, fcZero): case PackCategoriesIntoKey(fcNaN, fcNormal): case PackCategoriesIntoKey(fcNaN, fcInfinity): case PackCategoriesIntoKey(fcNaN, fcNaN): sign ^= rhs.sign; // restore the original sign if (isSignaling()) { makeQuiet(); return opInvalidOp; } return rhs.isSignaling() ? opInvalidOp : opOK; case PackCategoriesIntoKey(fcNormal, fcInfinity): case PackCategoriesIntoKey(fcInfinity, fcNormal): case PackCategoriesIntoKey(fcInfinity, fcInfinity): category = fcInfinity; return opOK; case PackCategoriesIntoKey(fcZero, fcNormal): case PackCategoriesIntoKey(fcNormal, fcZero): case PackCategoriesIntoKey(fcZero, fcZero): category = fcZero; return opOK; case PackCategoriesIntoKey(fcZero, fcInfinity): case PackCategoriesIntoKey(fcInfinity, fcZero): makeNaN(); return opInvalidOp; case PackCategoriesIntoKey(fcNormal, fcNormal): return opOK; } } IEEEFloat::opStatus IEEEFloat::divideSpecials(const IEEEFloat &rhs) { switch (PackCategoriesIntoKey(category, rhs.category)) { default: llvm_unreachable(nullptr); case PackCategoriesIntoKey(fcZero, fcNaN): case PackCategoriesIntoKey(fcNormal, fcNaN): case PackCategoriesIntoKey(fcInfinity, fcNaN): assign(rhs); sign = false; [[fallthrough]]; case PackCategoriesIntoKey(fcNaN, fcZero): case PackCategoriesIntoKey(fcNaN, fcNormal): case PackCategoriesIntoKey(fcNaN, fcInfinity): case PackCategoriesIntoKey(fcNaN, fcNaN): sign ^= rhs.sign; // restore the original sign if (isSignaling()) { makeQuiet(); return opInvalidOp; } return rhs.isSignaling() ? opInvalidOp : opOK; case PackCategoriesIntoKey(fcInfinity, fcZero): case PackCategoriesIntoKey(fcInfinity, fcNormal): case PackCategoriesIntoKey(fcZero, fcInfinity): case PackCategoriesIntoKey(fcZero, fcNormal): return opOK; case PackCategoriesIntoKey(fcNormal, fcInfinity): category = fcZero; return opOK; case PackCategoriesIntoKey(fcNormal, fcZero): if (semantics->nonFiniteBehavior == fltNonfiniteBehavior::NanOnly) makeNaN(false, sign); else category = fcInfinity; return opDivByZero; case PackCategoriesIntoKey(fcInfinity, fcInfinity): case PackCategoriesIntoKey(fcZero, fcZero): makeNaN(); return opInvalidOp; case PackCategoriesIntoKey(fcNormal, fcNormal): return opOK; } } IEEEFloat::opStatus IEEEFloat::modSpecials(const IEEEFloat &rhs) { switch (PackCategoriesIntoKey(category, rhs.category)) { default: llvm_unreachable(nullptr); case PackCategoriesIntoKey(fcZero, fcNaN): case PackCategoriesIntoKey(fcNormal, fcNaN): case PackCategoriesIntoKey(fcInfinity, fcNaN): assign(rhs); [[fallthrough]]; case PackCategoriesIntoKey(fcNaN, fcZero): case PackCategoriesIntoKey(fcNaN, fcNormal): case PackCategoriesIntoKey(fcNaN, fcInfinity): case PackCategoriesIntoKey(fcNaN, fcNaN): if (isSignaling()) { makeQuiet(); return opInvalidOp; } return rhs.isSignaling() ? opInvalidOp : opOK; case PackCategoriesIntoKey(fcZero, fcInfinity): case PackCategoriesIntoKey(fcZero, fcNormal): case PackCategoriesIntoKey(fcNormal, fcInfinity): return opOK; case PackCategoriesIntoKey(fcNormal, fcZero): case PackCategoriesIntoKey(fcInfinity, fcZero): case PackCategoriesIntoKey(fcInfinity, fcNormal): case PackCategoriesIntoKey(fcInfinity, fcInfinity): case PackCategoriesIntoKey(fcZero, fcZero): makeNaN(); return opInvalidOp; case PackCategoriesIntoKey(fcNormal, fcNormal): return opOK; } } IEEEFloat::opStatus IEEEFloat::remainderSpecials(const IEEEFloat &rhs) { switch (PackCategoriesIntoKey(category, rhs.category)) { default: llvm_unreachable(nullptr); case PackCategoriesIntoKey(fcZero, fcNaN): case PackCategoriesIntoKey(fcNormal, fcNaN): case PackCategoriesIntoKey(fcInfinity, fcNaN): assign(rhs); [[fallthrough]]; case PackCategoriesIntoKey(fcNaN, fcZero): case PackCategoriesIntoKey(fcNaN, fcNormal): case PackCategoriesIntoKey(fcNaN, fcInfinity): case PackCategoriesIntoKey(fcNaN, fcNaN): if (isSignaling()) { makeQuiet(); return opInvalidOp; } return rhs.isSignaling() ? opInvalidOp : opOK; case PackCategoriesIntoKey(fcZero, fcInfinity): case PackCategoriesIntoKey(fcZero, fcNormal): case PackCategoriesIntoKey(fcNormal, fcInfinity): return opOK; case PackCategoriesIntoKey(fcNormal, fcZero): case PackCategoriesIntoKey(fcInfinity, fcZero): case PackCategoriesIntoKey(fcInfinity, fcNormal): case PackCategoriesIntoKey(fcInfinity, fcInfinity): case PackCategoriesIntoKey(fcZero, fcZero): makeNaN(); return opInvalidOp; case PackCategoriesIntoKey(fcNormal, fcNormal): return opDivByZero; // fake status, indicating this is not a special case } } /* Change sign. */ void IEEEFloat::changeSign() { // With NaN-as-negative-zero, neither NaN or negative zero can change // their signs. if (semantics->nanEncoding == fltNanEncoding::NegativeZero && (isZero() || isNaN())) return; /* Look mummy, this one's easy. */ sign = !sign; } /* Normalized addition or subtraction. */ IEEEFloat::opStatus IEEEFloat::addOrSubtract(const IEEEFloat &rhs, roundingMode rounding_mode, bool subtract) { opStatus fs; fs = addOrSubtractSpecials(rhs, subtract); /* This return code means it was not a simple case. */ if (fs == opDivByZero) { lostFraction lost_fraction; lost_fraction = addOrSubtractSignificand(rhs, subtract); fs = normalize(rounding_mode, lost_fraction); /* Can only be zero if we lost no fraction. */ assert(category != fcZero || lost_fraction == lfExactlyZero); } /* If two numbers add (exactly) to zero, IEEE 754 decrees it is a positive zero unless rounding to minus infinity, except that adding two like-signed zeroes gives that zero. */ if (category == fcZero) { if (rhs.category != fcZero || (sign == rhs.sign) == subtract) sign = (rounding_mode == rmTowardNegative); // NaN-in-negative-zero means zeros need to be normalized to +0. if (semantics->nanEncoding == fltNanEncoding::NegativeZero) sign = false; } return fs; } /* Normalized addition. */ IEEEFloat::opStatus IEEEFloat::add(const IEEEFloat &rhs, roundingMode rounding_mode) { return addOrSubtract(rhs, rounding_mode, false); } /* Normalized subtraction. */ IEEEFloat::opStatus IEEEFloat::subtract(const IEEEFloat &rhs, roundingMode rounding_mode) { return addOrSubtract(rhs, rounding_mode, true); } /* Normalized multiply. */ IEEEFloat::opStatus IEEEFloat::multiply(const IEEEFloat &rhs, roundingMode rounding_mode) { opStatus fs; sign ^= rhs.sign; fs = multiplySpecials(rhs); if (isZero() && semantics->nanEncoding == fltNanEncoding::NegativeZero) sign = false; if (isFiniteNonZero()) { lostFraction lost_fraction = multiplySignificand(rhs); fs = normalize(rounding_mode, lost_fraction); if (lost_fraction != lfExactlyZero) fs = (opStatus) (fs | opInexact); } return fs; } /* Normalized divide. */ IEEEFloat::opStatus IEEEFloat::divide(const IEEEFloat &rhs, roundingMode rounding_mode) { opStatus fs; sign ^= rhs.sign; fs = divideSpecials(rhs); if (isZero() && semantics->nanEncoding == fltNanEncoding::NegativeZero) sign = false; if (isFiniteNonZero()) { lostFraction lost_fraction = divideSignificand(rhs); fs = normalize(rounding_mode, lost_fraction); if (lost_fraction != lfExactlyZero) fs = (opStatus) (fs | opInexact); } return fs; } /* Normalized remainder. */ IEEEFloat::opStatus IEEEFloat::remainder(const IEEEFloat &rhs) { opStatus fs; unsigned int origSign = sign; // First handle the special cases. fs = remainderSpecials(rhs); if (fs != opDivByZero) return fs; fs = opOK; // Make sure the current value is less than twice the denom. If the addition // did not succeed (an overflow has happened), which means that the finite // value we currently posses must be less than twice the denom (as we are // using the same semantics). IEEEFloat P2 = rhs; if (P2.add(rhs, rmNearestTiesToEven) == opOK) { fs = mod(P2); assert(fs == opOK); } // Lets work with absolute numbers. IEEEFloat P = rhs; P.sign = false; sign = false; // // To calculate the remainder we use the following scheme. // // The remainder is defained as follows: // // remainder = numer - rquot * denom = x - r * p // // Where r is the result of: x/p, rounded toward the nearest integral value // (with halfway cases rounded toward the even number). // // Currently, (after x mod 2p): // r is the number of 2p's present inside x, which is inherently, an even // number of p's. // // We may split the remaining calculation into 4 options: // - if x < 0.5p then we round to the nearest number with is 0, and are done. // - if x == 0.5p then we round to the nearest even number which is 0, and we // are done as well. // - if 0.5p < x < p then we round to nearest number which is 1, and we have // to subtract 1p at least once. // - if x >= p then we must subtract p at least once, as x must be a // remainder. // // By now, we were done, or we added 1 to r, which in turn, now an odd number. // // We can now split the remaining calculation to the following 3 options: // - if x < 0.5p then we round to the nearest number with is 0, and are done. // - if x == 0.5p then we round to the nearest even number. As r is odd, we // must round up to the next even number. so we must subtract p once more. // - if x > 0.5p (and inherently x < p) then we must round r up to the next // integral, and subtract p once more. // // Extend the semantics to prevent an overflow/underflow or inexact result. bool losesInfo; fltSemantics extendedSemantics = *semantics; extendedSemantics.maxExponent++; extendedSemantics.minExponent--; extendedSemantics.precision += 2; IEEEFloat VEx = *this; fs = VEx.convert(extendedSemantics, rmNearestTiesToEven, &losesInfo); assert(fs == opOK && !losesInfo); IEEEFloat PEx = P; fs = PEx.convert(extendedSemantics, rmNearestTiesToEven, &losesInfo); assert(fs == opOK && !losesInfo); // It is simpler to work with 2x instead of 0.5p, and we do not need to lose // any fraction. fs = VEx.add(VEx, rmNearestTiesToEven); assert(fs == opOK); if (VEx.compare(PEx) == cmpGreaterThan) { fs = subtract(P, rmNearestTiesToEven); assert(fs == opOK); // Make VEx = this.add(this), but because we have different semantics, we do // not want to `convert` again, so we just subtract PEx twice (which equals // to the desired value). fs = VEx.subtract(PEx, rmNearestTiesToEven); assert(fs == opOK); fs = VEx.subtract(PEx, rmNearestTiesToEven); assert(fs == opOK); cmpResult result = VEx.compare(PEx); if (result == cmpGreaterThan || result == cmpEqual) { fs = subtract(P, rmNearestTiesToEven); assert(fs == opOK); } } if (isZero()) { sign = origSign; // IEEE754 requires this if (semantics->nanEncoding == fltNanEncoding::NegativeZero) // But some 8-bit floats only have positive 0. sign = false; } else sign ^= origSign; return fs; } /* Normalized llvm frem (C fmod). */ IEEEFloat::opStatus IEEEFloat::mod(const IEEEFloat &rhs) { opStatus fs; fs = modSpecials(rhs); unsigned int origSign = sign; while (isFiniteNonZero() && rhs.isFiniteNonZero() && compareAbsoluteValue(rhs) != cmpLessThan) { int Exp = ilogb(*this) - ilogb(rhs); IEEEFloat V = scalbn(rhs, Exp, rmNearestTiesToEven); // V can overflow to NaN with fltNonfiniteBehavior::NanOnly, so explicitly // check for it. if (V.isNaN() || compareAbsoluteValue(V) == cmpLessThan) V = scalbn(rhs, Exp - 1, rmNearestTiesToEven); V.sign = sign; fs = subtract(V, rmNearestTiesToEven); assert(fs==opOK); } if (isZero()) { sign = origSign; // fmod requires this if (semantics->nanEncoding == fltNanEncoding::NegativeZero) sign = false; } return fs; } /* Normalized fused-multiply-add. */ IEEEFloat::opStatus IEEEFloat::fusedMultiplyAdd(const IEEEFloat &multiplicand, const IEEEFloat &addend, roundingMode rounding_mode) { opStatus fs; /* Post-multiplication sign, before addition. */ sign ^= multiplicand.sign; /* If and only if all arguments are normal do we need to do an extended-precision calculation. */ if (isFiniteNonZero() && multiplicand.isFiniteNonZero() && addend.isFinite()) { lostFraction lost_fraction; lost_fraction = multiplySignificand(multiplicand, addend); fs = normalize(rounding_mode, lost_fraction); if (lost_fraction != lfExactlyZero) fs = (opStatus) (fs | opInexact); /* If two numbers add (exactly) to zero, IEEE 754 decrees it is a positive zero unless rounding to minus infinity, except that adding two like-signed zeroes gives that zero. */ if (category == fcZero && !