摘要

Constant horizontal wavenumber approach is a simple method to model Periodic Boundary Conditions (PBC) in the Finite-Difference Time-Domain (FDTD) method proposed for efficient analysis of periodic structures; however, it requires execution of the FDTD simulations many times, each time for a different value of horizontal wavenumber to achieve useful results. Therefore, although each simulation may take a short time to complete, a sweep of simulations still takes a long time and there is a need to employ methods to speed-up the simulations. In this contribution we present an implementation of the FDTD/PBC algorithm using the Compute Unified Device Architecture (CUDA) to run the simulations on Graphics Processor Unit (GPU) devices to speed-up the the FDTD/PBC. simulations. We also present a method in which a problem space is extended by one padded cell on each of the four periodic sides. As a consequence, programming is simplified, especially for the GPU code for the field update process at the boundaries, the problem space and efficiency of calculations as well is improved.

  • 出版日期2014-12