xiangfu0 commented on PR #19264: URL: https://github.com/apache/pinot/pull/19264#issuecomment-6136137688
Rebenchmarked with the duplicate server response removed from the harness. The current BASE path is 4.0–8.6× faster than master with 75–88% lower allocation on the four low-cardinality workloads; full low/high result digests match master. | Query | Master ms/op | BASE ms/op | Speedup | Allocation reduction | |---|---:|---:|---:|---:| | ROLLUP 2 | 8.586 ± 0.315 | 1.003 ± 0.086 | 8.56× | 87.6% | | ROLLUP 3 | 13.267 ± 1.724 | 2.727 ± 0.123 | 4.87× | 74.5% | | CUBE 3 | 20.287 ± 1.688 | 3.225 ± 0.518 | 6.29× | 77.4% | | GROUPING SETS 3 | 11.617 ± 2.664 | 2.895 ± 0.638 | 4.01× | 74.7% | Retained a small pack-once optimization for the single-value packed expansion/fallback path. Low-cardinality paired batches improved 14–26%; high-cardinality timing is inconclusive because the trial ordering reversed the result. BASE execution and safety gates are unchanged. JDK25, shared Mac, 50×15k rows, 2 forks, 5×1s warmup / 5×2s measurement. Errors are JMH estimates; these are local SSE measurements. Existing packed-path integration test: passed (1 test, zero failures/errors/skips); style/license checks passed. -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
