Skip to content

Struct f16

struct f16 simd

struct f16 { … }

16-bit half-precision floating-point value stored as raw bits.

variable raw 

u16 raw

Raw bit representation of the half-precision value.

Fully qualified name: kfr::f16::raw

Declared at simd/types.hpp:312

constructor f16() 

f16() noexcept = default

Default constructor leaving the value uninitialized.

Fully qualified name: kfr::f16::f16()

Declared at simd/types.hpp:315

function from_raw(u16) 

static f16 from_raw(u16 raw) noexcept

Fully qualified name: kfr::f16::from_raw(u16)

Declared at simd/types.hpp:317

constructor f16(f32) 

f16(f32 value) noexcept

Converts a single-precision value to IEEE 754 binary16.

Uses F16C when the compilation target enables it; otherwise performs round-to-nearest, ties-to-even conversion in integer arithmetic.

Fully qualified name: kfr::f16::f16(f32)

Declared at simd/types.hpp:330

operator operator float() 

operator f32() const noexcept

Converts this IEEE 754 binary16 value to single precision.

Uses F16C when the compilation target enables it. The software path exactly preserves zeros, infinities, NaN payloads, and subnormals.

Fully qualified name: kfr::f16::operator float()

Declared at simd/types.hpp:346

Fully qualified name: kfr::f16

Declared at simd/types.hpp:309

This file was generated by cxxdox, a C++ documentation generator based on MkDocs Material and libclang.