hsivonen / hsivonen/encoding_rs
Experiment with load_select/store_select for tail remaining over stride size
Nobody has claimed this yet.
- Dominant language
- Rust
- Stars
- 464
- Forks
- 70
- Avg merge
- 2d 4h
- Merged PRs (30d)
- 2
Description
Currently, the tail that is left over after SIMD strides is handled using scalar iteration.
Experiment with https://doc.rust-lang.org/std/simd/type.u16x8.html#method.load_select and https://doc.rust-lang.org/std/simd/type.u16x8.html#method.store_select to see if they are faster with aarch64 NEON and AVX2, which don't have as good masked operations as AVX-512.
Three cases:
- Full SIMD load, masked store.
- Masked SIMD load, full store.
- Masked load and masked store.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start by locating the scalar tail handling after SIMD strides and review the Rust SIMD load_select and store_select documentation linked in the issue. Compare full-load/masked-store, masked-load/full-store, and masked-load/masked-store on aarch64 NEON and AVX2. Done means determining whether any approach is faster than scalar iteration on those targets.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- rust
- Domain
- performance
- Issue type
- Refactor
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Quiet
- Clarity
- Mostly clear
- Newbie friendliness
- 48/100