MachSMT: A Machine Learning-based Algorithm Selector for SMT Solvers

Abstract In this paper, we present MachSMT, an algorithm selection tool for Satisfiability Modulo Theories (SMT) solvers. MachSMT supports the entirety of SMT-LIB language. It employs machine learning (ML) methods to construct both empirical hardness models (EHMs) and pairwise ranking comparators (PWCs) over state-of-the-art SMT Given formula $$\mathcal {I}$$ I as input, leverages these learnt output a solvers based on predicted run time . We evaluate solvers, benchmarks, data obtained from SMT-COMP 2019 2020. observe frequently improves competition winners, winning $$54$$ 54 divisions outright up $$198.4$$ 198.4 % improvement in PAR-2 score, notably logics that have broad applications (e.g., BV, LIA, NRA, etc.) verification, program analysis, software engineering. The is designed be easily tuned extended any suitable solver application by users. not replacement means. Instead, it enables users leverage collective strength diverse set algorithms implemented part sophisticated