[Clp] Sparse matrices

[Daniel Herrero]

Hello everybody:

I'm researh at Politechnic University of Madrid. I've been working on solving LP problems with Nvidia GPU's using parallel methods. 


I'm starting now with sparse matrices, and I have read Suhl & Shul papers form 1990 and 1993. The results are much more time consuming than what I hope, for example a problem with 3500 constraints, 9625 continous variables and 90% sparsity, as is described in those papers, I spend 250 seconds using the usual Simplex method, and about 2500 seconds using the LU decomposition because the matrix multiplication is much more time consuming than row updates.

I'm not trying to parallelize the problem yet, just trying to understand the methodology of LU decomposition. Could you tell me what are you using for doing this?, I've read that you use Cholesky, but I'm a little lost and maybe not taking into account something important.


Thank you so much

 
Daniel Herrero Giner
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/clp/attachments/20120305/7105bf23/attachment.html>