> I want to set overall simplex parameters for an integer (linear) > optimization. I am little bit confused, that I “only” can use > glp_simplex (where I can set parameters like pricing, meth, etc.) > prior to the integer optimization. But then the MILP has no presolve > option and I seems that the simplex with standard settings (as far as > I can interpret the code/output) is applied during branch-and-cut. The > documentation doesn’t tell me more details about that.
> So, I thought to ask community for some explanation. Did it get the > situation right and might there be future enhancements? I am not a > native c developer, but after I inspected the code a little bit, it > might be possible to pass a smcp optionally also to integer > optimization or make parts of the parameters (at least meth, pricing) > to be available through iocp. I could imagine an overall performance > jump if dual simplex can be fully used during integer optimization for > a variety of models. In the mip solver the simplex solver is used internally and, strictly speaking, it is not available to the user. For example, the presolve option normally used in the simplex solver should not be used on reoptimization performed by the mip solver, because: i) it would be inefficient; ii) the mip solver has an internal presolver for this purpose. Of course, some options could be specified in iocp to pass to the simplex solver, but currently there is no need in that. Please note also that both the simplex and mip solvers are still under development. _______________________________________________ Help-glpk mailing list Help-glpk@gnu.org https://lists.gnu.org/mailman/listinfo/help-glpk
