[
https://issues.apache.org/jira/browse/QPID-6662?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14730998#comment-14730998
]
Lorenz Quack commented on QPID-6662:
------------------------------------
Today, I stumbled over the QpidByteBuffer#allocateDirectFromPool().
I find the name a bit unfortunate because:
* It might not allocate from a pool (if size > maxPooledBufferSize)
* QpidByteBuffer#allocateDirect() may allocate from a pool as well.
* It does not describe what the method is about (slicing a larger buffer to
efficiently handle smaller allocations)
IMHO, the method should go away and its functionality should be handled
transparently by QpidByteBuffer#allocateDirect(). There could be a threshold
(e.g., 0.25 * _maxPooledBufferSize) and for calls smaller than that it could
use a cached slicing buffer and for larger ones it could keep its current
behaviour.
> Use direct byte buffers
> -----------------------
>
> Key: QPID-6662
> URL: https://issues.apache.org/jira/browse/QPID-6662
> Project: Qpid
> Issue Type: Improvement
> Components: Java Broker
> Reporter: Keith Wall
> Fix For: qpid-java-6.0
>
>
> To improve performance of the Broker, direct ByteBuffers should be passed
> from the transport directly to the store, minimising copying whenever
> possible, and vice versa.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
