Google Git
Sign in
android / platform / external / FP16 / 654ee0f4d735470b425907508658b06cdc1c2d64
commit654ee0f4d735470b425907508658b06cdc1c2d64[log] [tgz]
authorHaibo Huang <[email protected]>Fri May 15 01:10:44 2020 +0000
committerAutomerger Merge Worker <[email protected]>Fri May 15 01:10:44 2020 +0000
tree30b48e320586c9d4e5beec06e7d903dc3fe55d94
parent572f521c8a2bbb60e8b10e90a3d1b87f9e727323 [diff]
parent169fc31aa8cbc9cb3ab21423ec01bbe60d3c8d0e [diff]
Upgrade FP16 to 4dfe081cf6bcd15db339cf2680b9281b8451eeb3 am: 7cea9a0f20 am: 6321f9ece6 am: 169fc31aa8

Change-Id: I7f5c2eb1fb6cc8d201036defebb6f14065c01688
tree: 30b48e320586c9d4e5beec06e7d903dc3fe55d94
  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.