Applying Program Visualization Techniques to Aid Parallel and Distributed Program Development

Parallel and distributed programming is intrinsically more difficult than sequential programming, yet few effective tools or methodologies have been developed to help programmers understand the behavior of their parallel programs. Browsing source code and tracing program output are tedious and often ineffective approaches for parallel program understanding. Program visualization, which relates a program's behavior to the programmer's model of the system's components and interactions, has been shown to be a novel and highly effective approach to program and algorithm comprehension. Extending and adapting program visualization to parallel programming can aid comprehension of the complex concurrent events and transitions that occur in parallel programs. We are defining a model for the capture and display of parallel program events and transitions, based upon the path-transition paradigm for animation, and partial ordering of events. Using this model, we are developing a prototype for visualizing parallel programs, and testing the model and prototype upon a suite of scientific parallel programs