Francesco Nigro

Feel free to review my branch and try it on a bigger server to see if you get the same or better improvement.

@anuraaga Thanks! I don't have a big machine to try it, but just a tuned i7 (without turbo boost, disabling frequency scaling etc etc) with 6 hw threads (12 with...

@anuraaga i have given a shot to the mentioned benchmark but it doesn't seem to stress the queue contention, because the XADD queue is unbounded and the test doesn't start...

@anuraaga In this case on JCTools we have an unbeateable queue for this: https://github.com/JCTools/JCTools/blob/master/jctools-core/src/main/java/org/jctools/queues/MpscCompoundQueue.java It uses a mesh of bounded queue to spread the contention: it should be the right...

@anuraaga > I tried the compound queue for that reason but couldn't find any noticable difference in our benchmarks on my laptop, so went with the simpler non-compound one And...

I've used a separate soft/lazy counter on https://github.com/franz1981/opentelemetry-java/tree/xadd_batch_span and I'm getting:  Consider that I don't know yet if the logic is sound

And for 20 threads (that doesn't makes sense on my poor 12 cores laptop):  Probably worth trying this one in an appropriate bigger machine IMO: consider that I didn't...

Great idea! Carbon could be interesting and relevant too!

@nbrady-techempower Hi! We (quarkus team) plan to send a PR to upgrade to the latest version (and other changes), but > We'll aim for mid-to-late May for an official preview...

Shared my thoughts on https://github.com/TechEmpower/FrameworkBenchmarks/issues/7475#issuecomment-1191789666 here https://github.com/TechEmpower/FrameworkBenchmarks/discussions/7358#discussioncomment-3200476 trying to both give my opinion but still convey that there are things, assuming what the purpose of the benchmark is, that should...