空时自适应处理的通用平台设计与实现

该文利用多个高性能数字信号处理器，结合FPGA和通用处理器，实现了一个空时自适应处理(STAP)的通用实时平台系统。借鉴Valiant(1990)提出的BSP模型, 采用多重流水线，提出一个空时自适应处理(STAP)计算模型。该模型可以弥补STAP算法和实际并行系统的差距，为开发提供了统一框架；同时，方便了对算法的性能评估。在基于该模型的具体开发过程中，选择可扩展簇式多处理机结构作为系统硬件架构，采用数据块静态分配方案进行算法的分解与映射，并采取一系列通信和程序优化的方法。结果表明，系统能满足实时要求，可扩展性好，方便类似系统的系列开发。

A Universal Real-Time Platform for Space-Time Adaptive Processing

A universal real-time platform for a Space-Time Adaptive Processing (STAP) is developed. The platform is composed of multi-DSPs. FPGA and a general-purpose processor. Refined from the Bulk Synchronous Parallel model (BSP) by Valiant(1990), an STAP computation model is brought forward. The model provides a bridge between STAP algorithms and real parallel systems. Moreover, it can be applied to performance evaluation. During the course of development, scalable cluster-organized multi-processors structure is adopted as hardware architecture. And data-block static allocating is taken as the mapping scheme. Afterwards, some optimization methods about communication and programming are introduced. This system can meet the real-time requirement, its scalability is good, and it facilitates the development of similar systems.