[Help-glpk] Re: glpk does not release all the memory?

[Andrew Makhorin]

>I've noticed tht GLPK consistently leaves 90 bytes of memory unfreed. 
>Even after calling glp_lib_fee_env().  It's not that important, but it
>rings a warning in my debugger, so that's a minor nuisance.  I've not
>been able to check were do these bytes come from.  Is it difficult to
>fix?

With probability 0.999999 there is no bug in glp_lib_free_env. I think
these 90 bytes were allocated by a non-glpk routine (for example, some
functions from the C run-time library can do that).

>Now just an unrelated question: what's the best way to discard redundant
>linear constraints in a linear problem using glpk?  Looking at the
>documentation I've not found a direct way to do that.  Maybe someone on
>the list have faced a similar problem when having a big number of
>constraints in a problem, most of them repeated or saying nothing. 

Removing redundant constraints is a presolver's task, so the best way
is to use a presolver. I'm going to finish an initial version of the lp
presolver soon (in 5-6 weeks) and include it in glpk.

Btw, redundant constraints being non-active during all the optimization
process don't affect significantly the performance of the simplex method.