spinscale opened a new pull request, #15729: URL: https://github.com/apache/lucene/pull/15729
This change requires some explanation. While trying to understand how doc value skippers work, I noticed on skipper generation `input.prefetch()` is always called - essentially translating to an madvise willneed, indicating that this will be needed in the future. While for most other usage in the doc values producer this seems correct, in the case of the doc value skipper there there is the possibility to exit early without doing any disk accesses for min/max value and doccount, so the prefetch is wasted. My understanding of the madvise usage is that it mostly makes sense if we are sure to read ahead, which is not the case for doc value skippers. In addtion I also don't think it makes sense to move prefetching within the `advance()` call, as I suppose those are called when the data is actually needed, so prefetching would not help. -- 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]
