数据项的并发处理是大幅度提高处理速度的有效途径。在诸如图象处理与天气预报计算等许多实时应用场合中,需要每秒执行10亿条以上浮点指令的运算速率。这样,就不可避免地要采用并发处理。当前,随着大规模集成技术的日益发展,将几百个甚至几千个市售的处理机与存储器模块互连起来,以构成一个并发处理系统是现实可行的,也是经济的。 The concurrent processing of data items is an effective way to significantly improve the processing speed. In many real-time applications such as image processing and weather forecasting computations, computational rates of more than 1 billion floating-point instructions per second are required. In this way, it is inevitable to use concurrent processing. At present, with the increasing development of large-scale integration technology, it is realistic and economical to interconnect hundreds and even thousands of commercially available processors and memory modules to form a concurrent processing system.