Interprocedural Dataflow Analysis in the Presence of Large Libraries

Interprocedural dataflow analysis has a large number of uses for software optimization, maintenance, testing, and verification. For software built with reusable components, the traditional approaches for whole-program analysis cannot be used directly. This paper considers component-level analysis of a main component which is built on top of a pre-existing library component. We propose an approach for computing summary information for the library and for using it to analyze the main component. The approach defines a general theoretical framework for dataflow analysis of programs built with large extensible library components, using pre-computed summary functions for library-local execution paths. Our experimental results indicate that the cost of component-level analysis could be substantially lower than the cost of the corresponding whole-program analysis, without any loss of precision. These results present a promising step towards practical analysis techniques for largescale software systems built with reusable components.