Divisibility is not properly optimized on aarch64

[ArhanChaudhary]

The x86 architecture recognizes how to do efficient divisibility testing (https://lomont.org/posts/2017/divisibility-testing/) but not aarch64: https://godbolt.org/z/MhbKvvc4b

It's possible this is not a portable-simd problem and instead somewhere else down the line.

Meta

rustc --version --verbose:

rustc 1.87.0-nightly (30f168ef8 2025-03-05)
binary: rustc
commit-hash: 30f168ef811aec63124eac677e14699baa9395bd
commit-date: 2025-03-05
host: x86_64-unknown-linux-gnu
release: 1.87.0-nightly
LLVM version: 20.1.0

crate version in Cargo.toml:

[dependencies]
stdsimd = 

(I don't know how to check on compiler explorer)

[programmerjake]

we intentionally scalarize div/rem on aarch64 to workaround backend bugs: #422 (comment)

@calebzulawski I didn't see a LLVM issue for that backend bug after some preliminary searching... do you want to create one?

[ArhanChaudhary]

Cool, I'll close this in favor of the aforementioned.

[programmerjake]

Cool, I'll close this in favor of the aforementioned.

#422 isn't an issue, it's a pull request for a completely different thing that was already merged...so this optimization issue is very much not solved, so I'm reopening it.