An Example of Deriving Performance Properties from a VisualRepresentation of Program Execution

Abstract

This paper demonstrates that one can use a visual representation of the execution of a program to analyze performance properties. The method starts with a visualization of the progress of processes, and then applies geometric properties of the visualization to analyze program execution behavior. The visualization is based on progress graphs from the literature, which map the progress of each process to one Cartesian graph axis. Line segments represent interprocess synchronization. A directed, continuous path that does not cross a segment represents a particular execution of a program, and can be found using computational geometric algorithms. A special case is also considered in which programs display periodic behavior. The relation of the program class studied to Petri nets is also examined.