I admit I haven’t read that much of the vertex program documentation, but:
I think one central question is whether the state of the temporary registers are kept from vertex to vertex (which I presume they are). This would allow for the results from the previous vertex processing to affect the processing of the next vertex, which I suspect would be much more powerful, since the result would both depend on previous results as well as parameters for the current vertex.
I’m not really that familiar with fractal graphics either, but is there some other serious limitations for doing these kinds of things with a vertex program as opposed to having the flexibility of doing everything on the cpu?
Is the lack of connectivity information that much of a limitation?
Comments greatly appreciated.
There’s no state kept from one vertex to another, you can however set a number of constants for the whole primitive and you can put other data than vertices/colors etc. in your vertex buffers.
So I can imagine that it would be possible to have some kind of seed number in the vertex buffer that combined with the constants form vertices. At nVidia they have created a nice particle system using this technique, but I’ll leave finding the trick for a fractal up to you.
Somebody on the nVidia site did a vertex program that drew the Mandelbrot Set.
It’s under the technical presentations section, I think.
j