bhaskargurram-ai commented on PR #11549: URL: https://github.com/apache/arrow-rs/pull/11549#issuecomment-6292963170
> @bhaskargurram-ai I ran the `decimal*` writer benchmarks on a dedicated aarch64 VM (3 interleaved rounds, pinned to fixed cores). I could not reproduce the slowdowns: every benchmark was within ±2% of base, and the decimal hot loops compile to identical assembly. I think the CI results were noise plus codegen lottery. The version with shared write paths is less code and the right long-term architecture. Two options: > > 1. (My preference) Split the write-path changes into a precursor PR. It would be a pure refactor, so we can review and benchmark it on its own. This PR then only adds `ArrowSbbf`, and both PRs are small. > 2. Add the shared write paths back here and rerun the benchmarks. @adriangb Thanks for running those, good to know it was noise. I’ll put the shared write paths back in this PR (keeping the every-type test as a guard against drift) and ping you for a benchmark run once it’s pushed. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
