[Help-glpk] Re: question about glpk's mip capabilities

[Andrew Makhorin]

> if we know an integer-feasible solution to a mip, is there a way
> to communicate this to the solver?  we are encountering situations
> wherein the solver takes far too long to get an upper bound; as you
> can imagine, this is frustrating when we happen to have a valid
> solution.  is there a way to communicate this solution to glpk? 
> more generally, what is the best we can make of a good (but not
> necessarily optimal) solution to a mip? 

I am working on a new implementation of glpk mip solver, which (among
other things) allows the user to provide mip solutions found with
a primal heuristic. I hope to include the new implementation in the
next version of the package.

In principle, you can provide the solver with an integer feasible
solution in the same way as the routine record_solution does (see
file glpmip02.c). If you are interested in details, please contact me.

Andrew Makhorin