Scaling CFL-Reachability-Based Points-To Analysis Using Context-Sensitive Must-Not-Alias Analysis

Pointer analyses derived from a Context-Free-Language (CFL) reachability formulation achieve very high precision, but they do not scale well to compute the points-to solution for an entire large program. Our goal is to increase significantly the scalability of the currently most precise points-to analysis for Java. This CFL-reachability analysis depends on determining whether two program variables may be aliases. We propose an efficient but less precise pre-analysis that computes context-sensitive must-not-alias information for all pairs of variables. Later, these results can be used to quickly filter out infeasible CFL-paths during the more precise points-to analysis. Several novel techniques are employed to achieve precision and efficiency, including a new approximate CFL-reachability formulation of alias analysis, as well as a carefully-chosen trade-off in context sensitivity. The approach effectively reduces the search space of the points-to analysis: the modified points-to analysi...

• Type: Conference
• Year: 2009
• Where: ECOOP
• Authors: Guoqing Xu, Atanas Rountev, Manu Sridharan
• Keywords: Alias Analysis, ECOOP 2009, Precise Points-to Analysis, Programming Languages, Reachability Formulation Achieve