Google Git
Sign in
android / platform / external / FP16 / refs/heads/android14-d1-release
commitbe8639608395396195769dadaf785a58b4d5d6df[log] [tgz]
authorAndroid Build Coastguard Worker <[email protected]>Wed Nov 02 03:06:52 2022 +0000
committerAndroid Build Coastguard Worker <[email protected]>Wed Nov 02 03:06:52 2022 +0000
treece0e0899872ba60de001e0ab3808516386ac81c2
parent7cbdbedccabdce9d78dd1e97a6481d15afc450ef [diff]
parent169d39b2978bc386396d57e00463fee62156fdb2 [diff]
Snap for 9245572 from 169d39b2978bc386396d57e00463fee62156fdb2 to udc-d1-release

Change-Id: Iddf1e39a69dbb0f974171263c2ded4b2c892b1c2
tree: ce0e0899872ba60de001e0ab3808516386ac81c2
  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.