[gecode-users] Memory consumption - element constraint

[Christian Schulte]

Hi Benoit,

Maybe you want to tell us how large your graphs are? 

To give you an idea, we have been playing a little bit with some word
problems (brought to us by Håkan Kjellerstrand). Here, in the small examples
we tried, the arrays have around 70,000 elements and there are plenty of
these element constraints and everything is solved in around 8 Megabytes of
main memory. I think, Håkan even successfully tried examples with ten times
as many elements.

Or, to put it into perspective: solving these puzzles with the large element
constraints takes a small fraction of the time it takes a C++ compiler to
compile the source code containing the word lists for the element
constraints as string constants...

Best
Christian

PS: Of course, the arrays for element constraints are never copied, they are
shared among copies!

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


-----Original Message-----
From: users-bounces at gecode.org [mailto:users-bounces at gecode.org] On Behalf
Of benoit
Sent: Monday, November 02, 2009 5:05 PM
To: users at gecode.org
Subject: [gecode-users] Memory consumption - element constraint

Hello,

I came across a problem of high memory consumption with the element 
constraints. According to this message 
(http://www.gecode.org/gecode-users/2009-February/002828.html), it would 
come from the large size of the employed arrays.

In my case, the array represents a directed multigraph. It stores the 
values hold by the arcs. These values are a priori computed, stored in a 
variable of type IntArgs and retrieved during execution. Here comes the 
problem of size.

A possible solution would consist in computing the needed values during 
execution. However, this calculation can be a bit complex and will 
require different temporary variables. I am afraid of producing a code 
difficult to maintain. Is there a better way of doing this ? Are there 
more appropriate structures ? (I should add that the considered 
multigraph is sparse).

Thanks in advance.

Benoît

_______________________________________________
Gecode users mailing list
users at gecode.org
https://www.gecode.org/mailman/listinfo/gecode-users