RFR: 8373118: Test java/lang/Thread/virtual/Starvation.java timed out [v5]
Viktor Klang
vklang at openjdk.org
Mon Dec 15 11:34:41 UTC 2025
On Sun, 14 Dec 2025 13:06:46 GMT, Doug Lea <dl at openjdk.org> wrote:
>> Changes signal filtering to avoid possible starvation
>
> Doug Lea has updated the pull request incrementally with one additional commit since the last revision:
>
> Adjust runworker for previous update
src/java.base/share/classes/java/util/concurrent/ForkJoinPool.java line 1865:
> 1863: long ac = (c + RC_UNIT) & RC_MASK, nc;
> 1864: int sp = (int)c, i = sp & SMASK;
> 1865: if ((short)(c >>> RC_SHIFT) >= pc)
I'm wondering if it might make sense to re-read `parallelism` on each new loop, since if we have a parallelism resize *during this loop* we want to exit early?
Suggestion:
for (long c = ctl;;) {
WorkQueue[] qs = queues;
long ac = (c + RC_UNIT) & RC_MASK, nc;
int sp = (int)c, i = sp & SMASK, pc;
if ((short)(c >>> RC_SHIFT) >= (pc = parallelism))
-------------
PR Review Comment: https://git.openjdk.org/jdk/pull/28797#discussion_r2619038375
More information about the core-libs-dev
mailing list