|
I have a Fortran code parallelized with OpenMP that I have been running on the Duke cluster.
The code runs in 30 min on 44 cores. Running in OpenMP requires me to use a single node on the cluster. Ideally, I would like to run it in more than one node (which I believe can be achieved with MPI). Thus I need some help to convert the OpenMP loops, to something that can be parallelized using MPI (and multiple nodes). I can send the code if you would like to take a look before starting the task (email me at: nd140@duke.edu) |