[Clp] Infinite loop in Clp

John J Forrest jjforre at us.ibm.com
Mon May 5 11:02:51 EDT 2008






Kim,

Even the run with an mps file does not look quite right.

a) Refresh your version of trunk - many changes have been made.
b) use writeMpsNative with formatType 2 to get an mps file with exactly the
same values

If you want you can send the mps fiel to me and I can look at run.

John Forrest


                                                                                                                           
  From:       "Kim Hansen" <kimhanse at gmail.com>                                                                            
                                                                                                                           
  To:         Clp <clp at list.coin-or.org>                                                                                   
                                                                                                                           
  Date:       05/05/2008 08:21 AM                                                                                          
                                                                                                                           
  Subject:    [Clp] Infinite loop in Clp                                                                                   
                                                                                                                           





I have a problem where Clp ends up in an infinite loop. I can
reproduce the loop when using my own program, but if I dump the
problem as mps and then reload it the problem is gone.

I am using trunk from 2008-04-18.

The code that calls Clp is something like:
  OsiClpSolverInterface solver;
  CoinPackedMatrix* matrix = new CoinPackedMatrix(true, rowIndices,
colIndices, elements, nzmax);
  matrix->setDimensions(rows, cols);
  solver.assignProblem(matrix, collb, colub, obj, rowlb, rowub);
  solver.setHintParam(OsiDoPresolveInInitial, true, OsiHintDo);
  solver.setHintParam(OsiDoCrash, true, OsiHintDo);
  solver.setHintParam(OsiDoDualInInitial, useDual(), OsiHintDo);
  solver.initialSolve();
  result = solver.getColSolution();

The log output when running the code on data from an mps file:
Coin0506I Presolve 109 (-75) rows, 120 (-137) columns and 1898 (-1121)
elements
Clp0028I Crash put 1 variables in basis, 0 dual infeasibilities
Clp0014I Perturbing problem by 0.001 % of 4.49537 - largest nonzero
change 1.44776e-05 (% 28.8042) - largest zero change 0.00122885
Clp0006I 0  Obj 4.78356e+11 Primal inf 1.22108e+18 (70) Dual inf 0.0031232
(4)
Clp0006I 31  Obj 1.28479e+15 Primal inf 1.74561e+07 (40)
Clp0001I Primal infeasible - objective value 1.4751e+15
Coin0505I Presolved problem not optimal, resolve after postsolve
Coin0511I After Postsolve, objective 1.60904e+15, infeasibilities -
dual 308431 (44), primal 1.88861e+14 (106)
Clp0006I 0  Obj 1.60904e+15 Primal inf 8.85833e+13 (47) Dual inf
3.41142e+13 (67)
Clp0029I End of values pass after 56 iterations
Clp0006I 56  Obj 1.29172e+11 Primal inf 1.48165e+08 (12) Dual inf
3.66647e+12 (20)
Clp0006I 172  Obj 7.49029e+06
Clp0000I Optimal - objective value 7.49029e+06

