A MASSIVELY-PARALLEL COMPUTATION STRATEGY FOR FDTD - TIME AND SPACE PARALLELISM APPLIED TO ELECTROMAGNETICS PROBLEMS

In this paper, we present a novel strategy for incorporating massive p arallelism into the solution of Maxwell's equations using finite-diffe rence time-domain methods. In a departure from previous techniques whe rein spatial parallelism is used, our approach exploits massive tempor al parallelism by computing all of the time steps in parallel. Further more, in contrast to other methods which appear to concentrate on expl icit schemes such as Yee's algorithm, our strategy uses the implicit C rank-Nicolson technique which provides superior numerical properties. We show that the use of temporal parallelism results in algorithms whi ch offer a massive degree of coarse grain parallelism with minimum com munication and synchronization requirements. Due to these features, th e time-parallel algorithms are particularly suitable for implementatio n on emerging massively parallel multiple instruction-multiple data (M IMD) architectures. The methodology is applied to a circular cylindric al configuration, which serves as a testbed problem for the approach, to demonstrate the massive parallelism that can be exploited. We also discuss the generalization of the methodology for more complex problem s.