I'm currently writing a planet renderer that renders a realistically scaled universe, everything works really well, LOD, full culling (horizon and view frustum). I'm currently using a modified quadtree too calculate the LOD of a planet terrain chunk and too determine whether or not a chunk is visible. Since my terrain is all ready calculated in fixed sized chunks i just render them as such (each terrain chunk has its own VB and a global index into a LOD Index Buffer), which means each terrain chunk is rendered explicitly with a draw call. When my planets are at the highest level of detail there will be a lot of Draw calls since there will be a lot of terrain chunks I figured I'd be able to reduce the number of draw calls by grouping terrain chunks together into larger buffers, i was just wondering what type of systems do people use for situations like this, all the geometry is static and is created when the quadtree sub divides. Is there a maximum number of primitives a draw call should render at once? At the moment my draw calls are rendering 4225 verts in one draw call, if there is 50 terrain chunks thats 50 draw calls, if i batched two VB together there would be 25 calls which i would assume is quicker.

Quote:Original post by jollyjeffers
I've had quite a lot of success going the opposite way around - single VB and many IB's. This makes LOD quite easy as you can have a heirarchy of LOD IB's, each skipping/stepping over different numbers of underlying vertices (for example)
Jack

This is the method I'm using at the moment for rendering the planet with no cracks and works very well, the only difference being i have multiple VB as well as multiple IB's, if i combined all the VB's into one VB and had separate IB's, i would still need to call the draw function multiple times since the IB's would be different like this

Set VB (Giant Buffer)
Set IB (Chunk 1)
Draw (Chunk 1)
Set IB (Chunk 2)
Draw ( Chunk 2)

as opposed to this (my current method)

Set VB (Chunk 1)
Set IB (Chunk 1)
Draw (Chunk 1)
Set VB (Chunk 2)
Set IB (Chunk 2)
Draw (Chunk 2)

Maybe there is a speed increase in calling set VB only once, but there should still be the same number of draw calls (please correct me if I'm wrong, I'd happily bang all these in a single VB if it would help). Would there be no performance increase if we put all the Indices in a giant IB (32bit Indices) as well?

Is rendering ~1M polygons in on call efficient, if so i'd happily throw all my VB's together :)

Quote:Original post by jollyjeffers
Have you read Ysaneya's developer journal? It's OpenGL based but sometime ago he did cover in depth the various terrain rendering algorithms and his results are simply jaw dropping
Jack

I haven't but i will :)

bump ? ;)