Parallel implementations of the finite difference time domain (FDTD) method

In this paper different parallel and sequential implementations of the finite difference time domain (FDTD) method are evaluated. The performance of current sequential architectures are first investigated. Possible improved performances from different parallel architectures are then examined. The parallel implementations consist of a network of workstations, a transputer network and an implementation on a single instruction stream multiple data stream (SIMD) architecture. The FDTD algorithm, a discretisation of Maxwell's curl equations used to solve electromagnetic boundary value problems is also discussed.<
>