Github user franz1981 commented on the issue:
https://github.com/apache/activemq-artemis/pull/1757
@michaelandrepearce
> What else you think we need do to merge all this? Be good to get it in,
before we crack on with the next thing tbh.
I will love to provide some basic unit tests for the `*Pool`s
implementation to strengthen it: anyway I can push the tests as a separate PR
too...
> On checking duplicate it seems as efficient as it can be as in its just
an object wrapping around the existing bytebuf, with new index's effectively,
so i reverted that change
Oh, just reading the name `duplicate` I've thought Netty was performing a
copy :(...BTW if we avoid that "sliced" allocation on an hot path I'm sure it
will help GC when many producers/consumers will be involved, but if we will let
it as it is I'm happy anyway considering that is not 100% the purpose of this PR
> (i get same perf result btw on load testing, its the pooling thats giving
the perf boost)
Good to know, if it will turn to be effective as it seems I will provide
something similar into `proton-j` too
> I think real solution is to get those core properties like routingtype to
the core message level, as such then they will be more efficient and
lazyproperties can be truely honoured server side and not decoded at all.
Agree, that's the real deal indeed
---
