Fluent Runtime After Adaption
Wondering what others experience is with running Fluent and hanging-node adaption.
I have a case with 50 million finite volumes that takes X sec/iteration. I adapted and have 65 million finite volumes and it now takes 4X sec/iteration. I would expect less, but any other experiences would be useful. Thanks Scott N |
hi scott
i presume you are running a distributed parallel run? (or you have very nice SMP hardware...) what i have seen is that the adapted mesh sometimes tend to stay on the partition where it was generated. You will start seeing unequal loads on the parallel cores - in effect you might have a few of your cores which now has 4 times as many cells to compute. you can either manually repartition in Fluent, or try activating the load balancing for parallel runs. just a thought... |
All times are GMT -4. The time now is 21:09. |