[Tarak]

Hii,

While decomposing my domain of 3.4 million cells into 64 processors, the cells do not get distributed evenly. While in some processors around 26000 cells are allocated, while for some around 70000. This makes it slower than that using 32 processors, and I assume this is due to this load impalance.

Can you please let me know a way to get around this, so that I can get almost equal number of cells in each processors.

[FG_HSRM]

Hi Tarak

Until now I have no idea, but you should post your decomposeDict.

[Tarak]

FoamFile
{
version 2.0;
format ascii;
class dictionary;
location "system";
object decomposeParDict;
}
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //

numberOfSubdomains 64;

method simple;

simpleCoeffs
{
n ( 8 4 2 );
delta 0.001;
}

hierarchicalCoeffs
{
n ( 2 1 1 );
delta 0.001;
order xyz;
}

metisCoeffs
{
}

manualCoeffs
{
dataFile "";
}

distributed no;
roots ( );


// ************************************************** *********************** //

[FG_HSRM]

Originally Posted by Tarak

Hii,
...almost equal number of cells in each processors.

Try to replace simple by metis, everything else should be fine.
You also can try scotch, but than you have to add:
scotchCoeffs
{
processorWeights ()
}
A detailed description you can find in the UserGuide.