commit: e54a2ce7ca79c168461f41ac7f0313d73f0b76ed
[log]
author: Marat Dukhan <maratek@gmail.com>
Fri Feb 17 01:24:56 2017 -0500
committer: Marat Dukhan <maratek@gmail.com>
Fri Feb 17 01:24:56 2017 -0500
tree: a6fc5b0e7dadf81a1a54e0336a2c5e8da037cf9b
parent: 7f1c3cfa77e9da6ae5b0b316fe8597d6fbcf99e5 [diff]

Benchmarks of FP16->FP32 conversion on array

bench/alt-array.cc[Added - diff]
bench/alt-element.cc[Renamed from bench/alt.cc - diff]
bench/ieee-array.cc[Added - diff]
bench/ieee-element.cc[Renamed from bench/ieee.cc - diff]
configure.py[diff]

5 files changed

tree: a6fc5b0e7dadf81a1a54e0336a2c5e8da037cf9b

README.md

FP16

Header-only library for conversion to/from half-precision floating point formats

Features

Supports IEEE and ARM alternative half-precision floating-point format
- Property converts infinities and NaNs
- Properly converts denormal numbers, even on systems without denormal support
Header-only library, no installation or build required
Compatible with C99 and C++11
Covered with unit tests and microbenchmarks