uschindler commented on PR #13907: URL: https://github.com/apache/lucene/pull/13907#issuecomment-2413307576
In the original code when I added madvise for the first time, I had an exclusion on the mapping function between ReadAdvise and the platform constant a NULL mapping for NORMAL. The later code then excluded the call when the mapping function returned NULL. The latter check is still there, so maybe we can clean this up, too. The updated code for compound files looks fine, although why should the madvise only called when read advice is NORMAL. The linux kernel has no problem when you open the CFS file with SEQUENTIAL and then you change the read advice to RANDOM. I was also thinking of generally supporting this in the API for merging. On the mailing list there was a complaint about merging being slower for docvalues and vectors, because those use RANDOM. So we could think about changing the readAdvice before merging for all files to SEQUENTIAL and restore it to the original value later. This could be done by adding an IndexInput method (similar to the new slice() method with advise). -- 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: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
