[Beowulf] Shared memory
Hakon.Bugge at scali.com
Mon Jul 4 01:49:46 PDT 2005
On Sun, 3 Jul 2005 20:12:10 -0700, Greg Lindahl wrote:
> > Unfortunately I can't recommend a simple established code or benchmark
> > which would allow transparent comparison of MPI versus OpenMP/MPI.
>MM5 runs both ways... and it's faster as pure MPI. If OpenMPI+MPI
>doesn't have some special benefit such as accellerating convergence,
>it's not going to be a win.
Hmm, last time I measured I got ~20% speedup using OpenMP+MPI on MM5.
Internode message exchange is less frequent and with larger messages;
both a good thing for most clusters...
More information about the Beowulf