Finite Difference Time Domain Method Based on GPU for Solving Quickly Maxwell’s Equations

Abstract

AbstractBased onGraphics Processing Units (GPUs) as main computational core, the Finite Difference Time Domain(FDTD) is presented for solving quickly Maxwell’s equations for electromagnetic. Firstly, the FDTD algorithm is developed by analyzing the direct time-domain solution like FDTD to Maxwell’s curl equations. Then, it is analyzed and compared with CPUs that how GPUs can be used to greatly speedup FDTD simulations, so enormous computation problem in FDTD simulations is resolved. At last, leveraging GPU processing power for FDTD update calculations, researchers can simulate much longer pulse lengths and larger models than was possible in the past, and computationally expensive simulations are completed in reasonable time. It is proved that FDTD simulations based on GPUs is accurate and high efficiency compared with CPUs.KeywordsGraphic Processing UnitFinite Difference Time DomainData ParallelismFinite Difference Time Domain SimulationLarge Model SpaceThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.