2010/12/3 Cedric Pradalier <cedric.pradalier@mavt.ethz.ch>

>
> On 12/03/10 10:18, Konrad Banachowicz wrote:
>
> I don't think that shm is right way of communication for ROS.
> It is extremely low-level approach to the problem and it's possibly
> duplicate existing in system messaging like UNIX domain socket or POSIQ MQ.
> Additional overhead for providing reliable communication would be difficult
> to achieve and would have significant impact on performance.
>
>
> True as well...
>
>
> In my opinion using UNIX domain socket would be much more beneficial to ROS
> and much simpler to implement.
>
>
> Good, it should be pretty trivial to implement within the framework.
>
> Do you have experience on how the performance compares with TCP over
> loopback?
>
I have done some benchmarks of these in the past and as far as I remember in
some cases it was more efficient.
But I don't have exact result now.
Look there :
http://lists.freebsd.org/pipermail/freebsd-performance/2005-February/001143.html


>
> I'm beginning to wonder if the biggest quality of the sharedmem plugin will
> only be to make sure nobody (else) loses time to implement it...
> Thanks for the feedback
>
>
> --
> Dr. Cedric Pradalierhttp://www.asl.ethz.ch/people/cedricp
>
>
> _______________________________________________
> ros-users mailing list
> ros-users@code.ros.org
> https://code.ros.org/mailman/listinfo/ros-users
>
>