[RegCNET] Regarding RegCM4.4.5.11 simulation stops due to CFL violation in-spite of decreasing dt

SUDIPTA GHOSH sudiptaghosh006 at gmail.com
Sat May 13 07:40:25 CEST 2017


Hello all,

I tried to run a 5year simulation with RegCM4.5 with dx=25km and dt=30 secs
but the model stopped due to CFL violation after 6month simulation. I
decreased the dt to 25, 20 and 15 still the job got killed due the same
reason, however the simulation period varied each time.

Next I tried to run a 2 year simulation with RegCM4.4 at the same
combination of dx and dt but again the problem occurred. I even tried
changing the vertical levels from 18 to 23 but no run was completed.

Suggestions regarding the same is essentially required.

I am submitting the job on 16nodes each containing 24cpus and the message
in the log file is as follows:

WHUUUUBBBASAAAGASDDWD!!!!!!!!!!!!!!!!
 No more atmosphere here....
 CFL violation detected, so model STOP
 #####################################
 #            DECREASE DT !!!!       #
 #####################################
 WHUUUUBBBASAAAGASDDWD!!!!!!!!!!!!!!!!
 No more atmosphere here....
 CFL violation detected, so model STOP
 #####################################
 #            DECREASE DT !!!!       #
 #####################################
 Abort called by computing node          345 at 2017-05-08 18:29:32.161
+0530
 Abort called by computing node          329 at 2017-05-08 18:29:32.162
+0530
--------------------------------------------------------------------------
MPI_ABORT was invoked on rank 345 in communicator MPI COMMUNICATOR 3 DUP
FROM 0
with errorcode 1.

NOTE: invoking MPI_ABORT causes Open MPI to kill all MPI processes.
You may or may not see output from other processes, depending on
exactly when Open MPI kills them.

Thanking you all in advance.

-- 
Regards

Sudipta Ghosh
PhD Scholar, Center for Atmospheric Sciences
Indian Institute of Technology Delhi,India.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ictp.it/pipermail/regcnet/attachments/20170513/3eb3300f/attachment.html>


More information about the RegCNET mailing list