(fs & opUnderflow) && sign != addend.sign) { sign = (rounding_mode == rmTowardNegative); if (semantics->nanEncoding == fltNanEncoding::NegativeZero) sign = false; } } else { fs = multiplySpecials(multiplicand); /* FS can only be opOK or opInvalidOp. There is no more work to do in the latter case. The IEEE-754R standard says it is implementation-defined in this case whether, if ADDEND is a quiet NaN, we raise invalid op; this implementation does so. If we need to do the addition we can do so with normal precision. */ if (fs == opOK) fs = addOrSubtract(addend, rounding_mode, false); } return fs; } /* Rounding-mode correct round to integral value. */ IEEEFloat::opStatus IEEEFloat::roundToIntegral(roundingMode rounding_mode) { opStatus fs; if (isInfinity()) // [IEEE Std 754-2008 6.1]: // The behavior of infinity in floating-point arithmetic is derived from the // limiting cases of real arithmetic with operands of arbitrarily // large magnitude, when such a limit exists. // ... // Operations on infinite operands are usually exact and therefore signal no // exceptions ... return opOK; if (isNaN()) { if (isSignaling()) { // [IEEE Std 754-2008 6.2]: // Under default exception handling, any operation signaling an invalid // operation exception and for which a floating-point result is to be // delivered shall deliver a quiet NaN. makeQuiet(); // [IEEE Std 754-2008 6.2]: // Signaling NaNs shall be reserved operands that, under default exception // handling, signal the invalid operation exception(see 7.2) for every // general-computational and signaling-computational operation except for // the conversions described in 5.12. return opInvalidOp; } else { // [IEEE Std 754-2008 6.2]: // For an operation with quiet NaN inputs, other than maximum and minimum // operations, if a floating-point result is to be delivered the result // shall be a quiet NaN which should be one of the input NaNs. // ... // Every general-computational and quiet-computational operation involving // one or more input NaNs, none of them signaling, shall signal no // exception, except fusedMultiplyAdd might signal the invalid operation // exception(see 7.2). return opOK; } } if (isZero()) { // [IEEE Std 754-2008 6.3]: // ... the sign of the result of conversions, the quantize operation, the // roundToIntegral operations, and the roundToIntegralExact(see 5.3.1) is // the sign of the first or only operand. return opOK; } // If the exponent is large enough, we know that this value is already // integral, and the arithmetic below would potentially cause it to saturate // to +/-Inf. Bail out early instead. if (exponent+1 >= (int)semanticsPrecision(*semantics)) return opOK; // The algorithm here is quite simple: we add 2^(p-1), where p is the // precision of our format, and then subtract it back off again. The choice // of rounding modes for the addition/subtraction determines the rounding mode // for our integral rounding as well. // NOTE: When the input value is negative, we do subtraction followed by // addition instead. APInt IntegerConstant(NextPowerOf2(semanticsPrecision(*semantics)), 1); IntegerConstant <<= semanticsPrecision(*semantics)-1; IEEEFloat MagicConstant(*semantics); fs = MagicConstant.convertFromAPInt(IntegerConstant, false, rmNearestTiesToEven); assert(fs == opOK); MagicConstant.sign = sign; // Preserve the input sign so that we can handle the case of zero result // correctly. bool inputSign = isNegative(); fs = add(MagicConstant, rounding_mode); // Current value and 'MagicConstant' are both integers, so the result of the // subtraction is always exact according to Sterbenz' lemma. subtract(MagicConstant, rounding_mode); // Restore the input sign. if (inputSign != isNegative()) changeSign(); return fs; } /* Comparison requires normalized numbers. */ IEEEFloat::cmpResult IEEEFloat::compare(const IEEEFloat &rhs) const { cmpResult result; assert(semantics == rhs.semantics); switch (PackCategoriesIntoKey(category, rhs.category)) { default: llvm_unreachable(nullptr); case PackCategoriesIntoKey(fcNaN, fcZero): case PackCategoriesIntoKey(fcNaN, fcNormal): case PackCategoriesIntoKey(fcNaN, fcInfinity): case PackCategoriesIntoKey(fcNaN, fcNaN): case PackCategoriesIntoKey(fcZero, fcNaN): case PackCategoriesIntoKey(fcNormal, fcNaN): case PackCategoriesIntoKey(fcInfinity, fcNaN): return cmpUnordered; case PackCategoriesIntoKey(fcInfinity, fcNormal): case PackCategoriesIntoKey(fcInfinity, fcZero): case PackCategoriesIntoKey(fcNormal, fcZero): if (sign) return cmpLessThan; else return cmpGreaterThan; case PackCategoriesIntoKey(fcNormal, fcInfinity): case PackCategoriesIntoKey(fcZero, fcInfinity): case PackCategoriesIntoKey(fcZero, fcNormal): if (rhs.sign) return cmpGreaterThan; else return cmpLessThan; case PackCategoriesIntoKey(fcInfinity, fcInfinity): if (sign == rhs.sign) return cmpEqual; else if (sign) return cmpLessThan; else return cmpGreaterThan; case PackCategoriesIntoKey(fcZero, fcZero): return cmpEqual; case PackCategoriesIntoKey(fcNormal, fcNormal): break; } /* Two normal numbers. Do they have the same sign? */ if (sign != rhs.sign) { if (sign) result = cmpLessThan; else result = cmpGreaterThan; } else { /* Compare absolute values; invert result if negative. */ result = compareAbsoluteValue(rhs); if (sign) { if (result == cmpLessThan) result = cmpGreaterThan; else if (result == cmpGreaterThan) result = cmpLessThan; } } return result; } /// IEEEFloat::convert - convert a value of one floating point type to another. /// The return value corresponds to the IEEE754 exceptions. *losesInfo /// records whether the transformation lost information, i.e. whether /// converting the result back to the original type will produce the /// original value (this is almost the same as return value==fsOK, but there /// are edge cases where this is not so). IEEEFloat::opStatus IEEEFloat::convert(const fltSemantics &toSemantics, roundingMode rounding_mode, bool *losesInfo) { lostFraction lostFraction; unsigned int newPartCount, oldPartCount; opStatus fs; int shift; const fltSemantics &fromSemantics = *semantics; bool is_signaling = isSignaling(); lostFraction = lfExactlyZero; newPartCount = partCountForBits(toSemantics.precision + 1); oldPartCount = partCount(); shift = toSemantics.precision - fromSemantics.precision; bool X86SpecialNan = false; if (&fromSemantics == &semX87DoubleExtended && &toSemantics != &semX87DoubleExtended && category == fcNaN && (!(*significandParts() & 0x8000000000000000ULL) || !(*significandParts() & 0x4000000000000000ULL))) { // x86 has some unusual NaNs which cannot be represented in any other // format; note them here. X86SpecialNan = true; } // If this is a truncation of a denormal number, and the target semantics // has larger exponent range than the source semantics (this can happen // when truncating from PowerPC double-double to double format), the // right shift could lose result mantissa bits. Adjust exponent instead // of performing excessive shift. // Also do a similar trick in case shifting denormal would produce zero // significand as this case isn't handled correctly by normalize. if (shift < 0 && isFiniteNonZero()) { int omsb = significandMSB() + 1; int exponentChange = omsb - fromSemantics.precision; if (exponent + exponentChange < toSemantics.minExponent) exponentChange = toSemantics.minExponent - exponent; if (exponentChange < shift) exponentChange = shift; if (exponentChange < 0) { shift -= exponentChange; exponent += exponentChange; } else if (omsb <= -shift) { exponentChange = omsb + shift - 1; // leave at least one bit set shift -= exponentChange; exponent += exponentChange; } } // If this is a truncation, perform the shift before we narrow the storage. if (shift < 0 && (isFiniteNonZero() || (category == fcNaN && semantics->nonFiniteBehavior != fltNonfiniteBehavior::NanOnly))) lostFraction = shiftRight(significandParts(), oldPartCount, -shift); // Fix the storage so it can hold to new value. if (newPartCount > oldPartCount) { // The new type requires more storage; make it available. integerPart *newParts; newParts = new integerPart[newPartCount]; APInt::tcSet(newParts, 0, newPartCount); if (isFiniteNonZero() || category==fcNaN) APInt::tcAssign(newParts, significandParts(), oldPartCount); freeSignificand(); significand.parts = newParts; } else if (newPartCount == 1 && oldPartCount != 1) { // Switch to built-in storage for a single part. integerPart newPart = 0; if (isFiniteNonZero() || category==fcNaN) newPart = significandParts()[0]; freeSignificand(); significand.part = newPart; } // Now that we have the right storage, switch the semantics. semantics = &toSemantics; // If this is an extension, perform the shift now that the storage is // available. if (shift > 0 && (isFiniteNonZero() || category==fcNaN)) APInt::tcShiftLeft(significandParts(), newPartCount, shift); if (isFiniteNonZero()) { fs = normalize(rounding_mode, lostFraction); *losesInfo = (fs != opOK); } else if (category == fcNaN) { if (semantics->nonFiniteBehavior == fltNonfiniteBehavior::NanOnly) { *losesInfo = fromSemantics.nonFiniteBehavior != fltNonfiniteBehavior::NanOnly; makeNaN(false, sign); return is_signaling ? opInvalidOp : opOK; } // If NaN is negative zero, we need to create a new NaN to avoid converting // NaN to -Inf. if (fromSemantics.nanEncoding == fltNanEncoding::NegativeZero && semantics->nanEncoding != fltNanEncoding::NegativeZero) makeNaN(false, false); *losesInfo = lostFraction != lfExactlyZero || X86SpecialNan; // For x87 extended precision, we want to make a NaN, not a special NaN if // the input wasn't special either. if (!X86SpecialNan && semantics == &semX87DoubleExtended) APInt::tcSetBit(significandParts(), semantics->precision - 1); // Convert of sNaN creates qNaN and raises an exception (invalid op). // This also guarantees that a sNaN does not become Inf on a truncation // that loses all payload bits. if (is_signaling) { makeQuiet(); fs = opInvalidOp; } else { fs = opOK; } } else if (category == fcInfinity && semantics->nonFiniteBehavior == fltNonfiniteBehavior::NanOnly) { makeNaN(false, sign); *losesInfo = true; fs = opInexact; } else if (category == fcZero && semantics->nanEncoding == fltNanEncoding::NegativeZero) { // Negative zero loses info, but positive zero doesn't. *losesInfo = fromSemantics.nanEncoding != fltNanEncoding::NegativeZero && sign; fs = *losesInfo ? opInexact : opOK; // NaN is negative zero means -0 -> +0, which can lose information sign = false; } else { *losesInfo = false; fs = opOK; } return fs; } /* Convert a floating point number to an integer according to the rounding mode. If the rounded integer value is out of range this returns an invalid operation exception and the contents of the destination parts are unspecified. If the rounded value is in range but the floating point number is not the exact integer, the C standard doesn't require an inexact exception to be raised. IEEE 854 does require it so we do that. Note that for conversions to integer type the C standard requires round-to-zero to always be used. */ IEEEFloat::opStatus IEEEFloat::convertToSignExtendedInteger( MutableArrayRef parts, unsigned int width, bool isSigned, roundingMode rounding_mode, bool *isExact) const { lostFraction lost_fraction; const integerPart *src; unsigned int dstPartsCount, truncatedBits; *isExact = false; /* Handle the three special cases first. */ if (category == fcInfinity || category == fcNaN) return opInvalidOp; dstPartsCount = partCountForBits(width); assert(dstPartsCount <= parts.size() && "Integer too big"); if (category == fcZero) { APInt::tcSet(parts.data(), 0, dstPartsCount); // Negative zero can't be represented as an int. *isExact = !sign; return opOK; } src = significandParts(); /* Step 1: place our absolute value, with any fraction truncated, in the destination. */ if (exponent < 0) { /* Our absolute value is less than one; truncate everything. */ APInt::tcSet(parts.data(), 0, dstPartsCount); /* For exponent -1 the integer bit represents .5, look at that. For smaller exponents leftmost truncated bit is 0. */ truncatedBits = semantics->precision -1U - exponent; } else { /* We want the most significant (exponent + 1) bits; the rest are truncated. */ unsigned int bits = exponent + 1U; /* Hopelessly large in magnitude? */ if (bits > width) return opInvalidOp; if (bits < semantics->precision) { /* We truncate (semantics->precision - bits) bits. */ truncatedBits = semantics->precision - bits; APInt::tcExtract(parts.data(), dstPartsCount, src, bits, truncatedBits); } else { /* We want at least as many bits as are available. */ APInt::tcExtract(parts.data(), dstPartsCount, src, semantics->precision, 0); APInt::tcShiftLeft(parts.data(), dstPartsCount, bits - semantics->precision); truncatedBits = 0; } } /* Step 2: work out any lost fraction, and increment the absolute value if we would round away from zero. */ if (truncatedBits) { lost_fraction = lostFractionThroughTruncation(src, partCount(), truncatedBits); if (lost_fraction != lfExactlyZero && roundAwayFromZero(rounding_mode, lost_fraction, truncatedBits)) { if (APInt::tcIncrement(parts.data(), dstPartsCount)) return opInvalidOp; /* Overflow. */ } } else { lost_fraction = lfExactlyZero; } /* Step 3: check if we fit in the destination. */ unsigned int omsb = APInt::tcMSB(parts.data(), dstPartsCount) + 1; if (sign) { if (!isSigned) { /* Negative numbers cannot be represented as unsigned. */ if (omsb != 0) return opInvalidOp; } else { /* It takes omsb bits to represent the unsigned integer value. We lose a bit for the sign, but care is needed as the maximally negative integer is a special case. */ if (omsb == width && APInt::tcLSB(parts.data(), dstPartsCount) + 1 != omsb) return opInvalidOp; /* This case can happen because of rounding. */ if (omsb > width) return opInvalidOp; } APInt::tcNegate (parts.data(), dstPartsCount); } else { if (omsb >= width + !isSigned) return opInvalidOp; } if (lost_fraction == lfExactlyZero) { *isExact = true; return opOK; } else return opInexact; } /* Same as