[Coin-discuss] How do I "speed up" CBC?

[John J Forrest]

Christian,

Very difficult to answer.  I have one integer model  which really annoys 
me as a tiny change to code or optimization level can change time for 
complete search from 10 seconds to nearly an hour.  How many nodes does 
each take?  If yours takes many more then it is not memory - but "bad 
luck".  If so try changing a few parameters e.g. to do with strong 
branching to see if you can make it faster.  It is also worth seeing how 
much objective moved at root due to cuts in case one cut generator had a 
better starting point.

John Forrest



WoelliUSM at aol.com 
Sent by: coin-discuss-bounces at list.coin-or.org
10/02/2006 02:18 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] How do I "speed up" CBC?






Dear all,
 
I got a question concerning the performance of the CBC-solver. I use 
FlopC++ in combination with CBC under Linux and my professor GAMS with CBC 
under Windows. We use the same test instances, but there are dramatic 
differences of the performances of CBC. 
After the generation of a test instance (does not matter whether FlopC++ 
or GAMS), it takes only seconds for him to solve the problem. I need much, 
much, much more time. 
One test instance has about 660000 non-zeros in the coefficient matrix. It 
takes about 250 seconds to solve on his computer, and I get the message 
"KILLED"! 
Where can be a software difference? Are there different starting 
parameters? We both did not change CBC after downloading, we just 
connected them to GAMS/FlopC++. 
There is only one difference of the computers we use. We both have a 3GHZ 
Inter Pentium 4, but I just got 1GB RAM, whereas he got 4GB RAM!
Is that the answer to the question above?
 
Thanks in advance,
 
Christian_______________________________________________
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/20061003/cc5e9302/attachment.html>