goSAT: Floating-point satisfiability as global optimization

Abstract

We introduce goSAT, a fast and publicly available SMT solver for the theory of floating-point arithmetic. We build on the recently proposed XSat solver [1] which casts the satisfiability problem to a corresponding global optimization problem. Compared to XSat, goSAT is an integrated tool combining JIT compilation of SMT formulas and NLopt, a feature-rich mathematical optimization backend. We evaluate our tool using several optimization algorithms and compare it to XSat, Z3, and MathSat. Our evaluation demonstrates promising results.