[
https://issues.apache.org/jira/browse/DISPATCH-1352?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16854746#comment-16854746
]
Francesco Nigro edited comment on DISPATCH-1352 at 6/3/19 4:00 PM:
-------------------------------------------------------------------
[~kgiusti] If the POC will work "maybe" we have the chance to do the same while
allocating the original qd_message_pvt_t too.
Annotations are being extracted by fields AFAIK: if they have a small-ish size
(< 512 bytes) maybe it would be more beneficial to just copy them into the ones
embedded into qd_message_pvt_t instead of "moving" into qd_message_pvt_t.
It would allow the allocated qd_buffer_t from the field to be short living and
quick released on the same allocating thread, improving the effectiveness of
the thread local qd_buffer_t pool.
was (Author: nigro....@gmail.com):
[~kgiusti] If the POC will work "maybe" we have the chance to do the same while
allocating the original qd_message_pvt_t too: annotations are being extracted
by fields AFAIK: if they have a smal size (< 512 bytes) maybe it would be more
beneficial to just copy them into the ones embedded into qd_message_pvt_t: it
would allow the qd_buffer_t allocated from the field to be short living on the
same allocating thread, improving the effectiveness of the thread local
qd_buffer_t pool.
> qd_buffer_list_clone cost is dominated by cache misses
> ------------------------------------------------------
>
> Key: DISPATCH-1352
> URL: https://issues.apache.org/jira/browse/DISPATCH-1352
> Project: Qpid Dispatch
> Issue Type: Improvement
> Components: Routing Engine
> Affects Versions: 1.7.0
> Reporter: Francesco Nigro
> Priority: Major
> Attachments: screenshot-1.png, screenshot-2.png
>
>
> qd_buffer_list_clone on qd_message_copy for
> qd_message_pvt_t.ma_to_override/ma_trace/ma_ingress is dominated by cache
> misses costs:
> * to "allocate" new qd_buffer_t
> * to reference any qd_buffer_t from the source qd_buffer_list_t
> Such cost is the main reason why the core thread is having a very low IPC (<
> 1 istr/cycle) and given the single threaded nature of the router while
> dealing with it, by solving it will bring a huge performance improvement to
> make the router able to scale better.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@qpid.apache.org
For additional commands, e-mail: dev-h...@qpid.apache.org
