Optimizing Pointer Analysis Using Bisimilarity

We introduce a new technique for dramatically improving the performance of inclusion-based points-to analysis, by using bisimilarity in order to detect pointer equivalences before constraint resolution. We present the design and correctness proof of this technique, along with an implementation prototype, and a series of benchmarks. The benchmarks indicate that our technique dramatically improves the scalability of inclusion-based points-to analysis, beating the current leading offline optimizations for inclusion-based points-to analysis.