GPU-Accelerated Primal Heuristics for Mixed Integer Programming

We introduce a fusion of GPU accelerated primal heuristics for Mixed Integer Programming. Leveraging GPU acceleration enables exploration of larger search regions and faster iterations. A GPU-accelerated PDLP serves as an approximate LP solver, while a new probing cache facilitates rapid roundings and early infeasibility detection. Several state-of-the-art heuristics, including Feasibility Pump, Feasibility Jump, and Fix-and-Propagate, are further accelerated and enhanced. The combined approach of these GPU-driven algorithms yields significant improvements over existing methods, both in the number of feasible solutions and the quality of objectives by achieving 221 feasible solutions and 22% objective gap in the MIPLIB2017 benchmark on a presolved dataset.