lazybvtoint at the SMT Competition 2020

Abstract

lazybv2int is a new prototype SMT-solver, that will participate in the incremental and non-incremental tracks of the QF BV logic.Overview. lazybv2int is a prototype SMT-solver for the theory of fixed-width bit-vectors and uninterpreted functions. This is the first year it will compete in SMT-COMP. It will participate in the incremental and non-incremental QF BV tracks. The basic algorithm of the tool relies on a translation from bit-vectors to non-linear integer arithmetic with uninterpreted functions, followed by a CEGAR loop [7] that lazily instantiates bit-vector axioms over the translation. The idea of using integer reasoning for bit-vector solving is not new (see, e.g., [1,4]), however, it is worth revisiting due to recent improvements in solvers for non-linear integer arithmetic [5,8,10]. We expect this solver to perform better on benchmarks that involve arithmetic bit-vector constraints and large bit-widths because the encoding of arithmetic constraints is straightforward and independent of bit-width, as opposed to the encoding of bit-wise constraints which is less natural and hindered by larger bit-widths. The tool is open-source and is available at https://github.com/yoni206/lazybv2int.