[CHiPPS] Questions on Blis
Yan.Xu at sas.com
Thu Mar 12 09:32:25 EDT 2009
1) The 0% printed in log is truncated. Its actual value is something like 0.0001%. A separated issue is that the gap parameters of Blis doesn't work. Need fix.
2) For parallel runs, the best bound will be updated periodically, so sometimes it doesn't move.
3) I tried to incorporate COIN's presolve, but couldn't make it work, so I quit. If you can contribute yours, that's great!
4) Blis stores nodes by differencing to save memory, subtrees are sorted of compressed. Probably, it need functionality to store subtrees in hard drive.
From: chipps-bounces at list.coin-or.org [mailto:chipps-bounces at list.coin-or.org] On Behalf Of Alessandro Tiberi
Sent: Thursday, March 12, 2009 7:26 AM
To: chipps at list.coin-or.org
Subject: [CHiPPS] Questions on Blis
I have been playing a little with Blis and I have a bunch of questions:
1) I would expect the search to stop as soon as a 0% gap is reached.
However, both in serial and parallel mode, even with a 0% gap, nodes
continue to be queued
and the search process does not stop until the node queue is empty (for
instance this happens with cap6000 of miplib3). Am I missing something?
2) While when running in serial mode the best bound is updated during
the search, in parallel mode it appears to stay put.
3) Presolve. Blis presolve appears to be only partially implemented and
hardcoded to be off. Is there any reason for not using it? (I tried to
fully implement it and it seems to work fine)
4) Large instances. Has anyone tried Blis with large problems? I am
trying to solve instances with about 500K columns and I am running into
Concerning memory, I was wondering if subtrees are or can be compressed...
CHiPPS mailing list
CHiPPS at list.coin-or.org
More information about the CHiPPS