Program Flow Graph Construction for Static Analysis of Explicitly Parallel Message-Passing Programs

摘要

In recent years, message-passing parallel codes have rallied around using the message passing interface (MPI). The parallelism in these codes is most often explicit; the developer must instrument the source code with calls to an optimized communications runtime library. MPI has been widely used for developing efficient and portable parallel programs, in particular for distributed memory multiprocessors and workstation/personal computer (PC) clusters, although its use in shared memory systems has been equally effective. This report presents algorithm for building a program flow graph representation of an MPI program. As an extension of the control flow graph representation of sequential codes, this representation provides a basis for important program analyses useful in software testing, debugging tools, and code optimization.