基于OpenMP与VALU硬件加速的表面积分方程矩量法混合并行求解技术

A Hybrid Parallelization Technique Based on OpenMP and VALU Acceleration for the Method of Moments Solution of the Surface Integral Equations

  • 摘要: 研究在共享内存式计算机架构下,基于OpenMP及矢量算术逻辑单元(VALU)硬件加速的表面积分方程矩量法的混合并行求解技术. 讨论了矩量法并行程序设计中的关键问题,并分析了影响并行计算的主要因素. 针对一些典型散射目标,给出了由混合并行程序得到的雷达散射截面数值结果. 通过对计算数据的对比分析,指出负载平衡及内存读取连续性是影响混合并行效率的关键问题所在.

     

    Abstract: In the framework of the shared-memory computer system, a hybrid parallelization technique based on OpenMP and VALU acceleration for the method of moments (MoM) solution of the surface integral equations (SIE) is proposed. In addition, several key issues in the hybrid parallel MoM code design and the major factors influencing the parallel efficiency are addressed in detail. Numerical results of the radar cross section of typical objects show that load-balance and memory-read continuity are the key factors influencing the parallel efficiency.

     

/

返回文章
返回
Baidu
map