[Coin-lpsolver] St9bad_alloc with ClpINterior

John J Forrest jjforre at us.ibm.com
Thu Sep 1 05:11:00 EDT 2005


Kish and fadeaway,

For wssmp you need to download from IBM's alphaworks.  If you have an old 
copy of osl you can use that with WSSMP_BARRIER=2.

If you have trouble using wssmp then try the University of Florida 
ordering code and modify COIN/Makefiles/Makefile.location.  WSSMP will be 
considerably faster.

using a dense factorization you need more than 4*n*n bytes which is a 
large number if n=50000.

The dense was just there for testing as there are good packages out there.

John Forrest



Kish Shen <ks15 at icparc.ic.ac.uk> 
Sent by: coin-lpsolver-bounces at list.coin-or.org
08/31/2005 09:18 PM

To
coin-lpsolver at list.coin-or.org
cc

Subject
[Coin-lpsolver] St9bad_alloc with ClpINterior







Hi,

I have been trying out the barrier solver of Clp, using the testBarrier
driver in Clp/Samples as a blueprint, but using ClpCholeskyDense instead 
of
ClpCholeskyWssmp.

For two moderately sized problems (one of them the MIPLIB problem nw04)
with about 200000 non-zeros and 50000 rows and columns, I got a 

terminate called after throwing an instance of 'std::bad_alloc'
  what():  St9bad_alloc

message and the execution aborted. This happened very soon after the
solving process started, and the process was using about 50M of memory
(which I was monitoring using 'top') immediately before this message, and
the machine I was running on has about 750M of real memory and 2G of swap
space, and most of the swap space was available, so it seems unlikely that
there was really insufficient memory.

Any ideas what the problem might be? Thanks in advance for any information
and help!

Cheers,

Kish Shen
IC-Parc
Imperial College London
U.K.
_______________________________________________
Coin-lpsolver mailing list
Coin-lpsolver at list.coin-or.org
http://list.coin-or.org/mailman/listinfo/coin-lpsolver

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/clp/attachments/20050901/388b3a4f/attachment.html>


More information about the Clp mailing list