RFR: 8325754: Dead AbstractQueuedSynchronizer$ConditionNodes survive minor garbage collections
Alan Bateman
alanb at openjdk.org
Thu Feb 22 11:59:54 UTC 2024
On Wed, 21 Feb 2024 15:01:15 GMT, Viktor Klang <vklang at openjdk.org> wrote:
> More aggressively breaking chains in order to prevent nodes promoted to older generations standing in the way for collecting younger nodes. I decided that it was most efficient to add this logic to the else-branch of updating the firstWaiter and lastWaiter.
>
> There's a race with unlinkCancelledWaiters() but according to @DougLea it should be a benign one.
>
> There's a performance impact of this, but as it is a plain write, and one to null at that, it should be acceptable.
If the race with unlinking non-waiting nodes is benign then this should be okay.
-------------
Marked as reviewed by alanb (Reviewer).
PR Review: https://git.openjdk.org/jdk/pull/17950#pullrequestreview-1895672384
More information about the core-libs-dev
mailing list