CFD Online Discussion Forums - multiregional russian dolls defeat decomposepar?

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)

- OpenFOAM Pre-Processing (https://www.cfd-online.com/Forums/openfoam-pre-processing/)

- - multiregional russian dolls defeat decomposepar? (https://www.cfd-online.com/Forums/openfoam-pre-processing/134676-multiregional-russian-dolls-defeat-decomposepar.html)

multiregional russian dolls defeat decomposepar?

II am using CHTmultiregion in op 2.3, with scotch. There are 5 regions nested like russian dolls. I find there is insufficient cells in the innermost region for to be split by the number of processors needed for the outermost region. This situation means vast processing times.
innermost region has 2 000 cells outermost region 700 000

Can i use different numbers of processors for each region? or is there another solution?

Greetings Derek,

You can control the processor weights. Have a look into the file indicated by this command:

Code:

echo $FOAM_UTILITIES/parallelProcessing/decomposePar/decomposeParDict

In the section for the scotch coefficients, you can control the weight to be assigned to each processor. I've never used it myself, so I don't know if a weight of 0.0 can be used or not.

Best regards,
Bruno

Quote:

Originally Posted by wyldckat (Post 489692)

Greetings Derek,

You can control the processor weights. Have a look into the file indicated by this command:

Code:

echo $FOAM_UTILITIES/parallelProcessing/decomposePar/decomposeParDict

many thanks ... changing the number of subdomains in a region from case level number causes a error in CHTmultiregionfoam.A 0 in the processor weights causes a segmentation fault in decomposepar. I then used a highnumber for processors I wanted (e.g. 10000) and 1 for those other processors in the regional decomposepardict. This seems to work

Code:



scotchCoeffs



{

    processorWeights (10000  1 1 1 1 1);

}