RFR: 8322732: ForkJoinPool may underutilize cores in async mode
Viktor Klang
vklang at openjdk.org
Sat May 11 23:26:07 UTC 2024
On Tue, 7 May 2024 22:50:18 GMT, Doug Lea <dl at openjdk.org> wrote:
> This set of changes address causes of poor utilization with small numbers of cores due to overly aggressive contention avoidance. A number of further adjustments were needed to still avoid most contention effects in deployments with large numbers of cores
src/java.base/share/classes/java/util/concurrent/ForkJoinPool.java line 599:
> 597: * Deactivation. When method scan indicates (twice) that no tasks
> 598: * are found by a worker, it tries to deactivate()), giving up
> 599: * (and rescanning) on ctl contention. To avoid missed signals
Suggestion:
* (and rescanning) on "ctl" contention. To avoid missed signals
src/java.base/share/classes/java/util/concurrent/ForkJoinPool.java line 880:
> 878: * updated by owners from those most commonly read by stealers or
> 879: * other management. For class WorkQueue, an embedded padded
> 880: * region segregates fields (all decalred as "int") most heavily
Suggestion:
* region segregates fields (all declared as "int") most heavily
-------------
PR Review Comment: https://git.openjdk.org/jdk/pull/19131#discussion_r1597520813
PR Review Comment: https://git.openjdk.org/jdk/pull/19131#discussion_r1597520717
More information about the core-libs-dev
mailing list