MPP系统模型化的广义随机Petri网技术  

GENERALIZED STOCHASTIC PETRI NET TECHNOLOGY OF MODELLING MPP SYSTEMS

作　　者：管惠维[1,2] 李志光[1,2] 

机构地区：[1]上海大学计算机工程系 [2]香港理工大学电子工程系

出　　处：《小型微型计算机系统》1997年第3期34-41,共8页Journal of Chinese Computer Systems

基　　金：上海市重点学科发展基金;上海市高教局科学发展基金;香港理工大学研究基金

摘　　要：本文提出一种模型化海量并行处理（ＭＰＰ）系统的广义随机Ｐｅｔｒｉ网技术ＧＳＰＮ。首先，通过外延优先权标记Ｐｅｔｒｉ网，提出了ＧＳＰＮ的定义、时序规范、变迁激发规则和计算公式。然后，讨论使用ＧＳＰＮ模型化ＭＰＰ系统的技术途径和实现方案。分别阐述了ＧＳＰＮ形式化描述共享存储器方式和信息传递方式ＭＰＰ系统的有效性，并给出了实例。此外，我们还分析和提出了ＣＳＰ（通信顺序进程）理论中主要进程与Ｐｅｔｒｉ网之间的映射关系。使用ＧＳＰＮ，能够方便地模型化ＭＰＰ结构，清晰地刻划和描述系统的动态操作和状态，为ＭＰＰ系统的设计、性能描述。A generalized stochastic petri net technology which can be used to model massively parallel processing systems, named as GSPN, is presented. At first, the definition and temporal specification of GSPN are described by extending the Marked Petri Net with Priorities, the firing rules of transition and calculus formula are given. Then, the technical approach and implementation methods of modelling MPP systems are described by using GSPN.The effectiveness of this modelling technique in the cases of share-memory structures and message passing architectures are discussed respectively, and the associated instances are given. In addition, the mapping relationship between main processes of CSP(Communicating Sequential Processes) and petri net is proposed. Using GSPN can conveniently model MPP architecture, precisely describe the dymantic operations and states of system, and it provide a new approach to model, specify, analyze and evaluate MPP systems.

关 键 词：PETRI网 并行处理系统 模型化 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]