RFR: 8343704: Bad GC parallelism with processing Cleaner queues [v4]

[Aleksey Shipilev]

On Thu, 14 Nov 2024 00:30:27 GMT, Brent Christian <bchristi at openjdk.org> wrote:

>> I pushed the array-based implementation, but ran out of time to properly test it. I'll circle back to it tomorrow. Review if you can, but it will probably have some touchups later.
>
> Speaking of testing, with this level of change, perhaps a new (white box?) regtest is warranted.

I see that existing `Cleaner` tests are quite rich already. Array-based implementation does rise question about testing the resizes, that's obvious. But I have discovered that copying arrays is not really that good in this scenario which balloons cleaner lists and shrinks them rapidly in our reproducers. So I have reimplemented the whole thing as the linked list of arrays, which makes capacity management code much less problematic. The new asserts around that code should IMO test enough with existing tests.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/22043#discussion_r1841914579