【摘要】 提出了一种用于并行程序符号调试系统的分布式监控方法。首先在并行程序中插入可监控的＂探针＂，并将并行程序的受控运行分为用户状态、活跃状态、通讯状态和等待状态，在分析这些状态的基础上设计了对并行程序进行分布式管理的监控算法。这一算法已经在ＰＤ－１００并行计算机上实现，实践表明这种监控机制是简便而有效的．更多还原

【Abstract】 Ref.[2] does not offer any information for monitoring 3L Parallel FORTRAN. To meet this monitoring nced, the authors have developed a method that is generally applicable to monitoring various parallel programs. In fact, due to the great difficulties involved, there exist, to the authors’ best know]cdgC, only special systems for monitoring specific parallel software such as GMAT of Cray and INSTANT of BBN.The authors’ strategy can be essentially described by what follows. The authors place a great number of probes into the user’s parallel programs. These probes can monitor any trouble occuring in any part of the parallel programs when it is running and report through a special communication network to a special management system. Thus the users can discover any trouble that occurs when running the parallel programs. As a step towards solving the problems associated with dynamic tracing of parallel programs, the authors developed an appropriate state transition mechanism for tasks. Recently the authors’ method was examined by a group of Chinese experts organizied by the funding agency. The method was deemed to be novel, dependable, and useful.更多还原