[
https://issues.apache.org/jira/browse/LUCENE-9165?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17022565#comment-17022565
]
Robert Muir commented on LUCENE-9165:
-------------------------------------
Yes, I agree with you about that.
But I think LUCENE-9160 fixed the bugs there: like i said it reduced load avg
while running tests substantially for me (more than 3x). Default
CICompilerCount on my jvm is 3... so we were really using (processorCount / 2)
* 4 due to the three extra always-busy C2 threads which was not good.
There are still GC threads and sometimes tests themselves use threads which can
cause the load to be more than the number of test jvms, but its no longer
absurdly more.
> change generate-defaults.gradle not to cap testsJvms at 4
> ---------------------------------------------------------
>
> Key: LUCENE-9165
> URL: https://issues.apache.org/jira/browse/LUCENE-9165
> Project: Lucene - Core
> Issue Type: Improvement
> Reporter: Robert Muir
> Priority: Major
>
> {code}
> def cpus = Runtime.runtime.availableProcessors()
> def testsJvms = (int) Math.max(1d, Math.min(cpus * 0.5d, 4))
> {code}
> Dividing by 2 is still good (hyperthreading is still a thing), but the
> artificial cap of 4 is a bottleneck if the hardware has more available.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org
