[mpich-discuss] MPICH2-1.0.8 performance issueson Opteron Cluster

[Dave Goodell]

On Jan 7, 2009, at 1:52 PM, Rajeev Thakur wrote:

> Are you using MPI-2 dynamic process functions (connect-accept or  
> spawn)? It
> is possible that for dynamically connected processes on the same  
> machine,
> Nemesis communication goes over TCP instead of shared memory (Darius  
> can
> confirm), whereas with ssm it does not.

The nemesis piece of this statement is true.  Nemesis dynamic  
processes will use tcp to communicate between different comm_worlds,  
although they will use shared memory still within a comm_world.  I  
don't know enough about ssm to say what it does (does it even support  
dynamic procs?).

-Dave