Closed lemire closed 3 years ago
@lemire Thanks! Would you mind rerunning this in a bit?
I'll finish a PR (hopefully within the hour) that (1) fixes the decimal bug, (2) improves the benching script and adds an all
option so you can get the full table at once for fast-float, for all benchmarks, (3) fix the benching itself a bit by removing needless addition operations (use volatile reads instead to prevent code folding), (4) adds a few micro-optimisations to parsing.
Also, it would be nice to get the updated parse_float
C++ benchmarks on the same machine (with big-ints changes, and also with the latest decimal changes - although those would only affect 2 of the benchmarks).
AMD Rome results also added to tables in the README, thanks!