A suitable computing architecture and its algorithm for parallel FDTD

Increasing need for solving complex or large scale electromagnetic problem makes parallel FDTD much important now. In this paper, two key factors for efficient parallel FDTD system are clarified first, which are “the number of parallel computing units” and “the number of parallel data exchanging units (PDEU)”. The reason why the traditional parallel FDTD system is not suitable for large scale parallel computing is disclosed. A more reasonable computing structure for overcoming the problem is presented based on PDEU. Scalable Coherent Boundary (SCI) protocol is introduced to realize PDEU and the new parallel FDTD system. The new system provides a 40% speed-up increase comparing with the older one.