Hi Kelvin, hi group!
Wow, it starts making sense to my stupid mind now.
> But D3D don\'t support QUADS, so we break the
> primitive into to draw two triangles at a time
>
> using
> DrawPrimitiveStrided(D3DPT_TRIANGLEFAN,
> vertexFormat, &strideData, 4, 0);
Strided? Uhhhh....
Lots of work for the driver I suppose.
> So for ColorCube there are 6000 DrawPrimitives call.
> That\'s why it hurts performance as Philip Taylor
> from Microsoft mention in his email.
Sure.
> There are several things that you can do right now to
> improve performance.
>
> (1) Use TriangleArray instead of QuadArray to define Cube
> (Only if D3D version is used)
How can I figure what driver is used? (Or should I know?)
> (2) In you scene graph there is a TransformGroup above every Cube.
> You can eliminate the 1000 TransformGroup by per-transforming
> the coordinate in your Cube definition. This can significantly
> increase performance under v1.2 (compare with v1.1.3).
Hm, in my original stuff, the actual GeomtryArray for the cube was kept static,
bc. I thought having the geometry/vertex data only once (only one vertexbuffer in the
future) being transformed from case to case should be faster. Vertexbuffers are a
limited resource bc. they use videomemory on T&L systems and a transform is cheap
there.
Do I have to construct my graph due to hardware capabilities now? How can I find out
about that? I will need to know wether I\'m on OGL or D3D and than on which version of
J3D?
Solutations,
- J
===========================================================================
To unsubscribe, send email to [EMAIL PROTECTED] and include in the body
of the message "signoff JAVA3D-INTEREST". For general help, send email to
[EMAIL PROTECTED] and include in the body of the message "help".
