[
https://issues.apache.org/jira/browse/CASSANDRA-10379?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15104977#comment-15104977
]
Sylvain Lebresne commented on CASSANDRA-10379:
----------------------------------------------
For what it's worth, I'm not in favor of adding that flag as it sounds
potentially dangerous and the benefits don't sound that big. That is, I can
kind of trust us to not modify final field by reflection, but we depend on a
lot of libraries and as much as we would hope they don't do it either, it's
hard to be sure without spending a lot of investigative time and even then we
can't be absolutely guaranteed this won't change behind our backs.
Hopefully the JVM can indeed do a better safe job at this, but in the meantime,
I'd suggest closing this as "won't fix".
> Consider using -XX:+TrustFinalNonStaticFields
> ---------------------------------------------
>
> Key: CASSANDRA-10379
> URL: https://issues.apache.org/jira/browse/CASSANDRA-10379
> Project: Cassandra
> Issue Type: Improvement
> Reporter: Robert Stupp
> Assignee: Robert Stupp
> Fix For: 3.x
>
> Attachments: 10379.txt
>
>
> The JVM option {{-XX:+TrustFinalNonStaticFields}}, although experimental,
> seems to improve performance a bit without any code change. Therefore I
> propose to include it in {{cassandra-env.sh/psl}}.
> [cstar perf
> benchmark|http://cstar.datastax.com/graph?stats=a6e75018-5ff4-11e5-bf84-42010af0688f&metric=op_rate&operation=1_user&smoothing=1&show_aggregates=true&xmin=0&xmax=865.59&ymin=0&ymax=145568.5]
> The cstar test was run with 8u45.
> {noformat}
> JVM_OPTS="$JVM_OPTS -XX:+UnlockExperimentalVMOptions"
> JVM_OPTS="$JVM_OPTS -XX:+TrustFinalNonStaticFields"
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
