2.4 Running CSIM Uncoupled

It is assumed that the user has downloaded the source code and input data from the web page described in Section 2.2. This section is intended to get uncoupled CSIM running "out of the box" with a minimal amount of information. More information on modifying the scripts is given in Section 3.1. The default configuration is a 10 day, startup run on the gx3v5 grid, using 8 processors and the message passing interface (MPI). The debugging option is turned on, history files are written out daily, and restart files are written out every 5 days. The debugging option should be turned off, and the output frequency should be decreased before starting any production runs. If your system does not have 8 available processors or MPI, see Section 2.4.3 on how to run the model on a single processor without MPI.

Running this configuration will verify that the library and compiler options are properly set in the Macros.$<$OS$>$ file, all the input data is in the correct place and the environment variables are set correctly before any further changes are made to the scripts or the source code. This will also provide benchmark output.

NOTE

If you are running this model a machine other than an IBM running AIX or an SGI running IRIX, you may need to make an equivalent Macros.$<$OS$>$ file with the paths and settings modified for your system.

Before you start, modifications will be needed in the run script csim_run to set the directories for the source code, input data, and executables. The following is a list of the environment variables that will need to be changed by the user:

setenv CSIMDIR  /home/$LOGNAME/csim5             # directory of scripts
setenv CSIMDATA  /fs/cgd/csm/inputdata/ice/csim4 # dir for input data
setenv CASE    test.me                           # Case name
setenv EXEROOT  /ptmp/$LOGNAME/$CASE             # run model here
setenv SHRCODE  $CSIMDIR                         # dir for share code

$CSIMDIR is the top directory of the source code, where the scripts are located. $CSIMDATA is the directory where the input data sets are located. $CASE is a string with a case name for the model run and should be kept short since it is used in path and file names. $EXEROOT is typically a large temporary disk where the executable files, and input data sets will reside during execution. Information output by the model will also be written to $EXEROOT.

The location of $SHRCODE will depend on where the source code was downloaded. If it was obtained with CCSM3.0 distribution, the $SHRCODE directory will be under ccsm3/models/csm_share. If you only have the source code for CSIM, the share code will be in the same directory as the rest of the ice model source code.

2.4.1 Multiple Processors with MPI

The default setting will use eight processors (two nodes with four processors each) and the Message Passing Interface (MPI), so simply submit the job. To submit a run to the batch queue on the IBM, type llsubmit csim_run. To submit the job to the batch queue on an SGI, type qsub $<$ csim_run, bsub $<$ csim_run or the appropriate command depending on your batch queueing system. On some systems it is possible to run multiple processor jobs interactively.

2.4.2 Single Processor with MPI

In the run script csim_run, change $NX and $NY to 1. If you are submitting to a batch queue, the number of processors you are requesting will also need to be modified in the batch queue environment information at the top of the script. For example, for the IBM, the following two lines should be modified to:

  # @ total_tasks = 1
  # @ node = 1

The model can also be run interactively by typing csim_run.

NOTE

When you change the number of processors, the output you get in the log file will be slightly different from that calculated with a different number of processors. This is due to changes in the order of operations in calculating the global sums.

2.4.3 Single Processor without MPI

In the run script csim_run, change $NX and $NY to 1, and $BINTYPE to 'single' (or anything except 'MPI'). This value of $BINTYPE will automatically change the preprocessor flags and the compiler name where necessary in the Macros.$<$OS$>$ file. The model can be run interactively by typing csim_run.