Hello

I found this old thread http://www.gamedev.net/topic/171751-texture-coordinates-and-index-buffers-can-it-be-done/ which describes a problem I wanted to know if newer DirectX versions has addressed.

Suppose you have a Cube, which has 8 corners. But depending on which side of the cube that is being rendered, different UVs and Normals would have to be used. The thread concludes that the vertices would have to be duplicated in the seams that use different components.

I immediately start thinking of Streams and how they could possibly be used for solving this issue, but I can't figure out exactly how it can be done, or if it can be done.

Is it possible to move UV and Normal components over to the Index buffer in some way, such that a vertex may be drawn multiple times with different UVs?

http://msdn.microsoft.com/en-us/library/windows/desktop/ff476335%28v=vs.85%29.aspx

You could put the vertex positions, normals and UVs each in a separate structured buffer, and the vertex buffer would simply contain indices into these three structured buffers.

Then, in the vertex or geometry shader, you simply use three structured buffers at the indices from the input vertex buffer.

You can indeed bind multiple vertex buffers, but you can only bind a single index buffer. You'll always read from the vertex streams with a shared index.

Mesh resolutions have become higher and thus smoother, resulting in less duplication needed, so in a way it has been addressed/minimized.

Compared to a few duplicated vertices, having multiple index buffers would only have higher demands on memory, because then every vertex is guaranteed to carry extra information (a fixed set of indices).

Mesh resolutions have become higher and thus smoother, resulting in less duplication needed, so in a way it has been addressed/minimized.

So smooth surfaces do not duplicate vertices at the same position? I figure the Normal must then be a blend of the 3 adjacent face normals?

So smooth surfaces do not duplicate vertices at the same position? I figure the Normal must then be a blend of the 3 adjacent face normals?

Yes, for a cube specifically, that would be problematic. If you tie the SV_VertexID semantic to a vertex shader input (D3D10+), you can forego vertex buffers altogether and construct a cube in its entirety in the vertex shader based of the index value that's passed in.

but you can only bind a single index buffer

That's why I said he can put the indices into a vertex buffer:

and the vertex buffer would simply contain indices into these three structured buffers