论文部分内容阅读
提出了在由微机互连构成的机群(COW)并行计算系统上应用信息传递的方式实现粗糙地面散射并行FDTD算法。综合考虑了区域分割和负载平衡因素,并详细分析了子区域在普通网格和吸收边界处与相邻子区域的场值的数据传递,提高了二维粗糙地面FDTD并行计算效率。解决了在计算电大尺寸粗糙地面散射时产生的内存不足和计算耗时长等瓶颈问题。理论分析和数值计算结果验证了该算法的正确性;当计算电大尺寸的粗糙地面散射时,并行效率提升明显,即当参与计算的处理器数量达到6个时,并行效率仍然可以保持在90%以上。
This paper presents a parallel FDTD algorithm for rough surface scattering based on the application of information transfer in the parallel computing system consisting of computer interconnects. Considering the factors of regional segmentation and load balancing, the data transfer between the sub-region and field values of adjacent sub-regions at normal grid and absorption boundary is analyzed in detail, and the parallel FDTD computational efficiency of two-dimensional rough terrain is improved. It solves the bottleneck problem of lack of memory and time-consuming calculation when calculating the scattering of rough surface of electric large size. The theoretical analysis and numerical results verify the correctness of the proposed algorithm. The parallel efficiency is obviously improved when the coarse scattering of electric large size is calculated. That is, when the number of processors involved in computation reaches 6, parallel efficiency can still be maintained at 90% the above.