ALPACA: A Large Portfolio-Based Alternating Conditional Analysis

Abstract

Program analysis tools typically compute either may or must information. By accumulating both kinds of information computed with respect to different portions of a program's state space, it is possible to collect a comprehensive view of how program inputs relate to some property. This can be done using the framework of an alternating conditional analysis (ACA). In this paper, we present a toolset that instantiates an ACA to analyze C programs. The toolset, dubbed ALPACA (A Large Portfolio-based ACA), computes a sound characterization of all the ways a program either may or must satisfy some property. It does so by alternating between over-and underapproximate analyses, conditioning analyses to ignore portions of the program that have already been analyzed, and combining the results of 14 state-of-the-art analysis tools in a portfolio run in parallel. Download ALPACA at https://bitbucket.org/mgerrard/alpaca. Its video demonstration is at https://youtu.be/H2yXtvODurQ.