[Trilinos-Users] Parallel computing with Trilinos

[Heroux, Michael A]

Ramakrishna,

Please give more details about your problem:


 *   Are you solving the same sized problem each time, or increasing it?  The CPU time is going up, but wall time is stable.
 *   How are you partitioning your problem?
 *   How are you executing your parallel job and on what type of system?

Mike


On 8/15/08 5:06 PM, "RAMAKRISHNA TIPIREDDY" <tipiredd at usc.edu> wrote:

Hi all,

I am trying to use the parallelization capability(with MPI) from Trilinos. I am solving the system of linear equations with Belos::Block Gmres.

I don't see the improvement in efficiency as I increase the number of processors. Follwing are the timings I observed,

1)with one processor:
Resources:        cput=02:19:47,mem=5220848kb,vmem=5508396kb,walltime=02:19:53

2)wth 2 processors
 Resources:        cput=04:44:55,mem=5267036kb,vmem=5718936kb,walltime=02:22:34

3)with 3 processors
Resources:        cput=09:35:22,mem=7185156kb,vmem=7990468kb,walltime=02:23:58.

It will be great if somebody can explain how to interpret these results.

Thanks in advance,

Ramakrishna



-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://software.sandia.gov/mailman/private/trilinos-users/attachments/20080815/a2a8e76e/attachment.html