Hello Aaron,

aaron bryden wrote:
> Has some change been made recently between the released 1.8-0 branch
> and 1.x head that would result in a drastic performance drop.

I'm not aware of any change that decreased performance and so far we had 
no reports about such an issue.

> I have
> a vrml model that after producing all the primitives in it results in
> about 930k triangles and 4600 nodes.
> 
> In windows using the QT based sceneviewer I get about 40 FPS and in OS
> X with a glut based app using SimpleSceneManager I get about 7 FPS.

Hm, a couple of things that come to mind:
- do you use release (optimized) libraries on both OS ?
- do your OSX graphics drivers have support for hardware acceleration ?
- what is the general quality of OpenGL on OSX, does someone know if 
there is a gap in driver quality ?
- is this on the same machine just with different OS or on different 
hardware as well ?

> If  I change the number of elements in the spheres and cylinders in
> the model by editing the VRML node desc in OS X on the same machine I
> can only get the frame rate to 25 going down to about 100k polygons
> and and about 20 FPS going to 250k polygons.  I have also tried
> various combinations of the merge and split operators to improve
> performance in OS X and have had little success.
> 
> Is this the real performance gap between OS X and windows xp or is
> there a certain revision that I should sync to before some performance
> damaging tinkering occured.

See above, I not aware of any patches that knowingly decreased 
performance in order to fix a correctness issue for example. Whether 
there is a difference in OpenGL performance between win and OSX I can 
not comment on, as I have not used either of them for development. Does 
somebody else have any insights ?

> Does anyone have a suggestion for how to
> deal with a model that is a bunch of cylinders and spheres located
> such that the cylinder + sphere appear to be capsid (I am dealing with
> VRML files exported from pymol).

hm, if the cylinder have end caps and the spheres are full spheres 
instead of just half-spheres there are quite a number of polygons being 
wasted, because they are inside the capsules. Do these capsules differ 
only in orientation/size/material or are they actually different 
geometries ? In the former case a single geometry that is just 
transformed should be sufficient for the scene and then you might also 
use a Level of Detail (LOD) node to switch to lower resolutions for 
geometry in the background.

        Hope it helps,
                Carsten



-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Opensg-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/opensg-users

Reply via email to