摘要
为简化并行FDTD算法实现,提出一种新的基于映射文件技术的解决方法。对于并行FDTD算法的两个关键环节:场值数组定义和子域间通信在该方法中借助操作系统的映射与重定向功能来实现,由于屏蔽了子域间显式的通信操作,将大幅度降低并行FDTD算法实现的复杂度。数值模拟实例和相关的性能比较验证了新方法的可行性与便利性,所得加速比、并行效率等性能指标参数与通常的MPI消息传递方法基本相当。
To simplify implementation of parallel FDTD algorithm, a novel approach based on mapped file technique is presented in this paper. Two key problems of the define of the EM field array and the communication between sub-domains in the implementation of parallel FDTD algorithm are solved by using the mapping and redirecting function of the operation system. By shielding the explicit message delivering operation, the complexity of parallel FDTD algorithm can reduced to a great extent. Some computed examples and comparisons with other methods are provided to prove the feasibility and facility of the presented approach. Numerical simulations show that the performances such as speedup ratio, parallel efficiency of the presented approach are near to those of the conventional method using MPI.
出处
《电波科学学报》
EI
CSCD
北大核心
2008年第4期634-639,共6页
Chinese Journal of Radio Science
基金
国家自然科学基金资助项目(60571020)
关键词
并行FDTD算法
内存映射数组
内存文件通信
并行性能
parallel FDTD algorithm
memory-mapped array
file communicating in memory
parallel performance
