Compiling and running an example configuration

Compiling

ROMS does not have a single, central executable, and each configuration must be compiled individually. This is because ROMS takes advantage of several configuration-dependent compile-time optimizations. ROMS includes a number of simple example configurations in the tests directory that can easily be compiled and run on a modern laptop. To verify everything has gone as expected up to this point, we’ll compile the Rivers_real example, which is a small simulation of a river outflow using netCDF input and output on 6 processors.

First, obtain the necessary netCDF input files for the examples:

cd $ROMS_ROOT/tests/input_data/
./get_input_files.sh

Next, compile the model:

cd $ROMS_ROOT/tests/Rivers_real
make

You should now have an executable, roms, in the directory.

Running

To run the model, do

mpirun -n 6 ./roms benchmark.in

where -n 6 is the number of processors, and benchmark.in is a text file defining runtime parameters such as the number of time steps. Unless edited, the model will run for 20 time steps.

Warning

On HPC systems, you will likely need to wrap the above MPI run command in a script and submit it to a scheduler. Consult your system’s documentation.

The screen should fill with live output from the model.

Processing the output

Once the run has finished, the directory will contain six files like rivers_his.20121209133435.0.nc, which are the model output. The six files each correspond to a different processor, each handling one sixth of the domain. We need to stitch these files together to view the domain in its entirety. This can be done using the ncjoin tool included with ROMS:

ncjoin rivers_his.20121209133435.*.nc

after which there will be an additional netCDF file rivers_his.20121209133435.nc containing the joined outputs from the 6 processors - this output file is now ready for analysis.