

MPI parallel run issue for RBF motion solver in foam extend 4.0
Hi,
I got to know that for running simulation in parallel for RBF solver their need to make sure each partition has atleast one boundary point that is on the moving boundary.
By this constraint, I can easily run my simulation for 4 cores (nsubDomain 4; method simple (2 2 1);). However, I want to run the simulation for more than 4 cores (such as for 8, 16, 32; I have system with 32 cores, yet).
So, to do that I have to provide atleast one boundary point of moving boundary at each partition. And I don't know how I can achieve this? Can anyone put some light on it.
Number of mesh elements: 4,00,000
Please, feel free to ask for more information. Any kind of Reply will be really helpful.
Thanks.
OpenFOAM


Hi, You could try decomposition using scotch method. Usually it automatically decompose domain smartly.


Thank you so much for doing some digging for me. However, by this paper I got to know that I need atleat one moving boundary points for each partion in parallel run. If you find something more please share. Thanks Best regards Ram
I tried but it did not work.
09-12-20, 4:11 p.m. guptaram@
I haven't worked much on RBF solver in parallel computing. But from a few forum discussions, I found this paper reference that might help you.
Paper: https://doi.org/10.1016/j.compfluid.2013.02.004
10-12-20, 5:37 p.m. divyesh
Login to add comment