goankur commented on PR #16145: URL: https://github.com/apache/lucene/pull/16145#issuecomment-6227383866
@michaeljmarshall I ran f6b3392 through the same setup, as is and with one change: every file starts confident (`new AtomicInteger(N)` in `PrefetchBackoff`). | | f6b3392 | + start confident | |---|---|---| | Hot luceneutil, AndHighLow vs main | −18.0% | −1.2% | | Hot, query types slower (p < 0.05) | 15 | 1 (main vs main: 0) | | Cold start, AndHighLow (arm takes cold JVM / main takes it) | −16.8% / −19.4% | −1.2% / −2.1% | | Rerank p99, 10 GB cgroup, 2 reps | 17.3 / 17.2 ms | 17.0 / 17.0 ms | The seeded counter works well: the shared `nextSeed()` increment doesn't show up in any of these. The remaining regression is the cold start, and starting confident removes it without a cost, even when every file opens cold. So I'd suggest keeping f6b3392 as it is, with just that one-line change. Branches, scripts and raw data: [followup-f6b3392](https://github.com/goankur/lucene/tree/pr16145-eval/results/followup-f6b3392). -- 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]
