[gecode-users] Implementing No-goods

Christian Schulte cschulte at kth.se
Wed Sep 25 21:17:59 CEST 2013


It uses the same master (well, an equivalent one).

Best
Christian

--
Christian Schulte, www.ict.kth.se/~cschulte/


> -----Original Message-----
> From: Kish Shen [mailto:kisshen at cisco.com]
> Sent: Wednesday, September 25, 2013 9:12 PM
> To: cschulte at kth.se
> Cc: gecode-users at gecode.org
> Subject: Re: [gecode-users] Implementing No-goods
> 
> On 13/09/2013 12:20, Christian Schulte wrote:
> > In your idea of doing several searches after each other I assume that
> > you use the next() function to get a space and then continue search.
> > As the no-goods are posted in the master space they will be contained
> > in the space returned by next() as well (unless they already have been
> > found out to be useless).
> 
> 
> Hi Christian,
> 
> I thought I better separate the two issues into different posts....
> 
> Thinking about the restart search more, I think my question may arise from
a
> (mis)understanding of what the master space is. Please let me know if the
> following is correct:
> 
> master(.., const Space* s, NoGoods& ng)
> 
> I had assumed that each time master is called, there is a fresh copy of
the space
> s, i.e. a clone of the space before search was started. In this case, then
you need
> to remember NoGoods for all the previous restarts in ng.
> 
> However, thinking about it more, it seems to make more sense that you use
the
> same master space (s) in all the restarts of a search; in this case ng
would only
> need to be the NoGoods in the single "pass" since the last restart,
because all
> the previous NoGoods would already be posted to s by previous calls to
> master().
> 
> If a different search is done after this, as I suggested previously, then
you have a
> different master space s from the previous search.
> 
> Cheers,
> 
> Kish
> 





More information about the users mailing list