[
https://issues.apache.org/jira/browse/ARTEMIS-1722?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16384102#comment-16384102
]
ASF GitHub Bot commented on ARTEMIS-1722:
-----------------------------------------
Github user tabish121 commented on the issue:
https://github.com/apache/activemq-artemis/pull/1918
@franz1981 @clebertsuconic I've sent an alternate PR #1928 which performs
the send without any buffer copy unless necessary due to message redelivery or
existence of original delivery annotations. Proton will be copying those bytes
anyway so an intermediate isn't really needed on the common path.
It could be argued that you could further change things beyond what I did
and just keep a single ReadableBuffer instance around in the consumer and give
it the outbound buffer to send as the proton send method will take that
readable buffer and call get(array, offset, length) on it to get a copy which
means netty should perform the most efficient write.
> Reduce pooled Netty ByteBuf usage
> ---------------------------------
>
> Key: ARTEMIS-1722
> URL: https://issues.apache.org/jira/browse/ARTEMIS-1722
> Project: ActiveMQ Artemis
> Issue Type: Improvement
> Components: AMQP, Broker
> Affects Versions: 2.4.0
> Reporter: Francesco Nigro
> Assignee: Francesco Nigro
> Priority: Major
> Fix For: 2.5.0
>
> Original Estimate: 3h
> Remaining Estimate: 3h
>
> Pooled Netty Heap ByteBuf could be avoided thanks to smart pooling on hot
> paths, improving scalability and memory footprint.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
