uschindler commented on PR #12632: URL: https://github.com/apache/lucene/pull/12632#issuecomment-1772952661
Hi, > The jvm already has these. For example a user can set max vector width and avx instructiom level already. I assume that avx 512 users who are running on downclock-susceptible CPUs would already set flags to use only 256bit vectors. So I am afraid of adding our own flags that conflict with those. The problem is that by default JVM enables those flags and then those machines get slow when they use Lucene and support cases are openend at Elasticsearch/Opensearch stuff. So my wish would be to have that opt-in. My idea was to add a sysprop like the MMap ones to enable/disbale manually or disable unsafe unmapping. In that case the sysprop would enable the AVX 512 bit case. It won't conflict with manual AVX=2 JVM override setting (because then the preferredBitSize=256 and our flag is a noop). -- 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
