|author||Marat Dukhan <firstname.lastname@example.org>||Sun Apr 12 02:31:17 2020|
|committer||Marat Dukhan <email@example.com>||Sun Apr 12 02:32:24 2020|
Fix bug in fp16_alt_to_fp32_value Wrong type for exp_offset constant didn't cause correctness issues as the constant fit into the wrong type, but potentially caused poor code-generation
Header-only library for conversion to/from half-precision floating point formats
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.