help abt CAM 4.0 installation

[v2dixit@rediffmail_com]

I am trying to install CAM 4.0 on IBM-AIX architecture. Though the code is compiled; it gives following error message while running at first time step.
I checked SST input file with help of GrADs but dataset seems ok.


3:(seq_mct_drv) : Initialize ocn component
3: Read in camdom namelist from file= ocn_in
3: SST dataset will be reused for each model year
3: (GETFIL): attempting to find local file domain.camocn.10x15_USGS_070807.nc
3: (GETFIL): using /home/caos1/mecdixit/cam4/ccsm4_0/models/atm/cam/cam4data/atm/cam/ocnfrac/domain.camocn.10x15_USGS_070807.nc
3: (GETFIL): attempting to find local file sst_HadOIBl_bc_10x15_clim_c050526.nc
3: (GETFIL): using /home/caos1/mecdixit/cam4/ccsm4_0/models/atm/cam/cam4data/atm/cam/sst/sst_HadOIBl_bc_10x15_clim_c050526.nc
3: WRAP_GET_VARA_REALX: error reading varid = 8
3: Index exceeds dimension bound
3: ENDRUN:HANDLE_ERROR

The configure and build namelist commands are as follows ( default test case ):
./configure -dyn fv -hgrid 10x15 -ntasks 6 nosmp
./build-namelist -test -config ./config_cache.xml

Any help in this regards shall be very helpful.

Thanks and Regards,
Vishal

 

[eaton]

Here is the output I get when I run the test you describe on the IBM power6 platform at NCAR:

0:(seq_mct_drv) : Initialize ocn component
0: Read in camdom namelist from file= ocn_in
0: SST dataset will be reused for each model year
0: (GETFIL): attempting to find local file domain.camocn.10x15_USGS_070807.nc
0: (GETFIL): using /fis/cgd/cseg/csm/inputdata/atm/cam/ocnfrac/domain.camocn.10x15_USGS_070807.nc
0: (GETFIL): attempting to find local file sst_HadOIBl_bc_10x15_clim_c050526.nc
0: (GETFIL): using /fis/cgd/cseg/csm/inputdata/atm/cam/sst/sst_HadOIBl_bc_10x15_clim_c050526.nc
0: SSTINI: Read sst data for dates 1216 43200 and 116 43200
0:(seq_mct_drv) : Initialize ice component

The run completes successfully.

The only difference that sticks out is that my log output is all coming from task 0 which is the typical "master task" while your log output is from task 3. I'm wondering if that is indicating a problem. I would try a serial run, and if that is successful then the problem would seem to be related to the mpi implementation.

 

[v2dixit@rediffmail_com]

Hi Eaton,

Thanks for the guidelines.
Any suggestions about how to handle changes in MPI implementation. Should I try run with different number of threads or SMP implementation ?

Thanks and Regards,
Vishal

 

[eaton]

I have no expertise with MPI installations. But changing the number of tasks and/or threads are good strategies for debugging this type of problem.