Transient Electromagnetic Simulation for Periodic Structures with a Modified Parallel DGTD Method

Due to natural block diagonal mass matrices, discontinuous Galerkin time domain (DGTD) method provides convenient condition for parallel computation. However, when the number of processes increases, the traffic between processes gradually increases, leading to the sharp decline of the parallelism efficiency. To solve these problems, non-blocking communication technology with the math kernel library (MKL) scheme is proposed to improve the efficiency of message passing interface (MPI) parallel computing. Thus, a typical periodic electromagnetic structure - frequency selective surface (FSS) is simulated, which achieves an acceleration ratio of more than 1.8 times, and further improves the performance of the DGTD algorithm.