[eclipse-clp-users] Strange behavior of the ic solver

From: Dario Campagna <dario.campagna_at_...184...>
Date: Mon, 10 Aug 2009 11:09:58 +0200
Hi,

	working with the ic library of ECLiPSe I discovered a strange  
behavior of the solver.
In particular, let us consider the following goal

[eclipse 2]: K #:: 0..80, X $= 10.2, X $= 3.2 + (K * 0.1).

K = 70
X = 10.2__10.2


Delayed goals:
	0 $= 8.8817841970012523e-16__1.7763568394002505e-15
Yes (0.00s cpu)

Let us now consider another goal with the same constraints of [eclipse  
2] but written in a different order

[eclipse 3]: K #:: 0..80, X $= 3.2 + (K * 0.1), X $= 10.2.

No (0.00s cpu)

Changing the order of the constraint we obtain different results. I  
think that this should not be the case for a solver.
Does this behavior depend only on the presence of the value 0.1 (that  
can not be exactly represented with binary floating-point arithmetic)  
in the goal or there are other causes?


Thanks.
Dario Campagna
Received on Mon Aug 10 2009 - 09:10:22 CEST

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