Infinite Projected Vertex

On NV’s paper(Robust VolumeShadow) ,it caculates the vertex ,projected away from the light to infinity by this:
(AxLw - LxAw , AyLw - LyAw, AzLw - AwLz,0). Is that from the formular: AwLw(Ax/Aw - Lx/Lw , Ay/Aw - Ly/Lw, Az/Aw - Lz/Lw,0) ??
And I think silhouette polygon edges caculation is a terrible thing cos it needs more time and more ram.and can mess up my pipeline at all…

All you have to do to find a possible silhouette is one dot product per poly and than compare this poly’s orientation to the neighbor’s one. Storing one byte or even one bit per poly isn’t too hard, is it ?

Hi thanks.
I mean, I have to use some spaces to caculate and store the result .all of my datas are strips even triangles .The problem is, I don’t have faces info but only strips so it’s not very happy to store all faces’ normals in ram.To get the topology also waste my time.

And can u answer my 1st question?thanks

what do you mean? your mesh is fragmented? I mean, it’s just a bunch of seperated strips and seperated tris?

if not, you can (and should) preprocess your geometry, find the edges, and generate a connectivity info.


Yes, but you have to take care that Lw is not zero.

Not that it comes up a whole lot, but our formulation assumes (and requires) Aw and Lw are non-negative.

Thanks -

Cass We all understood that the Aw and Lw coords are assumed to be non zero. Most likely 1. THe only problem I have had with my implementation of the paper on stenciled shodow volumes was that I had the FrontFace defined as CCW and I was drawing the shadow volumes CW. After all these years of doing opengl programming that was my only problem. These issues always come up. How is the time for everyone to laugh at me. Come on now Laugh it up.