Google Git
Sign in
android / platform / external / FP16 / 25d9d036414d3251432cc32be7e3b3c97214c9e9
commit25d9d036414d3251432cc32be7e3b3c97214c9e9[log] [tgz]
authorYimin Li <[email protected]>Tue Apr 26 11:30:45 2022 -0700
committerYimin Li <[email protected]>Tue Apr 26 11:30:45 2022 -0700
treedf61340909aacfee61ba9b551e63df7ae3d3eceb
parentfa719c9ca65bafbec85348690fc03af69309b994 [diff]
parentbea6512c6f1dbab79c9cfa898e677ec832bcf7f6 [diff]
Merge commit 'bea6512c6f1dbab79c9cfa898e677ec832bcf7f6' of sso://googleplex-android/platform/external/FP16 into HEAD
Change-Id: I0a6aac4cce6363e6f005c885b4a3dcc4bda06995
tree: df61340909aacfee61ba9b551e63df7ae3d3eceb
  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.