Cross product optimization

[rapid tartan]

that looks like code compiled without optimizations

which flag did you use to generate an optimized build?

[hallow quartz]

^^ if you're using build-exe (note that this will be what godbolt is using if you're doing this in godbolt) you want -OReleaseFast, whereas for zig build, if you're using standardOptimizeOption, you want -Doptimize=ReleaseFast

[woeful isle]

yep, this is with ReleaseFast https://godbolt.org/z/Eb4MrE3P3

[rapid tartan]

vpermd doesnt seem to have an xmm equiv https://www.felixcloutier.com/x86/vpermd:vpermw