Skip to main content

MIP root node suboptimal termination


1 comment

  • Matthias Miltenberger
    Gurobi Staff Gurobi Staff

    Hi Jennifer,

    This is a hard problem! You are using concurrent optimization for the root node relaxation, so primal simplex, dual simplex and barrier are running in parallel. Apparently, barrier is faster than the two simplex algorithms, which is not too surprising given the size of the problem. Still, it cannot solve the LP to optimality but is terminated earlier and Gurobi uses the crossover technique to first find a basic solution and then applies additional simplex steps to reach optimality.

    So, you don't need to worry about suboptimal termination here, because the infeasibilities are cleaned up afterwards.

    Did you try more aggressive presolving? While this may make the model numerically more difficult, it could also lead to a smaller problem size.

    Since the barrier seems to be fastest, you should consider using method=2 to disable the parallel simplex runs and provide the barrier with two more cores.



Please sign in to leave a comment.