Hi all: I just want to know if anyone can tell with clod terrain development, a book, a online tutorial, or any resource where i can get started with, i want to use VC++, please help!!! sorry about bad english

You could check here. Terrain LOD: Runtime Regular-Grid Approaches, here Continuous LOD Terrain Meshing Using Adaptive Quadtrees, also you could use google and search for LOD algorithms. There are plenty of them online.

But I have been seen a tendency of dropping such algorithms to use a brute force approach, since todays hardware can handle terrain been drawn that way.

Also search for geoclipmapping as MassacrerAL suggested

I hope that helps.



Edit: Fixed links, added info

I have also noticed that many terrain generation resources use a brute froce aproach but what happens when you try to render a huge terrain like in games like fsx, farcr, the brute force aproach is ok?, or it require another aproach? , which one?

but what aproach should i use to render an island surrounded by water?

You could start out by just breaking your terrain up into a quad tree. And since we're rendering an island, you can do some tricks to cull the terrain. Like, cull terrain below water when camera is above, cull terrain above water when camera is below, etc. And if you have your water height at half the height of the terrain, you can effectively cull half of the terrain, making it possible to have pretty big height maps.

Quote:Original post by MassacrerAL
google for geoclipmapping(geoclipmaps, whatever) or geomipmapping. the second one is quite easier than the first one, and i`m sure if you google for it or search gamedev you can find a lot of info on it.

This is just to note geoMipmapping is probably the simplest DLOD method out of there. I don't remember if terrain clipmaps (by Hoppe on Microsoft research site) qualify for being CLOD but the grid driving them is absolutely "discrete".

Quote:Original post by MassacrerAL
brute force when done effectively is not that bad. quake wars uses this approach and obviously it works well. i guess it wouldn`t work for far cry or oblivion. it depends on your needs. however i don`t know which technique is used in far cry or oblivion. anyone?

Good point. Whatever an algo scales or not is heavily dependant on the used data - the vertex shader complexity, the appearance, even the post-cache optimizations may give you a good boost.
I originally planned to have various algorithms but after an initial performance evaluation brute force and a very simple HLOD seemed to be strictly necessary. Even that way, NVPerfHud reports my program as pixel-processing bound!
I suppose that happens because my culling methods are fairly primitive but even cutting the pixel stage workload to a quarter it would still bottleneck...

I don't plan to use CLOD any time soon. Initial prototypes (admittedly pretty old right now) didn't prove much value and I'm scared about eating my PCIe bandwidth. As a last thing, their "progressive refinement" seemed worse than a DLOD popping. Think at it for a while: if progressive refinement is kicking in, it means the user is slowing down. It happens that pixel movement is a LOD metric by itself so if you refine when the user slows down you simply risk to get caught while cheating...

Quote:Original post by glaeken
And since we're rendering an island, you can do some tricks to cull the terrain. Like, cull terrain below water when camera is above, cull terrain above water when camera is below, etc. And if you have your water height at half the height of the terrain, you can effectively cull half of the terrain, making it possible to have pretty big height maps.

Take care: matching multi-terrain bounduaries wasn't easy for me and I suppose it won't be for you as well.