Heya,
Q: How much is your highest resoluted base-patch subdivided when giving it to the geometry shader? I read that there is a limitation of 256 faces? And playing around with glsl confirm this assertion. So i'm curious in understanding your technique. You are gaining patches with > a * pow(10,6) faces.
A small example will explain my confusion.
We assume a computed patch with 2 * pow(10,6) faces
Each face needs a maximum of 3 EmitVertex() commands. Which are limited by a maximum of 256.
fCount = 2 000 000 // Face Count
vCount = 6 000 000 * 3 = 6 000 000 //Vertex Count
Probably you wont need 6 * pow(10, 6) vertices for computing the faces by using of TriangleStrip Mode (or some other mode).
If you want to create this high resoluted patch with a geometry shader you would use a base patch with ~23400 faces
6 000 000 / 256 ~= 23437 vertices
Mh i see that this is not as much as i expected. So probably i did something wrong. Please could you shortly explain the technique behind the scenes?