[Bug-glpk] [ numerical instability & infinite loops ]

[GARDI , Frédéric]

Dear M. Makhorin,

Please find attached a CPXLP file describing a MIP whose resolution implies an infinite loop during branch and bound (the algorithm loops on a basis when solving a LP relaxation). The resolution is done with GLPK 4.24, compiled on Win32 x86 with VC 8.0, with the default parameters. The same problem occurs with different sets of parameters (for example with --nopresol, or with branching on most fractional variable --mostf). The same behavior is found when compiling with GCC 3.4.4 under Cygwin.

Note that in GLPK 4.24 the parameter  LPX_K_ITLIM has no more effect in MIP resolution. In effect, the iteration limit is not correctly passed from the IOS structure to the LP structure, when solving the MIP relaxation at each node. Idem for the time limit with parameter LPX_K_TMLIM. (I made some patches in my GLPK version, but I think that my corrections are quite dirty). Please feel free to contact me if my report is not enough detailed.

Yours sincerely,

Frédéric GARDI

Bouygues e-lab

Courriel : address@hidden

<<Numerical_Instability.txt>>

----------------

Les donnees et renseignements contenus dans ce message sont personnels, confidentiels et secrets. Toute publication, utilisation ou diffusion, meme partielle, doit etre autorisee. Si vous n'etes pas le bon destinataire, nous vous demandons de ne pas lire, copier, utiliser ou divulguer cette communication. Nous vous prions de notifier cette erreur a l'expediteur et d'effacer immediatement cette communication de votre systeme.

Any data and information contained in this electronic mail is personal, confidential and secret. Any total or partial publication, use or distribution must be authorized. If you are not the right addressee, we ask you not to read, copy, use or disclose this communication. Please notify this error to the sender and erase at once this communication from your system.

Numerical_Instability.txt
Description: Numerical_Instability.txt