[Click] Click SMP
Eddie Kohler
kohler at cs.ucla.edu
Wed May 26 23:45:55 EDT 2004
Hi Giudar,
First off, please report any issues you find with SMP Click. Click's SMP
support is in flux; it certainly doesn't work in 1.4pre1, although it probably
does work in the current CVS version.
We renamed PullToPush to Unqueue a while ago (name changes like this are often
on the changelog:
http://www.pdos.lcs.mit.edu/click/news.html
You've got the right idea for StaticThreadSched, but the details are a little
off. You must supply it with element _names_, not configurations. So not
StaticThreadSched(PollDevice(eth0) 0, ...);
but
pd::PollDevice(eth0);
StaticThreadSched(pd 0, ...);
And yes, Unqueue will fit in the same way.
Good luck!
Eddie
giudar at libero.it wrote:
> Hi all,
> I want to do some performance measurements with SMP Click, by using a dual Intel Xeon processor
> PC.
>
>>From the SMP Click paper, the possible ways to improve parallelism on a multiprocessor machine
> make use of the PullToPush element.
> Looking at the Click elements list, i have not found this element, but I have found another pull to push converter, that is, the "Burster".
> Is this the same of the PullToPush? If not, where can I find a similar element?
>
> In addition, I have used the static CPU assignment for the PollDevice and ToDevice schedulable elements in this way:
> StaticThreadSched(PollDevice(eth0) 0,
> ToDevice(eth1) 0,
> PollDevice(eth1) 1,
> ToDevice(eth0) 1);
>
> How could I statically assign the PullToPush element to a given CPU in a similar way?
> Is the instruction:
> StaticThreadSched(PullToPush 1,
> ..................
> ..................)
> correct?
>
> Any feedback will be very appreciated!
>
> Giuseppe
>
> _______________________________________________
> click mailing list
> click at amsterdam.lcs.mit.edu
> https://amsterdam.lcs.mit.edu/mailman/listinfo/click
More information about the click
mailing list