# Portable SIMD 2022-04-25 Meeting

###### tags: `Portable SIMD` `Minutes`

[Last Meeting] | [This Meeting's Chat] | [Next Meeting]

[Last Meeting]: https://hackmd.io/sNJHu0kXSe6s-PY8ymi95Q
[This Meeting's Chat]: https://rust-lang.zulipchat.com/#narrow/stream/257879-project-portable-simd/topic/Meeting.202022-04-25
<!-- [Next Meeting]: -->

## Agenda

- [Dot product example](https://github.com/rust-lang/portable-simd/pull/128) is just waiting on a rebase
- Jubilee will figure out after this what set of git incantations is needed to fix that up
- [Add dotproduct assembly docs](https://github.com/rust-lang/portable-simd/pull/270) hasn't been reviewed yet.
- Jacob reviewed it.
- [Move element-specific fn to traits](https://github.com/rust-lang/portable-simd/pull/278) hasn't been reviewed yet by Jubilee
- ...so Jubilee will review it, eh eh eh. *Jubilee poking Jubilee*
- [Libm work](https://github.com/rust-lang/portable-simd/pull/126) has two requests from Andy:
- `Simd::from(scalar)` instead of splat -> this would enable functions to take `Into<Simd>` everywhere.
- Much larger vector sizes f32x64 and f32x128 to gain more interleave.
- I don't remember if we had consensus on [Mask::cast](https://github.com/rust-lang/portable-simd/pull/251) or not
- we do actually!
- Jubilee will write a test
- Re-adding autosplats
- Jubilee: "no really I'll get around to it! <_<"
- concerns around using From::from vs. another trait for reifying "autosplat" logic?
- people will want to use "take a vector or a scalar" logic in their own code as well
- investigate using a new trait vs. extending an existing one?
- there might be something about libcore coherency that makes it harder to do in-repo
- Previous partially-resolved/unresolved questions:
- `simd::Float` + `Simd<T, N>` vs. `simd::Vector` + `SimdFloat<T,N>`
- ???
- What measure portability
- ???
- Conciseness in naming / avoiding repeating `lanes_` everywhere.
- ???

## Resolutions

0. それは秘密です。