ClSlaid commented on PR #9755: URL: https://github.com/apache/arrow-rs/pull/9755#issuecomment-4605831565
Pushed the rebased/squashed branch. Current head: `ae36adaf2`. Benchmark patch split out for independent CI/testing: #10050. This PR still keeps the benchmark patch so reviewers can run it directly from this branch. Local full-file smoke benchmark: - baseline: `apache/main` + benchmark-only patch from #10050 (`e9ec9c15e`) - feature: this PR head (`ae36adaf2`) - command: `cargo bench --features=arrow,async,test_common,experimental,object_store --bench coalesce_kernels -- --sample-size 10 --warm-up-time 0.2 --measurement-time 0.5` - scope: 178/178 `coalesce_kernels` cases matched on both sides Representative filter results, Criterion middle estimate: | case | baseline | this PR | ratio | |---|---:|---:|---:| | `single_binaryview max_len=8, nulls=0, sel=0.01` | 2.172 ms | 1.572 ms | 0.72x | | `mixed_binaryview max_len=8, nulls=0, sel=0.01` | 1.776 ms | 857.9 us | 0.48x | | `mixed_binaryview max_len=20, nulls=0, sel=0.01` | 2.349 ms | 2.342 ms | 1.00x | | `single_utf8view max_len=8, nulls=0, sel=0.01` | 2.161 ms | 1.557 ms | 0.72x | | `mixed_utf8view max_len=8, nulls=0, sel=0.01` | 1.792 ms | 844.5 us | 0.47x | Full-matrix smoke summary: filter 104 cases, median 0.99x / average 0.92x; take 74 cases, median 0.99x / average 0.99x. The short local run matches the earlier signal: low-selectivity fully-inline View cases improve substantially, while the `max_string_len=20` fallback case is around parity. I would still rely on #10050 / the benchmark runner for the final low-noise signal. -- 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]
