Journal of Systems Engineering and Electronics ›› 2009, Vol. 31 ›› Issue (4): 874-880.

• 系统工程 • 上一篇    下一篇

STAP并行处理系统的调度问题研究

任磊1,2, 王永良2, 陈辉2, 陈建文2   

  1. 1. 国防科技大学电子科学与工程学院, 湖南, 长沙, 410073;
    2. 空军雷达学院兵器运用工程军队重点实验室, 湖北, 武汉, 430019
  • 收稿日期:2008-04-21 修回日期:2008-09-03 出版日期:2009-04-20 发布日期:2010-01-03
  • 作者简介:任磊(1980- ),男,博士研究生,主要研究方向为阵列信号处理及其实时实现技术.E-mail:r18849@163.com
  • 基金资助:
    国家自然科学基金重点基目(60736009);总装“十一五”预研项目(513070503)资助课题

Research on the scheduling problems of STAP parallel processing system

REN Lei1,2, WANG Yong-liang2, CHEN Hui2, CHEN Jian-wen2   

  1. 1. School of Electronic Science and Engineering, National Univ. of Defense Technology, Changsha 410073, China;
    2. Key Laboratory of Radar Armament and Utilization Engineering, Air Force Radar Academy, Wuhan 430019, China
  • Received:2008-04-21 Revised:2008-09-03 Online:2009-04-20 Published:2010-01-03

摘要: 为了研究空时自适应处理(STAP)并行系统中的调度问题,以系统数据和任务为研究对象,以时延和吞吐率为目标函数参数,以均衡性和扩展性为系统设计理念,首先研究了STAP系统的数据划分和任务映射方法。然后提出了任务映射模型,建立了任务流程粗粒度有向非循环图(DAG),进一步分析了STAP系统中的两个重要性能指标。并基于以性能指标为参数的目标函数,研究了STAP系统中的调度策略,提出的五步映射法能够较好地均衡系统负载,满足系统的设计要求。最后给出了一组基于异构处理系统的测试基准和实际STAP系统的实现过程,验证了该研究的有效性。

Abstract: To study the scheduling problems of space-time adaptive processing(STAP) parallel system,the system data and tasks are investigated as objects,the latency and throughput are taken as parameters in the objective function,the proportionality and expansibility of system are considered.The data partition and tasks mapping methods are studied firstly,the tasks mapping model and the tasks coarse grain directed acyclic graph(DAG) are proposed for the analysis of the two performance indexes,latency and throughput.The scheduling strategies are studied based on the objective function with the two performance indexes as parameters,and a five-step mapping method is proposed to balance the computational load of the STAP system.Finally,a set of tasks benchmarks of heterogeneous processing board is given,and a real STAP processor is implemented to verify the validity of the study.

中图分类号: