wide

wide copied to clipboard

shssoichiro

Adds conversion functions from integers to floats

3

This PR adds four conversion functions from integers to floats: `f32x4::from_i32x4` `f32x8::from_i32x8` `f64x2::from_i32x4` `f64x4::from_i32x4`

Mottl

Adds boolean methods to u32x4 and u32x8

3

Closes #160

andyquinterom

Add methods from i32x4 for u32x4

2

I see there are some methods like all or any that are implemented for i32x4 but not u32x4. I could go ahead and implement them unless I am missing something.

andyquinterom

Nonoptimal `CmpLt`

1

https://github.com/Lokathor/wide/blob/729877b9a5458efb066c2731661ef0723971fba0/src/i32x8_.rs#L275 Much faster will be just reverse `CmpGt` ``` Self { avx2: cmp_gt_mask_i32_m256i(rhs.avx2, self.avx2) } ```

Melirius

Making new const simplified a bunch of code and allows users of this library to define their own constants without having to do anything hacky. It does mean that it...

mcroomp

Combination of `i8x16::move_mask()` and `.trailing_zeros()` is a footgun

3

For example, consider this code, extracted from one of my libraries: ``` for &group in samples.array_chunks::() { let mask = predicate(i8x16::new(group)); offset += mask.move_mask().trailing_zeros() as usize; if mask.any() { break...

whitequark

Add implementations for all the intrinsics @Melirius might be interested

mcroomp