I am writing a game that uses procedural textures. These textures are generated using a simple fractal technique. The fractal can be generated at an arbitrary level-of-detail. That is, I can repeatedly add higher and higher level-of-detail textures to the mipmap chain. I would like to know how D3D goes about selecting which mipmap level will use for each triangle it renders. Then, I would know to which level-of-detail I should generate the fractal textures every frame. I think this is a quite general problem that will be encountered by those using procedural textures. Thanks!

Also, do mipmaps help in speed, caus they draw smaller texture when farther away?

Esap1,

Mipmapping lets the renderer use lower-detail (smaller) textures for far away objects. One of the most important things they do is reduce aliasing. Without mipmapping, far away objects appear to "sparkle".

For my renderer, I generate higher and higher level-of-detail surfaces as I zoom in closer. By doing this on-the-fly, I save a lot of memory, at the expense of extra processor time. Actually, I''m using it for fractal terrain.

I suppose they do improve speed in a way. Because otherwise, the renderer would need to do more filtering in order to reduce aliasing.

Mipmap''s help in speed, because they increase the number of texel cache hits.

As for which one is selected, it is driver-dependent, but it should select the mipmap that results in a texel/pixel ratio that is closest to 1.