In this paper approaches to the evaluation of metaphors and view of Software Visualization systems are considered on the examples of representation of call graphs and execution traces of parallel programs. Execution traces map the dynamics of the certain program executions. Visualization and “replaying” of execution traces are important elements of debugging systems. The visual presentations of the Call Graph are widely used in the systems parallel program performance tuning systems. The visualization metaphors using to depict call graphs and execution traces are surveyed. “Traditional” (for example the City metaphor) and new (for example the Brain metaphor) metaphors are considered. The validity of visualization techniques is studied on basis of analysis of metaphor properties. It is important to understand what objects may be represented with one or another metaphor. The possibility of the visualization metaphors (for example City and Molecule metaphors) to represent large and huge volumes of data is analyzed. Shneiderman’s scheme is considered as method of evaluation of visualization. B. Shneiderman presents seven high level users needs that an information visualization application should support. (Overview, Zoom, Filter, Details-on-demand, Relate, History, Extract) The use of the Schneiderman's scheme presupposes the existence of large structured data volumes. Further some other approaches to the evaluation of metaphors and views are considered. Among them are the criterion mental structure conservation and the evaluation of system implementation efforts. The problems of Software Visualization for parallel computing are observed. The formalization of parallel computing and in particular performance tuning may be useful to resolve problems of Software Visualization.Ключевые слова: Software Visualization, Parallel Computing, call graphs, execution traces, visualization metaphors, Shneiderman’s scheme.
Предварительный вариант статьи
Вернуться в раздел Научная визуализация и визуализация программного обеспечения для параллельных вычислений