[Coin-bcpdiscuss] BCP: node buffering?

[Joerg Herbers]

I've got problems using BCP when very large search trees are explored. My BCP-based algorithm then runs out of memory - which may partly be attributed to the use of user data in the search nodes.

By default, application memory is limited to 2GB on WinXP 32bit (see http://www.brianmadden.com/content/content.asp?ID=69), but the situation is not much better on Linux (32 bit). In fact, some gigabytes of memory are quickly exhausted when exploring large search trees.

Has anybody thought about implementing a node buffer (similar to the CPLEX node file) before? Is there interest in having one?

Jörg



Dr. Jörg Herbers
INFORM GmbH, Pascalstraße 23, 52076 Aachen, Germany
Telefon: (+49) 2408 9456-24, Fax: (+49) 2408 9456-25
E-Mail: joerg.herbers at inform-ac.com, Web: http://www.inform-ac.com