[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Help-glpk] computational efforts, message level and KKT check

From: Ali Baharev
Subject: Re: [Help-glpk] computational efforts, message level and KKT check
Date: Wed, 1 Dec 2010 17:48:26 +0100

Dear Andrew,

Thank you for the detailed explanation.

> I decided not to include routines like lpx_get_parm/lpx_set_parm in the
> new api, because most of control parameters in the old api are solver
> properties rather than model properties. Probably such statistics could
> be reported through the parameter block passed to the solver, for
> example, glp_simplex could report the iteration count and other
> statistics through glp_smcp structure.

I am testing sophisticated algorithms and I have to track
computational efforts. This information is crucial for me.

Getting this info from the glp_smcp structure is good for me. Please
implement it.

> This bug was fixed. The messages from glp_scale_prob may appear only if
> you call it explicitly, i.e. out of glp_simplex, in which case you could
> use glp_term_out to disable undesirable output.

Is there a reason why the explicit call still produces output?

I still find it a bit confusing that the scaling routine does not obey
the message level.

> Lpx_check_kkt is the only routine that still has no equivalent in the
> new api, because I would like to replace it with several api routines to
> facilitate its usage; besides, the math used there needs some
> improvements.

I am looking forward to it a lot!

Many thanks,


reply via email to

[Prev in Thread] Current Thread [Next in Thread]