摘要
在Sun工作站群,SunHPCClusterTools系统和MPI消息传递接口环境下,研究了天波超视距雷达信号处理的并行化,对其两个核心部分二维FFT(距离-多普勒变换)和数字波束形成实现了并行.通过不同粒度并行方案的比较发现,粗粒度方案在MPI环境下具有较好的并行效果,同时在现有硬件和软件环境下进行了方案的比较、选择、可扩展性的分析和设计.实验结果表明,并行处理方案在数据处理的速度和实时性方面有较大提高.
This paper discussed parallel signal processing for high frequency over-the-horizon radar via cluster of workstations (COW), Sun HPC ClusterTools and message passing interface (MPI). There are two key parts of the signal processing algorithms, two-dimensional FFT (Range-Doppler transform) and digital beamforming (DBF). Of the different computing load distribution schemes, the coarse-grained scheme has better performance for MPI, and the scalability of the algorithm was analyzed and discussed for the special hardware and software configure. The simulation results demonstrate that the speed and performance of the parallel scheme are improved.
出处
《上海交通大学学报》
EI
CAS
CSCD
北大核心
2004年第2期236-239,247,共5页
Journal of Shanghai Jiaotong University
关键词
超视距雷达
信号处理
并行计算
工作站群
Computer simulation
Parallel processing systems
Signal processing
Two dimensional
