Ranking Abstractions

  • Aziem Chawdhary ,
  • Byron Cook ,
  • ,
  • Mooly Sagiv ,
  • Hongseok Yang

Programming Languages and Systems. ESOP 2008. Lecture Notes in Computer Science. |

Published by Springer, Berlin, Heidelberg

Publication | Publication | Publication

We propose an abstract interpretation algorithm for proving that a program terminates on all inputs. The algorithm uses a novel abstract domain which uses ranking relations to conservatively represent relations between intermediate program states. One of the attractive aspects of the algorithm is that it abstracts information that is usually not important for proving termination such as program invariants and yet it distinguishes between different reasons for termination which are not usually maintained in existing abstract domains. We have implemented a prototype of the algorithm and shown that in practice it is fast and precise.