[ros-users] Nodelet image transfer performance

Josh Faust jfaust at willowgarage.com
Mon Sep 20 17:34:56 UTC 2010


Ok, I think I have to put this down to differences in the machines (mine's a
4-core i7 desktop).  The CopyImageTest uses 0% CPU even at 5 copies, and I
know I'm looking at the right process because upping that to 50 makes it use
64%.

Josh

On Sat, Sep 18, 2010 at 11:04 AM, Joel Cartwright
<J.J.Cartwright at hw.ac.uk>wrote:

> Josh,
>
> I'm running the test on a 2 GHz Intel Core 2 Duo laptop. The CPU usage
> (around 12%) is also approximately what my CopyImageTest uses when
> numCopies is set to 1, and the fillimage call is doing the same size
> memcpy.
>
> Are you certain your 0% CPU use was for the 'nodelet manager' process,
> and not for the 'rosrun nodelet nodelet manager' process? The later does
> use 0% CPU on my system. I'd welcome any further performance increase,
> but given there is a memcpy to copy the raw data (~7 MB) into the image
> object, I wasn't expecting to do any better.
>
> Joel
>
> Josh Faust wrote, On 17/09/10 20:12:
> >
> >
> >     That's just what I was looking for - thanks. Performance is vastly
> >     improved now, down to about 12% CPU usage on the same within-nodelet
> >     test.
> >
> >
> > 12% still seems a bit high --  I'm running your test now and it's using
> > 0% of my CPU.  Then again I'm on a pretty beefy machine -- what hardware
> > are you running on?
> >
> > Also, I've updated the wiki to mention this form of publishing:
> >
> http://www.ros.org/wiki/roscpp/Overview/Publishers%20and%20Subscribers#Intraprocess_Publishing
> > http://www.ros.org/wiki/nodelet#Publishing_from_a_Nodelet
> >
> > Josh
> >
> >
> >
> > _______________________________________________
> > ros-users mailing list
> > ros-users at code.ros.org
> > https://code.ros.org/mailman/listinfo/ros-users
>
> --
> Research Assistant
> Ocean Systems Laboratory
> Heriot-Watt University, UK
> _______________________________________________
> ros-users mailing list
> ros-users at code.ros.org
> https://code.ros.org/mailman/listinfo/ros-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ros.org/pipermail/ros-users/attachments/20100920/cff11af4/attachment-0003.html>


More information about the ros-users mailing list