[gecode-users] Propagation order

[Christian Schulte]

Well,

You are right that A and B will go before C and C before D but this is just
in the current implementation. We might change the implementation in the
future, though. The reason why we just approximate the cost values is that
each cost value takes memory.

So, the idea is: very often the cost is respected by the system. But what
you can't do is: you implement D such that it breaks if C has not run
before.

Does that help?

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


-----Original Message-----
From: users-bounces at gecode.org [mailto:users-bounces at gecode.org] On Behalf
Of Peter Vanhee
Sent: Monday, June 14, 2010 10:39 AM
To: users at gecode.org
Subject: Re: [gecode-users] Propagation order

Hi Christian,

Most of the propagators I have implemented behave like filters: they get at
least one assigned variable and then accept or discard the assignment.

Some constraints are much more heavy than others and therefore I'd have 
them run after the less heavy constraints. 

For example

A -> PropCost::binary(PropCost::LO);
B -> PropCost::ternary(PropCost::LO);
C ->PropCost::quadratic(PropCost::LO);
D -> PropCost::crazy(PropCost::HI);

Can I rely on the fact that always A goes before B, B before D and so on?
It seemed to me that D would be run before all the rest, but that between A,
B and C there was not always a clear order.

Bye,
Peter


On 14 Jun 2010, at 10:18, Christian Schulte wrote:

> Hi,
> 
> There are of course the normal cost values: unary, binary, ternary,
linear,
> etc. The system picks the cheapest, however some cost values are treated
> alike. That is why there are no hard facts to be relied on.
> 
> But maybe you can give us an idea what you try to achieve?
> 
> Cheers
> Christian
> 
> --
> Christian Schulte, www.ict.kth.se/~cschulte/
> 
> 
> -----Original Message-----
> From: users-bounces at gecode.org [mailto:users-bounces at gecode.org] On Behalf
> Of Peter Vanhee
> Sent: Monday, June 14, 2010 7:23 AM
> To: users at gecode.org
> Subject: [gecode-users] Propagation order
> 
> Hi all,
> 
> I am trying to weight different propagators implementing a cost function
> "virtual PropCost cost(const Space&, const ModEventDelta&) const".
> 
> As indicated by the documentation "The only hard fact you can rely on is
> that a propagator 
> with a cost value using PropCost::HI is never given preference over a
> propagator with a cost 
> value using PropCost::LO". 
> 
> For my particular problem however I'd need to weigh propagators quite
> differently,
> so they get executed almost sequentially, and simply saying LO or HI isn't
> enough. 
> Does gecode offer another way to control the order of propagation?
> 
> Thanks,
> Peter 
> _______________________________________________
> Gecode users mailing list
> users at gecode.org
> https://www.gecode.org/mailman/listinfo/gecode-users
> 


_______________________________________________
Gecode users mailing list
users at gecode.org
https://www.gecode.org/mailman/listinfo/gecode-users