Soot-based implementation of a demand-driven reaching definitions analysis

Abstract

As a classical data-flow analysis, reaching definitions analysis is the corner stone of various techniques, such as code optimization and program slicing. The built-in data-flow analysis framework in Soot has been implemented in the traditional iterative style. While being able to meet general requirements for implementation of data flow analyses, the framework may be less efficient for a certain type of analyses in which the complete data-flow solution is unnecessary. In this paper, we introduce our Soot-based implementation of an inter-procedural demand-driven reaching definitions analysis. For a demand for reaching definitions facts, the analysis only explores relevant program points and variables, saving a considerable amount of computation. Preliminary results show that the implementation can be much more efficient than its traditional counterpart in several scenarios. The Soot framework has greatly facilitated the implementation by providing abundant basic analysis results via well designed APIs.