ok, i go to follow analysing a bit the main idea.
OBJECTIVE: we want a system that allows raydium create automatically, load and render terrains.
To
create automatically terrains we can use usual terrain generators code. However i'd like to make it better with a few improvements like precipices, undergrounds, rivers, and so.
The
loading and rendering methods are in the same block.
According the idea of megatextures we should:
1-generate a huge mesh
2-generate a huge bitmap
3-add extra info(sounds applied, friction) to the bitmap
4-use that mix-up.
People say that megatexture is really a implementation of clip-mapping, in other words, the game only uses a part of a bigger container of data. But this is not completely true since in Quake Wars is promised that (in certain places) you can see all the map from horizon to horizont(in the other side).
So, if this is true i have to think that:
The megatexture includes auto mipmaping + auto clipmapping
So:
If we can a texture of 32000x32000, it will have mipmapings from 16000x16000 to 1x1. And each level of mipmaping is realated to a level os clipmaping. If we fix a "part" with a size of 10 pixelsx10.
32000x32000 -> 3200x3200 parts
8000x8000 -> 800x800
10x10 -> 1 part
The intermediate parts could be usefull to draw far terrain sections. Obviously the 10x10(1 part) won't be of use in the game, you never will go to be enoguh away to see the full map like only 10x10 pixels.
The use of parts with mipmaps allow the engine a really high framerate avoid writind unneded data.
But, the thing is: Will ODE be comfortable with this?
Well, this is my particular vision of megatextures... Surely i'm completely wrong