Hi,
As Christian said, I am interested in having FloatVar in Gecode. I
contacted him last Spring and started work during Summer.
I try to follow Gecode standards as close as possible. All interval'
stuff are based on boost intervals.
Up to now, I designed FloatVar and views (FloatView, MinusView and
OffSetView) and a bunch of propagators :
- Linear (Lg, Gq, Eq)
- Arithmetic (Square, Sqrt, Abs, Pow, NthRoot, Mult, Div, Min, Max)
- Trigonometric (Sin, Cos, Tan, ASin, ACos, ATan)
- Transcendental (Exp, Log)
- Channel (to int)
For a first release it may be enough, so I concentrate on tests to
remove bugs. I plan to add tools for modeling (like other Gecode
Variables and constraints), but I must first solve an issue about the
rounding for trigonometric and transcendental functions. Indeed, boost
doesn't provide full rounding functions for trigonometric and
transcendental functions. To achieve correct rounding, it relies on
other libraries like MPFR. If I can't rely on standard C maths
functions to get a correct rounding, I have to investigate on other
alternatives. MPFR is not a good choice for Windows users, so maybe
another library (CRLibm ? others ?). I'd appreciate any advice on
this. Does someone known a good "light" maths library which gives
exact rounding ?
If you have any questions or remarks, don't hesitate to contact me!
Cheers,
Vincent
Christian Schulte <[email protected]> a écrit :
Hi Filip,
I started some time ago to look into the FloatVar stuff but then I got
contacted by Vincent Barichard about that. Vincent expressed his interest
and has already arrived at a second prototype. I am quite positive that this
will lead to a good module for 4.*.
Vincent, maybe you can describe the status so far.
Best
Christian
--
Christian Schulte, www.ict.kth.se/~cschulte/
-----Original Message-----
From: [email protected] [mailto:[email protected]] On Behalf
Of Filip Konvicka
Sent: Thursday, November 03, 2011 7:11 PM
To: [email protected]
Subject: Re: [gecode-users] Gecode 4.0.0
Christian,
Sorry to reply so late...
The list of features sounds great, especially the symmetry breaking (though
I admit that I haven't looked at the propagator groups yet).
Gonzalo from the AVISPA team contacted me in May regarding the Float Vars
work they were doing, but I haven't heard from them since. Do you have any
information on this? Are you counting on using their implementation? Or do
you plan to do this yourself?
Thanks,
Filip
Damn it, there are two pieces of information missing...
Timeline: we aim at the second quarter of 2012 for Gecode 4.0.0. But
today I went through some older emails (written when we discussed what
Gecode 3.* should be) and there I wrote that Gecode 4 should be around end
of 2009.
Okay, still trust me! This time everything is much different ;-) Just
kidding.
Then, of course: Is there something that we absolutely must do for our
users?
Christian
--
Christian Schulte, www.ict.kth.se/~cschulte/
-----Original Message-----
From: [email protected] [mailto:[email protected]] On
Behalf Of Christian Schulte
Sent: Friday, October 14, 2011 3:03 PM
To: [email protected]
Subject: [gecode-users] Gecode 4.0.0
Dear all,
As this year has already a number of releases of Gecode from the 3.*
version series we have decided to slow the pace and prepare for Gecode
4.0.0 as the next major release.
This mail is just to assure you that we are not dormant but that it
will take longer before the next version appears. If there is the need
for a bug fix release we will of course do that in the meantime.
Our current plan is to include the following features:
- Integrate propagator groups, see here:
http://web.it.kth.se/~cschulte/paper.php?id=LagerkvistSchulte:CP:2009
- Integrate lightweight dynamic symmetry breaking, see here:
Lightweight Dynamic Symmetry Breaking. C. Mears, M. Garcia de la
Banda, B. Demoen, M. Wallace.
SymCon'08.
LDSB is a dynamic symmetry breaking method designed to handle
common symmetries efficiently under any variable and value
ordering.
- Integrate float variables
- Provide modeling layer for float variables
- Support for restarts during search and large neighborhood search
- Make randomized tie-breaking more flexible
- Modeling layer for scheduling constraints
- Branching for scheduling problems
Some more speculative things that might or might not make it are:
- Integrate discrepancy-based search that is not LDS (?)
- Half-reification, see here:
Thibaut Feydy, Zoltan Somogyi, Peter J. Stuckey: Half Reification
and
Flattening. CP 2011: 286-301
- Redundancy analysis, see here:
http://web.it.kth.se/~cschulte/paper.php?id=SchulteStuckey:PPDP:2008
Best
Christian
--
Christian Schulte, www.ict.kth.se/~cschulte/
_______________________________________________
Gecode users mailing list
[email protected]
https://www.gecode.org/mailman/listinfo/gecode-users
_______________________________________________
Gecode users mailing list
[email protected]
https://www.gecode.org/mailman/listinfo/gecode-users
--
Vincent Barichard Université d'Angers (LERIA)
Tel: 02 41 73 52 06 Département Informatique
Fax: 02 41 73 50 73 H203
_______________________________________________
Gecode users mailing list
[email protected]
https://www.gecode.org/mailman/listinfo/gecode-users
