Program Analysis with Dynamic Change of Precision

Dirk Beyer, Thomas A. Henzinger, and Gregory Theoduloz

We present and evaluate a framework and tool for combining multiple program analyses which allows the dynamic (on-line) adjustment of the precision of each analysis depending on the accumulated results. For example, the explicit tracking of the values of a variable may be switched off in favor of a predicate abstraction when and where the number of different variable values that have been encountered has exceeded a specified threshold. The method is evaluated on verifying the SSH client/server software and shows significant gains compared with predicate abstraction-based model checking.

Proceedings of the 23rd International Conference on Automated Software Engineering (ASE), ACM Press, 2008, pp. 29-38.

Download inofficial, sometimes updated PostScript / PDF document. © 2008 ACM.