Skip to main content

redistributePar

Overview​

The redistributePar utility takes an existing case and redistributes its mesh and fields according to the settings of the <case>/system/decomposeParDict file.

The utility must be run in parallel

Key features:

  • Enables straightforward repartitioning of mesh and fields
  • Can be used to decompose and reconstruct cases in place of decomposePar and reconstructPar, reconstructParMesh
  • Combines reconstructing mesh with reconstructing fields, i.e. no need for reconstructParMesh after parallel mesh generation
  • Runs fully parallel (if supported by the fileHandler)

Constraints:

  • Must be used on the maximum number of processors
  • To run on distributed systems, the utility must be run locally and then the resulting processor directories copied manually to the remote system(s). (24.04.2023 this limitation is under development)

Usage​

redistributePar [OPTIONS]

To decompose the case in parallel

mpirun -np 2 redistributePar -decompose -parallel

To reconstruct the case in parallel

mpirun -np 2 redistributePar -reconstruct -parallel

To change the partitioning to a new decomposeParDict, e.g. number of processors or type of partitioning:

Contents of original decomposeParDict:

numberOfSubdomains 6;

method          hierarchical;

coeffs
{
    n           (6 1 1);
}

Contents of new decomposeParDict:

numberOfSubdomains 24;

method        scotch;

Note that the number of subdomains has increased from 6 to 24, and the method changed from hierarchical to scotch. The new settings are applied by executing:

mpirun -np 24 redistributePar -parallel

Further information​

Example usage:

  • decomposing using collated output file format:
mpirun -np 2 redistributePar -decompose -fileHandler collated -parallel

Tutorials:

Source code