nvneural::Float16 Class Reference

Float16 is a CPU-side implementation of half-precision floating point (aka FP16). More...

#include <nvneural/CoreTypes.h>

Public Member Functions
	Float16 ()=default
	Default constructor, initializes to zero.
	Float16 (const Float16 &copyFrom)
	Copy constructor.
	Float16 (float value)
	Converts a regular float to a Float16.
	operator float () const
	Converts the stored value to a float.
Float16 &	operator= (const Float16 &copyFrom)
	Assignment operator from Float16.
Float16 &	operator= (double value)
	Assignment operator from float/double.

Public Attributes
std::uint16_t	data = 0u
	Sized integer containing the float16 value.

Detailed Description

Float16 is a CPU-side implementation of half-precision floating point (aka FP16).

---

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

• Core/Inc/nvneural/CoreTypes.h
• Helpers/Src/Float16.cpp