Github user franz1981 commented on the issue:
https://github.com/apache/activemq-artemis/pull/1757
@michaelandrepearce
> Do we want to look address this also (NOT for this PR but something id
like to discuss with you), as its not helping matters on the front of GC, as
obviously avoidance is better?
Sure! I think there is a great margin of improvement GC wise (and maybe
latency/throughput too honestly) thanks to these changes, well done atfinding
this!
> This seems a little heavy handed, when could just get current reader and
writer index before decoding, then set them back after.
Looking at the current code I'm not seeing evident reasons why we can't
avoid that duplication: it is a WIN WIN IMO! Probably this optimization could
land into this same PR, but it is needed to change a lil' bit the PR
description...
This entire PR looks to me as a whole great job of garbage reduction and
I'm pretty about about it :+1:
---
