Cumulative Computing

In this paper we use the concept of resource cumulation to model various forms of computation. The space of cumulations (called a cumulator) is simply represented as a five tuple consisting of a well-founded partial order, a monoid and a volume function. The volume function is introduced to simplify reasoning about limit points and other topological properties. A specification command is a set of cumulations. Typical phenomena of concurrency such as reactiveness, safety and liveness, fairness, real time and branching time naturally arise from the model. In order to support a programming theory, we introduce a specification language that incorporates sequentiality, nondeterminism, simple parallelism, negation and general recursions. A new fixpoint technique is used to model general recursions. The language is applied to the case study on CSP, which becomes a special model of cumulative computing with a combination of four resource cumulators of alphabet, termination, trace and refusal. All laws of cumulative computing are also valid for CSP and the generalization from CSP to Timed CSP can be achieved by simply combining the four cumulators with real time. Loops whose bodies may take zero time can then be modeled more satisfactorily.