Google Git
Sign in
android / platform / external / FP16 / cad49b6d5a26a83ebb760bd5ecec5f45344acd4f
commitcad49b6d5a26a83ebb760bd5ecec5f45344acd4f[log] [tgz]
authorFlorian Mayer <fmayer@google.com>Fri Mar 18 15:29:13 2022 -0700
committerFlorian Mayer <fmayer@google.com>Fri Mar 18 15:29:13 2022 -0700
tree2c085b686688abfc7135e4727b9ff04a6e23a4b6
parent5484dbfabb408ee81ef8615182c9cc43ed9d885a [diff]
Add FP16 to hwasan-postsubmit

Test: Run tests under HWASan.
Bug: 193568145
Change-Id: Ife3dbf11219aaf159d12efde210adb7158084457
1 file changed
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.