非交叉语义并行计算模型及其语义组合规则

The trouble of interleaving assumption comes from the complete ignorance of intermediate states. It considers only how the initial states and the final states are related. The reason why many parallel computation models adopt interleaving assumption is that they have overlooked the interconnection of statements produced by shared variables. The net-shaped structures of Petri nets are just the means to reveal such interconnections, and furthermore, the local determinism principle of Petri nets provides a foundation for non-interleaving semantics. UNINET is a parallel computation model based on Petri nets. It enjoys non-interleaving semantics and provides a way for semantic combination. We have successfully made use of UNINET in analyzing the phenomina of inheritance anomaly of OO techniques. It has also been a success in workflow research to model process logic and to give semantics of workflow processes. It is a sound foundation for developing formal techniques in workflow analysis. PUNITY is another parallel computation model designed based on UNITY, which is suitable for parallel program design. We have developed an experimental transformation tool from PUNITY to UNINET and a simulator of UNINET. This will make it easier to map a PUNITY program to a given hardware architecture.

交叉语义假设是并行计算模型共同的语义基础，其合理性却没有得到论证。人们追求并行程序系统的组合语义描述，迄今也没有满意的结果。上述现象的根本原因是缺乏对并行程序系统静态结构的准确描述。本课题以Petri网结构为基础，结合面向对象技术和UNITY程序设计方法，提出一种不依赖交叉语义假设并具有语义组合性质的并行计算模型。.