To build Rosetta in MPI, simply add extras=mpi to your command line for compiling. This produced the MPI-enabled “parallel” executables. Note that almost all of Rosetta is not parallel; the Job Distributor layer is parallelized to allow many independent trajectories under the MPI umbrella. This means you won’t get a performance boost from running 100 Rosetta jobs in regular mode versus a 100-processor MPI job.
The executeables ought to have “mpi” in their name somewhere – if you can’t find them post what you have and I’ll figure it out.
I don’t know how to get MPI installed on your cluster architecture, of course…