Dramatic slowdown for -np 2

Martin Siegert siegert at sfu.ca
Tue Feb 12 11:21:23 PST 2002


Hi John,

On Mon, Feb 11, 2002 at 09:58:55PM -0800, J Harrop wrote:
> This may be an MPI problem, but I'm not sure so I'm posting it here and 
> comp.parallel.mpi
> 
> We are developing an application on a four node Beowulf while we wait for 
> the remaining nodes to arrive.  Speed-up has been close to predicted with 
> -np 3 and 4 in a master/slave mode.  But when I run at -np 2 the speed 
> drops to approximately 1/4 of the original serial application.  (On 4 nodes 
> - that is 1 master and 3 slaves, we get about 2.5 times speed-up relative 
> to the original application.)  All runs produce the same answer.
> 
> In the MPI code we have basic SEND, RECV, BCAST and REDUCE - nothing 
> fancy.  Does anyone know if any of these or other MPI functions run into 
> problems with one to one master slave ration?  Any other enlightenment 
> would be welcome.

If you are using mpich: compile using

mpif90 -mpilog ... (or mpicc -mpilog ...)

and then run your program as usual with "mpirun -np 2 ..."

Look at the generated log file with jumpshot (you may have to
convert the log file first: "clog2slog file.clog"). Then do the same with
"mpirun -np 3 ..." This should give you an idea in which part of the
communication pattern your job gets stuck.

Cheers,
Martin

========================================================================
Martin Siegert
Academic Computing Services                        phone: (604) 291-4691
Simon Fraser University                            fax:   (604) 291-4242
Burnaby, British Columbia                          email: siegert at sfu.ca
Canada  V5A 1S6
========================================================================



More information about the Beowulf mailing list