Google Git
Sign in
android / platform / external / FP16 / ef8c5d3523dd098895dc906b8ceddfcfc6a3754a
commitef8c5d3523dd098895dc906b8ceddfcfc6a3754a[log] [tgz]
authorXin Li <[email protected]>Fri Apr 10 03:35:18 2020 +0000
committerAutomerger Merge Worker <[email protected]>Fri Apr 10 03:35:18 2020 +0000
tree83259cd4128a6cb31b6249f2ffc1045f7a4f4d43
parentcc78f545aa4f756db520f4399b1a6e2073669e0f [diff]
parent72a2f9a48ed2a2cd70ec4261713b19df84f4f29e [diff]
[automerger skipped] DO NOT MERGE - Empty merge qt-qpr1-dev-plus-aosp into stag-aosp-master am: 6cc7c334c3 -s ours am: 660f9ffc45 am: 72a2f9a48e

Change-Id: I957d899c28ba77f8b8c999ef316e9852b0b47d5f
tree: 83259cd4128a6cb31b6249f2ffc1045f7a4f4d43
  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.