GPU-based acceleration of MPIE/MoM matrix calculation for the analysis of microstrip circuits

In this paper, we present a GPU-based algorithm which accelerates the MoM impedance matrix computation. Based on an efficient quasi-one-dimensional approximation of the reaction integrals, the MPIE formulation for the analysis of microstrip circuits is considered. We use NVIDIA CUDA as GPU development tool and choose an edge-connected line-fed patch antenna as reference problem. In order to demonstrate the validity of our work, we first show how the results provided by the GPU implementation, the corresponding CPU code and a well-known commercial simulator are indistinguishable. Second, we conduct a performance analysis that proves how the developed algorithm running on a cheap off-the-shelf GPU is about 17 times faster than the CPU code.