Skip to content

Comment on Using SIMD for Parallel Processing in Rustparent

Comments

The function has to be made pub so it doesn't get optimized out as unusued private function.

Godbolt is a better choice for looking at asm anyway. https://rust.godbolt.org/z/3Y9ovsoz9

Narrator: "The code did not, in fact, auto-vectorise."

(There's only addsd/movsd instructions, which are add/move scalar-double; we want addpd/movpd which are add/move packed-double in vectorised code.)

Ah, that worked, thanks!

Although I can now see why he didn't include the output directly.

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.