Regression verification for unbalanced recursive functions

Ofer Strichman, Maor Veitsman

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review


We address the problem of proving the equivalence of two recursive functions that have different base-cases and/or are not in lockstep. None of the existing software equivalence checkers (like RÊVE, RVT, SYMDIFF), or general unbounded software model-checkers (like SEAHORN, HSFC, AUTOMIZER) can prove such equivalences. We show a proof rule for the case of different base cases, based on separating the proof into two parts—inputs which result in the base case in at least one of the two compared functions, and all the rest. We also show how unbalanced unrolling of the functions can solve the case in which the functions are not in lock-step. In itself this type of unrolling may again introduce the problem of the different base cases, and we show a new proof rule for solving it. We implemented these rules in our regression-verification tool RVT. We conclude by comparing our approach to that of Felsig et al.’s counterexample-based refinement, which was implemented lately in their equivalence checker RÊVE.

Original languageEnglish
Title of host publicationFM 2016
Subtitle of host publicationFormal Methods - 21st International Symposium, Proceedings
EditorsConstance Heitmeyer, Anna Philippou, Stefania Gnesi, John Fitzgerald
Number of pages14
StatePublished - 2016
Event21st International Symposium on Formal Methods, FM 2016 - Limassol, Cyprus
Duration: 9 Nov 201611 Nov 2016

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume9995 LNCS


Conference21st International Symposium on Formal Methods, FM 2016

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)


Dive into the research topics of 'Regression verification for unbalanced recursive functions'. Together they form a unique fingerprint.

Cite this