Learning Splitting Heuristics for Parallel String Solvers

The paper presents a data-driven approach for automatically generating splitting heuristics in parallel string solvers, specifically implemented in Z3seq and Z3str4. By framing the selection of splitting atoms as a learning task, the method utilizes features from input formulas and dynamic solver execution data, resulting in improved performance over manually designed heuristics in terms of both the number of solved formulas and average solving time. This advancement is significant for practitioners as it enhances the efficiency of string constraint solving in multi-core environments, addressing the challenges posed by complex and undecidable constraints in real-world applications.