IEEE-754-float16

complete software implementation of float16 written in c99

Operation

• Arithmetic (+, -, *, /), (fp16_add, fp16_sub, fp16_mul, fp16_div)

• Compare (>, <, <=, >=, ==, !=), (fp16_gt, fp16_lt, fp16_lte, fp16_gte, fp16_eq, fp16_neq)

• Special case handle (inf, nan, subnormal)

• Rounding (rint, lrint, trunc, ceil, floor, round)

• Math operations (trigonometry, exponental, logarithmic, roots, etc..)

• Special functions (fp16_fma, fp16_ilogb, fp16_hypot, fp16_significand, fp16_scalb, fp16_scalbn)

Supported Exception

• INVALID

• OVERFLOW

• UNDERFLOW

• INEXACT

Supported Rounding

• TOWARDZERO

• UPWARD

• DOWNWARD

• TONEAREST

C++ Example

 #include "fp16/float16.hpp"
 #include <iostream>

 int main() {
  fp::float16 a = fp::float16(4.5f);
  fp::float16 b = fp::float16(78.9f);
  fp::float16 c = fp::sin(a + b);

  std::cout << static_cast<float>(c);
  return 0;
 }

C Example

 #include "fp16/float16.h"
 #include "fp16/math.h"
 #include <stdio.h>

 int main() {
  uint16_t a = fp16_tofloat16(33.2f);
  uint16_t b = fp16_tofloat16(12.0f);
  uint16_t c = fp16_div(a, b);

  uint16_t d = fp16_fma(a, b, c);
  float out = fp16_tofloat32(d);
  printf("%.3f", out);
  return 0;
 }