[clean-energy]

Hello,
At the moment I'm performing some parallel efficiency tests on a sandybridge cluster.
i noticed that the number of needed GAMG iteration depends heavily on the number of parallel processors. This shouldn't be the case (as far as i know).

Do you have any idea?
I'm using openFoam 2.1.x with ICC.

thanks in advance

[Chris Lucas]

Hi,

i'm not 100% sure, but the effect should be related to preconditioning. In a parallel run, the preconditioning is performed for each domain independently.

Best Regards,
Christian

[arjun]

Originally Posted by clean-energy

Hello,
At the moment I'm performing some parallel efficiency tests on a sandybridge cluster.
i noticed that the number of needed GAMG iteration depends heavily on the number of parallel processors. This shouldn't be the case (as far as i know).

Do you have any idea?
I'm using openFoam 2.1.x with ICC.

thanks in advance

Efficiency of multigrid goes down as number of partitions increase. It is known fact. (Not all smoothers do this though, some smoothers could really create multigrid independent of processors).

In case of GAMG, I do not know whether the fall in efficiency you observed is more than it should. But efficiency of multigrid does fall down.