convertToSignExtendedInteger, except we provide deterministic values in case of an invalid operation exception, namely zero for NaNs and the minimal or maximal value respectively for underflow or overflow. The *isExact output tells whether the result is exact, in the sense that converting it back to the original floating point type produces the original value. This is almost equivalent to result==opOK, except for negative zeroes. */ IEEEFloat::opStatus IEEEFloat::convertToInteger(MutableArrayRef parts, unsigned int width, bool isSigned, roundingMode rounding_mode, bool *isExact) const { opStatus fs; fs = convertToSignExtendedInteger(parts, width, isSigned, rounding_mode, isExact); if (fs == opInvalidOp) { unsigned int bits, dstPartsCount; dstPartsCount = partCountForBits(width); assert(dstPartsCount <= parts.size() && "Integer too big"); if (category == fcNaN) bits = 0; else if (sign) bits = isSigned; else bits = width - isSigned; tcSetLeastSignificantBits(parts.data(), dstPartsCount, bits); if (sign && isSigned) APInt::tcShiftLeft(parts.data(), dstPartsCount, width - 1); } return fs; } /* Convert an unsigned integer SRC to a floating point number, rounding according to ROUNDING_MODE. The sign of the floating point number is not modified. */ IEEEFloat::opStatus IEEEFloat::convertFromUnsignedParts( const integerPart *src, unsigned int srcCount, roundingMode rounding_mode) { unsigned int omsb, precision, dstCount; integerPart *dst; lostFraction lost_fraction; category = fcNormal; omsb = APInt::tcMSB(src, srcCount) + 1; dst = significandParts(); dstCount = partCount(); precision = semantics->precision; /* We want the most significant PRECISION bits of SRC. There may not be that many; extract what we can. */ if (precision <= omsb) { exponent = omsb - 1; lost_fraction = lostFractionThroughTruncation(src, srcCount, omsb - precision); APInt::tcExtract(dst, dstCount, src, precision, omsb - precision); } else { exponent = precision - 1; lost_fraction = lfExactlyZero; APInt::tcExtract(dst, dstCount, src, omsb, 0); } return normalize(rounding_mode, lost_fraction); } IEEEFloat::opStatus IEEEFloat::convertFromAPInt(const APInt &Val, bool isSigned, roundingMode rounding_mode) { unsigned int partCount = Val.getNumWords(); APInt api = Val; sign = false; if (isSigned && api.isNegative()) { sign = true; api = -api; } return convertFromUnsignedParts(api.getRawData(), partCount, rounding_mode); } /* Convert a two's complement integer SRC to a floating point number, rounding according to ROUNDING_MODE. ISSIGNED is true if the integer is signed, in which case it must be sign-extended. */ IEEEFloat::opStatus IEEEFloat::convertFromSignExtendedInteger(const integerPart *src, unsigned int srcCount, bool isSigned, roundingMode rounding_mode) { opStatus status; if (isSigned && APInt::tcExtractBit(src, srcCount * integerPartWidth - 1)) { integerPart *copy; /* If we're signed and negative negate a copy. */ sign = true; copy = new integerPart[srcCount]; APInt::tcAssign(copy, src, srcCount); APInt::tcNegate(copy, srcCount); status = convertFromUnsignedParts(copy, srcCount, rounding_mode); delete [] copy; } else { sign = false; status = convertFromUnsignedParts(src, srcCount, rounding_mode); } return status; } /* FIXME: should this just take a const APInt reference? */ IEEEFloat::opStatus IEEEFloat::convertFromZeroExtendedInteger(const integerPart *parts, unsigned int width, bool isSigned, roundingMode rounding_mode) { unsigned int partCount = partCountForBits(width); APInt api = APInt(width, ArrayRef(parts, partCount)); sign = false; if (isSigned && APInt::tcExtractBit(parts, width - 1)) { sign = true; api = -api; } return convertFromUnsignedParts(api.getRawData(), partCount, rounding_mode); } Expected IEEEFloat::convertFromHexadecimalString(StringRef s, roundingMode rounding_mode) { lostFraction lost_fraction = lfExactlyZero; category = fcNormal; zeroSignificand(); exponent = 0; integerPart *significand = significandParts(); unsigned partsCount = partCount(); unsigned bitPos = partsCount * integerPartWidth; bool computedTrailingFraction = false; // Skip leading zeroes and any (hexa)decimal point. StringRef::iterator begin = s.begin(); StringRef::iterator end = s.end(); StringRef::iterator dot; auto PtrOrErr = skipLeadingZeroesAndAnyDot(begin, end, &dot); if (!PtrOrErr) return PtrOrErr.takeError(); StringRef::iterator p = *PtrOrErr; StringRef::iterator firstSignificantDigit = p; while (p != end) { integerPart hex_value; if (*p == '.') { if (dot != end) return createError("String contains multiple dots"); dot = p++; continue; } hex_value = hexDigitValue(*p); if (hex_value == UINT_MAX) break; p++; // Store the number while we have space. if (bitPos) { bitPos -= 4; hex_value <<= bitPos % integerPartWidth; significand[bitPos / integerPartWidth] |= hex_value; } else if (!computedTrailingFraction) { auto FractOrErr = trailingHexadecimalFraction(p, end, hex_value); if (!FractOrErr) return FractOrErr.takeError(); lost_fraction = *FractOrErr; computedTrailingFraction = true; } } /* Hex floats require an exponent but not a hexadecimal point. */ if (p == end) return createError("Hex strings require an exponent"); if (*p != 'p' && *p != 'P') return createError("Invalid character in significand"); if (p == begin) return createError("Significand has no digits"); if (dot != end && p - begin == 1) return createError("Significand has no digits"); /* Ignore the exponent if we are zero. */ if (p != firstSignificantDigit) { int expAdjustment; /* Implicit hexadecimal point? */ if (dot == end) dot = p; /* Calculate the exponent adjustment implicit in the number of significant digits. */ expAdjustment = static_cast(dot - firstSignificantDigit); if (expAdjustment < 0) expAdjustment++; expAdjustment = expAdjustment * 4 - 1; /* Adjust for writing the significand starting at the most significant nibble. */ expAdjustment += semantics->precision; expAdjustment -= partsCount * integerPartWidth; /* Adjust for the given exponent. */ auto ExpOrErr = totalExponent(p + 1, end, expAdjustment); if (!ExpOrErr) return ExpOrErr.takeError(); exponent = *ExpOrErr; } return normalize(rounding_mode, lost_fraction); } IEEEFloat::opStatus IEEEFloat::roundSignificandWithExponent(const integerPart *decSigParts, unsigned sigPartCount, int exp, roundingMode rounding_mode) { unsigned int parts, pow5PartCount; fltSemantics calcSemantics = { 32767, -32767, 0, 0 }; integerPart pow5Parts[maxPowerOfFiveParts]; bool isNearest; isNearest = (rounding_mode == rmNearestTiesToEven || rounding_mode == rmNearestTiesToAway); parts = partCountForBits(semantics->precision + 11); /* Calculate pow(5, abs(exp)). */ pow5PartCount = powerOf5(pow5Parts, exp >= 0 ? exp: -exp); for (;; parts *= 2) { opStatus sigStatus, powStatus; unsigned int excessPrecision, truncatedBits; calcSemantics.precision = parts * integerPartWidth - 1; excessPrecision = calcSemantics.precision - semantics->precision; truncatedBits = excessPrecision; IEEEFloat decSig(calcSemantics, uninitialized); decSig.makeZero(sign); IEEEFloat pow5(calcSemantics); sigStatus = decSig.convertFromUnsignedParts(decSigParts, sigPartCount, rmNearestTiesToEven); powStatus = pow5.convertFromUnsignedParts(pow5Parts, pow5PartCount, rmNearestTiesToEven); /* Add exp, as 10^n = 5^n * 2^n. */ decSig.exponent += exp; lostFraction calcLostFraction; integerPart HUerr, HUdistance; unsigned int powHUerr; if (exp >= 0) { /* multiplySignificand leaves the precision-th bit set to 1. */ calcLostFraction = decSig.multiplySignificand(pow5); powHUerr = powStatus != opOK; } else { calcLostFraction = decSig.divideSignificand(pow5); /* Denormal numbers have less precision. */ if (decSig.exponent < semantics->minExponent) { excessPrecision += (semantics->minExponent - decSig.exponent); truncatedBits = excessPrecision; if (excessPrecision > calcSemantics.precision) excessPrecision = calcSemantics.precision; } /* Extra half-ulp lost in reciprocal of exponent. */ powHUerr = (powStatus == opOK && calcLostFraction == lfExactlyZero) ? 0:2; } /* Both multiplySignificand and divideSignificand return the result with the integer bit set. */ assert(APInt::tcExtractBit (decSig.significandParts(), calcSemantics.precision - 1) == 1); HUerr = HUerrBound(calcLostFraction != lfExactlyZero, sigStatus != opOK, powHUerr); HUdistance = 2 * ulpsFromBoundary(decSig.significandParts(), excessPrecision, isNearest); /* Are we guaranteed to round correctly if we truncate? */ if (HUdistance >= HUerr) { APInt::tcExtract(significandParts(), partCount(), decSig.significandParts(), calcSemantics.precision - excessPrecision, excessPrecision); /* Take the exponent of decSig. If we tcExtract-ed less bits above we must adjust our exponent to compensate for the implicit right shift. */ exponent = (decSig.exponent + semantics->precision - (calcSemantics.precision - excessPrecision)); calcLostFraction = lostFractionThroughTruncation(decSig.significandParts(), decSig.partCount(), truncatedBits); return normalize(rounding_mode, calcLostFraction); } } } Expected IEEEFloat::convertFromDecimalString(StringRef str, roundingMode rounding_mode) { decimalInfo D; opStatus fs; /* Scan the text. */ StringRef::iterator p = str.begin(); if (Error Err = interpretDecimal(p, str.end(), &D)) return std::move(Err); /* Handle the quick cases. First the case of no significant digits, i.e. zero, and then exponents that are obviously too large or too small. Writing L for log 10 / log 2, a number d.ddddd*10^exp definitely overflows if (exp - 1) * L >= maxExponent and definitely underflows to zero where (exp + 1) * L <= minExponent - precision With integer arithmetic the tightest bounds for L are 93/28 < L < 196/59 [ numerator <= 256 ] 42039/12655 < L < 28738/8651 [ numerator <= 65536 ] */ // Test if we have a zero number allowing for strings with no null terminators // and zero decimals with non-zero exponents. // // We computed firstSigDigit by ignoring all zeros and dots. Thus if // D->firstSigDigit equals str.end(), every digit must be a zero and there can // be at most one dot. On the other hand, if we have a zero with a non-zero // exponent, then we know that D.firstSigDigit will be non-numeric. if (D.firstSigDigit == str.end() || decDigitValue(*D.firstSigDigit) >= 10U) { category = fcZero; fs = opOK; if (semantics->nanEncoding == fltNanEncoding::NegativeZero) sign = false; /* Check whether the normalized exponent is high enough to overflow max during the log-rebasing in the max-exponent check below. */ } else if (D.normalizedExponent - 1 > INT_MAX / 42039) { fs = handleOverflow(rounding_mode); /* If it wasn't, then it also wasn't high enough to overflow max during the log-rebasing in the min-exponent check. Check that it won't overflow min in either check, then perform the min-exponent check. */ } else if (D.normalizedExponent - 1 < INT_MIN / 42039 || (D.normalizedExponent + 1) * 28738 <= 8651 * (semantics->minExponent - (int) semantics->precision)) { /* Underflow to zero and round. */ category = fcNormal; zeroSignificand(); fs = normalize(rounding_mode, lfLessThanHalf); /* We can finally safely perform the max-exponent check. */ } else if ((D.normalizedExponent - 1) * 42039 >= 12655 * semantics->maxExponent) { /* Overflow and round. */ fs = handleOverflow(rounding_mode); } else { integerPart *decSignificand; unsigned int partCount; /* A tight upper bound on number of bits required to hold an N-digit decimal integer is N * 196 / 59. Allocate enough space to hold the full significand, and an extra part required by tcMultiplyPart. */ partCount = static_cast(D.lastSigDigit - D.firstSigDigit) + 1; partCount = partCountForBits(1 + 196 * partCount / 59); decSignificand = new integerPart[partCount + 1]; partCount = 0; /* Convert to binary efficiently - we do almost all multiplication in an integerPart. When this would overflow do we do a single bignum multiplication, and then revert again to multiplication in an integerPart. */ do { integerPart decValue, val, multiplier; val = 0; multiplier = 1; do { if (*p == '.') { p++; if (p == str.end()) { break; } } decValue = decDigitValue(*p++); if (decValue >= 10U) { delete[] decSignificand; return createError("Invalid character in significand"); } multiplier *= 10; val = val * 10 + decValue; /* The maximum number that can be multiplied by ten with any digit added without overflowing an integerPart. */ } while (p <= D.lastSigDigit && multiplier <= (~ (integerPart) 0 - 9) / 10); /* Multiply out the current part. */ APInt::tcMultiplyPart(decSignificand, decSignificand, multiplier, val, partCount, partCount + 1, false); /* If we used another part (likely but not guaranteed), increase the count. */ if (decSignificand[partCount]) partCount++; } while (p <= D.lastSigDigit); category = fcNormal; fs = roundSignificandWithExponent(decSignificand, partCount, D.exponent, rounding_mode); delete [] decSignificand; } return fs; } bool IEEEFloat::convertFromStringSpecials(StringRef str) { const size_t MIN_NAME_SIZE = 3; if (str.size() < MIN_NAME_SIZE) return false; if (str == "inf" || str == "INFINITY" || str == "+Inf") { makeInf(false); return true; } bool IsNegative = str.front() == '-'; if (IsNegative) { str = str.drop_front(); if (str.size() < MIN_NAME_SIZE) return false; if (str == "inf" || str == "INFINITY" || str == "Inf") { makeInf(true); return true; } } // If we have a 's' (or 'S') prefix, then this is a Signaling NaN. bool IsSignaling = str.front() == 's' || str.front() == 'S'; if (IsSignaling) { str = str.drop_front(); if (str.size() < MIN_NAME_SIZE) return false; } if (str.starts_with("nan") || str.starts_with("NaN")) { str = str.drop_front(3); // A NaN without payload. if (str.empty()) { makeNaN(IsSignaling, IsNegative); return true; } // Allow the payload to be inside parentheses. if (str.front() == '(') { // Parentheses should be balanced (and not empty). if (str.size() <= 2 || str.back() != ')') return false; str = str.slice(1, str.size() - 1); } // Determine the payload number's radix. unsigned Radix = 10; if (str[0] == '0') { if (str.size() > 1 && tolower(str[1]) == 'x') { str = str.drop_front(2); Radix = 16; } else Radix = 8; } // Parse the payload and make the NaN. APInt Payload; if (!str.getAsInteger(Radix, Payload)) { makeNaN(IsSignaling, IsNegative, &Payload); return true; } } return false; } Expected IEEEFloat::convertFromString(StringRef str, roundingMode rounding_mode) { if (str.empty()) return createError("Invalid string length"); // Handle special cases. if (convertFromStringSpecials(str)) return opOK; /* Handle a leading minus sign. */ StringRef::iterator p = str.begin(); size_t slen = str.size(); sign = *p == '-' ? 