Release notes

What’s new in Knitro 13.2 ?

  • Knitro 13.2 offers improved robustness and efficiency when solving both linear and nonlinear MIP models through improved cuts, cut selection strategies, and other enhancements to the branch-and-bound method.

  • Knitro 13.2 offers several presolve improvements including improvements in bound and coefficient tightening, and detection of redundant constraints. A new user option presolveop_redundant was added to control the detection and removal of redundant constraints.

  • Knitro 13.2 offers improved heuristics for nonconvex mixed-integer problems through improved multistart procedures.

  • Knitro 13.2 adds the new mip_gomory user option for setting Gomory mixed-integer cuts.

  • Knitro 13.2 adds a default auto option setting for mip_clique, mip_knapsack, mip_liftproject and mip_zerohalf options. It also adds a new root setting and deprecates the nlp setting for the mip_mir option, and updates the option values for the mip_knapsack option.

  • Knitro 13.2 updates options for the Tuner when applied to mixed-integer models.

Bug Fixes in Knitro 13.2.0

  • Fixed issue that could cause Knitro to sometimes return an incorrect threadID value in user callbacks when using multi-threaded parallel features in Knitro.

  • Fixed issue that could cause a segfault on rare occasions when using the feasibility pump in the Knitro MIP solver.

  • Fixed issue with invalid option when setting mip_selectrule =1 on MIP models.

  • Fixed issue applying a quasi-Newton Hessian to MIP models.

  • Fixed issue that could cause a segfault when using the Knitro Active-Set algorithm (algorithm =3), or barrier crossover (bar_maxcrossit >0) on a model with complementarity constraints.

  • Fixed issue that could cause different Knitro behaviors depending on the setting of the outlev option.

  • Fixed issue solving model with nonlinear objective, but no nonlinear constraints when using MATLAB problem-based API.