Google Git
Sign in
android / platform / external / FP16 / b563b51b9540a719c2a29470a0b286d309fbc0e3
commitb563b51b9540a719c2a29470a0b286d309fbc0e3[log] [tgz]
authorAndroid Build Coastguard Worker <android-build-coastguard-worker@google.com>Fri Jul 07 04:50:44 2023 +0000
committerAndroid Build Coastguard Worker <android-build-coastguard-worker@google.com>Fri Jul 07 04:50:44 2023 +0000
treece0e0899872ba60de001e0ab3808516386ac81c2
parent8ba726615f86d6f6248b9d448ab61ea785c086f2 [diff]
parent7cbdbedccabdce9d78dd1e97a6481d15afc450ef [diff]
Snap for 10453563 from 7cbdbedccabdce9d78dd1e97a6481d15afc450ef to mainline-media-swcodec-release

Change-Id: I91dfe0535b6f7f2e794da7ec5d83c7d56eacee66
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.