摘要： 对星载合成孔径雷达(SAR)并行处理算法在分布式共享存储器(DSM)HPC平台下的实现作了深入研究,对比了用消息传递和OpenMP两种并行编程模型实现的并行方案,在此基础上提出了基于进程的共享变量并行模型。这种模型克服了前两种模型的缺点,经过实验测试和实际SAR成像应用,证明是一种高效、稳定的并行方案。

关键词: 合成孔径雷达, 高性能计算, 并行成像算法, 共享变量

Abstract: This paper examines the implementation issues of SAR parallel processing algorithm on distributed shared memory high performance computing(HPC )platforms. By comparing the parallelization scheme which uses message passing model with that uses OpenMP model, it implements a shared memory parallelization scheme based on processes. This model eliminates the drawbacks of the former models and is proved to be a high efficiency and stable parallel scheme by experiment and actual SAR imaging application.

中图分类号: 

• TN958