Closed ronniec95 closed 4 years ago
Do not merge!
okay, will not merge it
Now you can merge.
Please check the notes. There is a bug but it should only affect nan handling in some cases. If you can take a look and fix it, it would be appreciated. (f64x4,f64x2 sign_bits)
i can look at and probably merge this later today.
I'm not sure I'll have time for wide
bug fixing today or this weekend.
However, T-libs is starting a project group for portable simd in core
very soon, so with luck a lot of code either here or in packed_simd
will slot right in. The timeline goal is to have it done and stable by late 2021
packed_simd
is the more powerful library with it's comprehensive set of math functions and everything, and I would use it except that it doesn't work on stable. But yes, keep in the loop, I'd love to contribute
oops, looks like the other PR made a few conflicts here.
Conflicts resolved. Looks better. Feel free to merge. When's the next release? I want to push some of my own code up with depends on these new functions
Perhaps a 0.5.1 before a formal 0.6?
Have there been breaking changes yet? i think we can keep publishing under the 0.5 series at the moment
Great - will wait for your next 0.5.4 release.
Also had a look at packed_simd, it uses Sleef to implement the math functions; for a core library it would be strange to then depend on a C++ library. A pure rust implementation which you now have would be much better
okay, just put it out as 0.5.3
Added powf() functions for all the f32/f64 types. Also fixed ln() for i586 Disabled some tests in 0.5.2 for asin_acos which aren't working yet on i586