Conflict Points for Better Prefixes

Concurrent systems are usually studied using interleaving semantic models, which consider the events of a system to be totally ordered. An alternative to these models is to use non-interleaving (true concurrency) models, which consider events to be just partially ordered. An important exponent of the latter approach was introduced by McMillan and later refined by several others. It embodies the construction of the complete finite prefix of the (usually) infinite unfolding of a concurrent system. The algorithm to construct such a complete finite prefix is parameterized by a special kind of partial order on the configurations of the system. Historically, these orders have been defined on the basis of the syntactical structure of the system description. This means that it is impossible to have control on the structure of a prefix. For efficiently using a prefix in reachability analysis and model checking, it is very important to be able to generate prefixes with a “nice” structure. Therefore, in this paper, we present the family of such orders generated by a special enumeration of the nodes of an unfolding. The concept of conflict point is introduced and aids the definition of this family. Conflict points mark the actual diversions between configurations. This has the advantage that the adequate order does no need to be specified a priori, but instead can be constructed “on the fly”. Experimental evidence is provided.