Closed Ralith closed 5 years ago
I should've mentioned, credit goes to @sebcrozet for tracking this down all the way, I only confirmed that something was fishy.
I told you in https://github.com/bitshifter/mathbench-rs/issues/9 that you should just let each crate use its default features XD
I do understand "keep build time down" is a goal for a library, but this is a benchmark suite. Build times are nothing next to the execution time anyway. I think that all crates should be left in default mode.
Sometimes I want to rebuild with different settings like LTO, I care about how long that takes.
for things like that I'd check out the repo into two different folders
When I get back from vacation I'll rerun benchmarks and update the readme.
This was causing grossly misrepresentative results by forcing nalgebra to call out to libm for sqrt (for example) while other libraries were permitted to use the built-in version.