search

Lokathor / wide

A crate to help you go wide. By which I mean use SIMD stuff.
https://docs.rs/wide
zlib License
251 stars 22 forks source link

improve perf of reduce_min,reduce_max,reduce_add and add unsigned upcast from u16x8 #154

Closed mcroomp closed 2 months ago

mcroomp commented 2 months ago

From some reason the optimizer isn't vectorizing the reduction operations and ends up emitting a bunch of slow scalar code.

Lokathor commented 2 months ago

published in 0.7.20