Dear Ben, On Tue, 9 Sep 2008, Benjamin Kirk wrote:
> That is my thinking. The System::project_vector() code does some all-to-all > communication, and this seems to be scaling quite badly as you get to larger > processor counts. Running with 1 CPU/node will hopefully perform better > since you are not sharing a gigE connection between processors. > > By contrast, the ProjectVector::operator() does the actual projection and > contains no parallel communication. As you can see it is scaling ~linearly > to 40 processors: Ah, thank you very much for that enlightenment. That seems to explain things. I just noticed that it was quite silly to start the 1-CPU-per-node experiment with 1 node and restarted it now with 20 nodes. If your conjecture is true, it should be finished soon, and I can let you know today. Best Regards, Tim -- Dr. Tim Kroeger Phone +49-421-218-7710 [EMAIL PROTECTED], [EMAIL PROTECTED] Fax +49-421-218-4236 MeVis Research GmbH, Universitaetsallee 29, 28359 Bremen, Germany Amtsgericht Bremen HRB 16222 Geschaeftsfuehrer: Prof. Dr. H.-O. Peitgen ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/ _______________________________________________ Libmesh-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/libmesh-users
