邸瀚漪, 杨明林, 盛新庆. 电大均匀介质目标三维散射的并行多层快速多极子计算[J]. 电波科学学报, 2016, 31(4): 695-699. doi: 10.13443/j.cjors.2015110601
      引用本文: 邸瀚漪, 杨明林, 盛新庆. 电大均匀介质目标三维散射的并行多层快速多极子计算[J]. 电波科学学报, 2016, 31(4): 695-699. doi: 10.13443/j.cjors.2015110601
      DI Hanyi, YANG Minglin, SHENG Xinqing. An efficient parallel approach of MLFMA for solving 3D scattering by large homogeneous targets[J]. CHINESE JOURNAL OF RADIO SCIENCE, 2016, 31(4): 695-699. doi: 10.13443/j.cjors.2015110601
      Citation: DI Hanyi, YANG Minglin, SHENG Xinqing. An efficient parallel approach of MLFMA for solving 3D scattering by large homogeneous targets[J]. CHINESE JOURNAL OF RADIO SCIENCE, 2016, 31(4): 695-699. doi: 10.13443/j.cjors.2015110601

      电大均匀介质目标三维散射的并行多层快速多极子计算

      An efficient parallel approach of MLFMA for solving 3D scattering by large homogeneous targets

      • 摘要: 实现了计算电大均匀介质体散射问题的高效混合并行混合场积分方程(Electric and Magnetic Current Combined-Field Integral Equation, JMCFIE)求解, 在单纯消息传递接口(Message Passing Interface, MPI)并行基础上采用共享存储并行编程(Open Multi-Processing, OpenMP)进一步提升性能.该混合MPI与OpenMP的并行多层快速多极子技术通过灵活的进程和线程策略, 提升了负载平衡和可扩展性.数值实验展示了此混合MPI与OpenMP的并行多层快速多极子技术的计算能力, 计算了不同尺寸的电大目标体(包含一个半径120 m、1.1亿未知数目的介质球).

         

        Abstract: Based on the electric and magnetic current combined-field integral equation (JMCFIE), a fast algorithm is presented for calculating 3D scattering by large homogeneous objects. An efficient and flexible parallel implementation of the multilevel fast multipole algorithm (MLFMA) is employed to enhance efficiency and capability of the JMCFIE approach. This parallel approach is proposed by using open multi-processing(OpenMP) to further accelerate the pure MPI parallel MLFMA. This hybrid MPI-OpenMP paralleled MLFMA improves the load-balance and scalability by combing processes and threads flexibly, hence depress unavoidable load imbalance in the pure MPI parallel MLFMA. Numerical experiments on targets with different sizes including a dielectric sphere with radius 120 m modeled with about 110 million unknowns show the scalability and computational capability of the proposed MPI-OpenMP-MLFMA enhanced JMCFIE for extremely large targets.

         

      /

      返回文章
      返回