Four different attempts at parallelizing the calculations using the MPI 2 protocol resulted in no speeds up. In each case, the calculations were up to 2 times slower. It appears as though the data transfer of the PCS, atomics positions, vectors, etc. between nodes is slower than the calculations. As parallelization for speeding up the calculations can only achieve around two orders of magnitude faster calculations, the technique was abandoned.

The relax user manual (PDF), created 2016-10-28.