Exploiting Linearity in Sharing Analysis of Object-oriented Programs

We propose a new sharing analysis of object-oriented programs based on abstract interpretation. Two variables share when they are bound to data structures which overlap. We show that sharing analysis can greatly benefit from linearity analysis. We propose a combined domain including aliasing, linearity and sharing information. We use a graph-based representation of aliasing information which naturally encodes sharing and linearity information, and define all the necessary operators for the analysis of a Java-like language.