VADA: A transformation-based system for variable dependence analysis

Abstract

Variable dependence is an analysis problem in which the aim is to determine the set of input variables that can affect the values stored in a chosen set of intermediate program variables. This paper shows the relationship between the variable dependence analysis problem and slicing and describes VADA, a system that implements variable dependence analysis. In order to cover the full range of C constructs and features, a transformation to a core language is employed Thus, the full analysis is required only for the core language, which is relatively simple. This reduces the overall effort required for dependency analysis. The transformations used need preserve only the variable dependence relation, and therefore need not be meaning preserving in the traditional sense. The paper describes how this relaxed meaning further simplifies the transformation phase of the approach. Finally, the results of an empirical study into the performance of the system are presented.