Research on the GPU implementation of the FDTD method for thetwo-dimensional phononic band structure calculations

FDTD (Finite-difference time-domain) method is a commonly used method to compute the phononic crystal band gap. Due to the requirements of calculation precision and numerical stability, FDTD method generally requires a close space grid and smaller time step, when calculating the structure of the larger scale, which consume large amounts of computing resources, lead to the calculation more time-consuming. And based on CUDA (calculation Unified device architecture) multi-threaded parallel computing solution, can effectively applied the FDTD algorithm to the graphics processor (GPU). We present the technical implementation details, at the same time. compared with the traditional CPU efficiency, The results show that with spatial grid density increase, the GPU relative to the CPU to enhance the efficiency of calculation is also becoming distinct. for 3000 * 3000 spatial grid, computing speed compare to CPU can improve even more than 10 times.