The "category" of program structure and its applications in parallelization

We consider the parallelization of large numerical software. First, we introduce the concept of category of program structure and its features. Second, we concentrate on the category's roles in the parallelization of large numerical software. Finally we apply ideas of the category (and tag-array) to the programming of PRIS-1.0, a parallel reservoir numerical simulator, and give results of one million grid-data problems from the Daqing field with PRIS-1.0 on distributed memory PC-clusters (16 cpu PIII500) of R&DCPS. We cost no more than 60 hours on 16 PIII500s solving 1160 thousand grid-data, matching the 31 year history problem.