Are all your Geodes within the view frustum and not culled? If so, I'd
think the "fewer Geodes" case would be faster. But if some are culled,
then the "more Geodes" case would take advantage of host culling,
sending less data to the hardware, and would therefore run faster.
Harold Comere wrote:
Hi all,
I want to do a kind of performance benchmark for my application.
So i have done a class that display X triangles in Y geodes ( X and Y
varying ), render it few seconds, compute average fps then do it with
another X and Y.
Two things :
- I have noticed that for a given amount of triangles, fps will be
greater with any amount of geode than with another. For exemple, 2 000
000 triangles are displayed faster in 5000 geodes than in only one.
So i wonder if there is any ideal ratio triangles / geodes ?
But i do not understand how it can be faster with more geode than with
only one .. in my understanding, there is one call of display list /
VBO per geode drawable, so more geode do more calls and so is slower
than only one call, am i wrong ?
- I set X / Y triangles per geode and each geode is child of root node.
So all my tests are done with a one-stage graph without any
multiparent. I get very bad results, less than 10 average FPS for 2 000
000 triangles ... Therefor i wonder if this scene graph could be called
"worst scenario" ?
I guess that a balanced scene graph should give better results.
I imagine the best scenario should be a balanced graph with
multiparent. Maybe a forest with only one kind of tree, so there is
only transforms and on only geode containing the tree geometry ?
Well, i wonder how to do a reliable benchmark, testing worst, medium
and best scenario of scene graph.
Thank you for your attention,
Regards,
Harold
_______________________________________________
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
|
_______________________________________________
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org