high performance computing on graphics processing units: hgpu.org

In this paper, we employ the programmable graphics processing unit (GPU) to accelerate the IPO computation for analyzing the scattering of open cavities. Since the iterative strategy accounts for multiple reflections on the inner wall, the IPO method provides a more accurate solution than the other high frequency asymptotic methods. However, it suffers from a large requirement of simulation time. To date, the GPU featuring inherent parallelism and powerful floating-point capability has become an attractive alternative to the central processing unit (CPU) for some of computation tasks. Therefore, we map the time-consuming parts of IPO into the graphics hardware following the stream programming model, and the CPU carries out the computation of the far field scattering. In addition, the numerical results demonstrate the accuracy and effectiveness of our proposed method.