Program analysis for compiler validation

  • Authors:
  • Anna Zaks;Amir Pnueli

  • Affiliations:
  • New York University, New York;New York University, New York

  • Venue:
  • Proceedings of the 8th ACM SIGPLAN-SIGSOFT workshop on Program analysis for software tools and engineering
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

Translation Validation is an approach of ensuring compilation correctness in which each compiler run is followed by a validation pass that proves that the target code produced by the compiler is a correct translation (implementation) of the source code. It has been previously shown that the problem of translation validation can be reduced to checking if a single system - the corss-product of the source and target, satisfies a specific property. In this paper, we show how to adapt the existing program analysis techniques in the setting of translation validation. In addition, we present a novel invariant generation algorithm which strengthens our analysis when the input programs contain dynamically allocated data structures. Finally, we report on the prototype tool that applies the developed methodology to verification of the LLVM compiler. The tool handles many of the classical intraprocedural compiler optimizations such as constant folding, reassociation, common subexpression elimination, code motion, dead code elimination, and others.