[zeromq-dev] Efficiency for very few short messages

[dan smith]

Claudio,

Thanks for your answer.

When it comes to 8, the timing changes randomly, sometime times is less
than the time needed to solve 80 equations. I did 4 runs again, the times
were between 115000 and 125000. This program is very simple and each and
every thread does the very same thing using the very same code, no
difference between the threads. I do know how to debug that further. The
 main thread just sends the pointer and the worker thread solves it after
receiving it, 3 lines are relevant.The thread is a basic windows thread
like in inproc_lat.


On Mon, Jan 21, 2013 at 2:13 AM, Claudio Carbone <erupter at libero.it> wrote:

> dan smith <dan25311 at gmail.com> wrote:
> >80 times: 64614micros and 134429micros, the serial is already faster.
> >
> >Going down to 8: 6345 and 328286...
> >
>
> There must be something wrong if it takes less to solve 80eqs than 8, no
> matter what.
> Why don't you save/print split times for each eq and each phase of the
> program?
> With 8 it isn't crazy to analyze the numbers.
>
>
> Claudio
> -- Sent from my ParanoidAndroid Galaxy Nexus with K-9 Mail.
>
> _______________________________________________
> zeromq-dev mailing list
> zeromq-dev at lists.zeromq.org
> http://lists.zeromq.org/mailman/listinfo/zeromq-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.zeromq.org/pipermail/zeromq-dev/attachments/20130121/7a4f5b71/attachment.htm>