Visualizing program execution using user abstractions

Abstract

A practical system that uses visualization for understanding the execution of complex programs must offer the user views of the execution that are specific to the program being understood and the particular problem at hand without significantly slowing down or perturbing the system being examined. This paper describes a visualization data model and its implementation that accomplishes this. The model starts with program events that can be instrumented efficiently and with little overhead. It uses extended finite state automata to model program behaviors in terms of events. It builds time-varying data structures based on these automata. The data structures are then made available to various visualizations through a set of mappings that let the user dynamically control the visualization. The model and its implementation have been demonstrated on a range of specific understanding problems with a variety of different visualizations. © 2006 by the Association for Computing Machinery, Inc.