Re: [Help-glpk] restart interior point method

[Andrew Makhorin]

>I am using the interior point method to solve LPs which is much faster than
>the simplex method for the problem I am considering.
>I must solve the same problem several times with different coefficients for
>the objective function. The IP method does make a symbolic factorisation
>(well, that is what is printed during the preprocessing, before the
>iterations begin) that takes a long time.
>
>Is it possible to keep this work between problems (in an object that is
>passed to the solver like for glp_simplex1 ?
>It would save me a lot of time as the iterations themselves are quit fast.

Unfortunately you can't do that just because this feature is not
available on api level. On the other hand it could be easily done but
requires some hacking; so if you will decide to hack 'glpipm1.c', please
inform me and I will provide you with all necessary information.

Did you try glp_simplex2 instead glp_simplex1 ? This api routine also
allows restarting from the current basis, but for many lp problems it
works faster than glp_simplex1 and sometimes much faster than the
interior point solver.