types to use for mask parameters

[jackmott]

avx and avx512 have a lot of mask parameters.

For a function like _mm256_blendv_pd what should the type be for the mask parameter?
I've currently got it as:

pub fn _mm256_blendv_pd(a: f64x4, b: f64x4, mask: f64x4) -> f64x4 {
    unsafe { blendvpd256(a, b, mask) }
}

But should this be i64x4? __m256i?

[BurntSushi]

This is an interesting one. I'm not exactly sure. Both Clang and the Intel documentation say the mask should be a __m256d which is equivalent to f64x4. The documentation certainly seems to treat the mask parameter as if it were a normal bit vector, but the types clearly indicate otherwise.

Unless there's a compelling reason to the contrary, we should stick to the vendor interface. I will note this case down and mention in the RFC though. Thanks!

[alexcrichton]

@p32blo had a question about some mask types, @BurntSushi you may know better than I maybe?

[alexcrichton]

I'm going to close this now due to #263