[Coin-discuss] User exit tests in Cbc?
John J Forrest
jjforre at us.ibm.com
Fri May 5 14:01:30 EDT 2006
I am not sure how to recover from a crash. Are you checking memory usage
and then exiting if it gets close to 2Gb?
If so the simplest thing to do is to derive an event handler from
CbcEventHandler to do the checking and return stop if too much memory
being used. You use passIn EventHandler to tell CbcModel to use your one.
John Forrest
acw at ascent.com
Sent by: coin-discuss-bounces at list.coin-or.org
05/05/2006 12:29 PM
Please respond to
Discussions about open source software for Operations Research
<coin-discuss at list.coin-or.org>
To
Coin-discuss at list.coin-or.org
cc
Subject
[Coin-discuss] User exit tests in Cbc?
Our application using COIN/Cbc is going very well. We have successfully
ported to Windows and integrated an only-slightly-tweaked Cbc solver with
the rest of our application.
Recently, however, we encountered a case for which the solver ran out of
memory and crashed ignominiously in the middle of a run. (In the
benighted world of 32-bit Windows, a process's address space is limited to
about 2Gb.) We are, therefore, adding an extra exit test to the main loop
in CbcModel::branchAndBound. It isn't very tricky. But we would like to
do it without modifying CbcModel.cpp any more than is absolutely
necessary, so that we can stay synchronized with the main-branch sources
as easily as possible. So this led us to consider adding a little user
hook to the exit tests. Perhaps just after "!stoppedOnGap &&
!eventHappened" we could add "&& !userExitTest()".
Have the Cbc developers considered making this kind of user exit hook
available? What would you consider the most "Cbc-y" way to do it?
Thanks for your thoughts._______________________________________________
Coin-discuss mailing list
Coin-discuss at list.coin-or.org
http://list.coin-or.org/mailman/listinfo/coin-discuss
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/coin-discuss/attachments/20060505/abf28acb/attachment.html>
More information about the Coin-discuss
mailing list