[eclipse-clp-users] Integer Set Library: Performance related to initial domain size?

From: Ulrich Scholz <Ulrich.Scholz_at_eml-d.villa-bosch.de>
Date: Thu, 25 Sep 2008 14:58:22 +0200
Dear all,

I have a question about the Integer Sets library: How is the performance
related to the initial size of the domains?

Assume you have a Integer Sets constraint problem and the initial domains of
the variables are subsets of the set {1, ..., N1}.  

Furthermore, assume that the problem's solution is invariant regarding
increasing initial domain size.  So, for example, it would have the same
solution (*) if we chose the initial domain size to be subsets of the set
{1, ..., N2}, N2 > N1.

Now we increase N2.  How does the performance behave?


The reason I ask is: Set variables have to be defined with a domain.  In my
application, the number of variables and their relation (subset, \=, etc.)
are known before the actual domains of the variables.  So I have to "guess"
their domains.  If there is no performace penalty, I just choose the domains
"large enough". 


Thanks, Ulrich


(*) same solution: either both are unsolvable or each pair of corresponding
variables has the same range.

-- 
Ulrich Scholz            

Phone: +49-6221-533244
Email: ulrich.scholz at eml-d.villa-bosch.de

--

European Media Laboratory GmbH
Schloss-Wolfsbrunnenweg 33
69118 Heidelberg

Amtsgericht Mannheim / HRB 335719
Managing Partner: Dr. h.c. Klaus Tschira, Scientific and Managing
Director: Prof. Dr.-Ing. Andreas Reuter

www.eml-d.villa-bosch.de
   
Received on Thu Sep 25 2008 - 13:28:07 CEST

This archive was generated by hypermail 2.2.0 : Thu Feb 02 2012 - 02:31:58 CET