Parallel Processing ocean wave multiphase model
I have successfully produced free surface waves using a multiphase setup in both 2D and 3D using CFX 11. The models have very conservate parameters and run fine on 1 processor, but as soon as i submit the same model in parallel it fails. The out file shows successful partitioning, and implimentation of MPICH communication on the cluster. I have tried standard MeTiS partitioning and coupled (-part-coupled on linux script). I was wondering if there were any tips for this such as partition weighting or partition algorithym that anyone has for multiphase parallel solving?
At the moment the models run using a single processor, but as soon as i attempt to parallel solve the same model, it fails on the first few iterations, any idea what i am missing?
Check that your partition boundaries do not line up with the free surface. If you write out a backup/res file after 1 timestep, then plot isosurfaces of partition number in Post. If they do line up, then try a user specified direction partitioning option.
|All times are GMT -4. The time now is 23:16.|