It probably works the same way as vertex programs where if you modify the vertex position it does nothing to the vertex normal, and you would need to manually modify the vertex normal if needed. So you still need a normal map to get proper shading, unless you were doing something crazy and editing the normals along with…
You'd lock the vertex normals per-tesselation, the normals wouldn't be re-normalized after the displacement was added, so in theory, it wouldnt be an issue. You could skip the normal map and smooth the normals again in the shader, but I imagine this would be very slow.
Thank's for all the info so far guys. I love reading about this stuff. One thing bothers me however. What about using a tangent space normal map and a displacement map at the same time? Wouldn't tesselation destroy the vertex normal data used by the normal map?
From something I read a long time ago I remember that dx11 tessellation is done with a "tessellation program" similar to vertex and fragment programs in shaders. It goes in between the vertex and frag program I guess in there you tell it how much to tessellate and what you want to do with the extra geometry created. From…