manual struct target feature

[folkertdev]

this essentially inlines all avx2 codepaths.

this basically gives the same performance as -Ctarget-cpu=native on my machine.

but it's not the prettiest thing. I'm open to suggestions.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Files with missing lines	Coverage Δ
zlib-rs/src/cpu_features.rs	59.64% <ø> (ø)
zlib-rs/src/inflate.rs	91.17% <100.00%> (+0.04%)	⬆️
zlib-rs/src/inflate/writer.rs	89.17% <100.00%> (+1.23%)	⬆️

... and 1 file with indirect coverage changes

[folkertdev]

I tried specializing for avx512, but the target_feature(enable = "avx512f") etc are not stable, so the avx512 code would just use SSE loads, and that turns out to be worse. So, avx2 is the best we can do right now (until we start looking at avx512 seriously).