Speed up factor::numeric::gcd#1563
Conversation
|
Can someone rerun |
The GHA jobs will sometimes crash for various, seemingly chaotic, reasons. If the crashing job is just style, I'd ignore it and be about my day. For other jobs, you can use an amended commit (with no changes;
Different platforms expose/use different code. |
I'm aware, but that doesn't affect the code formatting and linting tools. |
I understand your point, but the "Style" check does frequently produce different warnings on different platforms (probably different compiler warnings; see "Annotations" for https://github.com/uutils/coreutils/actions/runs/177424066). |
The function had to be made `pub`, this is a [known limitation] of Criterion. [known limitation]: https://bheisler.github.io/criterion.rs/book/user_guide/known_limitations.html
Also add a property-based test against the Euclidean implementation. numeric::gcd got ~50-65% faster, according to criterion. The effect on the overall system is small, but later PRs will use a lot more GCD computations.
Exciting :( Anyhow, this discussion is probably way out of scope for this PR, but I should mention I did rebase the branch now that #1554 landed. :) |
|
Build failed on AppVeyor (Rust stable, Windows x64, GNU ABI) because the DNS resolution for |
7 to 10% faster, according to criterion.
|
I made |
2 participants
factor::numeric::gcd.~50-65% faster (on the GCD computations themselves)