Running Flow in parallel

Flow supports running simulations in parallel using MPI. To use this run Flow with mpi_run: (note that the next command should be a single line)

mpirun -np 4 flow NORNE_ATW2013.DATA --output-dir=out_parallel

The above command runs the simulator with four processes, so you should have at least four cores or hardware threads available to gain from this (adjust the parameter to suit your computer). The simulator should then run quite a bit faster (if not 4 times faster). It should be noted that since the input deck format and output file formats are serial in nature, input and output has not been parallelized.

The resulting output should be equivalent to the output from the serial run. It will not be identical however, since the linear solver will not behave identically between the serial and parallel runs.