Accelerating FDTD algorithm using GPU computing

Hardware acceleration of Finite-Difference Time-Domain (FDTD) algorithm has always been an important part of FDTD research. In this essay, we discussed the advantage and feasibility of accelerate FDTD algorithm using Graphics Processing Unit (GPU). With the implement of lattice-threads mapping and other techniques, we proposed a GPU FDTD programming model that is both efficient and accurate. Then we simulated two projects using proposed model, the result shows good agreement with CPU computing while simulation speedup is 20 at minimum, thus proves the efficiency and great potential of GPU accelerating in FDTD research.