aneesh-db commented on PR #16700: URL: https://github.com/apache/lucene/pull/16700#issuecomment-6516134843
@epotyom I created a synthetic JMH benchmark with 1 million documents and 100 scattered rescore hits. Half the documents contain both query terms, while phrase selectivity varies from rare to common. This measures the benefit of avoiding phrase verification on documents outside the rescore window. On this benchmark, rare phrases matching 0.1% of term-cooccurring documents improved from 5.42 ms to 0.095 ms for exact matching (57×), and from 9.11 ms to 0.095 ms for slop-1 matching (96×). More frequent exact phrases improved 1.8×, while the single-term control remained unchanged within measurement noise. -- 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]
