readme

CCSM Research Tools: Community Atmosphere Model (CAM)

A production version of CAM can be built by entering the following commands
from the command line:

% $CAM_ROOT/models/atm/cam/bld/configure
% make

This assumes:

1) The shell variable CAM_ROOT has been set to the root directory of the CAM
   distribution.  The root directory is the top level directory that is created
   when the distribution is untarred.  It contains the subdirectories
   models and scripts.
2) make is a GNU Make (check by issuing "make -v").

A namelist that will run CAM for a day is produced by issuing the command:

% setenv CSM_DATA top_level_data_dir
% $CAM_ROOT/models/atm/cam/bld/build-namelist

This assumes that the environment variable CSM_DATA has been set to the top
level directory where the CAM data distribution has been untarred.  The top
level directory contains the subdirectories atm and lnd.

To run CAM interactively for a day issue the command:

% ./cam < namelist

The configure and build-namelist scripts are fully documented in the CAM
User's Guide.


Example self-contained run scripts are provided for the three architectures
available at NCAR on which CAM is known to run.  These are IBM SP, SGI
Origin, and PC running Linux.  On a PC, either pgf90 (Portland Group) or lf95
(Lahey) compilers may be used.  The scripts are run-ibm.csh, run-pc.csh, and 
run-sgi.csh.

Each script is set up to run the production configuration (T42L26 with
Eulerian dynamics).  Minimal changes to the scripts should be necessary if they
are run with this configuration on most machines at NCAR.  In order to run
on machines at other locations, issues such as the location of the netcdf
library, differences in the queueing system, and interface to a data
archival device may need to be addressed.  Changing the model resolution or
underlying dynamical core will also require modifications.  Specifically,
changes to the .h files and/or Filepath are needed.  The configure script
produces these files.

Optimal execution speed of CAM is obtained on an IBM SP with a multitasking
strategy that is hybrid OpenMP/MPI.  On an SGI Origin, pure OpenMP yields the
best speed.  The run scripts for these architectures are configured
accordingly.  The script for a PC is configured to run pure MPI.