[
https://issues.apache.org/jira/browse/DISPATCH-1343?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16847779#comment-16847779
]
ASF GitHub Bot commented on DISPATCH-1343:
------------------------------------------
franz1981 commented on issue #510: DISPATCH-1343 Faster qd_buffer and
qd_message allocations
URL: https://github.com/apache/qpid-dispatch/pull/510#issuecomment-495730650
I've spent some time looking at the perf of this changes and it seems that
the new way to allocate qd_buffer without issuing a full memory barrier is
slower then the original one, because I was expecting the latter to use C11
variant of sys_atomic_init in atomic.h, while instead it was using the GNU
version.
The point is that some of atomic operation doesn't seem correct to me from a
memory model point of view and that's why the original version (that was just
doing a plain store!) was faster then this one, that was issuing a compiler
barrier for correctness.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
> Faster qd_buffer and qd_message allocations
> -------------------------------------------
>
> Key: DISPATCH-1343
> URL: https://issues.apache.org/jira/browse/DISPATCH-1343
> Project: Qpid Dispatch
> Issue Type: Improvement
> Affects Versions: 1.7.0
> Reporter: Francesco Nigro
> Priority: Major
>
> qd_buffer and qd_message allocations are using heavyweight initialization
> mechanisms for atomic ints fields, unnecessary for both atomicity and
> visibility purposes while allocating.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
