Program comprehension support for knowledge-based parallelization

Current compilation systems for distributed memory computers have to integrate new techniques to support the highly complex task of producing efficient programs for parallel systems. Two techniques, program comprehension and expert systems, although developed outside the scope of parallelization domain, are extremely useful to improve the quality of the parallel code generated and to make the parallelization process more convenient and automatic. We describe a parallelization environment consisting of three main components: Vienna Fortran Compilation System (VFCS), a tool for recognition of Parallelizable Algorithmic Patterns (PAP Recognizer), and a knowledge based parallelization support tool (Expert Adviser). After these main components are introduced, the paper focuses on integration issues of PAP Recognizer and Expert Adviser within the framework of VFCS. We outline the salient features of a new parallelization environment. The design of the XPA knowledge base for recognized program concepts (patterns) as presented, and the methodology of knowledge acquisition for program patterns is outlined.