Abstract

Speaker Alexey Gotsman

In recent years many ideas initially proposed in program logics have been applied to program analysis. For example, ideas from separation logic and its derivatives have been used to design analyses for checking a variety of properties (such as memory safety and noninterference) of programs with different porgramming models (such as procedures and concurrency). However, to date there has been no proper understanding of connections between such analyses and the logics they were inspired by. In this talk I will highlight conceptual and technical controversies arising from this lack of understanding. In order to resolve them, I will propose a method of co-designing logics and analyses that makes the connection between them explicit, enables using results of analyses in interactive theorem proving or proof-carrying code systems, and avoids giving separate soundness proofs to logics and corresponding analyses. As a case study, I will consider a modular analysis for concurrent programs inspired by concurrent separation logic and show that establishing its underlying logic leads to an interesting mathematical challenge: it requires constructing a non-standard model of concurrent separation logic in which resource invariants are not required to be precise and the conjunction rule does not hold.