<br><font size=2 face="sans-serif">Kim,</font>
<br>
<br><font size=2 face="sans-serif">Looks like an accuracy/tolerance issue
in CoinPreSolve. Presolve was creating elements with value 2.0e-12
and then believing them.</font>
<br>
<br><font size=2 face="sans-serif">I have modified tolerance in some places
(trunk). Is it OK now?</font>
<br>
<br><font size=2 face="sans-serif">John</font>
<br>
<br>
<br>
<table width=100%>
<tr>
<td><img src=cid:_1_0C0E84E80C0E8100007E62CE85257441>
<td width=100%>
<table width=100%>
<tr valign=top>
<td width=100%><font size=2 face="sans-serif"><b>Re: [Clp] Infinite loop
in Clp</b></font></table>
<br>
<table width=100%>
<tr>
<td><font size=2 color=#e26200 face="sans-serif"><b>Kim Hansen </b></font>
<td><font size=2 color=#8f8f8f face="sans-serif">to:</font>
<td><font size=2 face="sans-serif">John J Forrest</font>
<td>
<div align=right><font size=1 face="sans-serif">05/06/2008 12:22 PM</font></div></table>
<br>
<table width=100%>
<tr>
<td>
<table width=100%>
<tr>
<td><font size=2 color=#8f8f8f face="sans-serif">Sent by:</font>
<td width=100%><font size=2 color=#e26200 face="sans-serif"><b>clp-bounces@list.coin-or.org</b></font>
<tr>
<td><font size=1 color=#8f8f8f face="sans-serif">Cc:</font>
<td><font size=1 face="sans-serif">Clp</font></table>
<br>
<td>
<div align=right></div></table>
<br></table>
<br>
<br>
<hr>
<br>
<br>
<br><tt><font size=2>On Mon, May 5, 2008 at 5:02 PM, John J Forrest <jjforre@us.ibm.com>
wrote:<br>
> Kim,<br>
><br>
> Even the run with an mps file does not look quite right.<br>
><br>
> a) Refresh your version of trunk - many changes have been made.<br>
<br>
That fixed the problem, thanks.<br>
<br>
> b) use writeMpsNative with formatType 2 to get an mps file with exactly
the<br>
> same values<br>
><br>
> If you want you can send the mps fiel to me and I can look at run.<br>
<br>
I have attached the file in the standard format as the clp program<br>
finds the wrong solution when I use that file:<br>
<br>
kim@raph:~/orion/svn/trunk/solver$ clp x.mps<br>
Coin LP version 1.06.00, build May 6 2008<br>
command line - clp x.mps<br>
At line 1 NAME BLANK<br>
At line 2 ROWS<br>
At line 188 COLUMNS<br>
At line 1805 RHS<br>
At line 1896 BOUNDS<br>
At line 2073 ENDATA<br>
Problem BLANK has 184 rows, 257 columns and 3019 elements<br>
Model was imported from ./x.mps in 0.004 seconds<br>
Presolve 109 (-75) rows, 120 (-137) columns and 1921 (-1098) elements<br>
Perturbing problem by 0.001 % of 4.49537 - largest nonzero change<br>
1.44776e-05 (% 22.5473) - largest zero change 1.44096e-05<br>
0 Obj -1.22719e+11 Primal inf 1.22108e+18 (69)<br>
31 Obj 1.28446e+15 Primal inf 1.74144e+07 (39)<br>
33 Obj 1.47474e+15 Primal inf 2.05988e+13 (39)<br>
Primal infeasible - objective value 1.47474e+15<br>
Presolved problem not optimal, resolve after postsolve<br>
After Postsolve, objective 1.60904e+15, infeasibilities - dual 433739<br>
(35), primal 1.88861e+14 (114)<br>
PrimalInfeasible objective 1.609035234e+15 - 66 iterations time 0.022,<br>
Presolve 0.02<br>
kim@raph:~/orion/svn/trunk/solver$<br>
<br>
The real solution has an objective of 7.49029e+06, glpsol and the<br>
simple clp solver I have written agree on that.<br>
<br>
-- <br>
Kim Hansen<br>
Vadgårdsvej 3, 2.tv<br>
2860 Søborg<br>
Fastnet: 3956 2437 -- Mobil: 3091 2437<br>
[attachment "x.mps.bz2" deleted by John J Forrest/Watson/IBM]
_______________________________________________<br>
Clp mailing list<br>
Clp@list.coin-or.org<br>
http://list.coin-or.org/mailman/listinfo/clp<br>
</font></tt>
<br>