diff --git a/content/blog/2026-04-19-tmil-25/index.md b/content/blog/2026-04-19-tmil-25/index.md index ada19b4..072f70e 100644 --- a/content/blog/2026-04-19-tmil-25/index.md +++ b/content/blog/2026-04-19-tmil-25/index.md @@ -107,6 +107,21 @@ A screenshot of Scrolled Quran. +## Fearless SIMD + +Fearless SIMD is our SIMD infrastructure library. +It provides a solid, portable way to write SIMD code that compiles to Wasm, AArch64, x86, and x86_64. + +We released [fearless_simd 0.4.0], a large step towards the library being generally useful. +The major improvement is the use of AVX2 intrinsics on supported platforms. +That fulfills the vision of being able to write code that works efficiently in a variety of SIMD vector widths — 128 bit on Neon, and 256 bits on x86. + +We are starting to see adoption outside Linebender, notably [PhastFT], a high performance FFT library. +Applications that can benefit from SIMD acceleration should consider fearless_simd, and we are open to feedback on how to improve it further. + +There will be SIMD discussion at RustWeek, both as part of the Linebender unconference section and hopefully a session in the all-hands. +Get in touch if you'd be interested. + ## Related ecosystem projects There's a lot of interesting activity surrounding Linebender, including projects that use Linebender crates, and adjacent bits of infrastructure. @@ -160,3 +175,6 @@ They are expected to continue in January, keep an eye on Zulip for details. [Gosub engine]: https://gosub.io/ [drafft-ink]: https://github.com/PatWie/drafft-ink [Takumi]: https://github.com/kane50613/takumi + +[fearless_simd 0.4.0]: https://github.com/linebender/fearless_simd/releases/tag/v0.4.0 +[PhastFT]: https://github.com/QuState/PhastFT