MPI和Pthread混合编程在NCS算法中的设计和实现

根据星载合成孔径雷达（Synthetic Aperture Radar，SAR）实时和准实时成像的需求，论文设计和实现了一种混合编程模式的非线性CS（Nonlinear Chirp Scaling， NCS）成像算法。此方法基于SMP（Symmetric Multiprocessing）集群系统，利用MPI（Message Passing Interface）和Pthread混合编程，实现了多进程和多线程的两级流水线并行成像处理。通过在HP刀片系统上的实验验证和结果分析，这种混合编程的流水线NCS并行成像算法既能较好适应前端系统传输一景原始回波数据时间不小于8s的情况，又能较好适应大规模原始回波数据堆积并且等待处理的情况。

Parallel Imaging of MPI and Pthread Mixed Programming Based on Cluster

The mixed programming for Nonlinear Chirp Scaling (NCS) algorithm was designed according to the requirements of space borne Synthetic Aperture Radar (SAR) real time and near real time imaging. The two level pipelined parallel imaging of multi process and multi thread with Message Passing Interface (MPI) and Pthread was researched and implemented on SMP cluster. Through the experiments on HP cluster, it is proved that the mixed program imaging algorithm can not only adapt to the instance that the time of one block of original echo data transferred from the front system is more than or equal to eight seconds, but also the instance that a great number of original echo data has been already accumulated.