Debugger Visualizations for Shared-Memory Multiprocessors

  • Authors:
  • Cherri M. Pancake;Sue Utter

  • Affiliations:
  • -;-

  • Venue:
  • Debugger Visualizations for Shared-Memory Multiprocessors
  • Year:
  • 1991

Quantified Score

Hi-index 0.00

Visualization

Abstract

The production of error-free code has always been challenging, but the incorporation of parallelism makes the task truly formidable. Although the idea of exploiting multiple processors in order to decrease wall-clock time is simple conceptually, understanding the complexity of parallel behavior is not. This paper reviews recent developments in parallel debuggers, focusing on techniques for program behavior visualization. The exploitation of graphical techniques enhances the partnership between user and tool, making better use of human creative and analytic capabilities. Two general visualization approaches are contrasted. The first characterizes behavior by the flow of execution control, portraying changes in process state or sequences of lower-level events. The second employs the framework of data structures to describe patterns of access to shared data. Each approach is appropriate in certain debugging situations. Experience suggests that parallel debuggers need a variety of techniques and levels of presentation if they are to provide comprehensive support for applications programmers.