Open MPI logo

Open MPI User's Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Open MPI User's mailing list

From: Galen M. Shipman (gshipman_at_[hidden])
Date: 2005-10-31 11:20:49

On Oct 31, 2005, at 8:50 AM, Mike Houston wrote:

> When only sending a few messages, we get reasonably good IB
> performance,
> ~500MB/s (MVAPICH is 850MB/s).
What is your message size? Are you using the leave pinned option? If
not, specify -mca mpi_leave_pinned 1 option to mpirun. This tells
Open MPI to cache message buffers (this is what MVAPICH does),
otherwise we don't have a fair comparison.

> However, if I crank the number of
> messages up, we drop to 3MB/s(!!!). This is with the OSU NBCL
> mpi_bandwidth test.

When you say crank the number of messages up, do you mean the number
of messages of the same size? If so, what message size?



> We are running Mellanox IB Gold 1.8 with 3.3.3
> firmware on PCI-X (Couger) boards. Everything works with MVAPICH, but
> we really need the thread support in OpenMPI.
> Ideas? I noticed there are a plethora of runtime options configurable
> for mvapi. Do I need to tweak these to get performacne up?
> Thanks!
> -Mike
> _______________________________________________
> users mailing list
> users_at_[hidden]