CUDA Implementation of ${\rm TE}^{z}$-FDTD Solution of Maxwell's Equations in Dispersive Media

This letter presents the graphic processor unit (GPU) implementation of the finite-difference time domain (FDTD) method for the solution of the two-dimensional electromagnetic fields inside dispersive media. The FDTD is truncated by the convolutional perfectly matched layer (CPML) and the piecewise-linear recursive-convolution (PLRC) formulation is used for modeling dispersive media. By using the newly introduced Compute Unified Device Architecture (CUDA) technology, we illustrate the efficacy of GPUs in accelerating the FDTD computations by achieving significant speedup factors with great ease and at no extra hardware/software cost. We validate our approach by comparison to exact and other simulated results, which show favorable agreements. The effect of the GPU–CPU memory transfers on the speedup factor will be also studied.