Francesco Nigro

In the cases of 0 -> 15 pending elements, but the most benefit come when there are none or very few . ArrayDeque is not like ArrayList that defer the...

One of the adverse effects of this PR is that, in order to reduce the contention on the newly introduced size field, it's performing a batch update *after* the entries...

Sorry to join late the discussion, but if you need a fast unbounded queue on JCTools that work better under contention, if it turns to be the issue with the...

@tabish121 @gemmellr I'm going to share 2 flamegraph to show how this PR change the way completions are processed with some shared connections and many sessions per-connection.

This PR is still in draft because of: - no JIRA yet: still waiting maintainers feedbacks - add a configuration option to cap the max numbers of threads used in...

These are the results of this change. Running a benchmark with: - 10 non-durable producer sessions - 5 sessions per connection - 2 connections handled by 2 separate threads (ie...

There're still something to be considered while using this processing model in case JMS Sessions are used on different threads AND share the same connection: - on `main` the Netty...

Oops my bad, I should have used a synchronous queue instead of a linked blocking queue and it should behave as expected I would run the bench again to be...

> Oops my bad, I should have used a synchronous queue instead of a linked blocking queue and it should behave as expected And I was wrong, despite the 0-capacity...

> I dont believe it will ever work the way you want with a simple queue of any size. You have to add hoops to manipulate it into behaving that...