Speaker: Sudeep Chowdhary
Topic: SkipFlow: Improving the Precision of Points-to Analysis using Primitive Values and Predicate Edges (CGO ‘25)
Description: Classical points-to analysis techniques like Andersen’s or Steensgaard’s ignore the branching structure of the program being analysed due to which they may lose some precision. By keeping some additional information like primitive values along with traditional points-to sets, the precision may be increased if only the branches that are possible at runtime are analysed. The authors of this paper propose a novel lightweight and scalable points-to analysis called SkipFlow that interprocedurally tracks the flow of both primitives and objects, and explicitly captures the branching structure of the code using predicate edges. They have applied SkipFlow to GraalVM Native Image, a closed-world solution to building standalone binaries for Java applications. For the evaluation of SkipFlow, they show a reduction in the size of the applications/benchmarks like Dacapo and Renaissance in terms of reachable methods by 9% on average without significantly increasing the analysis time.
Time 5 pm to 6 pm (IST)
Presenter: Sudeep Chowdhary
About the Presenter: https://www.linkedin.com/in/sudeep-chowdhary/