Refinement in object-sensitivity points-to analysis via slicing
summary
Describes a points-to refinement technique built on k-object-sensitive analysis for Java that answers a given set of client queries by first running a cheap, low-k analysis and then locating imprecise (“bad”) points-to facts. It uses a program-slicing-style procedure to find allocation sites that likely cause those bad facts and re-runs object-sensitivity with longer object-name bounds only for those sites, implemented in Datalog on Petablox and proven complete, yielding substantial precision gains over baseline k-object-sensitivity and a recent SAT-based refinement.
AI-written summary · about