[gecode-users] Indexing into arrays

[Christian Schulte]

Uuups, the ES_FAILED is a different beast all together! This is what a
propagator will return to the kernel to signal failure but finding out which
propagator returned it will require hacking the kernel. Frankly, that will
require some stamina ;-)

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 Milt
Sent: Monday, November 21, 2011 4:11 PM
To: users at gecode.org
Subject: Re: [gecode-users] Indexing into arrays

Thanks Christian,

> ...failing a space does not
> stop execution of a space's method (or with other words, does not 
> throw a
> C++ exception or something like that). The reason for that is that it 
> C++ does
> not matter! The code defining the variables, constraints, and 
> propagators is executed just once initially in order to create the 
> space for the root node of the search tree. This is cheap. What takes 
> time is doing constraint propagation (triggered by status()) and this 
> is terminated immediately when failure occurs.

Good...so Gecode is right to keep going after failing the root node because
it has to finish setting up the model infrastructure once (to post all of
the constraints), but it's very efficient beyond the root node.  I'll read
further in the branch and search sections of MPG to better understand how
and in what order everything executes...that seems to be where I'm fuzziest
right now.

> I am not sure I understand your second point. After running status() 
> you should always check whether a space is failed (status() returns 
> that information). So, when testing why not run status() everytime to 
> make sure that things are not yet failed?

Ok, so I can check for the result from status() e.g., SS_FAILED, rather than
just running status() to force propagation. I just read some more of MPG
chapter 20 and see there's also ES_FAILED which can be used to determine
which propagator failed.

I'll give that a shot.
Thanks for your help!
-Milt


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