1 : 0; if (*p == '-' || *p == '+') { p++; slen--; if (!slen) return createError("String has no digits"); } if (slen >= 2 && p[0] == '0' && (p[1] == 'x' || p[1] == 'X')) { if (slen == 2) return createError("Invalid string"); return convertFromHexadecimalString(StringRef(p + 2, slen - 2), rounding_mode); } return convertFromDecimalString(StringRef(p, slen), rounding_mode); } /* Write out a hexadecimal representation of the floating point value to DST, which must be of sufficient size, in the C99 form [-]0xh.hhhhp[+-]d. Return the number of characters written, excluding the terminating NUL. If UPPERCASE, the output is in upper case, otherwise in lower case. HEXDIGITS digits appear altogether, rounding the value if necessary. If HEXDIGITS is 0, the minimal precision to display the number precisely is used instead. If nothing would appear after the decimal point it is suppressed. The decimal exponent is always printed and has at least one digit. Zero values display an exponent of zero. Infinities and NaNs appear as "infinity" or "nan" respectively. The above rules are as specified by C99. There is ambiguity about what the leading hexadecimal digit should be. This implementation uses whatever is necessary so that the exponent is displayed as stored. This implies the exponent will fall within the IEEE format range, and the leading hexadecimal digit will be 0 (for denormals), 1 (normal numbers) or 2 (normal numbers rounded-away-from-zero with any other digits zero). */ unsigned int IEEEFloat::convertToHexString(char *dst, unsigned int hexDigits, bool upperCase, roundingMode rounding_mode) const { char *p; p = dst; if (sign) *dst++ = '-'; switch (category) { case fcInfinity: memcpy (dst, upperCase ? infinityU: infinityL, sizeof infinityU - 1); dst += sizeof infinityL - 1; break; case fcNaN: memcpy (dst, upperCase ? NaNU: NaNL, sizeof NaNU - 1); dst += sizeof NaNU - 1; break; case fcZero: *dst++ = '0'; *dst++ = upperCase ? 'X': 'x'; *dst++ = '0'; if (hexDigits > 1) { *dst++ = '.'; memset (dst, '0', hexDigits - 1); dst += hexDigits - 1; } *dst++ = upperCase ? 'P': 'p'; *dst++ = '0'; break; case fcNormal: dst = convertNormalToHexString (dst, hexDigits, upperCase, rounding_mode); break; } *dst = 0; return static_cast(dst - p); } /* Does the hard work of outputting the correctly rounded hexadecimal form of a normal floating point number with the specified number of hexadecimal digits. If HEXDIGITS is zero the minimum number of digits necessary to print the value precisely is output. */ char *IEEEFloat::convertNormalToHexString(char *dst, unsigned int hexDigits, bool upperCase, roundingMode rounding_mode) const { unsigned int count, valueBits, shift, partsCount, outputDigits; const char *hexDigitChars; const integerPart *significand; char *p; bool roundUp; *dst++ = '0'; *dst++ = upperCase ? 'X': 'x'; roundUp = false; hexDigitChars = upperCase ? hexDigitsUpper: hexDigitsLower; significand = significandParts(); partsCount = partCount(); /* +3 because the first digit only uses the single integer bit, so we have 3 virtual zero most-significant-bits. */ valueBits = semantics->precision + 3; shift = integerPartWidth - valueBits % integerPartWidth; /* The natural number of digits required ignoring trailing insignificant zeroes. */ outputDigits = (valueBits - significandLSB () + 3) / 4; /* hexDigits of zero means use the required number for the precision. Otherwise, see if we are truncating. If we are, find out if we need to round away from zero. */ if (hexDigits) { if (hexDigits < outputDigits) { /* We are dropping non-zero bits, so need to check how to round. "bits" is the number of dropped bits. */ unsigned int bits; lostFraction fraction; bits = valueBits - hexDigits * 4; fraction = lostFractionThroughTruncation (significand, partsCount, bits); roundUp = roundAwayFromZero(rounding_mode, fraction, bits); } outputDigits = hexDigits; } /* Write the digits consecutively, and start writing in the location of the hexadecimal point. We move the most significant digit left and add the hexadecimal point later. */ p = ++dst; count = (valueBits + integerPartWidth - 1) / integerPartWidth; while (outputDigits && count) { integerPart part; /* Put the most significant integerPartWidth bits in "part". */ if (--count == partsCount) part = 0; /* An imaginary higher zero part. */ else part = significand[count] << shift; if (count && shift) part |= significand[count - 1] >> (integerPartWidth - shift); /* Convert as much of "part" to hexdigits as we can. */ unsigned int curDigits = integerPartWidth / 4; if (curDigits > outputDigits) curDigits = outputDigits; dst += partAsHex (dst, part, curDigits, hexDigitChars); outputDigits -= curDigits; } if (roundUp) { char *q = dst; /* Note that hexDigitChars has a trailing '0'. */ do { q--; *q = hexDigitChars[hexDigitValue (*q) + 1]; } while (*q == '0'); assert(q >= p); } else { /* Add trailing zeroes. */ memset (dst, '0', outputDigits); dst += outputDigits; } /* Move the most significant digit to before the point, and if there is something after the decimal point add it. This must come after rounding above. */ p[-1] = p[0]; if (dst -1 == p) dst--; else p[0] = '.'; /* Finally output the exponent. */ *dst++ = upperCase ? 'P': 'p'; return writeSignedDecimal (dst, exponent); } hash_code hash_value(const IEEEFloat &Arg) { if (!Arg.isFiniteNonZero()) return hash_combine((uint8_t)Arg.category, // NaN has no sign, fix it at zero. Arg.isNaN() ? (uint8_t)0 : (uint8_t)Arg.sign, Arg.semantics->precision); // Normal floats need their exponent and significand hashed. return hash_combine((uint8_t)Arg.category, (uint8_t)Arg.sign, Arg.semantics->precision, Arg.exponent, hash_combine_range( Arg.significandParts(), Arg.significandParts() + Arg.partCount())); } // Conversion from APFloat to/from host float/double. It may eventually be // possible to eliminate these and have everybody deal with APFloats, but that // will take a while. This approach will not easily extend to long double. // Current implementation requires integerPartWidth==64, which is correct at // the moment but could be made more general. // Denormals have exponent minExponent in APFloat, but minExponent-1 in // the actual IEEE respresentations. We compensate for that here. APInt IEEEFloat::convertF80LongDoubleAPFloatToAPInt() const { assert(semantics == (const llvm::fltSemantics*)&semX87DoubleExtended); assert(partCount()==2); uint64_t myexponent, mysignificand; if (isFiniteNonZero()) { myexponent = exponent+16383; //bias mysignificand = significandParts()[0]; if (myexponent==1 && !(mysignificand & 0x8000000000000000ULL)) myexponent = 0; // denormal } else if (category==fcZero) { myexponent = 0; mysignificand = 0; } else if (category==fcInfinity) { myexponent = 0x7fff; mysignificand = 0x8000000000000000ULL; } else { assert(category == fcNaN && "Unknown category"); myexponent = 0x7fff; mysignificand = significandParts()[0]; } uint64_t words[2]; words[0] = mysignificand; words[1] = ((uint64_t)(sign & 1) << 15) | (myexponent & 0x7fffLL); return APInt(80, words); } APInt IEEEFloat::convertPPCDoubleDoubleAPFloatToAPInt() const { assert(semantics == (const llvm::fltSemantics *)&semPPCDoubleDoubleLegacy); assert(partCount()==2); uint64_t words[2]; opStatus fs; bool losesInfo; // Convert number to double. To avoid spurious underflows, we re- // normalize against the "double" minExponent first, and only *then* // truncate the mantissa. The result of that second conversion // may be inexact, but should never underflow. // Declare fltSemantics before APFloat that uses it (and // saves pointer to it) to ensure correct destruction order. fltSemantics extendedSemantics = *semantics; extendedSemantics.minExponent = semIEEEdouble.minExponent; IEEEFloat extended(*this); fs = extended.convert(extendedSemantics, rmNearestTiesToEven, &losesInfo); assert(fs == opOK && !losesInfo); (void)fs; IEEEFloat u(extended); fs = u.convert(semIEEEdouble, rmNearestTiesToEven, &losesInfo); assert(fs == opOK || fs == opInexact); (void)fs; words[0] = *u.convertDoubleAPFloatToAPInt().getRawData(); // If conversion was exact or resulted in a special case, we're done; // just set the second double to zero. Otherwise, re-convert back to // the extended format and compute the difference. This now should // convert exactly to double. if (u.isFiniteNonZero() && losesInfo) { fs = u.convert(extendedSemantics, rmNearestTiesToEven, &losesInfo); assert(fs == opOK && !losesInfo); (void)fs; IEEEFloat v(extended); v.subtract(u, rmNearestTiesToEven); fs = v.convert(semIEEEdouble, rmNearestTiesToEven, &losesInfo); assert(fs == opOK && !losesInfo); (void)fs; words[1] = *v.convertDoubleAPFloatToAPInt().getRawData(); } else { words[1] = 0; } return APInt(128, words); } template APInt IEEEFloat::convertIEEEFloatToAPInt() const { assert(semantics == &S); constexpr int bias = -(S.minExponent - 1); constexpr unsigned int trailing_significand_bits = S.precision - 1; constexpr int integer_bit_part = trailing_significand_bits / integerPartWidth; constexpr integerPart integer_bit = integerPart{1} << (trailing_significand_bits % integerPartWidth); constexpr uint64_t significand_mask = integer_bit - 1; constexpr unsigned int exponent_bits = S.sizeInBits - 1 - trailing_significand_bits; static_assert(exponent_bits < 64); constexpr uint64_t exponent_mask = (uint64_t{1} << exponent_bits) - 1; uint64_t myexponent; std::array mysignificand; if (isFiniteNonZero()) { myexponent = exponent + bias; std::copy_n(significandParts(), mysignificand.size(), mysignificand.begin()); if (myexponent == 1 && !