Original Post
As far as vertex grouping for rendering, from what I''ve heard it sounds like most people group all vertices by texture, to reduce texture switching.
But lets say there are multiple objects (or fragments of objects) using a specific texture. Naturally all of the objects are in your main dynamic vertex buffer, but is each object still drawn separately? Or is it best to try to combine them all using linking tris (I forget what they''re called, the one-dimentional tris for connecting lists) so that its just one long triangle list for each texture?
If linking tris are used, some of the connected objects could potentially be very far apart. Is there any problem in having these linking tris stretched out so far, such as floating point errors or clipping complexities?
I hope this makes sense, any input would be great.