Nuno,
> I'm trying to run step 36 example with mpi but I can only manage to run
> it with only one processor.
> I'm using:
> - r20053 | bangerth | 2009-11-05 19:31:46 -0500 (Thu, 05 Nov 2009)
The step-36 tutorial does not run in parallel "out of the box". You will
have to make some minor modifications to the application to get him to
work on parallel architecture.
Fortunately, this really is quite simple ;-)
Start by comparing the step-36 code with the step-17 tutorial which
definately runs in parallel: First up: you'll be needing to set up an
MPICommunicator... also you'll need to transform the matrices and vectors
to their MPI counterparts, note PETScWrappers::MPI::SparseMatrix
stiffness_matrix. And so on.
Have fun there, if you make sure you are using the same compiler to
compile all of your components and setup the MPI correctly, I think
you'll find step-36 based things really can work in parallel and without
too much trouble.
> - ParMetis-3.1.1
A small note here:
I am not aware that any of the extended (ie. parallel) parts of ParMetis
are currently taken advantage of with deal.II. For safety, it may be worth
using Metis-4.0, which is well tried an tested and definately works well
with deal.II. It may be that ParMetis works in an equivalent manner as
Metis here, I do not know, I have never tried. Please let me know if
ParMetis works for you.
Best,
Toby
-----
Toby D. Young
Assistant Professor
Philosophy-Physics
Polish Academy of Sciences
Warszawa, Polska
www: http://www.ippt.gov.pl/~tyoung
skype: stenografia
_______________________________________________
dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
