Just tried these options and was able to increased shared memory throughput to approx. 1950 MB/s. Still I wonder what the bottleneck is. After all, the memory bandwidth of the systems is much higher.
Thanks -Andreas On 03:22 Fri 09 Jul , Ralph Castain wrote: > Did you remember to set --bind-to-core or --bind-to-socket on the cmd line? > Otherwise, the processes are running unbound, which makes a significant > difference to performance. > > > On Jul 9, 2010, at 3:15 AM, Andreas Schäfer wrote: > > > Maybe I should add that for tests I ran the benchmarks with two MPI > > processes: for InfiniBand one process per node and for shared memory > > both processes were located on one node. > > > > > > -- > > ========================================================== > > Andreas Schäfer > > HPC and Grid Computing > > Chair of Computer Science 3 > > Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany > > +49 9131 85-27910 > > PGP/GPG key via keyserver > > I'm a bright... http://www.the-brights.net > > ========================================================== > > > > (\___/) > > (+'.'+) > > (")_(") > > This is Bunny. Copy and paste Bunny into your > > signature to help him gain world domination! > > _______________________________________________ > > users mailing list > > us...@open-mpi.org > > http://www.open-mpi.org/mailman/listinfo.cgi/users > > > _______________________________________________ > users mailing list > us...@open-mpi.org > http://www.open-mpi.org/mailman/listinfo.cgi/users -- ========================================================== Andreas Schäfer HPC and Grid Computing Chair of Computer Science 3 Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany +49 9131 85-27910 PGP/GPG key via keyserver I'm a bright... http://www.the-brights.net ========================================================== (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination!
pgpsRtVWHglMu.pgp
Description: PGP signature
