Parallel ADI-BOR-FDTD Algorithm

We present a parallel implementation of the alternating direction implicit-body of revolution-finite difference time domain (ADI-BOR-FDTD) method on a high performance computer using a Message Passing Interface (MPI) library. In BOR-FDTD codes, the body of revolution symmetry is exploited to reduce the computational complexity by projecting the 3-D Yee-cell in cylindrical coordinates onto a 2-D plane. Adopting an implicit update scheme (ADI) frees the time-step size from the Courant-Friedrichs-Lewy time step constraint. We demonstrate further performance gains by parallelizing the algorithm, although the communication overhead between processors is proportional to the area of the domain. In our parallel ADI-BOR-FDTD method each tridiagonal matrix system is solved by a single processor, with the parallel computer architecture being exploited to solve multiple systems at the same time. We benchmarked our code on an IBM p690 symmetric multiprocessor revealing excellent scalability and efficiency.