[gecode-users] Heap memory exhaustion

[Guido Tack]

Joe Porter wrote:

> I'm having a problem with Gecode 2.0.0 on Windows XP Professional.   
> My application seems to work fairly well on many cases, but I have  
> two cases that run for about 9 minutes before declaring that the  
> heap memory is exhausted.

This is hard to analyze without some more facts about your model.   
E.g., how many variables do you create, and how big are the initial  
domains?  What kind of variables (IntVar, SetVar, CpltSetVar?) and  
constraints?

> Summary
>         propagations:  37440926
>         failures:      18387289
>         clones:        18387303
>         commits:       36777015
>         peak memory:   646672 KB

Apparently the search space is huge.  This is normally no problem for  
Gecode, unless the search tree is also very, very deep (memory  
consumption should be linear in the tree depth).  So if you have many  
variables with large domains and little propagation, the tree could be  
very deep, and the result is that the system gives up.  In this case,  
I'm not sure you can solve the problem by just using more memory,  
you're going to need a "stronger" model.  If, on the other hand, the  
tree is not so deep, it must be a bug (a memory leak) in Gecode, and  
we'd be most interested to get your actual model so that we can find  
out what's wrong.

Cheers,
	Guido