What version of the code are you using?
release-cesm2.1.5
Have you made any changes to files in the source tree?
No change in code
Describe every step you took leading up to the problem:
I made changes nhtfrq = -4 mfilt = 6 pertlim=1.0e-14 and pertlim=-2.0e-14 in user_nl_cam_0001 and user_nl_cam_0002
and STOP_N=1 to observe a simple run
Describe your problem or question:
I documented all the xml options in the text, and I would like to set a runtime of 2 hours (or even 1 hour?) for resubmit after dart assimilation. I found that the parameter for controlling the time step is atm_ncpl set to 96; is this sufficient for stop_n=2?
I am very concerned that a runtime that is too short will result in similar outputs for ensemble members due to the close proximity of disturbances. For the five-year resubmit-assimilation-resubmit-... circular, is 40 members sufficient, and maybe some preliminary practical suggestions regarding the selection of other parameters such as atm_ncpl is really useful.
Another question is that in experiments without disturbances, with NTASKS=64, the average CPU utilization is 6300% (core=64). However, after introducing disturbances, it fluctuates around 4500% (ntasks=32 -> run command is mpiexec -n 64). I would like to know the reason for the decrease in efficiency, as this is very important for estimating computational resources.
Thank you for your patience in reading and answering!
release-cesm2.1.5
Have you made any changes to files in the source tree?
No change in code
Describe every step you took leading up to the problem:
I made changes nhtfrq = -4 mfilt = 6 pertlim=1.0e-14 and pertlim=-2.0e-14 in user_nl_cam_0001 and user_nl_cam_0002
and STOP_N=1 to observe a simple run
Describe your problem or question:
I documented all the xml options in the text, and I would like to set a runtime of 2 hours (or even 1 hour?) for resubmit after dart assimilation. I found that the parameter for controlling the time step is atm_ncpl set to 96; is this sufficient for stop_n=2?
I am very concerned that a runtime that is too short will result in similar outputs for ensemble members due to the close proximity of disturbances. For the five-year resubmit-assimilation-resubmit-... circular, is 40 members sufficient, and maybe some preliminary practical suggestions regarding the selection of other parameters such as atm_ncpl is really useful.
Another question is that in experiments without disturbances, with NTASKS=64, the average CPU utilization is 6300% (core=64). However, after introducing disturbances, it fluctuates around 4500% (ntasks=32 -> run command is mpiexec -n 64). I would like to know the reason for the decrease in efficiency, as this is very important for estimating computational resources.
Thank you for your patience in reading and answering!