Original Post
I read this article on occlusion culling yesterday and, while the author admits that the method presented isn't that great, I've been having trouble finding other resources on what is a good method. I mean obviously, rendering the same mesh up to three times per frame is highly inefficient and since I've been having trouble finding other methods, I've decided to try my hand at it myself. I haven't actually tested this method yet, but I've been tossing it around in my head for about 6 hours now and I wanted to know what you guys think. What if we used basically the same method as presented in the above mentioned article, except instead of rendering bounding meshes, we render the actual mesh once, perform the D3DQUERY operations and use the results as the basis for rendering of the next frame. The only downfall I can think of for this is, obviously, results are going to be off a little, but I think the margin for error is small enough it might be a viable solution. Anybody have any thoughts? -AJ