Google Git
Sign in
android / platform / external / FP16 / refs/heads/android14-qpr2-s4-release
commit5df590299c8d310852505d6f347849146c338fcf[log] [tgz]
authorAndroid Build Coastguard Worker <[email protected]>Sat Oct 07 23:07:21 2023 +0000
committerAndroid Build Coastguard Worker <[email protected]>Sat Oct 07 23:07:21 2023 +0000
tree5861de0ba4651a75bafe7097b5665fed86eb6903
parentc295e338658ee77208a94f99f9c78d8257f31a94 [diff]
parent9e3376fdab33d108e5d2dc4918f3692dc9330f61 [diff]
Snap for 10917888 from 9e3376fdab33d108e5d2dc4918f3692dc9330f61 to 24Q1-release

Change-Id: I113b25c6bd94db8c83d9fdad76ad7eda59b76a12
tree: 5861de0ba4651a75bafe7097b5665fed86eb6903
  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.