Closed MichaReiser closed 2 months ago
Name | Link |
---|---|
Latest commit | 74a4de43dc5b41a23114f889b91f3f730269c369 |
Latest deploy log | https://app.netlify.com/sites/salsa-rs/deploys/66c6e90937271f0008dbd2be |
Comparing MichaReiser:reduce-set-cloning
(74a4de4) with master
(f608ff8)
✅ 8
untouched benchmarks
Okay, that's not what I expected. I do see a 1-2% perf improvement in local hyperfine benchmarks
Okay, that's not what I expected. I do see a 1-2% perf improvement in local hyperfine benchmarks
Using IndexSet::into_boxed_slice
seems to really hurt performance (at least in codspeed). I reverted that change.
I see about a 1% performance improvement on the Ruff benchmarks (second last benchmark). I hoped for more, but that's how it is with perf ;)
It seems that codspeed struggles a bit with matching the salsa frames. I suspect it doesn't help that too many frames have the same name :laughing:
A small performance improvement to avoid cloning hash sets where we can consume the value instead.