[Dip] non-empty intersection of column sets in subproblem

[Matthew Galati]

All shared columns must be in the master and not in subproblem blocks.

DIP will give an error if there are overlapping (column wise) blocks. The
example below will surely give overlapping blocks. There is an option to
turn this check off (for sake of speed), but, it is on by default.

   //---
   //--- sanity checks on user input
   //---
   if(m_param.DebugCheckBlocksColumns)
     checkBlocksColumns();


Let me know if it is not reporting or missing some overlap.

Matt



On Mon, Oct 22, 2012 at 7:47 AM, Shahin Gelareh <shahin.gelareh at gmail.com>wrote:

> hellp everybody
>
> Imagine we have a set of constraints like "x_{ijk}+x_{kji}+x_{jki} =1
> \forall i, j,\neq i , k\neq i,j" is left to be solved in teh subproblem.
>
> this set does not show a block angular structure and cannot be easily
> decomposed to different blocks.
>
> I wonder if we set a block for every ijk, say subproblem_{123} ==>
> {x_{123}+x_{321}+x_{231} =1}
> and given that I do not receive any error message from Dip, how is that
> managed in DIP, in general.
>
> How do the blocks communicate for the shared columns, do they actually
> need to communicate? how the contribution of the same column appearing in
> two or more blocks is calculated.
>
> thanks in advance
> Shahin
>
> _______________________________________________
> Dip mailing list
> Dip at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/dip
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/dip/attachments/20121022/2e262214/attachment.html>