Hello,
I am trying to run the "fixed sst" aquaplanet (compset QPC6 --res f19_f19_mg17) experiment on CESM 2.1.3. To explore if my control runs look the same or not I have conducted the same experiment twice without changing any configuration for 4 years and have discarded the first 6 months as spinup. To my surprise, I am observing significant differences in the meridional wind V of the two runs. The percentage difference for the area-averaged time series V-winds (100*(V1-V2/V1) comes around 300% to 1000% in magnitude (calculated using default monthly outputs h0).
Also, most of the time I am getting "forrtl: severe (174): SIGSEGV, segmentation fault occurred", even after using "ulimit -s unlimited" from the very beginning. I have to again submit the run to complete the run. Moreover, I have tried to conduct the experiment by increasing the total number of cores, but even after that the segmentation fault shows up sometimes.
To sum up: Is the large percentage deviation in V-wind for the two control runs is somehow associated with the segmentation fault or there are some other possible reasons?
I am attaching the log files for reference.
P.S: I am providing ./xmlchange NTASKS_ATM=384,NTASKS_OCN=384,DOUT_S=FALSE,STOP_OPTION=nmonths,STOP_N=6,RESUBMIT=7
I am trying to run the "fixed sst" aquaplanet (compset QPC6 --res f19_f19_mg17) experiment on CESM 2.1.3. To explore if my control runs look the same or not I have conducted the same experiment twice without changing any configuration for 4 years and have discarded the first 6 months as spinup. To my surprise, I am observing significant differences in the meridional wind V of the two runs. The percentage difference for the area-averaged time series V-winds (100*(V1-V2/V1) comes around 300% to 1000% in magnitude (calculated using default monthly outputs h0).
Also, most of the time I am getting "forrtl: severe (174): SIGSEGV, segmentation fault occurred", even after using "ulimit -s unlimited" from the very beginning. I have to again submit the run to complete the run. Moreover, I have tried to conduct the experiment by increasing the total number of cores, but even after that the segmentation fault shows up sometimes.
To sum up: Is the large percentage deviation in V-wind for the two control runs is somehow associated with the segmentation fault or there are some other possible reasons?
I am attaching the log files for reference.
P.S: I am providing ./xmlchange NTASKS_ATM=384,NTASKS_OCN=384,DOUT_S=FALSE,STOP_OPTION=nmonths,STOP_N=6,RESUBMIT=7