[osg-users] Optimizing for CAD Data

Paul Martz pmartz at skew-matrix.com
Mon Jul 28 11:43:11 PDT 2008

Hi Todd -- I'd think the Geode-per-Drawable might improve cull performance
in your case, as it's faster to cull against bounding spheres than boxes.

I've been working with complex CAD data for over a year and have had good
results with some models by inserting OcclusionQueryNodes at strategic
locations. You might take a look at that.

If you are working with PolyTrans to convert the CAD data, you should email
me offline. I've been developing some tools to ease that process.

Paul Martz
Skew Matrix Software LLC
+1 303 859 9466

> Hi folks,
> We deal with a lot of CAD data that has been translated to 
> OSG, and I am looking into ways to enhance the performance in OSG.
> One issue we are facing is that the CAD data is very poorly 
> organized with regards to the cull process.  We had hoped 
> that the "spatialize groups" optimization would help us out, 
> but our problem seems to go deeper than that.  Depending on 
> the source of our data, we get Geodes containing numerous 
> Drawables that are spread throughout the scene (all instances 
> of one part under one Geode).  As I understand the spatialize 
> optimization, it reorganizes Geodes but doesn't go to the 
> Drawable level.
> So, the idea now is to insert Geodes & enforce a rule of one 
> Drawable per Geode to give the optimizer a chance to succeed. 
>  Does that sound like it would work in our situation?
> Assuming the idea makes sense, would it be of value to the 
> community if we try to implement this as one of the 
> osgUtil::Optimizer options?  If so, would it make sense as an 
> improvement to the "spatialize groups" 
> option, or would it be better as an option unto itself?
> -Todd
> --
> Todd J. Furlong
> Inv3rsion, LLC
> http://www.inv3rsion.com
> _______________________________________________
> osg-users mailing list
> osg-users at lists.openscenegraph.org
> http://lists.openscenegraph.org/listinfo.cgi/osg-users-opensce

More information about the osg-users mailing list