RFR: 8343704: Bad GC parallelism with processing Cleaner queues [v7]

Fri Nov 15 07:34:42 UTC 2024

On Thu, 14 Nov 2024 17:26:31 GMT, Aleksey Shipilev <shade at openjdk.org> wrote:

> > That is probably correct. I was however thinking that it would only be pooled asymmetrically as some type of hystereses. So you pool when you remove a node (switch the head) and keep it far an arbitrary amount of removals. So it would only really waste memory for cleaners that have this behaviour that they keep adding and removing cleanable around a NODE_CAPACITY boundary.
> 
> I have been playing with [8343704-node-cache.txt](https://github.com/user-attachments/files/17753986/8343704-node-cache.txt) -- is that what you had in mind?

Yes. It amortises the allocation over at least NODE_CAPACITY inserts, instead of 1 in the worst case. 

I have very little experience how this plays out in practice, nor how the cleanable are used. This was purely an observation based on seen a symmetrical grow/shrink behaviour of some resource.

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

PR Comment: https://git.openjdk.org/jdk/pull/22043#issuecomment-2478136490