[Clp] (no subject)

[Jack Hagart]

Hello everyone from CLP mail list. I need a help.

I´m using the CLP to a software I developed. CLP works very well to me. The
problem is, when I started use some real data, with numbers greater, the
CLP output file is using round to the values of the variables in the
solution.

I implemented the same mathematical model in GAMS/CPLEX, with exactly same
input data, and see the result:


CLP GAMS/CPLEX  1 $43.597.637,00 $43.597.636,76  2 $23.108.017,00
$23.108.017,02  3 $44.053.085,00 $44.053.085,30  4 $7.624.834,80
$7.624.834,80  5 $17.312.266,00 $17.312.265,81  6 $11.199.682,00
$11.199.682,39  7 $16.631.305,00 $16.631.304,60  8 $9.903.795,80
$9.903.795,82  9 $10.465.548,00 $10.465.548,33  10 $12.723.051,00
$12.723.051,42  11 $8.210.366,70 $8.210.366,68

 For example, the line 1. The right solution, that one that respects all
the constraints, is $43.597.636,76, or the GAMS/CPLEX solution. The CLP
plot $43.597.637,00. Its the same solution, but CLP round the decimals, and
put 637 in the place of 636,76.

The same is happening with others values of the solution. But one thing is
strange. Take a look in the lines 4 and 8. In this cases, CLP didn´t round
the values!

This is happening just to numbers greater than 10 million. I tried change
the option output to 4 and 6 to improve precision but it didn´t worked to
CLP 1.14.4.

Can someone help me with new ideas?

A small piece of the original CLP solution report is bellow. The variable
f(1,13) is that one in line 1 in the table above.

optimal
Objective value -1.6527462e+008
      1 f(24,27)               1.6526604e+008                 0
      5 f(1,13)                      43597637                       0
      7 f(13,14)                     43597637                      0
     10 f(2,14)                      23108017                      0
     14 f(3,15)                      44053085                      0
     18 f(4,16)                     7624834.8                      0

Best Regards,
Jack
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/clp/attachments/20130715/67375b22/attachment.html>