Parallelization of multilevel fast multipole algorithm with open MPI for scattering by large scale targets

In this paper, we present the parallelization of multilevel fast multipole algorithm (MLFMA) for three-dimensional scattering problems of large scale objects. A new, full MPI-2 supported implementation, named Open MPI, is adopted in our code as the parallel communication library. We compare the performance of the parallel code based on Open MPI and MPICH, the numerical result shows that Open MPI performs better than MPICH as a parallel library for our code. Besides, a novel method is proposed for reducing the truncation number without any loss of accuracy. The capability and accuracy of the parallel code are demonstrated by considering a sphere of 150λ discretized with 20,919,633 unknowns and a square patch of 200λ discretized with 10,150,143 unknowns.