History

liiir1985 7f62dcda9f initial		2024-06-22 20:45:49 +08:00
..
bench	initial	2024-06-22 20:45:49 +08:00
cmake	initial	2024-06-22 20:45:49 +08:00
include	initial	2024-06-22 20:45:49 +08:00
test	initial	2024-06-22 20:45:49 +08:00
Android.bp	initial	2024-06-22 20:45:49 +08:00
CMakeLists.txt	initial	2024-06-22 20:45:49 +08:00
LICENSE	initial	2024-06-22 20:45:49 +08:00
METADATA	initial	2024-06-22 20:45:49 +08:00
MODULE_LICENSE_MIT	initial	2024-06-22 20:45:49 +08:00
OWNERS	initial	2024-06-22 20:45:49 +08:00
README.md	initial	2024-06-22 20:45:49 +08:00
TEST_MAPPING	initial	2024-06-22 20:45:49 +08:00
configure.py	initial	2024-06-22 20:45:49 +08:00
confu.yaml	initial	2024-06-22 20:45:49 +08:00

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
Fully covered with unit tests and microbenchmarks

Acknowledgements

The library is developed by Marat Dukhan of Georgia Tech. FP16 is a research project at Richard Vuduc's HPC Garage lab in the Georgia Institute of Technology, College of Computing, School of Computational Science and Engineering.

This material is based upon work supported by the U.S. National Science Foundation (NSF) Award Number 1339745. Any opinions, findings and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect those of NSF.