Re: [eclipse-users] Imposing over 1 Million Constraints

From: Stefano Novello <snovello_at_...6...>
Date: Mon, 18 Jun 2007 17:29:14 +0200
Hello David,
When posting such questions its always good to provide a litlle more detail.

For example you have not mentioned what library you are using for
this. Eplex and ic both handle the kinds of constraints you mentioned.

Many programs will post similar numbers of constraints one by one in
far less time than yours. A million constraints sounds less than
overwhelming when modern processors perform billions of instructions
per second.

Given the information you provided I have one suggestion for you. --
Use the profiler.

If you can cut down your problem to arbitary smaller sizes (its always
good to structure your applications so you can do that) then you might
see which are the predicates whose cputime contribution is growing
fastest with problem size, or there may simply be a dominant few
predicates you need to optimize away.

There are also libraries in eclipse for instrumenting your  code to
count invocations and for detecting where memory is used. They help
too, but the profiler is easiest to try first IMO.

Stefano Novello

> -----Original Message-----
> From:
> [
> om] On Behalf Of David Tian
> Sent: Friday, June 15, 2007 10:02 PM
> To: ECLiPSe Users List
> Subject: [eclipse-users] Imposing over 1 Million Constraints
> Hi,
> My problem is a data mining problem which is modelled as a
> CSP using linear integer constraints over boolean variables.
> The model consists of 1,261,570 constraints all of which are
> imposed in the usual way i.e., 1-by-1. It took a few days and
> nights to impose all of them.
> Is there a faster way to impose all the constraints?
> Is it possible to impose more than 1 constraints
> simultaneously on a PC? I am interested to find out how to do
> this on a PC.
> Many Thanks
> David
> _______________________________________________
> ECLiPSe-Users mailing list
> ECLiPSe-Users@...2...
Received on Mon Jun 18 2007 - 16:29:27 CEST

This archive was generated by hypermail 2.2.0 : Mon Jul 09 2018 - 02:05:29 CEST