is mumps inevitable?

This forum is for issues related to parallel processing
and OpenSees using the new interpreters OpenSeesSP and OpenSeesMP

Moderator: selimgunay

Post Reply
Unni Kartha G
Posts: 11
Joined: Sat Jul 18, 2009 11:16 pm
Location: Cochin University of Science and Technology
Contact:

is mumps inevitable?

Post by Unni Kartha G » Tue Aug 25, 2009 4:38 am

Myself is new to opensees.

I am installing opensees parallel on linux cluster (debian). Almost all the libraries like lapack, superlu etc are availabe in the debian repository. We are using openmpi. After quite a few trials, I could sucessfully install openseesSP. After installation i found that mumps is also a part of opensees.

I am planning to use opensees parallel for a research work in soil structure interaction.

My question is -
is mumps an inevitable part of opensees parallel?
or
In what all situation i will need mumps?

Please help. thanks in advance.
unnikartha

Unni Kartha G
Posts: 11
Joined: Sat Jul 18, 2009 11:16 pm
Location: Cochin University of Science and Technology
Contact:

regarding mumps

Post by Unni Kartha G » Sun Aug 30, 2009 3:37 am

I think i got the answer from the manual. it says....


"4.1 Additional Solvers for Parallel Processing
A number of additional solvers may be available for parallel processing on the local parallel
machine. These include Mumps, Petsc, and SuperLU. Of these, SuperLU will always be available,
Petsc will probably be available for your system, and Mumps you will have to install because of
copyright restrictions. Of the three, Mumps typically will perform the best and should be the
one that is tried first. If you specify SparseGEN in the script, when running on a parallel
machine you will automatically be switched to the parallel version of SuperLU. To use the
additional solvers, the system command is altered to include the parallel solvers.
system Mumps
system Petsc
"
unnikartha

Post Reply