On Fri, 22 Oct 2010, Tim Kroeger wrote:

> On Thu, 21 Oct 2010, Roy Stogner wrote:
> 
>> Would it make sense to loop over all active elements, not just local
>> ones?
>
> That would certainly solve the problem, but it wouldn't scale well, would it? 
> (At least as long as ParallelMesh is not stable.)

At the moment it won't scale much worse than looping over active local
elements.  The predicate iterators were implemented (by John, I
believe?) as nothing more complicated than iterator+predicate; they
don't do any data structure tricks and so they don't get you out of an
O(N_global) cost for iterating over every local element.  We've never
found much time being spent iterating over non-local elements when
profiling so we've never had a reason to try to optimize it.
---
Roy

------------------------------------------------------------------------------
Nokia and AT&T present the 2010 Calling All Innovators-North America contest
Create new apps & games for the Nokia N8 for consumers in  U.S. and Canada
$10 million total in prizes - $4M cash, 500 devices, nearly $6M in marketing
Develop with Nokia Qt SDK, Web Runtime, or Java and Publish to Ovi Store 
http://p.sf.net/sfu/nokia-dev2dev
_______________________________________________
Libmesh-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libmesh-users

Reply via email to