> 
> Ahhh premature optimization.  With performance optimization work one really 
> does need to profile against normal usage.
> 


But I think that kind of optimization on for example Android devices is useful.


> 
> Cull should normally be pretty cheap, if it's expensive it suggests that the 
> scene graph is sub-optimal and is too fine grained/contains too many in scene 
> graph transforms.  There are lots of techniques for addressing this.
> 


I know I was trying to improve it w. But when you have a lot of dynamic object 
on scene( 100-500 ). Lot's of animated object so you only optimize a root of 
problem which in my view is a cull traversal.


> 
> One has to do testing in context, doing specific tests such as running a test 
> within a small loop will give you quite different results than when the code 
> is used in normal scene graph usage.  As I mentioned, normally scene graphs 
> the cull and draw traversals are mostly effected by bandwidth limitations 
> that tend to swap CPU operations.
> 


As you says earlier that  OSG's focus is a high performance. So high 
performance should be not only in normal scene graph usage. I looking forward 
that OSG have a good performance in on heavy dynamic scene :). My personal goal 
is too help to achieve that.

To end this discusion I will try in the near future to prepare SEE versions of 
mathematics. And I hope that the performance boost of their use will that will 
be available as an option in the OSG.

Lukasz
 

------------------
Read this topic online here:
http://forum.openscenegraph.org/viewtopic.php?p=56726#56726





_______________________________________________
osg-submissions mailing list
[email protected]
http://lists.openscenegraph.org/listinfo.cgi/osg-submissions-openscenegraph.org

Reply via email to