Google Git
Sign in
android / platform / external / FP16 / 821f0883646b7a96bc199300f9583791dc39e3b9
commit821f0883646b7a96bc199300f9583791dc39e3b9[log] [tgz]
authorandroid-build-team Robot <android-build-team-robot@google.com>Wed Feb 17 03:14:51 2021 +0000
committerandroid-build-team Robot <android-build-team-robot@google.com>Wed Feb 17 03:14:51 2021 +0000
treeb299e45392655430c6cf1d183f304f65ac74f1ce
parentb3a2ec2204f036289d3df3063056207053db31f1 [diff]
parent9a78dff918d9ff9c13af0e9d271c60f85b2e1ebc [diff]
Snap for 7149879 from 9a78dff918d9ff9c13af0e9d271c60f85b2e1ebc to sc-release

Change-Id: I2ca55d3d2305166d0cbe0f91358fcefeb26feb9c
tree: b299e45392655430c6cf1d183f304f65ac74f1ce
  1. bench/
  2. cmake/
  3. include/
  4. test/
  5. .gitignore
  6. Android.bp
  7. CMakeLists.txt
  8. configure.py
  9. confu.yaml
  10. LICENSE
  11. METADATA
  12. MODULE_LICENSE_MIT
  13. OWNERS
  14. README.md
  15. TEST_MAPPING
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

HPC Garage logo Georgia Tech College of Computing logo

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.