[Clp] status wrong using OsiClp

John J Forrest jjforre at us.ibm.com
Fri Jan 29 05:03:53 EST 2010


Matt,

Can you get ClpSimplex*.cpp from svn stable/1.11 and try again.  By luck I
managed to get something close to your problem.  It was on a branch where
(luckily) I had not put in the fix I did to get round the assert you also
mentioned.  I had everything in debug mode apart from one function which I
had copied across optimized.  Interestingly this really stressed the code!
I think the reason was that some computations were always 64 bit floating
point, while optimized can have some at 80 bits.  This seemed to set up an
accuracy problem.

Anyway I got the assert and tracking it down I found a serious (but
unlikely to occur) bug which looks like your one.

So try it and tell me if it helps.

John


                                                                                                       
  From:       Matthew Galati <Matthew.Galati at sas.com>                                                  
                                                                                                       
  To:         "clp at list.coin-or.org" <clp at list.coin-or.org>                                            
                                                                                                       
  Date:       01/26/2010 10:12 PM                                                                      
                                                                                                       
  Subject:    [Clp] status wrong using OsiClp                                                          
                                                                                                       
  Sent by:    clp-bounces at list.coin-or.org                                                             
                                                                                                       





This is based on CLP release 1.11.0.

If I dump to MPS file, CLP claims the problem is infeasible - which seems
correct (cplex agrees). However, if it is run through OsiClp using resolve,
it returns that the problem is solved optimally. The log is below. Can you
tell anything from the log? If you want to debug this directly - I can try
to send instructions on how to repeat the issue - it happens when using my
DIP project.

Thanks,
Matt



Running CLP:

[magala at orclus71 build-O]$ ./bin/clp
Dip/examples/MILPBlock/PRICE_AND_CUT_master
Prob.n0.c6.p14.lp
Coin LP version 1.11.0, build Jan 26 2010
command line - ./bin/clp
Dip/examples/MILPBlock/PRICE_AND_CUT_masterProb.n0.c6.p
14.lp
Presolve 119 (-17416) rows, 76 (-17538) columns and 1511 (-20448) elements
0  Obj -1.0275e+07 Primal inf 160.178 (48) Dual inf 5.21844e+07 (71)
77  Obj -1.70728e+07 Primal inf 339.779 (36)
85  Obj -6.75452e+06 Primal inf 3729.11 (19)
87  Obj 2.00406e+07 Primal inf 11780.8 (21)
90  Obj 2.64131e+07 Primal inf 13955 (22)
91  Obj 3.32559e+07 Primal inf 18130.8 (31)
92  Obj 8.34552e+07 Primal inf 43072.8 (31)
94  Obj 4.23954e+08 Primal inf 180013 (41)
95  Obj 9.20098e+08 Primal inf 385843 (39)
96  Obj 9.48195e+08 Primal inf 402516 (43)
97  Obj 8.08035e+09 Primal inf 3.45018e+06 (48)
98  Obj 8.08035e+09 Primal inf 3.46527e+06 (50)
99  Obj 1.33977e+10 Primal inf 5.74173e+06 (51)
99  Obj 1.33977e+10 Primal inf 5.74173e+06 (51)
99  Obj 1.33977e+10 Primal inf 5.74173e+06 (51)
99  Obj 1.33977e+10 Primal inf 5.74173e+06 (51)
99  Obj 1.33977e+10 Primal inf 5.74173e+06 (51)
99  Obj 1.33977e+10 Primal inf 5.74173e+06 (51)
99  Obj 1.33977e+10 Primal inf 5.74173e+06 (51)
99  Obj 1.33977e+10 Primal inf 5.74173e+06 (51)
Primal infeasible - objective value 1.33977e+10
Presolved problem not optimal, resolve after postsolve
After Postsolve, objective 1.33977e+10, infeasibilities - dual -2.5429e+12
(50),                                                                primal
1.236e+08 (319)
PrimalInfeasible objective 1.339766625e+10 - 198 iterations time 0.032,
Presolve
0.02




Running OsiClp:

