Beatmup::GL::Float16 Class Reference

16 bit floating point representation. More...

#include <float16.h>

Public Member Functions
	Float16 (float input)
	Encodes a given floating point value in 16 bit. More...
	Float16 (uint8_t frac, uint8_t exp)
	Constructs a Float16 number from the fractional part and the exponent values given explicitly. More...
	operator float () const
	Converts the encoded value to a floating-point value. More...
uint8_t	getExp () const
	Returns encoded exponent. More...
uint8_t	getFrac () const
	Returns encoded fractional part. More...

Static Public Member Functions
static std::string	encodeGlsl (const std::string &name="encode")
	Returns a GLSL code defining the encoding function from a high precision floating point value into a low precision vec2. More...
static std::string	decodeGlsl (const std::string &name="decode")
	Returns a GLSL code defining the decoding function from a low precision vec2 to a high precision floating point value. More...

Private Attributes
uint8_t	frac
	fractional part More...
uint8_t	exp
	exponent part More...

Detailed Description

16 bit floating point representation.

The exponent and the signed fractional part are encoded to 8 bits each. x = frac * 2 ^ exp

Definition at line 31 of file float16.h.

Constructor & Destructor Documentation

Float16() [1/2]

Beatmup::GL::Float16::Float16 ( float  input )

inline

Encodes a given floating point value in 16 bit.

Parameters

input

The input floating point value

Definition at line 41 of file float16.h.

                                        {
                const int exponent = (int)(std::log2(std::abs(input)) + 1.0f);
                const float frac = std::ldexp(input, -exponent) * 128.0f + 128.5f;
                this->frac = frac < 0.0f ? 0 : frac > 255.0f ? 255 : (uint8_t)frac;
                this->exp = (uint8_t)(exponent + 128);
            }

Float16() [2/2]

Beatmup::GL::Float16::Float16 ( uint8_t  frac, uint8_t  exp  )

inline

Constructs a Float16 number from the fractional part and the exponent values given explicitly.

Parameters

frac	The fractional part
exponent	The exponent

Definition at line 53 of file float16.h.

: frac(frac), exp(exp) {}

Member Function Documentation

operator float()

Beatmup::GL::Float16::operator float ( ) const

inline

Converts the encoded value to a floating-point value.

Definition at line 58 of file float16.h.

                                          {
                return std::ldexp(frac / 128.0f - 1.0f, (int)exp - 128);
            }

getExp()

uint8_t Beatmup::GL::Float16::getExp ( ) const

inline

Returns encoded exponent.

Definition at line 65 of file float16.h.

                                          {
                return exp;
            }

getFrac()

uint8_t Beatmup::GL::Float16::getFrac ( ) const

inline

Returns encoded fractional part.

Definition at line 72 of file float16.h.

                                           {
                return frac;
            }

encodeGlsl()

static std::string Beatmup::GL::Float16::encodeGlsl ( const std::string &  name = "encode" )

inlinestatic

Returns a GLSL code defining the encoding function from a high precision floating point value into a low precision vec2.

Parameters

name	The function name

Definition at line 80 of file float16.h.

                                                                                 {
                return "lowp vec2 " + name + R"((highp float value) {
                    highp float e  = floor(log2(abs(value)) + 1.0);
                    return vec2(value * exp2(-e - 1.0) + (0.5 + 0.5/255.0), (e + 128.0) / 255.0);
                })";
            }

decodeGlsl()

static std::string Beatmup::GL::Float16::decodeGlsl ( const std::string &  name = "decode" )

inlinestatic

Returns a GLSL code defining the decoding function from a low precision vec2 to a high precision floating point value.

Parameters

name	The function name

Definition at line 91 of file float16.h.

                                                                                 {
                return "highp float " + name + R"((lowp vec2 pack) {
                    highp float e = floor(pack.y * 255.0 - 127.5);
                    return (floor(pack.x * 255.0) / 128.0 - 1.0) * exp2(e);
                })";
            }

Member Data Documentation

frac

uint8_t Beatmup::GL::Float16::frac

private

fractional part

Definition at line 33 of file float16.h.

exp

uint8_t Beatmup::GL::Float16::exp

private

exponent part

Definition at line 34 of file float16.h.

---

The documentation for this class was generated from the following file:

• core/gpu/float16.h