Google Git
Sign in
android / platform / external / FP16 / b4bb3c3b96688fbc81df2dad4854524ac2f5513e
commitb4bb3c3b96688fbc81df2dad4854524ac2f5513e[log] [tgz]
authorAndroid Build Coastguard Worker <[email protected]>Wed May 11 05:03:50 2022 +0000
committerAndroid Build Coastguard Worker <[email protected]>Wed May 11 05:03:50 2022 +0000
tree2c085b686688abfc7135e4727b9ff04a6e23a4b6
parent692d39568e39535f3772b5ffaa9be64e605ef24a [diff]
parenta2d590fa0617fcd0eba86d5545c4242ee3f37c24 [diff]
Snap for 8570526 from a2d590fa0617fcd0eba86d5545c4242ee3f37c24 to mainline-scheduling-release

Change-Id: Ieeb4e75d476c902c93ce2e979f873935da3170f6
tree: 2c085b686688abfc7135e4727b9ff04a6e23a4b6
  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.