cetra3 commented on PR #11505: URL: https://github.com/apache/arrow-rs/pull/11505#issuecomment-6104740993
Benchmark summary for `variant_get_unshredded_object_path_262k_rows`: | CPU | change | |---|---| | Neoverse V2 (bot, `c4a`) | +6% (slower) | | Intel Xeon 8581C (bot, `c4`, `arch: amd64`) | −5% | | Apple Silicon (local) | −3 to −5% | So it looks like this regresses on the ARM runner, but it's about a 5% improvement on x86 and macOS. The ARM result reproduces across several runs, including an A/A run (flat) and a run with base and PR swapped, so it doesn't look like runner noise. We're looking into why Neoverse V2 behaves differently. -- 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]
