[jdk26] RFR: 8375130: [BACKOUT] Scalability issue when submitting virtual threads with almost empty tasks [v2]

[Viktor Klang]

On Wed, 14 Jan 2026 08:13:02 GMT, Alan Bateman <alanb at openjdk.org> wrote:

>> This PR is for the jdk26 branch, not main line.
>> 
>> The changes in [JDK-8360046](https://bugs.openjdk.org/browse/JDK-8360046) improve the scalability of virtual threads but introduce a subtle issue with signal filtering and a regression in some SPECjbb2015 runs. [pull/28797](https://git.openjdk.org/jdk/pull/28797) is in progress for main line to address these issues.
>> 
>> For the jdk26 branch, the proposal is to backout the changes to ForkJoinPool from JDK-8360046 but leave the small change to setParallelism to trigger worker activation and the small change to VirtualThread to submit(ForkJoinTask) consistently.  As part of the change, the acquire fence change to address [JDK-8372835](https://bugs.openjdk.org/browse/JDK-8372835) are proposed to be included, to avoid bringing back an issue that was fixed by JDK-8360046. So a somewhat complicated story due to jdk26 being in RDP1.
>> 
>> 
>> Testing: test1-5, repeat testing of java/util/concurrent and java/lang/Thread/virtual tests.
>
> Alan Bateman has updated the pull request incrementally with one additional commit since the last revision:
> 
>   Change setParallelism to signalWork

Marked as reviewed by vklang (Reviewer).

-------------

PR Review: https://git.openjdk.org/jdk/pull/29187#pullrequestreview-3659639975