Cam5-som

[bing_pu@jsg_utexas_edu]

Hi all,

I'm a new user of CAM5 and I got a few questions about running it with the slab ocean model:

1. Is CAM5 (standalone) able to run with the mixed layer ocean model (DOCN-SOM), or it must be run under the CESM framework (e.g., compset E)?
2. If CAM can run standalone with SOM, what changes do I need to make when configuring CAM?
3. Is pop_frc.1x1d.090130.nc the right inputdata for the slab ocean model? For CAM model namelist, what should I set for 'bndtvs' ?

I've been struggling with these questions for a few days. Has any one run CAM5-SOM before? Any suggestions? Thanks in advance!

 

[eaton]

The CAM standalone scripts don't support running the SOM mode. The only supported way to do this is using the CESM scripts with an "E" compset.

 

[bing_pu@jsg_utexas_edu]

Thank you very much, eaton!

I start to run CESM1.0.2 in "E" compset, and have a problem with restart files.

No matter I set the restart options in env_run.xml file, REST_DATE, REST_N, and REST_OPTION, to some values, e.g., 6 months, or simply use the default values, REST_DATE=$STOP_DATE, REST_OPTION=$STOP_OPTION, simulation stopped right after producing one restart file XXX.docn.r.yyyy-mm-dd-hhmmss.nc, and history files of the month before the restart month are missing (history files are wrote out every month as default).

The error message in ccsm.log file is:
===========================
[cli_9]: aborting job:
Fatal error in MPI_Info_free: Invalid argument, error stack:
MPI_Info_free(102): MPI_Info_free(info=0x73109fc) failed
MPI_Info_free(62).: Invalid MPI_Info
[cli_10]: aborting job:
Fatal error in MPI_Info_free: Invalid argument, error stack:
MPI_Info_free(102): MPI_Info_free(info=0x73109fc) failed
MPI_Info_free(62).: Invalid MPI_Info
….
============================
Is there anything wrong in simulation setting? What should I do to get monthly restart files?

Thanks a bunch!

 

[eaton]

Errors encountered while writing restart files are often due to hitting a memory limit. Try running with more tasks and on more nodes to increase the total available memory and decrease the memory per task.

 

[bing_pu@jsg_utexas_edu]

Thanks, eaton! I increased 32 processors (on 4 nodes) to 64 (8 nodes), the maximum in our cluster, still get the same error message. Our IT support said that each node has 24G memory. How can I tell if it hits the limit?

Here are the memory use info from the ccsm.log file:

===============================
.....
Memory block size conversion in bytes is 4094.00
8 MB memory alloc in MB is 8.00
8 MB memory dealloc in MB is 8.00
Memory block size conversion in bytes is 4094.00
8 MB memory alloc in MB is 8.00
8 MB memory dealloc in MB is 8.00
Memory block size conversion in bytes is 4094.00
64 pes participating in computation
===============================

Thank you very much!!!

 

[eaton]

It seems that you have plenty of available memory.

The error message indicates a problem with an mpi_info_free call. I only see one reference to that routine in the source code, and it's in the pio code. What compiler (and version) are you using?

 

[jedwards]

This is fixed in a later version of pio. You can update to the latest version using svn here
http://parallelio.googlecode.com/svn/trunk_tags/pio1_2_6/pio

or (this is much easier) you can simply comment out the mpi_info_free call in
models/utils/pio/piolib_mod.F90

 

[bing_pu@jsg_utexas_edu]

I commented out the mpi_info_free call, and it works! Thank you all!!!

 

[1475922792@qq_com]

Hello all,Is pop_frc.1x1d.090130.nc the right inputdata for the slab ocean model? Can I do the parameterizations intercomparisons by using the slab-ocean model simulations? Hope someone can help.Thank you!Jocelyn