Computational insights into solving optimization instance sets with MIP solvers on multi-core hardware

High-performance computing systems offer ever-increasing core counts, yet most mixed-integer programming (MIP) solvers still rely on limited shared-memory parallelism with only modest speedups beyond a few cores. Many decision-support applications, however, require solving large collections of MIP instances. This talk reports on computational experiments with leading MIP solvers for managing such workloads on modern multi-core CPUs, offering practical guidance on thread allocation, degrees of job-level parallelism, and solver configurations to improve throughput and energy efficiency.