Running DG-SWEM basicsΒΆ

Running DG-SWEM should be quite simple. The code has been primarily developed to work with the intel fortran compilers, and python (with sympy functionality) is also necessary in order to run the sediment transport portion of the code – though the code may easily be run with this function turned off.

That being said, gfortran compilers can also be used if one has GNU Fortran (Ubuntu/Linaro 4.6.3-1ubuntu5) 4.6.3 or newer. For information on how to compile using this compiler see the online forum.

Once you have intel compilers and python (with sympy) loaded, go to the work directory, “dgswem/work/.” The code can be compiled to run in either serial, parallel, or both. To compile all of the above, a simple

$ make all

works. This is the best option, and in order to run the test case that ships with the code, you will need at least 64 processors, requiring MPI. We tend to use mvapich2, though any implementation will work.

Now, in order to run the code you can simply type:

$ ./dgswem_serial

or

$ ibrun -np 12 ./dgswem

and so on, depending on your submission system. A sample script is included in the repo for submitting to Stampede on TACC.

All parts of the code can be made separately as well. To see these options, one can simply open “makefile” in the work directory. For example, if one wants to compile DG-SWEM with short waves from SWAN, you can find the instructions in the online forum.

It should be noted that the test case that ships in the git repository requires large wind files that cannot be stored on github. As a consequence, these files are hosted elsewhere, and downloaded during compilation using the “get_winds.py” script in the work directory. When running a new (or different) test case, you need to make sure that get_winds is turned off in the “makefile.” This can be accomplished by removing “winds” from the makefile.

Previous topic

Download DG-SWEM and online content

Next topic

Input files