[Help-glpk] FW: trying to stop solver

[Din Rao]

Dear glpk help,

Here is the screenshot of glpsol trying to solve an LP (no binary or integer variables are present):

lpt_read_prob: reading LP data from `lpoct9_noint.txt'...
lpt_read_prob: 186719 variables, 58727 constraints
lpt_read_prob: 93665 lines were read
lpx_simplex: original LP has 58727 rows, 186719 columns, 476946 non-zeros
lpx_simplex: presolved LP has 56188 rows, 157793 columns, 417847 non-zeros
lpx_adv_basis: size of triangular part = 56163

      0:   objval =  -1.882780893e+05   infeas =   1.000000000e+00 (0)
    200:   objval =  -8.302602730e+04   infeas =   9.287234663e-01 (0)
    400:   objval =  -5.596677665e+04   infeas =   9.188069813e-01 (0)
    600:   objval =  -5.005017544e+04   infeas =   7.717075583e-01 (0)
    800:   objval =  -2.835981613e+04   infeas =   7.477124460e-01 (0)
   1000:   objval =  -2.822934262e+04   infeas =   7.320822590e-01 (0)
-----------------------------------------------------------------------------------------------------------------------------------------------

 

This process goes on and on and on..................

 

Please advise as to how to stop the solver from making NEGLIGIBLE progress AFTER the infeasibilities have reached 0.

 

I have noticed that the "objval" gets only less then 1% better after every 200 iterations AFTER the infeasibility has reached 0.

 

Notice that the LP is not exactly large and our UNIX server comparable to 800 MHz

(Only 476,946 vars and 58,727 constraints) 

 

Thank you,

Regards

 

Din Rao

Senior Engineer

Concerto Software

Bethesda, MD 20171

(301) 272-2251

www.cforcetech.com

-----Original Message-----
From: Din Rao [mailto:address@hidden]
Sent: Thursday, April 29, 2004 11:25 AM
To: glpk-GNU Help; Phillip Warner; Robert Horvath
Subject: trying to stop solver
Importance: High


I have glpk-4.4 for UNIX.

we noticed that even not so large LPs were taking forever to solve.

A problem with 450,000+ variables, 58,000+ constraints took more than 5 hours without offering a solution.

The infeasibilities were all 0 (ZERO), but still, the solver was running trying to get the best solution.

Is there anyway to stop the solver just a few iterations after the infeasibilities of 0 is reached?

Is there some code I have to hack? or some option in glpsol.exe?

Regards,

Din Rao
Senior Engineer
Concerto Software
(301) 272-2251
Bethesda, MD 20171