The log output with the loop, I have deleted a lot of Clp0006I lines:
Coin0506I Presolve 109 (-75) rows, 120 (-137) columns and 1898 (-1121)
elements
Clp0028I Crash put 1 variables in basis, 0 dual infeasibilities
Clp0014I Perturbing problem by 0.001 % of 3.32746 - largest nonzero
change 1.33111e-05 (% 26.7419) - largest zero change 0.00112984
Clp0006I 0  Obj 6.13795e+11 Primal inf 1.52176e+18 (70) Dual inf 0.00292548
(4)
Clp0006I 4  Obj 4.735e+06 Primal inf 3.25522e+18 (71) Dual inf 4703.47 (49)
Clp0029I End of values pass after 6 iterations
Clp0006I 6  Obj 4.735e+06 Primal inf 2.84103e+18 (69) Dual inf 4703.45 (47)
Clp0006I 59  Obj 4.735e+06 Primal inf 194552 (26)
Clp0006I 61  Obj 4.735e+06 Primal inf 194431 (26)
Clp0006I 63  Obj 4.735e+06 Primal inf 194050 (26) Dual inf 2.21203e-08 (1)
Clp0006I 63  Obj 4.735e+06 Primal inf 194050 (26)
Clp0006I 65  Obj 4.735e+06 Primal inf 193623 (26) Dual inf 3.06918e-07 (4)
[...]
Clp0006I 1330  Obj 4.735e+06 Primal inf 5.06975e+06 (48) Dual inf 457543
(45)
Clp0006I 1333  Obj 4.735e+06 Primal inf 4.59528e+06 (45) Dual inf 270058
(42)
Clp0006I 1337  Obj 4.735e+06 Primal inf 4.28724e+06 (41) Dual inf 53506.1
(37)
Clp0003I Stopped - objective value 2.59719e+09
Clp0006I 0  Obj 3.34228e+14 Primal inf 4.70856e+15 (51) Dual inf
1.74932e+12 (57)
Clp0006I 3  Obj 3.34233e+14 Primal inf 4.70856e+15 (48) Dual inf
1.93552e+11 (58)
Clp0006I 4  Obj 2.79111e+14 Primal inf 2.84194e+15 (47) Dual inf
2.11687e+11 (53)
[...]
Clp0006I 440  Obj 1.079e+14 Primal inf 7.1993e+06 (30) Dual inf 6.74896e+12
(21)
Clp0006I 449  Obj 4.87475e+13 Primal inf 6.6127e+06 (28) Dual inf
2.2495e+12 (20)
Clp0014I Perturbing problem by 0.001 % of 1.33856e+14 - largest
nonzero change 0.0685696 (% 0.00108395) - largest zero change 0
Clp0006I 461  Obj 4.86248e+13 Primal inf 6.63103e+06 (30) Dual inf
6.74836e+12 (24)
Clp0006I 464  Obj 4.87524e+13 Primal inf 6.61232e+06 (27) Dual inf
6.72637e+08 (15)
[...]
Clp0006I 2677  Obj 4.88298e+13 Primal inf 6.61218e+06 (26) Dual inf
4.53415e+08 (12)
Clp0006I 2678  Obj 4.88784e+13 Primal inf 6.61213e+06 (25) Dual inf
2.37719e+08 (11)
Clp0006I 2679  Obj 4.88847e+13 Primal inf 6.61213e+06 (24) Dual inf
2.55378e+07 (10)
Clp0003I Stopped - objective value -1.2448e+25
looks like real trouble - too many iterations in second clean up - giving
up
Clp0006I 0  Obj 1.079e+14 Primal inf 7.1993e+06 (41) Dual inf 2.99908e+12
(46)
Clp0006I 14  Obj 1.11277e+14 Primal inf 6.87105e+06 (27) Dual inf
3.14631e+08 (36)
Clp0006I 61  Obj 8.62841e+13 Primal inf 6.15898e+06 (21) Dual inf
2.42078e+07 (29)
[...]
Clp0006I 456  Obj 8.59858e+13 Primal inf 5.86308e+06 (23) Dual inf
2.34096e+08 (25)
Clp0006I 457  Obj 8.72097e+13 Primal inf 5.86245e+06 (22) Dual inf
1.53374e+08 (24)
Clp0014I Perturbing problem by 0.001 % of 1.33856e+14 - largest
nonzero change 3.35563e-05 (% 0.000298454) - largest zero change 0
Clp0006I 459  Obj 9.35777e+13 Primal inf 5.85974e+06 (20) Dual inf 77.832
(22)
Clp0006I 461  Obj 4.77082e+13 Primal inf 5.79746e+06 (25) Dual inf
1.9956e+12 (19)
[...]
Clp0006I 1034165  Obj 1.21543e+14 Primal inf 6.177e+06 (24) Dual inf
1.09929e+12 (22)
Clp0006I 1034166  Obj 1.23426e+14 Primal inf 6.02884e+06 (23) Dual inf
4.08519e+08 (24)
Clp0006I 1034167  Obj 1.2531e+14 Primal inf 6.02736e+06 (22) Dual inf
3.10325e+08 (23)
Clp0006I 1034169  Obj 1.36331e+14 Primal inf 6.02122e+06 (20) Dual inf
1.23784e+08 (21)
Clp0006I 1034173  Obj 4.30608e+13 Primal inf 5.9391e+06 (27) Dual inf
3.89958e+08 (18)
Clp0006I 1034174  Obj 4.39853e+13 Primal inf 5.93852e+06 (26) Dual inf
2.956e+08 (17)
Clp0006I 1034175  Obj 4.46011e+13 Primal inf 5.93821e+06 (25) Dual inf
2.04861e+08 (16)
Clp0006I 1034176  Obj 4.60052e+13 Primal inf 5.93668e+06 (24) Dual inf
1.0353e+08 (15)
Clp0006I 1034178  Obj 1.10442e+14 Primal inf 6.80034e+06 (27) Dual inf
4.39535e+12 (25)
Clp0006I 1034179  Obj 1.14629e+14 Primal inf 6.54034e+06 (26) Dual inf
3.29677e+12 (24)
Clp0006I 1034180  Obj 1.2146e+14 Primal inf 6.18597e+06 (25) Dual inf
2.19857e+12 (22)
Clp0006I 1034181  Obj 1.21543e+14 Primal inf 6.177e+06 (24) Dual inf
1.09929e+12 (22)
[...]

--
Kim Hansen
Vadgårdsvej 3, 2.tv
2860 Søborg
Fastnet: 3956 2437 -- Mobil: 3091 2437

_______________________________________________
Clp mailing list
Clp at list.coin-or.org
http://list.coin-or.org/mailman/listinfo/clp
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/clp/attachments/20080505/9476bd87/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: graycol.gif
Type: image/gif
Size: 105 bytes
Desc: not available
URL: <http://list.coin-or.org/pipermail/clp/attachments/20080505/9476bd87/attachment.gif>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ecblank.gif
Type: image/gif
Size: 45 bytes
Desc: not available
URL: <http://list.coin-or.org/pipermail/clp/attachments/20080505/9476bd87/attachment-0001.gif>


More information about the Clp mailing list