[gecode-users] Naive CCP ask in Gecode

[Christian Schulte]

The observation to be made is that you always should first propagate to a
fixpoint and then perform the ask. While not incorrect (asks are monotonic)
it would be inefficient anyway. So one why would be to do it as follows:

-          Give a space a lock

-          In order to do propagation, the lock must be acquired and hence
remains locked

-          In order to do an ask, also the lock must be acquired, so there
are no issues with synchronization

 

Then, of course, remains the question whether one wants to use OS threads.
They are made for big computations and not tiny asks, so that might incur a
tremendous overhead.

 

Then, one other thing: you might want to avoid pthreads as they are not
fully portable. Please check the boost threads package for a portable and
efficient alternative.

 

Cheers

Christian

 

--

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

 

From: users-bounces at gecode.org [mailto:users-bounces at gecode.org] On Behalf
Of Mauricio Toro
Sent: Tuesday, March 11, 2008 8:11 PM
To: cschulte at kth.se
Cc: gecode at gecode.org
Subject: Re: [gecode-users] Naive CCP ask in Gecode

 

Hello Mr. Schulte,

 

I had many threads doing something like this:

 

//pthread start

 while ( not root[j+n].assigned() ){ gecode::status()};

  gecode::status(); 

  if (root[j+n].val() == 1) 

      rel(this, root[i+n], IRT_EQ, 1);

//pthread end

 

But, since Gecode is not thread safe, I should

only have one thread asking if the variables are

assigned and calling the propagators, right?

 

Thank you very much,

 

Sorry for asking so much.

 

Mauricio

 

2008/3/11, Christian Schulte <cschulte at kth.se>:

What you can always do, is compute a fixpoint for a space (via status()) and
the find out.

 

Christian

 

--

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

 

From: users-bounces at gecode.org [mailto:users-bounces at gecode.org] On Behalf
Of Mauricio Toro
Sent: Monday, March 10, 2008 11:31 PM
To: gecode at gecode.org
Subject: Re: [gecode-users] Naive CCP ask in Gecode

 

Hello all,

 

I am trying to use Gecode for make some ccp processes.

something like:

 

  when a = 1 do tell  b = 1

|| when b = 1 do tell c = 1

|| tell a = 1

|| when b = 1 and c = 1 do tell answer = true

 

with a,b,c all in the same intvararray (in the same space),

how can i make that?

 

thanks

 

2008/3/10, Mikael Zayenz Lagerkvist <zayenz at gmail.com>:

Hi,

If you are using a Gecode Space in a concurrent setting, there are a
couple of things that you must observe.

First of all, a Space is not reentrant. Thus you must serialize all
accesses to the Space object.

Secondly, if you want to manipulate two different Spaces concurrently,
you must make sure that they do not share any data structures. This is
done by passing false as the share argument when cloning a Space.

Cheers,
Mikael


On Mon, Mar 10, 2008 at 12:54 PM, Mauricio Toro <mauriciotorob at gmail.com>
wrote:
> Hello,
>
> I am trying to do a naive CCP ask in Gecode and
> I am getting "Bus error".
>
> The idea is doing this process inside a thread
> ask (a = 1) do tell( b = 1)
>
> I am doing it like this:
>
>  while ( not root[i].assigned() ){ };
>    if (root[i].val() == 1)
>       rel(this, root[j], IRT_EQ, 1);
>
> Any clues why I am getting "Bus error",
>
> Thanks
>
> --
> Mauricio Toro Bermudez
>  Estudiante de Ingeniería de Sistemas
> Pontificia Universidad Javeriana, Colombia
>
> Stagiare à l'Ircam
> 1, place Igor-Stravinsky 75004 Paris,
> France de 2008 à 2009

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




--
Mikael Zayenz Lagerkvist, http://www.ict.kth.se/~zayenz/




-- 
Mauricio Toro Bermudez
Estudiante de Ingeniería de Sistemas
Pontificia Universidad Javeriana, Colombia

Stagiare à l'Ircam
1, place Igor-Stravinsky 75004 Paris, 
France de 2008 à 2009 




-- 
Mauricio Toro Bermudez
Estudiante de Ingeniería de Sistemas
Pontificia Universidad Javeriana, Colombia

Stagiare à l'Ircam
1, place Igor-Stravinsky 75004 Paris, 
France de 2008 à 2009 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.ps.uni-sb.de/pipermail/users/attachments/20080312/24900c5a/attachment.htm>