I'm attempting to create a gpu accelerated soft body simulation with shader model 3. I understand everything except when it comes to making a geometry map of the 3d positions for the simulation. Nvidia made a cloth simulation with a similar technique, but they made a geometry map from vertex positions.
From Nvidia's white paper:
"A geometry image is created from a given mesh by cutting it along a network of edge paths such that it becomes topologically equivalent to a disk and then sampling the result over a grid to get the positions of the vertices at each pixel of the image"
I am going to use the uv map, rather than trying to transform the mesh, and map the object space positions to the corresponding pixels on the uv map. It doesn't matter if this is done in the shader or not because only the initial positions are needed. If there is a way to do this on the cpu using bitmap manipulations in lite-c in a reasonable amount of time that would also work. The tricky part doing it on the cpu is that the pixels are not one to one with the vertices so it would have to find the vertex with the closes uv coordinate and I did not find any functions to do this and could be very slow.
My speculation is that there is some sort of transformation matrix I can use in the shader to do this, but I am not a shader expert so I don't know much about them.
Your help with this is much appreciated.