JDK-8322732 : ForkJoinPool may underutilize cores in async mode
  • Type: Bug
  • Component: core-libs
  • Sub-Component: java.util.concurrent
  • Affected Version: 22
  • Priority: P4
  • Status: Resolved
  • Resolution: Fixed
  • OS: generic
  • CPU: generic
  • Submitted: 2023-12-24
  • Updated: 2024-09-06
  • Resolved: 2024-06-05
The Version table provides details related to the release that this issue/RFE will be addressed.

Unresolved : Release in which this issue/RFE will be addressed.
Resolved: Release in which this issue/RFE has been resolved.
Fixed : Release in which this issue/RFE has been fixed. The release containing this fix may be available for download as an Early Access Release or a General Availability Release.

To download the current JDK release, click here.
JDK 23
23 b26Fixed
Related Reports
Duplicate :  
JDK-8328769 - ForkJoinPool trims worker threads too quickly after JDK-8319662
Duplicate :  
JDK-8321275 - java/util/stream/test/org/openjdk/tests/java/util/stream/CollectorsTest.java fails with jtreg timeout
Relates :  
JDK-8327854 - Test java/util/stream/test/org/openjdk/tests/java/util/stream/WhileOpStatefulTest.java failed with RuntimeException
Relates :  
JDK-8336707 - Contention of ForkJoinPool grows when stealing works
Description
Underutilization of available cores and poorer throughput  has been observed in some reports on Loom performance. There is not yet a standalone test case.
Comments
Peter: This is being tracked as https://bugs.openjdk.org/browse/JDK-8336707 . As it says there, it would be great to have help checking performance on different platforms.
30-07-2024
I find that after this commit our SPECjbb2015 critical-jOPS scores drop by 14%. I built JDK's from JDK-24 sources at commit 2c1b311f81319cee1af574526a91424c2577b78c (the one before this commit) and commit 789f704d9ab5aaf87193f508859c4c9a528d7779 (this commit), and ran SPECjbb2015 v1.04 in a multi-jvm configuration with 4 groups on 80 cores (all of the cores socket 0 of a 2P Altra), with each backend having a 31GiB heap. I also built a clone of JDK-24+8, and the performance is marginally (<0.5%) better than commit 789f704, but nothing to write home about. I have tested on Ampere Altra, but others have seen a similar critical-jOPS drop on Amazon G2. Anecdotally, more groups (e.g,. multi-jvm with 6 or 8 groups, or composite with many backends) shows more degradation from the change, but I have not run those configurations, yet, so I won't quote performance numbers.
29-07-2024
[~dl] the "backport" links in this JBS issue are not appropriate - those issues are not backports. I assume they should just be "relates to" links.
18-07-2024
Changeset: 789f704d Author: Doug Lea <dl@openjdk.org> Date: 2024-06-05 13:02:15 +0000 URL: https://git.openjdk.org/jdk/commit/789f704d9ab5aaf87193f508859c4c9a528d7779
05-06-2024
A pull request was submitted for review. URL: https://git.openjdk.org/jdk/pull/19131 Date: 2024-05-07 22:50:18 +0000
07-05-2024
Trial updates are at https://github.com/DougLea/jdk/tree/JDK-8322732
24-12-2023