Google Git
Sign in
android / platform / external / FP16 / refs/heads/android15-platform-release
commit7d0b441d2452787bd59031a63e3860c776368455[log] [tgz]
authorAndroid Build Coastguard Worker <[email protected]>Tue Apr 30 23:07:27 2024 +0000
committerAndroid Build Coastguard Worker <[email protected]>Tue Apr 30 23:07:27 2024 +0000
treecb376e49da4efb0db30089e589c64672c41dd208
parentc5217a62f33788774a977bd41b3adc306ad5a5b7 [diff]
parent5f32b7105932ea8520a0e8811c640f936367d707 [diff]
Snap for 11784291 from 5f32b7105932ea8520a0e8811c640f936367d707 to 24Q3-release

Change-Id: I6ef121b5f6bc6e8d35df7c532e62713399d56dc0
tree: cb376e49da4efb0db30089e589c64672c41dd208
  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.