Hello.
I am wondering wether rendering with VBO and keeping a second copy of vertices in system memory for stencil shadows would make sense?
I could imagine that the shadows would make the app so fillrate bound that the VBO-advantage never really would be apparent.
How does doom handle this for example?
Thanks for any help!
Why would you keep a seperate copy…?
Wouldn’t you just render the vertex array again with new transformations to render the shadow volume and perform stencilling?
Hmm, so far I can not use vertex programms.
Maybe what you describe is possible with them, but without them it is necessary to read your vertices (to my knowledge), which is not good with VBO and the like as I believe.
Some clarification here would be great!
EDIT: Can you find the silhouette and do all that stuff with vertex programms?
[This message has been edited by B_old (edited 07-27-2003).]
I may have misread your question. I think I was thinking of something else… I think.
What I mean.
Keeping your vertices in AGP memory is nice as long as you don’t read them, right?
So I thought I keep a copy in AGP memory for rendering and one in system memory for the stencil shadows. Don’t know, could be used for collision detection, too, maybe.
The question: might the stencil shadows make the gfx-card so busy that it really wouldn’t make any difference wether I use VBO or not?
Thanks for the help!
Originally posted by B_old:
The question: might the stencil shadows make the gfx-card so busy that it really wouldn’t make any difference wether I use VBO or not?
No. Using VBO is very efficient even in that case.
Ofcourse if you are very fillrate limited then it doesn’t help, but probably this mean also that your program is below 10fps (in a real scene case)…
OK then, thanks for the help.