I think I didn't send this to the mailing list, so here it is...
-------- Forwarded Message --------
Subject: Re: [gecode-users] propagation consistency of count (single
value) constraint
To: cschu...@kth.se
References: <55de1fb2.90...@cisco.com> <55df3132.8080...@cisco.com>
<00ba01d0e4a5$32dd8720$98989560$@kth.se>
From: Kish Shen <kiss...@cisco.com>
Message-ID: <55ebc8b7.3020...@cisco.com>
Date: Sun, 6 Sep 2015 06:01:43 +0100
User-Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:38.0) Gecko/20100101
Thunderbird/38.2.0
In-Reply-To: <00ba01d0e4a5$32dd8720$98989560$@kth.se>
Content-Type: text/plain; charset="windows-1252"; format=flowed
Content-Transfer-Encoding: 7bit
MIME-Version: 1.0
On 01/09/2015 11:58, Christian Schulte wrote:
Hmm, I had a quick look at the code and I think it should rewrite to the
int/int case. I am travelling right now, so I will look more carefully when
I am back in a week or so.
Hi Christian,
I looked at the program more closely, and it seems that the reason for
the apparent better performance with ICL_BND is because the constraint
succeeds when it is violated. The following does not fail when posted
with ICL_BND:
count(home, x, y, IRT_GR, z, ICL_BND)
x = {5,5,5,5,5,5,5}, y = 1, z = 0
y and z are IntVars with singleton domains.
This succeeds when posted with x set to all 5s, or as variables which
are then labelled to 5 during the search.
The constraint fails when posted with ICL_DOM.
Cheers,
Kish
_______________________________________________
Gecode users mailing list
users@gecode.org
https://www.gecode.org/mailman/listinfo/gecode-users
