// // FILE: float16.cpp // AUTHOR: Rob Tillaart // VERSION: 0.1.5 // PURPOSE: library for Float16s for Arduino // URL: http://en.wikipedia.org/wiki/Half-precision_floating-point_format // // HISTORY: // 0.1.00 2015-03-10 initial version // 0.1.01 2015-03-12 make base conversion separate functions // 0.1.02 2015-03-14 getting rounding right // 0.1.03 // 0.1.4 2021-11-26 setup repo to get it working again. // still experimental. // 0.1.5 2021-12-02 add basic math, optimize compare operators // 0.1.6 2021-12-18 update library.json, license, minor edits #include "float16.h" // #define DEBUG // CONSTRUCTOR float16::float16(double f) { _value = f32tof16(f); } // PRINTING size_t float16::printTo(Print& p) const { double d = this->f16tof32(_value); return p.print(d, _decimals); }; double float16::toDouble() const { return f16tof32(_value); } ////////////////////////////////////////////////////////// // // EQUALITIES // bool float16::operator == (const float16 &f) { return (_value == f._value); } bool float16::operator != (const float16 &f) { return (_value != f._value); } bool float16::operator > (const float16 &f) { if ((_value & 0x8000) && ( f._value & 0x8000)) return _value < f._value; if (_value & 0x8000) return false; if (f._value & 0x8000) return true; return _value > f._value; } bool float16::operator >= (const float16 &f) { if ((_value & 0x8000) && (f._value & 0x8000)) return _value <= f._value; if (_value & 0x8000) return false; if (f._value & 0x8000) return true; return _value >= f._value; } bool float16::operator < (const float16 &f) { if ((_value & 0x8000) && (f._value & 0x8000)) return _value > f._value; if (_value & 0x8000) return true; if (f._value & 0x8000) return false; return _value < f._value; } bool float16::operator <= (const float16 &f) { if ((_value & 0x8000) && (f._value & 0x8000)) return _value >= f._value; if (_value & 0x8000) return true; if (f._value & 0x8000) return false; return _value <= f._value; } ////////////////////////////////////////////////////////// // // NEGATION // float16 float16::operator - () { float16 f16; f16.setBinary(_value ^ 0x8000); return f16; } ////////////////////////////////////////////////////////// // // MATH // float16 float16::operator + (const float16 &f) { return float16(this->toDouble() + f.toDouble()); } float16 float16::operator - (const float16 &f) { return float16(this->toDouble() - f.toDouble()); } float16 float16::operator * (const float16 &f) { return float16(this->toDouble() * f.toDouble()); } float16 float16::operator / (const float16 &f) { return float16(this->toDouble() / f.toDouble()); } float16& float16::operator += (const float16 &f) { *this = this->toDouble() + f.toDouble(); return *this; } float16& float16::operator -= (const float16 &f) { *this = this->toDouble() - f.toDouble(); return *this; } float16& float16::operator *= (const float16 &f) { *this = this->toDouble() * f.toDouble(); return *this; } float16& float16::operator /= (const float16 &f) { *this = this->toDouble() / f.toDouble(); return *this; } ////////////////////////////////////////////////////////// // // MATH HELPER FUNCTIONS // int float16::sign() { if (_value & 0x8000) return -1; if (_value & 0xFFFF) return 1; return 0; } bool float16::isZero() { return ((_value & 0x7FFF) == 0x0000); } // bool float16::isNaN() // { // return ((_value & 0x7FFF) == 0x0000); // } bool float16::isInf() { return ((_value == 0x7C00) || (_value == 0xFC00)); } ////////////////////////////////////////////////////////// // // CORE CONVERSION // float float16::f16tof32(uint16_t _value) const { uint16_t sgn, man; int exp; double f; sgn = (_value & 0x8000) > 0; exp = (_value & 0x7C00) >> 10; man = (_value & 0x03FF); // ZERO if ((_value & 0x7FFF) == 0) { return sgn ? -0 : 0; } // NAN & INF if (exp == 0x001F) { if (man == 0) return sgn ? -INFINITY : INFINITY; else return NAN; } // SUBNORMAL/NORMAL if (exp == 0) f = 0; else f = 1; // PROCESS MANTISSE for (int i = 9; i >= 0; i--) { f *= 2; if (man & (1 << i)) f = f + 1; } f = f * pow(2.0, exp - 25); if (exp == 0) { f = f * pow(2.0, -13); // 5.96046447754e-8; } return sgn ? -f : f; } uint16_t float16::f32tof16(float f) const { uint32_t t = *(uint32_t *) &f; // man bits = 10; but we keep 11 for rounding uint16_t man = (t & 0x007FFFFF) >> 12; int16_t exp = (t & 0x7F800000) >> 23; bool sgn = (t & 0x80000000); // handle 0 if ((t & 0x7FFFFFFF) == 0) { return sgn ? 0x8000 : 0x0000; } // denormalized float32 does not fit in float16 if (exp == 0x00) { return sgn ? 0x8000 : 0x0000; } // handle infinity & NAN if (exp == 0x00FF) { if (man) return 0xFE00; // NAN return sgn ? 0xFC00 : 0x7C00; // -INF : INF } // normal numbers exp = exp - 127 + 15; // overflow does not fit => INF if (exp > 30) { return sgn ? 0xFC00 : 0x7C00; // -INF : INF } // subnormal numbers if (exp < -38) { return sgn ? 0x8000 : 0x0000; // -0 or 0 ? just 0 ? } if (exp <= 0) // subnormal { man >>= (exp + 14); // rounding man++; man >>= 1; if (sgn) return 0x8000 | man; return man; } // normal // TODO rounding exp <<= 10; man++; man >>= 1; if (sgn) return 0x8000 | exp | man; return exp | man; } // -- END OF FILE --