[gecode-users] Linear memory allocation

[Christian Schulte]

Hi,

1) The problem here is the number of variables, the number of constraints is
not that much of an issue. Are they Boolean or integer variables. Boolean
variables are much smaller. But still it takes too much memory. Could I have
a look at your model?

2) Domain consistent propagation for linear is exponential (that is, here it
is 2^49) Please read Tip 4.6 in MPG. Maybe MPG should have a big fat warning
(we had this issue before, that's why the reference doc has a big fat
warning).

2) What does "Gist crashes" mean. It should crash even though propagation
does not terminate. 

Cheers
Christian

--
Christian Schulte, www.ict.kth.se/~cschulte/

-----Original Message-----
From: users-bounces at gecode.org [mailto:users-bounces at gecode.org] On Behalf
Of Farshid Hassani Bijarbooneh
Sent: Tuesday, September 06, 2011 9:38 AM
To: users at gecode.org
Subject: [gecode-users] Linear memory allocation

Hello,

I have a model with 10^2 to 80^2 variables and linear constraints over 49
variables each. We have the following issues:
1. The code allocates at least 1Gb of memory for even N=50.
Changing the recomputation distance (-c-d) doesn't seem to help in this
case.
2. Changing the consistency to Domain (line 187) causes the code to process
endlessly! (Gist crashes) even for N=20 I suspected the domain consistency
problem is because of the use of IntVarArray instead of BoolVarArray and
changing to Bool fixed it, but still very strange that Int gets stuck.

Please find the code of the isolated problem attached.