Abstract

Handling procedures effectively is one of the most basic problems in static program analyses. A good method for analysing procedures can improve the precision and performance of a static analysis substantially. However, obtaining such a method is nontrivial and requires new insights into the type of the analysis and target programs.

In this talk, I will describe a few lessons that I recently learnt about interprocedural program analyses, while working with colleagues in the US and Korea. We found a new relationship between popular approaches for analysing procedures, which helped us to form a big picture about these approaches. We also came up with new ways of combining or adjusting existing interprocedural analyses, which led to the improvement in precision or performance in our experiments. I will explain these findings.

This is joint work with Mayur Naik, Ravi Mangal, Xin Zhang in Georgia Tech and Wonchan Lee, Hakjoo Oh, Kwangkeun Yi in Seoul National University.