Two-Level Parallel Implementation of FDTD Algorithm on CBE

To understand and predict electromagnetic behavior is one of the most important electrical engineering technologies. FDTD is the key algorithm for solving this kind of problems, which provides a direct, time domain solution to Maxwell's Equations in differential form with relatively good accuracy and flexibility. In this paper, we propose a novel hybrid parallelization model for FDTD method, which combines the domain distribution scheme and the Single Instruction Multiple Data (SIMD) parallelization. This model is able to exploit the computing potential of a new generation parallel machine named Cell Broadband Engine (CBE). The experiment results show that the algorithm obtains approximately linearly speedup with respect to the number of SPEs and has excellent scalability.