Difference between revisions of "Contrib multiSolver/parallel"

Revision as of 16:39, 1 July 2011

Parallel processing is now available with multiSolver.

1 decomposePar

To decompose the case directory:

1. Set it up as a usual multiSolver-enabled case directory;
2. Create a system/decomposeParDict file as you would with a regular parallel solver;
3. Instead of decomposePar, use:

multiSolver -preDecompose && decomposePar && multiSolver -postDecompose

2 reconstructPar

To reconstruct the case directory:

Instead of reconstructPar, use:

multiSolver -preReconstruct && reconstructPar && multiSolver -postReconstruct

3 Aliasing

If you are going to be doing this regularly, it might be a good idea to create a shorter alias for these two commands. To do this, add:

alias multiDecomposePar='multiSolver -preDecompose && decomposePar && multiSolver -postDecompose'
alias multiReconstructPar='multiSolver -preReconstruct && reconstructPar && multiSolver -postReconstruct'

To the end of your OpenFOAM-version/etc/aliases.sh or OpenFOAM-version/etc/aliases.csh file.

4 Parallel post processing

Apparently post processors are available that work with the data split across the processor directories / drives. multiSolver can be post processed in this way as well. To achieve this, use the commands as described on the post processing page, except run them in parallel.

For example, instead of:

multiSolver -load all

use:

mpirun -n 4 multiSolver -load all -parallel

substituting the correct options for mpirun. Then run your fancy parallel post processor.