GPU-ACCELERATED FUNDAMENTAL ADI-FDTD WITH COMPLEX FREQUENCY SHIFTED CONVOLUTIONAL PERFECTLY MATCHED LAYER

This paper presents the graphics processing unit (GPU) accelerated fundamental alternating-direction-implicit flnite-difierence time-domain (FADI-FDTD) with complex frequency shifted convolu- tional perfectly matched layer (CFS-CPML). The compact matrix form of the conventional ADI-FDTD method with CFS-CPML is formu- lated into FADI-FDTD with its right-hand-sides free of matrix oper- ators, resulting in simpler and more concise update equations. Using Compute Unifled Device Architecture (CUDA), the FADI-FDTD with CFS-CPML is further incorporated into the GPU to exploit data par- allelism. Numerical results show that a much higher e-ciency gain of up to 15 times can be achieved.