Abstract

Program analysis crucially depends on our ability to symbolically describe and manipulate sets of program behaviours in a mechanical way. Separation logic provides a promising foundation for reasoning about dynamic memory allocation in computer programs, but the development of efficient inference tools for separation logic remains a challenging problem. In this talk I will describe how my work led to the design of an efficient, sound and complete automated theorem prover for checking the validity of entailments in a fragment of separation logic.

Our prover is built from a modular integration of separation logic techniques—-to reason about the shape of structures in memory—-and superposition calculus—-to reason about equality/aliasing between program variables. I will also present some empirical evaluation of the efficiency of this procedure, which translated into speedups of several orders of magnitude with respect to previous work, and discuss the directions in which I’m currently working to further generalise these techniques.