Hi Ale,
Your scene tree is clearly overfragmented.
You may investigate an adapted osgUtil::Optimizer::SpatializeGroupsVisitor
usage
Cheers
Ale wrote:
> Hi,
>
> thanks for the aswer.
>
> Yes, sure. I am running in release.
>
> It is a big scene:
>
> State Graphs: 2321
> Fast Drawables:
Hi,
thanks for the aswer.
Yes, sure. I am running in release.
It is a big scene:
State Graphs: 2321
Fast Drawables: 14177
Primitive Sets: 52013
Vertices: 7215794
Triangles: 971664
Quads: 984448
Polygons: 362048
It is strange that Cull always go from 900ms to 300/400 ms doing nothing after
Hi Ale,
Having many thousands of transform in the scene graph will be very
expensive on cull performance as each transform require the view frustum,
modelview and projection matrices and other cull paramters to be cloned and
transformed into the local coordinate frame. Thosudans of transforms
897 ms culling time sounds excessive to me, even for a large scene. Is this
application built in a release configuration of the compiler?
You could attempt a hierarchical grouping of the scene graph nodes with
group objects to speed up the culling, using some kind of space
partitioning algorithm.
Hi,
until now I improved visualization performance of my application in several way
thanks to your suggestions.
Now I would like to go another step ahead.
May be this is common topics but it is not clear to me how to solve it.
I have a CAD-like application.
Potentially a scene can contain
5 matches
Mail list logo
