GPU-Accelerated Parallel Finite-Difference Time-Domain Method for Electromagnetic Waves Propagation in Unmagnetized Plasma Media

摘要

The finite-difference time-domain (FDTD) method has been commonly utilized inthe numerical solution of electromagnetic (EM) waves propagation through theplasma media. However, the FDTD method may bring about a significant incrementin additional run-times consuming for computationally large and complicated EMproblems. Graphics Processing Unit (GPU) computing based on Compute UnifiedDevice Architecture (CUDA) has grown in response to increased concern forreduction of run-times. We represent the CUDA-based FDTD method with theRunge-Kutta exponential time differencing scheme (RKETD) for the unmagnetizedplasma implemented on GPU. In the paper, we derive the RKETD-FDTD formulationfor the unmagnetized plasma comprehensively, and describe the detailedflowchart of CUDA-implemented RKETD-FDTD method on GPU. The accuracy andacceleration performance of the posed CUDA-based RKETD-FDTD method implementedon GPU are substantiated by the numerical experiment that simulates the EMwaves traveling through the unmagnetized plasma slab, compared with merelyCPU-based RKETD-FDTD method. The accuracy is validated by calculating thereflection and transmission coefficients for one-dimensional unmagnetizedplasma slab. Comparison between the elapsed times of two methods proves thatthe GPU-based RKETD-FDTD method can acquire better application accelerationperformance with sufficient accuracy.