(significandParts()[integer_bit_part] & integer_bit)) myexponent = 0; // denormal } else if (category == fcZero) { myexponent = ::exponentZero(S) + bias; mysignificand.fill(0); } else if (category == fcInfinity) { if (S.nonFiniteBehavior == fltNonfiniteBehavior::NanOnly || S.nonFiniteBehavior == fltNonfiniteBehavior::FiniteOnly) llvm_unreachable("semantics don't support inf!"); myexponent = ::exponentInf(S) + bias; mysignificand.fill(0); } else { assert(category == fcNaN && "Unknown category!"); if (S.nonFiniteBehavior == fltNonfiniteBehavior::FiniteOnly) llvm_unreachable("semantics don't support NaN!"); myexponent = ::exponentNaN(S) + bias; std::copy_n(significandParts(), mysignificand.size(), mysignificand.begin()); } std::array words; auto words_iter = std::copy_n(mysignificand.begin(), mysignificand.size(), words.begin()); if constexpr (significand_mask != 0) { // Clear the integer bit. words[mysignificand.size() - 1] &= significand_mask; } std::fill(words_iter, words.end(), uint64_t{0}); constexpr size_t last_word = words.size() - 1; uint64_t shifted_sign = static_cast(sign & 1) << ((S.sizeInBits - 1) % 64); words[last_word] |= shifted_sign; uint64_t shifted_exponent = (myexponent & exponent_mask) << (trailing_significand_bits % 64); words[last_word] |= shifted_exponent; if constexpr (last_word == 0) { return APInt(S.sizeInBits, words[0]); } return APInt(S.sizeInBits, words); } APInt IEEEFloat::convertQuadrupleAPFloatToAPInt() const { assert(partCount() == 2); return convertIEEEFloatToAPInt(); } APInt IEEEFloat::convertDoubleAPFloatToAPInt() const { assert(partCount()==1); return convertIEEEFloatToAPInt(); } APInt IEEEFloat::convertFloatAPFloatToAPInt() const { assert(partCount()==1); return convertIEEEFloatToAPInt(); } APInt IEEEFloat::convertBFloatAPFloatToAPInt() const { assert(partCount() == 1); return convertIEEEFloatToAPInt(); } APInt IEEEFloat::convertHalfAPFloatToAPInt() const { assert(partCount()==1); return convertIEEEFloatToAPInt(); } APInt IEEEFloat::convertFloat8E5M2APFloatToAPInt() const { assert(partCount() == 1); return convertIEEEFloatToAPInt(); } APInt IEEEFloat::convertFloat8E5M2FNUZAPFloatToAPInt() const { assert(partCount() == 1); return convertIEEEFloatToAPInt(); } APInt IEEEFloat::convertFloat8E4M3APFloatToAPInt() const { assert(partCount() == 1); return convertIEEEFloatToAPInt(); } APInt IEEEFloat::convertFloat8E4M3FNAPFloatToAPInt() const { assert(partCount() == 1); return convertIEEEFloatToAPInt(); } APInt IEEEFloat::convertFloat8E4M3FNUZAPFloatToAPInt() const { assert(partCount() == 1); return convertIEEEFloatToAPInt(); } APInt IEEEFloat::convertFloat8E4M3B11FNUZAPFloatToAPInt() const { assert(partCount() == 1); return convertIEEEFloatToAPInt(); } APInt IEEEFloat::convertFloatTF32APFloatToAPInt() const { assert(partCount() == 1); return convertIEEEFloatToAPInt(); } APInt IEEEFloat::convertFloat6E3M2FNAPFloatToAPInt() const { assert(partCount() == 1); return convertIEEEFloatToAPInt(); } APInt IEEEFloat::convertFloat6E2M3FNAPFloatToAPInt() const { assert(partCount() == 1); return convertIEEEFloatToAPInt(); } APInt IEEEFloat::convertFloat4E2M1FNAPFloatToAPInt() const { assert(partCount() == 1); return convertIEEEFloatToAPInt(); } // This function creates an APInt that is just a bit map of the floating // point constant as it would appear in memory. It is not a conversion, // and treating the result as a normal integer is unlikely to be useful. APInt IEEEFloat::bitcastToAPInt() const { if (semantics == (const llvm::fltSemantics*)&semIEEEhalf) return convertHalfAPFloatToAPInt(); if (semantics == (const llvm::fltSemantics *)&semBFloat) return convertBFloatAPFloatToAPInt(); if (semantics == (const llvm::fltSemantics*)&semIEEEsingle) return convertFloatAPFloatToAPInt(); if (semantics == (const llvm::fltSemantics*)&semIEEEdouble) return convertDoubleAPFloatToAPInt(); if (semantics == (const llvm::fltSemantics*)&semIEEEquad) return convertQuadrupleAPFloatToAPInt(); if (semantics == (const llvm::fltSemantics *)&semPPCDoubleDoubleLegacy) return convertPPCDoubleDoubleAPFloatToAPInt(); if (semantics == (const llvm::fltSemantics *)&semFloat8E5M2) return convertFloat8E5M2APFloatToAPInt(); if (semantics == (const llvm::fltSemantics *)&semFloat8E5M2FNUZ) return convertFloat8E5M2FNUZAPFloatToAPInt(); if (semantics == (const llvm::fltSemantics *)&semFloat8E4M3) return convertFloat8E4M3APFloatToAPInt(); if (semantics == (const llvm::fltSemantics *)&semFloat8E4M3FN) return convertFloat8E4M3FNAPFloatToAPInt(); if (semantics == (const llvm::fltSemantics *)&semFloat8E4M3FNUZ) return convertFloat8E4M3FNUZAPFloatToAPInt(); if (semantics == (const llvm::fltSemantics *)&semFloat8E4M3B11FNUZ) return convertFloat8E4M3B11FNUZAPFloatToAPInt(); if (semantics == (const llvm::fltSemantics *)&semFloatTF32) return convertFloatTF32APFloatToAPInt(); if (semantics == (const llvm::fltSemantics *)&semFloat6E3M2FN) return convertFloat6E3M2FNAPFloatToAPInt(); if (semantics == (const llvm::fltSemantics *)&semFloat6E2M3FN) return convertFloat6E2M3FNAPFloatToAPInt(); if (semantics == (const llvm::fltSemantics *)&semFloat4E2M1FN) return convertFloat4E2M1FNAPFloatToAPInt(); assert(semantics == (const llvm::fltSemantics*)&semX87DoubleExtended && "unknown format!"); return convertF80LongDoubleAPFloatToAPInt(); } float IEEEFloat::convertToFloat() const { assert(semantics == (const llvm::fltSemantics*)&semIEEEsingle && "Float semantics are not IEEEsingle"); APInt api = bitcastToAPInt(); return api.bitsToFloat(); } double IEEEFloat::convertToDouble() const { assert(semantics == (const llvm::fltSemantics*)&semIEEEdouble && "Float semantics are not IEEEdouble"); APInt api = bitcastToAPInt(); return api.bitsToDouble(); } #ifdef HAS_IEE754_FLOAT128 float128 IEEEFloat::convertToQuad() const { assert(semantics == (const llvm::fltSemantics *)&semIEEEquad && "Float semantics are not IEEEquads"); APInt api = bitcastToAPInt(); return api.bitsToQuad(); } #endif /// Integer bit is explicit in this format. Intel hardware (387 and later) /// does not support these bit patterns: /// exponent = all 1's, integer bit 0, significand 0 ("pseudoinfinity") /// exponent = all 1's, integer bit 0, significand nonzero ("pseudoNaN") /// exponent!=0 nor all 1's, integer bit 0 ("unnormal") /// exponent = 0, integer bit 1 ("pseudodenormal") /// At the moment, the first three are treated as NaNs, the last one as Normal. void IEEEFloat::initFromF80LongDoubleAPInt(const APInt &api) { uint64_t i1 = api.getRawData()[0]; uint64_t i2 = api.getRawData()[1]; uint64_t myexponent = (i2 & 0x7fff); uint64_t mysignificand = i1; uint8_t myintegerbit = mysignificand >> 63; initialize(&semX87DoubleExtended); assert(partCount()==2); sign = static_cast(i2>>15); if (myexponent == 0 && mysignificand == 0) { makeZero(sign); } else if (myexponent==0x7fff && mysignificand==0x8000000000000000ULL) { makeInf(sign); } else if ((myexponent == 0x7fff && mysignificand != 0x8000000000000000ULL) || (myexponent != 0x7fff && myexponent != 0 && myintegerbit == 0)) { category = fcNaN; exponent = exponentNaN(); significandParts()[0] = mysignificand; significandParts()[1] = 0; } else { category = fcNormal; exponent = myexponent - 16383; significandParts()[0] = mysignificand; significandParts()[1] = 0; if (myexponent==0) // denormal exponent = -16382; } } void IEEEFloat::initFromPPCDoubleDoubleAPInt(const APInt &api) { uint64_t i1 = api.getRawData()[0]; uint64_t i2 = api.getRawData()[1]; opStatus fs; bool losesInfo; // Get the first double and convert to our format. initFromDoubleAPInt(APInt(64, i1)); fs = convert(semPPCDoubleDoubleLegacy, rmNearestTiesToEven, &losesInfo); assert(fs == opOK && !losesInfo); (void)fs; // Unless we have a special case, add in second double. if (isFiniteNonZero()) { IEEEFloat v(semIEEEdouble, APInt(64, i2)); fs = v.convert(semPPCDoubleDoubleLegacy, rmNearestTiesToEven, &losesInfo); assert(fs == opOK && !losesInfo); (void)fs; add(v, rmNearestTiesToEven); } } template void IEEEFloat::initFromIEEEAPInt(const APInt &api) { assert(api.getBitWidth() == S.sizeInBits); constexpr integerPart integer_bit = integerPart{1} << ((S.precision - 1) % integerPartWidth); constexpr uint64_t significand_mask = integer_bit - 1; constexpr unsigned int trailing_significand_bits = S.precision - 1; constexpr unsigned int stored_significand_parts = partCountForBits(trailing_significand_bits); constexpr unsigned int exponent_bits = S.sizeInBits - 1 - trailing_significand_bits; static_assert(exponent_bits < 64); constexpr uint64_t exponent_mask = (uint64_t{1} << exponent_bits) - 1; constexpr int bias = -(S.minExponent - 1); // Copy the bits of the significand. We need to clear out the exponent and // sign bit in the last word. std::array mysignificand; std::copy_n(api.getRawData(), mysignificand.size(), mysignificand.begin()); if constexpr (significand_mask != 0) { mysignificand[mysignificand.size() - 1] &= significand_mask; } // We assume the last word holds the sign bit, the exponent, and potentially // some of the trailing significand field. uint64_t last_word = api.getRawData()[api.getNumWords() - 1]; uint64_t myexponent = (last_word >> (trailing_significand_bits % 64)) & exponent_mask; initialize(&S); assert(partCount() == mysignificand.size()); sign = static_cast(last_word >> ((S.sizeInBits - 1) % 64)); bool all_zero_significand = llvm::all_of(mysignificand, [](integerPart bits) { return bits == 0; }); bool is_zero = myexponent == 0 && all_zero_significand; if constexpr (S.nonFiniteBehavior == fltNonfiniteBehavior::IEEE754) { if (myexponent - bias == ::exponentInf(S) && all_zero_significand) { makeInf(sign); return; } } bool is_nan = false; if constexpr (S.nanEncoding == fltNanEncoding::IEEE) { is_nan = myexponent - bias == ::exponentNaN(S) && !all_zero_significand; } else if constexpr (S.nanEncoding == fltNanEncoding::AllOnes) { bool all_ones_significand = std::all_of(mysignificand.begin(), mysignificand.end() - 1, [](integerPart bits) { return bits == ~integerPart{0}; }) && (!significand_mask || mysignificand[mysignificand.size() - 1] == significand_mask); is_nan = myexponent - bias == ::exponentNaN(S) && all_ones_significand; } else if constexpr (S.nanEncoding == fltNanEncoding::NegativeZero) { is_nan = is_zero && sign; } if (is_nan) { category = fcNaN; exponent = ::exponentNaN(S); std::copy_n(mysignificand.begin(), mysignificand.size(), significandParts()); return; } if (is_zero) { makeZero(sign); return; } category = fcNormal; exponent = myexponent - bias; std::copy_n(mysignificand.begin(), mysignificand.size(), significandParts()); if (myexponent == 0) // denormal exponent = S.minExponent; else significandParts()[mysignificand.size()-1] |= integer_bit; // integer bit } void IEEEFloat::initFromQuadrupleAPInt(const APInt &api) { initFromIEEEAPInt(api); } void IEEEFloat::initFromDoubleAPInt(const APInt &api) { initFromIEEEAPInt(api); } void IEEEFloat::initFromFloatAPInt(const APInt &api) { initFromIEEEAPInt(api); } void IEEEFloat::initFromBFloatAPInt(const APInt &api) { initFromIEEEAPInt(api); } void IEEEFloat::initFromHalfAPInt(const APInt &api) { initFromIEEEAPInt(api); } void IEEEFloat::initFromFloat8E5M2APInt(const APInt &api) { initFromIEEEAPInt(api); } void IEEEFloat::initFromFloat8E5M2FNUZAPInt(const APInt &api) { initFromIEEEAPInt(api); } void IEEEFloat::initFromFloat8E4M3APInt(const APInt &api) { initFromIEEEAPInt(api); } void IEEEFloat::initFromFloat8E4M3FNAPInt(const APInt &api) { initFromIEEEAPInt(api); } void IEEEFloat::initFromFloat8E4M3FNUZAPInt(const APInt &api) { initFromIEEEAPInt(api); } void IEEEFloat::initFromFloat8E4M3B11FNUZAPInt(const APInt &api) { initFromIEEEAPInt(api); } void IEEEFloat::initFromFloatTF32APInt(const APInt &api) { initFromIEEEAPInt(api); } void IEEEFloat::initFromFloat6E3M2FNAPInt(const APInt &api) { initFromIEEEAPInt(api); } void IEEEFloat::initFromFloat6E2M3FNAPInt(const APInt &api) { initFromIEEEAPInt(api); } void IEEEFloat::initFromFloat4E2M1FNAPInt(const APInt &api) { initFromIEEEAPInt(api); } /// Treat api as containing the bits of a floating point number. void IEEEFloat::initFromAPInt(const fltSemantics *Sem, const APInt &api) { assert(api.getBitWidth() == Sem->sizeInBits); if (Sem == &semIEEEhalf) return initFromHalfAPInt(api); if (Sem == &semBFloat) return initFromBFloatAPInt(api); if (Sem == &semIEEEsingle) return initFromFloatAPInt(api); if (Sem == &semIEEEdouble) return initFromDoubleAPInt(api); if (Sem == &semX87DoubleExtended) return initFromF80LongDoubleAPInt(api); if (Sem == &semIEEEquad) return initFromQuadrupleAPInt(api); if (Sem == &semPPCDoubleDoubleLegacy) return initFromPPCDoubleDoubleAPInt(api); if (Sem == &semFloat8E5M2) return initFromFloat8E5M2APInt(api); if (Sem == &semFloat8E5M2FNUZ) return initFromFloat8E5M2FNUZAPInt(api); if (Sem == &semFloat8E4M3) return initFromFloat8E4M3APInt(api); if (Sem == &semFloat8E4M3FN) return initFromFloat8E4M3FNAPInt(api); if (Sem == &semFloat8E4M3FNUZ) return initFromFloat8E4M3FNUZAPInt(api); if (Sem == &semFloat8E4M3B11FNUZ) return initFromFloat8E4M3B11FNUZAPInt(api); if (Sem == &semFloatTF32) return initFromFloatTF32APInt(api); if (Sem == &semFloat6E3M2FN) return initFromFloat6E3M2FNAPInt(api); if (Sem == &semFloat6E2M3FN) return initFromFloat6E2M3FNAPInt(api); if (Sem == &semFloat4E2M1FN) return initFromFloat4E2M1FNAPInt(api); llvm_unreachable(nullptr); } /// Make this number the largest magnitude normal number in the given /// semantics. void IEEEFloat::makeLargest(bool Negative) { // We want (in interchange format): // sign = {Negative} // exponent = 1..10 // significand = 1..1 category = fcNormal; sign = Negative; exponent = semantics->maxExponent; // Use memset to set all but the highest integerPart to all ones. integerPart *significand = significandParts(); unsigned PartCount = partCount(); memset(significand, 0xFF, sizeof(integerPart)*(PartCount - 1)); // Set the high integerPart especially setting all unused top bits for // internal consistency. const unsigned NumUnusedHighBits = PartCount*integerPartWidth - semantics->precision; significand[PartCount - 1] = (NumUnusedHighBits < integerPartWidth) ? (~integerPart(0) >> NumUnusedHighBits) : 0; if (semantics->nonFiniteBehavior == fltNonfiniteBehavior::NanOnly && semantics->nanEncoding == fltNanEncoding::AllOnes) significand[0] &= ~integerPart(1); } /// Make this number the smallest magnitude denormal number in the given /// semantics. void IEEEFloat::makeSmallest(bool Negative) { // We want (in interchange format): // sign = {Negative} // exponent = 0..0 // significand = 0..01 category = fcNormal; sign = Negative; exponent = semantics->minExponent; APInt::tcSet(significandParts(), 1, partCount()); } void IEEEFloat::makeSmallestNormalized(bool Negative) { // We want (in interchange format): // sign = {Negative} // exponent = 0..0 // significand = 10..0 category = fcNormal; zeroSignificand(); sign = Negative; exponent = semantics->minExponent; APInt::tcSetBit(significandParts(), semantics->precision - 1); } IEEEFloat::IEEEFloat(const fltSemantics &Sem, const APInt &API) { initFromAPInt(&Sem, API); } IEEEFloat::IEEEFloat(float f) { initFromAPInt(&semIEEEsingle, APInt::floatToBits(f)); } IEEEFloat::IEEEFloat(double d) { initFromAPInt(&semIEEEdouble, APInt::doubleToBits(d)); } namespace { void append(SmallVectorImpl &Buffer, StringRef Str) { Buffer.append(Str.begin(), Str.end()); } /// Removes data from the given significand until it is no more /// precise than is required for the desired precision. void AdjustToPrecision(APInt &significand, int &exp, unsigned FormatPrecision) { unsigned bits = significand.getActiveBits(); // 196/59 is a very slight overestimate of lg_2(10). unsigned bitsRequired = (FormatPrecision * 196 + 58) / 59; if (bits <= bitsRequired) return; unsigned tensRemovable = (bits - bitsRequired) * 59 / 196; if (!tensRemovable) return; exp += tensRemovable; APInt divisor(significand.getBitWidth(), 1); APInt powten(significand.getBitWidth(), 10); while (true) { if (tensRemovable & 1) divisor *= powten; tensRemovable >>= 1; if (!tensRemovable) break; powten *= powten; } significand = significand.udiv(divisor); // Truncate the significand down to its active bit count. significand = significand.trunc(significand.getActiveBits()); } void AdjustToPrecision(SmallVectorImpl &buffer, int &exp, unsigned FormatPrecision) { unsigned N = buffer.size(); if (N <= FormatPrecision) return; // The most significant figures are the last ones in the buffer. unsigned FirstSignificant = N - FormatPrecision; // Round. // FIXME: this probably shouldn't use 'round half up'. // Rounding down is just a truncation, except we also want to drop // trailing zeros from the new result. if (buffer[FirstSignificant - 1] < '5') { while (FirstSignificant < N && buffer[FirstSignificant] == '0') FirstSignificant++; exp += FirstSignificant; buffer.erase(&buffer[0], &buffer[FirstSignificant]); return; } // Rounding up requires a decimal add-with-carry. If we continue // the carry, the newly-introduced zeros will just be truncated. for (unsigned I = FirstSignificant; I != N; ++I) { if (buffer[I] == '9') { FirstSignificant++; } else { buffer[I]++; break; } } // If we carried through, we have exactly one digit of precision. if (FirstSignificant == N) { exp += FirstSignificant; buffer.clear(); buffer.push_back('1'); return; } exp += FirstSignificant; buffer.erase(&buffer[0], &buffer[FirstSignificant]); } void toStringImpl(SmallVectorImpl &Str, const bool isNeg, int exp, APInt significand, unsigned FormatPrecision, unsigned FormatMaxPadding, bool TruncateZero) { const int semanticsPrecision = significand.getBitWidth(); if (isNeg) Str.push_back('-'); // Set FormatPrecision if zero. We want to do this before we // truncate trailing zeros, as those are part of the precision. if (!FormatPrecision) { // We use enough digits so the number can be round-tripped back to an // APFloat. The formula comes from "How to Print Floating-Point Numbers // Accurately" by Steele and White. // FIXME: Using a formula based purely on the precision is conservative; // we can print fewer digits depending on the actual value being printed. // FormatPrecision = 2 + floor(significandBits / lg_2(10)) FormatPrecision = 2 + semanticsPrecision * 59 / 196; } // Ignore trailing binary zeros. int trailingZeros = significand.countr_zero(); exp += trailingZeros; significand.lshrInPlace(trailingZeros); // Change the exponent from 2^e to 10^e. if (exp == 0) { // Nothing to do. } else if (exp > 0) { // Just shift left. significand = significand.zext(semanticsPrecision + exp); significand <<= exp; exp = 0; } else { /* exp < 0 */ int texp = -exp; // We transform this using the identity: // (N)(2^-e) == (N)(5^e)(10^-e) // This means we have to multiply N (the significand) by 5^e. // To avoid overflow, we have to operate on numbers large // enough to store N * 5^e: // log2(N * 5^e) == log2(N) + e * log2(5) // <= semantics->precision + e * 137 / 59 // (log_2(5) ~ 2.321928 < 2.322034 ~ 137/59) unsigned precision = semanticsPrecision + (137 * texp + 136) / 59; // Multiply significand by 5^e. // N * 5^0101 == N * 5^(1*1) * 5^(0*2) * 5^(1*4) * 5^(0*8) significand = significand.zext(precision); APInt five_to_the_i(precision, 5); while (true) { if (texp & 1) significand *= five_to_the_i; texp >>= 1; if (!texp) break; five_to_the_i *= five_to_the_i; } } AdjustToPrecision(significand, exp, FormatPrecision); SmallVector buffer; // Fill the buffer. unsigned precision = significand.getBitWidth(); if (precision < 4) { // We need enough precision to store the value 10. precision = 4; significand = significand.zext(precision); } APInt ten(precision, 10); APInt digit(precision, 0); bool inTrail = true; while (significand != 0) { // digit <- significand % 10 // significand <- significand / 10 APInt::udivrem(significand, ten, significand, digit); unsigned d = digit.getZExtValue(); // Drop trailing zeros. if (inTrail && !d) exp++; else { buffer.push_back((char) ('0' + d)); inTrail = false; } } assert(!buffer.empty() && "no characters in buffer!"); // Drop down to FormatPrecision. // TODO: don't do more precise calculations above than are required. AdjustToPrecision(buffer, exp, FormatPrecision); unsigned NDigits = buffer.size(); // Check whether we should use scientific notation. bool FormatScientific; if (!FormatMaxPadding) FormatScientific = true; else { if (exp >= 0) { // 765e3 --> 765000 // ^^^ // But we shouldn't make the number look more precise than it is. FormatScientific = ((unsigned) exp > FormatMaxPadding || NDigits + (unsigned) exp > FormatPrecision); } else { // Power of the most significant digit. int MSD = exp + (int) (NDigits - 1); if (MSD >= 0) { // 765e-2 == 7.65 FormatScientific = false; } else { // 765e-5 == 0.00765 // ^ ^^ FormatScientific = ((unsigned) -MSD) > FormatMaxPadding; } } } // Scientific formatting is pretty straightforward. if (FormatScientific) { exp += (NDigits - 1); Str.push_back(buffer[NDigits-1]); Str.push_back('.'); if (NDigits == 1 && TruncateZero) Str.push_back('0'); else for (unsigned I = 1; I != NDigits; ++I) Str.push_back(buffer[NDigits-1-I]); // Fill with zeros up to FormatPrecision. if (!TruncateZero && FormatPrecision > NDigits - 1) Str.append(FormatPrecision - NDigits + 1, '0'); // For !TruncateZero we use lower 'e'. Str.push_back(TruncateZero ? 'E' : 'e'); Str.push_back(exp >= 0 ? '+' : '-'); if (exp < 0) exp = -exp; SmallVector expbuf; do { expbuf.push_back((char) ('0' + (exp % 10))); exp /= 10; } while (exp); // Exponent always at least two digits if we do not truncate zeros. if (!TruncateZero && expbuf.size() < 2) expbuf.push_back('0'); for (unsigned I = 0, E = expbuf.size(); I != E; ++I) Str.push_back(expbuf[E-1-I]); return; } // Non-scientific, positive exponents. if (exp >= 0) { for (unsigned I = 0; I != NDigits; ++I) Str.push_back(buffer[NDigits-1-I]); for (unsigned I = 0; I != (unsigned) exp; ++I) Str.push_back('0'); return; } // Non-scientific, negative exponents. // The number of digits to the left of the decimal point. int NWholeDigits = exp + (int) NDigits; unsigned I = 0; if (NWholeDigits > 0) { for (; I != (unsigned) NWholeDigits; ++I) Str.push_back(buffer[NDigits-I-1]); Str.push_back('.'); } else { unsigned NZeros = 1 + (unsigned) -NWholeDigits; Str.push_back('0'); Str.push_back('.'); for (unsigned Z = 1; Z != NZeros; ++Z) Str.push_back('0'); } for (; I != NDigits; ++I) Str.push_back(buffer[NDigits-I-1]); } } // namespace void IEEEFloat::toString(SmallVectorImpl &Str, unsigned FormatPrecision, unsigned FormatMaxPadding, bool TruncateZero) const { switch (category) { case fcInfinity: if (isNegative()) return append(Str, "-Inf"); else return append(Str, "+Inf"); case fcNaN: return append(Str, "NaN"); case fcZero: if (isNegative()) Str.push_back('-'); if (!FormatMaxPadding) { if (TruncateZero) append(Str, "0.0E+0"); else { append(Str, "0.0"); if (FormatPrecision > 1) Str.append(FormatPrecision - 1, '0'); append(Str, "e+00"); } } else Str.push_back('0'); return; case fcNormal: break; } // Decompose the number into an APInt and an exponent. int exp = exponent - ((int) semantics->precision - 1); APInt significand( semantics->precision, ArrayRef(significandParts(), partCountForBits(semantics->precision))); toStringImpl(Str, isNegative(), exp, significand, FormatPrecision, FormatMaxPadding, TruncateZero); } bool IEEEFloat::getExactInverse(APFloat *inv) const { // Special floats and denormals have no exact inverse. if (!isFiniteNonZero()) return false; // Check that the number is a power of two by making sure that only the // integer bit is set in the significand. if (significandLSB() != semantics->precision - 1) return false; // Get the inverse. IEEEFloat reciprocal(*semantics, 1ULL); if (reciprocal.divide(*this, rmNearestTiesToEven) != opOK) return false; // Avoid multiplication with a denormal, it is not safe on all platforms and // may be slower than a normal division. if (reciprocal.isDenormal()) return false; assert(reciprocal.isFiniteNonZero() && reciprocal.significandLSB() == reciprocal.semantics->precision - 1); if (inv) *inv = APFloat(reciprocal, *semantics); return true; } int IEEEFloat::getExactLog2Abs() const { if (!isFinite() || isZero()) return INT_MIN; const integerPart *Parts = significandParts(); const int PartCount = partCountForBits(semantics->precision); int PopCount = 0; for (int i = 0; i < PartCount; ++i) { PopCount += llvm::popcount(Parts[i]); if (PopCount > 1) return INT_MIN; } if (exponent != semantics->minExponent) return exponent; int CountrParts = 0; for (int i = 0; i < PartCount; ++i, CountrParts += APInt::APINT_BITS_PER_WORD) { if (Parts[i] != 0) { return exponent - semantics->precision + CountrParts + llvm::countr_zero(Parts[i]) + 1; } } llvm_unreachable("didn't find the set bit"); } bool IEEEFloat::isSignaling() const { if (!isNaN()) return false; if (semantics->nonFiniteBehavior == fltNonfiniteBehavior::NanOnly || semantics->nonFiniteBehavior == fltNonfiniteBehavior::FiniteOnly) return false; // IEEE-754R 2008 6.2.1: A signaling NaN bit string should be encoded with the // first bit of the trailing significand being 0. return !APInt::tcExtractBit(significandParts(), semantics->precision - 2); } /// IEEE-754R 2008 5.3.1: nextUp/nextDown. /// /// *NOTE* since nextDown(x) = -nextUp(-x), we only implement nextUp with /// appropriate sign switching before/after the computation. IEEEFloat::opStatus IEEEFloat::next(bool nextDown) { // If we are performing nextDown, swap sign so we have -x. if (nextDown) changeSign(); // Compute nextUp(x) opStatus result = opOK; // Handle each float category separately. switch (category) { case fcInfinity: // nextUp(+inf) = +inf if (!isNegative()) break; // nextUp(-inf) = -getLargest() makeLargest(true); break; case fcNaN: // IEEE-754R 2008 6.2 Par 2: nextUp(sNaN) = qNaN. Set Invalid flag. // IEEE-754R 2008 6.2: nextUp(qNaN) = qNaN. Must be identity so we do not // change the payload. if (isSignaling()) { result = opInvalidOp; // For consistency, propagate the sign of the sNaN to the qNaN. makeNaN(false, isNegative(), nullptr); } break; case fcZero: // nextUp(pm 0) = +getSmallest() makeSmallest(false); break; case fcNormal: // nextUp(-getSmallest()) = -0 if (isSmallest() && isNegative()) { APInt::tcSet(significandParts(), 0, partCount()); category = fcZero; exponent = 0; if (semantics->nanEncoding == fltNanEncoding::NegativeZero) sign = false; break; } if (isLargest() && !isNegative()) { if (semantics->nonFiniteBehavior == fltNonfiniteBehavior::NanOnly) { // nextUp(getLargest()) == NAN makeNaN(); break; } else if (semantics->nonFiniteBehavior == fltNonfiniteBehavior::FiniteOnly) { // nextUp(getLargest()) == getLargest() break; } else { // nextUp(getLargest()) == INFINITY APInt::tcSet(significandParts(), 0, partCount()); category = fcInfinity; exponent = semantics->maxExponent + 1; break; } } // nextUp(normal) == normal + inc. if (isNegative()) { // If we are negative, we need to decrement the significand. // We only cross a binade boundary that requires adjusting the exponent // if: // 1. exponent != semantics->minExponent. This implies we are not in the // smallest binade or are dealing with denormals. // 2. Our significand excluding the integral bit is all zeros. bool WillCrossBinadeBoundary = exponent != semantics->minExponent && isSignificandAllZeros(); // Decrement the significand. // // We always do this since: // 1. If we are dealing with a non-binade decrement, by definition we // just decrement the significand. // 2. If we are dealing with a normal -> normal binade decrement, since // we have an explicit integral bit the fact that all bits but the // integral bit are zero implies that subtracting one will yield a // significand with 0 integral bit and 1 in all other spots. Thus we // must just adjust the exponent and set the integral bit to 1. // 3. If we are dealing with a normal -> denormal binade decrement, // since we set the integral bit to 0 when we represent denormals, we // just decrement the significand. integerPart *Parts = significandParts(); APInt::tcDecrement(Parts, partCount()); if (WillCrossBinadeBoundary) { // Our result is a normal number. Do the following: // 1. Set the integral bit to 1. // 2. Decrement the exponent. APInt::tcSetBit(Parts, semantics->precision - 1); exponent--; } } else { // If we are positive, we need to increment the significand. // We only cross a binade boundary that requires adjusting the exponent if // the input is not a denormal and all of said input's significand bits // are set. If all of said conditions are true: clear the significand, set // the integral bit to 1, and increment the exponent. If we have a // denormal always increment since moving denormals and the numbers in the // smallest normal binade have the same exponent in our representation. bool WillCrossBinadeBoundary = !isDenormal() && isSignificandAllOnes(); if (WillCrossBinadeBoundary) { integerPart *Parts = significandParts(); APInt::tcSet(Parts, 0, partCount()); APInt::tcSetBit(Parts, semantics->precision - 1); assert(exponent != semantics->maxExponent && "We can not increment an exponent beyond the maxExponent allowed" " by the given floating point semantics."); exponent++; } else { incrementSignificand(); } } break; } // If we are performing nextDown, swap sign so we have -nextUp(-x) if (nextDown) changeSign(); return result; } APFloatBase::ExponentType IEEEFloat::exponentNaN() const { return ::exponentNaN(*semantics); } APFloatBase::ExponentType IEEEFloat::exponentInf() const { return ::exponentInf(*semantics); } APFloatBase::ExponentType IEEEFloat::exponentZero() const { return ::exponentZero(*semantics); } void IEEEFloat::makeInf(bool Negative) { if (semantics->nonFiniteBehavior == fltNonfiniteBehavior::FiniteOnly) llvm_unreachable("This floating point format does not support Inf"); if (semantics->nonFiniteBehavior == fltNonfiniteBehavior::NanOnly) { // There is no Inf, so make NaN instead. makeNaN(false, Negative); return; } category = fcInfinity; sign = Negative; exponent = exponentInf(); APInt::tcSet(significandParts(), 0, partCount()); } void IEEEFloat::makeZero(bool Negative) { category = fcZero; sign = Negative; if (semantics->nanEncoding == fltNanEncoding::NegativeZero) { // Merge negative zero to positive because 0b10000...000 is used for NaN sign = false; } exponent = exponentZero(); APInt::tcSet(significandParts(), 0, partCount()); } void IEEEFloat::makeQuiet() { assert(isNaN()); if (semantics->nonFiniteBehavior != fltNonfiniteBehavior::NanOnly) APInt::tcSetBit(significandParts(), semantics->precision - 2); } int ilogb(const IEEEFloat &Arg) { if (Arg.isNaN()) return IEEEFloat::IEK_NaN; if (Arg.isZero()) return IEEEFloat::IEK_Zero; if (Arg.isInfinity()) return IEEEFloat::IEK_Inf; if (!Arg.isDenormal()) return Arg.exponent; IEEEFloat Normalized(Arg); int SignificandBits = Arg.getSemantics().precision - 1; Normalized.exponent += SignificandBits; Normalized.normalize(IEEEFloat::rmNearestTiesToEven, lfExactlyZero); return Normalized.exponent - SignificandBits; } IEEEFloat scalbn(IEEEFloat X, int Exp, IEEEFloat::roundingMode RoundingMode) { auto MaxExp = X.getSemantics().maxExponent; auto MinExp = X.getSemantics().minExponent; // If Exp is wildly out-of-scale, simply adding it to X.exponent will // overflow; clamp it to a safe range before adding, but ensure that the range // is large enough that the clamp does not change the result. The range we // need to support is the difference between the largest possible exponent and // the normalized exponent of half the smallest denormal. int SignificandBits = X.getSemantics().precision - 1; int MaxIncrement = MaxExp - (MinExp - SignificandBits) + 1; // Clamp to one past the range ends to let normalize handle overlflow. X.exponent += std::clamp(Exp, -MaxIncrement - 1, MaxIncrement); X.normalize(RoundingMode, lfExactlyZero); if (X.isNaN()) X.makeQuiet(); return X; } IEEEFloat frexp(const IEEEFloat &Val, int &Exp, IEEEFloat::roundingMode RM) { Exp = ilogb(Val); // Quiet signalling nans. if (Exp == IEEEFloat::IEK_NaN) { IEEEFloat Quiet(Val); Quiet.makeQuiet(); return Quiet; } if (Exp == IEEEFloat::IEK_Inf) return Val; // 1 is added because frexp is defined to return a normalized fraction in // +/-[0.5, 1.0), rather than the usual +/-[1.0, 2.0). Exp = Exp == IEEEFloat::IEK_Zero ? 0 : Exp + 1; return scalbn(Val, -Exp, RM); } DoubleAPFloat::DoubleAPFloat(const fltSemantics &S) : Semantics(&S), Floats(new APFloat[2]{APFloat(semIEEEdouble), APFloat(semIEEEdouble)}) { assert(Semantics == &semPPCDoubleDouble); } DoubleAPFloat::DoubleAPFloat(const fltSemantics &S, uninitializedTag) : Semantics(&S), Floats(new APFloat[2]{APFloat(semIEEEdouble, uninitialized), APFloat(semIEEEdouble, uninitialized)}) { assert(Semantics == &semPPCDoubleDouble); } DoubleAPFloat::DoubleAPFloat(const fltSemantics &S, integerPart I) : Semantics(&S), Floats(new APFloat[2]{APFloat(semIEEEdouble, I), APFloat(semIEEEdouble)}) { assert(Semantics == &semPPCDoubleDouble); } DoubleAPFloat::DoubleAPFloat(const fltSemantics &S, const APInt &I) : Semantics(&S), Floats(new APFloat[2]{ APFloat(semIEEEdouble, APInt(64, I.getRawData()[0])), APFloat(semIEEEdouble, APInt(64, I.getRawData()[1]))}) { assert(Semantics == &semPPCDoubleDouble); } DoubleAPFloat::DoubleAPFloat(const fltSemantics &S, APFloat &&First, APFloat &&Second) : Semantics(&S), Floats(new APFloat[2]{std::move(First), std::move(Second)}) { assert(Semantics == &semPPCDoubleDouble); assert(&Floats[0].getSemantics() == &semIEEEdouble); assert(&Floats[1].getSemantics() == &semIEEEdouble); } DoubleAPFloat::DoubleAPFloat(const DoubleAPFloat &RHS) : Semantics(RHS.Semantics), Floats(RHS.Floats ? new APFloat[2]{APFloat(RHS.Floats[0]), APFloat(RHS.Floats[1])} : nullptr) { assert(Semantics == &semPPCDoubleDouble); } DoubleAPFloat::DoubleAPFloat(DoubleAPFloat &&RHS) : Semantics(RHS.Semantics), Floats(std::move(RHS.Floats)) { RHS.Semantics = &semBogus; assert(Semantics == &semPPCDoubleDouble); } DoubleAPFloat &DoubleAPFloat::operator=(const DoubleAPFloat &RHS) { if (Semantics == RHS.Semantics && RHS.Floats) { Floats[0] = RHS.Floats[0]; Floats[1] = RHS.Floats[1]; } else if (this != &RHS) { this->~DoubleAPFloat(); new (this) DoubleAPFloat(RHS); } return *this; } // Implement addition, subtraction, multiplication and division based on: // "Software for Doubled-Precision Floating-Point Computations", // by Seppo Linnainmaa, ACM TOMS vol 7 no 3, September 1981, pages 272-283. APFloat::opStatus DoubleAPFloat::addImpl(const APFloat &a, const APFloat &aa, const APFloat &c, const APFloat &cc, roundingMode RM) { int Status = opOK; APFloat z = a; Status |= z.add(c, RM); if (!z.isFinite()) { if (!z.isInfinity()) { Floats[0] = std::move(z); Floats[1].makeZero(/* Neg = */ false); return (opStatus)Status; } Status = opOK; auto AComparedToC = a.compareAbsoluteValue(c); z = cc; Status |= z.add(aa, RM); if (AComparedToC == APFloat::cmpGreaterThan) { // z = cc + aa + c + a; Status |= z.add(c, RM); Status |= z.add(a, RM); } else { // z = cc + aa + a + c; Status |= z.add(a, RM); Status |= z.add(c, RM); } if (!z.isFinite()) { Floats[0] = std::move(z); Floats[1].makeZero(/* Neg = */ false); return (opStatus)Status; } Floats[0] = z; APFloat zz = aa; Status |= zz.add(cc, RM); if (AComparedToC == APFloat::cmpGreaterThan) { // Floats[1] = a - z + c + zz; Floats[1] = a; Status |= Floats[1].subtract(z, RM); Status |= Floats[1].add(c, RM); Status |= Floats[1].add(zz, RM); } else { // Floats[1] = c - z + a + zz; Floats[1] = c; Status |= Floats[1].subtract(z, RM); Status |= Floats[1].add(a, RM); Status |= Floats[1].add(zz, RM); } } else { // q = a - z; APFloat q = a; Status |= q.subtract(z, RM); // zz = q + c + (a - (q + z)) + aa + cc; // Compute a - (q + z) as -((q + z) - a) to avoid temporary copies. auto zz = q; Status |= zz.add(c, RM); Status |= q.add(z, RM); Status |= q.subtract(a, RM); q.changeSign(); Status |= zz.add(q, RM); Status |= zz.add(aa, RM); Status |= zz.add(cc, RM); if (zz.isZero() && !zz.isNegative()) { Floats[0] = std::move(z); Floats[1].makeZero(/* Neg = */ false); return opOK; } Floats[0] = z; Status |= Floats[0].add(zz, RM); if (!Floats[0].isFinite()) { Floats[1].makeZero(/* Neg = */ false); return (opStatus)Status; } Floats[1] = std::move(z); Status |= Floats[1].subtract(Floats[0], RM); Status |= Floats[1].add(zz, RM); } return (opStatus)Status; } APFloat::opStatus DoubleAPFloat::addWithSpecial(const DoubleAPFloat &LHS, const DoubleAPFloat &RHS, DoubleAPFloat &Out, roundingMode RM) { if (LHS.getCategory() == fcNaN) { Out = LHS; return opOK; } if (RHS.getCategory() == fcNaN) { Out = RHS; return opOK; } if (LHS.getCategory() == fcZero) { Out = RHS; return opOK; } if (RHS.getCategory() == fcZero) { Out = LHS; return opOK; } if (LHS.getCategory() == fcInfinity && RHS.getCategory() == fcInfinity && LHS.isNegative() != RHS.isNegative()) { Out.makeNaN(false, Out.isNegative(), nullptr); return opInvalidOp; } if (LHS.getCategory() == fcInfinity) { Out = LHS; return opOK; } if (RHS.getCategory() == fcInfinity) { Out = RHS; return opOK; } assert(LHS.getCategory() == fcNormal && RHS.getCategory() == fcNormal); APFloat A(LHS.Floats[0]), AA(LHS.Floats[1]), C(RHS.Floats[0]), CC(RHS.Floats[1]); assert(&A.getSemantics() == &semIEEEdouble); assert(&AA.getSemantics() == &semIEEEdouble); assert(&C.getSemantics() == &semIEEEdouble); assert(&CC.getSemantics() == &semIEEEdouble); assert(&Out.Floats[0].getSemantics() == &semIEEEdouble); assert(&Out.Floats[1].getSemantics() == &semIEEEdouble); return Out.addImpl(A, AA, C, CC, RM); } APFloat::opStatus DoubleAPFloat::add(const DoubleAPFloat &RHS, roundingMode RM) { return addWithSpecial(*this, RHS, *this, RM); } APFloat::opStatus DoubleAPFloat::subtract(const DoubleAPFloat &RHS, roundingMode RM) { changeSign(); auto Ret = add(RHS, RM); changeSign(); return Ret; } APFloat::opStatus DoubleAPFloat::multiply(const DoubleAPFloat &RHS, APFloat::roundingMode RM) { const auto &LHS = *this; auto &Out = *this; /* Interesting observation: For special categories, finding the lowest common ancestor of the following layered graph gives the correct return category: NaN / \ Zero Inf \ / Normal e.g. NaN * NaN = NaN Zero * Inf = NaN Normal * Zero = Zero Normal * Inf = Inf */ if (LHS.getCategory() == fcNaN) { Out = LHS; return opOK; } if (RHS.getCategory() == fcNaN) { Out = RHS; return opOK; } if ((LHS.getCategory() == fcZero && RHS.getCategory() == fcInfinity) || (LHS.getCategory() == fcInfinity && RHS.getCategory() == fcZero)) { Out.makeNaN(false, false, nullptr); return opOK; } if (LHS.getCategory() == fcZero || LHS.getCategory() == fcInfinity) { Out = LHS; return opOK; } if (RHS.getCategory() == fcZero || RHS.getCategory() == fcInfinity) { Out = RHS; return opOK; } assert(LHS.getCategory() == fcNormal && RHS.getCategory() == fcNormal && "Special cases not handled exhaustively"); int Status = opOK; APFloat A = Floats[0], B = Floats[1], C = RHS.Floats[0], D = RHS.Floats[1]; // t = a * c APFloat T = A; Status |= T.multiply(C, RM); if (!T.isFiniteNonZero()) { Floats[0] = T; Floats[1].makeZero(/* Neg = */ false); return (opStatus)Status; } // tau = fmsub(a, c, t), that is -fmadd(-a, c, t). APFloat Tau = A; T.changeSign(); Status |= Tau.fusedMultiplyAdd(C, T, RM); T.changeSign(); { // v = a * d APFloat V = A; Status |= V.multiply(D, RM); // w = b * c APFloat W = B; Status |= W.multiply(C, RM); Status |= V.add(W, RM); // tau += v + w Status |= Tau.add(V, RM); } // u = t + tau APFloat U = T; Status |= U.add(Tau, RM); Floats[0] = U; if (!U.isFinite()) { Floats[1].makeZero(/* Neg = */ false); } else { // Floats[1] = (t - u) + tau Status |= T.subtract(U, RM); Status |= T.add(Tau, RM); Floats[1] = T; } return (opStatus)Status; } APFloat::opStatus DoubleAPFloat::divide(const DoubleAPFloat &RHS, APFloat::roundingMode RM) { assert(Semantics == &semPPCDoubleDouble && "Unexpected Semantics"); APFloat Tmp(semPPCDoubleDoubleLegacy, bitcastToAPInt()); auto Ret = Tmp.divide(APFloat(semPPCDoubleDoubleLegacy, RHS.bitcastToAPInt()), RM); *this = DoubleAPFloat(semPPCDoubleDouble, Tmp.bitcastToAPInt()); return Ret; } APFloat::opStatus DoubleAPFloat::remainder(const DoubleAPFloat &RHS) { assert(Semantics == &semPPCDoubleDouble && "Unexpected Semantics"); APFloat Tmp(semPPCDoubleDoubleLegacy, bitcastToAPInt()); auto Ret = Tmp.remainder(APFloat(semPPCDoubleDoubleLegacy, RHS.bitcastToAPInt())); *this = DoubleAPFloat(semPPCDoubleDouble, Tmp.bitcastToAPInt()); return Ret; } APFloat::opStatus DoubleAPFloat::mod(const DoubleAPFloat &RHS) { assert(Semantics == &semPPCDoubleDouble && "Unexpected Semantics"); APFloat Tmp(semPPCDoubleDoubleLegacy, bitcastToAPInt()); auto Ret = Tmp.mod(APFloat(semPPCDoubleDoubleLegacy, RHS.bitcastToAPInt())); *this = DoubleAPFloat(semPPCDoubleDouble, Tmp.bitcastToAPInt()); return Ret; } APFloat::opStatus DoubleAPFloat::fusedMultiplyAdd(const DoubleAPFloat &Multiplicand, const DoubleAPFloat &Addend, APFloat::roundingMode RM) { assert(Semantics == &semPPCDoubleDouble && "Unexpected Semantics"); APFloat Tmp(semPPCDoubleDoubleLegacy, bitcastToAPInt()); auto Ret = Tmp.fusedMultiplyAdd( APFloat(semPPCDoubleDoubleLegacy, Multiplicand.bitcastToAPInt()), APFloat(semPPCDoubleDoubleLegacy, Addend.bitcastToAPInt()), RM); *this = DoubleAPFloat(semPPCDoubleDouble, Tmp.bitcastToAPInt()); return Ret; } APFloat::opStatus DoubleAPFloat::roundToIntegral(APFloat::roundingMode RM) { assert(Semantics == &semPPCDoubleDouble && "Unexpected Semantics"); APFloat Tmp(semPPCDoubleDoubleLegacy, bitcastToAPInt()); auto Ret = Tmp.roundToIntegral(RM); *this = DoubleAPFloat(semPPCDoubleDouble, Tmp.bitcastToAPInt()); return Ret; } void DoubleAPFloat::changeSign() { Floats[0].changeSign(); Floats[1].changeSign(); } APFloat::cmpResult DoubleAPFloat::compareAbsoluteValue(const DoubleAPFloat &RHS) const { auto Result = Floats[0].compareAbsoluteValue(RHS.Floats[0]); if (Result != cmpEqual) return Result; Result = Floats[1].compareAbsoluteValue(RHS.Floats[1]); if (Result == cmpLessThan || Result == cmpGreaterThan) { auto Against = Floats[0].isNegative() ^ Floats[1].isNegative(); auto RHSAgainst = RHS.Floats[0].isNegative() ^ RHS.Floats[1].isNegative(); if (Against && !RHSAgainst) return cmpLessThan; if (!Against && RHSAgainst) return cmpGreaterThan; if (!Against && !RHSAgainst) return Result; if (Against && RHSAgainst) return (cmpResult)(cmpLessThan + cmpGreaterThan - Result); } return Result; } APFloat::fltCategory DoubleAPFloat::getCategory() const { return Floats[0].getCategory(); } bool DoubleAPFloat::isNegative() const { return Floats[0].isNegative(); } void DoubleAPFloat::makeInf(bool Neg) { Floats[0].makeInf(Neg); Floats[1].makeZero(/* Neg = */ false); } void DoubleAPFloat::makeZero(bool Neg) { Floats[0].makeZero(Neg); Floats[1].makeZero(/* Neg = */ false); } void DoubleAPFloat::makeLargest(bool Neg) { assert(Semantics == &semPPCDoubleDouble && "Unexpected Semantics"); Floats[0] = APFloat(semIEEEdouble, APInt(64, 0x7fefffffffffffffull)); Floats[1] = APFloat(semIEEEdouble, APInt(64, 0x7c8ffffffffffffeull)); if (Neg) changeSign(); } void DoubleAPFloat::makeSmallest(bool Neg) { assert(Semantics == &semPPCDoubleDouble && "Unexpected Semantics"); Floats[0].makeSmallest(Neg); Floats[1].makeZero(/* Neg = */ false); } void DoubleAPFloat::makeSmallestNormalized(bool Neg) { assert(Semantics == &semPPCDoubleDouble && "Unexpected Semantics"); Floats[0] = APFloat(semIEEEdouble, APInt(64, 0x0360000000000000ull)); if (Neg) Floats[0].changeSign(); Floats[1].makeZero(/* Neg = */ false); } void DoubleAPFloat::makeNaN(bool SNaN, bool Neg, const APInt *fill) { Floats[0].makeNaN(SNaN, Neg, fill); Floats[1].makeZero(/* Neg = */ false); } APFloat::cmpResult DoubleAPFloat::compare(const DoubleAPFloat &RHS) const { auto Result = Floats[0].compare(RHS.Floats[0]); // |Float[0]| > |Float[1]| if (Result == APFloat::cmpEqual) return Floats[1].compare(RHS.Floats[1]); return Result; } bool DoubleAPFloat::bitwiseIsEqual(const DoubleAPFloat &RHS) const { return Floats[0].bitwiseIsEqual(RHS.Floats[0]) && Floats[1].bitwiseIsEqual(RHS.Floats[1]); } hash_code hash_value(const DoubleAPFloat &Arg) { if (Arg.Floats) return hash_combine(hash_value(Arg.Floats[0]), hash_value(Arg.Floats[1])); return hash_combine(Arg.Semantics); } APInt DoubleAPFloat::bitcastToAPInt() const { assert(Semantics == &semPPCDoubleDouble && "Unexpected Semantics"); uint64_t Data[] = { Floats[0].bitcastToAPInt().getRawData()[0], Floats[1].bitcastToAPInt().getRawData()[0], }; return APInt(128, 2, Data); } Expected DoubleAPFloat::convertFromString(StringRef S, roundingMode RM) { assert(Semantics == &semPPCDoubleDouble && "Unexpected Semantics"); APFloat Tmp(semPPCDoubleDoubleLegacy); auto Ret = Tmp.convertFromString(S, RM); *this = DoubleAPFloat(semPPCDoubleDouble, Tmp.bitcastToAPInt()); return Ret; } APFloat::opStatus DoubleAPFloat::next(bool nextDown) { assert(Semantics == &semPPCDoubleDouble && "Unexpected Semantics"); APFloat Tmp(semPPCDoubleDoubleLegacy, bitcastToAPInt()); auto Ret = Tmp.next(nextDown); *this = DoubleAPFloat(semPPCDoubleDouble, Tmp.bitcastToAPInt()); return Ret; } APFloat::opStatus DoubleAPFloat::convertToInteger(MutableArrayRef Input, unsigned int Width, bool IsSigned, roundingMode RM, bool *IsExact) const { assert(Semantics == &semPPCDoubleDouble && "Unexpected Semantics"); return APFloat(semPPCDoubleDoubleLegacy, bitcastToAPInt()) .convertToInteger(Input, Width, IsSigned, RM, IsExact); } APFloat::opStatus DoubleAPFloat::convertFromAPInt(const APInt &Input, bool IsSigned, roundingMode RM) { assert(Semantics == &semPPCDoubleDouble && "Unexpected Semantics"); APFloat Tmp(semPPCDoubleDoubleLegacy); auto Ret = Tmp.convertFromAPInt(Input, IsSigned, RM); *this = DoubleAPFloat(semPPCDoubleDouble, Tmp.bitcastToAPInt()); return Ret; } APFloat::opStatus DoubleAPFloat::convertFromSignExtendedInteger(const integerPart *Input, unsigned int InputSize, bool IsSigned, roundingMode RM) { assert(Semantics == &semPPCDoubleDouble && "Unexpected Semantics"); APFloat Tmp(semPPCDoubleDoubleLegacy); auto Ret = Tmp.convertFromSignExtendedInteger(Input, InputSize, IsSigned, RM); *this = DoubleAPFloat(semPPCDoubleDouble, Tmp.bitcastToAPInt()); return Ret; } APFloat::opStatus DoubleAPFloat::convertFromZeroExtendedInteger(const integerPart *Input, unsigned int InputSize, bool IsSigned, roundingMode RM) { assert(Semantics == &semPPCDoubleDouble && "Unexpected Semantics"); APFloat Tmp(semPPCDoubleDoubleLegacy); auto Ret = Tmp.convertFromZeroExtendedInteger(Input, InputSize, IsSigned, RM); *this = DoubleAPFloat(semPPCDoubleDouble, Tmp.bitcastToAPInt()); return Ret; } unsigned int DoubleAPFloat::convertToHexString(char *DST, unsigned int HexDigits, bool UpperCase, roundingMode RM) const { assert(Semantics == &semPPCDoubleDouble && "Unexpected Semantics"); return APFloat(semPPCDoubleDoubleLegacy, bitcastToAPInt()) .convertToHexString(DST, HexDigits, UpperCase, RM); } bool DoubleAPFloat::isDenormal() const { return getCategory() == fcNormal && (Floats[0].isDenormal() || Floats[1].isDenormal() || // (double)(Hi + Lo) == Hi defines a normal number. Floats[0] != Floats[0] + Floats[1]); } bool DoubleAPFloat::isSmallest() const { if (getCategory() != fcNormal) return false; DoubleAPFloat Tmp(*this); Tmp.makeSmallest(this->isNegative()); return Tmp.compare(*this) == cmpEqual; } bool DoubleAPFloat::isSmallestNormalized() const { if (getCategory() != fcNormal) return false; DoubleAPFloat Tmp(*this); Tmp.makeSmallestNormalized(this->isNegative()); return Tmp.compare(*this) == cmpEqual; } bool DoubleAPFloat::isLargest() const { if (getCategory() != fcNormal) return false; DoubleAPFloat Tmp(*this); Tmp.makeLargest(this->isNegative()); return Tmp.compare(*this) == cmpEqual; } bool DoubleAPFloat::isInteger() const { assert(Semantics == &semPPCDoubleDouble && "Unexpected Semantics"); return Floats[0].isInteger() && Floats[1].isInteger(); } void DoubleAPFloat::toString(SmallVectorImpl &Str, unsigned FormatPrecision, unsigned FormatMaxPadding, bool TruncateZero) const { assert(Semantics == &semPPCDoubleDouble && "Unexpected Semantics"); APFloat(semPPCDoubleDoubleLegacy, bitcastToAPInt()) .toString(Str, FormatPrecision, FormatMaxPadding, TruncateZero); } bool DoubleAPFloat::getExactInverse(APFloat *inv) const { assert(Semantics == &semPPCDoubleDouble && "Unexpected Semantics"); APFloat Tmp(semPPCDoubleDoubleLegacy, bitcastToAPInt()); if (!inv) return Tmp.getExactInverse(nullptr); APFloat Inv(semPPCDoubleDoubleLegacy); auto Ret = Tmp.getExactInverse(&Inv); *inv = APFloat(semPPCDoubleDouble, Inv.bitcastToAPInt()); return Ret; } int DoubleAPFloat::getExactLog2() const { // TODO: Implement me return INT_MIN; } int DoubleAPFloat::getExactLog2Abs() const { // TODO: Implement me return INT_MIN; } DoubleAPFloat scalbn(const DoubleAPFloat &Arg, int Exp, APFloat::roundingMode RM) { assert(Arg.Semantics == &semPPCDoubleDouble && "Unexpected Semantics"); return DoubleAPFloat(semPPCDoubleDouble, scalbn(Arg.Floats[0], Exp, RM), scalbn(Arg.Floats[1], Exp, RM)); } DoubleAPFloat frexp(const DoubleAPFloat &Arg, int &Exp, APFloat::roundingMode RM) { assert(Arg.Semantics == &semPPCDoubleDouble && "Unexpected Semantics"); APFloat First = frexp(Arg.Floats[0], Exp, RM); APFloat Second = Arg.Floats[1]; if (Arg.getCategory() == APFloat::fcNormal) Second = scalbn(Second, -Exp, RM); return DoubleAPFloat(semPPCDoubleDouble, std::move(First), std::move(Second)); } } // namespace detail APFloat::Storage::Storage(IEEEFloat F, const fltSemantics &Semantics) { if (usesLayout(Semantics)) { new (&IEEE) IEEEFloat(std::move(F)); return; } if (usesLayout(Semantics)) { const fltSemantics& S = F.getSemantics(); new (&Double) DoubleAPFloat(Semantics, APFloat(std::move(F), S), APFloat(semIEEEdouble)); return; } llvm_unreachable("Unexpected semantics"); } Expected APFloat::convertFromString(StringRef Str, roundingMode RM) { APFLOAT_DISPATCH_ON_SEMANTICS(convertFromString(Str, RM)); } hash_code hash_value(const APFloat &Arg) { if (APFloat::usesLayout(Arg.getSemantics())) return hash_value(Arg.U.IEEE); if (APFloat::usesLayout(Arg.getSemantics())) return hash_value(Arg.U.Double); llvm_unreachable("Unexpected semantics"); } APFloat::APFloat(const fltSemantics &Semantics, StringRef S) : APFloat(Semantics) { auto StatusOrErr = convertFromString(S, rmNearestTiesToEven); assert(StatusOrErr && "Invalid floating point representation"); consumeError(StatusOrErr.takeError()); } FPClassTest APFloat::classify() const { if (isZero()) return isNegative() ? fcNegZero : fcPosZero; if (isNormal()) return isNegative() ? fcNegNormal : fcPosNormal; if (isDenormal()) return isNegative() ? fcNegSubnormal : fcPosSubnormal; if (isInfinity()) return isNegative() ? fcNegInf : fcPosInf; assert(isNaN() && "Other class of FP constant"); return isSignaling() ? fcSNan : fcQNan; } APFloat::opStatus APFloat::convert(const fltSemantics &ToSemantics, roundingMode RM, bool *losesInfo) { if (&getSemantics() == &ToSemantics) { *losesInfo = false; return opOK; } if (usesLayout(getSemantics()) && usesLayout(ToSemantics)) return U.IEEE.convert(ToSemantics, RM, losesInfo); if (usesLayout(getSemantics()) && usesLayout(ToSemantics)) { assert(&ToSemantics == &semPPCDoubleDouble); auto Ret = U.IEEE.convert(semPPCDoubleDoubleLegacy, RM, losesInfo); *this = APFloat(ToSemantics, U.IEEE.bitcastToAPInt()); return Ret; } if (usesLayout(getSemantics()) && usesLayout(ToSemantics)) { auto Ret = getIEEE().convert(ToSemantics, RM, losesInfo); *this = APFloat(std::move(getIEEE()), ToSemantics); return Ret; } llvm_unreachable("Unexpected semantics"); } APFloat APFloat::getAllOnesValue(const fltSemantics &Semantics) { return APFloat(Semantics, APInt::getAllOnes(Semantics.sizeInBits)); } void APFloat::print(raw_ostream &OS) const { SmallVector Buffer; toString(Buffer); OS << Buffer << "\n"; } #if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP) LLVM_DUMP_METHOD void APFloat::dump() const { print(dbgs()); } #endif void APFloat::Profile(FoldingSetNodeID &NID) const { NID.Add(bitcastToAPInt()); } /* Same as convertToInteger(integerPart*, ...), except the result is returned in an APSInt, whose initial bit-width and signed-ness are used to determine the precision of the conversion. */ APFloat::opStatus APFloat::convertToInteger(APSInt &result, roundingMode rounding_mode, bool *isExact) const { unsigned bitWidth = result.getBitWidth(); SmallVector parts(result.getNumWords()); opStatus status = convertToInteger(parts, bitWidth, result.isSigned(), rounding_mode, isExact); // Keeps the original signed-ness. result = APInt(bitWidth, parts); return status; } double APFloat::convertToDouble() const { if (&getSemantics() == (const llvm::fltSemantics *)&semIEEEdouble) return getIEEE().convertToDouble(); assert(getSemantics().isRepresentableBy(semIEEEdouble) && "Float semantics is not representable by IEEEdouble"); APFloat Temp = *this; bool LosesInfo; opStatus St = Temp.convert(semIEEEdouble, rmNearestTiesToEven, &LosesInfo); assert(!(St & opInexact) && !LosesInfo && "Unexpected imprecision"); (void)St; return Temp.getIEEE().convertToDouble(); } #ifdef HAS_IEE754_FLOAT128 float128 APFloat::convertToQuad() const { if (&getSemantics() == (const llvm::fltSemantics *)&semIEEEquad) return getIEEE().convertToQuad(); assert(getSemantics().isRepresentableBy(semIEEEquad) && "Float semantics is not representable by IEEEquad"); APFloat Temp = *this; bool LosesInfo; opStatus St = Temp.convert(semIEEEquad, rmNearestTiesToEven, &LosesInfo); assert(!(St & opInexact) && !LosesInfo && "Unexpected imprecision"); (void)St; return Temp.getIEEE().convertToQuad(); } #endif float APFloat::convertToFloat() const { if (&getSemantics() == (const llvm::fltSemantics *)&semIEEEsingle) return getIEEE().convertToFloat(); assert(getSemantics().isRepresentableBy(semIEEEsingle) && "Float semantics is not representable by IEEEsingle"); APFloat Temp = *this; bool LosesInfo; opStatus St = Temp.convert(semIEEEsingle, rmNearestTiesToEven, &LosesInfo); assert(!(St & opInexact) && !LosesInfo && "Unexpected imprecision"); (void)St; return Temp.getIEEE().convertToFloat(); } } // namespace llvm #undef APFLOAT_DISPATCH_ON_SEMANTICS