Google Git
Sign in
android / platform / external / FP16 / c6087aa5d1f70b4d46b3f9306b3057514c40c88d
commitc6087aa5d1f70b4d46b3f9306b3057514c40c88d[log] [tgz]
authorAndroid Build Coastguard Worker <[email protected]>Sat Nov 13 04:02:58 2021 +0000
committerAndroid Build Coastguard Worker <[email protected]>Sat Nov 13 04:02:58 2021 +0000
treedf61340909aacfee61ba9b551e63df7ae3d3eceb
parent174e833f96e5c123bde20fcc43062bec66942407 [diff]
parentc968563343fb4dd480e7765fa1e8bfa7ab783a96 [diff]
Snap for 7912506 from c968563343fb4dd480e7765fa1e8bfa7ab783a96 to tm-release

Change-Id: Ibf25b9ca2fe8d2c182865d9ff9717ce85cdc4d6f
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.