muu::half struct
#include <muu/half.h>

A 16-bit "half-precision" IEEE754 floating point type.

This type is equipped with the full set of operators you'd expect from a float type, and is capable of being converted to other floats and integers, as well as direct construction using the _f16 literal:

// creation from other floats and integers is explicit:
auto f1 = muu::half{ 1.0f };
auto f2 = static_cast<muu::half>(2);

// promotion to larger float types is implicit (to mimic the behaviour of the built-ins):
double f3 = f2;

// using the _f16 literal
using namespace muu::literals;
auto f4 = 4.0_f16;

// arithmetic operators
auto f5 = 4.0_f16 * 5.0_f16; // result is a half
auto f6 = 4.0_f16 * 5.0;     // result is a double because of promotion
auto f7 = 4.0_f16 * 5;       // result is a half because of promotion

// comparison operators
auto b1 = 4.0_f16 <= 5.0_f16;
auto b2 = 4.0_f16 <= 5.0;
auto b3 = 4.0_f16 <= 5;

Public types

using constants = muu::constants<half>: Static constants for this type.

Public static functions

static auto from_bits(uint16_t val) -> half constexpr noexcept: Creates a half-precision float from its raw bit equivalent.

Function documentation

half operator""_f16(long double val) consteval noexcept

Literal for creating a half-precision float.

const half val = 42.5_f16;

half operator""_f16(unsigned long long val) consteval noexcept