Coin0513I 17416 rows dropped after presolve pass 1
Coin0513I 17416 rows dropped after presolve pass 2
Coin0506I Presolve 119 (-17416) rows, 76 (-17538) columns and 1511 (-20459)
elements
Clp1001I Initial range of elements is 0.0600375 to 11508
Clp1002I Range of elements is 0.0654858 to 15.2705
Clp1002I Range of elements is 0.108171 to 9.24462
Clp1003I Final range of elements is 0.112844 to 8.86181
Clp0022I Absolute values of scaled rhs range from 0.0927259 to 22.6474,
minimum gap 1e+100
Clp0020I Absolute values of scaled objective range from 10250.2 to 1.61378e
+07
Clp0021I Absolute values of scaled bounds range from 0.0710216 to 1.80219,
minimum gap 0.0710216
Clp0018I 1 total structurals rejected in initial factorization
Clp0006I 0  Obj -1.69954e+07 Primal inf 875.907 (53) Dual inf 2.62043e+06
(24)
Clp0006I 102  Obj -7.0894e+06 Primal inf 3599.86 (21)
Clp0006I 104  Obj -6.90525e+06 Primal inf 16284.6 (37)
Clp0010I Flagging variable C33
Clp0010I Flagging variable C61
Clp0010I Flagging variable R1
Clp0010I Flagging variable C34
Clp0010I Flagging variable R53
Clp0006I 106  Obj 8.66818e+06 Primal inf 40216.1 (34)
Clp0010I Flagging variable R114
Clp0010I Flagging variable R78
Clp0006I 107  Obj 8.66818e+06 Primal inf 40156.6 (34)
Clp0010I Flagging variable C48
Clp0010I Flagging variable R102
Clp0010I Flagging variable R38
Clp0006I 115  Obj 2.93615e+07 Primal inf 74000.2 (44)
Clp0010I Flagging variable C45
Clp0010I Flagging variable C62
Clp0006I 117  Obj 7.36568e+07 Primal inf 142995 (41)
Clp0010I Flagging variable R50
Clp0006I 119  Obj 1.7023e+08 Primal inf 293559 (38)
Clp0010I Flagging variable C21
Clp0006I 124  Obj 3.50591e+08 Primal inf 573525 (37)
...
Clp0010I Flagging variable C17470
maybe forcing re-factorization - sum 1.31472e+09  1 pivots
maybe forcing re-factorization - sum 1.43916e+09  2 pivots
Clp0060I Primal error 1.07374e+09, dual error 3.50794e+06
Clp0006I 145  Obj 1.23847e+20 Primal inf 1.98872e+24 (188) Dual inf 127.725
(1)
Clp0060I Primal error 1.07374e+09, dual error 3.50794e+06
Clp0060I Primal error 4.29497e+09, dual error 9.66162e+06
Clp0006I 148  Obj 1.29644e+20 Primal inf 2.08182e+24 (162) Dual inf
0.0783458 (1)
Clp0060I Primal error 1.07374e+09, dual error 9.66162e+06
maybe forcing re-factorization - sum 6.87571e+13  1 pivots
Clp0060I Primal error 16384, dual error 1.06329e+07
Clp0006I 150  Obj 1.37555e+20 Primal inf 2.20884e+24 (106) Dual inf 3360
(1)
Clp0060I Primal error 2.14748e+09, dual error 1.06329e+07
Clp0060I Primal error 2.14748e+09, dual error 4.38545e+06
...
Clp0010I Flagging variable R16579
Clp0006I 285  Obj 1.98074e+21 Primal inf 3.18064e+25 (196)
Solution update n_cols:17614      n_rows: 17535      n_iter: 419
time: 0.871
clp status        = 0
clp prob status   = 0
clp second status = 0
Iteration Count               : 419
isAbandoned()                 : 0
isProvenOptimal()             : 1
isProvenPrimalInfeasible()    : 0
isProvenDualInfeasible()      : 0
isPrimalObjectiveLimitReached : 0
isDualObjectiveLimitReached   : 0
isIterationLimitReached       : 0




_______________________________________________
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/20100129/6910bbd1/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/20100129/6910bbd1/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/20100129/6910bbd1/attachment-0001.gif>


More information about the Clp mailing list