[Bug-glpk] [Fwd: Simplex termination 32 vs. 64 bit]

[Andrew Makhorin]

-------- Forwarded Message --------
From: Horst Schirmeier <address@hidden>
To: address@hidden
Subject: Simplex termination 32 vs. 64 bit
Date: Sat, 7 Apr 2012 18:56:09 +0200

Hi,

for the attached CPLEX-format ILP I'm observing a different termination
behaviour on 32-Bit (doesn't terminate within the observed time frame of
several hours) vs. 64-Bit (terminates after 0.4s) Linux.  I'm running
unpatched glpsol 4.47 (.tar.gz taken from gnu.org, built from sources)
on Debian 6.0:

Debian 6.0, 32-Bit:
$ glpsol --lp input.cplex
GLPSOL: GLPK LP/MIP Solver, v4.47
Parameter(s) specified in the command line:
 --lp input.cplex
Reading problem data from `input.cplex'...
332 rows, 2301 columns, 6903 non-zeros
2301 integer variables, all of which are binary
7228 lines were read
GLPK Integer Optimizer, v4.47
332 rows, 2301 columns, 6903 non-zeros
2301 integer variables, all of which are binary
Preprocessing...
314 rows, 2301 columns, 4602 non-zeros
2301 integer variables, all of which are binary
Scaling...
 A: min|aij| =  1.000e+00  max|aij| =  3.000e+00  ratio =  3.000e+00
Problem data seem to be well scaled
Constructing initial basis...
Size of triangular part = 314
Solving LP relaxation...
GLPK Simplex Optimizer, v4.47
314 rows, 2301 columns, 4602 non-zeros
      0: obj =   3.372000000e+03  infeas =  2.430e+02 (0)
*   315: obj =   3.153000000e+03  infeas =  0.000e+00 (0)
*   500: obj =   1.321000000e+03  infeas =  1.092e-15 (0)
*   851: obj =   5.020000000e+02  infeas =  7.401e-17 (0)
OPTIMAL SOLUTION FOUND
Integer optimization begins...
+   851: mip =     not found yet >=              -inf        (1; 0)
+  1437: >>>>>   5.040000000e+02 >=   5.020000000e+02   0.4% (50; 0)
+  1631: >>>>>   5.030000000e+02 >=   5.020000000e+02   0.2% (65; 21)
+ 40041: mip =   5.030000000e+02 >=   5.020000000e+02   0.2% (51; 2232)
+ 76140: mip =   5.030000000e+02 >=   5.020000000e+02   0.2% (137; 4202)
+116304: mip =   5.030000000e+02 >=   5.020000000e+02   0.2% (116; 6415)
[... continues for hours without ever terminating ...]

Debian 6.0, 64-Bit:
$ glpsol --lp input.cplex
GLPSOL: GLPK LP/MIP Solver, v4.47
Parameter(s) specified in the command line:
 --lp input.cplex
Reading problem data from `input.cplex'...
332 rows, 2301 columns, 6903 non-zeros
2301 integer variables, all of which are binary
7228 lines were read
GLPK Integer Optimizer, v4.47
332 rows, 2301 columns, 6903 non-zeros
2301 integer variables, all of which are binary
Preprocessing...
314 rows, 2301 columns, 4602 non-zeros
2301 integer variables, all of which are binary
Scaling...
 A: min|aij| =  1.000e+00  max|aij| =  3.000e+00  ratio =  3.000e+00
Problem data seem to be well scaled
Constructing initial basis...
Size of triangular part = 314
Solving LP relaxation...
GLPK Simplex Optimizer, v4.47
314 rows, 2301 columns, 4602 non-zeros
      0: obj =   3.372000000e+03  infeas =  2.430e+02 (0)
*   315: obj =   3.153000000e+03  infeas =  0.000e+00 (0)
*   500: obj =   1.488000000e+03  infeas =  0.000e+00 (0)
*   844: obj =   5.020000000e+02  infeas =  8.327e-17 (0)
OPTIMAL SOLUTION FOUND
Integer optimization begins...
+   844: mip =     not found yet >=              -inf        (1; 0)
+  2394: >>>>>   5.040000000e+02 >=   5.020000000e+02   0.4% (42; 0)
+  3560: >>>>>   5.030000000e+02 >=   5.030000000e+02   0.0% (37; 88)
+  3560: mip =   5.030000000e+02 >=     tree is empty   0.0% (0; 171)
INTEGER OPTIMAL SOLUTION FOUND
Time used:   0.4 secs
Memory used: 2.1 Mb (2252288 bytes)

The input data is attached to this email.  Is it somewhat malformed, or
did I indeed step on a GLPK bug?

Kind regards,
 Horst Schirmeier

input.cplex.bz2
Description: Binary data