Computation Improvement for the Rigorous Coupled-wave Analysis with GPU

The rigorous coupled-wave analysis (RCWA) technique needs to obtain the exact solution of Maxwell´s equations, a huge number of matrix operations may be involved in the process. The graphics process unit (GPU) has inherent advantage in large-scale parallel computation compared with CPU. In this paper, matrix operations are put on GPU using NVIDIA´s Compute Unified Device Architecture (CUDA). It is shown that by exploitation of the General Purpose Computing based on GPU, several-fold speed up in the numerical efficiency is produced. The numerical experiment is run on NVIDIA´s GPU GTX460, and the results show that GPU performs better than CPU.