Demand-Driven Construction of Call Graphs

Call graph construction has been an important area of research within the compilers and programming languages community. However, all existing techniques focus on exhaustive analysis of all the call-sites in the program. With increasing importance of just-in-time or dynamic compilation and use of program analysis as part of the software development environments, we believe that there is a need for techniques for demand-driven construction of the call graph. We present a demanddriven call graph construction framework in this paper, focusing on the dynamic calls due to polymorphism in object-oriented languages. We use a variant of Callahan’s Program Summary Graph (PSG) and perform analysis over a set of influencing nodes. We show that our demanddriven technique has the same accuracy as the corresponding exhaustive technique. The reduction in the graph construction time depends upon the ratio of the cardinality of the set of influencing nodes to the set of all nodes.