Google Git
Sign in
android / platform / external / FP16 / 80907cb9bef73c460932da760aff5745cf8cb409
commit80907cb9bef73c460932da760aff5745cf8cb409[log] [tgz]
authorIan Hua <[email protected]>Thu Aug 12 11:16:47 2021 +0100
committerIan Hua <[email protected]>Fri Aug 13 13:56:08 2021 +0100
treedf61340909aacfee61ba9b551e63df7ae3d3eceb
parenta738f586f15db2566055279ac2de08559d40b8d1 [diff]
Add host_supported, min_sdk_version and apex_available configs to Android.bp for FP16.

This would solve build erros such as
"missing variant:os:linux_glibc,arch:x86_64" when adding
fp16_headers to external/tensorflow repo.

Bug: 196532440
Test: mma
Change-Id: Ia8ac144df6a18f17cc9c181ed8618a5d365abaa6
1 file changed
tree: df61340909aacfee61ba9b551e63df7ae3d3eceb
  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.