Problem with Parallel not with Serial

iyer_arvind

i am using coodles for LES simulation for a mixing layer. After several thousand iterations, (when i guess that the gradients are large over the partition boundary) the solver crashes due to temperature iterations not converging (typically due to unphysical temperatures). if i go back a few steps and start the process in serial mode the there is absolutely no problem, it does not cause any unphysical temperatures and the courant numbers are stable, as opposed to the former. Is there any problem in the parallel solver?