Improving the Precision of Equality-Based Dataflow Analyses

We present two new, orthogonal techniques for improving the precision of equality-based dataflow analyses. Subtype expansion models objects at a per-type granularity, enabling a form of subtype-restricted equality constraint, while mutation tracking uses a simple effect analysis to avoid a class of false aliases induced by the bidirectional nature of equality constraints. The utility and costs of these techniques are demonstrated in a context-sensitive interprocedural optimization whose static precision improves by 6-600% when our techniques are applied.